Forebrain neural progenitors effectively integrate into host brain circuits and improve neural function after ischemic stroke

Abstract

Human cortical neural progenitor cell transplantation holds significant potential in cortical stroke treatment by replacing lost cortical neurons and repairing damaged brain circuits. However, commonly utilized human cortical neural progenitors are limited in yield a substantial proportion of diverse cortical neurons and require an extended period to achieve functional maturation and synaptic integration, thereby potentially diminishing the optimal therapeutic benefits of cell transplantation for cortical stroke. Here, we generated forkhead box G1 (FOXG1)-positive forebrain progenitors from human inducible pluripotent stem cells, which can differentiate into diverse and balanced cortical neurons including upper- and deep-layer excitatory and inhibitory neurons, achieving early functional maturation simultaneously in vitro. Furthermore, these FOXG1 forebrain progenitor cells demonstrate robust cortical neuronal differentiation, rapid functional maturation and efficient synaptic integration after transplantation into the sensory cortex of stroke-injured adult rats. Notably, we have successfully utilized the non-invasive ¹⁸F-SynVesT-1 PET imaging technique to assess alterations in synapse count before and after transplantation therapy of FOXG1 progenitors in vivo. Moreover, the transplanted FOXG1 progenitors improve sensory and motor function recovery following stroke. These findings provide systematic and compelling evidence for the suitability of these FOXG1 progenitors for neuronal replacement in ischemic cortical stroke.

Introduction

Ischemic stroke is a cerebrovascular event that results in neuronal damage due to a lack of glucose and oxygen supply to a specific region of the brain. At present, mechanical thrombectomy and pharmacological thrombolysis remain the only approved treatments aimed at dissolving obstructive agents during the acute phase (within 6 h) after the event¹. Meanwhile, clinical trials have demonstrated that the drug and rehabilitation training therapies have not yielded favorable outcomes in the majority of chronic cases^2,3. Consequently, there is an urgent need for novel strategies to facilitate tissue repair and neurological recovery.

Recently, transplantation of human neural progenitor cells (NPCs) has emerged as a promising therapeutic approach for stroke⁴. Numerous studies on stem cell transplantation for cerebral stroke, particularly cortical stroke, have identified two primary mechanisms mediating behavioral improvements¹. One mechanism involves the so-called bystander effects, such as nutritional support, modulation of inflammation, and stimulation of plasticity⁵. The other, more direct approach involves the transplantation of cells to replace lost neurons, achieving functional integration to repair damaged circuits⁶. To date, clinical trials of stem cell transplantation in ischemic stroke have primarily focused on mesenchymal stem cells or immortalized neural stem cells⁷. The behavioral improvements observed in these trials have been attributed mainly to bystander effects rather than cell replacement mechanisms⁴. This is because such cells are theoretically difficult to differentiate into diverse functional neurons capable of circuit repair. In recent years, a growing body of research utilizing animal models has demonstrated that transplanted cortical neural progenitor cells can differentiate into mature cortical neurons within adult cortices damaged by stroke^4,8,9,10 or chemically injured by ibotenic acid¹¹. These cells functionally integrate into the host brain’s neural network, reconstruct damaged neural circuits,^4,8,9,10,11 ameliorate neurological deficits^4,8,10. Collectively, these studies strongly support the concept that replacing dead neurons and reconstructing the stroke-injured adult brain is possible, and they raise the possibility that this approach may also be achievable in a clinical setting in the future. Assuming that transplanted cortical neurons can achieve fundamentally repair of damaged circuits in stroke patients through cell replacement mechanisms, it becomes particularly important to explore and identify the optimal source of neural precursor cells. Regrettably, several pioneering studies in the field of cell replacement therapy for cortical strokes continue to focus solely on and utilize specific differentiation method for cortical precursor cells^4,8,9. In reality, significant advancements have been made in the induced differentiation of cortical neural precursor cells. However, the role played by these cortical neuronal precursors, which exhibit varying degrees of differentiation and maturation capabilities, in neuronal replacement within the stroke-damaged cortex remains a significant unknown.¹²

Over the past decade, a plethora of techniques have been developed to generate specific neuronal lineages^{13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28}. Among these, the strategy of dual SMAD signaling inhibition was initially developed to induce the differentiation of human pluripotent stem cells into PAX6⁺ neural precursors²⁹. Moreover, exposure to compounds that inhibit WNT signaling under dual SMAD inhibition has been demonstrated to promote the expression of forkhead box G1 (FOXG1), a marker for forebrain precursors²². This provides access to an early intermediate progenitor capable of generating various neuronal lineages, making it the prevailing approach for promoting efficient and thorough neural conversion³⁰. Based on the aforementioned methodology, a unique differentiation approach has been developed through the isolation of rosette neural aggregates (termed RONAs). This approach generates human FOXG1-positive forebrain neural progenitor cells that are capable of differentiating into six-layer cortical neurons and forming functional excitatory-inhibitory (E-I) neuronal networks resembling those found in the human cortex²⁸. However, the differentiation of RONA-derived NPCs into functional cortical neurons necessitates a relatively protracted timetable (8 weeks of neuronal differentiation in culture or 10 weeks of transplantation in newborn rats)^28,31,32, which impedes their potential benefits in neuronal replacement therapy. Several pioneering studies have suggested that the use of combinatorial small molecules in culture can expedite the induction and maturation of peripheral sensory neurons and cortical neurons within the central nervous system^33,34. While this approach has great potential to generate deep-layer cortical neurons with mature electrophysiological properties at a very early stage (16 days of neural differentiation) of differentiation³⁴, no inhibitory neurons are generated, and thus the diverse types of cortical neurons needed to replace those lost in brain diseases affecting the human cerebral cortex are lacking.

To achieve efficient neuronal replacement and functional repair of the injured cortex, we utilized a neural differentiation strategy to generate FOXG1-positive forebrain progenitors with the potential to differentiate into mixed cortical neurons, including excitatory and inhibitory neurons, and to undergo expedited functional maturation simultaneously by integrating the strengths of the two differentiation strategies discussed above^28,34. Furthermore, we investigated the in vivo cortical neuronal differentiation, functional maturation, synaptic integration, and therapeutic benefits of these FOXG1 precursors in stroke-damaged brains. Single-nucleus RNA sequencing (snRNA-seq), immunoelectron microscopy, electrophysiology, and viral tracing collectively confirmed the efficient cortical neuronal differentiation, functional maturation, and synaptic integration of this unique forebrain NPC at an early stage of transplantation (as early as 7 weeks). Importantly, transplanted FOXG1 NPCs promoted sensorimotor function recovery and reduced the incidence of chronic secondary cortical seizure development in a rat model of stroke. This neural differentiation strategy thus provides a valuable source of cortically fated cells for neuronal replacement therapy in ischemic cortical stroke and other neurological diseases affecting the cortex.

Results

Generation of forebrain NPCs derived from an optimized neural differentiation strategy

The FOXG1-positive forebrain NPCs were derived from the human iPSC line DYR0100, which is based on the enhancement (CN patents 202110505838.6 and 201810298488.9) of the RONA differentiation method previously reported by Xu et al.²⁸. Briefly, human iPSCs were cultured in mTeSR1 medium for 4 days at 37 °C, then disassociated and cultured in mTeSR1 medium at 37 °C to form embryonic bodies (EBs). On day 0 (EB formation), dual SMAD inhibition with noggin and SB431542²⁹ was initiated for 7 days, allowing for the robust formation of neural rosettes. The EBs were cultured in suspension for 7 days at 37 °C, then transferred and attached to Matrigel-coated plates. The attached EBs formed rosette neural aggregates after 14 days of culture at 37 °C in neural induction medium. These aggregates were subsequently manually separated using the same procedures reported by Xu et al.²⁸, and then cultured to form neurospheres in neurobasal medium with 2% B27 supplement and 1 mM GlutaMAX-I. The next day, the neurospheres were disassociated into single cells and seeded onto poly-D-lysine-coated plates, where dissociated NPCs were cultured in differentiation and maintenance medium containing a cocktail of small-molecule inhibitors, including SU5402, BIBF1120, and IBMX, for rapid maturation (Fig. 1a, b). Specifically, SU5402 and BIBF1120 were introduced to inhibit the fibroblast growth factor (FGF) and vascular endothelial growth factor (VEGF) receptor signaling pathways, thereby preventing cell aggregation, promoting the migration of immature neurons, and facilitating neurite outgrowth and subsequent maturation into functional neurons. Furthermore, given that neuronal differentiation and maturation require substantial levels of cAMP for energy provision³⁵, and IBMX acts as a non-specific inhibitor of the intracellular cAMP-degrading enzyme phosphodiesterase (PDE), which functions optimally in the presence of glucose, the culture medium was supplemented with IBMX and glucose to elevate intracellular cAMP and support rapid NPC maturation. Additionally, timed administration of retinoic acid in this stage was used to facilitate the development of an appropriate balance between excitatory and inhibitory neuronal populations according to the previously reported RONA differentiation protocol²⁸. After 4–7 days of culture at 37 °C, the NPCs were separated into single cells and harvested in neurobasal medium at a density of 1.0 × 10⁵ cells/μL for intracranial injection.

Fig. 1: Induction of forebrain NPCs from human iPSC based on an optimized neural differentiation strategy.

a Schematic summary of optimized neural differentiation approach and culture timeline. mTeSR, modified TeSR™ medium. Nog, noggin. SB, SB431542. EB, embryonic body. NIM, neural induction medium. NB, Neurobasal. RA, retinoic acid. E / I network, excitatory / inhibitory network. The time when human iPSC (hiPSC) clones are dissociated to form EBs is defined as Day 0 of the entire differentiation process. To facilitate comparison with the previous RONA method, the time of forebrain NPC formation (D28) is designated as the starting point (t = 0) for neural precursor differentiation into neurons. b Representative bright-field images of the same rosette neural aggregates captured on days 0, 1, 21, 22, and 30. Scale bar, 200 μm. The experiment was independently repeated 5 times with similar results. c–g Expression and quantification of neural stem cell marker SOX2 (c) and early forebrain regionalization markers FOXG1 (d), PAX6 (e), LHX2 (f), NKX2.1 (g) in the stage of NPCs formation. Colors are indicated in images. Scale bar, 50 μm. Data are means ± SEM (n = 3 samples). h Neuronal differentiation of forebrain NPCs produced by our differentiation strategy. DAPI (blue) was used to counterstain nuclei. Colors are indicated in the images. Scale bar, 50 μm. i Time-course immunostaining analysis of TUJ1⁺ cells in total cell population at weeks 1, 2, 4, and 8 of differentiation (n = 3 samples in each group). Data are means ± SEM. **P = 0.003, ***P < 0.001, one-way ANOVA followed by Tukey’s post hoc test. Source data are provided as a Source Data file.

Before grafting, we first evaluated the potential of these FOXG1 NPCs to differentiate into excitatory and inhibitory cortical neurons using multiple neural precursor subtype markers. Immunofluorescent staining at 32 days after neural differentiation demonstrated the overall distribution pattern of SRY (sex determining region Y)-box 2 (SOX2) (pan-neural precursor cell marker, 87 ± 3%), FOXG1- (forebrain marker, 96 ± 1%), paired box 6 (PAX6, 79 ± 3%) and LIM homeobox 2 (LHX2, 82 ± 18%) (dorsal excitatory neuronal lineage marker) as well as NK2 homeobox 1 (NKX2.1, 33 ± 7%) (medial ganglionic eminence marker) positive neural progenitors, indicating that the human iPSCs transformed into high-purity FOXG1-positive forebrain neural progenitors (more than 95%) with the potential capability of differentiating into both excitatory and inhibitory neurons (Fig. 1c–g).

Using the aforementioned differentiation method, we observed similar expression patterns of SOX2 (95 ± 3%), FOXG1 (97 ± 1%), and PAX6 (88 ± 1%) in another iPSC line, HOP06-C8-06, strengthening the reproducibility of this differentiation approach across different iPSC cell lines (Supplementary Fig. 3a–d). At the same time, we evaluated an alternative cortical neural induction method in our experimental system, which employs markedly different differentiation procedures, specifically the monolayer differentiation method. This approach allowed us to directly compare its neural differentiation efficiency with that of our differentiation strategy. Unfortunately, the proportion of PAX6⁺ cortical precursors was significantly lower (30 ± 5%) compared to the previously reported 95%²⁶. Additionally, consistent with prior studies²⁶, we did not detect the expression of NKX2.1, a marker for inhibitory neuron precursors, in our monolayer differentiation system. (Supplementary Fig. 5a–f).

Here, for facilitating comparison with the previous RONA method²⁸, the formation time of forebrain NPCs (D28) has been set as the starting point (t = 0) for their subsequent cortical neuronal differentiation. Four weeks after neuronal differentiation, the cultures were composed primarily of β-tubulin III (TUJ1)-positive neuronal cells (90 ± 3%) (Fig. 1h, i). Additionally, at 2 weeks of neuronal differentiation, our reported method in another human iPSC line, HOP06-C8-06, robustly yielded approximately 90% TUJ1-positive neurons (Supplementary Fig. 3e–f). Furthermore, at 8 weeks of neuronal differentiation, we observed ~70% MAP2-positive mature neurons (Supplementary Fig. 3g, h). In contrast, the monolayer differentiation method produced an average neuron proportion of approximately 70% (Supplementary Fig. 5g and h), which is consistent with the proportion previously reported²⁶. This suggests that our reported differentiation method may have a stronger potential for neuronal differentiation.

Differentiation of balanced cortical subtype composition and functional E-I networks

Two weeks after in vitro neuronal differentiation, TUJ1-positive progeny neurons began to express a significant proportion of deep-layer cortical markers, specifically TBR1 (layers I, V, and VI; 65 ± 10%), while only a minor fraction expressed upper-layer cortical markers, such as BRN2 (layers II-IV; 21 ± 7%) (Fig. 2a). These findings indicate that in vitro-derived neurons follow a temporal order of projection neuron production, with deep-layer neurons preceding upper-layer neurons, consistent with several previous cortical precursor differentiation studies^15,26,34. Over time, the expression of TBR1 decreased (28% ± 10%), whereas BRN2 (41% ± 5%) and SATB2 (40% ± 8%) showed an increasing trend, ultimately resulting in relatively balanced distribution of deep and superficial layer neurons within this system (Fig. 2a and Supplementary Fig. 4b), in contrast to many differentiation protocols that result in reduced production of upper-layer neurons. At 8 weeks of neuronal differentiation, the proportions of differentiated cells expressing the excitatory neuronal marker VGLUT and the inhibitory neuronal marker VGAT were 83 ± 10% and 16 ± 1%, respectively (Fig. 2b), closely resembling the intrinsic composition of the cerebral cortex. These results are similar to the proportion of cortical subtypes obtained by the RONA differentiation method reported previously²⁸. Furthermore, by using our developed differentiation protocol, we observed a similar proportion of cortical layer markers expression in neurons differentiated from the HOP06-C8-06 line compared with those derived from our primary DYR0100 line, while minor statistical differences exist in the expression levels of certain cortical markers (SATB2 for superficial layers and TBR1 for deep layers) (Supplementary Fig. 4b–d). This result strengthens the reproducibility of our developed differentiation method.

Fig. 2: Generation of balanced cortical subtypes and functional E-I networks.

a Expression and quantification of markers of superficial (BRN2) and deep (TBR1) cortical-layer neurons at 2 and 8 weeks of neuronal differentiation. Colors are indicated in images. Data are means ± SEM (n = 3 samples in each group). *P = 0.0156 for BRN2 and *P = 0.0107 for TBR1, two-tailed unpaired t-tests. Scale bar, 50 μm. b Expression and quantification of markers of excitatory (VGLUT) and inhibitory (VGAT) neurons at 4 and 8 weeks of neuronal differentiation. The zoomed-in view shows detailed fluorescence signal information. Colors are indicated in images. Data are means ± SEM (n = 3 samples in each group). *P = 0.047, ***P = 0.0004, two-tailed unpaired t-tests. Scale bar, 50 μm. c Action potential firing traces at 1, 2, and 4 weeks of neuronal differentiation without current injection and with −10 pA step current injection (down). d Quantification of percentage of cells with indicated firing frequencies at 1, 2, and 4 weeks of neuronal differentiation without current injection and with −10 pA step current injection (down). n = 11, 14, and 15 cells from three batches of independent cell cultures at 1, 2, and 4 weeks of neuronal differentiation. e Quantification of single-cell electrophysiological properties at 1, 2, and 4 weeks of neuronal differentiation. Solid dots denote individual cells. n = 11, 14, and 15 cells from three batches of independent cell cultures at 1, 2, and 4 weeks of neuronal differentiation. Data are means ± SEM. **P = 0.0022, ****P < 0.0001, one-way ANOVA followed by Tukey’s post hoc test. f Left: Representative traces of spontaneous excitatory current (sEPSC) in cultured human cortical neurons. Right: Summary pie chart of proportion of cells receiving sEPSCs (n = 11 cells from three samples) at 2 weeks of neuronal differentiation. g Left: Representative traces of miniature inhibitory postsynaptic current (mIPSC) in cultured human cortical neurons. By selectively blocking GABA receptors, bicuculine (10 μM) blocks mIPSCs, indicating inhibitory synaptic currents. Right: Summary pie chart of proportion of cells receiving mIPSCs (n = 9 cells from three samples) at 4 weeks of neuronal differentiation. Source data are provided as a Source Data file.

We next addressed whether rapid in vitro maturation occurs at 1, 2, and 4 weeks of neuronal differentiation, such as the ability to fire action potentials spontaneously or following induction by current injection. Previous research has shown that RONA-derived neurons are functionally mature after 8–10 weeks of neuronal differentiation^28,31,32. Here, single spontaneously fired action potentials were detected within one week of in vitro differentiation. After extending the culture time to 2 and 4 weeks, neuronal spike firing gradually increased, with more regular spontaneous firing (Fig. 2c, d). The results obtained from multi-electrode assay recordings are consistent with those obtained from single-cell patch-clamp recordings (Supplementary Fig. 2a). Additional parameters of neuronal maturation included hyperpolarized resting membrane potential, reduced membrane input resistance, and increased maximum action potential firing frequency (Fig. 2e). Using whole-cell patch-clamp recordings, postsynaptic currents were detected to determine excitatory and inhibitory synaptic function. Spontaneous excitatory postsynaptic currents were recorded 2 weeks after NPC differentiation, indicating that excitatory synaptic inputs into neurons (Fig. 2f). At 4 weeks after NPC differentiation, miniature inhibitory postsynaptic currents were also recorded, indicating that relative slower differentiation of interneurons comparing with excitatory neurons (Fig. 2g). Further morphological analysis showed co-localized expression of pre- and post-synaptic markers (synapsin and PSD95, respectively) at 8 weeks post-differentiation (Supplementary Fig. 2b), confirming the establishment of an excitatory and inhibitory neuronal network in vitro. Electrophysiological detection was also performed on neural precursor cells using RONA at the same time points (Supplementary Fig. 1a-e), and the results showed that the cells began to show significant functional maturation at 4 weeks of neuronal differentiation, while cells capable of generating action potentials were very rare at 1–2 weeks of neuronal differentiation (Supplementary Fig. 1b and c). In contrast, with the differentiation method reported in this study, more than 50% of the patched cells exhibited the ability to generate action potentials at 2 weeks (Supplementary Table 1). Additionally, compared to the RONA method, the progeny neurons generated by this method had lower membrane input resistance and higher resting membrane potential (Supplementary Table 1. The chi-square test supported the statistical differences in the occurrence probabilities of action potentials in progeny neurons undergoing the two differentiation methods at the same time point of differentiation (Supplementary Table 1). The above evidence strongly demonstrates the potential of this optimized method to accelerate the maturation of progeny neurons.

Single-nucleus RNA-sequencing (snRNA-seq) analysis of forebrain NPC-derived cell subtypes in vivo in rats after stroke

We next examined the differentiation of NPC-derived cortical neurons in a rat model of stroke. On days 4–5 after photothrombotic infarction, magnetic resonance imaging (MRI) was carried out to confirm the precise location and extent of the infarct region in vivo. on day 7 after the establishment of the focal cortical stroke model, we transplanted 2 × 10⁵ forebrain NPCs into four injection sites 1 mm to the left of the boundary of the stroke lesion identified by T2-MRI scanning (Fig. 3a and Supplementary Fig. 15b). At 11 weeks post-transplantation, we conducted snRNA-seq to characterize the fate of forebrain NPC-derived cells in a pre-clinical rat model of stroke. In three stroke-injured brains, we obtained 15, 399 NPC-derived human single nuclei and performed Uniform Manifold Approximation and Projection (UMAP) and graph-based clustering analysis (Fig. 3a).

Fig. 3: In vivo differentiation of grafted forebrain NPCs in focal photothrombotic model of stroke.

a Schematic timeline for experimental design. PT, photothrombotic. PET, positron emission tomography. MRI, magnetic resonance imaging. BLI, bioluminescence imaging. IEM, immunoelectron microscopy. EEG, electroencephalography. b Graph-based clustering of grafted cells from forebrain NPCs by snRNA-seq (n = 3 rats) at 11 weeks post-transplantation. c, d Five (c) and 7 (d) clusters showing detailed cell-type annotation utilizing singleR. RG, radial glia cell; DPC, diving progenitor cell; IPC, intermediate progenitor cell; EN, excitatory neuron; IN, inhibitory neuron. ULN, upper layer cortical neuron. DLN, deep layer cortical neuron. The pie chart showing the proportion of seven different cell types derived from forebrain NPCs (n = 3 rats) at 11 weeks post-transplantation. e UMAP plot showing selected marker gene expression of different cell types derived from grafted forebrain NPCs. f, g Representative images (f) and quantification (g) of NeuN-positive neurons derived from forebrain NPCs at 8 weeks after transplantation. Data are means ± SEM (n = 3 rats). Scale bar, 20 μm. h, i Pseudo-time trajectories of engrafted forebrain NPCs using diffusion map, colored by identified subpopulation. Diffusion map was constructed using Monocle2 (h) and Slingshot (i). Source data are provided as a Source Data file.

In total, 21 cell clusters (clusters 0–20) were identified (Fig. 3b, Supplementary Fig. 6a and Supplementary Data 1) based on the snRNA-seq data. The vast majority of clusters exhibited the expression of forebrain markers (FOXG1, Supplementary Fig. 7b), thus confirming the remarkable potential of the transplanted cells to differentiate into forebrain neurons with great efficiency. Most clusters (clusters 0, 2–9, 11, 12, 14–18, and 20) were highly expressed cortical layer-specific marker genes (SATB2, CUX2, ROBO1 and BCL11B) and inhibitory marker genes (GAD1, GAD2 and ERBB4), and were assigned to cortical neurons (CN, 84.1%) (Fig. 3b–e). The remaining clusters were assigned to a non-neuronal group. Among all cortical neuronal clusters, one cluster (cluster 12) enriched in GAD1, GAD2, and ERBB4 was annotated as interneurons (IN, 3.0%). Sixteen clusters (clusters 0, 2–9, 11, 14–18, and 20) showing high SLC17A6 (vGLUT2) and SLC17A7 (vGLUT1) expression were annotated as excitatory neurons (EN). Among all EN clusters, cluster 2, 3, 5, and 14 showed high and common expression of upper-layer neuronal (ULN, 25.7%) markers of SATB2 and CUX2 expression. Cluster 4, 8, 9, 16, 18, and 20 showed high and common expression of deep-layer neuronal (DLN, 26.7%) markers of ROBO1 and BCL11B (CTIP2). Among all non-CN clusters, clusters 1 and 19 showed high expression of HES1, HOPX, NOTCH, GFAP and AQP4, were assigned to radial glia cells (RG, 9.9%). Cluster 10 showed enrichment in cell-cycle genes (RRM2, MKI67, TOP2A and ASPM) and was assigned to dividing progenitor cells (DPC, 3.0%). Cluster 13 showed high expression of EOMES and PPP1R17 was annotated as intermediate progenitor cells (IPC, 3.0%). (Fig. 3b–e). In addition, all three independent samples consisted of consistent major cell clusters, including radial glial cells, dividing progenitor cells, intermediate progenitor cells, and various types of cortical neurons (Supplementary Fig. 7a). Staining results revealed that the majority of transplanted human cells, as identified by HuNu (a human-specific nuclear marker), expressed the neuronal NeuN marker (84% ± 4.3%, Fig. 3f, g). This proportion is similar to that of cortical neurons in total cells, indicating a high efficiency of forebrain NPC differentiation into cortical neurons. Furthermore, pseudo-time analysis using Monocle2 and Slingshot confirmed that the differentiation trajectory of the transplanted FOXG1 NPCs began with radial glia cells and dividing progenitor cells, then differentiated into intermediate progenitor cells, and finally terminated in excitatory neurons (Fig. 3h, i).

To further validate the cell types differentiated from forebrain NPCs after transplantation, we performed immunohistological staining of forebrain and cortical neuronal markers in NPC-derived cells. Immunostaining results showed that transplanted NPC-derived cells highly expressed forebrain and cortical layer-specific neuronal markers (Supplementary Fig. 8a-e). A small number of forebrain NPC-derived cells expressed interneuronal markers GABA (Supplementary Fig. 8f). These snRNA-seq and histological results demonstrate that the optimized differentiation approach-based human iPSC-derived FOXG1 NPCs possess excellent cortical neuronal differentiation potential after transplantation in a focal cerebral infarction rat model.

Functional maturation and integration of forebrain NPC-derived neurons in host neural network post-transplantation

In the snRNA-seq experiments, we also found that cortical neurons expressed KCND2 (clusters 0, 8, and 15) and KCNIP4 (clusters 0, 8, 15–18) (Supplementary Fig. 7c), which encode the voltage gated potassium channel Kv4.2 and mediate the rapid activation and inactivation of voltage-dependent outward K⁺ currents, and are thus important for neuronal action potential firing properties and excitability^36,37. In addition, cortical neurons also expressed genes important for synapse development (TENM2, clusters 2–9, 11, 15–18, and 20)³⁸, postsynaptic density formation (DLGAP2, clusters 2–9, 11–12, 15–18, and 20)³⁹, AMPA glutamatergic synapse function (GRIA2, clusters 0, 2–9, 11, 14–18, and 20)⁴⁰ (Supplementary Fig. 7d), neuronal dendritic patterning (CNTN5, clusters 0, 2, 4, 5, 6, 8, 9, 12, 14–18, and 20)⁴¹, and axonal outgrowth (CDH12, clusters 3, 7, 11, 16–17; SLIT3, clusters 3, 6, 7, 8, 9, 11, 16–18; EFNA5, clusters 3, 4, 6, 7, 9, 11–12, 15, 18, and 20)^42,43,44 (Supplementary Fig. 7e). Moreover, the dynamic genes in pseudo trajectory were enriched in synaptic (Supplementary Figs. 9 and 10) and axonogenesis pathways (Supplementary Figs. 11 and 12). These results imply that transplanted human forebrain progenitors may possess the capacities of action potential firing, neurite growth, and synapse formation of functional neurons.

Prior research has predominantly concentrated on evaluating synaptic alterations subsequent to NPC transplantation through ex vivo methodologies. However, there exists a dearth of noninvasive in vivo techniques to appraise synaptic changes following NPC transplantation, thereby limiting the possibility of clinical translational studies targeting synaptic regeneration and repair post-NPC transplantation. The advent of a category of PET molecular imaging agents that specifically target synaptic vesicle glycoprotein 2 A (SV2A) presents a promising opportunity for non-invasive observation of synapses within the brain⁴⁵. In this study, ¹⁸F-SynVesT-1 was prepared using trimethyltin precursors according to the reported method⁴⁶ (Fig. 4a). Furthermore, SV2A PET imaging results demonstrated that ¹⁸F-SynVesT-1 accumulation increased markedly in the NPC group compared with the vehicle control at 4 weeks post-transplantation, indicating that forebrain NPC transplantation can promote the regeneration of synapses in the transplanted region (Fig. 4b, c).

Fig. 4: Functional maturation and synaptic integration of NPC-derived cortical neurons post-transplantation.

a Chemical formula for the synthetic radiotracer ¹⁸F-SynVest-1 and a schematic diagram of its microPET scan. b, c Representative cerebral ¹⁸F-SynVest-1 PET images (b) and semiquantitative analysis (c) of the SUVr in vehicle and NPC groups (n = 3 rats in vehicle group, n = 3 rats in NPC group; Data are means ± SEM. *P = 0.021, two-way ANOVA followed by Sidak’s post hoc test). Images are shown in coronal view. The white lines depict the midline of the rat brain, while the black circles denote the transplanted regions. TP, transplantation. SUVr = SUV (transplanted region) / SUV (brain stem region). d Gene Ontology (GO) analysis of NPC-derived cell subtypes after transplantation (See also Supplementary Data 2). e Engrafted GFP/gold nanoparticle-positive grafted human neurons (marked in black, dense particles with red arrows) with Golgi apparatus (red arrowheads) and mitochondria (red asterisk). The experiment was independently repeated 6 times in two rats, yielding similar results. Scale bar, 2 μm. f–h Engrafted GFP/gold nanoparticle-positive grafted human neurons (marked in black, dense particles with red arrows) established asymmetric (f) and symmetric (g) synaptic contacts with rat host neurons. Scale bar, 500 nm. h Summary pie chart of proportion of synaptic types (n = 251 synapses from two rats) at 6 weeks after transplantation. i–k Engrafted GFP/gold nanoparticle-positive grafted human neurons (marked in black, dense particles with red arrows) established afferent (i) and efferent (j) synaptic contacts with rat host neurons. Scale bar = 500 nm. k Summary pie chart of proportion of synaptic types (n = 251 synapses from two rats) at 6 weeks after transplantation. l Left: Representative trace of action potential firing pattern of NPC-derived cortical neurons elicited by 50 pA depolarizing current injection. Right: Summary pie chart of proportion of recorded NPC-derived cells firing action potentials (n = 27 cells from a total of 4 rats) at 7 weeks post-transplantation (2 rats) and 11 weeks post-transplantation (2 rats). m Left: Representative traces of spontaneous excitatory (sEPSC) or inhibitory postsynaptic current (sIPSC), which were blocked by glutamate receptor antagonist NBQX (AMPAR antagonist, 5 μM) or GABA_A receptor antagonist bicuculine (10 μM), respectively. Right: Summary pie chart of proportion of cells displaying postsynaptic currents (n = 23 cells from a total of 4 rats) at 7 weeks post-transplantation (2 rats) and 11 weeks post-transplantation (2 rats). Source data are provided as a Source Data file.

Although the SV2A PET results indicate that the transplantation of NPCs leads to an augmentation in the quantity of synapses within the transplanted area, it is still unclear whether transplanted NPCs can re-establish new synaptic connections with the host neurons. In snRNA-seq studies, the top expressed gene in transplanted human cells is associated with synapse organization, axonogenesis, and synapse assembly according to GO enrichment analysis (Fig. 4d). In order to investigate the extent to which the human grafts had matured and integrated into the rat host brain circuits, we employed immunoelectron microscopy to identify synaptic connections between the grafts and host neurons. We analyzed 251 immunogold nanoparticle-labeled green fluorescent protein (GFP)-positive synapses in two rats. Most forebrain NPC-derived cells displayed the ultrastructures of mature neurons at 6 weeks post-transplantation, including mitochondrial-rich cytoplasm, rough endoplasmic reticulum, Golgi apparatus, and free ribosomes (Fig. 4e). Interestingly, we found most synaptic contacts (83%) were asymmetric with excitatory synapse structural characteristics, e.g., broad synaptic cleft, high postsynaptic density, and spherical synaptic vesicles (Fig. 4f–h). In addition, the engrafted neurons established bidirectional synaptic contacts with the host rat neurons, with more afferent synaptic inputs (61%) than efferent outputs (39%) (Fig. 4i–k). Thus, the above ultrastructural data suggested that the forebrain NPC-derived cortical neurons mainly established excitatory synaptic connections with host neurons, thereby receiving afferent synaptic inputs from the transplanted host rat brain.

We next explored whether forebrain NPC-derived cells could differentiate into functional neurons following cortical transplantation in rats with focal cerebral infarction. Whole-cell patch-clamp recordings were performed in acute brain slices at 7 and 11 weeks post-transplantation (Fig. 4l-m and Supplementary Table 2). Results showed that 93% of the 27 recorded forebrain NPC-derived cells fired action potentials upon depolarizing current injection (Fig. 4l). In addition, most recorded forebrain NPC-derived cells received postsynaptic currents, including spontaneous excitatory (sEPSCs, 78%) and inhibitory postsynaptic currents (sIPSCs, 4%) in 23 recorded forebrain NPC-derived neurons (Fig. 4m). These results demonstrated that transplanted forebrain NPC-derived neurons can establish effective functional synaptic connections with the host brain.

The thalamus sends prominent excitatory projections to the cortex and thalamocortical projections exert important physiological functions⁴⁷. Thus, we next explored whether transplanted forebrain NPC-derived neurons could receive direct projections from the thalamus in the host rat brain by AAV-mediated transsynaptic anterograde tracing⁴⁸. Initially, we injected Cre virus into the ventral thalamus ipsilateral to the infarct. Approximately 4–5 days later, we administered another loxp virus carrying mCherry into the transplantation site (Supplementary Fig. 13a). Notably, distinct mCherry signal expression was observed in host cells outside the transplantation zone (Supplementary Fig. 13b), indicating the effectiveness of this approach in tracing the inherent neural circuits between host neurons beyond the transplanted area and the ipsilateral thalamus. However, reliable mCherry signal expression was not detected within the transplantation zone (Supplementary Fig. 13c, some minor positive signals were attributed to nonspecific staining from dead transplanted cells). We then adopted various strategies to validate the feasibility of this research approach in tracing the circuit connectivity between the host thalamus and transplanted cells. Pre-transduction of neural progenitor cells with LOXP-mCherry via lentivirus was performed, followed by cre virus injection into the host thalamus six weeks after transplantation (Supplementary Fig. 13d). This strategy aimed to eliminate the possibility of ineffective loxp virus infection in progenitor cells. However, we still did not observe a reliable mCherry signal expression in the transplantation area (Supplementary Fig. 13e and f). To further investigate, we employed a more extreme method by directly injecting cre virus into the transplanted cell region previously infected with LOXP-mCherry (Supplementary Fig. 13g), eliminating the likelihood of failed neural circuit reconstruction across brain regions. Yet, effective signal expression remained undetected in the transplantation zone (Supplementary Fig. 13h and i). Considering the effectiveness of the AAV tracing strategy in inherent neural circuit tracing of the host, we hypothesized that the absence of mCherry signal expression in the transplantation area was not due to the failure of circuit reconnection but rather the possibility that AAV viruses carrying cre could not effectively infect or express cre recombinase in human cells.

Meanwhile, we adopted another classical retrograde circuit tracing scheme based on rabies virus (Fig. 5a). The starting neurons (GFP⁺/tdTomato⁺) and tracer neurons (GFP^-/tdTomato⁺) can be easily distinguished based on the different combinations of GFP and tdTomato signals (Fig. 5b). Our analysis revealed that, across different grafts (n = 3 rats), approximately 30% to 40% of GFP⁺ neurons were identified as starting neurons (tdTomato⁺). Interestingly, the presence of traced host cells (HuNu^-/tdTomato⁺) among the transplanted human cells (HuNu⁺) suggests that the transplanted cells can integrate into the host cortical tissue. Distinct tdTomato signal expression was observed in multiple brain regions adjacent to the transplantation site, including the contralateral cortex, ipsilateral thalamus, bilateral claustrum, globus pallidus, and ipsilateral amygdala (Fig. 5c–e and Supplementary Fig. 14). This finding suggests the establishment of afferent neural circuit connections between multiple host brain regions and transplanted cells, aligning with previous similar studies. The success of the RV tracing strategy further underscores the non-feasibility of the AAV tracing approach in tracing host-human neuron circuits, although the underlying reasons remain unknown.

Fig. 5: Rabies-mediated tracing of synaptic inputs from host neurons to engrafted human neurons.

a The strategy for tracing direct synaptic inputs to human neurons transplanted into the stroke-injured cortex. Green: the transplanted region; gray: the infarcted region. b Fluorescence photomicrographs show the expression patterns of GFP, HuNu, and tdTomato in human or host neurons at the graft site. The white arrowheads indicate GFP⁺/tdTomato⁺ starter neurons, the white long arrows indicate GFP^-/HuNu⁺/tdTomato⁺ rabies virus-traced human neurons, and the white short arrows indicate HuNu^-/tdTomato⁺ rabies virus-traced host neurons. The experiment was independently repeated 3 times with similar results. Scale bar = 20 μm. c–e Traced host neurons in the ipsilateral cortex (c), contralateral cortex (d), and multiple subnuclei of ipsilateral thalamus (e). Edge of the thalamus subnuclei are highlighted with white dotted line. VL: ventrolateral; VM: ventromedial. VPL: Ventral posterolateral. The experiment was independently repeated 3 times with similar results. Scale bars = 100 μm.

Furthermore, we examined another maturity indicator of transplanted cells, namely, cellular morphological characteristics (Fig. 6a-d). Cortical excitatory neurons exhibit a characteristic pyramidal morphology, marked by a prominent apical dendrite, while the majority of GABAergic interneurons display unpolarized or bipolar structures. To objectively determine neuronal morphology, we utilized the Pyramidal Morphology Index (PMI), as described in previous related studies^8,15. Quantitative results from morphological observations revealed that over 50% (53.36% ± 0.79%) of the cells exhibited pyramidal cell morphology, approximately 2% displayed bipolar cell morphology, and the remaining cells presented other types of morphologies or atypical pyramidal and bipolar cell shapes (Fig. 6d). Additionally, in the study of axonal projections from transplanted cells, we observed that eight weeks after transplantation, the transplanted human neurons projected fibers to multiple brain regions, with the frontal cortex, sensory cortex, claustrum, and white matter including the corpus callosum and internal capsule receiving the most projections (Fig. 6e and Supplementary Table 3). Moreover, a small number of fiber projections were detected in the striatum and hippocampal CA1 region (Supplementary Table 3). Immunofluorescence staining for cytoskeletal protein tau also confirmed the high expression of cytoskeletal protein in the transplanted cells (Fig. 6f).These distribution patterns are highly consistent with previous research⁴.

Fig. 6: Forebrain NPC transplanted in stroke-injured cortex exhibit cortical neuronal morphology and result in distant axonal projections across the brain.

a An overview of the relative positions of the infarct and the graft. Scale bar = 1 mm. b, c Representative fluorescent images of pyramidal-like cells (b) and bipolar cells (c) are shown with low (Scale bar = 50 um) and high (Scale bar = 20 um) magnification. The dashed rectangle at the bottom is shown at higher magnification above. The experiment was independently repeated 3 times with similar results. d Summary pie chart of proportion of grafted cells with different morphology at 8–10 weeks after transplantation (n = 3 rats). e Engrafted GFP/STEM121-positive grafted human neurons formed extensive axonal outgrowth in corpus callosum (CC), ipsilateral claustrum, bilateral frontal cortex (FrC), and bilateral internal capsule (IC). Green: the transplanted region; gray: the infarcted region. Scale bar, 100 μm. The experiment was independently repeated 3 times with similar results. f Axonal outgrowth of grafted human neurons, as measured by TAU and GFP co-localization. Scale bar, 50 μm. The experiment was independently repeated 3 times with similar results. Source data are provided as a Source Data file.

Taken together, SV2A PET imaging, snRNA-seq and neurobiological analysis confirmed that the optimized differentiation approach-based NPC-derived cells achieved early functional maturation and efficient integration into the stroke-injured cortex.

Therapeutic effects of transplanted forebrain NPCs on focal cerebral infarction

To assess whether these transplanted FOXG1 NPCs have therapeutic effects in the stroke-injured brain, we performed molecular imaging and behavioral studies showing the long-term survival and therapeutic value of the transplanted cells (Fig. 3a). Molecular imaging is of great value for the accurate diagnosis and assessment of the therapeutic effects of major neurological diseases^49,50. Here, ¹⁸F-FDG uptake and accumulation increased markedly in the NPC group compared with the vehicle control at 5 weeks post-transplantation, indicating that forebrain NPC treatment increased glucose metabolism in the infarcted area (Supplementary Fig. 15c and d). The NPCs group exhibited a significant reduction in infarct volume 4 weeks post-transplantation when compared to the vehicle control group, thus providing additional evidence of the therapeutic benefits of cell transplantation for focal cerebral infarction (Fig. 7a, b). Furthermore, compared with the vehicle control, the grafted forebrain NPCs significantly improved sensory and motor performance at 8 weeks post-transplantation, confirming the therapeutic effects of cell transplantation on neurological dysfunction (Fig. 7c, d).

Fig. 7: Therapeutic effects of transplanted forebrain NPCs on focal cerebral infarction.

a, b Representative MRI images of coronal brain slides (a) and quantitative analysis (b) of infarct volume in each group (n = 4 rats in vehicle group, n = 6 rats in NPC group; Data are means ± SEM. *P = 0.0277, two-way ANOVA followed by Sidak’s post hoc test). Coronal brain slides are arranged in sequence. The areas indicated by white arrow represent infarcted regions in ipsilateral hemisphere. TP, transplantation. c Sensorimotor function is measured by the percentage of touches made by the impaired limb compared to the ipsilateral and contralateral paw touches in the cylinder test. (n = 5 rats in sham group, n = 6 rats in NPC group, n = 7 rats in vehicle group; Data are means ± SEM. *P = 0.0233, two-way ANOVA followed by Tukey’s post hoc test). TP, transplantation. d Fine motor improvement, as determined by success rates (%) of single pellet reaching (SPR) (n = 5 rats in sham group, n = 6 rats in NPC group, n = 7 rats in vehicle group; Data are means ± SEM. *P = 0.0365, two-way ANOVA followed by Tukey’s post hoc test). TP, transplantation. e Typical EEGs and power spectrogram recorded from primary motor cortex (M1) adjacent to infarct in vehicle and NPC groups during post-stroke seizure. f Proportion of rats with EEG seizures and without EEG seizure in vehicle (n = 5 rats) and NPC transplantation (n = 8 rats) groups, respectively. g Representative PET images of rats with EEG seizures and without EEG seizure. White arrow points to M1 (primary motor cortex) area. h Comparison of ¹⁸F-FDG uptake in EEG recording area (M1 brain area) in rats with EEG seizures (n = 4 rats) and without EEG seizure (n = 13 rats, containing an additional 3 sham rats). Data are means ± SEM. ***P < 0.001, two-tailed unpaired t-tests. Source data are provided as a Source Data file.

Secondary epilepsy after cerebral infarction is a chronic brain dysfunction caused by ischemic injury⁵¹. To explore the potential enduring benefits of forebrain NPC transplantation on injured brain function, epileptic discharge activity was monitored in the peri-infarct area in rats at 12, 24, and 48 weeks post-NPC transplantation (Fig. 3a). The electroencephalogram (EEG) data demonstrated a 12% probability (1 out of 8 rats) of epilepsy onset in the forebrain NPC transplantation group, compared to 60% (3 out of 5 rats) in the vehicle control group, indicating great potential for the prevention of epilepsy after focal cortical stroke (Fig. 7e and f). In the control group, epileptic seizures occurred in 3 out of 5 rats, with a frequency of 491 ± 210 seizures per day. The typical waveform of individual ictal discharges had a frequency of 9 Hz and a duration of 28.8 ± 3.9 seconds. In the transplantation group, epileptic seizures occurred in 1 out of 8 rats, with a frequency of 221 seizures per day. The typical waveform of individual ictal discharges in this group had a frequency of 8 Hz and a duration of 23.7 seconds. Although the chi-square test (P > 0.05, Fisher’s exact test) did not reveal a statistically significant difference in the probability of post-infarction epilepsy between the two groups, the hazard ratio test suggested that the NPC group could reduce the onset probability by 21% compared to the vehicle group in animals with cerebral stroke. Interestingly, glucose uptake at the peri-infarct areas (main transplantation location) was higher in rats with seizures (P < 0.001) than in rats without seizure activity recorded by EEG (Fig. 7g, h).

Bioluminescence imaging allows real-time long-term tracking of engrafted NPCs⁵². Here, bioluminescence imaging demonstrated that the luciferase signals of forebrain NPCs and their progeny gradually increased and peaked at 12 weeks post-transplantation (Supplementary Fig. 16a–f), as supported by the presence of dividing progenitors based on snRNA-seq analysis (Fig. 3b–d). We conducted brain slice staining 8–10 months after transplantation and found that the grafts were still alive in most of (n = 7/9) the animals. Meanwhile, graft disappearance or remnants were also observed in a few (n = 2/9) animals (Supplementary Fig. 17). We speculate that this was not due to degenerative changes in the grafts themselves, but rather to the possibility that in some individuals of the SD rats we used, the immune rejection response could not be suppressed by immunosuppressants in the long term, leading to the death of the transplanted cells. Additionally, no tumor ¹⁸F-FDG uptake was detected in various organs of the rats through whole-body micro-PET scanning 48 weeks after transplantation and H&E staining did not find any suspicious heterotypic tumor cells (Supplementary Fig. 18 a–h, a′-h′ and Supplementary Movie 1), indicating the long-term safety of forebrain NPC transplantation in the brain. Together, these findings demonstrated the long-term survival and safety of forebrain NPC-derived neurons following transplantation into brains with stroke injury.

Discussion

Here, we reported an efficient strategy to generate human cortical fated neural precursor cells that exhibit substantial capacity for differentiation into cortical neurons, encompassing both excitatory and inhibitory neurons, and for undergoing expedited functional maturation concurrently in vitro (Fig. 8). Furthermore, we systematically evaluated the fate of the transplanted cells in stroke-injured brains, including their differentiation, maturation, survival, and synaptic integration, as well as their therapeutic value. Notably, we employed the non-invasive ¹⁸F-SynVesT-1 PET imaging technique with success to evaluate changes in synapse count pre- and post-transplantation therapy of FOXG1 progenitors in vivo, demonstrating the feasibility of this methodology for in vivo synaptic quantification in transplantation studies. Although the transplanted cells did not exhibit significant migration to the injury site or fill the cavity, the human neurons transplanted into the infarct-adjacent region were able to establish circuit connections with multiple regions of the damaged adult brain. Additionally, the transplanted FOXG1 forebrain progenitor cells not only promoted sensorimotor function recovery, but also reduce the onset probability by 21% compared to the vehicle group in animals with cerebral stroke. This study demonstrates the capacity of these FOXG1 precursors to serve as a viable option for neuronal replacement therapy in cases of ischemic cortical stroke.

Fig. 8: Summary of the improved strategy for neural differentiation and transplantation therapy using forebrain NPC.

In this study, the ROCK inhibitor, Y-27632, was used to promote the survival of hiPSC as single cells in this study. Dual SMAD inhibition by noggin and SB431542 promoted the differentiation of embryonic bodies (EBs) towards forebrain fates. Rosette neural aggregates were collected and maintained as neurospheres in neural induction medium (NIM). Then, timed administration of retinoic acid promoted the generation of an appropriate proportion of excitatory and inhibitory neuronal populations. Meanwhile, the additions of SU5402, BIBF1120, and IBMX facilitated neurite outgrowth and maturation into truly functional neurons. At 2 weeks after NPC differentiation in vitro, the derived cortical neurons exhibited action potential firing properties. At 4 weeks after NPC differentiation in vitro, the derived cortical neurons exhibited excitatory-inhibitory postsynaptic current activity. At 7 to 11 weeks after transplantation of NPCs into the stroke-injured cortex, the differentiated progeny cortical neurons matured and established functional synaptic connections with host cells (a). Transplanted cells (colored in green) are located adjacent to the cortical ischemic lesion (colored in gray). Red cells: host neurons located in the ipsilateral thalamus, ipsilateral and contralateral cortex.TH: thalamus. For clarity and as an example, thalamo-cortical afferent projections to the graft and bidirectional interhemispheric projections between the graft and host cortical neurons are presented. Different techniques are distributed in the areas of the brain where they are applied: (b) snRNA-seq, (c) electron microscope, (d) electrophysiology, (e) virus tracing.

Within our system of neural induction differentiation, we have successfully isolated a highly enriched population (>95%) of FOXG1-positive forebrain neural progenitors by utilizing the previous rosette neural aggregate differentiation approach. Additionally, timed administration of retinoic acid has facilitated the development of an appropriate balance between excitatory (~80%) and inhibitory neuronal populations (~20%) after 8 weeks of neuronal differentiation (Fig. 2b). Consistent with previously reported RONA methods²⁸, these findings indicate that our differentiation system effectively generates diverse and balanced cortical subtypes, including upper- and deep-layer excitatory neurons as well as inhibitory neurons. Comparing two distinct iPSC lines (HOP06-C8-06 and DYR0100) under identical differentiation conditions reveals minor statistical differences in the expression levels of certain cortical markers (SATB2 for superficial layers and TBR1 for deep layers), yet the overall proportion of differentiated cortical neuron subtypes remains consistent (Supplementary Fig. 4b–d). The balanced expression of diverse cortical neuron subtype markers in neurons derived from RONA methods represents a significant advancement compared to many other reported methods (Supplementary Data 3).

Furthermore, we incorporated several patented small-molecule inhibitors into the differentiation step to enhance neuronal maturation (CN patent 201810298488.9). The whole-cell patch-clamp recordings revealed that the cortical neurons derived from the optimized protocol exhibited a significantly higher proportion of action potential firing and spontaneous excitatory postsynaptic currents compared to those generated using the previous RONA method during the same time frame (1–4 weeks post-induction of neuronal differentiation) (Fig. 2c–f, and Supplementary Fig. 1a–f). This suggests that the optimized differentiation strategy facilitates more robust functional maturation of in vitro differentiated cortical neurons compared to the previous RONA method. Regarding cortical neurons derived from other differentiation methods, we recognize that each differentiation protocol varies greatly in the steps and timing required to induce pluripotent stem cells into cortical precursor cells (ranging from 9 to 32 days in vitro, Supplementary Data 3). Therefore, it is still challenging to make a direct comparison of the maturity level of cortical neurons in vitro.

The exceptional in vitro performance of FOXG1 NPCs must be translated into animal models, particularly disease models, to evaluate their in vivo fate and therapeutic potential of these cells. Here, the snRNA sequencing results showed the presence of progenitor cell subsets including RG, DPC and IPC, which was consistent with the previous scRNA-seq results of human cortical organoids⁵³, indicating that our forebrain NPC possess multidirectional differentiation potential similar to that observed in organoids. Moreover, snRNA-seq and histological analyzes demonstrated that human iPSC-derived FOXG1 NPCs based on the updated differentiation strategy predominately (>80%) differentiated into NeuN-positive cortical neurons, including upper- and deep-cortical layer neurons, 11 weeks after transplantation in rats with stroke, indicating robust cortical neuronal differentiation of engrafted FOXG1 precursors. Notably, a recent study using human cortical neuronal progenitors derived from human iPSC-derived long-term expandable neuroepithelial-like stem cells (lt-NES) showed that at 6 months post-transplantation in a stroke model, 41% of grafted human cells express the mature neuronal marker NeuN⁴ (Supplementary Data 4). Very rare, in a study where cortical neurons derived from an alternative differentiation protocol were transplanted into the healthy cortex of newborn mice, we noticed that another mature neuronal marker MAP2 proportion at 6 months after transplantation was close to our NeuN proportion¹⁵ (only qualitative fluorescence data) (Supplementary Data 4). Another study using combination therapy showed that physical exercise can promote the differentiation and maturation of transplanted NPCs⁵⁴. At 35 days after transplantation, the proportion of NeuN-positive mature neurons in the exercise group increased 2.5-fold (nearly 18%) compared with the transplantation group alone (nearly 7%). These results demonstrated that a higher proportion of mature cortical neurons can be obtained by simply optimizing neuronal differentiation before transplantation instead of applying additional conventional therapies such as physical exercise after transplantation into the infarcted brain.

There are differences in the proportions of cell subtypes between different samples by single-nucleus RNA sequencing analysis, for example, the proportion of deep-layer cortical neuron cluster (DLN) is higher in sample 2, while the inhibitory neuron cluster (IN) is not present in sample 3 (Supplementary Fig. 7a), possibly due to the relative small proportion of interneurons post-transplantation (Supplementary Fig. 7a and 8f). The variability could stem from differences in the efficiency of capturing various cell subtypes by 10x Genomics during individual single-nucleus RNA sequencing experiments. We further demonstrated a relatively balanced expression of upper- and deep-layer cortical neurons in the stroke-damaged cortex by immunofluorescent staining. Interestingly, the proportion of inhibitory neurons was far lower than that observed after 8 weeks of differentiation in vitro. Similar findings were also reported in a study on stroke using lt-NES-derived cortical precursor cells⁴ (Supplementary Data 4), suggesting that the microenvironment within the brain, especially under diseased conditions, is significantly different from that of the culture medium and may ultimately influence the differentiation trajectory of certain neuronal subtypes.

In the present study, we utilizes SV2A PET to assess the global alteration in synaptic amount within the infarct-adjacent region following transplantation of human neural progenitor cells. The SV2A is predominantly found on the presynaptic membrane of various synaptic structures within the brain. A positive correlation exists between the content of SV2A and the number of synapses, thereby rendering it a suitable imaging marker for synapses visualization and quantification⁴⁵. It should be noted that the changes in SUVr within SV2A PET before and after cell transplantation in the target region (Fig. 4b, c) suggest an increase in the number of synapses in the target area due to cell transplantation. However, this does not directly imply the establishment of direct synaptic connections between the transplanted neurons and host neurons, as SV2A is a broad-spectrum indicator of synaptic structure⁴⁵. In other words, the change in SUVr within SV2A PET could be attributed to the direct contribution of transplanted cells to synapse regeneration, including synapses established between the transplanted cells themselves and between transplanted cells and host cells. Alternatively, it could also represent an indirect contribution of transplanted cells to synapse regeneration, such as promoting the regeneration of endogenous synapses. Therefore, following the SV2A PET imaging, this study employed a series of more specific ex vivo research methods, such as immunoelectron microscopy, to confirm the establishment of effective synaptic connections between transplanted cells and the host. Nonetheless, SV2A PET remains the most ideal non-invasive, in vivo, and clinically applicable assessment tool for visualizing synaptic density⁵⁵. Additionally, the nanomolar and picomolar concentrations of PET molecular imaging probes, along with their ligand-receptor binding characteristics, confer high safety, sensitivity, and specificity⁵⁶. These attributes enhance the likelihood of regulatory approval for in vivo use, thereby increasing the potential for SV2A PET molecular imaging to become a preferred method for assessing synaptic regeneration of forebrain NPCs in clinical investigations.

In a series of pioneering cell replacement studies of focal cortical stroke, the authors demonstrated that the cortical-fated human lt-NES cells they used require at least 16–20 weeks post-transplantation to achieve efficient differentiation and integration into the stroke-damaged adult brain^4,8,9. In this study, we genereate a source of forebrain NPCs by optimizing previously established RONA differentiation strategy. These NPCs can differentiated into functional cortical neurons and achieve efficient synaptic integration into the stroked-injured adult brain at an earlier stage after transplantation (as early as 7–11 weeks). The electrophysiological data provided direct evidence for evaluating the maturation characteristics of grafted human neurons. Between 7 and 11 weeks post-transplantation, we observed that forebrain hNPCs exhibited a very high incidence of action potentials (93%) and postsynaptic currents (82% total) (Fig. 4l, m), which corroborated with the high proportion of mature neuronal marker (NeuN, 84%) expression. These findings also aligned with the broad expression of marker genes associated with neuronal maturation and synapse development identified through single-nucleus RNA sequencing (Supplementary Fig. 7c–e and Supplementary Fig. 9–12). This proportion was significantly higher than that observed in cortical-fated lt-NES cells at 20–25 weeks post-transplantation (21% incidence of action potentials, with only evoked postsynaptic currents detected and spontaneous postsynaptic currents almost undetectable) (Supplementary Data 4).

At 7 weeks post-transplantation, the resting membrane potential (RMP) of human neurons stabilized and approached the typical potassium equilibrium potential (approximately −70 mV), indicating maturation of the potassium channel system and energy metabolism system. The widespread expression of potassium channel genes such as KCND2 and KCNIP4 demonstrated through single-nucleus RNA sequencing (Supplementary Fig. 7c) provided powerful evidence at the genetic level. Conversely, the resting membrane potential of cortical-fated lt-NES cells remained below (−50.8 ± 5.7 mV) the typical potassium equilibrium potential even at 20–25 weeks post-transplantation. During early neuronal development, membrane input resistance (R_in) is typically high due to smaller cell size and incompletely developed ion channels. As neurons mature, cell size increases, and the type and density of ion channels change, typically resulting in a decrease in R_in⁵⁷. In this study, R_in was significantly reduced at 11 weeks post-transplantation of forebrain hNPCs, approaching the hundreds of MΩ observed in mature host neurons. However, cortical-fated lt-NES cells maintained R_in in the GΩ range even at 20–25 weeks post-transplantation⁸. Developing neurons have a relatively small membrane surface area and low capacitance. As neuronal morphology matures, the membrane surface area increases, leading to an increase in membrane capacitance⁵⁸. Notably, the membrane capacitance of our forebrain hNPCs had already reached levels similar to those of mature host neurons by 7 weeks post-transplantation. Conversely, transplanted cortical-fated lt-NES cells remained at a lower level (even slightly lower than non-fated lt-NES cells, although no significant statistical difference was observed) at 20–25 weeks. Additionally, Daniele Linaro et al. presented a detailed timeline of the electrophysiological properties of transplanted human cortical neurons using a different differentiation protocol (monolayer + dual SMAD inhibition + removal of FGF2) in newborn mouse cortex⁵⁹. Compared to their data at 1–2 months post-transplantation, our forebrain hNPCs also demonstrated superior performance in the aforementioned parameters.

Besides the passive membrane properties of neurons, the after-hyperpolarization potential (AHP) of action potential (AP) in our forebrain hNPCs had already approached the level of mature host neurons by 7 weeks post-transplantation. The AHP contributes to neuronal signal regulation, preventing excessive neuronal firing. Due to its close dependence on recording conditions, the AHP reported in different studies often varies considerably⁶⁰, making direct comparisons across different recording systems challenging. Moreover, the threshold potential and amplitude of APs in our forebrain hNPCs at 7–11 weeks post-transplantation (Supplementary Table 2) were similar to those observed in cortical-fated lt-NES cells at 20–25 weeks post-transplantation⁸. Taken together, the forebrain hNPCs generated using the differentiation protocol reported here exhibit faster and more excellent maturation characteristics of electrophysiological function.

It’s important to note that while our forebrain hNPCs showed excellent performance in parameters such as passive membrane properties and after-hyperpolarization potential of APs, they had not yet reached similar levels to mature host neurons in terms of the threshold potential and amplitude of APs, which are crucial for physical function. In contrast, the electrophysiological results recorded after 4 weeks of neuronal differentiation in vitro showed that the threshold potential and amplitude of APs were close to the level of mature neurons (Supplementary Table 1). This might reflect that the in vivo environment, especially the damaged cortex, could delay the complete maturation of transplanted neurons. A similar phenomenon has been confirmed in a comparative study at the same differentiation time point (10 weeks) in vitro and in vivo³¹. Additionally, the threshold potential and amplitude of APs observed by the pioneering research group at 20–25 weeks after transplantation into the stroke-injured cortex⁸ were significantly lower than the results recorded by another team in the neonatal healthy cortex at the same time point⁵⁹.

The results of FDG PET imaging revealed a noteworthy elevation in metabolism within the infarcted foci in the forebrain NPC transplantation group as compared to the control group. Despite the absence of significant transplanted forebrain NPC migration into the infarct foci (Supplementary Fig. 17a), it is hypothesized that the augmented metabolism in the infarct region may be attributed to compensatory recanalization of blood vessels⁶¹. Furthermore, it is suggested that the cell types present within the infarct region primarily consist of inflammation-associated glial cell proliferation rather than transplanted forebrain NPC and their progeny cells.

In terms of the impact of transplanted cells on neurological function, we examined two paradigms for evaluating sensory and motor function. One is the cylinder test, which assesses both motor and sensory forelimb function in stroke animals⁴. The other is the singel pellet retrieval test, which evaluates fine motor function, focusing on grasping behavior with paws. Both behavioral tests detected significant improvements at 8 weeks after transplantation. In a series of pioneering studies investigating cell replacement therapy for focal cortical stroke^4,8,9,62, transplanted cortical-fated lt-NES cells have demonstrated the ability to differentiate into relatively mature (both morphologically and functionally) cortical neurons. These neurons can partially reconstruct bidirectional circuit connections with host neurons and promote the recovery of neurological functions. However, the evidence chain linking graft activity to behavioral recovery remains incomplete. In the early stages of transplantation (8 weeks), cortical-fated lt-NES cells have not yet achieved functional maturity, although they can already receive neural projections from multiple brain regions of the host. Therefore, the authors speculate that the behavioral recovery observed in the early stages of transplantation is likely not due to a cell replacement mechanism. Instead, it is more probably attributed to non-cell replacement mechanisms such as the secretion of trophic factors, inflammation regulation, and stimulation of plasticity, collectively known as the bystander effect^8,9. This research group attempted to clarify the role of the cell replacement mechanism in long-term behavioral improvement during the later stages of transplantation (6 months), when the grafts achieve a higher degree of functional maturity as indicated in their previous study⁸. However, their use of optogenetic techniques to specifically inhibit graft cell activity did not reverse the expected recovery of sensorimotor function in the affected limbs. This suggests that other dominant mechanisms may be playing a key role at this stage⁴. In contrast, in the field of cell transplantation for Parkinson’s disease, multiple studies have demonstrated that graft activity directly regulates the improvement of rodent motor behavior at six months post-transplantation^63,64,65. We posit that focal cortical stroke involves a more diverse range of neuron types and more complex circuits compared to Parkinson’s disease, which specifically affects dopaminergic neurons in the substantia nigra and the nigrostriatal pathway. Furthermore, based on the electrophysiological data presented in the aforementioned studies, the AP threshold and amplitude of the grafts are still far from those of fully mature cortical neurons at six months post-transplantation in this study. This implies that the transplanted cells at this stage are still not fully capable of exerting their physiological functions. In other words, longer transplantation periods are required to observe the impact of graft activity on functional improvement.

Furtherly, we observed that at 16 weeks post-transplantation, the sensory and motor functions of the vehicle group (assessed via the cylinder test and SRP test) gradually recovered to levels similar to those of the NPC group (Fig. 7c, d). As mentioned above^8,66, the rodent’s sensorimotor functions exhibit a significant self-compensatory effect that enhances over time. It is pointed out that this is distinctly different from both human and rodent models of Parkinson’s disease. Indeed, the typical behavioral observation period in rodent models of focal cortical stroke is 5–8 weeks^8,10,54. The discordance between this relatively short behavioral observation window and the prolonged maturation time of grafts poses a significant challenge to studying the cell replacement mechanism in focal cortical stroke⁵⁴, making it a difficult and hotly debated topic in the field. Future efforts should be directed towards two main areas: developing more refined sensorimotor behavioral evaluation models that are less prone to self-compensation, and exploring more effective differentiation methods to accelerate graft functional maturity, particularly in enhancing action potential firing amplitude and frequency.

Although there is currently insufficient evidence in rodent models of focal cortical stroke to support the contribution of the cell replacement mechanism to behavioral improvement, it cannot be concluded that transplanted cortical neurons cannot exert a fundamental influence on behavioral improvement through this mechanism in humans affected by stroke. This is because in chronic stroke patients, self-compensation is almost impossible, and even rehabilitation training often fails to achieve substantial changes in motor function⁵⁴. Meanwhile, current clinical stem cell research, which primarily relies on the bystander effect⁴, has also not significantly improved patients’ motor symptoms. Therefore, cautious and continuous research on cortical neuron replacement remains crucial. Studies involving cell transplantation in primates appear to be a promising approach, as primates, like humans, have difficulty achieving self-compensation of motor function after stroke⁶⁷, making them suitable for long-term observations of the maturation of transplanted human cells.

Besides, it is pointed out that post-stroke epilepsy is very common in clinical practice and a good indicator of long-term disability after stroke⁶⁸, but has received little attention in stem cell therapy for cerebral infarction. Thus, we evaluated the incidence of seizure after stroke with or without NPC transplantation. After video-EEG monitoring at 12, 24, and 48 weeks poster stroke, the NPC group reduce the onset probability by 21% compared to the vehicle group in animals with cerebral stroke (Fig. 7e, f). Unfortunately, in this study, no statistically significant difference in the incidence of secondary epilepsy between the transplantation group and the vehicle group was observed through a chi-square test on a small sample of the overall incidence within one year.

Last but not least, forebrain neural precursor cells (NPCs) exhibited a degree of division and proliferation, as evidenced by single-cell sequencing analysis following 11 weeks of transplantation with a limited number of precursor cells. However, over time, cell numbers decreased according to bioluminescence data. Notably, neither whole-body FDG PET-CT nor H&E staining detected any evidence of tumor formation at the 12-month post-transplantation mark, indicating that the forebrain NPCs utilized in this study were safe for long-term engraftment within the brain.

In summary, we present an optimized strategy for cortically neural differentiation that integrates the benefits of existing neuronal differentiation protocols^28,29,34 to generate FOXG1-positive forebrain NPC with the potential to subdivide into dorsal (PAX6 and LHX2) and ventral (NKX2.1) territories. This FOXG1-positive NPC has the ability to differentiate into a high proportion of balanced upper-and deep-layer cortical neurons, as well as inhibitory neurons, exhibiting early functional maturation in vitro. Subsequently, we transplanted FOXG1-positive NPCs into the cortex affected by stroke. Through multiple methods, we comprehensively demonstrated that these forebrain NPCs robustly differentiated into a substantial proportion of balanced cortical neurons (including superficial and deep excitatory neurons as well as inhibitory neurons) and achieved excellent maturation characteristics in terms of NeuN marker expression, morphology and axonal projection, input circuit reconstruction, as well as electrophysiological passive membrane parameters and action potential after-hyperpolarization potential (7–11 weeks). Furthermore, these grafts contributed to a certain extent to the improvement of neurological function after stroke in the early stages of treatment, including sensory and motor impairments. The generation and utilization of FOXG1 forebrain progenitors present promising opportunities for neuronal replacement in various neurological disorders that impact the cortex, such as stroke, Huntington’s disease, Alzheimer’s disease, and amyotrophic lateral sclerosis.

Despite the aforementioned advantages of forebrain hNPCs derived from the optimized differentiation strategy, we also found that the infarcted adult cortex seemed to be unfavorable for the differentiation of grafts into inhibitory neuronal populations, and it also hindered the accelerated maturation of active membrane properties of action potentials (such as threshold potential and amplitude of action potentials). The mechanisms involved in these observations are currently unclear, and further research is needed to identify the key molecules or pathways that play a critical role and to provide specific regulation to maximize the advantages of forebrain hNPCs in differentiation and maturation. Additionally, this study only included male animals, and the potential impact of gender differences still needs to be considered. That is, the results and findings of this study are only applicable to male rats, and caution should be exercised when extrapolating these findings to female rats.

Methods

Animals

Sprague-Dawley (SD) rats (male, 2 months old, 300–350 g) were purchased from the SLAC Experimental Animal Company (Shanghai, China), then housed under a 12 h light/dark cycle in an air-conditioned room (21 ± 2 °C). Only SD male rats were used based on previous experience from related research and literature^4,8,9,10,54. All animals were provided with ad libitum access to tap water and standard chow. All experimental procedures and animal welfare protocols were approved by the Institutional Animal Care and Use Committee at Zhejiang University School of Medicine (Protocol No. ZJU20190068) and strictly adhered to the National Guidelines for Animal Protection.

Stem cell preparation

All neural progenitor cells (NPCs) in this study were generated from two distinct human iPSC lines: DYR0100 iPSC line (Catalog Number: SCSP-1301) obtained from the National Collection of Authenticated Cell Cultures in China; HOP06-C8-06 iPSC line established from a healthy female donor by Hopstem Bioengineering Co., Ltd., in China. Experimental procedures involving human pluripotent stem cells were conducted in accordance with ethical guidelines approved by the Institutional Animal Care and Use Committee of Zhejiang University School of Medicine (Protocol ZJU20190068).

Generation of forebrain NPCs using our developed protocol

The Foxg1-positive forebrain NPCs were derived from the human iPSC line DYR0100 or HOP06-C8-06 based on the updating differentiation method modified from Xu et al.²⁸. Briefly, human iPSCs were cultured on Matrigel (Cat # 354277, Corning) with mTeSR1 medium (Cat # 85850, Stemcell Technologies) for 4 days, then disassociated with TrypLE™ Express Enzyme (Cat # 12605010, Gibco) and cultured in low attachment plates with mTeSR1 medium to form embryonic bodies (EBs). The EBs were cultured for 7 days in suspension and then transferred and attached to Matrigel coated plates. The attached EBs formed Rosette neural aggregates with 14 days culture in neural induction medium containing Dulbecco’s Modified Eagle Medium/Nutrient Mixture F-12 (DMEM/F12, Cat # 11320082, Gibco), 1% N2 supplement (Cat # 17502048, Gibco), 0.1 mM NEAA (Cat # 11140050, Gibco), 1 mM GlutaMAX-I (Cat # 35050061, Gibco), and 2 μg/mL heparin (Cat # H3149, Sigma). The Rosette neural aggregates were then manually isolated to form neurospheres in neurobasal medium (Cat # 21103049, Gibco) containing 2% B27 supplement (Cat # 12587010, Gibco) and 1 mM GlutaMAX-I (Cat # 35050061, Gibco). The following day, the neurospheres were disassociated into single cells using Accutase (Cat # A1110501, Gibco) and placed on poly-D-lysine (Cat # P7886, Sigma) coated plates where NPCs were cultured in differentiation and maturation medium (CN patent 201810298488.9) for further development. The patent medium consisted of neurobasal medium, supplement without B27, 1 × 2-mercaptoenthanol, 0.2 mM ascorbic acid, 20 ng/ml brain-derived neurotrophic factor (BDNF), 20 ng/ml glial cell-derived neurotrophic factor (GDNF), 100 nM SU5402, 200 ng/ml BIBF1120, 10 μM IBMX, and 5 mM glucose. After 4–7 days of culture, the NPCs were disassociated into single cells using Accutase and collected in neurobasal medium at a density of 1.0 × 10⁵ cells/μL, which is suitable for intracranial injections. Specifically, the neural precursor cells used for transplantation in this study were at the 25th or 28th day of neural induction and differentiation (the 25th day when lentivirus transfection is not required, and when lentivirus transfection is needed, the culture period is extended by 3 days).

Generation of forebrain NPCs via the control RONA protocol

Neural differentiation of hiPSCs was based on the previously reported rosette neural aggregates (RONAs) method²⁸. Briefly, human iPSCs (specifically, the human iPSC line DYR0100) were cultured on Matrigel (Cat # 354277, Corning) with mTeSR1 medium (Cat # 85850, Stemcell Technologies) for 4 days, then disassociated with TrypLE™ Express Enzyme (Cat # 12605010, Gibco) and cultured in low attachment plates with mTeSR1 medium to form embryonic bodies (EBs). The EBs were cultured for 7 days in suspension and then transferred and attached to Matrigel coated plates. The attached EBs formed Rosette neural aggregates with 14 days culture in neural induction medium containing Dulbecco’s Modified Eagle Medium/Nutrient Mixture F-12 (DMEM/F12, Cat # 11320082, Gibco), 1% N2 supplement (Cat # 17502048, Gibco), 0.1 mM NEAA (Cat # 11140050, Gibco), 1 mM GlutaMAX-I (Cat # 35050061, Gibco), and 2 μg/mL heparin (Cat # H3149, Sigma). The Rosette neural aggregates were then manually isolated to form neurospheres in neurobasal medium (Cat # 21103049, Gibco) containing 2% B27 supplement (Cat # 12587010, Gibco) and 1 mM GlutaMAX-I (Cat # 35050061, Gibco). The following day, the neurospheres were disassociated into single cells using Accutase (Cat # A1110501, Gibco) and placed on poly-D-lysine (Cat # P7886, Sigma) coated plates where NPCs were cultured in differentiation medium for further experiments. The medium consisted of neurobasal medium, supplement with B27, 1 × 2-mercaptoenthanol, 0.2 mM retinoic acid, 20 ng/ml brain-derived neurotrophic factor (BDNF), 20 ng/ml glial cell-derived neurotrophic factor (GDNF), and 5 mM glucose. After 4–7 days of culture, the NPCs were disassociated into single cells using Accutase and collected in neurobasal medium. Maintain the NPCs for an additional 1, 2, and 4 weeks before characterization.

Generation of NPCs through the control monolayer protocol

Induction of forebrain neurons from induced pluripotent stem cells (iPSCs) was achieved using another well-established monolayer differentiation protocol as described by Shi et al.²⁶. Briefly, human iPSCs (specifically, the human iPSC line DYR0100) were cultured on Matrigel (Cat # 354277, Corning) with mTeSR1 medium (Cat # 85850, Stemcell Technologies) for 4 days. After cells reached ~100% confluency, media was changed to neural maintenance medium supplemented with 1 μM Dorsomorphin and 10 μM SB431542. Neuroepithelial cells were split at day 12 using dispase and re-plated in neural maintenance medium. Maintain the cells for an additional 1, 2, and 4 weeks before characterization.

Virus transfection

For viral transfection, NPCs (1.5 × 10⁵) differentiated from DYR0100 iPSCs through our optimized protocol were added to a 48-well plate, and lentivirus transfection was performed after 3 days of culture at 37 °C. Four lentiviruses carrying different target genes were used, including pLenti-CMV-EGFP-3Flag (titer: 1.3 × 10⁹ vg/mL, ObioTechnology, Shanghai, China) and pLenti-CMV-MCS-EF1a-copGFP (titer: 1.3 × 10⁹ vg/mL, TaiTool Bioscience, Shanghai, China) for labeling grafted progeny cells, pLenti-pCDH-EF1a-Firefly Luciferase-T2A-copGFP (titer: 1.3 × 10⁹ vg/mL, TaiTool Bioscience, Shanghai, China) for bioluminescence imaging tracking, VSVG-LENTAI-EF1Aht-DIO-mCherry-WPRE-pA (titer: 1.0 × 10⁹ vg/mL, TaiTool Bioscience, Shanghai, China) for AAV-mediated anterograde transsynaptic tagging, and VSVG-LENTAI-UbiC-Cre-EGFP-WPRE-PA (titer: 1.0 × 10⁹ vg/mL, TaiTool Bioscience, Shanghai, China) for rabies vector-mediated retrograde transsynaptic tagging. Briefly, 6.98 μL of viral vector was diluted to 400 μL in fresh medium. The vector was transferred evenly to two wells (200 μL/well) in a 48-well plate, followed by incubation for 2 h at 37 °C, and the addition of 1 mL/well of fresh medium. The plate was returned to the incubator for cultivation at 37 °C, with the culture medium exchanged with fresh medium after 24 h. During the preliminary lentiviral transfection experiment, we designed different gradients of multiplicity of infection (MOI) values for each lentivirus. The optimal MOI value for each lentivirus was set when the GFP expression efficiency (excluding VSVG-LENTAI-EF1Aht-DIO-mCherry-WPRE-pA) was close to 80% without exhibiting cytotoxicity. For VSVG-LENTAI-EF1Aht-DIO-mCherry-WPRE-pA, the mCherry fluorescent gene cannot be expressed due to the absence of cre enzyme. Our approach is to first infect NPC with VSVG-LENTAI-UbiC-Cre-EGFP-WPRE-PA and obtain the optimal MOI value for this virus through the expressed EGFP fluorescent tag. Then, we infect the NPCs, which have been determined with the optimal MOI value for VSVG-LENTAI-UbiC-Cre-EGFP-WPRE-PA, with different gradients of VSVG-LENTAI-EF1Aht-DIO-mCherry-WPRE-pA. The optimal MOI value for VSVG-LENTAI-EF1Aht-DIO-mCherry-WPRE-pA can be determined based on the expression of mCherry.

Ischemic brain injury

Cortical photothrombosis was induced in the sensorimotor cortex based on published protocols⁶⁹. In brief, after anaesthetization with 3% isoflurane, the rats were positioned in a stereotaxic frame. A cranial window (3 × 4 mm) was established over the unilateral sensorimotor cortex (3.5 mm lateral and 0.5 mm anterior to bregma), ensuring that the dura was kept intact. A small piece of opaque tin foil was used to allow light to pass through a 3 × 4-mm hole in the center to irradiate brain tissue. At 2 min before laser irradiation, the rats received an intravenous injection of 1 mL of rose bengal solution (15 mg/mL) via the tail vein. After the cranial window was covered by a masking sheet (3 × 4 mm), focal illumination (3.5-mm diameter focal spot, 25 mW, 532-nm CNI Laser (Changchun, China)) was applied for 15 min to induce occlusion, with medical adhesive then used to cover the skull. The experimental groups included stroke induction + vehicle injection, stroke induction + NPC injection, and sham (rose bengal injection only) groups.

TTC staining

Before the formal cell transplantation experiment, we utilized the 2,3,5-triphenyltetrazolium chloride (TTC) staining method to confirm the location and extent of the infarct. On the second day after cortical stroke modeling, cardiac perfusion was performed, and the brain was extracted. The brain tissue was then placed in a −20 °C freezer for 20 min to harden it, facilitating subsequent manual slicing. Simultaneously, the TTC solution (2 mg/mL) was incubated in a 37 °C oven for 20 min. The frozen brain was quickly removed, and the brain tissue within the corresponding range of the infarct was sliced into 3–4 sections, each with a thickness of approximately 1–2 mm. The brain slices were submerged in TTC solution, returned to the oven, and flipped for staining after approximately 15 min, ensuring that the staining process was conducted in a dark environment. The brain slices were then removed, placed on a solid-colored background, and photographed.

Cell transplantation under MRI guidance

Cell transplantation was performed in accordance with previous research¹⁰, with minor adjustments. Before transplantation, all rats underwent MRI scanning to determine the location and extent of infarction (see MRI scanning section). At 7 days after infarction, rats were anesthetized and positioned in a stereotactic device (RWD Instruments, China). In total, 2 × 10⁵ NPCs differentiated from DYR0100 iPSCs through our optimized protocol were injected into the cortex at four sites along the anterior-posterior axis (1 mm to the left of the boundary of the stroke lesion) in a volume of 2.0 μL using a microinjection pump (KD Scientific, USA) at a rate of 0.1 μL/min. Before each injection, the stem cell suspension was gently remixed using a 2.5-μL pipette to ensure a constant number of transplanted cells. After each injection, a 33-G Hamilton needle was maintained in place for 5 min, then gradually withdrawn to ensure maximum cell retention at the injection site. The injection coordinates were located at four different points: i.e., 2.5 mm anterior to bregma and 2.8 mm ventral to skull surface; 1.5 mm anterior to bregma and 2.8 mm ventral to skull surface; 0.5 mm posterior to bregma and 2.6 mm ventral to skull surface; and 1.5 mm posterior to bregma and 2.6 mm ventral to skull surface. Body temperature during all operative procedures was monitored (rectal probe) and maintained (heating pad). In this study, all rats in the sham, vehicle and NPC group received an intraperitoneal injection of the immunosuppressant cyclosporine A (MCE, 10 mg/kg) each day for one month and every other day for the remaining months of the experiment⁹.

Immunostaining

For culture immunostaining, the cells were rinsed using phosphate-buffered saline (PBS), fixed in 4% paraformaldehyde (PFA, 15 min), blocked using blocking buffer (10% donkey serum and 0.2% Triton X-100 in PBS), and incubated overnight (4 °C) with blocking buffer-diluted primary antibodies. The cells were then washed with blocking buffer (three times), incubated (1 h at room temperature) with corresponding donkey anti-mouse Alexa Fluor 647- or donkey anti-rabbit Alexa Fluor 546-conjugated secondary antibodies (1:500; ThermoFisher), and washed with blocking buffer (three times).

For brain slice immunostaining, the rats were first anesthetized with 1% pentobarbital sodium (0.5 mL/100 g), then rinsed transcardially with 0.9% saline and 4% PFA. The brains were removed and post-fixed overnight in 4% PFA at 4 °C. Coronal brain slices (40-μm thick) were excised with a sliding microtome (Leica VT1200S, Germany), permeabilized with 0.2% Triton X-100 for 10 min, blocked with PBS solution with 0.5% bovine serum albumin (BSA) and 10% normal goat serum for 1 h at room temperature, and incubated overnight with primary antibodies at 4 °C.

The primary antibodies included goat anti-SOX2 (1:200, SC-17320, Santa), mouse anti-NESTIN (1:1 000, MAB5326, Millipore), rabbit anti-FOXG1 (1:200, ab18259, Abcam), rabbit anti-PAX6 (1:200, 901301, BioLegend), rabbit anti-LHX2 (1:500, AB184337, Abcam), rabbit anti-NKX2.1 (TTF1, 1:200, ab76013, Abcam), mouse anti-TUJ1 (1:500, AB1637, Millipore), rabbit anti-GFAP (1:1 000, AB5804, Millipore), rabbit anti-MAP2 (1:500, ab183830, Abcam), rabbit anti-BRN2 (1:200, ab137469, Abcam), rabbit anti-SATB2 (1:200, ab51502, Abcam), rabbit anti-CTIP2 (1:200, ab187668, Abcam), rabbit anti-TBR1 (1:250, ab183032, Abcam), rabbit anti-VGLUT (1:500, ab77822, Abcam), rat anti-VGAT (1:500, sc-270411, Santo Cruz), rabbit anti-NeuN (1:200, ab177487, Abcam), rabbit anti-synapsin (1:500, ab64581, Abcam), mouse anti-PSD95 (1:500, M1511-4, Huabio), rabbit anti-GABA (1:500, Nbp2-43558, NOVUS), and mouse anti-Tau (1:500, Ab80579, Abcam). For identification of transplanted NPCs, the brain slices were co-stained with mouse monoclonal antibodies against human nuclei (HuNu) (1:20, MAB1281, Millipore) or mouse monoclonal antibodies against human cytoplasmic protein (STEM121) (1:500, Y40410, TaKaRa). After rinsing with PBS, the brain slices were incubated with the corresponding donkey anti-rabbit Alexa Fluor 546- or donkey anti-mouse Alexa Fluor 647-conjugated secondary antibodies (1:500, A10040 or A − 31571, ThermoFisher) and donkey anti-mouse Alexa Fluor 488, donkey anti-rabbit Alexa Fluor 546- or donkey anti-mouse Alexa Fluor 647-conjugated secondary antibodies (1:500, R37114, A10040, or A32787, Invitrogen) for 1 h at room temperature. After washing (PBS) and counterstaining (4, 6-diamino-2-phenylindole (DAPI) nuclear dye), the slices were fixed on glass slides and fluorescence images were taken using an inverted microscope (Axio Observer 3; Zeiss, Germany) or a Nikon A1 Ti confocal scanning microscope (Germany). The ratio of FOXG1, NeuN, BRN2, SATN2, CTIP2, TBR1 and GABA positive cells in the NPCs graft was determined based on the positive cell numbers within a sampling box (size: 1024 × 1024 pixels in confocal images at ×60 magnification), divided by the total number of human cells labeled with HuNu or the constitutive fluorescence within the same sampling box. For a given marker, three randomly selected fields from three randomly selected coronal sections were enumerated, thereby a total of 9 images for each subject were quantified. For the morphological analysis, we utilized the Pyramidal Morphology Index (PMI)^8,15. Specifically, using Fiji software, a 25-micron circular boundary (based on the scale bar in the image) was delineated around the soma (cell body) of each neuron. We subsequently quantified the number of neurites (N) intersecting this boundary and measured the width (L, μm) of the thickest dendrite at the point of intersection. The PMI was calculated as L divided by (N − 1). Neurons with a PMI cutoff of 1.2 were classified as pyramidal neurons. Neurons with N = 2 and a PMI less than 1.2 were classified as bipolar. The remaining cases were uniformly classified as neurons of other morphologies.

Single-cell nuclear extraction and snRNA-seq library generation and sequencing

At 11 weeks after grafting, the animals were anesthetized with pentobarbital sodium (1%, 0.5 mL/100 g) and perfused with ice-cold PBS. Brains were immediately removed and coronally sectioned with a vibratome (Leica, VT1200S) in ice-cold oxygenated (5% O₂ and 95% CO₂) artificial cerebral fluid (aCSF) containing (in mM): NaCl 92, KCl 2.5, MgSO₄ 2, CaCl₂·2H₂O 2, NaH₂PO₄·2H₂O 1.2, NaHCO₃ 30, HEPES 20, Na-ascorbate 5, Na-pyruvate 3, thiourea 2, and glucose 25. The GFP-positive regions in brain slices were micro-dissected under a fluorescence stereomicroscope (Leica VT1200S, Germany). To ensure the integrity of nuclear RNA, all solutions were kept on ice and centrifugation was operated at 4 °C. The dissected brain tissue was then homogenized in ice-cold RNAase-free homogenization buffer containing: 0.32 M sucrose, 5 mM CaCl₂, 3 mM MgAc₂, 0.1 mM EDTA, 10 mM Tris-HCl pH 7.6, 0.4 U/μL recombinant RNA inhibitor, 0.1 mM PMSF phenylmethanesulfonyl fluoride, 0.1 mM β-mercaptoethanol, 1% BSA, and 0.01% NP-40 in ultra-pure distilled water. Subsequently, 1.5 mL of 0.32 M sucrose solution containing 0.32 M sucrose, 5 mM CaCl₂, 3 mM MgAc₂, 0.1 mM EDTA, 10 mM Tris-HCl pH 7.6, 0.4 U/μL recombinant RNA inhibitor, 0.1 mM β-mercaptoethanol, and 1% BSA in ultra-pure distilled water were added to the homogenates, followed by centrifugation at 900 g for 10 min at 4 °C. After withdrawing the supernatant, 1.5 mL of 0.32 M sucrose solution was added to the tube, then resuspended, filtered (35-mm cell-strainer, Cat. No. 352235, Falcon, Corning, USA) and centrifuged at 900 g for 8 min at 4 °C. The precipitate was resuspended in 0.32 M sucrose solution and diluted with an equal volume of 50% OptiPrep density gradient medium (60% OptiPrep density gradient medium, 0.32 M sucrose solution) to give a final concentration of 25% medium solution and centrifuged by 3 000 g for 20 min at 4 °C. After removal of the supernatant, the nuclear pellets were dissolved in 100 mL of PBS. Nuclear density was analyzed using cell counting plates (Cat. No. 177-112 C, Watson, Fukae-Kasei, Japan). Single-nucleus capture (target 8 000 nuclei/sample) was performed using a single-cell 3’ Library and Gel Bead Kit V3.1 on the 10X Genomics platform (USA). Single-nucleus libraries from individual samples were pulled and sequenced using the Illumina HiSeq X Ten platform. Nucleus capture and library preparation protocols were performed following the manufacturer’s recommendations (10X Genomics, USA).

snRNA-seq data processing

Reads were aligned to a mixed reference genome (human (hg38) and rat (rn7)) using STAR in CellRanger v7.2.0 with default parameters. Utilizing the distinct genomic expression profiles of the transplanted human cells and the host rat cells, we categorized the extracted single-nucleus RNA sequencing (snRNA-seq) data into two distinct groups: one representing the transplanted human cells and the other representing the host rat cells. We ensured that only reads uniquely aligned to the human transcriptome were preserved for further analysis. A unique molecular identifier (UMI) counting matrix was generated and loaded into the R package Seurat (v4.0.0). Cells with less than 500 expressed genes or a minimum of 25% mitochondrial genes were discarded. Genes with less than five detected cells were removed. After filtering, a gene-barcode matrix of 15 399 cells and 41 905 genes was used for downstream analysis. Briefly, the raw UMI counts were normalized to total reads with log-normalization and scaled using a factor of 10,000. The top 2 000 highly variable genes selected by Variance Stabilizing Transformation (VST) were used to perform principal component analysis (PCA). Using 20 principal components (PCs), dimension reduction was implemented by Uniform Manifold Approximation and Projection (UMAP). Further cell clusters were identified based on Louvain-clustering. Differentially expressed genes (DEGs) in each cluster were ascertained using Seurat (“FindAllMarkers” function) with default parameters. Cell populations were identified using gene enrichment analysis based on cell-type and layer-specific marker gene sets obtained from published snRNA-seq datasets of the cerebral organoids, developing and adult human cerebral cortices^{53,70,71,72,73}.

Pseudo-time analysis across all clusters, except inhibitory neurons, was performed using the R package Monocle2. Briefly, ‘DDRTree’ dimension reduction was performed utilizing 15 PCs based on the top 2 000 highly variable genes. The top 1 000 DEGs across clusters (identified by Seurat) were used to order cells and construct the pseudo-time trajectory. Each ordered cell was assigned a pseudo-time value across pseudo-time trajectory. Pseudo-time genes were identified by Monocle2 with gene expression fit along pseudo-time, with significance at a false discovery rate (FDR) < 0.05. Smooth scatter plots were generated to show dynamic gene expression changes across pseudo-time trajectory. To confirm the accuracy of pseudo-time trajectory, we reperformed pseudo-time analysis using the R package Slingshot. In brief, a lineage dimension reduction implemented by ‘diffusion map’ was generated and loaded into Slingshot. Summarizing cell positions in ‘diffusion map’ embeddings and expression patterns of the top 2 000 highly variable genes, we constructed a pseudo-time trajectory across 21 clusters without supervision.

Electrophysiology

For cell culture electrophysiology, cells were cultured on coverslips, then transferred to a recording chamber perfused with oxygenated (95% O₂ and 5% CO₂) aCSF containing (in mM): 145 NaCl, 5 KCl, 2 CaCl₂, 10 HEPES, 2 CaCl₂, 2 MgCl₂, and 10 glucose for whole cell patch-clamp recording. In addition, neuronal activity was recorded using the microelectrode array⁶⁰ system with AxIS software (Axion Biosystems) according to manufacturer’s protocols. Spontaneous activity was recorded for 10 min and the spike-detection threshold was 5.5-fold the standard deviation of noise. The CytoView MEA 48-well plate (Cat # M768-tMEA-48W) for the Maestro MEA system contained 16 electrodes per well. Active electrodes were classified as those electrodes with an average of at least 5 spikes/min.

For brain slice electrophysiology, rats were anesthetized using 1% pentobarbital sodium (0.5 mL/100 g) and perfused with ice-cold oxygenated cutting aCSF containing (in mM): 194 sucrose, 30 NaCl, 4.5 KCl, 0.2 CaCl₂, 2 MgCl₂, 1.2 NaH₂PO₄, 26 NaHCO₃, and 10 glucose. After quick removal, the brains were sliced (300-μm coronal sections) with a vibratome (Leica VT1200S, Germany) in aCSF, treated for 25–35 min at 34 °C in oxygenated routine aCSF containing (in mM): 119 NaCl, 2.5 KCl, 11 glucose, 1.0 NaH₂PO₄, 26.2 NaHCO₃, 2.5 CaCl₂, and 1.3 MgSO₄, then maintained at room temperature until the experiment. Cell fluorescence was imaged using an Eclipse FN1 microscope (Nikon, Japan), with a 40× water-immersion lens and mercury lamp illumination.

To determine cultured cell and brain slice properties, recording pipettes were loaded with an intracellular solution consisting of (in mM): 125 K-gluconate, 15 KCl, 10 HEPES, 4 MgCl₂, 4 Na₂ATP, 0.4 Na₃GTP, 10 Tris-phosphocreatine, and 0.2 ethylene glycol tetraacetic acid (EGTA). Action potential recordings were obtained with current clamp configuration. Spontaneous EPSCs were obtained under voltage-clamp configuration, with membrane potential held at −70 mV. For miniature iPSC recordings, the extracellular solution contained 1 μM tetrodotoxin (TTX), 5 μM NBQX (AMPAR antagonist) and 50 μM D-AP5 (NMDAR antagonist).

A MultiClamp 700B amplifier and 1440 A digitizer (Molecular Devices, USA) were used for whole-cell patch-clamp recordings at room temperature. A series of current steps (50-pA steps, 400 ms) were applied to provoke action potentials. The N-methyl-D-aspartate (NMDA), aminomethylphosphonic acid, and γ-aminobutyric acid A (GABAA) receptors were blocked using D-AP5 (50 μM), NBQX (5 μM), and bicuculine (10 μM), respectively. Responses were filtered (2 kHz), digitized (10 kHz), and evaluated (pClampfit v10.4, Molecular Devices, USA; Mini-Analysis v6.0 Synaptosoft, USA).

Immunoelectron microscopy

Immunoelectron microscopy was conducted based on previously reported protocols⁷⁴. Rats were first anaesthetized with 1% pentobarbital sodium (0.5 mL/100 g), then perfused transcardially with saline (200 mL), ice-cold 0.05% glutaraldehyde, and 4% PFA (200 mL, pH 7.4) in 0.1 M PBS. Brains were rapidly taken and postfixed by submersion in the same fixative as 4% PFA overnight at 4 °C. Sequential coronal brain sections (50-μm thick), including the cortical region with grafted human cells, were excised with a vibratome (Leica, VT1200S, Germany) and transferred to the same fixative for another 2 h.

Grafted cells were distinguished via immunogold-silver staining. In brief, brain slices were blocked with buffer (1% Triton X-100 in 0.1 M PB and 0.1% BSA-cTM) for 30 min, followed by overnight (4 °C) incubation with primary antibodies (1:200, rabbit anti-GFP, Abcam) and 1-h (room temperature) incubation with secondary antibodies (1:50, Nanogold-labeled goat anti rabbit IgG (H + L), Nanoprobes) diluted with 0.1 M PB and 0.1% BSA-cTM. After rinsing, the sections were postfixed in 2.5% glutaraldehyde in 0.1 M PB for 2 h. Silver enrichment was carried out for 6–8 min in the dark for visualization of GFP immunoreactivity.

Immunolabelled slices were postfixed with 1% buffered osmium tetroxide (OsO₄) in 0.1 M PB for 30 min at 4 °C, dehydrated in a graded ethanol series (15 min per grade) and then in 100% acetone, and flat-embedded in Epon 812 amongst plastic sheets. Three to four selected slices from each brain encompassing grafted cell immunoreactivity in the frontal cortex were pruned under a stereomicroscope and fixed on blank resin stubs. Sequential ultra-thin slices were created with an Ultramicrotome diamond knife (Leica EM UC6, Wetzlar, Germany), then placed on formvar-coated mesh grids (6–8 slices per grid) and observed using a 120-kV frozen transmission electron microscope (Tecnai G2 spirit, Czech) equipped with a CCD camera. Synapses were identified based on ≥2–3 synaptic vesicles in a presynaptic terminal, presence of a synaptic cleft, and high postsynaptic density in the postsynaptic structure⁹.

Adeno-associated virus (AAV)-mediated anterograde transsynaptic tagging

AAV tracing vectors (TaiTool Bioscience, Shanghai, China) were used for anterograde transsynaptic tagging. A volume of 1.5 μL of AAV2/1-hSyn-Cre (titer: 1.23 × 10¹³ vg/mL) was first injected into the ipsilateral ventral thalamus (AP: −2.85 mm, ML: +3.0 mm, DV: −6.5 mm) 2 months after NPC transplantation, as serum type AAV2/1 exhibits a certain neural anterograde labeling potential⁴⁸. Then, 2.0 μL of AAV2/8-EF1a-DIO-mCherry-WPRE-HGHpA (titer: 1.27 × 10¹³ vg/mL) was injected into the four sites (0.5 μL/site) matching the previous NPC transplantation sites. The promoter carried by the AAV2/8-EF1-DIO-mCherry vector is a broad-spectrum EF1α, so all cell types (including host cells and transplanted NPCs) are theoretically tagged when connected to the thalamus. Then, the engrafted progenies could be distinguished from host cells as they carried the GFP marker. We also conducted HuNu staining to distinguish human cells from host cells more reliably, considering the possibility of gene silencing with lentiviral-mediated GFP. Animals were allowed 3–4 weeks to recover following all injections.

Viral injections and rabies tracing experiments

For rabies tracing experiments, 0.8 μL of AAV expressing Cre-dependent TVA (AAV2/9-Ef1a-DIO-TVA, titer 2.00 × 10¹² genome copies (gc)/ml), or 0.8 μL of AAV expressing Cre-dependent Rabies Glycoprotein (AAV2/9-Ef1a-DIO-G, titer 2.00 × 1012 gc/ml) were co-injected into the four graft sites (0.2 μL per site) of stroke animals six weeks after transplantation. Three weeks later, EnVA-pseudotyped, rabies G deleted, tdTomato-expressing rabies virus (RVdG-tdTomato, 1.6 μL, titer 2 × 10⁸ pfu/ml) was injected into the same four sites (0.4 μL per site) for trans-synaptic labeling. One week later, the rats were sacrificed for histological analysis. After fixation, the brain was sectioned (40 μm thick) using a freezing microtome. All coronal sections without staining were imaged with a fluorescence microscope (Zeiss Observer3). The locations of the labeled neurons and the outlines of the brain areas were manually labeled using Photoshop in accordance with the Rat Brain in Stereotaxic Coordinates (Fifth Edition). Some sections underwent HuNu immunostaining to clarify the identity of human cells.

Behavioral test

Single-pellet retrieval (SPR) is a widely used task for the precise evaluation of fine sensorimotor function in rat stroke studies⁶⁹. During the entire SPR experiment, which included pre-training, training, and testing periods, the animals were positioned in a clear Plexiglass box that contained a vertical slit opening (10 × 1 cm) on each side of the front wall. A height-fixed shelf was installed in front of the slit. The rats were trained to pass through the opening to obtain a food pellet (dustless precision pellets/45 mg, banana flavor, Bio-Serv, USA) placed on the shelf. The pre-training period was conducted to identify the dominant forelimb for each rat. After pre-training, rats underwent daily training for 3 weeks before surgery with 50 pellets until a success rate > 60% was achieved over three continuous days. Post-surgical testing was conducted on day 2 after stroke and weeks 1, 2, 4, and 8 after NPC transplantation, with 20 pellets per session per day. The reaching success rate (only considered successful if the pellet was eaten) was used as an indicator for recovery, calculated as [(number of successful reaches / total number of reaches) × 100].

The cylinder test was conducted in the manner previously described^3,4. The animal was positioned in a glass cylinder with a diameter of 20 cm, and the forelimb activity was recorded using a digital video camera. The quantification of forelimb usage was accomplished by placing a mirror beneath the cylinder at a 45-degree tilt to make the entire surface of the cylinder clearly visible. Forelimb use was determined by the placement of the entire palm on the wall, and the contacts were counted offline by an observer who was blinded to the group identity of the animals for up to a total of 20 contacts or over 20 min. The use of each paw was then calculated as a percentage of the ipsilateral and contralateral paw touches.

Video-electroencephalography (EEG) monitoring

Video-EEG monitoring was conducted as per earlier research⁷⁵. EEG screws were attached to stainless steel wires. A stainless-steel screw electrode was implanted in the skull over the transplantation area (M1) adjacent to the infarcted focus as a recording electrode (1 mm anterior and 2 mm lateral from bregma). Two other screws were implanted above the cerebellum as a reference electrode (11 mm posterior and 2 mm lateral to bregma, contralateral to recording electrode) and a ground electrode (11 mm posterior and 2 mm lateral to bregma, ipsilateral to recording electrode). The recording, reference, and ground electrodes were connected to a small homemade plug and stabilized using medical dental cement. All experiments were conducted using freely behaving animals and were performed at 12, 24 and 48 weeks after cell transplantation. All rats were placed in a glass-walled chamber with a multi-channel physiological signal acquisition system (RM-6240B, Chengyi, China) and synchronized video recording (Gz-MG330, Dahua, China). The electrocorticographic signals recorded with the skull screws were defined as EEG recordings. The EEG recordings were visually analyzed using a computer to detect spontaneous seizures, defined as a paroxysmal discharge with rhythmic repetitive waveforms lasting for at least 10 s, with a clear start and end and temporal evolution in amplitude and frequency⁷⁶. If a seizure was detected, rat behavior was evaluated using video recordings. At the final week of cell transplantation therapy, the daily seizure frequency was determined. The seizure frequency was calculated as the ratio of the number of seizures to the number of video-EEG monitoring days during the session in the final week. We analyzed data using custom Matlab software and performed spectral analysis with Matlab (MathWorks) using the wavelet method.

MRI scanning

The MRI data were obtained from a GE 3.0 T MR scanner equipped with a rat coil at 3 days before and at 1, 2, and 4 weeks after NPC transplantation. The rats were secured in the prone position and anaesthetized with isoflurane in oxygen (4% for induction, 1.5%–2.5% for maintenance). T2-weighted MRI images were measured using a fast recovery-fast spin echo (FRFSE) sequence with the following parameters: TR = 2 075 ms, TE = 80 ms, FOV = 80 × 80 mm, matrix = 256 × 256, slice thickness = 1.3 mm, spacing = 0.2 mm, voxel size = 0.2 × 0.2 × 1.5 mm.

¹⁸F-FDG PET/CT imaging and image analysis

The PET/CT data were obtained via an Argus small-animal PET/CT scanner (Sedecal, Madrid, Spain). Rats were fixed in the prone position while anaesthetized with isoflurane gas anesthesia (4% for induction, 1.5%–2.5% for maintenance). The animals were deprived of food but allowed access to water for 12–20 etc., hours prior to the ¹⁸F-FDG injections for PET scans. For PET imaging, ~350 μCi of ¹⁸F-FDG (400 μL final volume) was injected intravenously into each rat via the tail vein under 1.5%–2.5% isoflurane gas anesthesia. The PET images were acquired in 3D and reconstructed using OSEM (ordered-subset expectation maximization) algorithm (calculation factor 2.29 MBq/cps), with 16 subsets and 25 iterations. Images corrected for random and scatter events. The CT imaging data were obtained at standard resolution using the scanning parameters: continuous mode, tube voltage 50 kV, tube current 300 µA, number of projections 360, number of shots 9, and axial field-of-view 120 mm. The CT images were restructured with 0.84 Hounsfield for correction of attenuation. All scans were recorded without respiratory gating. ¹⁸F-FDG accumulation was calculated as the percentage injected dose/gram of tissue using PMOD v.3.902 (PMOD Technologies Ltd., Switzerland).

For metabolic recovery evaluated by PET imaging and analysis, PET data were acquired 6 days before and 1, 3, and 5 weeks after NPC transplantation. Brain PET data were acquired for 10 min using static acquisition mode 40 min after ¹⁸F-FDG injection. To assess changes in brain metabolism before and 1, 3, and 5 weeks after NPC transplantation, 3D regions of interest (ROIs) were depicted manually around infarct lesions based on the MRI scans obtained before NPC transplantation. The ROIs in the infarct lesion and the pons normal areas were identified from transverse brain section images. The lesion-to-pons (L/P) ratio was used for semiquantitative analysis, calculated as: L/P ratio = mean counts per pixel of lesion ROI / mean counts per pixel of pons region.

For evaluation of animals with EEG and without-EEG seizures, PET data were acquired 12 months after cell transplantation. To assess changes in brain metabolism from animals with EEG and without-EEG seizures post-stroke, the whole-brain cortex was partitioned automatically using an improved rat brain template (Tohoku). ROIs in different cortical regions and pons normal areas were identified from transverse brain section images. For semiquantitative analysis, the M1 brain area was used as the EEG recording region. The region-to-pons (R/P) ratio was calculated as: R/P ratio = mean counts per pixel of M1 brain area / mean counts per pixel of pons region.

For safety evaluation, whole-body PET/CT data were acquired using a whole-body emission protocol for 15 min in two bed positions 1 h after ¹⁸F-FDG injection. To assess tumorigenicity, we visually assessed the accumulation of ¹⁸F-FDG in various organs, including the brain, heart, lung, stomach, intestine, kidney, liver, and spleen, 12 months after NPC transplantation. Following whole-body ¹⁸F-FDG PET scanning, these organs were stained with hematoxylin and eosin and observed under a microscope to rule out any pathological evidence of tumor formation subsequent to forebrain NPC transplantation.

¹⁸F-SynVesT-1 radiochemistry, PET imaging and image analysis

No-carried added [¹⁸F]fluoride was produced by cyclotron bombardment and transferred to an automated radiosynthesizer, where the [¹⁸F]fluoride anion in target water (H₂¹⁸O) was trapped on an anionic exchange cartridge Chromafix 45-PS-HCO₃, which was preconditioned sequentially with ethanol (5 mL), an aqueous solution of potassium triflate (KOTf, 90 mg/mL, 5 mL), and deionized (DI) water (5 mL). Then the [¹⁸F]fluoride was eluted off from the cartridge into a 5 mL reaction V-vial with a mixture of acetonitrile (500 µL), an aqueous solution of potassium carbonate (1 mg/mL, 50 µl), and an aqueous solution of KOTf (10 mg/mL, 450 µL). The eluent was azeotropically dried with anhydrous acetonitrile triplicate under 110 °C and a nitrogen airflow. A solution of trimethyltin precursor Me₃Sn-SynVest-1 (3 mg) dissolved in anhydrous N,N-dimethylacetamide (DMA, 0.4 mL) was then added to the reaction V-vial, followed by the mixture of pyridine (1 M in DMA, 0.1 mL) and copper(II) triflate (0.2 M in DMA, 67 μL). The reaction mixture was then heated at 110 °C for 20 min, after which the mixture was cooled down and quenched with high-performance liquid chromatography (HPLC) mobile phase (3 mL) and injected into a semi-preparative HPLC for purification (Phenomenex Synergi Hydro-RP column, 10 × 250 mm, 4 µm; mobile phase: 1.25 (ml/min) acetonitrile and 3.75 (ml/min) 0.1% aqueous trifluoroacetic acid). The fraction containing purified ¹⁸F-SynVesT-1 was collected by monitoring with a radioactivity detector and then diluted with DI water (40 mL). The dilution was passed through a C18 cartridge, preconditioned with ethanol (5 mL) and DI water (10 mL), and the product was trapped on the C18 cartridge. The C18 cartridge was washed with DI water (10 mL) and USP-grade ethanol (0.4 mL). Subsequently, the purified radiotracer was then eluted off with USP-grade ethanol (0.6 mL) into a sterile vial and diluted with USP-grade saline (6 mL). Finally, the mixture was sterilized by a sterile membrane filter (Millex-GV, 0.22 μm) and collected in another sterile vial to afford a formulated solution ready for administration.

The PET/CT data were acquired using an Argus small-animal PET/CT scanner (Sedecal, Madrid, Spain). The rats were immobilized in a prone position following induction with isoflurane gas anesthesia (4% for induction, 1.5%–2.5% for maintenance). Prior to the ¹⁸F-SynVest-1 injections for PET scans, the animals were fasted but allowed access to water for a period of 12–20 h. For PET imaging, each rat was intravenously injected with 300-350 μCi of ¹⁸F-SynVest-1 (400 μL final volume) via the tail vein under 1.5%–2.5% isoflurane gas anesthesia. The acquisition of PET images was performed utilizing a 3D and subsequently reconstructed through implementation of the OSEM algorithm (with a calculation factor of 2.29 MBq/cps), utilizing 16 subsets and 25 iterations. The resulting images were corrected for both random and scatter events. The CT imaging data were obtained at a standard resolution, utilizing scanning parameters that consisted of continuous mode, a tube voltage of 50 kV, a tube current of 300 µA, 360 projections, 9 shots, and an axial field-of-view of 120 mm. The CT images underwent reconstruction with a correction attenuation of 0.84 Hounsfield. Notably, all scans were conducted without respiratory gating. To assess alterations in synapse intensity, PET data were collected prior to and at 1, 2, and 4 weeks post NPC transplantation. Brain PET data were obtained through static acquisition mode, 30 min after ¹⁸F-SynVest-1 injection, for a duration of 30 min. 3D regions of interest (ROIs) were manually delineated around the NPC transplantation sites, based on MRI scans acquired prior to transplantation. The ROIs within the NPC transplantation regions were identified from transverse brain section images and quantified as the standard uptake value (SUV) utilizing PMOD v.3.902 (PMOD Technologies Ltd., Switzerland). For semiquantitative analysis, the standardized uptake value ratio (SUVr) of ROI was calculated using the brain stem as a reference region, based on a previous quantification study of SV2A binding in rodents⁷⁷.

In vitro and in vivo bioluminescence imaging

The IVIS Lumina II Imaging System (PerkinElmer, USA) was used for bioluminescence imaging following previously reported protocols⁷⁸. For in vitro imaging, the cells were seeded in a 48-well plate, washed with Dulbecco’s Phosphate-Buffered Saline (D-PBS), and incubated at 37 °C with medium containing 150 μg/mL D-luciferin (Gold Biotechnology, USA) for 5 min before analysis, with signals detected using the above imaging system. For in vivo imaging, the rats first received an intraperitoneal injection of D-Luciferin (150 mg/kg) (Synchem, Germany) and were then anesthetized with oxygenated 2% isoflurane. A series of images were acquired at 5–30 min post-injection. The bioluminescence signals in a fixed ROI were detected using IVIS image analysis software (PerkinElmer, USA), with data quantified in units of p/s/cm²/sr.

Quantification and statistical analysis

Data are presented as mean ± standard error of the mean (SEM), with P < 0.05 considered statistically significant. Semiquantitative analysis of PET images was conducted using AMIDE v9.2 (Stanford University). GraphPad Prism v6 was used to determine statistical differences, with a Two-tailed unpaired t-test and one-way or two-way analysis of variance (ANOVA) applied for comparisons between two groups and among multiple groups, respectively. The chi-square test was utilized to compare the rates or proportions between two groups. Detailed statistical information is provided in Supplementary Table 4.

Reporting summary

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