A robust culture system to generate neural progenitors with gliogenic competence from clinically relevant induced pluripotent stem cells for treatment of spinal cord injury

Abstract Cell‐based therapy targeting spinal cord injury (SCI) is an attractive approach to promote functional recovery by replacing damaged tissue. We and other groups have reported the effectiveness of transplanting neural stem/progenitor cells (NS/PCs) derived from human induced pluripotent stem cells (hiPSCs) in SCI animal models for neuronal replacement. Glial replacement is an additional approach for tissue repair; however, the lack of robust procedures to drive iPSCs into NS/PCs which can produce glial cells has hindered the development of glial cell transplantation for the restoration of neuronal functions after SCI. Here, we established a method to generate NS/PCs with gliogenic competence (gNS/PCs) optimized for clinical relevance and utilized them as a source of therapeutic NS/PCs for SCI. We could successfully generate gNS/PCs from clinically relevant hiPSCs, which efficiently produced astrocytes and oligodendrocytes in vitro. We also performed comparison between gNS/PCs and neurogenic NS/PCs based on single cell RNA‐seq analysis and found that gNS/PCs were distinguished by expression of several transcription factors including HEY2 and NFIB. After gNS/PC transplantation, the graft did not exhibit tumor‐like tissue formation, indicating the safety of them as a source of cell therapy. Importantly, the gNS/PCs triggered functional recovery in an SCI animal model, with remyelination of demyelinated axons and improved motor function. Given the inherent safety of gNS/PCs and favorable outcomes observed after their transplantation, cell‐based medicine using the gNS/PCs‐induction procedure described here together with clinically relevant iPSCs is realistic and would be beneficial for SCI patients.


| INTRODUCTION
The central nervous system (CNS) has a limited capacity to regenerate; therefore, any loss of function by traumatic injury is mostly irreversible. Spinal cord injury (SCI) is a traumatic injury to the CNS that results in severe physical impairments, including permanent paralysis, sensory disturbance, and bladder dysfunction. 1 The number of new cases annually is approximately 120 000 in the United States and 5000 in Japan. 2,3 However, approaches to treat SCI patients are currently limited.
In recent years, cell transplantation therapies are considered to be effective in SCI animal models, and preclinical evaluations have been initiated. [4][5][6][7][8][9][10] As a cell source for the therapies, human induced pluripotent stem cells (hiPSCs) have attracted particular interest due to a limited number of ethical concerns, as they are directly generated from somatic cells. 11 We have reported the efficacy of cell transplantation in SCI animal models using neural stem/progenitor cells (NS/PCs) derived from hiPSCs and the integration of hiPSC-derived neurons into the host neuronal networks. [12][13][14][15] However, mouse gliogenic NS/PCs were reported to produce more favorable outcomes than neurogenic NS/PCs after transplantation into a SCI animal model, indicating additional processes mediated by gliogenic NS/PCs can be therapeutic in addition to neuronal replacement by grafted NS/PCs. 16 Moreover, transplantation of NS/PCs derived from myelindeficient shiverer mice, produced less functional recovery after transplantation into injured spinal cord, indicating the importance of remyelination in the repair process by grafted NS/PC-derived oligodendrocytes. 17 These results indicate the importance of glial cells including oligodendrocytes for treatment of SCI. Since hiPSCs can serve as an unlimited and scalable source of somatic cells, successful derivation of glial cells would benefit a broad range of neurological disorders including SCI. However, it is noteworthy that hiPSC-NS/PCs predominantly differentiate into neurons both in vitro and in vivo. 12,13,18 Accordingly, numerous attempts have been made to improve the generation of glial cells including oligodendrocytes from hiPSCs. [19][20][21][22][23] However, these procedures often required long duration to generate glial cells and/or complicated techniques. 19,20,24 Moreover, most procedures are optimized for iPSCs generated in individual laboratories. There is limited information for reproducibility of differentiation over multiple iPSCs. Therefore, a standardized protocol to generate glial cells from iPSCs has not been developed.
In addition, there is another issue to be in account. Numerous studies have demonstrated the potency of iPSCs in regenerative medicine; however, most of these studies used iPSCs that were not generated for clinical use. Generation and validation of clinically relevant iPSCs is costly and time-consuming, which hinders the preclinical evaluation of iPSCs. Thus, there is a need to develop a cell bank of iPSCs, which provides qualified iPSCs to researchers. Then, each researcher would conduct preclinical evaluation of iPSC-derivatives without generating and validating iPSCs by himself. The Center for iPS Cell Research and Application (CiRA) at Kyoto University, a core center for iPSC research in Japan, was funded by the Japanese government to prepare clinical grade iPSCs from donors who were homozygous for the three major human leucocyte antigen (HLA) loci haplotypes (HLA-A, HLA-B, and HLA-DR) (HLA-homozygous), which would reduce immune rejection after allogenic transplantation. 25,26 However, there is no established protocols to generate NS/PCs optimized for therapeutic intervention, and, most importantly, their safety and therapeutic efficacy after transplantation remain unknown.
Here, we aimed to establish and optimize a robust protocol to produce NS/PCs with gliogenic competence (gNS/PCs) from clinically relevant iPSCs, from which oligodendrocytes can be generated. We also assessed the efficacy and safety of the NS/PCs in a rodent model of SCI ( Figure 1A). This study is the first evaluation of clinically relevant HLA-homozygous iPSCs in an SCI model.

See further details in the Supporting Information Materials and
Methods.

Significance statement
This study demonstrated a successful and robust procedure to generate neural stem/progenitor cells (NS/PCs) with gliogenic competence from clinically relevant human induced pluripotent stem cells, which in turn efficiently differentiated into astrocytes and oligodendrocytes in vitro and in vivo. Transplantation of the NS/PCs triggered remyelination and functional recovery in a spinal cord injury (SCI) animal model. These findings indicate that this method to produce gliogenic NS/PCs will be useful for developing cell-based therapies for SCI. Stemgent) at 37 C and 5% CO 2 /5% O 2 for 7 days. On day 7, the medium was switched to StemFit AK03N without FGF2 supplemented with 1 μM retinoic acid (RA, Sigma). The medium was changed every day. On day 14, EBs were transferred into ultralow attachment culture flasks (Corning) and cultured in proliferation medium supplemented with 1 μM RA and 1 μM purmorphamine (PM, Merck Millipore) at 37 C and 5% CO 2 /20% O 2 . Proliferation medium is a mixture of DMEM/F12 (Wako) containing 1% N2 (Thermo Fisher Scientific), 2% B-27 without Vitamin A (Thermo Fisher Scientific), 60 ng/mL triiodo-L-thyronine (T3, Sigma), 10 ng/mL Platelet-derived growth factor (PDGF)-AA(PeproTech), 20 ng/mL FGF2 (PeproTech), 10 ng/mL epidermal growth factor (EGF, PeproTech), 10 ng/mL insulin growth factor 1 (IGF-1, R&D systems), and 10 ng/mL neurotrophin 3 (NT-3, R&D Systems). On day 21, EBs were enzymatically dissociated into single cells using TrypLE Select, and the dissociated cells were cultured in suspension at a density of 1.0 × 10 5 cells/mL in F I G U R E 1 Legend on next page. proliferation medium supplemented with 1 μM PM to generate the first neurospheres. The medium was changed every 7 days. After 14 days, the first neurospheres were dissociated in the same manner as described above and cultured in proliferation medium without PM. Passaging was carried out once every 14 to 28 days thereafter.
The third passage neurospheres were used for transplantation. Neuronal and glial differentiation was performed as previously described. 23 Undissociated 15 to 20 neurospheres were plated on 8-well chamber glass slides coated with 1% growth factor reduced Matrigel (Corning), and cultured in differentiation medium that consisted of KBM Neural Stem Cell (KOHJIN BIO) supplemented with 2% B27, 1% nonessential amino acids (Thermo Fisher Scientific), 60 ng/mL T3, 10 ng/mL NT-3, 10 ng/mL Leukemia Inhibitory Factor (Millipore), and 25 ng/mL Ciliary Neurotrophic Factor (Peprotech) for 4 weeks in a humidified atmosphere of 5% CO 2 . Half of the medium was changed every 2 or 3 days. The neurogenic NS/PCs were kindly provided by Dr Yonehiro Kanemura and cultured as previously described. 29 2.2 | SCI animal model and feeder-free hiPSCs (ffiPSC)-gNS/PCs transplantation Eight-week-old female nonobese diabetic-severe combined immunodeficiency mice (Charles River Laboratories) were anesthetized with an intraperitoneal injection of ketamine (100 mg/kg) and xylazine (10 mg/kg). Contusive SCI was induced at the level of the tenth thoracic spinal vertebra using an Infinite Horizon impactor (60-70 kdyn; Precision Systems and Instrumentation), as described previously. 30 After SCI, 12.5 mg/kg ampicillin was administered intramuscularly.
Seven days after injury, all the mice were evaluated for motor function, and mice with a Basso Mouse Scale (BMS) 31 score of 2.5 or higher were excluded from the study. The SCI mice were randomly assigned to each group (15 mice per group). For gNS/PCs group mice, 9 days after injury, ffiPSC-gNS/PCs (5 ×

| Motor function analyses
Hind limb locomotor function was evaluated for 12 weeks after transplantation using the BMS. At 12 weeks after transplantation, a rotating rod apparatus, a treadmill gait analysis, and a kinematics were performed. Detailed methods are described in the Supporting Information.
Repeated-measures two-way analysis of variance (ANOVA), followed by the Turkey-Kramer test, was used for the BMS analysis (control group, n = 12; ffiPSC-gNS/PCs group, n = 14). In each case, *P < .05 and **P < .01 were statistically significant.

| Generation of glial progenitor-enriched NS/ PCs from feeder-free HLA-homozygous hiPSCs
To realize clinical application of NS/PCs derived from iPSCs, we focused on clinically relevant feeder-free hiPSCs (ffiPSCs) provided by iPSC bank in CiRA. 25,26 These iPSCs had been established from HLA-homozygous donors to lower the immunogenicity of iPSC-derivatives 28,32 ( Figure 1A). Since neural differentiation protocols usually depend on the innate differentiation capacity of the individual iPSC clones and are successful for a limited number of iPSC clones, 33 we were motivated to modify previously reported neural induction procedures for NS/PCs, which had been examined over multiple iPSC clones established from different somatic tissue origins. 29,34 Given that the NS/PCs generated from this procedure were less gliogenic, 29 we attempted to determine experimental conditions to generate NS/PCs containing glial progenitors using the ffiPSCs. Initially, we applied this protocol on the ffiPSCs derived from umbilical cord blood cells (WJ14s01) and examined differentiation capacity of NS/PCs. As shown in Figure S1, we successfully differentiated ffiPSCs into NS/PCs which mainly produced neurons and, as expected, glial differentiation capacity was limited: any oligodendrocytes judged by the expression of oligodendrocyte marker O4 was not detected. Then, we modified this protocol to the trigger precocious glial differentiation from NS/PCs and determined the protocols as follows.
We further applied the protocol to three additional clones of ffiPSCs (WJ14s02, WJ23s01, and WJ23s02) and consistently obtained NS/PCs that produced astrocytes and oligodendrocytes, indicating that the procedure was robust enough to generate NS/PCs from the ffiPSCs (data not shown).
Tumorigenicity is one of potential barriers to put iPS-derived cells into clinical setting. 46 Accordingly, we examined the safety of ffiPSC-gNS/PCs by transplanting them into the striatum of immunodeficient rats as previously described 34 Figure 1F). Therefore, we found that ffiPSC-gNS/PCs were less tumorigenic.

| Transcriptome signature of ffiPSC-gNS/PCs
To characterize the ffiPSC-gNS/PCs, we performed RNA-seq analysis on ffiPSCs, ffiPSC-derived EBs, and ffiPSC-gNS/PCs (P2 and P3) to that of human fetal NS/PCs while that of EBs was more similar to neurogenic iPSC-NS/PCs ( Figure 2D). Notably, the trend of the gNS/PCs transcriptome becoming more similar to that of human fetal NS/PCs was apparent during cellular passage, revealing a transcriptional transition in ffiPSC-gNS/PCs from P2 to P3. We selected differentially expressed genes between these cells (306 upregulated and 714 downregulated genes in P3 ffiPSC-gNS/PCs) ( Figure 2E) and performed Gene Ontology (GO) analysis to illuminate the biological functions ( Figure 2F). While the GO term "cell adhesion" was identified in the downregulated gene set, GO terms obtained from upregulated genes included "nervous system development," "neurotransmitter secretion," "chemical synaptic transmission," and "axon guidance," indicating neuronal maturation during passaging.  Figures S1 and 3A). We sorted the NS/PCs into 96-well plates and processed them for scRNA-seq analysis. 50 After quality control, 150 ffiPSC-gNS/PCs and 91 ffiPSC-nNS/PCs were further analyzed. Cell-clustering analysis by uniform manifold approximation and projection revealed clear separation of ffiPSC-nNS/PCs and ffiPSC-gNS/PCs ( Figure 3B). The differential distribution of the clusters between these NS/PCs suggested that each group was substantially different at the transcriptome level ( Figure 3B). For example, both NS/PCs shared similar SOX2 expression profiles; however, MKI67, a marker for proliferation, was mainly expressed in ffiPSC-nNS/PCs, indicative of the progressive expandable capacity of neurogenic NS/PCs ( Figure 3C). Since activation of Notch signaling is essential to acquire gliogenic potential of NS/PCs during mouse CNS development, 51 we examined F I G U R E 5 Legend on next page. expression of HES5, a downstream effector of Notch signaling, and found that HES5 was mainly observed in the ffiPSC-gNS/PCs, suggesting gliogenic properties of ffiPSC-gNS/PCs ( Figure 3C).
Given that in our procedure to generate ffiPSC-gNS/PCs required RA, well-known for caudalizing activity on neural tissues during development, 39 we examined difference in the regional identity. As shown in Figure S2, we examined expression of regional specific neural markers 52-54 and observed ffiPSC-gNS/PCs hold regional properties as caudalized neural tissues compared to ffiPSC-nNS/PCs, indicating regional identity in addition to differentiation capacity was altered. To characterize difference between these NS/PCs, we classified them based on transcriptome, and identified three clusters (C1-C3) ( Figure 3D). Each cluster was characterized by marker genes that were  Figure 3G). In contrast, the expression of glial genes, including HEY2 and NFIB, was observed in Cluster C2 ( Figure 3E,F). Interestingly, the Cluster C2 was highly enriched in ffiPSC-gNS/PCs, further supporting the glia-committed properties of ffiPSC-gNS/PCs compared with ffiPSC-nNS/PCs ( Figure 3G).

| Transplanted ffiPSC-gNS/PCs survive, migrate, and differentiate into three neural lineages in vivo
We next examined whether the NS/PCs could promote functional recovery after CNS injury. We generated an SCI animal model by inducing contusive SCI with moderate severity at the 10th thoracic vertebrae level in adult female immunodeficient mice. 14 neurons, 27.23% ± 1.92% into GFAP + astrocytes, and 36.56% ± 2.82% into APC + oligodendrocytes ( Figure 4E).
transplantation. Next, we further examined oligodendrocyte differentiation of ffiPSC-gNS/PCs by immunohistochemistry using MBP and found that there were several STEM121 + /MBP + areas in the ffiPSC-gNS/PCs group ( Figure 5C,D). Immunoelectron microscopy showed that myelin sheaths were strongly associated with myelin cytoplasm with nanogold-labeled STEM121 + spots ( Figure 5E,F). These results indicate that transplanted ffiPSC-gNS/PCs-derived oligodendrocytes formed mature myelin sheaths on spared axons. To examine the myelinated region in cross-sections of the spinal cord, we visualized myelin sheaths in the injured spinal cord by LFB staining at 12 weeks after transplantation. The LFB + area was dramatically reduced in the injured spinal cord, indicative of demyelination in the SCI model ( Figure 5G).
In contrast, quantitative myelinated areas compared with the PBS group from the epicenter to +0.96 mm rostrocaudal ( Figure 5H). Consistently, we found that MBP + areas in the transplanted group were also significantly larger than those of the PBS group ( Figure S4A,B).
Taken together, these results indicate that these NS/PCs harbor the capacity to differentiate into mature oligodendrocytes and form myelin sheaths.
Since reconstitution of neuronal circuit is important for functional recovery after SCI 14

| Transplanted ff-hiPSC-gNS/PCs enhance functional recovery following SCI
Finally, recovery of locomotor function was evaluated by the BMS score, 31 rotarod test, 58,62 treadmill gait analysis using the DigiGait system and kinematic analysis. The BMS score showed significantly improved motor function in the ffiPSC-gNS/PCs group compared with that in the control PBS group at 14 days after SCI, and functional recovery was significant thereafter ( Figure 6A). Body weight was also significantly increased in the ffiPSC-gNS/PCs group ( Figure S3C), consistent with a previous report. 63 The gait performance of the mice in the two groups was examined using the rotarod test and the DigiGait system at 12 weeks after transplantation. In the ffiPSC-gNS/PCs group, treadmill gait analysis showed a significantly longer stride length and smaller paw angle ( Figure 6B,C).
The mice transplanted with the NS/PCs ran on the rotating rod for a significantly longer time than those given PBS ( Figure 6D). To evaluate the gait and movement of each joint in more detail, kinematic analysis was performed at 12 weeks after transplantation. Representative stick diagrams of hind limb movements of the swing phase indicate that the ffiPSC-gNS/PCs group had a smoother step and a more consistent step cycle than the control group ( Figure 6E,F). Furthermore, each joint moved with less variation in the ffiPSC-gNS/ PCs group ( Figure 6G,H), and various types of joint function in the ffiPSC-gNS/PCs group were significantly improved compared with those in the control group ( Figure S6).
To evaluate the spinal stretch reflex, which plays an important role in locomotion, electrophysiological examination was performed using the H-reflex at 12 weeks after transplantation. The amplitude of the H-reflex was significantly lower in the ffiPSC-gNS/PCs group compared with that in the PBS group ( Figure 6I). Together, these results indicate that the gait in the ffiPSC-gNS/PCs group was also improved compared with that in the control group, suggesting that ffiPSC-gNS/PCs restore function of residual neuronal fibers to enable precise gait coordination.
F I G U R E 6 Motor function analyses after transplantation of feeder-free hiPSCs (ffiPSC)-gNS/PCs. A, Comparison of Basso Mouse Scale (BMS) scores between the PBS and ffiPSC-gNS/PCs groups. The BMS scores showed significantly functional recovery in the ffiPSC-gNS/PCs group compared with that in the PBS group at 14 days after SCI and thereafter. Values are means ± SEM (control group, n = 12; ffiPSC-gNS/PCs group, n = 14; *P < .05). B,C, Comparison of stride lengths (B) and paw angles (C) between the PBS and ffiPSC-gNS/PCs groups. Treadmill gait analysis was performed at 12 weeks after transplantation using the DigiGait system. Stride length and stance angle were significantly longer and more parallel, respectively, in the ffiPSC-gNS/PCs group than in the PBS group. Values are means ± SEM (control group, n = 12; ffiPSC-gNS/PCs group, n = 14; *P < .05 **P < .01  (Table S1). Although we did not directly compare these protocols, our protocol has certain advantages in terms of the required culture period to obtain MBP + oligodendrocytes and higher yield to obtain glial cells.
As for oligodendrocyte differentiation, we observed contribution of gNS/PCs to remyelination in injured spinal cord ( Figure 5D-F). It is worth noting that gNS/PCs-derived OLIG2 + GST-π − immature oligodendrocytes were still present among the grafts ( Figure 5A,B). Given that OLIG2 is indicative of migrating oligodendrocytes 68 and that 77% of cells in the oligodendrocyte lineages still expressed OLIG2 ( Figure 5B) at 12 weeks after transplantation, the gNS/PCs-derived cells that had entered oligodendrocyte lineages still theoretically harbor the potential to promote functional recovery by further remyelination.
Oligodendrocyte precursors migrate toward demyelinated lesions 69 ; therefore, the residual OLIG2 + cells would remyelinate the demyelinated axons over time. Further functional recovery would be enhanced in SCI animal models by rehabilitation, possibly through ongoing myelination together with continued axonal regeneration and synaptogenesis as previously described. 70 There is a series of reports demonstrating restoration and/or reorganization of neuronal circuits by graft-derived neurons in SCI animals after NS/PC-transplantation. 71 Although we did not observe significant regeneration of the RtST, gNS/PCs still hold the capacity of neuronal differentiation and functional motor recovery might be acquired by neurons differentiated from gNS/PCs ( Figure 4D). Indeed, we observed synapse formation between host and gNS/PCs-derived neurons ( Figure S5A,B), providing mechanistic insight into the functional recovery observed after transplantation.
Taken together, remyelination of damaged tissues in addition to neuronal replacement would be mediated by gNS/PCs to enhance functional recovery of SCI after transplantation.
One might ask whether gNS/PCs exerted functional recovery more efficiently than neurogenic NS/PCs. In our previous reports using integration-free iPSCs, 34 functional recovery after transplantation was not obvious using neurogenic NS/PCs, indicating that gNS/PCs are more beneficial for SCI patients in clinical settings.
Indeed, we transplanted nNS/PCs derived from same parental iPSCs and did not observe functional recovery after the transplantation, indicating that gNS/PCs are superior cell sources for cell-based therapy for SCI ( Figure S7). Importantly, this observation might be supported by the previous finding using directly reprogrammed NS/PC from somatic cells. 9 They also found more beneficial aspects of NS/PCs which had capacity to differentiate into oligodendrocytes.
To acquire cell source for transplantation via directly reprogramming approach might be more realistic in the setting of autologous transplantation. Although, it would be challenging to prepare large amounts of cells with quality assurance for autologous without bypassing iPSCs, cell-therapy relying on NS/PCs with gliogenic differentiation capacity hold promising therapeutic potential. However, gNS/PCsderived oligodendrocytes still need residual host neurons, which are demyelinated upon SCI. Therefore, criteria to identify SCI patients who will display functional recovery with gNS/PCs should be determined and examined in future studies.
The current study has several limitations that should be discussed.
First, although iPSCs provided by CiRA were generated and maintained under good manufacture practice (GMP)-conditions, 27  Future studies should seek to develop a procedure to bias NS/PCs toward selected cell lineages that exhibit greater therapeutic potential against neurotraumatic injuries.

| CONCLUSION
We established a robust protocol to generate gNS/PCs from clinically relevant iPSCs. The gNS/PCs derived from HLA-homozygous iPSCs exhibited prominent favorable outcomes after transplantation into an SCI animal model, providing a promising cell source for treating SCI in future clinical settings.