Human induced pluripotent stem cells integrate, create synapses and extend long axons after spinal cord injury

Abstract Numerous interventions have been explored in animal models using cells differentiated from human induced pluripotent stem cells (iPSCs) in the context of neural injury with some success. Our work seeks to transplant cells that are generated from hiPSCs into regionally specific spinal neural progenitor cells (sNPCs) utilizing a novel accelerated differentiation protocol designed for clinical translation. We chose a xenotransplantation model because our laboratory is focused on the behaviour of human cells in order to bring this potential therapy to translation. Cells were transplanted into adult immunodeficient rats after moderate contusion spinal cord injury (SCI). Twelve weeks later, cells derived from the transplanted sNPCs survived and differentiated into neurons and glia that filled the lesion cavity and produced a thoracic spinal cord transcriptional program in vivo. Furthermore, neurogenesis and ionic channel expression were promoted within the adjacent host spinal cord tissue. Transplanted cells displayed robust integration properties including synapse formation and myelination by host oligodendrocytes. Axons from transplanted hiPSC sNPC‐derived cells extended both rostrally and caudally from the SCI transplant site, rostrally approximately 6 cm into supraspinal structures. Thus, iPSC‐derived sNPCs may provide a patient‐specific cell source for patients with SCI that could provide a relay system across the site of injury.

cord, and it has been demonstrated that matching regional specificity of the exogenous cells to the injury site is critical to the integration of transplanted neurons. Both spinal cord-derived (sNPCs) and neocortex-derived NPCs survive after transplantation into a rodent model of SCI. However, in the spinal cord environment neocortex-derived NPCs fail to acquire mature neuronal markers, do not integrate and will not extend neurites in vivo. In contrast, sNPCs transplanted into injured spinal cords differentiate into neurons that display mature markers and form functional connections with local neurons. 8 NPCs derived from foetal, post-mortem and pluripotent stem cells (ESC and iPSCs) have all been utilized with promising results, however, of these options, only autologous iPSC-derived NPCs can avoid the need for immune suppressing agents in human clinical trials. 9 We have previously designed an improved, rapid and clinically relevant method to generate regionally specific spinal neural progenitor cells (sNPCs) from human iPSCs to utilize in transplantation studies 10 (These iPSC-sNPCs were originally termed human ventral spinal neural progenitors (hVSNPs) in the previous publication). In this current study, we now evaluate the survival, fate and integration of these iPSC-sNPCs in a rat model of subacute thoracic contusion injury.
There is evidence that regionally specific human sNPCs transplanted into the spinal cord extend axons over several centimetres into the brain and can potentially provide functional recovery, 8,11,12 Here, we determine whether our transplanted iPSC-sNPCs are also capable of forming functional connections with the native rat spinal cord after transplantation and investigate the effects of the sNPCs on the host rat spinal cord environment. This work utilizes a novel protocol designed to generate human iPSC-sNPCs for clinical transplantation after SCI and investigates whether a relay mechanism is a possible strategy for the treatment of SCI using these cells.

| Human iPSC culture and sNPC preparation
Human iPSCs were maintained in adherent culture at 37°C in 5% CO 2 on human vitronectin (rhVTN, AF-140-09; PeproTech, Rocky Hill, NJ) in Essential 8 Media (A2858501; Thermo Fisher Scientific, Waltham, MA) and passaged using hypertonic citrate according to our previously published protocol. 13 The derivation of sNPCs has been previ-  23 adult female athymic nude (ATN) rats (200-220 g) received a thoracic 8/9 (T8/9) moderate contusion injury. Animals had free access to food and water throughout the study and were housed in a 25°C specific pathogen free facility. All protocols were approved by the Institutional Animal Care and Use Committee (IACUC) at the University of Minnesota. ATN rats were selected because they can readily accept xenografts due to their immune deficiency and our previous experience with RNA-seq in these rats. 14 9 days after SCI, rats were randomized to either sNPCs (n = 13) or culture medium only (n = 10). 3 rats receiving sNPCs were sacrificed at 4 weeks and tissue examined for short-term cell survival, fate and integration. The remaining 20 rats underwent functional testing as described below. Two rats in each group did not survive; therefore, there were 8 rats per group utilized in the final analysis. These were sacrificed at 12 weeks and tissue examined for cell survival, fate and integration. Immunohistochemistry (IHC) was performed on 5 rats per group and 3 short-term rats. Additionally, laser microdissection and RNA-seq were performed (n = 3 rats per group) to further examine the cells and the surrounding microenvironment. Nine days after injury, rats were anaesthetized by inhalation of isoflurane, and the laminectomy site was reopened. A 10-ul Hamilton syringe with a 32-gauge needle (0.5 inch long, 30° bevelled tip) was used to inject culture media (10 μl) or sNPCs (10 μl, 50,000 cells/μl, total = 500,000 cells) divided into three separate and equal injection sites at the epicentre and 1 mm rostral and caudal to the epicentre of the lesion. Culture media or cells were injected at the rate of 1 μl/min using a microinjector (Stoelting, Wood Dale, IL, USA). The injection needle was left in situ for five minutes after injection to minimize cellular regurgitation. The surgeon was not blinded to group allocation, but subsequent assessments were performed by blinded examiners.

| Assessment of functional recovery
Functional tests were performed before and after the initial injury, after transplantation, and then weekly from injury until sacrifice.
Locomotor activity was evaluated using the Basso, Beattie and Bresnahan (BBB) locomotor rating scale, 15 by two independent blinded examiners. Motor subscores were determined according to the method of. 16 Ladder walk analysis was recorded and analysed weekly according to a modification of the method of Metz. 17 All remaining rats were sacrificed at 12 weeks after transplantation.

| Tissue harvesting
Rats were fully anaesthetized with intraperitoneal injection of ketamine hydrochloride and transcardially perfused with 4% paraformal-

| Cavitation analysis
Cavitation analysis was performed for the long-term rats only (n = 5 rats per group). To analyse cavitation, every eighth section was processed with Nissl & Eosin Y staining. Ten sections per rat (rostro-caudally 1.6 mm from the epicentre of cavity) were imaged on a Leica DMi8 inverted microscope. To compare cavity area between groups, a modified protocol which combines Nissl and Eosin Y staining was developed. The area of maximum cavitation (epicentre) of each section was traced using Image J software from Fiji (v.1.45) (NIH; Bethesda, MA). The measurements obtained were used to generate values for the cavity volume for each of the cords from each treatment group. The total cavity volume and total spinal cord volume were calculated using the Cavalieri method, 19 and the percentage cavitation determined.

| Immunohistochemistry (IHC)
IHC analysis was performed for both the short-term (n=3 rats per group) and the long-term rats (n = 5 per group). For fluorescence immunostaining, antibodies were utilized to identify the transplanted cells with either human nuclear antigen (hNA, MAB1281; 1:250; Millipore, Billerica, MA) or human cytoplasmic marker (Stem121, AB-121-U-050; 1:250; SC Proven, Newark, CA). We used two different antibodies because one stains nuclei whereas the other stains cytoplasm. When counting cells, a nuclear stain can be more useful; however, a cytoplasmic stain can also demonstrate dendrites and axons.
In any given assay, only one of the two antibodies was utilized. Other

| Laser dissection microscopy (LDM) and RNA preparation
LDM was performed on long-term rats only (n = 3 rats per group).

| Transcriptomic analysis
Samples for RNA-seq analysis included in vitro sNPCs, in vivo transplanted sNPCs prepared using LDM and host spinal cord microenvironment prepared using LDM. RNA samples were prepared in triplicate for RNA-sequencing. Briefly, total RNA samples were quantified using the RiboGreen fluorometry assay and RNA integrity was confirmed using capillary electrophoresis. Integrity of submitted RNA was also assessed using Nanodrop and Bioagilent Analyzer 2100 (RINs ranged from 8.4 to 10.0). Total DNA contamination was quantified using the Picogreen fluorometry assay.
Libraries were generated from 250 ng of total RNA. Polyadenylated coding mRNA in each sample was isolated and reverse transcribed using random primers. The resulting paired-end cDNA libraries were subsequently sequenced using an Illumina HiSeq 2500. For each sample, at least 20 million paired-end reads of 50 base pairs were performed in four lanes. FASTQ files for each sample were combined, and raw sequences were analysed using a customized pipeline developed and maintained by the Minnesota Supercomputing Institute (gopher-pipelines; https://bitbu cket.org/jgarb e/gophe rpipel ines/overview). 22 Briefly, quality controls were performed on each FASTQ files using FastQC (v0.11.5) before and after trimming with Trimmomatic (v.033). Trimmed sequences were aligned using HISAT2 (v2.02). Transcript abundance was then estimated and differential gene expression was determined using Cufflinks (v2.2.1).
Read counts generated by subread were filtered and the remaining reads normalized and log transformed using edgeR. Heat maps were generated using the log-transformed values with pheatmap package.
Hierarchical clustering was performed using Euclidean distances and average linkage clustering method, and principal component analysis was performed in R using prcomp and visualized using ggplot2.

| qPCR probe design
Total cellular RNA was isolated with TRIzol (Invitrogen) extraction.

| Data analysis
All quantitative data are presented as mean ± SEM. Open field locomotor scores (BBB) for the hind limbs from the same group were averaged to yield one score for each time point. BBB scores, motor subscores and ladderwalk data were compared using re- 4.2% ± 1.0% of transplanted cells expressed the mature astrocyte marker GFAP at 4 weeks compared with 3.5% ± 1.2% at 12 weeks (p < 0.005) indicating a small contribution of astrocytes from the transplanted cells in these experiments ( Figure 1H).
In addition, 29.1% ± ± 2.0% of the transplanted cells expressed the immature marker nestin at 4 weeks and 23.2% ± 2.0% still expressed nestin at 12 weeks (p < 0.001) ( Figure 1G,H). Proliferation in the transplanted cells was low, indicated by Ki-67 expression in only 6.3% ± 1.6% of HNA expressing cells at 4 weeks and 3.2% ± 0.72% at 12 weeks, with a trend towards decreasing proliferation over time.
TUNEL staining was performed to determine whether there was ongoing apoptotic cell death, and less that 5% of the transplanted cells showed this at both 4 and 12 weeks (data not shown). No teratomas or other abnormal tissue growth was observed at any time point in this study.

| Transplanted human iPSC-sNPCs demonstrate a spinal cord transcriptional program in vivo
We utilized RNA-sequencing (RNA-seq) to analyse molecular differences between the in vitro cell identity of the sNPCs and expression patterns in these cells 12 weeks after transplantation in vivo. We utilized laser microdissection to specifically isolate cells from the transplantation site (Figure 2A). Unsupervised hierarchical clustering revealed distinct separation between the two groups ( Figure 2B). The top 500 differentially expressed genes revealed that transplanted sNPCs developed a regional identity associated with the thoracic spinal cord after transplantation, including HOXA7 and HOXB9 (top 20 differentially expressed genes) ( Figure 2C, left). The majority of the contents in the cyst cavities stained for SC121 after 12 weeks which is suggestive of viable tissue. Furthermore, we performed TUNEL staining to identify apoptosis and this was very low.

| Transplanted human iPSC-sNPCs generated neurons that extended axons both rostrally and caudally from the transplant site and formed synapses
The majority of cell bodies derived from the transplanted sNPCs were located at or adjacent to the injection sites 12 weeks after injury, with processes extending away from the site of injury. These human cells Additionally, our data support previous findings 11,28 that transplanted cells require direct cell-to-cell contact with the host in order to project axons ( Figure 4D). We found that the majority of In addition, host oligodendrocytes were directly adjacent to transplanted cell axonal projections, indicating myelination of the transplanted neurons by the host (Figure 4G,H). Furthermore, host myelin basic protein (MBP) was located in close proximity to SC121/MAP2positive neurons, suggesting that axonal extension in transplanted cells was not inhibited by host myelin in either the grey matter or white matter. Despite these promising findings, no functional recovery was observed in these experiments.

| DISCUSS ION
There has been renewed interest in the role of new neurons generated following NPC transplantation at CNS injury sites after recent publications have demonstrated that they can produce functional benefit in both rodent and primate models. 29,30 Earlier studies explor- transplanted post-mortem-derived sNPCs can also develop into neurons with axons that extend over the entire length of the rat spinal cord. Using our human iPSC-sNPCs, we also observed similar differentiation and long-distance axon extension. We found that the presence of a 'gap' between the transplant and the host tissue results in a failure of these axons to extend across the lesion border. However, when cell-to-cell contact between transplanted cells and host tissue was present, dendrites and axons were seen to extend into the native rat spinal cord and some terminated within the spinal cord in the grey matter. Furthermore, there was evidence of synapse formation at the end of these axons, supporting the concept of functional integration [36][37][38][39][40][41][42][43] and this corresponds with the recent report by Ceto et al. 44 showing that dissociated murine embryological sNPC grafts can form extensive networks when transplanted into a mouse spinal cord injury model. Our work helps advance these findings towards the clinic by utilizing a regionalized source of human neural progenitor cells suitable for potential autologous transplantation.
In this study, we did not find functional benefit, and other studies have produced mixed results. There are several possible explanations for this. It is possible that the studies have not been carried out long enough for the cells to fully mature, form appropriate functional synapses, and provide benefit. It has been shown that it could take many months, even more than a year, for human progenitor cells to fully differentiate following transplantation. 11,12,45,46 In these experiments, we did find that the majority of cells (~90%) expressed mature neural markers 12 weeks post-transplantation; however, we do not know what further capabilities a mature neuron might have in terms of neuroplasticity. Another possible explanation is that the study was underpowered to detect any subtle differences in locomotor change or that the functional synapses require reinforcement through activation such as that provided by physical therapy or electrical stimulation.

ACK N OWLED G EM ENTS
AMP and JRD acknowledge the Minnesota Spinal Cord Injury and Traumatic Brain Injury Research Grant Program, the Spinal Cord Society, and an anonymous philanthropic donor. We also acknowledge the University of Minnesota Imaging Center (UIC) and the University of Minnesota Genomics Center (UMGC).

CO N FLI C T S O F I NTE R E S T
The authors declare no competing interests.

CO NTAC T FO R R E AG E NT A N D R E S O U RCE S H A R I N G
Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Ann M. Parr (amparr@umn.edu).

DATA AVA I L A B I L I T Y S TAT E M E N T
The data sets and code utilized in this study are available upon request.