Transplantation of Directly Reprogrammed Human Neural Precursor Cells Following Stroke Promotes Synaptogenesis and Functional Recovery

The experimental design was a controlled laboratory experiment. Male and female animals were used and separated into groups via random assignment by a blinded third party until appropriate numbers of samples were achieved for each group. Behavioral analysis was conducted by an observer blinded to the treatment groups. Tissue and cellular outcomes were evaluated by three separate observers blinded to the experimental groups.

For functional analysis, 10–15 mice per group were analyzed. Using a sensitivity analysis on G*Power (version 3.1) with power = 0.8, we determined the treatment effect size to be ~ f(V) = 0.65 (η²_partial = 0.3) when analyzing all four treatment groups, f(V) = 0.42 when comparing between drNPC and vehicle groups, and f(V) = 0.93 when comparing between brains that had surviving drNPCs and those that did not. For tissue outcome comparisons between mice that received drNPCs or vehicle alone, the effect size was always above d = 0.9 (Gliosis, 0.97; Lesion Volume, 1.51; Synaptophysin, 1.60) with a power = 0.8. All effect size (sensitivity) calculations were based on Cohen’s d [27]. Excluded animals were not considered for power and sensitivity analyses.

Immunocompromised Fox Chase SCID/Beige (8–16 weeks old; CB17.Cg-Prkdc^scidLyst^bg-J/Crl; Charles River Laboratories, Wilmington, MA) mice were singly housed on a 12-h light/dark cycle with food and water provided ad libitum for the duration of testing, starting 3 days prior to stroke, until sacrifice. A total of 87 mice (establishment of stroke, n = 13, [sex not tracked]; long-term deficit analysis for ET-1 stroke, n = 13 [7 males, 6 females]; confirmation of stable measures in long-term testing of naïve mice, n = 8 [6 males, 2 females]; therapeutic evaluation of drNPC transplants, n = 53 [30 males, 23 females]) were used in this study. Outliers and mice that did not meet our inclusion criteria were removed from the study as described in the supplementary materials.

ET-1 stroke was performed in SCID/Beige mice as previously described [17, 28, 29]. Briefly, the skull was exposed, a small burr hole was drilled at the site of the right sensorimotor cortex at + 0.6 mm anterior and − 2.25 mm lateral to bregma. Mice received a 1-μL injection of ET-1 (Calbiochem, 800 picomolars) 1 mm deep from the surface of the brain at a rate of 0.1 μL/min using a 2.5 μL Hamilton Syringe with a 26 gage, 0.375″ long needle (Hamilton, Reno, NV). The needle was kept in place for 10 min following the injection and then slowly withdrawn.

Human bone marrow cells were reprogrammed with transient expression of transcription factors musashi-1 (Msi1), neurogenin-2 (Ngn2), and methyl-CpG binding domain protein 2 (MBD2), as described in detail in the supplementary material. The cells were cultured until they reached ~ 80% confluence in each passage and collected for transplantation after 4–9 passages. Sister cultures were prepared for in vitro analysis to characterize the cells using immunocytochemistry and PCR.

For transplantation, drNPCs were suspended in artificial cerebrospinal fluid (aCSF) or Hyaluronan Methylcellulose (HAMC; supplementary methods) hydrogel (100,000 cells/μL). Cells were transplanted into the stroke site 4 days following stroke with the same surgical procedures used for ET-1-induced ischemia in the sensorimotor cortex. Control animals received 1 μL injections of aCSF or HAMC only.

A live/dead assay was performed to determine the percent of surviving cells post-injection through the syringe [30]. Using the same protocol as was used for transplantation, 100,000 cells in 1 μL of vehicle were injected into a well containing 14 μL of warm (37 °C) aCSF at a rate of 0.1 μL/min. Following injection, 15 μL of the live/dead stain solution (0.2% Ethidium Homodimer and 0.05% calcein AM) (L3224, ThermoFisher) was added, followed by a 5-min incubation period after which 220 μL aCSF was added for a final volume of 250 μL. The solution was imaged using an AxioVision Zeiss UV microscope (5× magnification) and images were visualized in FIJI [31]. The percent of live/dead cells was calculated using the “analyze particles” feature on the FIJI software.

Fixed tissue and cells were rinsed with 1× phosphate buffered saline (PBS), permeabilized with 0.3% Triton-X100 in 0.01 M PBS for 20 min and blocked in 10% Normal Goat Serum (NGS) with 0.3 M glycine for 1 h at room temperature. Samples were then treated with primary antibodies (Table 1) in 0.01 M PBS and left overnight at 4 °C in a humid chamber. Samples were washed with 1× PBS and exposed to secondary antibodies for 1 h at room temperature (Table 2). The samples were then washed, cover-slipped with mowiol® 4-88 (Sigma-Aldrich), and imaged using an AxioVision Zeiss UV microscope, an Olympus FV1000 confocal point-scanning microscope, or a ZEN Zeiss spinning disk confocal microscope.

Following cresyl violet staining (see Supplementary methods), serial 20- μm thick coronal sections (200 μm apart) spanning a total of 3–4 mm surrounding the injury site were imaged at 5× magnification using an AxioCam ICc1 camera. The cortical lesion was measured on FIJI using the lasso and polygon tools to outline and quantify the total cortical lesion infarct area, as defined by the area with atypical tissue morphology including pale areas with lost Nissl staining and areas filled with dark pyknotic stained debris [32]. The total volume of the injury was estimated by averaging the area measured in each coronal section and multiplying by the total length of the scar, which was calculated from the number of sections in which the lesion was present.

A set of serial coronal sections (20 μm thick) immunostained for GFAP⁺ expression were visualized at 5× magnification at 200 μm intervals using FIJI [31]. The total area of cortical GFAP⁺ expression was measured in each section. Measurements were taken from anterior to posterior through the scar and the maximal GFAP expression, as well as the total gliosis volume, was calculated per brain.

All of the images were taken with identical parameters using confocal microscopy on a ZEN Zeiss spinning disk confocal microscope to generate z-stacks comprised of eight optical sections at 0.49 μm per section. The channel exposure was fixed at 1000 ms throughout the imaging of the entire set. Quantification of total synaptophysin-positive pixels per analyzed brain section was conducted by using FIJI [31] to measure the number of positive pixels in the perilesional areas. The mean pixel intensity in two perilesional regions of interest (ROIs) was used to measure the amount of staining (Fig. 7a), as this measure represents the sum of all detected bright pixels (gray values) divided by the total number of pixels within the channel. Imaging, ROI selection, and analysis were conducted by a blinded observer.

Three cell culture wells per biological replicate were stained for each specific antibody and were counted within the field of view in five areas within each well at 20× magnification. The percentage of each cell type was calculated as a percent of all DAPI or Hoechst labeled cells.

For in vivo analysis, coronal sections 20 μm thick at 200 μm intervals were immunostained for HuNu or STEM121 and antibodies found in Table 1. Total numbers of surviving transplanted cells were calculated by extrapolating the average number of surviving drNPCs per section over the total number of sections that contained drNPCs (ranging from 15 to 20 sections). To analyze proliferation, the numbers of Ki67⁺/HuNu⁺ cells were counted in the same representative sections and calculated as a percent of all HuNu⁺ cells. Cell differentiation in vivo post-transplantation was analyzed by immunohistochemistry in brains that had surviving drNPCs.

Cultured drNPCs were collected into Buffer RL (Norgen Biotek) with β-mercapthenol and then processed according to the manufacturer’s directions using Total RNA Purification Kit (Norgen Biotek — Cat#17200). Cycling conditions consisted of polymerase activation and DNA denaturation (3 min at 98 °C), followed by 35 cycles of 10 s at 95 °C and 30 s at 60 °C. Primer sequences used are listed in Table 3.

Samples were collected into Buffer RL (Norgen Biotek) and processed according to the manufacturer’s directions using Total RNA Purification Kit (Norgen Biotek — Cat#17200). cDNA synthesis was carried out with iScript gDNA Clear cDNA Synthesis Kit (Bio-Rad — Cat# 1725034). RT-qPCR reactions were prepared according to the manufacturer’s directions using SsoAdvanced Universal SYBR Green Supermix (Bio-Rad — Cat# 172-5270). RT-qPCR was carried out on Bio-Rad CFX384 Touch Real-Time PCR System (Bio-Rad). Cycling conditions consisted of polymerase activation and DNA denaturation (3 min at 98 °C), followed by 40 cycles of 10 s at 95 °C and 30 s at 60 °C. All reactions were concluded by incubation at 65 °C and increasing the temperature (at 0.5 °C increments) to 95 °C for melting-curve analysis. Prior to performing relative expression analyses, standard curves were generated for targets (see below) via the serial dilutions of pooled cDNA. In accordance with MIQE (Minimum Information for Publication of Quantitative Real-Time PCR Experiments) guidelines, the amplification efficiencies (E) of reported runs were between 97% and 113% and R² > 0.9 with minimum two technical replicates per reaction. The Bio-Rad SYBR Green Assays used were Nestin (qHsaCED0044457), Tuj1 (qHsaCED0005794), Olig2 (qHsaCED0007834), Gfap (qHsaCID0022307), BDNF (qHsaCED0047199), and Gapdh (qHsaCED0038674). Relative expression data were normalized to the reference gene Gapdh to control for variability in expression levels and were analyzed using the Livak and Schmittgen (i.e., 2^−ΔΔCT) and Pfaffl methods. The relative expression of each target was assessed by unpaired two-tailed t test. A p value of less than 0.05 was considered significant.

Brain-derived neurotrophic factor (BDNF) released into the conditioned medium of drNPC cultures that were differentiated towards a neural lineage was measured by antigen-capture ELISA at different time points and compared to the release of BDNF in the conditioned medium of mature neurons (cat #1520, ScienCell). Conditioned medium from each group was collected, centrifuged, and then stored at − 80 °C until assaying. BDNF concentrations were measured by ELISA kit (BDNF E_max Immunoassay System, Promega Corporation, USA), according to the manufacturer’s instructions. Briefly, 96-well ELISA immunoplates were coated with Anti-BDNF (CatNb#G700B) diluted 1/1000 in carbonate buffer (pH 9.7), and incubated at 4 °C overnight. The following day, all wells were washed with TBS-Tween 0.5% before incubation with Block/Sample buffer 1× at room temperature for 1 h without shaking. After blocking, standards and samples were added to the plates and incubated and shaken (450 ± 100 rpm) for 2 h at room temperature. Subsequently, after washing with TBS-Tween wash buffer, plates were incubated for 2 h with Anti-Human BDNF (1:500 dilution in Block & Sample 1× Buffer) at 4 °C. After incubation, plates were washed five times with TBS-Tween 0.5% wash buffer and 100 μl of diluted Anti-IgYHRP Conjugate was added to each well (1:200 dilution in Block & Sample 1X Buffer) and incubated for 1 h at room temperature with shaking (450 ± 100 rpm). Then, plates were washed five times with TBS-Tween 0.5% wash buffer and 100 μl of TMB One Solution was added to each well. Following 10 min incubation at room temperature with shaking (450 ± 100 rpm) for the BDNF plate, a blue color formed in the wells. After stopping the reaction by adding 100 μl of 1 N hydrochloric acid, the absorbance was read at 450 nm on a microplate reader (Synergy 4) within 30 min of stopping the reactions. Concentration of released BDNF in the supernatants was determined according to the standard curves. BDNF concentrations were compared using an unpaired, two-tailed t test for each time point.

Behavioral analysis was performed using the foot fault task, measuring gross motor functions such as coordination and balance, as well as fine sensorimotor function like reaching and stepping [33], 3 days prior to injury (baseline) and at 3, 8, 18, and 32 days post-stroke; and the cylinder test at 3 days prior to injury and at 3 and 32 days post-stroke. Detailed methods of behavioral tests can be found in the supplementary material. All behavioral tests were recorded with a digital camera (SX 60 HS, Canon) and viewed on VLC Media Player (Version 2.2.1, VideoLAN Organizarion). Videos were scored by a blinded observer.

Statistical analysis was performed using Prism 6 (GraphPad Software, San Diego, CA) and IBM SPSS Statistics (International Business Machines Corp., Armonk, NY). Data was analyzed using a variety of statistical methods which can be found in the supplementary material. All data is reported as mean ± SEM.

To confirm the identity of drNPCs, we characterized their expression of pluripotency and neural markers using immunostaining and polymerase chain reaction (PCR). We examined Oct4 expression as a measure of pluripotency; Sox2 and Nestin for NPCs; Ki67 to assess proliferation; GFAP for astrocytes; TUJ1 for neurons; and Olig2 for oligodendrocytes. We found that drNPCs do not express Oct4, and ubiquitously express NPC markers Sox2 and Nestin, indicating a precursor phenotype (Fig. 1a, Supplementary Fig. 1). In addition, we determined that 71.8 ± 4.0% of drNPCs were Ki67⁺ in vitro (Fig. 1a). We also confirmed that drNPCs have the ability to differentiate into the three neural subtypes and express GFAP, TUJ1, and Olig2 (Fig. 1b). PCR analysis confirmed the expression of neural lineage markers including Nestin, Sox2, Ascl1, Pax6, MAP2, and CD133 and the absence of pluripotency markers Nanog and Oct4 (Fig. 1c). RT-qPCR analysis comparing the expression levels of neural markers Nestin, Tuj1, Olig2, and Gfap in drNPCs prior to culturing (from frozen vials) and drNPCs that were cultured prior to transplantation (cultured drNPCs) revealed similar expression of Nestin and Gfap between the two cell populations and increased expression of Tuj1 and Olig2 in cultured drNPCs (Fig. 1d). Collectively, these results confirm that drNPCs are neurally committed, remain in a precursor state at time of transplant, and can give rise to all three neural cell types.

With the goal of enhancing cell viability at the time of transplant and within the host, we tested two transplant vehicles; (1) artificial cerebrospinal fluid (aCSF), a buffer to mimic circulating CSF, and (2) a hyaluronan methylcellulose hydrogel (HAMC), which has been shown to improve xenograft survival in the CNS [17, 22]. To determine the effect of vehicle on the number of live cells at the time of transplantation, we performed a live/dead assay on cells prior to and following injection through the syringe. Cell viability was determined at 0 h and 2 h (aCSF) or 6 h (HAMC) after cell preparation (Fig. 2a), which was reflective of the longest elapsed time from preparation to transplanting in vivo. Cell viability was always > 88% of the original cell population (aCSF 0 h = 95.5 ± 0.2%, 2 h = 94.5 ± 1.5%; HAMC 0 h = 95.2 ± 0.8%, 6 h = 88.2 ± 0.3%). Cell viability following injection through the syringe, relative to the numbers of viable cells placed in the syringe, was > 95% in all conditions and was not significantly different between vehicles at any time point (Fig. 2b). Thus, all mice received a minimum of ~ 85,000 viable drNPCs at the time of transplantation.

To determine the efficacy of drNPCs for stroke recovery, we used a clinically relevant model of ET-1 stroke in the sensorimotor cortex. This resulted in consistent tissue damage, gliosis, and functional impairment in the foot fault task as early as 4 days post-stroke, which was maintained up to 32 days post-stroke, the longest time point examined (Supplementary Fig. 2). Cultured drNPCs were injected directly into the stroke lesion at 4 days post-stroke based on previous work [17]. Mice were tested in the foot fault task at 3 days prior to stroke to establish a baseline measure, and at 3, 8, 18, and 32 days post-stroke (Fig. 3a). Only stroke-injured mice that had significant motor impairments on the foot fault task by 8 days post-stroke (deficits present at day 3 or day 8) were included in our analysis (aCSF = 11/12 mice; HAMC = 11/13 mice; drNPCs+aCSF = 10/13 mice; drNPCs+HAMC = 14/15 mice). Animals not showing any deficits on either day 3 or 8 post-stroke were excluded from our analysis and completely removed from the study. These exclusion criteria were used to prevent animals that did not actually have a deficit following stroke from skewing the outcomes of the study to falsely showing improved performance.

Stroke-injured mice that received vehicle-only (aCSF or HAMC) injections displayed functional impairments at all times examined, relative to their own baseline performance as well as compared to naïve (uninjured) controls (Fig. 3b). Interestingly, mice that received drNPC transplants in either HAMC or aCSF recovered to their baseline performance by 32 days post-stroke and were not significantly different from naïve performance (Fig. 3b). Our analysis revealed that the observed functional recovery was due to drNPC transplants, irrespective of vehicle (Fig. 3c). We investigated this relationship further by comparing the performance of mice that received drNPCs vs vehicle-alone injections and found that mice that received drNPCs performed significantly better than those that received vehicle-only injections at day 32 post-stroke (Fig. 3d). Functional recovery following drNPC transplantation was seen in both male and female mice (Supplementary Fig. 3a), indicating that drNPC transplants promote recovery regardless of recipient sex.

To further assess functional outcomes, we performed the cylinder test in a subset of mice that demonstrated impairments in the foot fault task post-stroke. Similar to what we observed in the foot fault task, only mice that received drNPCs recovered back to baseline levels, whereas mice that received vehicle-only injections did not recover (Supplementary Fig. 2b). Taken together, these data support the conclusion that drNPC transplants promote functional recovery.

To assess drNPC survival post-transplant, we sacrificed mice at 8 and 32 days post-stroke and stained for human cell markers HuNu and/or STEM121. At 8 days post-stroke (i.e., 4 days post-transplant), all mice (100%) had drNPCs at the site of injection. By 32 days post-stroke (i.e., 28 days post-transplant), the time when functional recovery was observed, drNPCs were only observed in 71% (HAMC, 8/14; aCSF, 9/10) of transplanted brains. A Fisher’s exact test and chi-square test revealed no significant association between vehicle and cell survival (p > 0.05). Interestingly, in all brains that had drNPCs present, irrespective of the time of sacrifice, transplanted cells were confined within the lesion boundary demarcated by NeuN expression, and did not penetrate deep into the uninjured tissue (Fig. 4d).

The total number of viable HuNu⁺ drNPCs within the transplanted brains 8 days post-stroke was 12,780 ± 2963 when delivered in aCSF and 21,633 ± 9880 when delivered with HAMC (Fig. 4a, b). A decline in the numbers of surviving HuNu⁺ cells was observed by day 28 post-transplantation in both vehicles; by 32 days post-stroke, the total numbers of HuNu⁺ cells were 4961 ± 1266 for cells transplanted in aCSF and 5130 ± 1815 for cells transplanted in HAMC (Fig. 4b). However, this decrease was only significant for brains that received cells in HAMC. No significant difference in drNPC survival was observed between vehicles at 8 or 32 days following stroke (Fig. 4b).

Proliferation of drNPCs at 8 and 32 days post-stroke was measured by counting the number of Ki67⁺/HuNu⁺ cells as a percent of all HuNu⁺ cells in brains (Fig. 4a). We found no significant difference between the two times examined in either vehicle and no significant effect of transplant vehicle (Fig. 4c).

We further characterized the in vivo profile of surviving drNPCs (HuNu⁺ or STEM121⁺ cells) at 32 days post-stroke using immunostaining for Sox2 and Nestin (undifferentiated NPCs), GFAP (astrocytes), TUJ1 (immature neurons), NeuN (mature neurons), and Olig2 (oligodendrocytes). Irrespective of the transplant vehicle, the vast majority of surviving drNPCs remain undifferentiated (Sox2⁺, Nestin⁺). A subpopulation of drNPCs expressed GFAP and a small minority expressed the immature neuronal marker TUJ1, but no mature neurons (NeuN⁺) or oligodendrocytes (Olig2⁺) were observed at 28 days post-transplant (Fig. 4e).

Approximately 30% of brains that received drNPC transplants did not have drNPCs present at 28 days post-transplant when functional recovery was observed. We asked whether cell survival was necessary for maintaining functional recovery by comparing behavioral performance in mice with (n = 17) and without (n = 7) drNPCs present at day 32 post-stroke. Our findings reveal that functional recovery occurs irrespective of the presence of cells at day 32 post-stroke (Fig. 5a). Interestingly, mice without cells at 32 days post-stroke had already recovered by 18 days post-stroke. There was no significant difference in performance between the two groups at any time-point tested. Moreover, Pearson’s and Point-Biseral analyses revealed a significant correlation between functional performance at day 32 (compared to baseline) and the injection of drNPCs into the cortex (r = 0.37, n = 46, p = 0.011), but no correlation between whether the drNPCs were present on day 32 post-stroke (r = − 0.17, n = 24, p = 0.43) or with the absolute number of surviving drNPCs (r = 0.21, n = 21, p = 0.36) (Fig. 5b). These findings reveal that long-term transplanted cell survival is not necessary for maintaining functional recovery.

We asked whether drNPC transplants affected the extent of gliosis and the size of the lesion following ET-1 stroke, and whether these outcomes were related to the observed motor recovery. The extent of gliosis was determined using GFAP staining in vehicle and drNPC transplanted mice at 32 days post-stroke (Fig. 6a) by measuring the maximal cortical GFAP⁺ area, which was strongly correlated (r = 0.89, n = 35, p < 0.001) to total gliosis volume per brain (Supplementary Fig. 4). A comparison between the gliotic response in mice that received drNPCs versus vehicle-only injections (n ≥ 16 per group) revealed no significant difference between groups (Fig. 6b). There was also no significant difference between vehicle type in drNPC or vehicle-only treated groups (Supplementary Fig. 5a). Furthermore, a Pearson’s correlation revealed no significant correlation (r = − 0.102, n = 36, p = 0.554) between the size of the glial scar and functional performance at 32 days post-stroke (Fig. 6c).

A similar observation was made comparing the ischemic lesion volume between vehicle-only and drNPC-transplanted mice. Cresyl violet staining revealed no difference in the size of the lesion between mice that received vehicle only versus drNPCs (Fig. 6d, e) or between the vehicle type within the groups (Supplementary Fig. 5b). A Pearson’s correlation also revealed no significant correlation (r = − 0.38, n = 16, p = 0.15) between behavioral recovery and lesion volumes (Fig. 6f). Hence, drNPC transplants do not affect the extent of gliosis or size of the lesion volume post-ET-1 stroke and these tissue outcomes are not correlated with functional improvement.

To investigate the possibility that drNPC transplants influenced recovery by promoting synaptic plasticity, we examined the expression of synaptophysin, a presynaptic vesicle protein, in treated stroke-injured brains. Using immunohistochemistry and confocal imaging (Fig. 7a), we compared the mean pixel intensity (MPI) of synaptophysin expression in the perilesional tissue of stroke-injured mice that received drNPCs (n = 6) and those that received vehicle-only injections (n = 9). Mice that received drNPC transplants had a significant increase (p = 0.012) in synaptophysin expression compared to vehicle-treated mice (Fig. 7b). A Pearson’s correlation also revealed that decreased functional impairment is strongly correlated (r = − 0.80, n = 15, p = 0.0003) with increased synaptophysin MPI (Fig. 7c). This supports the hypothesis that drNPC transplants lead to increased synaptic plasticity which may underlie the observed functional recovery.

To elucidate a potential mechanism by which drNPC transplants exert their beneficial effects, we asked whether drNPCs express and secrete brain-derived neurotrophic factor (BDNF), which is known to promote neuroplasticity [34]. RT-qPCR analysis revealed that BDNF is expressed in both frozen and cultured drNPCs (Fig. 7d). An ELISA analysis of culture medium derived from drNPC cultures that were differentiated towards a neural lineage over time reveals that BDNF is released at levels comparable to mature neurons (Fig. 7e). Hence, BDNF-mediated plasticity may play a role in the functional recovery observed following drNPC transplantation.