Multichannel bridges and NSC synergize to enhance axon regeneration, myelination, synaptic reconnection, and recovery after SCI

Regeneration in the injured spinal cord is limited by physical and chemical barriers. Acute implantation of a multichannel poly(lactide-co-glycolide) (PLG) bridge mechanically stabilizes the injury, modulates inflammation, and provides a permissive environment for rapid cellularization and robust axonal regrowth through this otherwise inhibitory milieu. However, without additional intervention, regenerated axons remain largely unmyelinated (<10%), limiting functional repair. While transplanted human neural stem cells (hNSC) myelinate axons after spinal cord injury (SCI), hNSC fate is highly influenced by the SCI inflammatory microenvironment, also limiting functional repair. Accordingly, we investigated the combination of PLG scaffold bridges with hNSC to improve histological and functional outcome after SCI. In vitro, hNSC culture on a PLG scaffold increased oligodendroglial lineage selection after inflammatory challenge. In vivo, acute PLG bridge implantation followed by chronic hNSC transplantation demonstrated a robust capacity of donor human cells to migrate into PLG bridge channels along regenerating axons and integrate into the host spinal cord as myelinating oligodendrocytes and synaptically integrated neurons. Axons that regenerated through the PLG bridge formed synaptic circuits that connected ipsilateral forelimb muscle to contralateral motor cortex. hNSC transplantation significantly enhanced the total number of regenerating and myelinated axons identified within the PLG bridge. Finally, the combination of acute bridge implantation and hNSC transplantation exhibited robust improvement in locomotor recovery vs. control and hNSC transplant alone. These data identify a successful novel strategy to enhance neurorepair through a temporally layered approach using acute bridge implantation and chronic cell transplantation to spare tissue, promote regeneration, and maximize the function of new axonal connections.

inflammation, and provides a permissive environment for rapid cellularization and robust axonal regrowth 23 through this otherwise inhibitory milieu. However, without additional intervention, regenerated axons 24 remain largely unmyelinated (<10%), limiting functional repair. While transplanted human neural stem 25 cells (hNSC) myelinate axons after spinal cord injury (SCI), hNSC fate is highly influenced by the SCI 26 inflammatory microenvironment, also limiting functional repair. Accordingly, we investigated the 27 combination of PLG scaffold bridges with hNSC to improve histological and functional outcome after SCI.

28
In vitro, hNSC culture on a PLG scaffold increased oligodendroglial lineage selection after inflammatory 29 challenge. In vivo, acute PLG bridge implantation followed by chronic hNSC transplantation demonstrated 30 a robust capacity of donor human cells to migrate into PLG bridge channels along regenerating axons and 31 integrate into the host spinal cord as myelinating oligodendrocytes and synaptically integrated neurons.

32
Axons that regenerated through the PLG bridge formed synaptic circuits that connected ipsilateral forelimb 33 muscle to contralateral motor cortex. hNSC transplantation significantly enhanced the total number of 34 regenerating and myelinated axons identified within the PLG bridge. Finally, the combination of acute 35 bridge implantation and hNSC transplantation exhibited robust improvement in locomotor recovery vs. 36 control and hNSC transplant alone. These data identify a successful novel strategy to enhance neurorepair 37 through a temporally layered approach using acute bridge implantation and chronic cell transplantation to 38 spare tissue, promote regeneration, and maximize the function of new axonal connections.

40
Penetrating spinal cord injury (SCI), which accounts for a significantly smaller proportion of injuries, offers 41 little opportunity for acute neuroprotection, and may require true axonal regeneration for repair. Axons in 4 PLG bridge implantation modulates the cellular immune response in vivo by prolonging the 88 macrophage/microglial response.

89
We have previously shown that SCI is associated with a multiphasic immune cell response and SCI 90 inflammatory microenvironment modulates hNSC fate, migration, and potential for repair after in vivo 91 transplantation [31,32,34]. In parallel, we have reported that macrophages migrate to and phagocytose 92 implanted biodegradable PLG bridges [14,28]. Critically, no deleterious effects of PLG are observed in 93 vitro, or in vivo after bridge implantation [14,16,17,28,[35][36][37], and lactate, a PLG biodegradation product, 94 polarizes immune cells towards tolerogenic phenotypes [38,39]. Accordingly, we first tested how PLG 95 bridge implantation modulates the innate immune cell response within the bridge and surrounding spared 96 spinal cord tissue at different time points following injury. We utilized a quantitative flow cytometry-based 97 method to characterize this multiphasic response in a hemisection injury model for PLG bridge and SCI 98 control animals (Fig. 1A).

99
PLG bridge and SCI control groups behaved similarly until 24 weeks post-injury (WPI), where the number 100 of myeloid cells were significantly reduced in PLG vs. SCI control mice at 24 WPI ( Fig. 2A). We next 101 investigated the proportion of Ly6G + polymorphonuclear leukocytes (PMN) and CD68 + macrophages 102 (MØ)/microglia in the total myeloid population. As demonstrated previously after contusion SCI in rodents 103 [34], the epicenter environment was dominated by PMN in both SCI control and PLG groups at 1 day post-104 injury (DPI) (Fig. 2B). PMNs were dramatically reduced at 1 and 4 WPI. Similarly, there a delayed re-105 emergence of the PMN population observed at 8 and 24 WPI. This second phase of PMN infiltration was 106 still present, but significantly reduced, in PLG bridge vs. SCI control mice (Fig. 2B). As expected from 107 previous analysis in contusion SCI, MØ/microglia populations were sparsely detected at 1 DPI (Fig. 2C) 108 and exhibited a later peak in response to injury. However, while the PLG bridge and SCI control groups 109 were similar at 1 WPI, the MØ/microglia subpopulation peak was significantly extended in PLG bridge vs.

134
PMN-CM enhanced astroglial fate and suppressed neuronal fate, while MØ-CM had no effect on astroglial 135 fate and enhanced neuronal fate [32]. hNSC cultured on PLG substrate modulated these effects, reducing 136 astroglial fate at baseline and in response to PMN-CM and MØ-CM, and enhancing neuronal fate in 137 response to PMN-CM and MØ-CM (Fig. 3H-U). However, PLG showed no significant effect vs.

138
PLO/LAM on the total number of cells after 14 days in vitro (DIV) in differentiation media (DM) via nuclei 139 (Fig. 3V). Collectively, these data indicate that PLG substrate alters hNSC fate at baseline and in response 140 to immune cues.

141
The enhancement of baseline oligodendroglial fate, and suppression of astroglial fate, by culture on PLG

159
We selected the timing of chronic hNSC transplantation based on the following considerations. First, axons 160 enter and extend into the channels of the PLG bridge by 4WPI. We predicted that these newly formed axons 161 would strongly cue hNSC to migrate and undergo oligodendroglial lineage selection, enabling myelination 162 of axons within the PLG bridge [48,49]. Second, the SCI epicenter environment is dominated by MØ rather 163 than PMN at 4WPI (Fig. 2). In vitro testing showed that, in the presence of MØ-derived cues, hNSC 164 oligodendroglial lineage selection was rescued, astroglial lineage selection was reduced, and neural lineage 165 selection was maximized by PLG scaffold -optimizing potential for repair. Third, chronic transplants are 166 a more clinically relevant time period for delivery of cellular therapies in humans, enabling both improved 167 informed consent and a more medically stable population given the requirement for immunosuppression in 168 an allogeneic transplant setting, minimizing serious adverse effects [50].

169
We have reported that immunodeficient models lacking T cells are required to enable both xenogeneic and 170 allogeneic CNS engraftment at levels sufficient to analyze sustained effects of donor NSC in the host 171 environment [29,51]. These models are similar to the translational setting, in which allogeneic stem cell 172 transplantation requires long-term administration of pharmacological immunosuppressants in humans, 173 which similarly target lymphoid cells. Accordingly, these studies utilize Rag1 mice, which lack mature T 174 and B cells but retain a functional innate immune cell response [52]. We first confirmed that the innate 175 immune cell profile in Rag1 mice for the hemisection injury (Supplemental Fig. 1) is similar to the profile 176 observed in C57BL/6 mice (shown in Fig. 2). As expected, Rag1 mice have fewer innate immune cells 177 (CD45 + CD11b + cells) in comparison to C57BL/6 mice, since they lack T-cells which orchestrate the innate 6 immune response ( Fig. 2A, Supplemental Fig. 1A). Importantly however, PMN and MØ/microglia cell 179 proportions closely paralleled that of C57BL/6 mice ( Fig. 2D-E, Supplemental Fig.1B-C) after SCI, from 180 1 DPI through 24 WPI, demonstrating that innate immune responses are, in essence, conserved between 181 these models, which would be predicted to be similar to pharmacological immunosuppression in humans.

182
hNSC were transplanted into the spared parenchyma rostral and caudal to the SCI control or PLG bridge 183 implanted lesion site, we therefore first evaluated hNSC engraftment, migration, and fate in these regions 184 at 16 weeks post-transplantation (WPT) (Fig. 1C, Fig. 4). Spinal cord transverse sections were aligned

199
We next determined whether donor hNSC were competent to migrate into the SCI control or PLG bridge

214
We have previously shown that PLG bridge implantation stabilizes adjacent spinal cord tissues [14]. We 215 sought to test whether hNSC that engrafted in the spared tissue directly contralateral to the site of PLG 216 bridge implantation vs. SCI control exhibited differences engraftment or fate. The number of total engrafted 217 STEM121 + hNSC was significantly increased in the PLG group compared to the SCI control (Fig. 5I), as 218 was the number of STEM121 + /Olig2 + cells (Fig. 5J). While there was a parallel trend for an increase in the 219 number of immature STEM121 + /DCX + cells (p-value =0.06, Fig. 5K), this was not observed for either 220 mature STEM121 + /NeuN + cells (Fig. 5L) or STEM121 + /GFAP + hNSC (Fig. 5M). Critically, the proportion 221 of hNSC adopting different lineages was not significantly different between the PLG bridge vs. SCI control 7 groups (data not shown). These data suggest that the principal effect of the PLG bridge was on overall 223 survival/engraftment of hNSC in the spared tissue, which is consistent with the stabilization of this region 224 in the acute period after injury. This difference in engraftment cannot be ascribed to an increased 225 area/volume but may rather reflect a change in the molecular microenvironment and/or mechanostability

229
In previous studies using contusion SCI models, we have shown that hNSC that are localized adjacent to 230 the SCI epicenter are directed towards the astroglial lineage, but that these cells do not migrate into the SCI 231 epicenter per se [29, 31, 44]. As described above, hNSC that migrated along the PLG channels retained tri-232 lineage potential. Here, we asked whether the fate of hNSC that migrated into the PLG bridge and 233 inflammatory environment of the SCI epicenter was similar to or different from hNSC that engrafted in 234 contralateral spared tissue in the same tissue sections. While hNSC that migrated into the PLG bridge 235 retained the capacity to differentiate along the oligodendroglial lineage (STEM121 + /Olig2 + ; Fig. 5N

257
Although histological quantification was predominantly performed at 16 WPT, a subset of animals was 258 assessed at 26 WPT. We hypothesized that this additional engraftment time would optimize for the 259 maturation and integration of transplanted hNSC as myelinating cells within the PLG bridge. We therefore 260 tested whether hNSC that entered the bridge channels (white dotted lines in Fig. 6J) were capable of 261 myelinating regenerated axons. Indeed, there was an abundant association of STEM121 + hNSC labeling 262 aligned with and in close proximity to NF-H + axons (Fig. 6J, arrowheads). Further, high magnification 263 clearly identifies MBP + /STEM121 + co-labeling of NF-H + axons (Fig. 6K), indicating that hNSC contributes 264 to myelination of axons that regenerate into PLG bridge channels.

8
We confirmed this observation and tested whether donor NSC myelinated regenerating descending motor 266 axons using CRYM reporter mice. We have previously employed these mice to specifically visualize CST

271
This approach allowed mature, compact myelin from donor mT-mNSC to be visualized in association with 272 regenerated CST axon targets within the PLG bridge. We employed pharmacological immunosuppression 273 for these allogeneic transplants because CST-GFP reporter mice were not on a constitutively 274 immunodeficient background. While this resulted in reduced mT-mNSC survival and migration in 275 comparison with transplantation into Rag-1 mice, we were able to identify mT-mNSC aligning with and 276 ensheathing NF-H + (Fig. 6L) and GFP + CST (Fig. 6M) axons in the PLG bridge. Orthogonal projection 277 clearly indicated that mT-mNSC fully surrounds GFP + CST fibers (Fig. 6M1, blue arrowhead).

278
The PLG bridge supports regeneration of a synaptic circuit connecting forelimb muscle to motor 279 cortex.

280
An advantage of the PLG bridge implantation model is that axons can only reach the channels by 281 regenerating; they cannot represent spared fibers. However, observation of axons in this region, even CST 282 reporter axons, does not mean that these fibers have integrated into synaptic circuitry. We sought to test 283 this key aspect of regeneration via transsynaptic PRV tracing. We injected GFP-reporter PRV virus into the 284 forelimb triceps muscle ipsilateral to the SCI hemisection (Fig. 1C). PRV is retrogradely transported from 285 the neuromuscular junction, sequentially infecting synaptically connected neurons to enable circuit tracing;  Fig. 4B). Surprisingly, we found a reduction of retrograde PRV 300 labeling in 10-month-old naive mice (Supplemental Fig. 4B, B1-3

314
The contralateral motor cortex exhibited clear evidence of PRV + neurons in PLG bridge implanted vs. SCI 315 control groups (Fig. 7D-F). While not significant, there was a trend for fewer PRV + neurons in the PLG 316 bridge + hNSC combination vs. PLG bridge + vehicle group (Figs. 7F). hNSC in the spinal cord were also 317 PRV + , suggesting stable integration of donor human cells into mouse host circuitry (Fig. 7H), which could 318 have diluted PRV transport to the brain because of an increased number of intermediate synaptic In sum, therefore, we interpret these data to demonstrate that motor cortex axons can regenerate 337 through the PLG bridge and form synaptic connections with forelimb muscle, reconstituting motor circuitry.

338
Moreover, we demonstrate novel evidence for the capacity of donor human NSC to similarly exhibit 339 synaptic integration within a mouse host. Consistent with findings in Crym reporter mice (Fig. 6M), PRV + 340 axons exhibited close proximity to STEM121 + hNSC labeling, suggesting donor cell interaction and 341 myelination of synaptically connected host axons (Fig. 7I).

342
PLG bridge implantation and hNSC transplantation improve locomotor recovery.

369
We also conducted CatWalk kinematic gait analysis (Fig. 8C), applying an unbiased multivariate approach 370 to test whether the effect of PLG bridge implantation on locomotion could be separated from that of hNSC 371 transplantation on locomotion. This approach avoids a priori assumptions about which variables are 372 meaningful to recovery. Univariate two-way ANOVAs were conducted in R using the "aov" (Analysis of 373 Variance) function to analyze the independent effects of PLG bridge implantation and hNSC 374 transplantation, identifying variables that had a p-value ≤ 0.05 (Fig. 8C). PLG bridge implantation and

436
In sum, we demonstrate that hNSC transplanted in the chronic period after acute PLG bridge implantation 437 are capable of migrating along regenerating axons into the PLG bridge channels, and enabling both donor

452
Animal models

472
New York, NY) to prevent muscle adhesion to the PLG bridge. SCI control groups received only gelfoam 473 at the lesion site. Rag1 mice were 11-19 weeks of age and CRYM-ZsGreen1 transgenic mice were 9-15 474 weeks of age at the time of injury. Mice were randomly allocated to different treatment groups. The exposed 475 muscle was sutured using 5-0 chromic gut, and the skin was closed using wound clips. All mice were placed 476 in cages on top of heating pads at 37°C overnight with clean Alpha-Dri bedding. The following 477 subcutaneous injections were administered post-op: Baytril (2.5mg/kg) once a day for two weeks, lactated 478 ringers (50mL/kg) once a day for five days, and buprenorphine (0.05mg/kg) every 12 hours for three days.

479
Throughout the course of the study, bladder expression was performed manually twice per day. Since Rag1     Table 1. Images were captured with 10X objective using random sampling with a ZEISS 562 Axio Imager II light microscope with an Apotome2 image processor (Zeiss, Oberkochen, Germany). Imaris 563 software (Oxford Instruments, Abingdon, United Kingdom) was used to quantify hNSC fate. Image 564 acquisition and quantification were conducted by researchers that were blinded to the experimental groups.

577
ZsGreen1 transgenic mice as previously described [29,31,44]. After re-exposure of the spinal cord at C5, 578 cells were injected using a NanoInjector with micromanipulator (World Precision Instruments, Sarasota, 579 FL) and siliconized beveled glass micropipettes (outer diameter = 100-110 mm; inner diameter = 70 mm; 580 Sutter Instruments, Novato, CA). A total of 75,000 cells or vehicle (X-VIVO media) in 1uL total volume 581 was delivered to the spared spinal parenchyma at four sites, two rostral and two caudal to the lesion, 250nL 582 per site (schematic in Fig. 1C). All mice received post-operative care as described above.

611
Hampton, NH) to be processed into 30 μm thick sections with a cryostat and Cryo-Jane tape transfer system 612 (Leica Biosystems, Wetzlar, Germany). Spinal cords were sectioned either transversely or horizontally, and 613 the slides were stored at -20˚C until processed for immunohistochemistry.

614
Immunohistochemistry and imaging 615 Slides with spinal cord tissue sections (CryoJane tape transfer slides) were incubated at 60 0 C on a hot plate 616 for 2 hours, dipped in Histo-Clear II clearing agent (National Diagnostics, Atlanta, Georgia) for 20 minutes, 617 rehydrated in a descending ethanol gradient (100%, 95%, 80%, and 70%) 5 minutes in each solution, and 16 hydrated in distilled water for 10 minutes. Then for antigen retrieval, slides were immersed in a preheated 619 sodium citrate antigen retrieval solution (10 mM sodium citrate, 0.05% Tween 20, pH 6.0) for 30 min at 96 620 °C in a water bath. Following a 20-minute cool down at room temperature, the slides were rinsed in water 621 and processed for immunohistochemistry as described below. Spinal cord sections (CryoJane tape transfer 622 slides) and brain sections (microtome, free-floating) were incubated with blocking buffer (1.5% donkey 623 serum and 0.1% Triton X in PBS) for an hour at room temperature and then incubated with primary antibody 624 (diluted in blocking solution) overnight at room temperature. After 3 five-minute washes (0.1% Triton-X 625 100 in PBS), tissue was incubated with appropriate fluorescent-dye conjugated secondary antibodies and 626 nuclear dye Hoechst 33342 (1 μg/mL dilution; Invitrogen, Waltham, MA) for 2 hours at room temperature.

627
Following 3 final five-minute washes (0.1% Triton-X 100 in PBS), slides were mounted with Fluoromount 628 G (SouthernBiotech, Birmingham, AL). Images were captured using a ZEISS Axio Imager II light 629 microscope with an Apotome2 image processor or a ZEISS LSM 900 with Airyscan 2 microscope (Zeiss,

630
Oberkochen, Germany).   Table 1. For quantification, four to six random 657 optical fields within the lesion per section were imaged using a Zeiss LSM 900 with Airyscan super 658 resolution microscope. For each optical field, 6 μm Z stack images (0.3 μm Z-step) were captured using a 659 60X oil objective. 3D surface volume rendering was performed using Imaris v9.6 (Oxford Instruments,

660
Abingdon, United Kingdom). Briefly, all three volumes of total neurofilament, total MBP positive 661 myelinated neurofilament, and total P0 positive myelinated neurofilament were masked using the Surface 662 feature. To exclude excess noise, a filter was made to a minimum voxel size of around 1000 for each image.

663
when masking the surface volumes. Next, the NF-H positive surface volume of neurofilament that was associated with the oligodendrocyte-derived myelin (MBP + POvolume) and Schwann cell-derived myelin 665 (MBP + P0 + and MBP -P0 + ) was masked using the object-to-object shortest distance (0.4 um) filter in Imaris 666 software.

668
Spinal cord tissue for these analyses was collected at 30 WPI and sectioned horizontally at 30μm thickness 669 using CryoJane tape transfer method. Immunostaining for ⅙ sampling interval sections was performed to 670 detect PRV-GFP labeled fibers. For quantification, five to six random optical fields within the lesion site 671 per section were imaged using Zeiss LSM 900 with Airyscan super resolution microscope. Antibody source 672 and the dilutions were used as listed in Supplemental Table 1. For each optical field, 24 μm Z stack images 673 (0.4 μm Z-step) were captured using 60X oil objective. PRV-GFP labeled filament volumes were manually 674 traced using Imaris v9.6 filament manual tracing software (Oxford Instruments, Abingdon, United 675 Kingdom). Brain sections were sectioned coronally at 30μm using a sliding microtome. Immunostaining

676
(all sections containing motor cortex, no sampling) was performed to detect PRV-GFP cell bodies. All 677 PRV + cell bodies within the motor cortex were counted manually. Quantification was performed at 20X 678 magnification using ZEISS Axio Imager II light microscope with an Apotome2 image processor. Image 679 acquisition and quantifications were performed by investigators blinded to the experimental groups.

681
All behavioral data were collected and analyzed by observers blinded to the experimental groups. Ns as 682 indicated under statistical analysis and exclusions below. Mice were handled daily for two weeks prior to 683 pre-injury behavioral testing to acclimate animals to human contact. Horizontal ladder beam and CatWalk 684 Gait Acquisition were performed to quantify changes in motor recovery and kinetic parameters after injury.

685
For Rag1 mice, locomotor assessments were acquired pre-injury, pre-transplantation, and 16 WPT  randomized across groups to ensure an unbiased distribution by age and weight by an investigator not 704 involved directly in the study. A total of 17 Rag-1 mice in the main transplantation study were excluded 705 using pre-hoc criteria by an investigator blinded to the experimental groups and not involved directly in the 706 study. Exclusion criteria: excessive weight loss, autophagy, staph infection of the hair follicles, or surgical 707 complication (n =12); anatomical defect (n =2); failed bridge apposition (n = 1); Exclusion criteria: 708 18 excessive weight loss, autophagy, staph infection of the hair follicles, or surgical complication (n =12); 709 anatomical defect (n =2); failed bridge apposition (n = 1); failed transplantation/engraftment (n =1); and 710 Grubbs' test for outliers (n=2). Final Ns for behavioral analysis were as follows. Ladder beam: SCI control 711 + Vehicle (n=10), SCI control + hNSC (n=10), PLG bridge + Vehicle (n=9), and PLG bridge + hNSC 712 (n=12). Catwalk analysis: SCI control + Vehicle (n=8), SCI control + hNSC (n=9), PLG bridge + Vehicle 713 (n=9), and PLG bridge + hNSC (n=12). For Catwalk, two mice in SCI control + Vehicle group and one 714 mouse in the SCI control + hNSC group could not perform the task and were excluded from statistical 715 analysis for this task but were retained for horizontal ladder beam analysis and histology.

717
The data that support the findings of this study are available from the corresponding author upon 718 reasonable request.