Delayed viral vector mediated delivery of neurotrophin-3 improves skilled hindlimb function and stability after thoracic contusion

Intramuscular injection of an Adeno-associated viral vector serotype 1 (AAV1) encoding Neurotrophin-3 (NT3) into hindlimb muscles 24 hours after a severe T9 spinal level contusion in rats has been shown to induce lumbar spinal neuroplasticity, partially restore locomotive function and reduce spasms during swimming. Here we investigate whether a targeted delivery of NT3 to lumbar and thoracic motor neurons 48 hours following a severe contusive injury aids locomotive recovery in rats. AAV1-NT3 was injected bilaterally into the tibialis anterior, gastrocnemius and rectus abdominus muscles 48-hours following trauma, persistently elevating serum levels of the neurotrophin. NT3 modestly improved trunk stability, accuracy of stepping during skilled locomotion, and alternation of the hindlimbs during swimming, but it had no effect on gross locomotor function in the open field. The number of vGlut1 + boutons, likely arising from proprioceptive afferents, on gastrocnemius α-motor neurons was increased after injury but normalised following NT3 treatment, suggestive of a mechanism in which functional benefits may be mediated through proprioceptive feedback. Ex vivo MRI revealed substantial loss of grey and white matter at the lesion epicentre but no effect of delayed NT3 treatment to induce neuroprotection. Spasms and hyperreflexia were not reliably induced in this severe injury model suggesting a more complex anatomical or physiological cause to their induction. We have shown that delayed intramuscular AAV-NT3 treatment can promote recovery in skilled stepping and coordinated swimming, supporting a role for NT3 as a therapeutic strategy for spinal injuries potentially through modulation of somatosensory feedback.

reduced the presence of hindlimb spasms during swimming and normalised hyperreflexia in the hind paw. There was, however, no synergistic effect of rehabilitation and NT3 treatment on functional and electrophysiological outcomes.
We hypothesised that intramuscular injection of AAV1-NT3 into muscles innervated by thoracic and lumbar spinal motor neurons would enhance skilled motor function and potentially alleviate hindlimb hyperreflexia and spasticity if delivered at a clinically relevant 48-hours following the initial trauma. We show that severe thoracic contusion led to persistent hindlimb deficits but little, if any, spasm, or hyper-reflexivity. Nevertheless, 48h-delayed NT3 treatment improved skilled locomotor function and hindlimb alternation during swimming potentially due to modifications of vGlut1 signalling on proprioceptive afferents.

Ethical approval and animal welfare
Experiments were approved by the King's College London Welfare and Ethics Committee and were conducted in accordance with UK Animals (Scientific Procedures) Act 1986 (ASPA) under Home Office Project Licence number P53631BC2. During all experiments, data processing, and analysis the investigators were blind to the treatment group of each animal. An independent third party coded and randomly assigned animals to treatment groups in an alternating pattern prior to injury without knowledge of baseline behavioural assessments and electrophysiological recordings.
Animals were housed in groups of four or five, exposed to a normal 12-hour dark-light cycle at 21°C with environmental enrichment and access to food and water ad libitum. The health and welfare of all animals was monitored daily by veterinary staff and the study investigators at King's College London and was in accordance with the Animal Welfare Act 2006. 28 female Wistar rats (256g ± 29g; Envigo; RccHan:WIST) were used in this study. Of the 28 animals, one died three days after contusion injury, with all other animals making uneventful recoveries. One animal was excluded from the study due to regaining near perfect locomotion in open field (BBB score of 20) within three weeks of injury. Animals were divided into three groups: 1) contusion + NT3 (NT3, n=11); 2) contusion + PBS (PBS, n=12); and 3) uninjured + no treatment (sham, n=4). PBS was used as a control group as it was the vehicle used for treatment application (Tang et al., 2020, Johnston et al., 2021, Rghei et al., 2021, Aguilà et al., 2020, Sahenk et al., 2014b. Our previous work, and that of others, has shown that intramuscular PBS treated animals for the duration of the study. Sham animals were housed together in a separate cage.

Thoracic contusion injury
Rats were induced with Isoflurane (5%; Zoetis, UK Ltd) at 1 L.min -1 O2 flow and maintained with 2.5% isoflurane throughout surgery. Preoperatively 5mg/kg Enrofloxacin (Baytril) and 5 mg/kg carprofen (Carprieve) were injected subcutaneously (s.c.) and a homeothermic heat pad (Harvard Apparatus) maintained body temperature at 37±1C. Upon reaching a surgical plane of anaesthesia, eye ointment (Viscotears) was applied and the region around the lower thoracic vertebrae was shaved and the skin sterilised.
A 2cm dorsal midline incision from T8-T11 vertebrae was performed and the spinotrapezius muscles retracted to expose the vertebral column. A complete laminectomy of the T8 vertebra was performed and the vertebral column stabilised in the Infinite Horizon contusion impactor (Precision Systems and Instrumentation). Animals received a single bilateral contusion, of 250 kDyn force with 0s dwell time using a 2.5mm diameter impact tip (Precision Systems Instrumentation), at the T9 spinal segment.
Muscle layers and the skin were sutured in two layers (4-0 Vicryl; Ethicon) and animals given saline (5mL; s.c.). Sham animals received a full T8 laminectomy, without clamping in the Infinite Horizons impactor, and all pre-and post-operative drugs.
For five days after injury animals continued to receive; warm saline (s.c.), 5 mg/kg Baytril, and 5mg/kg Carprieve. Animals were unable to void their bladders naturally for the first week after injury, necessitating manual bladder emptying three times per day. Once spontaneous voiding was recovered, bladder volume and voiding was monitored daily to ensure no infections occurred.

Adeno-associated virus encoding neurotrophin-3
Forty-eight hours after contusion, the tibialis anterior and gastrocnemius muscles of both hindlimbs as well as rectus abdominus were injected with AAV-NT3 or PBS ( Figure 1A). Animals were prepared similarly to above with both hindlimbs and the abdomen shaved. Injections were made using a 26G non-coring beveled needle attached to a Hamilton syringe. A total of 2.57 x10 12 gc of AAV-NT3 diluted in 220μl of PBS (or equivalent amount of PBS vehicle) was injected into each animal. Targeting motor endplates was achieved by deep injections positioned in a circumferential arc across the tibialis anterior and gastrocnemius muscles (Yin et al., 2019) (Figure 1B). For the rectus abdominus all injections were superficial, with care taken not to penetrate through the muscle into the peritoneal cavity.
Tibialis Anterior: Injections were made across the proximal aspect of the muscle, approximately two thirds the distance from the ankle to the knee. A total of 6x5μl of AAV-NT3 (3.50x10 11 gc) or PBS was delivered per muscle ( Figure 1C).
Gastrocnemius: A total of 6x5μl of AAV-NT3 (3.5x10 11 gc) or PBS were injected into both the medial and lateral heads, totalling 7.1 x10 11 gc in each gastrocnemius.
Rectus Abdominus: Rectus Abdominus was exposed from the xiphoid process down to the pubic symphysis. Two 5μl injections of AAV-NT3 (4.67 x10 11 gc) or PBS were made between the 4 th -5 th and 5 th -6 th tendinous insertions, to target the region innervated at or below the level of thoracic contusion (Hijikata et al., 1992) (Figure 1C). In total, 20μl was delivered to each side.

Behavioural functional assessment
Open field locomotor assessments Locomotor functional recovery was monitored using the 0-21 point Basso, Beattie and Bresnahan scale (BBB) (Basso et al., 1995). One week prior to injury, animals were habituated to a circular Perspex open field assessment area (100cm diameter, 20cm high) for ten minutes each day. Testing was done in the afternoon at day -1, and then weekly following injury during a five-minute period by two assessors.
Analysis: The BBB score for each hindlimb were averaged between the two assessors and the BBB sub-scores similarly calculated (Popovich et al., 2012).

Horizontal ladder
Accuracy of stepping was assessed through the animals' ability to cross a 1m horizontal ladder with rungs spaced randomly 1-3cm apart. Prior to injury, animals were trained to cross the ladder for 15 min periods each day and baseline values obtained within 5 days prior to injury. Testing comprised video recording three complete runs weekly following injury.
Analysis: Footage was slowed 5x original speed in VLC media player (VideoLan) and steps by both hindlimbs was assessed using the seven-point horizontal ladder scoring system (Metz and Whishaw, J o u r n a l P r e -p r o o f 2002). This 7-point category scale assessed the accuracy of foot placement on ladder rungs (including fall, slip (deep or slight), replacement, correction, imperfect partial placement, and perfect placement) with each type of positioning demonstrating an ability to perform independent skilled functions. Briefly; a 'miss' occurs when the hindlimb fails to connect with a rung, and a 'slip' when the limb bears weight but slips off the rung with or without interruption to gait cycle. A "corrective step" is defined as when the hindpaw contacts the rung but is then repositioned. Finally, a 'hit' is a placement which results in a successful step.

Swimming
To assess non-weight bearing locomotor function, animals swam in a rectangular glass chamber (120cm x 12cm x 50cm) filled with 25cm water at 20-23°C (Ryu et al., 2017. A Perspex mirror was placed at a 45° angle on the base, allowing recording of ventral and lateral views of the animal (60 frame per second at 1920x1080 pixels, HERO7 Black, GoPro).
Following a two-week habituation period, baseline recordings were collected within 5 days prior to injury. Rats placed at one end of the chamber would swim 80cm, equalling one run, to an accessible platform ( Figure 3A) at the other end. Animals were assessed weekly from three weeks following injury with five complete runs, recorded during the same time of day.
Analysis: Hind-limb swim strokes from all runs were analysed in slow motion using VLC media player (5x original speed, VideoLan). The percentage of left-right alternating and total number of strokes was calculated for each limb. Hindlimb strokes were excluded if they: occurred outside the 80cm assessment area; were associated with a change in direction; or were the initial stroke following an interruption to swimming. The number of coordinated hind limb strokes over all the runs performed by each animal was displayed as a proportion of total number of hindlimb strokes performed.

H reflex recordings
Animals were anaesthetised with ketamine and medetomidine (30mg/kg and 0.10 mg/kg i.p respectively) and treated with atropine sulphate (0.1 mg/kg s.c) and carprofen s.c (Carprieve 5 mg/kg s.c.). Body temperature was maintained at 37±1C using a homeothermic heat pad (Harvard Apparatus) and the ankle and hindpaw cleaned with 4% chlorhexidine. Recordings were taken from the left hindlimb at final timepoint, or from the limb with the higher BBB subscore (n=4). Two 26-gauge needles were positioned, subcutaneously in parallel and approximately 2mm apart, over the medial malleolus of the ankle to stimulate the medial plantar nerve. To record compound muscle action potentials (CMAPs) from the intrinsic hindpaw muscles innervated by the medial plantar nerve, two non-insulated needle recording electrodes were positioned at the base of first digit between the two prominent paw pads and placed superficially, 1mm apart and 2mm deep (Fig 4B).
A monophasic 100µs square wave stimulus was applied using an isolated constant current pulse stimulator (DS3, Digitimer) and the CMAP signal (amplified 2000x, band-pass filtered between 300Hz -6kHz) digitized via a Power1401-3A unit (Cambridge Electrical Design Ltd, CED) and visualised in J o u r n a l P r e -p r o o f Signal software (v5, CED), recording in 5000ms sweeps. Motor threshold was defined as the stimulus at which the M wave was elicited in three out of four stimuli. A recruitment curve was generated by recording 15 stimuli at 0.2Hz, starting at 1x motor threshold (MT) to a maximum of 2x MT in 10% increments up to 1.5xMT and finally at 2xMT. Subsequent rate dependent depression (RDD) was assessed with a paired stimulation protocol (4) at 1.3xMT stimulus intensity. Fifteen paired pulses were delivered 300ms into the sweep at various interstimulus intervals (ISI. Fig 4A) Once recordings were finished, 2mg/kg (i.p.) atipamezole hydrochloride and 5mL saline (s.c.) was injected.
Analysis: Amplitude of the M and H waves was calculated in Signal software (v5, CED). The recruitment curve yielded the maximum H wave amplitude (Hmax) and maximum M wave amplitude (Mmax). For each ISI, the 15x M wave and H wave amplitudes corresponding to the conditional stimulus and the 15 for the test stimulus were pooled and averaged. These averages were then normalised to the Mmax for that testing session. Finally, RDD was calculated by representing the normalised H wave of the test stimulus as a percentage of the H wave of the conditioning stimulus.

Retrograde tracing
Cholera Toxin Subunit B (CTb, #103, List Biologicals) was used to retrogradely trace motor neurons innervating the lateral head of the gastrocnemius in the left hindlimb. Animals were prepared and the gastrocnemius visualised as per the previous muscle injections. A total of 5x1μl of 0.5% Cholera Toxin B (diluted in ddH20) was injected into the lateral muscle head, following a similar trajectory to treatment injections, using a 26G needle on a Hamilton syringe. Following surgery, animals were given 5ml of saline and the following day a single dose of Carprieve.

Tissue collection
Animals were deeply anaesthetized with sodium pentobarbital (Euthatal, 80 mg/kg i.p.). 1.5ml of blood was taken through cardiac puncture of the left ventricle and stored overnight at 4°C before processing. Animals were transcardially perfused with PBS. The left gastrocnemius, including both the medial and lateral heads, was removed, snap frozen, and stored at -80°C. Animals were perfused with 4% paraformaldehyde in 0.1M phosphate buffer. The spinal cord from C3 to the cauda equina was removed and post fixed in fresh 4% PFA in PBS for 24 hrs at 4°C.

ELISA & BCA quantification
Levels of NT3 protein in the serum were assessed using a Human Neurotrophin-3 ELISA Kit (#ab100615 Abcam), which can detect vector-derived and endogenous NT3. Whole blood was centrifuged (7200g, 25C), the serum removed and stored at -80°C. ELISA was performed as per the manufacturer's instructions and the plate read immediately at 450nm using a BMG LabTech FLUOStar (Omega). NT3 levels were normalised to the total amount of protein extracted from each sample determined by a Bicinchoninic acid assay (BCA) assay (#71285-3, Novagen, Millipore).

J o u r n a l P r e -p r o o f
Serum samples were diluted 1:150 in PBS, and a four-parameter standard curve was generated (above r>0.99).

Ex vivo MRI
For lesion quantification the epicentre of the lesion was imaged using a 9.4 T MRI scanner (Brucker Biospec). T2 weighted (T2W) images were acquired using a fast spin-echo sequence: echo train length = 4, effective TE = 38 ms, TR = 3000 ms, FOV = 40 x 20 x 20 mm, acquisition matrix = 400 x 200 x 200, acquisition time = 9 h 20m. A custom-made device enabled simultaneous scanning of the T6-L1 regions of 14 spinal cords. Tissue was fully submerged in Fomblin (Solvay) and all air bubbles removed prior to scanning. The spinal cords were prepared for scanning as described previously (Sydney-Smith et al., 2021a) with additional care taken due to the reduced stability of the spinal cord around the epicentre.
Analysis: Convert3D (ITK-SNAP) was used to alter the voxel dimensions to isotropic voxel size of 50μm. The composite image was separated into individual spinal cords to final voxel dimensions of 30x100x30μm in the X x Y x Z axis respectively. Within ITK-SNAP, the contrast was increased (minimum adjustment:0.03, maximum adjustment:0.82, levels:0.41, window:0.82) and automatic segmentation performed for: presumptive spared white matter (lower limit=0.18, upper limit=0.28) and total spinal cord (lower limit=0.00, upper limit=0.29). The segmentation parameters were: radius 2.5, competition force =1.000, smoothing force = 0.2, α=1.000, β=0.100, speed = 1.00. All parameters were identical for each spinal cord. A screenshot series of the transverse views of the mask and images every 100μm were quantified in FIJI 9 (U.S. National Institutes of Health). Volume for each parameter was calculated as the thickness of the voxel multiplied by the cross-sectional area. The cross-sectional area of the tissue was measured in a region 4.1mm in length and encompassing the epicentre.
Analysis: Analysis was performed in FIJI (NIH) by measuring raw integrated density within a 2.5μm band surrounding the soma, and manually counting distinct vGlut1+ puncta abutting the motor neuron.
Locomotion behavioural assessments were analysed using a linear model with a suitable covariance structure as described previously (Duricki et al., 2016c). For analysis of the horizontal ladder, for each dependent variable (e.g., Slips, Hits or Miss) in turn, we specified a linear mixed model in which we included limb (Left or Right) as a factor. Because these analyses often revealed a difference between limbs (Supplementary Table 1), we also analysed each limb separately and report these values in the Results section.
All data sets were found to be normally distributed by viewing frequency histograms and/or satisfying Kolmogorov Smirnov and Shapiro-Wilk tests. Error bars in graphs are mean ± Standard error of mean. Divergences were considered significant if P<0.05. In all figure panels: NT3 = blue triangles, controls = orange squares, sham = black diamonds, * = P<0.05, ** = P<0.01, *** = P<0.001, and **** = P<0.0001. If no post-hoc result is shown, comparison was not significant.

NT3 treatment recovered skilled motor function and coordination following injury
A contusion injury of 250kdyn at the T9 spinal level produced a severe SCI in 23 adult female Wistar rats. Animals were randomised into either an NT3 group (n=11, receiving a total of 2.57 x10 12 gc of AAV-NT3 in 220μl PBS) or a control group (n= 12, receiving 220μl PBS). Uninjured rats served as the sham control (sham group, n=4) and received all surgical procedures and after care apart from the displacement and the muscle injections. No differences in the force (NT3 =256 ± 4.7 kDyn, PBS =255 ± 3.3 kDyn) or displacement (NT3=1417 ± 157 µm, PBS=1428 ± 147 µm) of the spinal cord were observed between the control and treatment groups (Unpaired T test, Force p=0.63, t=0.49; displacement p=0.86, t=0.18). Behavioural assessments (BBB, horizontal ladder, and swimming) week vs 4 weeks p=0.0007, 4 weeks vs 9 weeks p= 0.01, 9 weeks vs 10 weeks p=0.02) and a consistent but non-significant trend for the NT3 group to exhibit higher average functional capacity

NT3 treatment restores coordinated limb movement during swimming
To see if NT3 treatment facilitated motor tasks that do not require weight support, animals were assessed during an 80cm swim in a customised tank kept between 20-23 °C to avoid temperatureevoked spasms ( Figure 3A) . At baseline, animals consistently used the hindlimbs for propulsion in swimming with forepaws kept close to the jaw (Figure 3Bi). Contused animals were able to swim but showed abnormal use of the forelimbs for propulsion with the tail J o u r n a l P r e -p r o o f submerged (Figure 3Bii). These abnormalities are consistent with previous studies (Smith et al., 2006, Xu et al., 2015 (Figure 3Biii-iv).
The effects of injury on swimming gait are reflected in the quantification of animals' swimming activity.
The number of hindlimb strokes four weeks after injury was reduced by approximately 50% ( Figure   3C;

Contusion did not cause spasms or hyperreflexia of the propriospinal circuity
Only one animal showed noticeable spasm, clonus or otherwise abnormal activation of the hindlimbs or trunk during swimming (during one single instance; Figure 3E) and only five animals showed presumptive spasm activity once following swimming ( Figure 3F). Further, the animals did not show any spontaneous spasm during meticulous observation during behavioural tests, when in the open field, and in their home cage. The rareness of spasms was unexpected as this is activity associated with this injury (Ryu et al., 2021).
To investigate hyperreflexia, the H reflex of the animals was recorded from the hindpaw at 11 weeks post injury during paired stimulation of the medial plantar nerve over a defined range of frequencies ( Figure 4A-B) , Lee-Kubli and Calcutt, 2014, Chang et al., 2019. The amplitude of the normalised H wave from the conditional stimulus remained comparable to the H wave from the test stimulus in all three groups (NT3=103.1 ±5%, PBS 102.2 ±2.6%, sham 101.6 ±1.3%; Figure 4C

NT3 treatment did not affect lesion volume
Ex vivo Magnetic Resonance Imaging (MRI) was used to acquire high resolution T2 weighted (T2W) images of the thoracic spinal cord ( Figure 5B-I) following termination of the experiment to assess any differences in lesion volume between groups. At this resolution, there was a clear distinction between the grey and white matter from T7 to T11 on both sham ( Figure 5B-C&F) and contused cords ( Figure   5D-E&G). The epicentre was determined as the spinal cord transverse section with the smallest cross-sectional area of presumptive spared white matter. Within the epicentre, there was a mixture of fragmented hypointense signal, likely corresponding to fluid filled cavitation, interspersed with scar tissue and/or oedema. Sagittal sections showed that some injured spinal cords were deformed at the contused region ( Figure 5E&G).

Journal Pre-proof
Total spinal cord transverse cross-sectional area and presumptive spared white matter were measured following thresholding for the respective tissue and automatic detection in ITK-SNAP software. Presumptive spared white matter was defined as regions with a T2 weighted signal comparable to, and continuous with, the signal in the lateral funiculus in the rostral and caudal most spinal level imaged ( Figure 5H-I). Injury reduced the total transverse cross-sectional area in the regions extending at least 2mm rostral and caudal from the lesion epicentre ( Figure 5J hoc LSD NT3 vs PBS p=0.94). These data suggest that NT3 treatment did not reduce lesion volumes.

NT3 normalised vGlut1+ excitatory input onto motor neurons
To determine treatment induced proprioceptive sprouting, immunohistochemistry was used to assess direct glutamatergic (vGlut1+) connections from Ia afferent fibres on α-motor neurons. Following all assessments, motor neurons innervating the lateral head of the gastrocnemius were retrogradely traced with Cholera Toxin B (CtB, Figures 1A&6A). The L4-L6 spinal cord region, encompassing the gastrocnemius motor pool (Mohan et al., 2014), was stained for vGlut1, a marker of Ia afferent derived motor neuron input and CTb to identify motor neurons innervating the lateral head of the gastrocnemius ( Figure 6B). NeuN was used as an additional marker to exclude connections onto homonymous γ-motor neurons (Friese et al., 2009). vGlut1+ immunoreactive puncta seen on the motor neuron perimeter were quantified: 1) as individual counts and 2) as integrated pixel intensity. A near doubling of vGlut1+ puncta close to motor neuron somas was observed following contusion injury (average puncta in Sham=2.6±0.17, PBS=4.9±0.45 puncta; Figure 6D). However, NT3 treatment was able to reduce this increased marker of excitatory input back to a level comparable with Sham animals (NT3 group: 3.4±0.23 puncta; one way ANOVA,

Discussion
The aim of this study was to determine whether NT3, delivered with a clinically feasible delay to the muscles of the ankle, tibialis anterior and gastrocnemius, and lower portion of the rectus abdominus following a T9 contusion, would modulate the lumbar neural circuitry facilitating functional recovery.
We demonstrated that treated animals had a 75-fold increase in circulating NT3 levels at 11 weeks post injury. Animals did not demonstrate gross improvement in behavioural function but showed, through numerous motor tasks, that the treatment strategy enhanced skilled hindlimb stepping, coordination of these limbs when swimming and trunk stability in the open field. We suggest that this increase in function is due to modulated proprioceptive input innervating these muscles. Interestingly, the injury did not result in measurable signs of spasticity, such as hindlimb spasms and hyperreflexia, suggesting that these typical SCI outcomes are not reliably induced through severe thoracic contusion. This work demonstrates that a clinically feasible 48-hour delay in AAV1-NT3 treatment application can help restore fine functional motor control and supports its use as a therapeutic strategy for SCI.
We used an AAV1-mediated gene therapy to deliver NT3 via injection into peripheral muscle. This is an effective and safe delivery system used previously in human patients though the Glybera therapy and numerous preclinical studies (Sydney-Smith et al., 2021b, Kakanos and Moon, 2019, Wang et al., 2018, Zhou et al., 2003, Chen et al., 2006, Fortun et al., 2009, Petruska et al., 2010. NT3 is safe and well-tolerated in human adults when given peripherally at high doses (Sahenk et al., 2005, Parkman et al., 2003, Coulie et al., 2000 and an AAV-tMCK-NT3 treatment applied through injection into the leg muscle has recently been approved for use in a phase 1 clinical trial (NCT03520751). These data would suggest that the AAV1-CMV-NT3 vector we used would be safe for use in clinical patients.
However, future work will introduce an off switch in the AAV vector to inhibition long-term transgene expression in case of safety concerns (Nguyen et al., 2021, Zhong et al., 2020. Here, we confirmed the AAV serotype 1 produced high quantities of circulating NT3 long after initial injection. We have previously shown that elevated circulating blood serum levels of NT3 protein following AAV1-NT3 injection into muscle occurs concurrently with elevated protein levels of the neurotrophin specifically in the treated muscle, innovating dorsal root ganglia (DRG), and spinal cord dorsal horn and motor neurons , Duricki et al., 2016a. Our work, and that of others, has shown that following intramuscular or intraneural injection of AAV1-CMV-NT3, retrograde transport of NT3 protein J o u r n a l P r e -p r o o f from muscle to DRG, large sensory and motor neurons can occur (Curtis et al., 1998, DiStefano et al., 1992, Wang et al., 2018.

Ex vivo imaging of the spinal cord showed an extensive injury was produced
Ex vivo MRI revealed an extensive lesion. Several comparative studies have shown quantification of injury size from MRI closely matches that from conventional histology (Ditor et al., 2008, Byrnes et al., 2010, Mihai et al., 2008. We report sparing of the lateral and ventral white matter likely encompassing the reticulospinal tract (Schucht et al., 2002), propriospinal neurons (Flynn, Conn et al. 2017Sheikh, Keefe et al. 2018) and possibly the rubrospinal tract (Morris and Whishaw, 2016). The corticospinal tract, providing fine motor control input, is predominantly located in the dorsal columns, which have considerable white matter loss. Tracing experiments after a similar severity of contusion to the cervical spinal cord disrupts all but a few CST fibres in the dorsal lateral white matter (Anderson et al., 2009), and given the smaller diameter of the thoracic cord impacted in this study, it is expected that even fewer CST fibres remain caudal to injury in our animals.

NT3 treatment promoted recovery in skilled control of the hindlimb
Our results showed a beneficial effect of NT3 treatment on trunk stability, swimming coordination, and stepping accuracy. The contusion produced a severe motor deficit which, following endogenous recovery, the animals were unable to regain consistent plantar stepping and interlimb coordination . This is consistent with previous studies following moderate thoracic contusion from a weight drop mechanism (Bose et al., 2012, Basso et al., 1995. The locomotor impairments we report were, unexpectedly, not as severe as studies using a 250 kDyn contusion at T7 (Park et al., 2016), T8 (Ryu et al., 2017) or T9 spinal levels (Chang et al., 2019), where injured animals had limited hindlimb movement early in recovery and plateaued from week four, being unable to perform weight-supportive plantar stepping (e.g., at BBB=9 in Chang et al 2019 versus BBB=10 in ours). These subtle differences between our study and theirs may suggest the contusion performed here did not induce as severe a neurological deficit as expected despite similar force of contusion. Similarly, the functional improvements we saw following NT3 treatment were not as extensive as others have demonstrated despite the same dose of AAV-NT3 being used for hindlimb muscles (Chang et al., 2019) and considering that we additionally treated rectus abdominis muscles. This may suggest that delivery of NT3 at 24 hours is superior to delivery at 48 hours. This should be investigated in further experiments.
Data regarding toe clearance and trunk stability was of particular significance to the study as NT3 was applied to the main ankle flexor and extensor muscles and the rectus abdominus muscles. NT3 treated animals consistently scored higher averages above PBS treated rats for toe clearance and we identified an effect of NT3 treatment on trunk stability during open field walking. Numerous muscles are involved in trunk stability including the rectus abdominus, obliques and muscles of the vertebral J o u r n a l P r e -p r o o f column such as erector spinae (Musienko et al., 2014). Here only the caudal half of the rectus abdominus was injected with AAV-NT3 (although NT3 circulating in the blood might have affected non-injected neuromuscular circuits). Improved trunk stability has been achieved after thoracic contusion experiments following stair training (Singh et al., 2011). These data may suggest that NT3 treatment can mimic some of the effects of regular training in specific circumstances. Further experimentation where all muscles involved in trunk stability are injected with NT3 should be considered.
The effect of NT3 improving precision hindlimb stepping on the horizontal ladder is consistent with other work (Duricki et al., 2016a, Fortun et al., 2009. This behavioural effect occurred at two weeks following treatment administration, indicative of the rapid expression (Zincarelli et al., 2008, Dang et al., 2017, Petruska et al., 2010 and effect of the transgene. We demonstrate the ability of the NT3 to aid the animals' ability to correct misplaced steps, while other skilled functions (e.g. accurate initial placement) is not improved. This may suggest restoration of function through proprioceptive afferents which relay information regarding inappropriate positioning during weight baring steps. Administration of a viral vector encoding NT3 into the sciatic nerve partially recovered error free hindlimb stepping following thoracic contusion , likely a result of NT3 mediated regrowth of dendritic arbors of caudal motor neurons. This was accompanied by elevated numbers of synaptic-like contacts from long descending propriospinal neuron terminals onto motor neuron soma and dendrites. It is possible the recovery we see in hindlimb performance during multiple behaviours is due to a similar enhancement of spinal connectivity since retrograde transport of NT3 occurs from peripheral tissues to motor neurons (DiStefano et al., 1992. NT3 mediating connectivity to these circuits may explain the improved coordination of hindlimb strokes and postural corrections, as contralaterally projecting propriospinal interneurons or commissural V0 interneurons may be affected (Talpalar et al., 2013). However, NT3 treatment is known to mediate the function of sensory neurons (Oakley et al., 1997). As such dissection of the role NT3 plays in functional recovery of neural pathways should be further elucidated.
Several transcriptional changes in the DRG occur because of intramuscular delivery of AAV1-CMV-NT3 in rats with pyramidotomy (Kathe and Moon, 2018). These include upregulation of Sema4c, which belongs to a family of axon guidance molecules and cytoskeletal remodelling associated genes like Gas7. Such transcriptional changes could stabilise afferent contacts on motor neurons, strengthen proprioceptive circuitry and be responsible for reported improved skilled motor function . NT3 treatment reduced afferent input onto gastrocnemius α-motor neurons in our study, suggesting recovery of precision ladder stepping may require, in part, normalisation of vGlut1+ input. Our previous work has shown that despite elevated circulatory NT3 following treatment, spinal reflexes, and the number of vGlut1+ boutons only normalised on circuits where the afferent nerve extended into the treated muscle . As such, we would expect the normalisation of vGlut1+ input in this study to be specific to afferent connections of the injected J o u r n a l P r e -p r o o f muscles. Of note is the increase we show in corrective steps. This may be explained by more effective proprioceptive feedback, enabling detection of a slip early and repositioning of the paw before transfer of weight support and slippage. As slipping would involve activation of Ia afferents innervating flexors and extensors of the ankle joints, it is possible that proprioceptive circuitry involving the ankle flexors and extensors was abnormal after injury. However, note that in our study the H-reflex was recorded instead from an intrinsic hindpaw muscle.

T9 contusion model did not reliably induce spasms or hyperreflexia
Our study reveals that hindlimb spasms are not reliably generated by a 250kDyn contusion at the T9 spinal level in female Wistar rats, contrasting with previous reports using the same severe contusion injury model (Ryu et al., 2017, Chang et al., 2019 or a T-lesion at the same level . Others have demonstrated plantar hindlimb muscle hyperreflexia and visible hindlimb spasms during swimming, walking and in their home cage using female Sprague Dawley rats using this severe contusion injury model (Ryu et al., 2017, Ryu et al., 2021. Reasons for the differences could be slight alterations in equipment, methodologies employed, or strains used (Mills et al., 2001). Our injuries were not as functionally extensive as those previously reported with this injury (Chang et al., 2019), suggesting the CST was not as comprehensively injured, which may account for the lack of visible spasms . Alternatively, spared reticulospinal function is positively correlated with the presence of hindlimb spasticity in SCI patients (Sangari and Perez, 2019), suggesting we may have excessively damaged this pathway. It is possible that altering the methods employed may have yielded different data. For example, EMG recordings from affected muscles may detect mild spasms, or a train of stimuli may fully depress the H reflex  compared to paired pulses (Chang et al., 2019, Ryu et al., 2017. However, as the methods employed have been successful in several prior studies, this study highlights that production of visually identifiable spasms and hyperreflexia of the hindpaw is not as a reliable outcome of severe T9 contusive injuries as previously reported.

Conclusion
We show that targeted, delayed delivery of NT3 to hindlimb and trunk muscles 48h following a severe contusive injury to the thoracic spinal cord improved trunk stability, accuracy of stepping during skilled locomotion and alternation of the hindlimbs during swimming but had no effect on gross locomotor function in the open field. We suggest that this increase was through modulated proprioceptive feedback, as vGlut1+ boutons on presumptive proprioceptive afferents innervating treated muscles were normalised. Further, here we show that a T9 250kDyn contusion does not reliably result in measurable signs of spasticity, such as hindlimb spasms and hyperreflexia. This is the first demonstration that a clinically feasible 48-hour delay in AAV1-NT3 treatment application can help J o u r n a l P r e -p r o o f restore precise motor control in functionally compromised hindlimbs, supporting its use as a therapeutic strategy for SCI and continued clinical development.