Non-viral-mediated suppression of AMIGO3 promotes disinhibited NT3-mediated regeneration of spinal cord dorsal column axons

After injury to the mature central nervous system (CNS), myelin-derived inhibitory ligands bind to the Nogo-66 tripartite receptor complex expressed on axonal growth cones, comprised of LINGO-1 and p75NTR/TROY and induce growth cone collapse through the RhoA pathway. We have also shown that amphoterin-induced gene and open reading frame-3 (AMIGO3) substitutes for LINGO-1 and can signal axon growth cone collapse. Here, we investigated the regeneration of dorsal root ganglion neuron (DRGN) axons/neurites after treatment with a short hairpin RNA (sh) AMIGO3 plasmid delivered with a non-viral in vivo-jetPEI vector, and the pro-survival/axogenic neurotrophin (NT) 3 in vitro and in vivo. A bicistronic plasmid, containing both shAMIGO3 and NT3 knocked down >75% of AMIGO3 mRNA in cultured DRGN and significantly overexpressed NT3 production. In vivo, intra-DRG injection of in vivo-jetPEI plasmids containing shAMIGO3/gfp and shAMIGO3/nt3 both knocked down AMIGO3 expression in DRGN and, in combination with NT3 overexpression, promoted DC axon regeneration, recovery of conduction of compound action potentials across the lesion site and improvements in sensory and locomotor function. These findings demonstrate that in vivo-jetPEI is a potential non-viral, translatable DRGN delivery vehicle in vivo and that suppression of AMIGO3 disinhibits the growth of axotomised DRGN enabling NT3 to stimulate the regeneration of their DC axons and enhances functional recovery.

transfection is limited in acute conditions. Non-viral gene delivery vectors include cationic lipid agents and a more recently formulated non-lipid polymer, polyethylenimine (in vivo-jetPEI), which transfects cells both in vitro and in vivo 15,16 . In vivo-jetPEI has high transfection efficiency and yields up to 4× higher transfection rates than naked DNA 15 . Moreover, in vivo-jetPEI is easy to prepare, stable and safe, and in vivo-jetPEI/shRNA transfection induces faster suppression of mRNA translation than viral vectors, giving immediate therapeutic benefit 17 . For example, in the CNS, Melanopsin expression is rapidly abolished in RGC after intravitreal injection of in vivo-jetPEI-shMelanopsin 18 .
In this study, we show that in vivo-jetPEI successfully delivers a bicistronic plasmid DNA to DRGN encoding shAMIGO3 to suppress AMIGO3 and perturb axon growth inhibitory signalling while, at the same time inducing the overexpression of neurotrophin-3 (NT3). We show that suppression of AMIGO3 and concomitant stimulation of DRGN with NT3 enhanced disinhibited DRGN neurite/axon regeneration and improved functional recovery. These results suggest that in vivo-jetPEI PEI delivers genes to DRGN without the disadvantages of viral vectors, rendering in vivo-jetPEI PEI as translationally relevant.

Results
In vitro experiments. In vivo-jetPEI delivered plasmids knock down AMIGO3 mRNA. In untreated, sham and non-specific shgfp control transfected DRGN, there was no change in mRNA for AMIGO3 suggesting that none of these treatments had any non-specific effects on amigo3 mRNA (Fig. 1A). Treatment with increasing amounts of shAMIGO3 plasmid delivered by in vivo-jetPEI (referred to as PEI in the results section from herein) caused a dose-dependent decrease in amigo3 mRNA to a minimum at 2 µg of plasmid DNA, correlating with 80% knockdown compared to untreated, sham or shgpf controls (Fig. 1A). Increasing the amount of plasmid DNA above 2 µg did not decrease amigo3 mRNA levels further, confirming that 2 µg of plasmid DNA gave optimal knockdown.

PEI-delivered plasmids did not invoke off-target effects after intra-DRG injection.
Intra-DRG injection of any of the plasmids that were delivered by PEI did not invoke changes in the CD68 and GFAP immunoreactivity in DRG (Fig. 6A). In contrast, intra-DRG injection of T7-sip75 NTR , a chemically synthesized siRNA that induces Low levels of nt3 mRNA were detected in IC, DC and DC + PEI-gfp groups, whilst significantly higher levels (P < 0.0001) were detected in DC + PEI-nt3/gfp and DC + PEI-shAMIGO3/nt3 groups. (B) Western blot and subsequent densitometry for NT3 protein reflected the changes in mRNA. β-actin was used as a loading control (full scans shown). (C) Immunohistochemistry for NT3 showed low levels of NT3 (red) in DRGN after DC. (C(i)), which were gfp − (C(ii)) and in DC + PEI-gfp groups (C(iv)), with gfp expression (green) in some DRGN (C(v)). NT3 levels (Cvii)) were high in both gfp + and gfp − DRGN (C(viii)). Scale bar = 100 μm. C(iii), C(vi) and C(ix) are merged images showing NT3 (red), gfp (green) and DAPI stained nuclei (blue). ***P < 0.0001, ANOVA.
The mean sensing time for the tape removal test which assess sensory function 20 remained between 12-25 s throughout the 6-week experimental time period in sham treated animals (Fig. 7D). The mean sensing time increased to 80-85 s in DC + PEI-nt3 and DC + PEI-shAMIGO3/gfp-treated groups at 2d after DC lesion (Fig. 7D). However, in DC + PEI-shAMIGO3/nt3-treated animals, the mean sensing time at 2d was significantly Figure 6. Response to PEI-delivered plasmids. (A) CD68 and GFAP immunoreactivity in intact and after PEIdelivered plasmids in section of DRG to show a lack of induction of immunological responses to these markers. In contrast, T7-sip75 NTR invoked significant changes in CD68 and GFAP + immunoreactivity in DRG. (B) Lack of activation of off-target cytokines and innate immunity related genes in DRG after PEI-delivered plasmid transduction in DRG. Once again, T7-sip75 NTR showed significant upregulation of mRNA for all of these genes. (C) ED1 immunoreactivity was low in DC + PEI-nt3/gfp-treated animals whilst abundant ED1 + cells were present within the lesion site (*) in DC + PEI-shAMIGO3/nt3-treated animals. (D) GFAP + astrocytes were restricted to the lesion edge in DC + PEI-nt3/gfp-treated animals whilst GFAP + astrocytes invaded the lesion site in DC + PEI-shAMIGO3/nt3-treated animals. Scale bars in A = 100 µm; C and D = 500 µm.
Scientific REPORTs | (2018) 8:10707 | DOI:10.1038/s41598-018-29124-z lower (P < 0.001), taking 53 ± 8 s to detect the tape (Fig. 7D). The mean sensing time improved over the 6-week time period in all animals, however, in DC + PEI-shAMIGO3/nt3, the mean sensing time returned to sham treated baseline levels by 4 weeks after DC lesion. Over the 6-week time period, DC + PEI-shAMIGO3/nt3-treated animals were significantly better at detecting the tape and removing it compared to all other treatment groups (P < 0.0012, linear mixed model). The mean ratio (slips/total steps) in the horizontal ladder crossing test, which detects locomotor function 20 remained between 0.1 and 0 throughout the 6-week time period in sham treated animals (Fig. 7E). The mean ratio increased to 0.58 ± 0.05 in DC + PEI-in DC + PEI-nt3 and DC + PEI-shAMIGO3/gfp-treated groups at 2d after DC lesion (Fig. 7E). However, in DC + PEI-shAMIGO3/nt3-treated animals, the mean ratio at 2d was significantly lower (P < 0.001), at 0.33 ± 0.02 (Fig. 7D). The mean ratio improved over the 6-week time period in all animals, however, in DC + PEI-shAMIGO3/nt3, the mean ratio returned to sham treated baseline levels by 4 weeks after DC lesion. Over the 6-week time period, DC + PEI-shAMIGO3/nt3-treated animals were significantly better at ladder crossing compared to all other treatment groups (P < 0.01, generalized linear mixed models).
Taken together, these results show that suppression of AMIGO3 and simultaneous NT3 stimulation of DC axon regeneration promotes sensory and locomotor function recovery.

Discussion
In this study, we investigated the use of a non-viral vector for gene delivery to DRGN and showed in vitro and that shAMIGO3 knockdown combined with NT3 over-expression, induced DRGN to grow neurites on a CME substrate. We also showed in vivo that in vivo-jetPEI-delivered shAMIGO3/gfp plasmid DNA: (1), increased transduction in all sizes of DRGN; (2), efficiently knocked down AMIGO3 while simultaneously increasing NT3 production; (3), stimulated DC axon regeneration which conducted action potentials across the lesion site and led to enhanced sensory and locomotor function.
In vitro findings. We tested the efficiency of in vivo-jetPEI-delivered plasmids to suppress AMIGO3 and upregulate NT3 in DRGN cultures. It has been recognised by us and others that a combinatorial strategy will be required to promote optimal CNS axon regeneration due to the complexity of the injury. For example, we have shown that suppression of axon growth inhibitory molecules combined with neurotrophic factor stimulation promotes significant disinhibited DRGN neurite outgrowth in the presence of CME [21][22][23][24] . In this study, 2 µg of plasmid DNA encoding shAMIGO3 was sufficient to significantly knock down AMIGO3, to similar levels that we observed with siAMIGO3 8 . The same amount of DNA was also able to promote over-expression of NT3 in DRGN and significantly increased titres of NT3 secretion into the culture medium.
However, similar to our previous observations with siAMIGO3, knockdown of AMIGO3 using an shAMIGO3 alone did not overcome CME-mediated neurite growth inhibition. Added simultaneous stimulation with NT3 to promote DRGN neurite outgrowth was required, confirming that disinhibition alone in insufficient to promote DRGN neurite outgrowth in the presence of CME but that concomitant stimulation of growth is required to drive regeneration 22,25,26 . In vivo findings. We compared the targeting efficiency of non-viral plasmids delivered to DRGN through intra-DRG injection with in vivo-jetPEI. There is a high heterogeneity of DRGN in size distribution 27,28 and our plasmids, especially in the presence of shAMIGO3, increased transduction in all sizes of DRGN. In a previous study, we found that preferential targeting of large diameter DRGN was achieved with AAV8 delivered gfp 28 . It is possible that the large surface area exposed to the plasmids results in an increased incidence of transduction 28 . It is well established that the majority of the DRGN that project their axons through the DC are large diameter DRGN and hence their increased transduction rates by our plasmids is actually beneficial, after DC injury. It is clear from our study that NT3 production by DRGN through the local production of NT3 may also have paracrine effects on neighbouring DRGN, since both GFP + (i.e. transduced DRGN) and GFP − DRGN (i.e. non-transduced DRGN) were NT3 + .
A common concern with RNA interference (RNAi) is RNAi-associated immune stimulation, which is a major hurdle to the safety and efficacy of RNAi. The induction of innate immunity may arise from sequence-dependent effects, delivery vehicles and the RNAi process itself leading to undesired effects and misinterpretation of the data 29,30 . However, our study found no evidence of changes in the most common off-target cytokines and innate immunity genes after transduction of DRGN with in vivo-jetPEI-delivered plasmids. Others have also reported that in vivo-jetPEI vector is safe and does not induce off-target immune-mediated cytokines such as TNF-α, IFN-γ, IL-6, IL-12/IL-23, IL-1β, and IFN-β 31 . Our study therefore suggests that in vivo-jetPEI may be safe for clinical applications.
Not only did the in vivo-jetPEI-delivered shAMIGO3 suppress AMIGO3 levels in >80% DRGN, immunoreactivity for AMIGO3 was reduced in both gfp + and gfp − DRGN. This suggests that the plasmid efficiently suppressed AMIGO3 levels in both transduced and non-transduced DRGN. At present, we cannot explain why non-transduced DRGN showed reduced AMIGO3 levels but, to our knowledge, this is the first report to show such an effect. It may suggest that suppression of AMIGO3 in DRGN had paracrine effects on neighbouring DRGN, representing an intriguing possibility that not all DRGN require targeting to promote therapeutically effective outcomes. Similarly, over-expression of NT3 was also apparent in both gfp + and gfp − DRGN suggesting that both transfected and non-transfected DRGN produced high titres of NT3. Our results therefore demonstrate that in vivo-jetPEI efficiently knocked down AMIGO3 and up-regulated NT3 in DRGN.
Knockdown of AMIGO3 and simultaneous stimulation with NT3 promoted significant DRGN axon regeneration after DC injury. This was correlated with significant improvements in CAP amplitudes across the lesion site and sensory and locomotor function. This is the first demonstration that knockdown of AMIGO3 promotes spinal cord axon regeneration in vivo and demonstrates that AMIGO3 may require targeting to promote CNS axon regeneration. We previously showed that AMIGO3 could substitute for LINGO-1 in binding to p75 NTR and NgR1 to induce RhoA activation in response to CME. Our current results are consistent with these findings. Axons regenerated for up to 6 mm rostral to the lesion site and conducted action potentials across the lesion site correlating with improved sensory and locomotor function. Other types of behavioural analyses however, Scientific REPORTs | (2018) 8:10707 | DOI:10.1038/s41598-018-29124-z have been shown to be inconclusive with the T8 DC lesion model, since measurable functional improvements in mechanical and thermal hyperplasia do not occur 32,33 .
Our current study demonstrates in vivo proof-of-principle that in DRGN cultures grown in the presence of CNS myelin extracts, knockdown of AMIGO3 promoted signifcant DRGN neurite outgrowth 8 and that knockdown of AMIGO3 together with co-incident growth stimulation (here by NT3) promoted signficant DC axon regeneration and improved functional recovery. In addition, spinal cord cavitation was reduced and the lesion site was filled with invading cells, suggesting a further therapeutic use of AMIGO3 and NT-3 in spinal cord injury. It remains to be elucidated why knockdown of AMIGO3 suppressed cavitation that normally results after DC injury in rats 33 . One possibility is that since AMIGO3 belongs to a family of adhesion molecules, knocking down AMIGO3 may have limited the influx of inflammatory cells after DC injury and hence reduced inflammation 34 limiting secondary damage and cavitation 35,36 .
Although ED1 + cells were upregulated in the lesion site of shAMIGO3/nt3-treated animals, we did not identify whether these cells were of the M1 (classical activation) or M2 (alternatively activated macrophages) lineage 37 . However, pro-regenerative M2 macrophage lineages are only present for the first week whilst M1 macrophages persist for more than 4 weeks. Therefore, it is unlikely that ED1 + cells were pro-regenerative M2 phenyotypes. However, we are unsure why there are more ED1 + cells after treatment with shAMIGO3/nt3 and are currently investigating why this is so. Astrocytes on the other hand play multifaceted roles after spinal cord injury. Normally astrocytes are the major component of the glial scar and prevent axon regeneration by forming a physical and chemical barrier at the lesion borders. However, it appears that shAMIGO3/nt3 treatment promoted astrocyte invasion into the lesion site, transforming the lesion border from a dense network of tightly woven and thickened processes to a more open meshwork, allowing astrocytes to invade the lesion site. It is possible that astrocyte invasion allowed greater axon regeneration in these animals since they are able to restore the extracellular ionic environment, sequester extracellular glutamate and produce neurotrophic factors [38][39][40][41] . Astrocytes secrete axon growth inhibitory CSPG but also adhesive extracellular molecules such as laminin that are supportive to axon growth [42][43][44][45] . TGFα infusion into the spinal cord lesion site in mouse also allowed similar astrocyte invasion and new cells as was observed by us with shAMIGO3/nt3 treatment. For example, the centre of the lesion site contained greater number of cells and increased astrocyte invasion, which correlated with increased numbers of GAP43 + axons 46 . These results suggest that astrocytes may provide axon growth supportive roles by migrating into the lesion site.

Conclusion
Our results demonstrated that non-viral plasmid delivery enhanced transduction in all sizes of DRGN, significantly downregulated AMIGO3 and upregulated NT-3 production from DRGN. This reduced cavity formation and promoted axon regeneration and functional recovery in the damaged spinal cord. These results suggest that AMIGO3 might comprise a future therapy to promote axon regeneration and functional recovery after spinal cord injury. Plasmid transfection is likely to be more translational since viruses need at least 2 weeks to start producing effective titres of the shRNA of choice, whereas plasmids begin to express the relevant payload within hours. Moreover, in vivo-jetPEI vector is safe and does not induce off-target immune-mediated cytokines. Therefore, the lack of immunogenic responses after in vivo-jetPEI injection is to be expected and is a further advantage over viral vectors.

Methods
Preparation of CNS myelin extracts. CNS myelin extracts (CME), containing significant titres of Nogo-A, OMgp, MAG and CSPG were prepared according to our previously published methods 21 .
To monitor NT3 production and disinhibited DRGN neurite outgrowth after shAMIGO3 knockdown and NT3 stimulation, DRGN were transfected with the optimal dose of plasmid DNA (2 μg) and incubated for 3d in the presence of 100 μg/ml CME 21 , before harvesting culture supernatant to detect NT3 production by ELISA and neurite outgrowth by immunocytochemistry.
Scientific REPORTs | (2018) 8:10707 | DOI:10.1038/s41598-018-29124-z Quantitative real-time PCR to confirm shAMIGO3 knockdown: Total RNA from treated DRGN (n = 6 wells/treatment) was prepared using Trizol according to the manufacturer's instructions (Invitrogen). AMIGO3 knockdown and nt3 over-expression were validated, using qRT-PCR from complementary DNA prepared from extracted mRNA, with appropriate forward and reverse primers (Table 1) using the LightCycler real time qRT-PCR system (Roche, Burgess Hill, UK) according to our previously published methods 19 . Fold-changes were computed using the ΔΔCt method.
NT-3 ELISA: NT-3 was assayed in supernatant from n = 6 wells/treatment using the NT-3 DuoSet ELISA kit according to the manufacturer's instructions (R&D Systems, Abingdon, UK).
Immunocytochemistry: Cells were fixed in 4% paraformaldehyde, washed in 3 changes of PBS before immunocytochemistry as described by us previously 21 . To visualise neurites, DRGN were stained with monoclonal anti-βIII tubulin (Sigma) detected with Alexa-488 anti-mouse antibodies (Invitrogen). Slides were then viewed with an epi-fluorescent Axioplan 2 microscope, equipped with an AxioCam HRc and running Axiovision Software (all from Carl Zeiss, Hertfordshire, UK). The proportion of DRGN with neurites and mean neurite lengths were calculated using Axiovision Software, as previously described 21 . All experiments were performed in duplicate and repeated on 3 independent occasions (i.e. n = 6 wells/condition).
In vivo experiments. Surgical procedures: All animal surgeries were carried out in strict accordance to the guidelines of the UK Animals Scientific Procedures Act, 1986 and the Revised European Directive 1010/63/EU and conformed to the guidelines and recommendation of the use of animals by the Federation of the European Laboratory Animal Science Associations. Experiments were licensed by the UK Home Office and all experimental protocols were approved by the University of Birmingham's Animal Welfare and Ethical Review Board. Adult female Sprague-Dawley rats weighing 170-220 g (Charles River, Margate, UK) were used in all experiments. The experiments comprised 8 groups (n = 12 rats/group), including 4 control groups comprising: Intact controls (Group 1: IC; to detect baseline levels), partial laminectomy but no DC lesion; Group 2: sham; to detect surgery-induced changes), DC transected controls (Group 3: DC; to detect injury-mediated changes), and DC + intra-DRG injection of in vivo-jetPEI-gfp (Group 4: DC + PEI-gfp; to monitor DRGN transduction efficiencies) and 4 experimental groups in which DC were transected and DRG injected with plasmid-containing in vivo-jetPEI complexes, comprised of: DC + intra-DRG injection of in vivo-jetPEI-gfp (Group 1: DC + PEI-gfp); DC + intra-DRG injection of in vivo-jetPEI-nt3/gfp (Group 2: DC + PEI-nt3/gfp; to monitor DRGN transduction and NT3 over-expression); DC + in vivo-jetPEI-shAMIGO3/gfp (Group 3: PEI-shAMIGO3/gfp; to monitor AMIGO3 knockdown in DRGN and to establish if AMIGO3 knockdown without concomitant stimulation with NT3 promoted DRGN axon regeneration); and DC + intra-DRG injection of in vivo-jetPEI-shAMIGO3/nt3 (Group 4: DC + PEI-shAMIGO3/nt3; to determine if AMIGO3 knockdown and simultaneous stimulation with NT3 was required for DRGN axon regeneration).
Before surgery, rats were injected subcutaneously with 0.05 ml Buprenorphine and anaesthetised using 5% of Isoflurane in 1.8 ml/l of O 2 with body temperature and heart rate monitored throughout surgery. After partial T8 laminectomy, DC were crushed bilaterally using calibrated watchmaker's forceps and, after injection of in vivo jet-PEI/plasmids into L4/L5 DRG 28,33 , rats were killed at either 28d (western blot, immunohistochemistry, qRT-PCR and ELISA) or 6 weeks (electrophysiology and functional tests) in rising concentrations of CO 2 .
Analysis of off-target cytokines and innate immune response genes: Total RNA from n = DRG/treatment were extracted using Trizol reagent as described above. A selection off-target cytokines and innate immune response genes (Table 1) were analysed by qRT-PCR from complementary DNA prepared from extracted mRNA as described above 19 . Fold-changes were computed using the ΔΔCt method.
Tissue preparation and cryo-sectioning: Rats were intracardially perfused with 4% formaldehyde and DRG and segments of T8 cord containing the DC injury sites were post-fixed for 2 h at RT, prepared for cryoprotection and sectioned using a cryostat as described by us previously 33 .  Table 1. Primers used in this study for qRT-PCR.
Protein extraction, western blot and densitometry: Total protein from DRG was extracted and western blot followed by densitometry was performed according to our previously published methods 49,50 . Briefly, 40 µg of total protein extract was resolved on 12% SDS gels, transferred to polyvinylidene fluoride (PVDF) membranes (Millipore, Watford, UK) and probed with relevant primary antibodies: anti-AMIGO3 (1:400 dilution, Santa Cruz); anti-NT-3 (1:1000 dilution, Abcam) and β-actin (1:1000 dilution, Sigma). Membranes were then incubated with relevant HRP-labelled secondary antibodies and bands were detected using the enhanced chemiluminescence kit (GE Healthcare, Buckinghamshire, UK). Anti β-actin antibodies were used as a protein loading control.
For densitometry, western blots were scanned into Adobe Photoshop (Adobe Systems Inc, San Jose, CA, USA) keeping all scanning parameters the same between blots and the integrated density of bands analysed using the built-in-macros for gel analysis in ImageJ (NIH, USA, http://imagej.nih.gov/ij) 21,[50][51][52] . Means ± SEM were plotted in Microsoft Excel (Microsoft Corporation, CA, USA).
Analysis of GFP expression: Images of sections (n = 160) of L4/L5 DRG from each animal (n = 10) were captured at x10 magnification using the Zeiss Axioplan 200 (Carl Zeiss). Images were merged in Adobe Photoshop (Adobe Systems Incorporated) using Photomerge and the total number of GFP + DRGN and their diameters recorded, as described previously 28 .
Quantification of axons: Axon regeneration in the spinal cord was quantified according to previously published methods 53 . Briefly, serial parasagittal sections of cords were reconstructed by collecting all serial 50 μm-thick sections (∼70-80 sections/animal; n = 10 rats/treatment). In each section, the numbers of intersections of GAP43 + fibers through a dorsoventral orientated line was counted from 4 mm rostral to 4 mm caudal to the lesion site. Axon number was calculated as a percentage of the fibers seen 4 mm above the lesion, where the DC was intact.
Electrophysiology: Four weeks after surgery or treatment, compound action potentials (CAP) were recorded after shAMIGO3/nt3 treatment as previously described 54,55 . The experimenter was blinded to the treatment status of the animals. The amplitude of CAP was calculated between the negative deflection after the stimulus artifact and the next peak of the wave. The area of the CAP was calculated by rectifying the CAP component (full-wave rectification) and measuring its area. At the end of each experiment, the dorsal half of the spinal cord was transected between stimulating and recording electrodes to confirm that a CAP could not be detected.
Functional tests. Functional testing after DC lesions was carried out as described by others 20 . Animals (n = 18/ group) were trained to master traversing the rope and ladder for 1w before functional testing. All functional tests were performed 2-3d before injury to establish baseline parameters. Animals were then tested 2d after DC lesion + treatment and then weekly for 6 weeks. Experiments were performed by 2 observers (from which the treatment conditions were masked) in the same order, the same time of day and each test performed for 3 individual trials.
Horizontal ladder test: This tests the animals locomoter function and is performed on a 0.9-meter-long horizontal ladder with a diameter of 15.5 cm and randomly adjusted rungs with variable gaps of 3.5-5.0 cm. Animals were assessed traversing the ladder and the left and right rear paw slips were recorded along with the total number of steps. To calculate the mean error rate: the number of slips were divided by the total number of steps.
Tape removal test (sensory function): The tape removal test determines touch perception from the left hind paw. Animals were held with both hind-paws extended and a piece of tape of 15 × 15 mm (Kip Hochkrepp, Bocholt, Germany) was affixed to the palm of the left hind-paw. The time it took for the animals to detect and remove the tape was recorded and used to calculate the mean sensing time.
Statistical analysis. Data are presented as means ± SEM. Significant differences were calculated using SPSS Version 22 (IBM, NJ, USA) software by one-way analysis of variance (ANOVA), with Bonferroni post hoc tests, set at P < 0.05.
For the horizontal ladder crossing functional tests, data was analysed according to Fagoe et al. 20 using R package (www.r-project.org). Whole time-course of lesioned and sham-treated animals were compared using binomial generalized linear mixed models (GLMM), with: lesion/sham set to true in lesioned animals post-surgery (set to false otherwise); operated/unoperated set to false before surgery, true after surgery as fixed factors; animals as random factors and time as a continuous covariate. Binomial GLMMs were fitted in R using package lme4 with the glmer function. P values were then calculated using parametric bootstrap. For the tape removal test, linear mixed models (LMM) were calculated by model comparison in R using the package pbkrtest, with the Kenward-Roger method 20 .
Data availability. All data generated or analysed during this study are included in this article.