Timed GDNF gene therapy using an immune-evasive gene switch promotes long distance axon regeneration

Abstract

Neurosurgical repair in patients with proximal nerve lesions results in unsatisfactory recovery of function. Gene therapy for neurotrophic factors is a powerful strategy to promote axon regeneration. Glial cell line-derived neurotrophic factor (GDNF) gene therapy promotes motor neuron survival and axon outgrowth; however, uncontrolled delivery of GDNF results in axon entrapment. We report that time-restricted GDNF expression (1 month) using an immune-evasive doxycycline-inducible gene switch attenuated local axon entrapment in avulsed reimplanted ventral spinal roots, was sufficient to promote long-term motor neuron survival (24 weeks) and facilitated the recovery of compound muscle action potentials by 8 weeks. These improvements were associated with an increase in long-distance regeneration of motor axons. In contrast, persistent GDNF expression impaired axon regeneration by inducing axon entrapment. These findings demonstrate that timed expression can resolve the deleterious effect of uncontrolled growth factor delivery and shows that inducible growth factor gene therapy can be employed to enhance the efficacy of axon regeneration after neurosurgical repair of a proximal nerve lesion in rats. This preclinical study is an important step in the ongoing development of a neurotrophic factor gene therapy for patients with severe proximal nerve lesions.

Introduction

Patients with severe traction lesions of the brachial plexus frequently suffer from concurrent spinal root avulsions. In the majority of patients, an avulsion lesion involves extensive damage to multiple adjacent spinal roots and has therefore been designated a longitudinal spinal cord lesion (Carlstedt and Havton, 2012). Despite advances in nerve repair techniques (Carlstedt et al., 1995; Malessy and Thomeer, 1998), recovery of function is often minimal and typically limited to the proximal musculature.

The two main contributing factors to poor functional recovery following a spinal root avulsion lesion are: (i) the progressive degeneration of injured ventral spinal motor neurons (Koliatsos et al., 1994); and (ii) the decline in the capacity of motor neurons to regenerate their axons towards the denervated muscles. Experimental ventral root reimplantation results in a modest improvement of motor neuron survival and some degree of motor axon regeneration. These beneficial effects are attributed to Schwann cells in the reimplanted ventral root and distal nerve, which convert to a regeneration-promoting phenotype (Jessen and Mirsky, 2016). The proregenerative features of Schwann cells include upregulation of the neurotrophic factors glial cell line-derived neurotrophic factor (GDNF) and brain-derived neurotrophic factor (BDNF), which are important for neuronal survival and axon regeneration (Henderson et al., 1994; Li et al., 1995). However, following lumbar ventral root reimplantation, regenerating motor axons need to bridge distances of many centimetres before reaching their targets. Although the fastest axons entering the reimplanted root initially grow at a velocity of 2.8 mm/day, around 8 weeks post-lesion the speed of axon growth has declined to 0.7 mm/day (Eggers et al., 2010). Staggered axon regeneration into the nerve and a decline in growth speed result in prolonged denervation of the distal nerve and hind limb muscles (Al-Majed et al., 2000; Witzel et al., 2005). In a chronically denervated nerve, Schwann cells lose their capacity to support regeneration and fibrosis creates an inhibitory environment for axon extension (Sulaiman and Gordon, 2009; Jonsson et al., 2013; Ronchi et al., 2017). The endogenous increase in the expression of GDNF and BDNF following an avulsion and reimplantation is transient and declines between 2 and 4 weeks after injury (Eggers et al., 2010), contributing to progressive motor neuron degeneration and the failure of long-distance axon regeneration.

Exogenous GDNF protein delivery (Li et al., 1995; Yuan et al., 2000; Wu et al., 2003; Bergerot et al., 2004; Zhou and Wu, 2006a; Chu et al., 2012; Ruven et al., 2018) consistently shows beneficial effects on motor neuron survival and axon regeneration into the reimplanted nerve (reviewed in Eggers et al., 2016). However, neurotrophic proteins have a short half-life and exhibit poor tissue penetration (Anderson et al., 1995; Dittrich et al., 1996; Hadaczek et al., 2010). Gene therapy ensures local availability of the biologically active neurotrophic factor for the entire regeneration period (Eggers et al., 2008; Tannemaat et al., 2008; Mason et al., 2011; Allodi et al., 2014; Hoyng et al., 2015; Marquardt et al., 2015). However, uncontrolled delivery of GDNF causes side effects, including aberrant sprouting (Hudson et al., 1995; Georgievska et al., 2002; Blesch and Tuszynski, 2003; Love et al., 2005; Su et al., 2009), axon entrapment (Blits et al., 2004; Eggers et al., 2008; Tannemaat et al., 2008; Santosa et al., 2013; Hoyng et al., 2014a; Ee et al., 2017), nerve hypertrophy (Hoke et al., 2003) and weight loss (Manfredsson et al., 2009). This limits the use of GDNF and other neurotrophic factors as therapeutic agents. Efforts to develop inducible viral vectors have shown promising results with a large selection of transgenes, including GDNF, BDNF and NGF (Blesch et al., 2001; Blesch and Tuszynski, 2007; Chtarto et al., 2007; Shakhbazau et al., 2013; Marquardt et al., 2015; Chtarto et al., 2016; Liu et al., 2017). Recent proof of concept studies showed that transplantation of Schwann cells with doxycycline-controlled expression of GDNF has beneficial effects on axon regeneration, whereas uncontrolled GDNF delivery inhibited regeneration (Shakhbazau et al., 2013; Marquardt et al., 2015). This shows that it is of paramount importance to have the ability to control the time window of neurotrophic factor delivery (Kemp et al., 2011; Pajenda et al., 2014; Marquardt et al., 2015; Santos et al., 2016; Tajdaran et al., 2016). The doxycycline-inducible system is a robust platform for in vivo therapeutic gene regulation (Das et al., 2016). However, preclinical studies in rodents (Ginhoux et al., 2004; Markusic et al., 2010) and in non-human primates (Favre et al., 2002; Latta-Mahieu et al., 2002; Le Guiner et al., 2014) demonstrated that long-term regulation of a therapeutic gene was hampered by the loss of transduced cells caused by an immune response to the foreign transactivator. Hence, experimental in vivo use and clinical translation is compromised by immune-mediated cell removal. Here we used an immune-evasive transactivator generated by fusing the original transactivator (rtTA) to the glycine-alanine repeat (GAr) of the Epstein-Barr virus nuclear antigen 1 (Zaldumbide et al., 2010), which is known to protect against T cell-mediated cell removal (Levitskaya et al., 1997; Ossevoort et al., 2003). This enabled the development of a doxycycline-inducible lentiviral vector-based delivery system with immune-evasive properties in a bioassay for human cytotoxic T cell presentation (Ossevoort et al., 2003; Zaldumbide et al., 2010; Hoyng et al., 2014b) and following in vivo application in the injured rat spinal cord (Burnside et al., 2018).

Here, we used a lentiviral immune-evasive doxycycline-inducible GDNF vector system (dox-i-GDNF) to assess the efficacy of time-restricted GDNF expression on motor neuron survival and axon regeneration following ventral root avulsion and reimplantation. Dox-i-GDNF allowed effective regulation of GDNF in the injured reimplanted roots and improved motor neuron survival and long-distance axon regeneration. This preclinical study represents the first use of an immune-evasive gene therapy vector to deliver and regulate a neurotrophic factor in the injured nervous system and is an important step towards the ongoing development of a neurotrophic factor gene therapy as an adjunct treatment for surgical nerve repair.

Materials and methods

Study design

First we assessed the performance of dox-i-GDNF in cultured rat Schwann cells. Subsequently, two in vivo experiments were performed. The aims of the first experiment were (i) to test the ON/OFF characteristics of dox-i-GDNF in reimplanted spinal ventral roots; and (ii) to compare the effects of temporal (4 weeks) and continuous expression (12 weeks) of GDNF on motor neuron survival, axon trapping and axon regeneration. The second experiment builds on the first and was designed to perform a comparison between temporal (4 weeks) and long-term (24 weeks) GDNF expression using electrophysiological and histological analysis.

Production of lentiviral vectors and quantification of GDNF expression in vitro and in vivo

Lentiviral (LV) vectors were generated using standard procedures (Supplementary material). Vectors used here are dox-i-GDNF (1:1 mix of LV-TRE-GDNF and LV-CMV-GArrtTA), LV-CMV-GDNF and LV-CMV-GArGFP (Fig. 1A). Quantification of GDNF protein expression was performed in vitro in primary Schwann cells by performing a titre-matched comparison (Fig. 1B–D) and in avulsed and reimplanted ventral roots (Fig. 3) using an ELISA (Supplementary material).

Surgical procedures

Young adult female Wistar rats (180–200 g Charles River) were housed under standard conditions at a 12:12 h light/dark cycle with ad libitum access to water and chow. All experimental procedures were performed in accordance with the European guidelines for laboratory animals (86\609\EEC) and were approved by the local committee for laboratory animal welfare. Animals across all treatment groups were randomized over the cages. The continuously ON groups were by necessity in the same cage containing doxycycline-supplemented food. The experimenter performing the electrophysiological and histological analysis was blind to the treatment groups.

In Experiment 1, 45 rats were assigned to four treatment groups including a CMV-GFP control (n = 12), an uncontrolled expression group injected with CMV-GDNF (n = 12) and two groups injected with dox-i-GDNF comprising the 4 week ON/OFF group (n = 9) and 12 week ON group (n = 12) (Fig. 2G). All rats received a unilateral avulsion of lumbar ventral roots L3–L5 (Eggers et al., 2008) under anaesthesia (fentanyl/fluanisone; Janssen Pharmaceuticals; and midazolam, Roche). Directly following avulsion, roots were injected with 1.5 µl of lentiviral vector using a glass needle. Fast Green (Sigma) was added to the viral vector solution to aid in per-operative visualization. The roots were implanted ventrolaterally into their corresponding spinal cord segment and fixed in place using fibrin glue (Tissucol®, Baxter). Buprenorphine (Temgesic®, Schering-Plough) was provided as post-surgery analgesic.

In Experiment 2, 75 rats received an L3–L5 ventral root avulsion. In addition, the L6 ventral root was avulsed and not reimplanted to ensure the small subset of motor axons that remained intact in Experiment 1 was also lesioned. Animals were randomly assigned to either avulsion only (n = 10), CMV-GFP (n = 23), dox-i-GDNF ON (n = 23) or ON/OFF (n = 19) groups.

Doxycycline food supplementation regimen and rationale

Directly following surgery, all rats received doxycycline-supplemented chow (6 g doxycycline/kg, TD.09282, Envigo). At 4 weeks post-surgery all groups returned to regular chow except the ON group, which was kept on doxycycline food for 12 (Experiment 1) or 24 (Experiment 2) weeks. The 4-week timed GDNF expression period is based on three previous findings. First, we found that 6 to 8 weeks of GDNF expression leads to axon entrapment, while at 4 weeks this was not yet observed (Eggers et al., 2008). Second, although the first regenerating axons reach the reimplanted ventral root within 1 week, the majority of axon ingrowth occurs between Weeks 2 to 4 (Eggers et al., 2010). Third, GDNF expression was switched ON directly following surgery to promote optimal motor neuron survival, as motor neurons become progressively unresponsive to reimplantation or GDNF treatment beyond the critical period of 4–6 weeks post-avulsion (Chai et al., 1999; Zhou and Wu, 2006b).

Assessment of compound muscle action potential

The compound muscle action potential (CMAP) was recorded in the gastrocnemic and plantar musculature under isoflurane anaesthesia (Abbott) by an observer blinded to the treatment. For the gastrocnemic CMAP, a supramaximal percutaneous stimulus was applied using an EMG (Nicolet Viking), delivered at the sciatic notch and the CMAP was recorded in the belly of the gastrocnemic muscle. To obtain plantar muscle CMAPs, stimulation was performed retro-malleolar and recorded subcutaneously in the dorsal aspect of the foot between the third and fourth metatarsal. CMAP measurements were performed preoperative to obtain baseline values and at 1 week postoperative to confirm completeness of the avulsion lesion. Recordings of the CMAP amplitude were performed every 2 weeks.

End-point motor nerve conduction velocities (MNCV) of the peripheral nerve were measured prior to perfusion at 24 weeks in animals that did not undergo retrograde tracing. Under deep sodium pentobarbital anaesthesia (Sanofi Sante), gastrocnemic CMAP recordings were performed as described above. Proximal- and distal nerve stimulation was performed at the lumbar plexus rostral to the iliac crest, and at the sciatic nerve 20 mm distal from the sciatic notch (average distance 49.6 mm). The distance and peak latencies between these stimulation sites were subsequently used to calculate the MNCVs.

Retrograde motor neuron labelling

To quantify the motor neurons that regenerated an axon in the sciatic nerve up to 10 cm distal from the implantation site a subset of animals was subjected to retrograde tracing 24 weeks after reimplantation. To ensure equal distribution between groups, animals in each intervention group were ranked according to their CMAP value at 23 weeks. Subsequently odd numbered animals were subjected to retrograde tracing. Under isoflurane anaesthesia, the sciatic nerve was transected 20 mm distal from the sciatic notch and the proximal nerve stump was submerged in 5% FastBlue (EMS-Chemie) for 30 min. Simultaneously a 10 mm distal sciatic nerve segment was dissected for histological analysis 7 days post-tracing.

Tissue preparation and histological procedures

At 12 (Experiment 1) and 24 (Experiment 2) weeks post-surgery, lumbar spinal cords, sciatic nerve and gastrocnemic muscles were harvested and processed for histological analysis (Supplementary material). To visualize motor neurons and their axons, immunohistochemical staining (Supplementary material) of spinal cord and sciatic nerves was performed using primary antibody against choline acetyl transferase (ChAT, 1:200; AB144P Chemicon). Biotinylated secondary antibody (1:200; anti-goat; Jackson ImmunoResearch) was used for DAB immunohistochemistry, amplified by Vector Labs ABC kit (PK-6100, Vector laboratories).

Gene expression and coil formation was assessed after immunofluorescence staining of sections (Supplementary material) incubated with primary antibodies against S100 (1:600; Dako), MBP (1:100, AB9348 Merck Millipore), GFP (1:400, MAB3580 Merck Millipore), GDNF (1:500, AF212NA, R&D systems) and neurofilament (1:200; 2H3, DSHB) and appropriate secondary antibodies (1:800; anti-rabbit Alexa-594, anti-chicken Cy5, anti-mouse Alexa-488, biotinylated anti-goat and streptavidin Cy3 Jackson ImmunoResearch).

Quantification of motor neuron survival and retrograde labelling

Longitudinal spinal cord sections stained for ChAT were used to quantify the number of motor neurons by an observer blinded to the treatment group (Eggers et al., 2010). Only ChAT-positive profiles present in the L3 to L6 ventral horn with a visible nucleus were included. At the 24-week time point (Experiment 2), both traced and non-traced animals were included in the quantification. In a subset of retrogradely traced animals, the number of Fast Blue-positive motor neurons was quantified in every fourth section using a fluorescence microscope.

Assessment of coil formation

To investigate the effects of constitutive or timed GDNF expression on the regenerating motor axons and coil formation after 12 weeks, high resolution images of MBP, S100, NF stained ventral roots were taken 2–8 mm distal from the implantation site. In previous studies, the area of continuous GDNF expression contained dense fibre coils, defined as fibres growing with a circular and/or swerving orientation (Eggers et al., 2008). Quantification of reimplanted ventral root diameters was performed in ChAT stained spinal cord sections. High resolution tiled images of the reimplanted ventral roots were captured using a Zeiss axioplan microscope. Analysis was performed on four systematically selected spinal cord sections per animal, resulting in a controlled dorso-ventral sampling of the ventral root over a 600-µm range. Two outlines were drawn delineating the inner and outer borders of the ventral roots using ImagePro software, with an average outline length of 6 mm. Over the entire length of these lines perpendicular measurements yielded an average diameter per section. Finally, the average ventral root diameter per animal was calculated.

Sciatic nerve motor axon quantification

At 12 (Experiment 1) and 24 (Experiment 2) weeks post-surgery, the number of motor axons located 10 cm distal from the implantation sites was quantified in transverse sciatic nerve sections stained for ChAT. Axons were counted manually by an observer blind to the treatment groups. In Experiment 1, a cluster of relatively large calibre motor axons was frequently found. These motor axons most likely originate from the intact L6 ventral root (Eggers et al., 2016). All axons were included in the quantification due to the inability to discriminate between intact or regenerating axons. In Experiment 2 (24 week) the L6 ventral root was avulsed to avoid the presence of this small subset of intact axons and allow reliable quantification of regenerating motor axons. Distal sciatic nerves harvested during retrograde tracing and of non-traced animals were used for quantification.

Data analysis and statistics

All data are expressed as mean ± standard error of the mean (SEM). Statistical analysis was performed using Graphpad Prism (Graphpad software) and a value of P < 0.05 was considered significant. Normality was assessed with the Shapiro-Wilk test. One-tailed Student’s t-test analysis was performed to determine statistical significance. Time-dependent data were analysed using two-way ANOVA for repeated measures followed by post hoc Bonferroni.

Data availability

The datasets generated during the experimental procedures and analysis are available from the corresponding author on reasonable request.

Results

Dox-i-GDNF confers dox-sensitive GDNF expression in cultured Schwann cells

The performance of our lentiviral vector system was assessed in cultured rat Schwann cells transduced with titre matched CMV-GFP, CMV-GDNF or dox-i-GDNF. Cells transduced with dox-i-GDNF were divided into two groups: cells that received doxycycline for 4 days (ON/OFF) and for 11 days (ON). Addition of doxycycline to cells transduced with dox-i-GDNF results in a significant increase in GDNF protein levels at 4 days, which is maintained over the entire 11-day culture period in the ON group. Removal of doxycycline at 4 days (ON/OFF) results in a gradual decline in GDNF concentration reaching baseline levels at 8 days in culture (Fig. 1B). Schwann cell density was not affected by the exposure to either doxycycline or GDNF (Fig. 1D). Cells transduced with a titre-matched constitutively active CMV-GDNF produced four times more GDNF independent of dox. These observations show that dox-i-GDNF confers dox-sensitive GDNF expression in Schwann cells with no detectable leaky expression in the OFF state.

Dox-i-GDNF enables temporal regulation of GDNF in avulsed reimplanted ventral roots

To assess whether the dox-i-GDNF system regulates GDNF expression in avulsed and reimplanted spinal ventral roots, GDNF expression was analysed using ELISA in contra- and ipsilateral spinal cord segments containing the injected ventral roots (Fig. 3A and B). In the presence of doxycycline, dox-i-GDNF expression was successfully induced and persisted up to 24 weeks. When doxycycline administration was terminated at 4 weeks GDNF levels returned to baseline with low-level leaky expression at 12 weeks (P < 0.02 versus CMV-GFP or control). In Experiment 1 we included our previously used CMV-GDNF vector as a positive control (Eggers et al., 2013). Based upon the titre-matched in vitro results, four times more viral vector particles were injected in the reimplanted ventral roots of the dox-i-GDNF groups. This resulted in levels of GDNF obtained with dox-i-GDNF ON that were comparable to those observed with CMV-GDNF (Fig. 3; not significant P = 0.08). In longitudinal spinal cord sections, immunofluorescence staining for GDNF and GFP protein confirms persistent 24-week GDNF and GFP expression in the reimplanted ventral roots (Fig. 3C–E). GFP-positive nuclei of transduced cells in the reimplanted root of CMV-GFP animals are distributed in a proximo-distal fashion [Fig. 3C(i and ii)]. Secreted extracellular GDNF is noticeably present throughout the reimplanted roots in the dox-i-GDNF ON animals [Fig. 3D(i and ii)], in contrast to the CMV-GFP and dox-i-GDNF ON/OFF group where GDNF expression is very low. These results confirmed the functionality of our dox-i-GDNF gene switch in vivo.

Experiment 1: Timed GDNF expression reduces axonal coil formation, enhances motor neuron survival and axonal outgrowth

To investigate whether time-restricted expression of GDNF prevents the formation of fibre coils and axon entrapment, axon fibre growth was assessed at 12 weeks following either transient (4 weeks; ON/OFF group) or long-term (12 weeks) GDNF overexpression. In the CMV-GFP control group, no coil formation occurred (Fig. 4A, E and G). Consistent with previous studies, in the CMV-GDNF and ON group, roots were hypertrophic, axon growth was chaotic and large nerve coils were present (Fig. 4B). Schwann cells showed a lack of longitudinal alignment and had a disrupted morphology (Fig. 4C, H and I). A significant increase in the ventral root diameter in both CMV-GDNF and ON groups is found compared to the CMV-GFP control (Fig. 4E; P < 0.0001). In contrast, in the ON/OFF group the diameter was significantly reduced and many regenerating axons were oriented longitudinally (Fig. 4D and J; P < 0.002). In the ON/OFF group smaller isolated coils were occasionally observed (Fig. 4D, arrow). These smaller coils were composed of thin axons similar to those seen in the coil areas in the constitutive GDNF expression groups (Fig. 4K). Thus, temporally controlled GDNF gene therapy attenuates excessive axon growth and coil formation in reimplanted ventral roots.

To determine the effects of timed GDNF expression on motor neuron death, we quantified the number of ChAT-positive motor neurons in the spinal ventral horn. In the CMV-GDNF, ON and ON/OFF groups, a significant improvement in motor neuron survival occurs (62%, 75% and 63% of the ipsilateral side, respectively; Fig. 4L; P < 0.01) compared to CMV-GFP. This indicates that timed 4-week GDNF expression is sufficient to achieve motor neuron survival up to 12 weeks post-lesion.

To assess whether timed GDNF expression improved distal axon regeneration after 12 weeks, the total number of ChAT positive axons 10 cm distal from the implantation sites were quantified. During quantification, a distinct patch of apparently intact large diameter motor axons was frequently observed, most likely originating from the intact L6 ventral root. Because it was not possible to reliably exclude these fibres from the quantification, all axons were counted. In both CMV-GDNF and ON groups the average number of fibres (261 ± 60 and 218 ± 73, respectively) was not significantly different from the CMV-GFP group (165 ± 26, Fig. 4M). A significant increase of almost 2-fold was found in the ON/OFF group (412 ± 71) compared to the CMV-GDNF and ON groups (P < 0.008 versus CMV-GFP, P < 0.05 versus CMV-GDNF and ON).

The results of Experiment 1 show that timed 4-week overexpression of GDNF was (i) sufficient to promote survival of spinal motor neurons; (ii) resulted in a reduction in misdirected axon growth; and (iii) stimulated axon regeneration into the distal nerve.

Experiment 2: Timed GDNF treatment improved electrophysiological recovery

Based on the encouraging observations in Experiment 1, a second experiment was designed to investigate whether timed GDNF expression leads to improved recovery during a 24-week long-term survival period. Since in Experiment 1 the results in the CMV-GDNF and ON groups were comparable, we omitted the CMV-GDNF group and introduced an extra avulsion only control group. Furthermore, we performed an additional L6 ventral root avulsion to ensure that spontaneous recovery from the small subset of intact motor axons in Experiment 1 was abolished.

At 7 days after surgery, CMAPs in the large majority of animals were absent as expected. Despite the avulsion of four roots, 17% of the animals displayed a CMAP response. This observation corroborates a recent study in which a postoperative CMAP is found in 13% of the animals after a comparable root avulsion procedure (Torres-Espin et al., 2013). These animals were excluded from CMAP and histological analysis. In the avulsion only group, CMAPs remained absent with the exception of one animal, which regained a small amplitude starting at 12 weeks (Fig. 5B). The first small CMAPs in the CMV-GFP group occurred at 14 weeks in the gastrocnemic muscle, gradually increasing to reach a final amplitude of 0.30 ± 0.07 mV. The first signs of innervation of the distal plantar musculature occur 11 weeks later (Fig. 5B and D). A similar recovery pattern was present in animals in the ON group with a gradual increase in amplitude up to 0.38 ± 0.07 mV at 24 weeks. In contrast, in the ON/OFF group the first CMAPs in the gastrocnemic muscle were found as early as 4 weeks post-reimplantation. A progressive increase occurred over time resulting in significantly higher CMAP amplitudes of 0.72 ± 0.14 mV at 24 weeks (P < 0.001). At 18 weeks 100% of the animals in the ON/OFF group displayed a CMAP, whereas six additional weeks were required for all animals in both CMV-GFP and ON groups to respond (Fig. 5B). Innervation of the distal musculature in the ON/OFF group occurred 8 weeks earlier than in all other groups (Fig. 5D), with a significantly higher amplitude during the last 6 weeks of the experiment (Fig. 5C, P < 0.01). These results indicate a significant benefit of timed GDNF expression over uncontrolled GDNF overexpression on axonal regeneration and reinnervation of the hind paw muscles.

End-point MNCV was measured prior to sacrifice in animals that were not retrogradely traced by stimulation at a proximal- and distal segment of the peripheral nerve. A clear reduction of motor nerve conduction velocity to 20 m/s was observed in all animals compared to 73 m/s in intact contralateral nerves. Specifically, in the ON group, many animals had a reduced conduction velocity, however the reduced average velocity of 12 m/s in this group did not reach significance compared to CMV-GFP and ON/OFF groups (P = 0.052).

Timed GDNF treatment promotes motor neuron survival and distal axon regeneration

Quantification of motor neuron survival was performed 24 weeks post-surgery in longitudinal ChAT stained spinal cord sections in all animals. Severe atrophy of motor neurons occurred in the ipsilateral side of the avulsion only group (Fig. 6A and E) and quantitative analysis revealed a 65% motor neuron loss (Fig. 6I). Reimplantation of ventral roots transduced with the control vector (CMV-GFP) promoted a small but significant degree of motor neuron survival (Fig. 6B, F and I; P < 0.001). The number of surviving motor neurons in both the ON and ON/OFF groups strongly increased compared to avulsion only and CMV-GFP groups (P < 0.0001). Moreover, motor neurons in both ON and ON/OFF groups appeared healthier and larger (Fig. 6G and H). A small but significant difference was observed in the survival of motor neurons between the ON and the ON/OFF group with 87% and 77% motor neuron survival, respectively (P < 0.04).

To assess the number of surviving motor neurons that were able to regenerate one or more axons into the distal sciatic nerve, retrograde tracing was performed in a subset of animals. In the avulsion only group retrogradely-traced motor neurons were rarely observed (4.2 ± 2.0 motor neurons). Reimplantation of the avulsed ventral roots resulted in an average of 44.7 ± 6.0 motor neurons that regenerated an axon up to 10 cm distal from the implantation site. In the ON group, 49.4 ± 13.1 motor neurons were retrogradely labelled, with the majority of animals ranking below the GFP average. In contrast, in the ON/OFF group, the number of retrogradely-labelled motor neurons doubled (87.4 ± 14.4) compared to the CMV-GFP or the ON groups (Fig. 6J; P < 0.03). Further calculations showed that of the surviving motor neurons, 1.4% and 6.0% were retrogradely traced in the avulsion only and CMV-GFP groups, respectively. This fraction is lower in the ON group (4.8%) in comparison to the ON/OFF group (8.8%, P < 0.03), whereas compared to GFP this was not statistically significant. Taken together, both sustained and 4 week time-restricted dox-i-GDNF gene therapy significantly promoted motor neuron survival, but only time-restricted expression of GDNF enhanced long-distance axon regeneration into the sciatic nerve.

Quantitative assessment of axonal outgrowth towards the target muscle 24 weeks post-implantation was performed in transverse sciatic nerve sections 10 cm distal from the implantation site. In the second experiment the L6 ventral root was additionally avulsed. Axon counts in the avulsion only group confirmed the complete absence of intact large diameter motor axons. Despite the avulsion of all four ventral roots, incidentally small diameter motor axons were observed after avulsion only (Fig. 7D). On average, sciatic nerves from these animals contained 17.5 ± 5.2 axons, which were homogeneously distributed throughout the nerve. In both the CMV-GFP and ON groups on average 61.8 ± 10.1 and 71.0 ± 12.2 axons were present, respectively. In contrast, timed GDNF expression in the ON/OFF group leads to a significant 1.8- and 2-fold increase in axons compared to CMV-GFP and ON animals with an average of 124.6 ± 16.1 axonal profiles (Fig. 7D; P < 0.006). Together with the higher number of retrogradely-traced motor neurons observed in the ON/OFF group, this indicates that significantly more motor neurons regenerated axons into the distal nerve.

GDNF treatment reduced muscle atrophy

To obtain insight into hind paw muscle atrophy, the gastrocnemic muscles were dissected from the animals that did not undergo retrograde tracing. In many animals, hind limb contractures were present and in all animals atrophy was macroscopically evident. In avulsion only animals, a severe 86% loss of muscle weight occurred compared to the contralateral muscle (Fig. 8). No significant muscle weight increase is observed in the CMV-GFP or ON groups compared to the avulsion only group (P = 0.06 avulsion only versus ON). Gastrocnemic muscle weight in the ON/OFF group was significantly higher compared to avulsion only or CMV-GFP (P < 0.03 and P < 0.01, respectively) but not to the ON group, although atrophy remained severe in the ON/OFF group (79% loss).

Discussion

In this article we demonstrate the application of an immune-evasive doxycycline-inducible GDNF gene switch (dox-i-GDNF) in a longitudinal spinal cord lesion. We show that time-restricted expression of GDNF for a post-lesion period of 4 weeks promotes the survival of spinal motor neurons, attenuates excessive axon growth and the entrapment of axons, enhances the number of axons that regenerate into the nerve over a distance of 10 cm by 2-fold and facilitates electromyographical recovery by 8 weeks. Persistent doxycycline-mediated delivery of GDNF for 12 and 24 weeks promotes motor neuron survival but, in contrast to time-restricted GDNF expression, impairs axon regeneration by inducing axon entrapment in the roots. These results show that to achieve the goal of an effective GDNF-based gene therapy for ventral root repair the timing of GDNF delivery is critical.

Regulating GDNF protein expression in spinal ventral roots

GDNF is a potent neuron survival factor and has been used to prevent neuronal degeneration and promote axon regeneration (Airaksinen and Saarma, 2002). However, uncontrolled delivery of GDNF causes side effects, including aberrant sprouting and axon entrapment (Blesch and Tuszynski, 2003; Blits et al., 2004; Love et al., 2005; Eggers et al., 2008; Santosa et al., 2013). Although a regulatable viral vector could potentially overcome these problems, preclinical studies demonstrated an immune response against the transactivator resulting in the removal of cells expressing the therapeutic gene (Favre et al., 2002; Ginhoux et al., 2004).

Our study is the first to apply a ‘stealth’ transactivator to create an immune-evasive gene switch for GDNF, reliably regulating short- and sustained long-term GDNF expression in vivo in injured, reimplanted nerve roots. The fusion of glycine-alanine repeat (Gar) to a foreign protein confers immune-evasive properties in vitro protecting cells in a bioassay for T cell-mediated immune attack (Ossevoort et al., 2003; Zaldumbide et al., 2010; Hoyng et al., 2014b). In vivo, this ‘stealth’ component attenuated interferon-γ and CD8b expression in a study investigating the effects of regulated expression of chondroitinase following spinal injury (Burnside et al., 2018). Thus the ‘stealth’ gene switch represents a significant advance for controlling therapeutic gene expression in the injured nervous system.

Attenuated axon entrapment following time-restricted GDNF delivery

Transient overexpression of GDNF for a period of 10 days from adipose-derived stem cells transfected with a GDNF-expression plasmid and transplanted extra-radicularly improved regeneration and was not associated with axon entrapment. However, in this approach a lack over the control of GDNF-expression prevents studying the effect of timing on axon entrapment and regeneration (Pajenda et al., 2014). We found that time-restricted expression of GDNF for 4 weeks significantly reduced nerve hypertrophy and axon entrapment observed after sustained GDNF expression. Regenerating axons displayed a longitudinal direction of growth. Thus, by controlling GDNF expression the ‘candy store’ effect was effectively resolved. However, incidental isolated nerve coils were observed in some nerve roots. These small pockets of aberrant axon growth may be caused by residual GDNF, which was detectable in the nerve roots of ON/OFF animals at 12 weeks after application of dox-i-GDNF in the absence of doxycycline treatment. This possibility is supported by the observation that the number and size of the axon coils in a nerve transduced with a non-inducible vector was dose-dependent and was significantly diminished, but not completely abolished, at lower dosages of GDNF (Eggers et al., 2013). It is well known that the classical rtTA system exhibits some level of leaky expression (Loew et al., 2010). In cultured primary Schwann cells, dox-i-GDNF confers dox-sensitive expression of GDNF with no detectable leaky expression in the OFF state. Cultured Schwann cells do produce endogenous GDNF and leaky expression may only represent a small proportion of endogenous GDNF expression and would therefore not have been detected in the ELISA. The stealth vector had reduced leaky expression in HEK cells (Hoyng et al., 2014b). However, in vivo some residual expression of chondroitinase was detected following treatment with dox-i-CHABC (Burnside et al., 2018). Leaky expression can potentially be reduced by modification of the transactivator (Das et al., 2016; Roney et al., 2016) and/or the tetO sequences in the tetracycline/doxycycline inducible promoter (Loew et al., 2010). Efforts are ongoing to test the potential benefit of these elements in the ‘stealth’ vector.

Regulated GDNF expression promotes long-term motor neuron survival and regeneration

Severe progressive motor neuron degeneration following root avulsion is a major cause of the failure to recover hind limb function. Longer motor neuron survival increases the chance of extending axons into the reimplanted roots, which can then continue their way towards the muscles. We show for the first time that a 1 month timed GDNF treatment enhanced motor neuron survival beyond 12 weeks with only a relatively small decline in survival compared to constitutive expression for 24 weeks. Thus, the ability to regulate GDNF enabled us to show that a 1 month time window of GDNF delivery is sufficient to exert long-lasting neuroprotective effects on motor neurons. This extends earlier observations showing that a single per-operative dose of GDNF protein was sufficient to enhance motor neuron survival for 3 to 12 weeks (Li et al., 1995; Yuan et al., 2000; Wu et al., 2003; Zhou and Wu, 2006b; Chu and Wu, 2009; Chu et al., 2012; Ruven et al., 2018). One-month GDNF delivery is therefore an effective treatment period to protect spinal motor neurons against avulsion-induced death for many months. This is clinically relevant because it significantly extends the time window for long-distance axonal regeneration towards the periphery.

CMAP measurements show that target muscle reinnervation was significantly facilitated by 1-month GDNF treatment. Electromyographic improvements were accompanied by a 2-fold increase in the number of regenerating motor axons in the sciatic nerve. These observations, together with an almost 2-fold increase in the number of retrogradely traced motor neurons, show that the increase in motor axons is not due to sprouting but represents genuine GDNF-induced long-distance regeneration. Despite the improved CMAPs and the increased number of regenerating axons, no improvements in voluntary hind paw function were observed. This is in agreement with the lack of correlation between improved electrophysiological CMAP responses and the degree of voluntary movement (Valero-Cabre et al., 2004; Torres-Espin et al., 2013). CMAPs, fibre counts and retrograde tracing all show that even under the most optimal circumstances, only a small fraction of motor neurons (<10%) are able to regenerate towards the distal nerve. Two possible mechanisms underlying the lack of voluntary hind paw function are the misrouting of regenerating axons and prolonged denervation of the peripheral nerve. Following ventral root reimplantation, regenerating motor axons grow towards ectopic spinal cord areas (Risling et al., 1992; Gramsbergen et al., 2000; Eggers et al., 2010). Despite careful reconstructive surgery, axonal misrouting can lead to reinnervation of incorrect target muscles, impeding the recovery of function (Brushart et al., 2002; Valero-Cabre and Navarro, 2002; English, 2005; Torres-Espin et al., 2013). In addition, lack of function despite the formation of structurally sound neuromuscular junctions is further attributed to synaptic failure (Sakuma et al., 2016). After prolonged periods of denervation, the Schwann cells in the nerve lose their proregenerative properties and growth inhibitory proteins accumulate in the nerve (Fu and Gordon, 1995; Zuo et al., 1998; Sulaiman and Gordon, 2000; Ronchi et al., 2017). Over time, the velocity of axonal growth declines significantly when long distances and chronically denervated nerve stumps have to be bridged (Hoke et al., 2002; Gordon et al., 2003; Eggers et al., 2010; Torres-Espin et al., 2013). The longer the distance between the lesion site and the target the longer it takes for an axon to extend distally and the less favourable the cellular environment of the nerve becomes for successful regeneration. As a consequence, many regenerating axons will not reach their correct target muscle. In the future, approaches are required that prolong the post-lesion period during which Schwann cells in a chronically denervated nerve continue to support axon regeneration, including the expression of neurotrophic factors along the entire denervated nerve pathway, the neutralization of inhibitory molecules in the chronically denervated nerve and reprogramming of Schwann cells into a proregenerative state (Arthur-Farraj et al., 2012; Li et al., 2015).

Perspective

Many studies revealed the unprecedented potency of neurotrophic factors as agents to promote neuronal survival and axon regeneration. However, to take full advantage of the therapeutic effects of neurotrophic factors and to provide a safeguard against unwanted side effects, it is necessary to carefully control their expression. Our immune-evasive gene switch represents a powerful tool to control the timing of expression of a neurotrophic factor gene in the injured nervous system. The preclinical data presented here support the development of a treatment where neurosurgical ventral root repair in patients is combined with temporally restricted GDNF gene therapy. In patients with a longitudinal spinal cord lesion, progressive motor neuron death and the subsequent failure of these motor neurons to regenerate towards distal targets is the cause of life-long functional deficits, even if they undergo surgical repair. The sustained survival of motor neurons would significantly extend the time that is available for axons to regenerate towards and establish functional connections in the periphery.

Funding

This work was supported by grants from the Wings for Life Spinal Cord Research Foundation (WFL-NL-17/16 to J.V.), International Spinal Research Trust (TRI004_03 to J.V.) and a gift from the Dwarslaesiefonds (to J.V.).

Competing interests

The authors report no competing interests.

Abbreviations

Abbreviations
 
• ChAT

  choline acetyl transferase

 
• CMAP

  compound muscle action potential

 
• dox-i-GDNF

  doxycycline inducible GDNF vector system

References