SUMMARY & General Discussion

Spinal cord injury is a severely disabling condition, rendering many patients paralysed as a result from damage to motor and sensory neurons in the spinal cord. Injured spinal cord neurons do sprout at the site of the injury but cannot regenerate over extended distances and therefore do not reconnect to their distal targets, which is necessary to regain function. This phenomenon, termed abortive regeneration, was first described by Ramon y Cajal (1928). However, depending on whether spinal cord injury is complete or incomplete, partial spontaneous recovery of function is observed in spinal cord-injured patients (reviewed by Raineteau and Schwab, 2001b). Thus, although neuroregeneration of severed spinal cord nerve tracts does not occur, collateral formation and compensatory sprouting of spared spinal cord axons does occur (Bareyre et al., 2004; Fouad et al., 2001; Raineteau and Schwab, 2001b), resulting in some degree of locomotor function.

We found that after a bilateral lesion of the rat dorsal corticospinal tract (CST) there was only a very limited decline in locomotor function ( Chapter 2; Figs. 3 and 4). Because lesioning the dorsal CST leaves the ventral CST intact, ventral fibers start to form collaterals that project to inter- and motoneurons located in the spinal cord gray matter (Brosamle and Schwab, 2000). Furthermore, severed dorsal CST fibers proximal from the lesion also form collateral sprouts that project to the gray matter (Fouad et al., 2001), a process we observed as well ( Chapter 2; Figs. 1C and D). Bilateral lesioning of the rubrospinal tract (RST) revealed an immediate severe loss of motor control that improved significantly over a long time period, indicating that plastic changes in the neural tracts or circuitries take place over a long period of time after the lesion. Finally, a bilateral dorsal hemi-section severing both the CST and RST, results in a sharp decline in locomotor function that only partially returns in time. These results ( Chapter 2; Figs. 3 and 4; see also Weidner et al., 2001) showed that there is substantial plasticity of spinal nerve tracts, but when too much spinal cord tissue is damaged after injury, functionally significant spontaneous recovery of motor function will not occur. Although the above described studies have all been performed in rats, lesioning of the CST at the thoracic level in monkeys have been shown to result in spontaneous recovery of locomotor function as well (Courtine et al., 2005).

Although CNS neurons are intrinsically able to regenerate (reviewed in Bregman, 1998; Hagg and Oudega, 2006)and show a certain degree of plasticity after injury (reviewed in Raineteau and Schwab, 200 1b)injury to the spinal cord does nearly always lead to permanent loss of function. The lack of autonomous regeneration following CNS injury is thought to be due to a lack of neurotrophic support at the injury site (Li et al., 2007; Liebl et al., 2001; Widenfalk et al., 2001) and the development of a neurite outgrowth-inhibiting environment (reviewed in Fawcett and Asher, 1999; Hagg and Oudega, 2006; Schwab and Bartholdi, 1996). Since the discovery of nerve growth factor (NGF; Levi-Montalcini, 1964), as the first protein that can promote neurite outgrowth, a number of additional neurotrophic molecules (NTs) have been identified that enhance neuronal survival and neurite outgrowth in vitro and in vivo. Brain-derived neurotrophic factor (BDNF), glial cell-line-derived neurotrophic factor (GDNF), neurotrophic factor 3 (NT-3) and ciliary neurotrophic factor (CNTF) (reviewed by Lewin and Barde, 1996) applied in different spinal cord injury paradigms all provide neurotrophic support and stimulate axonal regeneration. Strategies to deliver NTs to the injured spinal cord include i) intracerebral or intrathecal delivery with a minipump; ii) implantation of genetically altered cells that secrete NTs; iii) liposome- or non-viral vector-mediated gene delivery and iv) viral vector-mediated gene delivery. These delivery strategies and their applications in spinal cord injury research are reviewed in the introduction of this thesis ( Chapter 1). An important disadvantage of intracerebral or intrathecal delivery of axonal growth-promoting proteins by injection is that the short half- life of most growth-promoting molecules necessitates repeated administration. Delivery with a minipump does allow continuous application of growth promoting proteins but clogging of the minipump does occur over time and the application of minipumps can also results in compression and scarring of the spinal cord at the site of delivery (Jones and Tuszynski, 2001). An alternative, very promising approach to deliver NTs is gene transfer in which genes encoding the neurite outgrowth-promoting proteins overexpress these factors in the area of the injured spinal cord. This can be performed by implantation of genetically altered NT-producing cells, e.g., Schwann cells or olfactory ensheathing glia (OEG) implanted in the lesion area after spinal cord injury, which is particularly effective in contusion injuries where fluid-filled cysts exist (reviewed by Oudega and Xu, 2006; Ruitenberg et al., 2006). Non viral vector-mediated delivery of NT genes with cationic liposomes to the injured spinal cord suffers from low transfection efficiencies and short term expression of the delivered protein even in post-mitotic cells such as neurons (Lu et al., 2002). A direct way to deliver NTs to the injured spinal cord is injection of a viral vector encoding an NT in CNS tissue. This results in the transduction of cells at the site of injection and in the expression of an NT at the site where regeneration is to be stimulated. The additional advantages of the use of a viral vector to overexpress NTs compared to other delivery methods is that i) endogenous cells are genetically modified to overexpress NTs at physiological levels; and ii) a single injection with a viral vector results in long-term transduction of targeted cells. Hence, viral vector- mediated delivery is a minimally invasive technique to transfer NT-encoding genes to the injured spinal cord.

This thesis describes the use of lentiviral (LV) vectors in different CNS injury models in the rat ( Chapter 3-6) either to overexpress a reporter gene to investigate the transduction efficiency of this vector ( Chapter 3 and 5) or to overexpress the growth- promoting molecules CNTF ( Chapter 4), BDNF and GDNF ( Chapter 6) in an attempt to enhance neuronal survival and axonal regeneration. LV vectors transduce both dividing and non-dividing cells and incorporate their genetic material in the genome of the transduced cells (Naldini et al., 1996b), resulting in stable transgene expression without gene silencing commonly observed with the use of retroviral vectors (Pannell and Ellis, 2001; Zufferey et al., 1998). This means that neurons as well as glia cells can be transduced. This provides the possibility to express neurotrophic factors locally at the site of the injury to enhance neurotrophic support and render the inhibitory environment of the neural scar more permissive for axonal regrowth.

In order to create a neurite outgrowth-permissive environment at the level of spinal cord injury after a dorsal hemisection, we first analysed transgene expression in neural scar tissue after direct application of a LV vector encoding the reporter green fluorescent protein (GFP). As described in Chapter 3, GFP expression was found up to 14 days after delivery of LV-GFP in different cell types constituting the neural scar. Although the LV vector is a pan-tropic vector (Naldini et al., 1996a; see also Chapter 1), this study revealed predominant transgene expression in astrocytes, whereas no meningeal fibroblasts in the core of the neural scar were transduced. This was an unexpected result given the observation that cultured meningeal fibroblasts can readily be transduced ( Chapter 3, Fig. 2). The absence of transduction of meningeal fibroblast might be due to the formation of an extensive extracellular matrix (reviewed in Busch and Silver, 2007) preventing diffusion of the LV vector particles into the core of the neural scar or may be due to phagocytosis of viral vector particles by macrophages invading the core of the injury site. Nevertheless, since the transduced astrocytes are in close proximity to the injured neurons, these cells form an ideal target for LV vector-mediated expression of NTs to enhance regeneration of injured spinal cord neurons.

In Chapter 4, a LV vector encoding CNTF was used to transduce Schwann cells in vitro, which were subsequently used to reconstitute a freeze-thawed peripheral nerve graft. The reconstituted nerve was then neurosurgically inserted into the lesioned optic nerve in order to support regeneration of lesioned retinal ganglion cell (RGC) axons and of survival of RGC bodies in the retina. It has previously been shown that a freeze-dried peripheral nerve graft repopulated with autologous Schwann cells supports regeneration of RGC axons when grafted in the lesioned rat optic nerve (Cui et al., 2003a). The transduced Schwann cells expressed high levels of biological CNTF ( Chapter 4, Fig. 1) and implantation resulted in better survival of RGC and in enhanced regeneration of its axons. This result successfully demonstrated that LV vector-mediated transduction of a peripheral nerve graft with a gene encoding a neuronal outgrowth-promoting molecule can further enhance regeneration of damaged CNS fibers through a transduced peripheral nerve graft.

The feasibility to transduce Schwann cells in vivo in a peripheral nerve with a LV vector to eventually support regeneration of lesioned motoneuron axons was subsequently investigated in a ventral root avulsion model ( Chapters 5 and 6). It has been shown that motoneurons in the ventral horn after avulsion of the ventral root of the spinal cord display retrograde cell death (Koliatsos et al., 1994) and that reimplantation of the avulsed ventral root enhances motoneuron survival and even stimulates regeneration of lesioned motoneuron axons (Carlstedt et al., 1986; Gu et al., 2004). In Chapter 5 the possibility of transducing Schwann cells in the avulsed ventral root was investigated. Strong GFP expression was observed in Schwann cells and macrophages at earlier time points after LV-GFP injection, whereas at later time points GFP-expressing cells had virtually disappeared ( Chapter 5, Fig. 7). Since large numbers of GFP-containing/expressing macrophages were present in the avulsed, injected and reimplanted root, it was assumed that an immune response was triggered by GFP, a foreign protein, which might have been the cause of the decline in transgene expression. Therefore a LV vector was used encoding a GFP protein fused to the GlyAla-repeat (GAr) of Epstein Barr Nuclear Antigen 1, which prevents an immune response against GFP. Avulsed and reimplanted ventral roots treated with LV-GArGFP contained GFP-expressing Schwann cells even 16 weeks after injection ( Chapter 5, Fig. 8). With the use of the GlyAla repeat fused to an exogenous transgene it is thus possible to induce long-term transgene expression in the peripheral nerve without triggering an immune response reducing transgene expression. This is important in studies where long-term expression of endogenous genes is required and viral vector-mediated gene transfer needs to be compared with a control vector encoding an exogenous foreign reporter protein.

In a follow-up study ( Chapter 6), LV vectors encoding BDNF or GDNF were injected in vivo into the avulsed ventral root and subsequently reimplanted into the spinal cord. Previous work has shown that adeno-associated viral vector (AAV)- mediated expression of BDNF or GDNF into the ventral motoneuron pool after avulsion and reimplantation of the ventral root, resulted in an enhanced survival of the ventral horn motoneurons (Blits et al., 2004). Although increased outgrowth of damaged motoneuron axons was observed, the regenerating axons did not leave the area of NT expression and did not grow into the avulsed and reimplanted ventral root. Injection of LV-BDNF or -GDNF into to the avulsed ventral root rather than to the site of the affected motoneurons in the spinal cord was therefore investigated. This approach also prevented motoneuron atrophy, but only after treatment with LV-


GDNF and not following application of LV-BDNF ( Chapter 6, Fig. 4). Prominent regeneration of axons into the reinserted root was observed, but again they failed to extend further into the distal parts of the reimplanted root and formed large, coiled structures at the sites of local high concentrations of GDNF ( Chapter 6, Fig. 6). Trapping of regenerating motoneuron axons is presumably also responsible for the observed lack of functional recovery in this study. A possible solution for this problem lies in the application of viral vectors with regulatable expression (Blesch and Tuszynski, 2007; Gossen and Bujard, 1992).

When to intervene?

Despite the advances in spinal cord injury research there is at present no treatment for spinal cord injury. An important issue with intervening strategies, including those explored in this thesis, and designed to enhance the functional outcome after spinal cord injury is choosing the optimal post-lesion time point to apply potential therapeutic interventions. As already mentioned in the above SUMMARY, the degree of spontaneous functional (locomotor) recovery depends to a large extent on the severity of the spinal cord trauma. Therefore, clinicians treating spinal cord-injured patients will be reluctant to intervene in a very early stage after the initial trauma, because they want to allow as much spontaneous functional recovery to occur as possible. Furthermore, additional damage to spinal cord tissue or side effects of the treatment is a great concern of clinicians involved in clinical trails for spinal cord injury. On the other hand, intervening strategies to enhance functional outcome after spinal cord injury should probably be applied in an early stage in order to render maximal benefit.

Spontaneous recovery after partial or incomplete lesions in the rat has been described in Chapter 2 of this thesis and was clearly demonstrated following a bilateral RST lesion. After a RST lesion there is a sharp decline in locomotor performance that without any intervention other than repeated testing limb function gradually improved in time and nearly reached control levels as far as over ground locomotor control is concerned, indicating the potential plasticity of the CNS after spinal cord injury. Naturally occurring plasticity in the form of sprouting and collateral formation of intact fibers and/or severed fibers above the level of the lesion has also been described (Bareyre et al., 2004; Fouad et al., 2001; Weidner et al., 2001) and contributes to spontaneous functional recovery. We observed similar spontaneous sprouting and collateral formation in hemisected animals ( Chapter 2, Figs. 1 and 7).

In human, spontaneous recovery of spinal cord injury very much depends on the extent of the lesion. The degree of recovery is typically better in subjects with milder deficits at first presentation. Moreover, the degree of recovery is different for each patient in terms of the gradual return of particular sensory and motor functions (reviewed in Blight and Tuszynski, 2006). Thus any clinical trial in this patient group will have a substantial number of patients with spontaneous recovery and will thus bias the outcome measures. Together with the risks of surgery in an area that is already damaged, the neurosurgeon will not readily perform any intervention early following the lesion. People in chronic stages of injury in whom essentially no improvement in function can be established (typically 12-18 month post-injury; Burns et al., 2003) and thus with a relatively stable base-line of functional defects may be more eligible for any clinical test trial to initiate neuroregenerative axon growth. This holds also for complete spinal cord injured patients, with practically no probability of spontaneous recovery. Most interventions in spinal cord injury animal models that resulted in functional recovery were applied shortly after injury whilst so far only a few animal studies have been performed studying the efficacy of interventions in chronic spinal cord lesion models. In future studies aimed at enhanced neuroregeneration and improvement of functional outcome, more chronic spinal cord injury animal models need to be used in order to design safe and efficient treatments for spinal cord injury (Courtine et al., 2007).

Future perspectives

The most effective strategy to successfully promote functional recovery after spinal cord injury will probably be multifaceted and has to include the stimulation of outgrowth by applying NTs, neutralizing neurite outgrowth inhibitors and promoting plasticity of intact neural pathways (Bradbury and McMahon, 2006). Most of the current approaches do address only one of these aspects to enhance functional outcome and are based on either stimulation of neuronal outgrowth of lesioned nerve tracts (reviewed in Chapter 1) or on counteracting the activity or the expression of inhibitory molecules present at the site of the lesion (reviewed in Yiu and He, 2006). Approaches using overexpression of NTs to enhance axonal regeneration after spinal cord injury are based on the observation that there is a lack in neurotrophic support after injury (Liebl et al., 2001; Widenfalk et al., 2001), which in combination with gliosis and the appearance of a fibrous scar results in abortive regeneration of CNS nerve fibres (Fawcett and Asher, 1999; Yiu and He, 2006).

Viral vectors are very efficient tools to express genes encoding NTs both in vivo and in vitro. Viral vectors can either be used to directly transduce CNS tissue (in vivo) or to transduce cells in vitro, which are subsequently implanted in the injured CNS (ex vivo; reviewed in Chapter 1). During the past decade viral vector systems have been improved continuously and long-term gene expression in the CNS has now been established especially with the use of AAV and LV vectors (reviewed in Jakobsson et al., 2003b; Ruitenberg et al., 2002a). Depending on the location and cell types to be transduced in the CNS, viral vectors can be used containing cell-specific promoters or be pseudo-typed with viral envelopes that confer differential specificity for the transduction of particular cell types (Jakobsson and Lundberg, 2006; Shevtsova et al., 2005). In order to transduce cells in vitro or CNS tissue in vivo regardless of the cell type targeted, LV vectors containing the strong CMV promoter driving transgene expression are amongst the most widely used. In addition, LV vectors are commonly pseudo typed with the VSV-G envelope conferring a wide range of host cell transducability (Burns et al., 1993), although predominant neuronal transduction has been reported after in vivo injection in the rat brain (Naldini et al., 1996a) or predominant astroglial transduction after injection in the rat injured spinal cord ( Chapter 2).

During recent years it has become clear that to optimize viral vector-mediated gene transfer to the injured CNS, only certain cell types or certain populations of cells need to be transduced. To confer cell specific expression after viral vector- mediated gene transfer, vectors can be pseudo typed with a variety of viral envelopes. Pseudotyping LV vectors with the glycoprotein of the rabies-related Mokola virus or murine leukaemia virus (MuLV) resulted in strong transgene expression after injection in the rat striatum and hippocampus, with expression confined to neurons and oligodendrocytes (Watson et al., 2002). Pseudotyping LV vectors with Ross River glycoprotein resulted in efficient transduction of both neurons and astrocytes after injection into different rat brain regions (Jakobsson et al., 2006a). Cell-type specific transduction can also be achieved with the use of different AAV serotypes. AAV-2 transduces predominantly neurons after injection in the rat CNS (Kaplitt et al., 1994; McCown et al., 1996), as do AAV-2 vectors serotyped with AAV-1 or -5 (Burger et al., 2004). An advantage of AAV-2/5 serotyped vectors is that they diffuse much further away from the point of injection and thus transduce a larger area (Davidson et al., 2000; Shevtsova et al., 2005) Although sero- or pseudotyping confers cell-specificity to some extent, additional control of transgene expression in specific cell types can be obtained with the use of cell-specific promoters. Astrocyte- or oligodendrocyte-limited expression has been shown after injections into the rat brain with LV vectors encoding transgenes under control of the human GFAP or MBP promoter, respectively (Jakobsson et al., 2003a; McIver et al., 2005). Thus by combining pseudotyped viral vectors encoding transgenes under control of cell-specific promoters, long-term transgene expression can be achieved in only one particular cell type. This would be a powerful tool to express NTs or siRNAs targeting outgrowth inhibitory molecules in specific cell populations in the injured spinal cord, when a direct viral vector-mediated gene transfer approach is used.

Another important issue pertinent to viral vector-mediated delivery of NTs to the injured CNS is the post-lesion time window needed for NT expression. As described above, AAV and LV vector-mediated gene delivery results in long-term expression. In studies using these vectors to overexpress NTs in order to enhance axonal regeneration, a major problem encountered is the trapping of regenerating nerve fibers due to local high concentrations of NT (Blesch and Tuszynski, 2007; Blits et al., 2004). In Chapter 6 this trapping or “candy store” effect is described for LV vector-mediated overexpression of GDNF in the avulsed and reimplanted root, where regenerating motor fibers remain at the local high concentration of GDNF. In fact the GDNF expression resulted in a lower axon density distally from the avulsion/ reinsertion side as compared to control intervention. A solution may be the use of inducible promoters to control the expression of NTs. An inducible system frequently used, is based on a promoter that can be activated by tetracycline or doxycycline (Gossen and Bujard, 1992). Furthermore, a number of AAV and LV viral vectors have been reported that contain more advanced tetracycline-inducible promoters reducing the potential “leakiness” of tetracycline-controlled transgene expression (Chtarto et al., 2007; Pluta et al., 2007; Szulc et al., 2006). BDNF expression in the injured spinal cord under control of a doxycycline-inducible promoter resulted in sustained growth of regenerating axons into the lesion area, which remained there even when NT expression was turned off by removal of doxycycline (Blesch and Tuszynski, 2007). This shows that transient expression of NT molecules is sufficient to sustain regenerated axons in a cellular graft without a decrease in axonal density. For the trapping phenomenon observed in Chapter 6, an auto-regulatory promoter would be a very powerful tool to improve the outgrowth of motor fibers to their distal targets. One such candidate may be the GFAP promoter. GFAP is expressed in Schwann cells that detach from the severed axon after sciatic nerve injury (Cheng and Zochodne, 2002). NTs expressed under the control of the GFAP promoter are then turned on, which has been shown for astrocytes in the rat brain (Jakobsson et al., 2006b). As soon as Schwann cells along the regenerating sciatic nerve start to myelinate the axon, the GFAP promoter is turned off and consequently the expression of transgenic NTs is turned off. This would hypothetically result in a “wave” of NT expression guiding the regenerating axon tip to its distal target, resulting in a functional connection. Thus, future approaches on injury-induced transgene expression that is auto-regulatory, and that would overcome the need for complex regulatory transgene expression systems, would be a big step forward for the application of gene therapy in neurotrauma.

The experiments described in this thesis were conducted in order to characterize a LV vector gene transfer approach for CNS trauma. The results show the feasibility of LV vector- mediated gene transfer of NT to the injured CNS in a number of different CNS injury models. However, as discussed a number of aspects still have to be optimized such as targeting specific cell populations and regulating NT expression.

Furthermore, with the identification of additional genes and proteins involved in the process of regeneration (e.g. transcription factors; (Stam et al., 2007), matrix metalloproteinases and proteoglycans and developmentally active neurite outgrowth inhibitors (Niclou et al., 2006; Pizzi and Crowe, 2007), viral vector-mediated overexpression or knock down (siRNA) of specific novel target genes involved in neuroregeneration will develop into an increasingly important and more and more sophisticated tool in neuroregeneration research.