|Home | About | Journals | Submit | Contact Us | Français|
This review highlights current tissue engineering and novel therapeutic approaches to axonal regeneration following spinal cord injury. The concept of developing 3-dimensional polymer scaffolds for placement into a spinal cord transection model has recently been more extensively explored as a solution for restoring neurologic function after injury. Given the patient morbidity associated with respiratory compromise, the discrete tracts in the spinal cord conveying innervation for breathing represent an important and achievable therapeutic target. The aim is to derive new neuronal tissue from the surrounding, healthy cord that will be guided by the polymer implant through the injured area to make functional reconnections. A variety of naturally derived and synthetic biomaterial polymers have been developed for placement in the injured spinal cord. Axonal growth is supported by inherent properties of the selected polymer, the architecture of the scaffold, permissive microstructures such as pores, grooves or polymer fibres, and surface modifications to provide improved adherence and growth directionality. Structural support of axonal regeneration is combined with integrated polymeric and cellular delivery systems for therapeutic drugs and for neurotrophic molecules to regionalize growth of specific nerve populations.
Current interventions offer little hope of functional recovery for patients after a spinal cord injury (SCI). In the United States, the incidence of SCI is 32 injuries per million population, approximately 11,000 new injuries per year, affecting a young group of people of median age 26, that is predominately male (82%). Road traffic accidents, acts of violence, falls, and sports injuries account for the majority of injuries. The average inpatient stay is 9 months, during and following which the patient’s life, in virtually all aspects, is profoundly changed (Lali et al., 2001). 45.7% of the 253,000 persons living in the United States with the residual of spinal cord injury have permanent and complete paraplegia or tetraplegia, irreversible loss of neurologic function below the level of the injury (National Spinal Cord Injury Association, 2009).
The majority of patients die from respiratory complications. Injury at any level of the spinal cord will impair respiratory function, through the destruction of motor nuclei and descending motor tracts innervating diaphragmatic, thoracic, intercostal and abdominal accessory muscles. Equally impaired are ascending sensory signals for muscle control via stretch reflexes, for cough, vomit and secretion clearance and from peripheral respiratory chemoreceptors. These axons project through the spinal cord to and from a neural network in the brainstem comprising three interconnected centers, the pontine group, and the medullary dorsal and ventral respiratory group. The pontine group (parabrachial/Kölliker-Fuse complex) controls respiratory timing, receives input from lung stretch receptors, and links respiration to behavioural cues; the dorsal group receives afferents from respiratory chemo and mechanoreceptors, and coordinates respiratory-cardiac reflexes; the ventral group (Bötzinger complex) projects inspiratory neurons, expiratory motor neurons rostrally, and includes a pre-complex generating the respiratory rhythm. Axons descend in the spinal cord in the anterolateral white matter to phrenic, intercostal and abdominal motor neurons, laterally in the high cervical cord near the spinothalamic tract for autonomic function and with the corticospinal tracts for voluntary respiratory control (Nogues and Benarroch, 2008).
Accordingly, respiratory failure with spinal cord injury occurs as a consequence of alternations in tidal volume, ventilation and its pattern, diminished responses to hypercapnia, reduced lung and chest wall compliance, and progressive respiratory muscle fatigue due to compensatory breathing rates. Hypoxia from respiratory compromise can further the neurologic injury. Common secondary pathology includes (aspiration) pneumonia, atelectasis and the complications of mechanical ventilation (Lane et al., 2008). Injury to the cord can also induce paralytic ileus worsening aspiration. More severe respiratory compromise occurs with higher levels of injury with risk to phrenic motor nuclei located in cervical spinal cord segments C3-C5 (occasionally as low as C7) (Zimmer et al., 2007).
Pathological (Quencer and Bunge, 1996) and imaging studies (Bodley, 2002) demonstrate tissue destruction with cysts and gliosis in the area of injury, along with atrophy in adjacent segments of cord. Strategies aimed at preventing immediate and delayed secondary damage need to be administered within minutes or hours of injury. Even if ideal protective agents were available, many patients would not be in circumstances where this would be available or successful. The area of cysts and glial scarring does not contain cells or tissue that contribute to regeneration and is consequently both a gap and a barrier to regeneration. There are, therefore, only two ways to re-establish neurologic function below the block: bypassing the area or rebuilding functional tissue within the cyst/scar. A functional bypass might be established by nerve autograft connections from areas above the lesion to distal effectors (cord or muscle) (Tadie et al., 2002). The second approach is to replace the cyst/scar with functional tissue, promoting the development of neural tissue bridges to carry regenerating axons from above to roots or muscles below the lesion (Friedman et al., 2002). For future use in patients, replacement of a segment of cord would be suitable for those with massive damage to the cord with no evidence of residual functional tissue in the area. Unfortunately, this accounts for a significant number of patients.
Animal models of spinal cord injury include complete transection (thoracic), hemisection (dorsal or unilateral), and contusion injuries (forceps and computer-controlled weight impact). These models approximate common human pathology, open cord laceration (1/4 of injuries) and closed compression/contusion injuries (3/4). Biomaterial polymers may be delivered as gels, suitable for contusion and small tears, as devices designed to fill larger defects (sponges) or to bridge large gaps and traverse the glial scar (tubes and multichannel scaffolds (Hiroshi et al., 2006). While deep tears or transections are rare in human injury, complete or partial transections in animal models are useful as proof of concept, and for the controlled study of axonal regeneration (Talac et al., 2004). Animal models of respiratory dysfunction focus on high cervical injury producing diaphragm hemiplegia, but no studies to date have employed polymer-based tissue engineering strategies specifically in this context. Given the severity of patient morbidity and the rates of mortality associated with respiratory compromise, neurologic repair is an important therapeutic goal. A relatively short distance, from the medulla to phrenic C3-C5 or within the phrenic segments for example, needs be bridged by new neuronal tissue. Equally, respiratory innervation associates with discrete tracts, corticospinal and spinothalamic, and repair may often be unilateral given a lateral injury and diaphragmic hemiplegia. Such tracts represent ideal targets for polymer scaffold implantation given their limited scope and clinical importance.
The classification of biomaterials for the spinal cord is based on whether the materials are naturally derived or synthetic, whether they are hydrogels, whether or not they are biodegradable, as well as other sub-classifications based on specific modification or functional adaptation (surface charged, drug-delivery etc.). Regardless of the source or application, the material must have properties which are biocompatible specifically within the spinal cord environment (Kohane and Langer, 2008). These properties in turn influence the regenerative capacity of engineered structural support for neurite outgrowth at a macro (i.e. fascicular) and micro (axonal) level.
In relation to blood, cerebral spinal fluid is low in cellular nutrients. Scaffold permeability to various molecular sizes becomes crucial for access to oxygen and nutrients and removal of metabolic wastes. The degree to which a material swells within the aqueous environment of the spinal cord must be known if the scaffold is to maintain an appropriate alignment and not compress regenerating nerves. Degradation kinetics may be accelerated by the ingrowth of axons and by the deposition of extracellular matrix by support cells of the CNS or by the therapeutic cell line seeded within the scaffold. Stiffness, permeability, swelling, strength and degradation are of course specific to the particular polymer employed, are all readily modified through changes in polymer concentration or constituent ratios. de Ruiter et al. (2008) exemplify the type of in vitro characterization and methodology required to develop an implant of any polymer type. In this study, the authors present a series of methods to characterize multichannel nerve tubes for properties of bending, deformation, swelling, and degradation and introduce a new method to test the permeability of multichannel nerve tubes from the rate of diffusion of different-sized fluorescent dextran molecules. Equally, the implantation methodology must be developed. While the material is placed within rigid spinal column, the spine may require further fixation (Fig. 1) (Rooney et al., 2008). Scaffolds in unfixed spines have a greater tendency to produce scoliosis and become displaced (Fig. 2). The material should be of sufficient softness not to physically damage the cord as the animal moves. The degradation products of the polymer, and any residual agents used in its fabrication, cannot be locally or systemically cytotoxic or elicit an immune response which will further gliosis, and be destructive both to the scaffold complex and any regenerating axons (Liu and Cao, 2007). Cells detect mechanical characteristics of the environment through adhesion complexes and the actin cytoskeleton, and the stiffness of the substrate may be of critical importance (Discher et al., 2005). Finally the tensile strength of the material – its ability to hold a suture for example – will contribute to its clinical use.
Scaffold biomaterials for spinal cord placement are fabricated by dissolving the monomer in an aqueous or organic solvent to produce a liquid state that is polymerized into macromers by a chemical, thermal or photo-crosslinking reaction. Additional reagents may be added to enhance the crosslinking reaction. Synthetic polymers often employ the use of chemical initiators and accelerators to fine-tune polymerization rates. The majority of spinal cord scaffolds is made by injection molding. To create pores in the structure, porogens are incorporated into the polymer mix. Commonly used porogens include sodium chloride crystals, ice crystals, gas bubbles introduced by peroxides or air-foaming, and gelatin composite materials. In each case, the polymer forms around the porogen, which itself is removed from the final structure leaving only the space it occupied (Fig. 3). The size of the pore therefore can be controlled by adjusting the variables that control crystal size and the direction of crystal growth (Madaghiele et al., 2008), or by adjusting the water content of a hydrogel.
A continuous porous structure closely mimicking the intrinsic mechanical characteristics of the original tissue may provide a better environment for regeneration (Discher et al., 2005; Deguchi et al., 2006). Material porosity is essential for cell attachments, allows for greater distances that may be bridged, and improves functional recovery following transection (Jeng and Coggeshall, 1985; Vleggeert-Lankamp et al., 2007; Reynolds et al., 2008). Porosity also allows for tissue vascularization of the avascular scaffold implant, influences cell migration and phenotype, and will improve implant stability at the cord-scaffold interface through interlocking between implant and cord tissue (Dadsetan et al., 2008).
Nerve tracts in the human spinal cord have diameters of 100–1000 μm. It is the role of scaffold macroengineering to design conduits whose architecture optimizes axonal growth potential through the alignment of fascicular groups (Fig. 4). Channel sizes depicted in the templated agarose scaffold are on the order of 200 μm (Fig. 4A), and 450 μm in the PLGA scaffold (Fig. 4C). In the first study of its kind, Wong et al. (2008) directly compared porous poly (ε-caprolactone) synthetic polymer scaffolds cast in five different architectures, cylinder, tube, multichannel, and open-path design with and without a central core (Fig. 4B). The findings demonstrated not only that open path designs were improvements over the other three designs in terms of regenerative capacity, but also that the other more closed designs adversely affected the surrounding cord, doubling the defect length. In a multichannel model, the channel size is of importance. Spinal cord scaffolds with multiple longitudinally aligned channels of 450 and 660 μm were constructed from PLGA using injection molding. When seeded with Schwann cells, this scaffold design supported robust axonal growth (Moore et al., 2006), although also demonstrated the formation of a rim of fibrous tissue surrounding the core of regenerated neurons. Further work evaluating the relationship between scaffold channel diameter and the number of axons regenerating showed a larger area of fibrous tissue and a reduced axon number in the larger channel size (Krych et al., 2009).
A large myelinated axon in the CNS has a diameter of 15–20 μm. In order to create functional reconnections, the scaffold must also regionally align separated neuron groups at an axonal and growth cone level. Scaffold microengineering refers to designing features that are in the order of a few microns in at least one dimension (Khademhosseini and Langer, 2007). In the context of the spinal cord, this includes micro-structures that will provide precise directionality to growth and improve axonal adherence at the level of the advancing growth cone. It is conceptually useful to consider the re-alignment of nerve fascicles as 3-dimensional bundles of axons which themselves prefer to grow or elongate along the planes of 2 dimensions (Bellamkonda, 2006). Microgrooves can be placed in the polymer surface by laser etching, affecting contact guidance and alignment of neurites. In vitro work has demonstrated optimal sizes of 2 μm minimum groove depth (Clark et al., 1991), with more narrow ridges (5 μm versus 10 μm) improving the number neurites aligned as well as the number of focal contact adhesions in a given cell (Goldner et al., 2006). Further improvements in neurite outgrowth are seen with coating the grooved surface with collagen or laminin peptides, Fig. 5(A and B) (Yao et al., 2009). Axonal growth cones preferentially advance up gradients of laminin (Adams et al., 2005). For the spinal cord, gradients have been made in natural and synthetic polymers with laminin (Dodla and Bellamkonda, 2006).
Axons also show a preference to grow along the length of micro and nanofibres of polymers such as collagen and synthetic polymers, Fig. 5(C and D). Polymer fibres are made by monomer self-assembly, producing a randomized fibre orientation of larger caliber fibres, or by electrospinning for parallel alignments of nanometer scale fibres (50 nm–30 μm). This technique uses electric charge to draw and elongate threads of liquid polymer (collagen, PLGA, PCL) from a syringe pump source. The solvent evaporates from the airborne filament which is laid down upon an electrically grounded plate or rod, or onto a spindle apparatus (Yang et al., 2005). Collagen filament bridges with fibres of 20 μm show promise as 2-dimensional surfaces to guide axons, improving axonal density and motor function, but appear also to have significant cytotoxicity with high animal mortality in a rabbit model (Yoshii et al., 2004; Yoshii et al., 2009). A goal of scaffold microengineering is to pack larger channels of scaffolds with these fibrous substrates to approximate normal axonal densities (Schnell et al., 2007).
The spinal cord lacks a support matrix equivalent to the endoneurium and perineurium in peripheral nerve, one that can act as conduits to approximate disconnected axonal groups. Furthermore the axonal density of the spinal cord far surpasses that of peripheral nerve with far less extracellular matrix support. The rationale for polymer implants is to replace a damaged area of the cord with just such a structural matrix. Natural polymers are biological fibrillar protein, polysaccharide, or glycosylaminoglycan (GAG) carbohydrates which form hydrogels. These polymers already have an intrinsic function such as extracellular matrix or structural support, and a degradation profile by in vivo enzymes that is in keeping with their natural role. Hydrogels are mesh networks of insoluble polymer fibres, through which water can freely flow to osmotically swell or shrink the overall structure. Hydrogel polymers are attractive materials for use in the spinal cord. They are macroporous, soft materials readily allowing cell adhesion and migration, while nutrients and wastes are easily exchanged. They can be easily shaped to fit the defect, their elasticity and degradation may be adjusted by component density. Table 1 details the main natural polymers and highlights examples of their application within the spinal cord. Two spinal cord extracellular matrix (ECM) components are used, collagen and hyaluronic acid (HA); two polymers are derived from marine plants, agarose and alginate, and chitosan is derived from insect or crustacean shells.
Being the predominant extracellular matrix protein, type I collagen has intrinsic properties including molecular sites for cell adhesion and migration, inherent signaling transduction for proliferation and differentiation, and mechanical properties similar to soft tissue. Antigenicity is low, provided the origin species is the same as the host. Solutions of collagen are polymerized by adjusting pH, or with the addition of ionic salts. Whereas early use focused on its application as a 3-dimensional matrix growth, it was soon realized that collagen itself has a limited capacity to support axon growth (Marchand et al., 1993), and that its use required further functionalization. An important trend for the use of collagen is in combined strategies, particularly as a growth or elution matrix within or on the surface of other polymer types (Tsai et al., 2006) or as a composite material (Cheng et al., 2003). Axonal extension onto collagen can be improved through covalent modification, or the incorporation of cell-adhesion molecules such as laminin, to provide directional guidance as a gradient along collagen fibres (Yao et al., 2009). Collagen also has the advantage of being thermoresponsive, gelling at physiologic temperatures, making it attractive for use as an injectible polymer delivery system. This property enables the incorporation of neurotrophic factors, drugs or cells at the time of gelation without thermal damage to the factor or cell line. Incorporation of neurotrophin-3 (NT-3) and brain-derived neurotrophic factor (BDNF) (Houweling et al., 1998) improved axonal counts and animal function, including a specific regional improvement of corticospinal tract density with the use of NT-3. Degradation of the collagen in situ allows for sustained release of the growth factor, as well as an improved surface for cell attachments.
The glycosaminoglycan HA was thought to be a good material candidate given its role as extracellular matrix in the brain, but its success in supporting axonal growth is modest. As a scaffold material, HA has been used with benefit as a matrix for cultured embryonic spinal cord tissue placed into transected cord (Rochkind et al., 2002). HA has however been developed into an extremely useful co-polymer gel with methylcellulose for intrathecal drug delivery (Gupta et al., 2006). Whereas methylcellulose gels at increasing temperature, unmodified HA quickly disperses in vivo. The combination of acetate-modified HA with methylcellulose (HAMC) has the distinctive property of already gelling at room and physiologic temperature prior to its delivery, but become liquid when subjected to the mechanical sheer forces involved with syringe and needle injection (Katz and Burdick, 2009). Collagen embedded epidermal growth factor (EGF) (Shoichet et al., 2007), and erythropoietin (Kang et al., 2009), have been safely delivered with sustained release in situ from HAMC. The latter agent enhanced neuroprotection with reduced cavitation size and increased neuron numbers following clip compression of the spinal cord.
Agarose is used in many of the same ways as collagen for spinal cord repair. It is a linear polysaccharide derived from seaweed and crosslinked by temperature gradients through hydrogen bonding. Agarose is thermoresponsive, but at temperatures lower than 37 °C. It has been used as an injectable system when it can be rapidly cooled in situ using liquid nitrogen vapour (Jain et al., 2006). Such a system is now being developed for direct topical delivery of dexamethasone onto the injury site, from drug eluting nanoparticles suspended within an agarose implant (Chvatal et al., 2008). Early inflammatory infiltrates and lesion size were reduced by day 7. Like collagen, agarose itself is relatively impenetrable by axons, but serves as an excellent axonal growth substrate, particularly when functionalized with laminin gradients (Dodla and Bellamkonda, 2006). Tuszynski and co-worker have used a freeze drying method to form agarose scaffolds containing linear guidance pores with a mean diameter of 120 μm (Stokols and Tuszynski, 2004, 2006). This process involves the formation of ice crystals whose size and direction of growth can be controlled by the temperature gradient (Tabesh et al., 2009). Pore size in the scaffold can also controlled by the freezing rate and pH, with the faster rate creating smaller sizes (Sachlos and Czernuszka, 2003). Integrating BDNF into the scaffold material (Fig. 6) and in separate experiments, BDNF-secreting mesenchymal stem cells scaffold channels, significantly improved the scaffold’s capacity to promote regeneration (Stokols and Tuszynski, 2006). BDNF within lipid microtubules has also been incorporated into agarose scaffolds, enhancing axonal growth for the length of the scaffold but not into the distal cord (Jain et al., 2006).
Alginate is obtained from algae and the polymer solution is crosslinked by calcium into a sponge-like structure. Such a structure supported axonal extension in the spinal cord and limited gliosis (Kataoka et al., 2001). Hippocampal neurospheres and BDNF-secreting fibroblasts have been seeded onto alginate and placed into the transected cord (Nomura et al., 2006). Agarose and alginate requires ultrapurification prior to use, given that commercial preparations often contain mitogens and cytotoxic byproducts.
Chitosan is a GAG carbohydrate polymer derived by chemical deacetylation of chitin, the major structural polysaccharide found in crustacean, shellfish and insect shells. Cell adhesion to the structure is determined by the extent of its positive charge, itself a function of the degree of alkaline deacetylation (Nisbet et al., 2008). Further improvements in cell attachment are seen with the addition of poly-L-lysine to the polymer mix, and a thermoresponsive polymer can be made with the addition of glycerol phosphate salts (Crompton et al., 2007). Chitosan scaffolds support axonal growth in the spinal cord (Freier et al., 2005), and the polymer may be used to encapsulate therapeutic cell lines (Yuan et al., 2004), Fig. 7(lower). Recent work by Shoicet and co-workers demonstrates chitosan’s use as extramedullary and intramedullary conduits capable of supporting neural stem cell differentiation in the transected cord (Nomura et al., 2008; Zahir et al., 2008).
Plasma derived polymers, fibronectin and fibrin, are being used as spinal cord scaffolds. Fibronectin mats are formed with linearly aligned fibres which can orient axonal growth (King et al., 2003) and sequester growth factor within its pores for gradual release (Phillips et al., 2004). Fibrin scaffolds are formed from monomers following fibrinogen cleavage by thrombin and crosslinked with Factor XIIIa (Willerth and Sakiyama-Elbert, 2007). It is a natural matrix for wound repair having inherent cell-binding sites. Heparin has been crosslinked to fibrin scaffolds using a bidomain Factor XIIIa-heparin linker peptide for use as an affinity-based delivery system for growth factors, including NT-3 (Taylor and Sakiyama-Elbert, 2006) and for factors to differentiate embryonic stem cells seeded within the matrix. A recent study implanting fibrin polymer scaffolds into a dorsal hemisection model demonstrated delayed astrocytosis and improved neuron fibre extension (Johnson et al., 2009), Fig. 7(upper panels).
Whereas the use of natural polymers in the spinal cord takes advantage of their inherent properties, their natural role to some extent being the basis of their function or their modification, using synthetic polymers offers wider scope to design and control the characteristics of the material. The synthetic polymers used thus far in the spinal cord are either biodegradable materials based around polyesters of lactic and glycolic acid (PLA and PGA), are biodegradable hydrogels based on polyethylene glycol (PEG), or are non-biodegradable hydrogels based on methacrylate. Early spinal cord scaffolds were made from the same materials as were in common clinical use for surgical repair of peripheral nerve and skin grafting. Rapid advances in hydrogel chemistry have produced materials more suited to the spinal cord in their mechanical properties. The trend now is towards using these highly aqueous, soft polymers given the similarity of their properties to spinal cord tissue, and the versatility with which their chemistry and architecture can be adjusted. Functionalized synthetic polymers have included gradients and surface charge modification for cell adhension, neurotrophic gradients, and have opened the field to using scaffolds themselves as drug delivery and gene delivery vehicles to novel extent. The technical ability to rapidly photocrosslink synthetic hydrogels has also enabled a remarkable degree of sophistication in macro and micro-architecture through the use of photolithography. Please see Table 2 for a listing of synthetic polymers and examples of their use within the spinal cord.
At the time of our previous review (Friedman et al., 2002), much of the work in scaffold design focused on the use of biodegradable synthetic polymers, particularly the poly (α-hydroxy acids) PLA, PGA and their co-polymer PLGA. These polymers are polyester links of lactic and glycolic acid which are hydrolyzed in vivo to release lactide and glycolide, dissolving the material. The pH around the grafted site accordingly will become more acidic. These compounds were a good initial choice for spinal cord placement, having a long track record of FDA approved clinical use as an absorbable suture material, and as grafting material for skin and for peripheral nerve repair (Mackinnon and Dellon, 1990; den Dunnen et al., 1993). The transition to CNS applications was based on the idea that such scaffolds could provide a corresponding PNS-like endoneurial and perineurial guidance structure to regenerating spinal cord axons, improved with the addition of myelinating Schwann cells within the scaffold (Hadlock et al., 2000).
PLA has been used to make scaffolds. Having shown the resorbability and biocompatibility of PLA with Schwann cells and the spinal cord (Gautier et al., 1998), the University of Miami group placed single channel Schwann cell loaded scaffolds into the transected rat spinal cord. The PLA materials used were structurally unstable, fragmenting and collapsing, but proved to support the extension of axons and vascular growth into the graft (Oudega et al., 2001). More recently, macroporous PLA foam scaffolds made with longitudinally aligned pores were fabricated using a freeze drying technique. BDNF was dissolved into the scaffold matrix but did not improve an overall low yield in axon numbers through the graft (Patist et al., 2004). This study was extended to incorporate Schwann cells into the foam that had been genetically engineered to secrete a bi-functional neurotrophin (D15A) with BDNF and NT-3 activity (Hurtado et al., 2006). Axonal regeneration was modest at 6 weeks, and few Schwann cells survived the first week of scaffold placement. There has been recent interest in the use of PLA nanofibres as a cell substrate (Wang et al., 2009).
The degradation rate of PLA can be somewhat controlled in the co-polymer PLGA by altering the ratios of the PLA and PGA in composite. The in vitro characteristics (bending, swelling, deformation, degradation and permeability) of varying PLA:PGA ratios have been extensively determined for PLGA spinal cord scaffolds implants (de Ruiter et al., 2008). Multichannel PLGA scaffolds have been demonstrated to support robust axonal regeneration when seeded with Schwann cells without functional improvements (Moore et al., 2006). As detailed in the section below, PLGA degradation kinetics are the basis for drug delivery via microspheres embedded in polymer scaffolds.
Polyethylene glycol is a biodegradable synthetic polymer of ethylene oxide units. Its role is somewhat unique in that it is an ‘exclusionary’ compound, immunoprotecting the areas to which it is applied by keeping out cell infiltrates, and equally is used as a delivery system for cells, neurotrophins and genetic constructs. It has been formulated into gels for topical application onto the injured spinal cord (Borgens et al., 2002), combined with PLA for neurotrophin delivery of a photo-inducible gel, or as a formulation of ‘pegylated’ BDNF as an intrathecal infusion (Soderquist et al., 2008). Intravenous solutions with magnesium sulfate are being investigated as a first line immunomodulatory therapy in acute spinal cord injury (Kwon et al., 2009).
Poly(2-hydroxyethyl methacrylate) (pHEMA) compounds were initially used as non-biodegradable materials for soft contact lenses. Professor Shoicet’s group in Toronto has extensively developed pHEMA and pHEMA-co-methyl methacrylate (pHEMA-MMA) for use as spinal cord scaffolds. Early applications included pHEMA sponges (Giannetti et al., 2001), evolving to guidance channels (Tsai et al., 2004), with a variety of surface modifications to improve cell adherence and axonal extension (Moore et al., 2006; Tsai et al., 2006) including neurotrophic gradients and adhension gradients within a chitosan composite (Yu et al., 2007). Microfluidic techniques are employed to create functional concentration gradients in scaffolds, again providing cues for directionality of axonal growth. The technique involves casting a scaffold using 2 or more inlet ports coupled with rapid polymerization. An important advantage of this polymer class is the modification of surface charge with the addition of quaternary amine groups or of a second methacrylate subtype. Many cell types adhere better to a positively charged surface. Lesny et al. (2006) evaluated the in growth of neural tissue in a dorsal hemisection injury bridged with four pHEMA composites, demonstrating an improvement in axonal regeneration into the core of the scaffold and a reduction in astrocyte infiltration in the positively charged scaffold.
Very sophisticated architecture is possible with these compounds. The scaffolds have been reinforced with coils, as well as being made as with neurotrophin PLGA microspheres. Multilayered macroporous pHEMA composites, including innermost channel layers that elute neurotrophic factors, have been made by liquid–liquid centrifugal casting. Thin layers of the polymer liquid are forced against the side of the mold by the centrifugal force and polymerized. The polymers can also be crosslinked with light with the addition of a photoinitiator. Using photolithography, in which discrete sections of the polymer mix are exposed to light while others are masked, open channel size and pore size were recently controlled with remarkable precision (Bryant et al., 2007).
As one of the field’s principle contributors, Bellamkonda advocates that scaffold strategies for nerve regeneration should combine four main components, a permissive growth substrate (hydrogel or micro/nanofibre), a neurostimulatory extracellular matrix (protein or peptide), the provision of neurotrophic factors, and glial or support cells (Schwann cells, neural stem cells). Cell therapies can be delivered to the spinal cord by direct injection into the cord substance, intrathecal infusion, or by polymeric microspheres or scaffolds. It has been shown that a variety of cells support axonal regeneration within polymer scaffold models in the cord. Cells can be loaded onto polymer scaffolds when suspended in a support matrix such as fibrin or Matrigel™. If the scaffold is a macroporous hydrogel, the cells migrate and become resident within that porous structure, and if the scaffold is designed to be multichannel, select channels can be seeded with different cell types allowing for regional topography (Friedman et al., 2002).
Early work in the spinal cord was again derived from peripheral nerve repair strategies. Schwann cells myelinate peripheral nerve, and will naturally migrate from peripheral-central nerve junctions such as the dorsal root ganglia during times of cord injury to support repair (Oudega et al., 2005). Here they play a structural role, lay down extracellular matrix proteins like laminin, and provide paracrine trophic support through secretion of nerve growth factor (NGF), NT-3, BDNF, ciliary neurotrophic factor (CNTF) and basic fibroblast growth factor (bFGF) (Willerth and Sakiyama-Elbert, 2008). Xu et al. (1995) placed Schwann cells expanded in culture from harvested rat sciatic nerve into a poly(acrylonitrile-co-vinylchloride) (PAN/PVC) nerve conduit and into a complete cord transection model. The study demonstrated axonal extension and myelination of about a quarter of the regenerated sensory nerves as seen with electron microscopy, and that the extent of the repair could be enhanced by administering methylprednisolone in the acute phase of the injury (Chen et al., 1996). The Miami group also utilized peripheral nerve grafts placed directly into the transected spinal cord serving as natural conduits for axonal extension and as a source of Schwann cells (Oudega and Hagg, 1996, 1999). This work was extended to cell-seeding polymer scaffolds, demonstrating evoked nerve conduction potentials across the graft only in the Schwann cell group (Pinzon et al., 2001). In all applications, it is important to understand whether cells survive in the grafted material (natural or synthetic). We have demonstrated cell survival for up to 6 weeks after transplantation through a multichannel scaffold, and that in this model, Schwann cells have a higher capacity than stem cell neurospheres to enhance axonal regeneration in the transected cord, Fig. 8 (Olson et al., 2009).
Olfactory ensheathing glia are cells within the peripheral and central component of the olfactory system that contribute to the regenerative capacity of olfactory neurons. The average lifespan of olfactory receptor neurons is 4 weeks. These neurons are bipolar cells projecting to the nasal epithelium and to the olfactory bulb. New receptor neurons are derived from a stem cell layer at the base of the nasal epithelium, from where axons are projected through the cribriform plates into olfactory bulbs. The role of OEC is to enfold and guide growing axons to the bulb in bundles of unmyelinated axons, and once at the bulb to interact with resident astrocytes and fibroblasts to finalize the connections (Franssen et al., 2007). While extensive work (over 40 in vivo studies) has been done with cell injections into the cord, few studies have utilized polymer scaffolds. In one of the first, and most successful studies Bunge’s group used Schwann-cell seeded PAN/PVC conduits in a thoracic transection model, now in combination with stereotactic injection of OECs at four midline depths. The OECs induced axonal growth in through the SC-containing channel and for distances of 2.5 cm, the longest distance observed thus far for OECs (Ramon-Cueto et al., 1998). Followup studies showed functional recovery without the use of scaffolds (Ramon-Cueto et al., 2000). Lu and Ashwell (2002) used collagen matrix soaked with OECs to bridge a dorsal transection and with similar efficacy to pieces of olfactory lamina propria and OEC injections. Chuah et al. (2004) encapsulated OECs into polyvinylidene fluoride particles for injection following dorsal transection which increased the numbers of collateral branches from the intact ventral cord. A strategy which used Schwann cells embedded in a Matrigel™ bridge, injections of OECs into the distal and proximal cord stumps following transection, and alternate day delivery of Chondroitinase ABC via an intrathecal catheter, resulted in select fibre regeneration and some recovery of function (Fouad et al., 2005).
Neural stem cells (NSCs) are a pluripotent, self-renewing population of precursor cells that give rise to astrocytes, oligodendrocytes, and neurons in the CNS. Their role in regenerative medicine is therefore to remyelinate axons growing from the injured cord, to themselves become neuronal links within the injury (Lowry and Temple, 2007), and to elaborate neurotrophic factors to stimulate regrowth (Lu et al., 2003). In animal models, they are derived from fetal brain homogenates, from which they are cultured as spherical aggregates (neurospheres), which can be in turn be subcultured for in vitro differentiation or in vivo transplantation. Stem cells for use in humans pose several problems. Their nature as multipotent cells risks unrestrained proliferation to desired or to unwanted cell lineages. They must be derived from fetal tissue in order to preserve the full range of pluripotency, or be obtained in the adult from brain or spinal cord biopsy; the latter as a source may provide cells whose lineage is more restricted to a neuronal phenotype. Cross-species implantation will not avoid immune surveillance. When injected into the cord following contusion injury (Cao et al., 2001) or a hemi-section model (Chow et al., 2000), neurospheres tend to adopt an astrocyte morphology and the neurite connections which do form tend to create pain circuits (Hofstetter et al., 2005). That indeed no study has shown large scale neuronal differentiation of engrafted stem cells suggests the injured cord, regardless of the injury model, is not a permissive environment (Enzmann et al., 2006). There have been numerous efforts to control cell lineage with growth factors, location and timing of cell harvest, and genetic transduction prior to implantation.
Many researchers have turned to polymer scaffold substrates as a biomechanism to differentiate the cells into neurons either in vitro prior to implantation, in vivo in conjunction with a scaffold implant, or to encapsulate the cells with polymer for immune protection (Zhong and Bellamkonda, 2008). Spinal cord scaffolds then may well play a central role should neural stem cells become useful for human clinical therapy. Recent in vitro work has set out to define which polymer compounds and structures are permissive or inhibitory to stem cell viability and differentiation. The most promising substrates include the use of 3-dimensional fibrin scaffolds, seen in vitro to support and differentiate mouse embryonic stem cells when cell culture and scaffold fabrication conditions were optimized (Willerth et al., 2006). Further work evaluated the effects of growth factors, their doses and combinations on stem cell differentiation to neurons and oligodendrocytes (Willerth et al., 2007). PLGA has also been shown to support NSC viability and neurite outgrowth to a greater extent than poly-ε-caprolactone (PCL) and PLA (Bhang et al., 2007). In vivo work done by the Langer lab has used PLGA scaffolds seeded with NSCs in a hemisection model: this scaffold had complex architecture, with an inner layer seeded with NSCs approximating the gray matter, and an outer layer whose pores were oriented longitudinally for axon guidance and radially for permeabililty. Scaffolds were left in the animal for up to a year with persistent functional improvement, including hindlimb weight bearing and improved coordination seen between 2 and 3 months post implantation (Teng et al., 2002). Our group has also seeded neurospheres into multichannel PLGA scaffolds in a direct comparative study with Schwann cells, Fig. 9, demonstrating the capacity of PLGA to support neurosphere differentiation (Olson et al., 2009).
Micro and nanostructures influence stem cell growth and differentiation in vitro. A micropatterned polystyrene surface provided growth differentiation and direction to hippocampal stem cells with laminin (Recknor et al., 2006) and with growth factor cues without adhension molecules (Oh et al., 2008). Fibrous nanostructures differentiate NSCs and may even control cell lineage through the structure itself. Silva et al. showed self-assembled nanofibres could rapidly differentiate NSCs to neurons and inhibit the development of astrocytes (Silva et al., 2004) Similarly, electrospun nanofibres of PCL polymer differentiated cells and provided a stimulus and directionality to neuron growth (Xie et al., 2009).
The therapeutic potential of spinal cord scaffolds is enhanced by the development of integrated polymer drug delivery systems and cell lines that are genetically modified to secrete neurotrophic factors. Polymeric delivery from scaffolds is achieved by means of the materials inherent properties: its porosity and permeability for the sequestration and diffusion of a drug its degradation kinetics for release of the entrapped drug, its chemical affinity to a drug by means of a linker moiety, or by being a non-biodegradable material for the preservation of a gradient (Willerth and Sakiyama-Elbert, 2007). We have seen in the discussion of individual polymer types a number of ways each polymer type can be used to deliver neurotrophins or drugs. The drug or factor can be released from the material itself, from integrated micro- or nano-spheres or tubules of a different material, or by means of a scaffold’s capacity to support a genetically modified cell line in vivo.
Micro- and nano-spheres refer to particulate synthetic polymer of the order of microns or nanometers in diameter. By far the most frequently used polymer is PLGA, as its degradation rate and thus drug release is readily controlled by the proportion of PLA to PGA. For example, 85:15 PLA to PGA will degrade significantly more slowly than a 50:50 ratio. PLGA microspheres are typically 2–45 μm in size. Alginate-chitosan microspheres are also used. Micro and nanospheres are produced by microemulsion techniques, whereby an aqueous solution is emulsified in an organic phase polymer solution to create spherical droplets which are then extracted into another external aqueous phase (Benoit et al., 2000). The size of the droplet can be controlled by the emulsion agitation rate, the aqueous and organic phases used, and the addition of surfactants to modify the surface tension between the phases (Khademhosseini and Langer, 2007). The drug is usually stabilized with protein (bovine serum albumen) or with zinc, added to the aqueous phase and becomes encapsulated upon polymerization of the droplets. Other techniques include aerosol freeze drying. The rate of drug delivery will depend on the initial concentration in the sphere and the size of the sphere for a polymer of given degradation kinetics. The particles are injected as a suspension, or can be suspended in a scaffold of another material, in which case the pore size of the material must enable diffusion or cell access within. Alternatively, as shown in chitosan scaffolds (Fig. 10), the spheres can be localized directly to the channel wall by centrifugation (spin-coating) techniques (Kim et al., 2008). Nanoshells may be useful in the delivery of hydrophobic drugs from scaffolds. These spherical particles combine the benefits of PEG liposomes with polymeric shell, and are fabricated from PLGA polymer, lecithin phospholipid and a PEG core using modified emulsion techniques (Chan et al., 2009).
Microspheres in scaffolds allow the opportunity for sustained local delivery of therapeutic molecules within the blood–brain barrier. Neurotrophins embedded in PLGA microspheres have been used in a variety of polymer scaffolds with NGF (Mahoney et al., 2006) being the most extensively characterized. Cyclic AMP has been shown in many models to induce or enhance axonal growth (Murray and Shewan, 2008). Similarly the work of Bunge and co-workers has demonstrated that neurotrophins enhance axonal growth in regenerating spinal cord (Blits et al., 2003) and that this may be further enhanced by c-AMP (Pearse et al., 2004). Chondroitinase ABC (ch-ABC) an enzyme which degrades a variety of chondroitin and heparin sulfate proteoglycans (Plant et al., 1998) as well as hyaluronan, disrupts the glial scar matrix and facilitates axonal repair. Treatment with ch-ABC significantly enhanced axon regeneration in vitro (Plant et al., 1995) and following brain (Nash et al., 2002) or peripheral nerve lesions (Richardson et al., 1980) and, most importantly, SCI in rats (Stichel et al., 1999). Furthermore, functional recovery in rats after SCI has been observed (Bradbury et al., 2002). We have demonstrated sustained release of dibutyryl cAMP for over 3 weeks when incorporated into microspheres. Finally, while the efficacy of systemic delivery of methyprednisolone in the acute setting for SCI remains an area of contention, polymeric delivery of the steroid is more beneficial in animal models than systemic steroid in reducing the lesion volume when delivered from a scaffold (Kim and Martin, 2006) or from an injectible nanoparticle colloid (Kim et al., 2009).
A common fabrication technique is to incorporate a neurotrophic factor into the polymer mix prior to polymerization, whereby the entire scaffold elutes the drug in an isotropic fashion. Natural materials are particularly well suited to this approach, given their polymerization rarely involves processes that will denature or destroy the factor. This approach was among the earliest used in the spinal cord, whereby a collagen matrix with NT-3 was placed into a transection model (Houweling et al., 1998). The collagen construct specifically attracted corticospinal sensory tracts, promoting partial recovery, and in doing so exemplified important principles of polymer scaffold design. The chemotrophic effect of neurotrophins locally delivered from polymers enhanced the regenerative capacity of the scaffold, and did so in a manner that can be regionally controlled on the basis of the neurotrophin’s target axonal population.
A second delivery approach is to use synthetic multichannel channel scaffolds whose conduits are filled with eluting polymer. Iannotti et al. (2003) filled a minichannel PAN/PVC with Matrigel™ containing recombinant glial cell derived neurotrophic factor (GDNF), or GDNF Matrigel™ with additional Schwann cells. GDNF attracted axons primarily of propriospinal phenotype, and the effect of Schwann cells was synergistic. The polymer matrix in the channel itself can also significantly impact the axonal phenotype regenerated. When filled with either Matrigel™, collagen, fibrin, or methylcellulose, fibrin matrix within pHEMA-MMA channels demonstrated preferential regeneration of descending reticular neurons, while methylcellulose promoted growth of vestibular and red nucleus neurons (Tsai et al., 2006).
Thirdly, genetically modified cells that over-express neurotrophic factors have been placed in the injured cord. Initial work, and the vast majority of work to date, involves the injection-engraftment of cell suspensions with or without combinations of ‘cell-bridges’ of cell suspension matrices. Early on, fibroblasts and Schwann cells were transfected with retrovirus to express NT-3 (Grill et al., 1997), NGF (Grill et al., 1997) and BDNF (Menei et al., 1998) respectively. In these 3 studies, NT-3 cells supported corticospinal tract growth in a bilateral dorsal column transection model. NGF secreting cells engrafted by the same group in the same model supported dorsolateral primary sensory axons and cerulospinal axons, but did not support the growth of corticospinal or motor axons. Suspensions of BDNF-Schwann cells implanted into a complete transection whose ends were re-apposed, enhanced the growth primarily of reticular and raphe neurons.
In the context of scaffold delivery, lentivirus was used to transfect Schwann cells to secrete a bifunctional neurotrophic protein (D15A) with NT-3 and BDNF functionality, but cell survival was poor in a PLA scaffold. BDNF-secreting MSCs in templated agarose scaffolds survived to significantly improve regeneration over that supported by the scaffold alone (Stokols et al., 2006). Different cell types and thus different neurotrophic delivery systems can be placed within the same scaffold framework to align corresponding axon groups (Fig. 9A). From the initial studies cited above, a great deal of work has clarified the sensitivity of select axons to specific neurotrophic support (see Lu and Tuszynski (2008) for a recent review). NGF is involved in the growth of nociceptive cells, BDNF those that modify the activity of motor neurons, NT-3 the corticospinal tracts and dorsal sensory axons, NT4/5 proprioceptive and motor modifying inputs, and GDNF the proprioceptive, dorsal sensory, and nociceptive neurons. Despite these approaches to augment neurotrophic support and phenotypic selection, there remains no convincing evidence that axonal populations are extending beyond the cellular or polymer bridges to form lasting functional connections. Furthermore, there is evidence that the inherent regenerative capacity of the axonal subtypes may be different. As recently suggested, within the spinal cord, propriospinal may exhibit a greater regenerative capacity than rubrospinal and reticulospinal, each of which is superior to corticospinal (Blesch and Tuszynski, 2009).
This review has highlighted some of the novel tissue engineering approaches to spinal cord repair using implantable polymers in thoracic transection models. Restoration of respiratory function will be a critical application for biomaterial approaches given the degree of mortality and morbidity associated with respiratory compromise after spinal cord injury. Important bioengineering considerations, fabrication techniques for macroengineering and microengineering, along with some prominent applications and modifications of the main materials in use for scaffolds, have been presented as an overview of the field. Regarding polymer materials, these continue to evolve rapidly in the sophistication of their chemistry and design. New applications with the polymer Polypyrrole (Ppy) represent an exciting avenue for nanofibre and drug delivery systems. Used in the past as coating for neuroelectrodes, Ppy is electrically conductive, and voltage stimulus can induce axonal growth and its direction, along with the release of integrated neurotrophic factors and dexamethasone from the polymer surface.
The integration of Schwann cells, olfactory ensheathing cells and neural stem cells into the polymer structures have improved their regenerative capacity, as has the use of scaffolds as delivery devices for therapeutic agents, particularly neurotrophic factors. Designing scaffolds and seeded cell lines to release neurotrophins for specific axonal locations and sensitive phenotypes may regionalize repair. The use of human embryonic stem cell implantation in polymer scaffolds will be directed by recent FDA approval for Phase 1 clinical trials, and the outcomes of renewed studies in animals. The absence of their use has the influenced development of alternate stem cell types for scaffolds including mesenchymal stem cells and induced pluripotent stem cells (iPSCs) as potential sources both of new neurons as well as drug delivery cells. Further advances in polymers as genetic delivery tools, particularly for siRNA, non-viral and viral gene delivery with or without targeted genomic integration, is an exciting therapeutic prospect. It is hoped that combination strategies will maximize our ability to regenerate spinal cord tissue through the glial scar and recreate connections. The caveat of course is that the investigation of many possible combinations of material, geometry, functionalization strategies, and integrated drug or cell based therapies will need to be properly controlled and systematic. The ability to control these variables with increasing precision, through rapid technological advances such as were highlighted here, enables scaffold models of spinal cord repair to be highly informative about individualized facets of the repair process.
This paper is part of a special issue entitled ‘Spinal cord injury—Neuroplasticity and recovery of respiratory function’, guest-edited by Gary C. Sieck and Carlos B. Mantilla.