BACKGROUND: The design of implants comprised of biodegradable electrospun nanofibers for the purpose of bridging injuries in damaged spinal cord is discussed. Electrospun nanofibers structurally mimic the extracellular matrix on which neurons and other cell types grow in vivo. This property has created great interest for their use in tissue engineering applications. However, their employment as biomimetic surfaces for such in vivo applications is still in its infancy. RESULTS: A nonwoven fabric comprised of electrospun polyamide nanofibers supported modest axonal regeneration in injured adult rat spinal cord. Covalent modification of the nanofibers with a bioactive peptide derived from the neuroregulatory extracellular matrix molecule tenascin-C enhanced the ability of the nanofibers to facilitate axonal regrowth. However, the random orientation of the nanofibrillar fabric folds was an impediment to the forward movement of axons. CONCLUSIONS: Polyamide nanofibers covalently modified with neuroactive molecules provide a promising material for grafts to promote spinal cord regeneration. However, for the proper guidance of regrowing axons, attention must be paid to the engineering of ordered nanofibrillar structures. 2007 Society of Chemical Industry
During development of and following injury to the nervous system, the growth and regeneration of axons is strongly influenced by astrocyte-derived extracellular matrix molecules 1 and Schwann cell-derived basement membrane molecules. 2 (The basement membrane is a structurally compact form of the extracellular matrix.) Examples of such molecules include thrombospondin in the extracellular matrix 3 and laminin-1 in the basement membrane. 4 Therefore, artificial scaffolds that are extracellular matrix/basement membranemimetic might provide ideal substrates for neuronal growth, particularly for in vivo applications that require regenerative scaffolds to be compatible with tissue. Although the extracellular matrix produced by astrocytes includes both positive (e.g. thrombospondin, fibronectin) and negative (e.g. chondroitin and keratan sulfate proteoglycans) effectors of neuronal growth, 1,4 -6 the underlying hypothesis is that nanofibers will form a scaffold that mimics extracellular matrix-based supportive as opposed to inhibitory cues.
Work with electrospun nanofibers has supported this hypothesis. 7 Nanofibers produced via the process of electrospinning have unprecedented porosity, a high surface to volume ratio and high interconnectivity, all physical properties that are ideal for cellular attachment and growth. 7 The nanofiber aggregates can be deposited in either a random or aligned array, to result in random or oriented axonal growth. 8 Moreover, the geometry of the nanofibers mimics the fibrillar organization of the extracellular matrix that forms a network for neuronal attachment and axonal growth during central nervous system (CNS) development. Nanofibers electrospun from a variety of synthetic and naturally occurring polymers have thus generated tremendous interest due to their potential as scaffolds for regenerating tissue, 9 -12 with a recent study indicating that silk nanofibers improved bone regeneration in the rabbit. 13 Chemistry as well as geometry of the extracellular matrix is critical for proper neuronal function, necessitating that an optimized biomimetic surface for CNS repair incorporate both types of cues. Tenascin-C, a multidomain, multifunctional extracellular matrix glycoprotein with neuroregulatory actions, 14 -18 provides an example of a chemical cue that might enhance the function of a nanofibrillar scaffold. The alternatively spliced fibronectin type III region of human tenascin-C has demonstrated growth-promoting actions for a variety of neuronal types. 14, 16 The active site for neurite outgrowth in this region from cerebellar granule, cerebral cortical, spinal cord motor and dorsal root ganglion neurons was localized to a peptide with amino acid sequence VFDNFVLKIRDTKK, 18, 19 called the D5 peptide. Covalent modification of electrospun polyamide nanofibers with the D5 peptide promoted more in vivo-like growth patterns for neurons, with long, well-elaborated processes. 19 This in turn suggests that the peptide-modified nanofibers may find therapeutic applications in the repair of spinal cord injury and other CNS disorders.
Despite interest in using nanofibers in applications for spinal cord repair, 8 there has been little discussion about how this might actually be accomplished. Therefore, we turned our attention to this question. Given our initial success with the use of polyamide nanofibers as modifiable scaffolds for the growth of neurons in vitro, the hypothesis was explored that polyamide nanofibrillar matrices, alone or derivatized with the D5 peptide, can serve as biomimetic, regenerative implants following injury to the spinal cord.
Experimental Polyamide Nanofibers
Randomly oriented polyamide nanofibers (continuous fibers that collect as a nonwoven fabric) were electrospun by Donaldson Co., Inc. (Minneapolis, MN). They were electrospun from a blend of two polymers, (C 28 O 4 N 4 H 47 ) n and (C 27 O 4.4 N 4 H 50 ) n , either onto plastic (Aclar) coverslips for imaging studies or as a free fabric for spinal cord studies. The polymeric nanofiber mat was crosslinked in the presence of an acid catalyst and formed a network of filaments on average 180 nm in diameter interspersed with pores 100-800 nm in diameter. 20 In some cases the nanofibers were covalently coated with a proprietary polyamine polymer by Surmodics, Inc. (Eden Prairie, MN) to provide functional groups for further covalent modification of the nanofibers with bioactive peptides.
Nanofibers for the spinal cord studies were either unmodified or covalently modified with a neurite outgrowth-promoting peptide derived from tenascin-C, called the D5 peptide. (The complete covalent modification method is described in Ahmed et al. 19 ) The amino acid sequence of the D5 peptide is ADEGVFDNFVLKIRDTKKQ, where FD and FV are required for facilitation of neurite extension. 18 Atomic force microscopy Nanofibers (unmodified and peptide-modified) for the imaging studies were visualized using AFM as follows. Contact mode, friction force and tapping mode AFM were performed in ambient air using a Nanoscope IIIa MulitMode system from Veeco Instruments, Inc (Woodbury, NY). A J-scanner with a maximum 125 × 125 µm 2 x-y scan range was used, with 512 × 512 pixels recorded for each image. Silicon nitride tips with a nominal 20-25 nm tip radius of curvature were used for contact mode AFM, and etched silicon tips with a nominal 5-10 nm tip radius of curvature were used for tapping mode AFM. On each sample, a series of corresponding 100, 25 and 10 µm 2 areas were examined in contact mode AFM. On each sample a further series of corresponding 10, 5 and ca 2.5 µm 2 areas were examined in tapping mode AFM.
Surgical Procedure And Postoperative Care
To investigate whether polyamide nanofibers allow axonal regrowth in vivo following CNS injury, unmodified or peptide-modified nanofibers were incorporated into an over-hemisection spinal cord injury model. The subjects of this study included 5 animals with unmodified nanofibrillar implants, 5 animals with peptide-modified nanofibrillar implants and 5 injury-only control animals for the visualization of axons (3 weeks survival time); and 10 animals with unmodified nanofibrillar implants and 10 injury-only control animals for the visualization of cell bodies (3 and 5 weeks survival time; 5 animals for each group and each time point).
Adult female Sprague Dawley rats (250-260 g) were anesthetized using ketamine/xylazine (75 mg kg −1 + 10 mg kg −1 , intraperitoneal (IP)). They were also given buprenorphine (0.05 mg kg −1 , delivered subcutaneously) on a pre-emptive basis for postoperative pain. A laminectomy was made at the thoracic level 8 (T 8 ) spinal vertebrae. Iridectomy scissors were stereotaxically positioned at the dorsal midline of the spine and lowered to a depth of 2.0 mm. A piece of polyamide nanofiber fabric (ca 5 × 5 mm 2 ) comprised of randomly deposited fibers with a thickness of 2-4 µm 2 was gently crumpled with a pair of micro-forceps and carefully implanted into the lesion site. Control animals received no nanofibers. The 2-4 µm thickness of the fabric was chosen because thinner layers of nanofibers were very difficult to handle, while thicker layers of nanofibers lost their flexibility. The musculature was sutured and the skin was closed with surgical clips. All animal procedures were performed in strict accordance with institutional guidelines (Institutional Animal Care and Use Committee Approval no. PI04-063-06).
Animals were allowed to recover in clean cages atop heating pads. They were observed for infection or lethargy and were treated for 4 days postsurgery with buprenorphine analgesic (0.05 mg kg −1 , 2-3 times daily, delivered subcutaneously). Animals were also injected subcutaneously with enrofloxacin (Baytril) antibiotic for at least 3 days to manage bladder infections (5 mg kg −1 , 2 times daily) and with Ringer’s lactate solution (5 mL, 3 times daily, delivered subcutaneously) for 2-3 days (or until animals were observed to be drinking on their own) to manage dehydration. Bladders were evacuated 3 times daily until autonomic bladder function was restored (usually 4-7 days).
Following postoperative periods of 3 or 5 weeks, animals were transcardially perfused under deep terminal anesthesia (sodium pentobarbital, 100 mg kg −1 , IP) with 4% paraformaldehyde in 0.1 mol L −1 sodium phosphate buffer. Spinal cords were removed and sagittally sectioned on a cryostat at 20 µm. Sections were mounted on poly(L-lysine)-coated glass slides and stained for cell bodies using Nissl stain or immunolabeled for axons using monoclonal mouse antibody against neurofilament-M (Chemicon, Temecula, CA) (1:400 dilution overnight at room Figure 1 . Network of electrospun polyamide nanofibers. Images of nanofibers electrospun onto plastic coverslips. Contact mode AFM images of (A) unmodified nanofibers and (B) peptide-modified nanofibers revealed a fibrous network of filaments interspersed with pores. Line tracking for both images indicated only minor differences in how well the tip tracked the surfaces. Scale bar, 5 µm. Friction force AFM images of (C) unmodified nanofibers and (D) peptide-modified nanofibers revealed the effect of the different surface chemistries, indicated by the increased torsioning of the cantilever as the tip moved across the peptide-modified nanofibers. Scale bar, 5 µm. temperature) followed by a CY3-conjugated goat anti-mouse secondary antibody (Jackson ImmunoResearch, West Grove, PA) (1:500 dilution for 1 h at room temperature). At least 30-40 sections per spinal cord were examined. Slides were examined using a Zeiss Axioplan microscope equipped with an epifluorescence illuminator.
Atomic Force Microscopy
For each sample, a series of corresponding 100, 25 and 10 µm 2 areas were examined in contact mode AFM. Height mode and fiction force images were both investigated. 21 Height mode AFM measures the tip-sample interaction forces normal to the sample surface and provides a representation of the surface topography through Hooke’s law: F z = k z. Friction force AFM measures cantilever torsion and bending forces lateral to the tip motion across the sample surface and provides a representation of the surface friction through Amonton’s law:
F friction = µF loading .
Wide-area 100 × 100 µm scans (not shown) were first performed to assess the uniformity of the nanofiber mesh surfaces. Both unmodified and peptide-modified nanofiber meshes were uniform over 100 µm 2 areas. A corresponding series of 25 × 25 µm 2 , then 10 × 10 µm 2 images, taken from within the 100 µm 2 scans, were examined. The 25 × 25 µm 2 height and friction force images are shown in Fig. 1 . Trace/retrace scans included below Figs 1(A) and (B) indicated good tracking for both unmodified and peptide-modified nanofibers. Height mode AFM images of unmodified nanofibers ( Fig. 1(A) ) and peptide-modified nanofibers ( Fig. 1(B) ) revealed a fibrous network of filaments interspersed with pores. Friction force AFM images of unmodified nanofibers (Fig. 1(C) ) and peptide-modified nanofibers (Fig. 1(D) ) revealed the effect of the different surface chemistries, indicated by the increased torsioning of the cantilever as the tip moved across the peptide-modified nanofibers. The difference in surface friction may be due to both steric and chemical effects given that the peptide was covalently bound to the nanofiber at the N-terminus end, but was otherwise free to pivot and interact as the tip moved across the sample surface.
A further series of corresponding 10 µm 2 and 5 µm 2 areas, and a close-up image ca 2.5 µm 2 , were examined in tapping mode AFM. 22 In tapping mode AFM, a root mean square (RMS) height mode image is taken using an oscillating tip that lightly ‘taps’ the surface. Lateral friction forces are greatly reduced, as x-y motion occurs while the tip is raised. Furthermore, the phase difference between the natural frequency of the oscillating cantilever and the changed frequency that develops when the tip interacts with a sample surface provides a sensitive measure of the local environment.
Tapping mode AFM images (height and phase) of the unmodified nanofibers and peptide-modified nanofibers are shown in Fig. 2 . The difference in the nanofiber surfaces was most apparent in the phase images (Figs 2(B) and (D)). The roughness of the peptide-modified nanofibers (Fig. 2(D) ) in comparison to that of the unmodified nanofibers (Fig. 2(B) ) resulted from factors that affect the RMS oscillation of the cantilever and may be due to both topographical surface roughness and to surface chemistry. Similar effects have been observed by Karakecili et al. 23 for unmodified versus RGDmodified polycaprolactone membranes.
Biocompatibility Of Polyamide Nanofibers
To evaluate the biocompatibility of polyamide nanofibers and their potential as a regenerative matrix in injured spinal cord, a piece of nanofibrillar fabric comprised of randomly oriented nanofibers was implanted into an over-hemisection wound as described above. The fabric was well integrated into the host tissue 3 and 5 weeks after injury, with no evidence of cavitation (illustrated by Nissl staining in Fig. 3(A) for the 5 week time point), whereas a large cavity was present in injury-only, control animals ( Fig. 3(B) ). Imaging of sequential sections showed that the lesion site was completely filled with nanofibers. (Compare Figs 5(B) , 5(C) and 6(A), all sections from the same animal.)
Axonal Growth On Unmodified And Peptide-Modified Nanofibers
Axons were visualized 3 weeks after injury utilizing an antibody against neurofilament-M and epi-fluorescent microscopy. No axons were seen in the injury site in control animals (injury only, no nanofibers) ( Fig. 4(A) ), whereas they extended onto the nanofibrillar fabric (arrows), detectable in the injury site as a consequence of its low level of autofluorescence ( Fig. 4(B) ). No labeling was detected when animals were sacrificed immediately after implantation of nanofibers, indicating that the observed axons represented regrowth, as opposed to axons spared in the lesion process ( Fig. 4(C) ).
Covalent modification of the nanofibers with an extended version of the D5 peptide, called the D5 peptide (amino acid sequence ADEGVFD-NFVLKIRDTKKQ), enhanced their ability to facilitate neuronal process extension in vitro 19 four additional amino acids at its N-terminus to allow for better presentation of the FD/FV active site to neurons. 19 ) However, despite this encouraging result, this initial method of introducing nanofibers into the damaged spinal cord has proven to be unsatisfactory. The fabric itself was randomly folded during implantation, and regenerating axons appeared to follow along with the surface contours of the random folds, with forward motion at times restricted (Figs 5(B) and (C)) On the other hand, when the folds in the nanofibrillar fabric by chance fell parallel to the spinal cord axis, axons also grew with a parallel organization. This is illustrated in Figs 6(A) and (B), which focus in on the nanofibrillar implant within the lesion site. Therefore, these results suggest a necessity for engineering a nanofibrillar device with an ordered structure to appropriately and reproducibly direct axonal growth.
This work explores the utility of polyamide nanofibers as implantable substrates for the regeneration of axons in the lesioned spinal cord. Significantly, the polyamide composition of the nanofibers (proprietary, Donaldson Company, Inc., Minneapolis, MN) is not rapidly degradable in vivo (S. Meiners, personal observation) and has no apparent neurotoxicity, 19 and the nanofibers maintain their structural integrity in vivo for several weeks. It seems likely that the formation and then maintenance of the reformed neuronal circuitry within the spine might be best facilitated and maintained by scaffolds that only slowly degrade within the body. Moreover, breakdown of rapidly biodegradable materials such as polyglycolate or polylactate can result in the release of monomers (e.g. lactate) that can significantly lower the local pH 24 and negatively impact the viability of regenerating neurons. In addition, a recent study has highlighted the utility of non-or slowly degrading materials in general, and polyamide in particular, in the promotion of axonal regrowth in the peripheral nervous system (PNS). 25 Third, thin layers of polyamide nonwoven fabrics are extremely flexible, 26 allowing for excellent incorporation into the damaged spinal cord.
Because of the problems encountered with randomly folded polyamide nanofiber implants illustrated above, the nanofibrillar fabric is instead envisioned fashioned into a multilayered tubular construct, reminiscent of the bilayered tubular construct fabricated by Kidoaki et al. 9 as a prototype scaffold for implantation into arteries. A view of one type of multilayered polyamide nanofibers (engineered by Donaldson Co., Inc., using proprietary technology) which could potentially be employed is shown in Fig. 6(C) . In this example, each layer (ca 0.5 µm thick) was separated by about 10 µm from the next. However, nanofibrillar layers will need to be designed that are separated by 50-200 µm to allow more room for cellular growth. 27, 28 Moreover, parameters can be adjusted to yield a device with desired mechanical properties.
For experimental testing, the tubular construct will ideally be implanted into a completely transected or near-completely transected spinal cord instead of the more modest over-hemisected spinal cord used in our studies, since very little sparing in the lateral corticospinal tract can lead to large effects on spared locomotion. 29 This can in turn lead to difficulty in behavioral testing to assess the effectiveness of the implant in promoting functional recovery. If nanofibers are to be used in actual human medicine, layers of nanofibers are envisioned as cut and stacked to fit the size of the injury following, if necessary, trimming away of the scar tissue, which impedes axonal elongation. In the experimental model, the diameter of the multilayered tube will need to be fashioned to match the diameter of the transected spinal cord with the orientation of the layers parallel to the spinal cord axis (Fig. 7) , allowing for longitudinally oriented axon growth. However, since neurite extension on randomly deposited nanofibers is itself random, 8, 19 this design is not likely to limit lateral or dorsal-ventral axonal growth.
To overcome this problem, the nanofibrillar fabric should incorporate parallel, aligned nanofibers. Aligned nanofibers provide contact guidance cues to growing neurites, with neurite outgrowth exactly parallel to the nanofiber axis. 8 Thus aligned nanofibers will potentially allow directed growth from one end of the lesion to the other. However, a possible disadvantage of aligned nanofibers for tissue engineering in the spinal cord is their considerable rigidity in comparison to randomly deposited nanofibers, 30 with the potential to cause further trauma during implantation. Therefore, an aligned nanofibrillar fabric is pictured that consists of a very thin layer of nanofibers (ca 1 nanofiber thick) deposited in an aligned array on either side of a core of randomly deposited nanofibers. The aligned nanofibers should ideally be spaced ca 0.5-1 µm apart to maintain porosity. When formed into a multilayered tube and implanted such that the aligned fibers are parallel to the spinal cord, this nanofiber fabric will restrict longitudinal, lateral and dorsal-ventral growth of axons, while still retaining flexibility.
Of course, further experiments are required to test the hypothesis that polyamide nanofibers or nanofibers of other compositions will be superior materials for spinal cord repair. For example, tract tracing studies must be done to evaluate whether axons in fact leave the implant, a necessity for the formation of functional connections. Furthermore, one must consider the inflammatory response and the scarring response. Injury to the CNS triggers an inflammatory reaction with potentially devastating consequences. These include the production of proinflammatory cytokines by recruited inflammatory cells and activated endogenous CNS cells, 31 -33 formation of a glial/fibrotic scar that impedes axonal regeneration, 34 -36 axon dieback, 37 death of neurons and oligodendrocytes and production of inhibitory myelin products 38, 39 and formation of cystic cavities. 40 Many biomaterials provoke a considerable foreign body response, with heightened inflammation and encapsulation of the implant by scar tissue 41 and ultimate rejection of the implant. As such, these biomaterials are unsuitable for tissue engineering.
However, polyamide nanofibers do not appear to provoke a considerable tissue response in the spinal cord as evidenced by low levels of scarring three and five weeks after injury in comparison to injury-only controls. Early experiments suggest that they are associated with reduced levels of glial fibrillary acidic protein, a marker for reactive astrocytes, and chondroitin sulfate proteoglycans, a group of molecules expressed by both glial and fibroblastic components of the scar and largely believed to be inhibitory to axonal regrowth 34, 36 (not shown). Why this should be the case is at this point unclear. Attenuation of scarring may be due to the nanofibrillar nanostructure and porosity, which closely resemble features of naturally occurring extracellular matrix/basement membrane scaffolds and may therefore cause the nanofibers to look like ‘self’, or to the nanofibrillar chemistry, or to both.
Electrospun nanofibers have recently demonstrated widespread application in tissue culture due to their biomimetic properties and their structural resemblance to the extracellular matrix. The success of nanofibers for the culture of cells has generated great enthusiasm for their potential use in tissue engineering and regenerative medicine. However, to date very few in vivo applications of nanofiber technology have in fact been demonstrated. It is becoming increasingly clear that nanofibers will have to be electrospun into specialized scaffolds and structures in order to allow tissue regeneration within the body under conditions that allow physiologically relevant tissue formation. This is most obvious in the spinal cord, where axons follow a strict topography, much of which needs to be recreated in order to permit reconnection of appropriate circuits and maximal functional recovery following injury. Such structures must also incorporate within their design the ability to be ‘stealthy’ and not elicit the foreign body response or recognition by the immune system.
Polym Int 56:1340-1348 (2007) DOI: 10.1002/pi