3D-engineering of Cellularized Conduits for Peripheral Nerve Regeneration

Tissue engineered conduits have great promise for bridging peripheral nerve defects by providing physical guiding and biological cues. A flexible method for integrating support cells into a conduit with desired architectures is wanted. Here, a 3D-printing technology is adopted to prepare a bio-conduit with designer structures for peripheral nerve regeneration. This bio-conduit is consisted of a cryopolymerized gelatin methacryloyl (cryoGelMA) gel cellularized with adipose-derived stem cells (ASCs). By modeling using 3D-printed “lock and key” moulds, the cryoGelMA gel is structured into conduits with different geometries, such as the designed multichannel or bifurcating and the personalized structures. The cryoGelMA conduit is degradable and could be completely degraded in 2-4 months in vivo. The cryoGelMA scaffold supports the attachment, proliferation and survival of the seeded ASCs, and up-regulates the expression of their neurotrophic factors mRNA in vitro. After implanted in a rat model, the bio-conduit is capable of supporting the re-innervation across a 10 mm sciatic nerve gap, with results close to that of the autografts in terms of functional and histological assessments. The study describes an indirect 3D-printing technology for fabricating cellularized designer conduits for peripheral nerve regeneration, and could lead to the development of future nerve bio-conduits for clinical use.

biomedical applications, such as engineering of bone, cartilage, and vascular tissues 10,13 . The gelatin derivative is an inherently cell-adhesive material comprised of modified natural extracellular matrix (ECM) components that can provide functional cues to the resident cells within the scaffolds and thus can aid in nerve regeneration 14,15 . Moreover, cell delivery can be facilitated by priming the support cells in vitro before transplanted into the injury site, which reduces the rate of cell loss and leakage to surrounding tissues 11,15 .
Peripheral nerve injury may involve one or more nerves with different lengths and injury degrees (ranging from neurapraxia to neurotmesis) 2, 16 . In addition, the injured nervesmay be varied in geometries 17 . The inherent variance in patient anatomies and injury profiles have motivated the development of personalized treatment for peripheral nerve injury 18 . However, challenges remain for conventional manufacturing methodologies to fabricate nerve conduits with complex or custom geometries, because the conduits are mainly manufactured around cylindrical substrates. The acquired nerve conduits with simpler architectures may not be sufficient to promote structural and functional nerve regeneration 14,18,19 . Recent advances in materials and manufacturing enable the fabrication of complex or patient-specific devices 20 . Three-dimensional (3D) printing, an additive manufacturing technology, is a powerful tool for fabricating 3D constructs with intricate geometries under computer-aided design/computer aided manufacture (CAD/CAM) system control 21 . The technique currently has been applied to manufacture customized medical devices, such as amputee prosthetics, airway splints, and scaffolds for tissue engineering 22 . However, the current 3D-printed devices are often made from synthetic materials that lack of biological functionality for cell delivery 23 .
The aim of this study was to establish a novel manufacturing method for the fabrication of nerve conduits with designer geometries for peripheral nerve regeneration, using cell-adhesive gelatin cryogels that prepared by cryopolymerization of GelMA (cryoGelMA). CryoGelMA NGCs with complex or custom architectures (multichannel, bifurcating and anatomically accurate) could be manufactured using a commercially available inexpensive desktop 3D printer. Then the degradability and biocompatibility of the cryoGelMA NGCs were evaluated. The introduction of support cells, an important biological cue, into nerve conduits holds great promise for nerve repair by providing a beneficial local microenvironment for the regenerating axons 24,25 . Adipose-derived stem cells (ASCs) have attracted considerable attention for nerve tissue engineering, because they are multipotent, and can be easily obtained via minimally invasive techniques and be rapidly expanded in vitro 26,27 . Thus the cryo-GelMA NGCs were loaded with ASCs to repair a 10 mm nerve defects in rats, and the ability of promoting nerve regeneration was assessed by walking track analysis, electrophysiological assessment, and histological examination. Our strategy to construct a nerve guidance bio-conduit with designer geometries for peripheral nerve regeneration is outlined in Fig. 1.

Results
Fabrication of cryoGelMA NGCs. Moulds with a "lock and key" structure were fabricated by a 3D printing technique, as shown in Fig. 1. Assisted by the 3D-printed moulds, cryoGelMA NGCs with intricate architectures could be obtained. A multichannel NGC that mimics nerve fascicles and a bifurcating NGC that mimics nerve plexus were successfully constructed ( Fig. 2A,B). The appearance of NGCs is white and practically in accordance with the original CAD models. To create a patient-specific NGC that was anatomically accurate, we integrated the indirect 3D printing technique with magnetic resonance (MR) neurography and successfully fabricated a Moulds with a "lock and key" structure were fabricated by an indirect 3D printing technique. Then a cryoGelMA NGC was fabricated and seeded with ASCs to bridge a 10 mm sciatic nerve defect.
Scientific RepoRts | 6:32184 | DOI: 10.1038/srep32184 personalized NGC with architectures matching the sciatic nerve of a patient (Fig. 2C). The 3D-printed moulds for molding geometrically complex and patient-specific NGCs are shown in Figure S1A-C.
A cryoGelMA NGC based on the general anatomical features of a rat sciatic nerve was fabricated to test its degradability, biocompatibility, and efficacy in promoting nerve regeneration. A rat sciatic nerve was transected from a Sprague-Dawley (SD) rat weighted 220-250 g to provide a tissue template. A NGC with inner diameter of 1.5 mm, outer diameter of 4.0 mm, and length of 15 mm was fabricated (Fig. 3A). The 3D-printed moulds for molding the NGC are available in Figure S1D. SEM revealed that the NGC had highly porous microstructure that was essential for cell attachment, proliferation, and survival (Fig. 3B). The wall thickness of the fabricated conduits matched with the original CAD model, but the inner diameter was approximately 0.1 mm smaller than the CAD model. Degradability of cryoGelMA NGCs. Non-biodegradable conduits are associated with several disadvantages, such as chronic inflammation and a secondary surgery for conduits removal 9 . Thus, biodegradable materials are advocated for the preparation of NGCs. Firstly, we tested whether the cryoGelMA NGCs could be enzymatically degraded in vitro. As shown in Fig. 3C, the NGCs could be degraded completely in the presence of 1 mg/ml collagenase type II within 20 h. Then the NGCs were implanted subcutaneously on the dorsal site of rats to evaluate the degradability in vivo (Fig. 3D). The NGCs degraded throughout the implantation period, and remained structurally stable for 2 weeks after implantation. The NGCs were collapsed at 4 th week, and did not completely degrade in 2 months. The infiltration of fibroblasts and inflammation cells were evaluated by H&E staining (Fig. 3E). Rare fibroblasts and inflammatory cells could be seen in the NGCs within 2 weeks, suggesting the NGCs prevented the fibrous tissue ingrowth effectively and provoked mild acute inflammatory responses. At 4 th and 8 th weeks, the NGCs elicited a foreign body reaction, and fibrous tissue with dispersing neocapillaries could be observed in the NGCs.
In vitro cell compatibility of cryoGelMA NGCs. The suitability of the cryoGelMA NGCs as a substrate for cell attachment and growth was further investigated. Rat ASCs were found distributed throughout the NGCs with rare dead cells as revealed by the live/dead staining (Fig. 4A). The ASCs attached to the surface of NGCs and maintained their fibroblast-like morphology after 2 days of culture (Fig. 4B,D). The deposition of ECMs by the ASCs was more apparent on the NGCs than that on the tissue culture polystyrene (TCP) (Fig. 4C,D). AlamarBlue assay was carried out to evaluate the cell viability, and a lower proliferation rate was observed on the NGCs (Fig. 4E). The gene expression of neurotrophic factors of the ASCs cultured on the NGCs and TCP was further tested. As shown in Fig. 4F, the expression of BDNF and GDNF was significantly up-regulated after cultured on the NGCs (p < 0.05).
In vivo nerve regeneration. The rat sciatic nerve transection model has been commonly used for the evaluation of tissue engineered NGCs in promoting peripheral nerve regeneration in vivo 28 . The cryoGelMA NGCs were implanted into the injured sciatic nerve to evaluate the nerve regeneration behaviors through the prepared four groups (sham surgery, autograft, NGCs, NGCs + ASCs). All rats were in good health condition throughout the experiment. To assess the functional recovery of the sciatic nerve, walking track analysis was performed on all operated animals at 2 nd , 4 th , 8 th and 16 th weeks post-surgery (Fig. 5). Sham surgery did not affect the SFI value significantly. The autograft group showed a significantly higher SFI value than the NGCs group at 8 th and 16 th weeks (p < 0.05), but there was no significant difference between the autograft group and the NGCs + ASCs group. Though repair with NGCs had a lower SFI value than that of NGCs+ ASCs, no significant difference was observed.
Electrophysiological analysis was performed at 4 th , 8 th , and 16 th weeks postoperatively (Fig. 6). CMAP has been commonly used to determine the numbers of regenerated motor nerve fibers 29 . The action potentials were observed in all groups at 4 th week and still intensive at 8 th and 16 th weeks after surgery, indicating a gradual enhancement of functional recovery for the injured nerves 30,31 . In consistence with the SFI findings, the CMAP in the NGCs + ASCs group was close to the autograft group, and there was no significant difference between the NGCs + ASCs and the NGCs group. NCV is an objective index that offers important insight for evaluating the conduction of action potentials in peripheral nerve 31 . The NCV value of the NGCs group was significantly lower than that of the autograft and NGCs + ASCs groups at 4 th and 8 th weeks after implantation (p < 0.05), but no statistically significant difference was observed at 16 weeks among the three groups. Latency of CMAP onset was decreased over time for the autograft, NGCs, and NGCs + ASCs groups, but no significant difference was observed among the three groups.
Macroscopically, the cryoGelMA NGCs were all degraded completely, and the proximal and distal stumps were successfully reconnected by the regenerated nerve at 16 th week after surgery (Fig. 7). H&E staining showed the general regenerated nerve fibers morphology in the distal segment at 16 th week after implantation (Fig. 8A). The axon diameter and myelin sheath thickness of the distal segments of regenerated nerves were then evaluated by toluidine blue staining and transmission electron microscope (TEM) (Fig. 8B-E). The sham surgery group has the largest axon diameter and thickest myelin sheath. The axon diameter and myelin sheath thickness of the NGCs + ASCs group were significant longer than that of the NGCs group (p < 0.05). Although myelin sheath in the NGCs + ASCs group was thinner than that of the autograft group (p < 0.05), there was no significant difference in the axon diameter.
After sciatic nerve transection, the gastrocnemius muscles demonstrated degradations of muscle fibers and became fragments, and then regained nerve re-innervation 32 . H&E staining was performed to evaluate the recovery of nerve function in all groups (Fig. 9A-D). Compared to normal muscle morphology in the sham surgery group, gastrocnemius muscles were degenerated in the autograft and NGCs + ASCs groups, and prominently in the NGCs group. The muscle fiber diameter in the sham surgery, autograft, NGCs, and NGCs + ASCs groups was 43.0 ± 8.3 μ m, 37.8 ± 8.6 μ m, 33.9 ± 4.5 μ m, and 36.0 ± 5.5 μ m, respectively (Fig. 9E). The relative gastrocnemius muscle weight increase was similar to the increase in muscle fiber diameter in the four groups (Fig. 9F). The muscle fiber diameter and weight observed in the NGCs + ASCs group significantly increased when compared with that of the NGCs group.

Discussion
In this work, we demonstrated a flexible method for the fabrication of cryoGelMA NGCs with desired geometries by a low cost desktop 3D printer. At present, the desktop 3D printer is being widely used in education, manufacturing, and industry for its low cost and versatility in fabrication 33 . With the help of the desktop 3D printer, rationally designed "lock and key" moulds were printed to fabricate NGCs with complex or personalized geometries for peripheral nerve regeneration. Using the "lock and key" moulds, the prepared NGCs could be acquired without dissolving the negative moulds by organic solvent, thus reducing the manufacturing time and  avoiding organic solvent influences the internal structural characteristics of the conduits, such as pore size 34,35 . Conventionally, the NGCs are simple cylindrical structures, and it's difficult and time-consuming to fabricate NGCs with complex architectures 18,19,36 . The fabrication of NGCs with advanced structures, such as multichannel, has been reported in previous studies using a wire mesh method [36][37][38][39] . However, the conventional method is infeasible to fabricate NGCs with versatile structure parameters, such as the internal diameter or wall thickness, that might affect the efficacy of nerve regeneration 5,6,40 . Take advantages of the indirect 3D printing technique, we can reliably fabricate NGCs with advanced structures and control the structure parameters. Our technique has advantages in terms of simplicity, flexibility and low cost in fabricating designer NGCs with complex or personalized geometries for peripheral nerve regeneration, which may lead to the development of future NGCs for clinical use.
Recently, microstereolithography was used for patterning NGCs with intricate geometries, and a 3D printing methodology was also applied to construct anatomical nerve regeneration pathways based on 3D scanning of a rat bifurcating nerve 14,18,19 . However, to our knowledge, the fabrication of a NGC based on patient-specific anatomy has never been explored. MR neurography has been increasingly used in recent years to evaluate abnormalities such as nerve injury, entrapment, and neoplasm 41,42 . The complex 3D anatomy of peripheral nerve is resolvable with multiplanar reconstruction, which provides prudently information of the length of injured nerve and the geometries of the proximal and distal discontinuous nerve stumps 2 . In this work, we performed an attempt to integrated 3D printing methodology with MR neurography to engineer a NGC based on the anatomic geometries of a patient without any aggressive manipulations, which allows the customization of a NGC that precisely match a particular nerve defects of a patient 18 .
Both in vitro and in vivo degradation studies show that the cryoGelMA NGCs were degradable. The NGCs have a proper degradation rate (completely degraded at 2-4 months), which is similar to several FDA approved NGCs, such as Neurotube ® and AxoGuard TM Nerve Connector 6 . We found that ASCs cultured on the NGCs gave rise to more ECMs accumulation than that on the TCP. Furthermore, the expression of several neurotrophic factors mRNAs, such as BDNF, was unregulated after seeding the ASCs on the NGCs. The enhancement of cell-cell and cell-matrix interactions on the NGCs provides a more physiologically relevant environment to the ASCs, and thus enhances their functions 43 . Those results suggested that the introduction of ASCs into cryoGelMA NGCs may provide a favorable repair-conducive environment for nerve regeneration 26 . In vivo experiments also confirmed a benefit of applying the ASCs in promoting nerve regeneration after transplanted into the injured site. Several previous studies have demonstrated the utilization of ASCs improved axonal regeneration in vivo, and this effect has been attributed mainly to the environmental support provided by the ASCs during nerve regeneration [44][45][46][47] .
Walking track assessment and electrophysiological analysis are widely accepted techniques for the functional evaluation of sciatic nerve repair in rats 24,32 . We found that the cryoGelMA NGCs cellularized with ASCs showed similar results to the autograft group in function recovery and axonal regeneration as evidenced by no statistical difference in SFI, electrophysiological results, and nerve and muscle fiber diameters between the two groups. Although we did not observe significant beneficial effects on gait and electrophysiological analysis in ASCs-seeded NGCs compared to bare conduits, the significantly increased axonal regeneration and nerve re-innervation of gastrocnemius muscle suggested the application of ASCs promoted nerve regeneration in histomorphological parameters. The discrepancy between the outcomes concerning motion analysis and tissue structures is in line with several previous study 32,45 . A higher dose of ASCs may show improved functional recovery as a relatively low dose of ASCs (1 × 10 6 ) was used in our study 46,48 .

Conclusion
We described an indirect 3D-printing technology for fabricating cellularized designer cryoGelMA NGCs for peripheral nerve regeneration. By molding with 3D-printed "lock and key" moulds, NGCs with desired geometry, such as multichannel, bifurcating and the personalized structures, could be obtained by a low cost desktop 3D printer. The ASCs-cellularized NGCs were comparable with that of the autografts in repairing a peripheral nerve defect, showing potential clinical application in peripheral nerve regeneration.

Materials and Methods
Synthesis of GelMA. GelMA was synthesized following the procedures described elsewhere 13 . Briefly, gelatin type A (300 bloom from porcine skin, Sigma) at 10% (w/v) was dissolved into stirred Dulbecco's phosphate buffered saline (DPBS) at 60 °C. Methacrylation of gelatin was achieved by adding 20% (v/v) of methacrylic anhydride (Sigma) at a rate of 0.5 mL/min and reacting at 50 °C for 1 h. Following a 5× dilution with warm DPBS (40 °C), the mixture was dialyzed against distilled water at 40 °C for 2 weeks. The sample was then lyophilized and stored at − 20 °C until further use. CryoGelMA NGCs fabrication. CryoGelMA NGCs with desired structures, such as multichannel and bifurcating, were prepared by molding based on "lock and key" moulds that were designed via SolidWork version 2012 (SolidWorks Corp., Waltham, Massachusetts, USA) and manufactured by a commercially available desktop ink-jet printer (TD-IIA, TD ARTIST, Chengdu, China). To fabricate a cryoGelMA NGC based on anatomic geometry of a patient, the shape of sciatic nerve was isolated from the MR neurography by a commercially available software (Mimics, Materialise). Based on the image data, "lock and key" moulds were designed and converted into an STL file followed by 3D printing 49 . To further investigate the degradability, biocompatibility, and efficacy of nerve regeneration of the cryoGelMA NGCs, a 10 mm right sciatic nerve was acquired from the mid-thigh level of SD rats (220-250 g). The diameters of proximal and distal stumps of the injured nerve were measured for designing moulds.
GelMA prepolymer solution was formed by dissolving 5% (w/v) GelMA in dH2O and maintained at 60 °C to dissolve adequately. After incubated on ice for 5 min, 0.5% (w/v) ammonium persulfate (APS; Sigma) and 0.1% (w/v) tetramethylethylenediamine (TEMED; Sigma) were added to the prepolymer solution. The solution was then pipetted into the 3D-printed moulds cavity and underwent cryopolymerization for 24 h in a − 20 °C refrigerator. The resulting cryogels were hydrated and harvested, and washed by deionized water extensively. After immersed in 75% ethyl alcohol for one hour, the gels were washed and collected into cell culture dish, lyophilized overnight (Boyikang), and stored in − 20 °C prior to use.

SEM.
The morphology of the cryoGelMA NGCs were observed by using scanning electron microscopy (SEM, Hitachi S4800) operated at 5.0 kV. For this observation, the lyophilized conduits were mounted on aluminum stubs and gold-coated for 90 s before SEM imaging.
Degradability. CryoGelMA NGCs were incubated with 1 mg/ml collagenase type II (Sigma) in DPBS and were placed on an orbital shaker at 100 rpm and 37 °C. At predefined time points (t = 4, 8, 12, 16, and 20 hours), the conduits were harvested. The degradation of the NGCs was calculated as a ratio of loss of dried weight to the primary weight. For in vivo degradation study, the conduits were implanted subcutaneously on the back of SD rats. At predefined time points (1, 2, 4 and 8 weeks) after implantation, the implants were photographed and harvested together with surrounding tissue for histological evaluation. We confirm that all experiments were performed in accordance with the guidelines and regulations of Sichuan University Committee on Animal Research and Ethics and were approved by the Institutional Animal Care and Use Committee of West China Hospital of Sichuan University.
Cell adhesion and proliferation. Rat ASCs were isolated from subcutaneous adipose tissue of inguinal region of SD rats. ASCs were then maintained in Dulbeccos' modified Eagles' medium (DMEM, Invitrogen) with low glucose, containing 10% (v/v) fetal bovine serum (FBS, Invitrogen), 100 U/ml penicillin, 100 μ g/ml streptomycin. Cells were trypsinized upon 80% confluency and each conduit was seeded with 1 × 10 5 cells. 2 days after seeding, Live/Dead assay (Invitrogen) was performed for visualization of cell viability using a Zeiss confocal microscope (Zeiss). Moreover, F-actin was stained using rhodamine phalloidin (Cytoskeleton) and cell nucleus was stained using DAPI (Invitrogen), and images were taken on the confocal microscope. The morphology of ASCs on the cryoGelMA NGCs was visualized using SEM. After fixed with 3% glutaraldhyde for 2 h, specimens were rinsed with PBS and dehydrated with graded concentrations (30,50,70,90, 100% v/v) of ethanol. Then the samples were coated with gold and observed with SEM.
Cell proliferation was determined by AlamarBlue assay (Invitrogen). Briefly, 5 × 10 4 cells in 50 μ l were seeded on each conduit and incubated for 1.5 h to allow for cell attachment. Then 1 ml fresh medium was added and the composites were cultured in 24-well plates for 24, 48 and 72 h. Meanwhile, ASCs of same density were seeded on the TCP as the control. After each time point of cell seeding, the culture medium was removed and 1 ml of 10% (v/v) Alamar blue in culture solution was added. After 3 h of incubation, aliquots were pipette into a 96-well plate and the absorbance at 570 nm and 595 nm was measured. Cell numbers were determined by a standard curve.

RT-PCR.
Total RNA was extracted using the Trizol ® reagent (Invitrogen) form ASCs cultured on conduits and TCP for 2 days. RNA strand was reverse transcribed into cDNA and amplified using PrimeScript II 1st Strand cDNA Synthesis Kit (TaKaRa). Polymerase chain reactions (PCRs) were performed using primers for glyceraldehyde 3-phosphate dehydrogenase (GAPDH), nerve growth factor (NGF), brain-derived neurotrophic factor (BDNF), and glial cell-derived neurotrophic factor (GDNF) as previously described 50 . PCRs were carried out with SYBR Premix Ex Taq (TaKaRa). The resulting amplification was monitored with the CFX96 Real-Time System (Bio-Rad). The expression levels were normalized against the reference gene GAPDH, and the relative gene expression was analyzed. The primer sequences for each gene used in this study are shown in Table S1.
Animal and surgical procedure. SD rats weighing 220-250 g were used and randomly divided into four experimental groups: sham surgery group (n = 6), autograft group (n = 12), NGCs group (n = 12) and NGCs + ASCs group (n = 12). 1 × 10 6 ASCs at early passages (P3-P6) were seeded on each conduit and cultured for 3 h at 37 °C to allow cells fully attached to the cryoGelMA scaffolds. After anesthetized by intraperitoneal chloral hydrate, the hair on the right femur was removed. Under aseptic conditions, the right sciatic nerve was exposed, and a 10 mm segment of the sciatic nerve was removed at the mid-thigh level. The NGCs were interposed between the proximal and distal nerve stumps. The nerve stumps were inserted into the NGCs to a depth of 2.5 mm, and fixed with 8-0 absorbable vicryl sutures. For autograft, the 10 mm transected nerve was re-implanted under microscope. Following the implantation, the muscle incision was closed using 5-0 vicryl sutures and the skin was closed with 2-0 silk sutures. Postoperatively, animals were free access to food and water, and housed in a controlled room with 12 h light cycles.
Walking track analysis. The motor functional recovery was evaluated by walking track analysis at 2, 4, 8 and 16 weeks post-operatively. Pre-operatively, all rats were underwent conditioning training in a 50 cm × 8 cm wooden track. After the hind feet dipped in black ink, rats were allowed to walk down the track. Five measurable footprints were collected for each rat. Sciatic Function Index (SFI) was calculated by the formula proposed by Bain et al. 51  The print length (PL) is the distance from the heel to the third toe, the toe spread (TS) is the distance from the first to the fifth toe, and the intermediary toe spread (IT) is the distance from the second to the fourth toe. EPL, ETS, and EIT represent recordings from the operated foot, as well as NPL, NTS and NIT from the non-operated foot. A value of − 100 implies total impairment. Electrophysiological analysis. Electrophysiological tests were performed using a previous developed method at 4, 8 and 16 weeks after implantation 30 . An electromyograph machine (Nuocheng, Shanghai, China) was used to measure all the experimental animals. After the sciatic nerve was exposed under anesthetization, a stimulating electrode was placed at the proximal side of regenerated nerve, and a recording electrode was inserted into the gastrocnemius muscle. The reference electrodes were positioned between the stimulating electrode and the recording electrode. The ground electrode was placed in the tail. For quantitative analysis, the peak amplitude of compound muscle action potential (CMAP), nerve conduction velocity (NCV), and latency of CMAP onset values were calculated respectively.
Histological assessment. The distal nerve segments were harvested immediately after electrophysiological examination at 16 weeks after implantation. The paraformaldehyde-fixed nerves were dehydrated, embedded in olefin, cut into 5 μ m-thick slices, and stained with hematoxylin/eosin (H&E). The other samples were embedded in Epon 812 epoxy resin and stained with toluidine blue after cut into semi-thin sections (0.5 μ m). The images of histological sections were captured and analyzed using a digital image analysis system (Nikon E600 Microscope with a Nikon Digital Camera DXM 1200, Nikon Corporation, Japan). To observe the ultrastructure of myelin sheath, ultrathin sections (50 nm) were viewed and photographed with a Hitachi H7650 TEM (Tokyo, Japan). The diameters of myelinated axons and thickness of myelin sheath were quantified from TEM images using the Image J software. For each specimen, a total of 50-60 random axons were analyzed. The gastrocnemius of both limbs was harvested from the bone attachments at 16 weeks after surgery. The moist weights of the gastrocnemius muscle were weighed on an analytic scale, and the ratio of wet weight on the experimental to contralateral sides in each group was calculated to evaluate the target muscle reinnervation. Then the middle portion of the muscle was dissected and placed in 4% paraformaldehyde for H&E staining. The diameters of muscle fibers were calculated from 5 random fields using the Image J software.

Statistical analysis.
Values are expressed as means ± standard deviation. Significant differences among groups were analyzed by single-factor analysis of variance (ANOVA) followed by Bonferroni's post-hoc test using SPSS 16.00 software. Statistically significant differences between medians were determined with a Mann-Whitney U test. A value of p < 0.05 was considered statistically significant.