A Novel 3D‐Printed/Porous Conduit with Tunable Properties to Enhance Nerve Regeneration Over the Limiting Gap Length

Engineered grafts constitute an alternative to autologous transplant for repairing severe peripheral nerve injuries. However, current clinically available solutions have substantial limitations and are not suited for the repair of long nerve defects. A novel design of nerve conduit is presented here, which consists of a chitosan porous matrix embedding a 3D‐printed poly‐ε‐caprolactone mesh. These materials are selected due to their high biocompatibility, safe degradability, and ability to support the nerve regeneration process. The proposed design allows high control over geometrical features, pores morphology, compression resistance, and bending stiffness, yielding tunable and easy‐to‐manipulate grafts. The conduits are tested in chronic animal experiments, aiming to repair a 15‐mm long gap in the sciatic nerve of rats, and the results are compared with an autograft. Electrophysiological and nociception tests performed monthly during a 4‐month follow‐up show that these conduits allow a good degree of muscle functional recovery. Histological analyses show abundant cellularization in the wall and in the lumen of the conduits and regenerated axons within all rats treated with these grafts. It is suggested that the proposed conduits have the potential to repair nerves over the limiting gap length and can be proposed as strategy to overcome the limitations of autograft.


Introduction
Peripheral nerve injuries represent a relevant clinical issue with an incidence of 14 cases per 100 000 inhabitants per year in Europe. [1] Depending on the severity of the lesion, peripheral nerve injuries result in a partial or total loss of motor, sensory, and autonomic functions of the area innervated by the injured nerve, decreasing the quality of life of the affected patients. Neurotmesis is the most severe injury, and it is featured by a complete loss of axons continuity with the formation of a proximal and a distal nerve stump separated by a gap. Epineural suturing of the two stumps is the classical method of repair. When the length of the gap is too long, autograft (AG) is the gold standard repair, since the presence of autologous Schwann cells (SCs) and the same connective tissue of the native tissue support nerve regeneration with remarkable performances. [2] However, this technique has several drawbacks such as impairment of sensation in the area where the autograft is removed, possible neuroma formation, and limited number of donor nerves. [3] In the last decades, tissue engineering technologies helped to develop alternatives to AG through the development of tubular scaffolds named nerve conduits (NCs). NCs aim to provide a favorable environment for nerve regeneration and hold the two nerve stumps coaxially at a certain distance between them, leaving an intraluminal gap that depends on the entity of the nerve damage. [4] Tubulization repair aims to isolate growing axons from the fibrotic tissue, thus protecting the regenerating nerve from mechanical stresses, guiding it longitudinally and condensing the growth factors secreted by migrating SCs from the two nerve stumps. [5,6] For this purpose, the conduits must have certain characteristics such as biocompatibility, biodegradability, possibility to adapt their size and diameter to the nerve ones, sufficient wall strength thus to be sutured, but not so rigid as to damage the nearby tissue, and permeability to nutrients, yet avoiding scar tissue formation. [7] The most limiting factor for tubulization is the distance between the stumps of the resected nerve held by the NCs. Depending on the entity of the injury, this gap has different lengths, which strongly affect regeneration capabilities. Nerve regeneration within an NC occurs spontaneously when the gap is shorter that a certain distance called limiting gap length. On the contrary, beyond this value, no nerve regeneration will occur when using standard polymeric NCs, in contrast with AG that, in principle, has no length limit. This gap is dependent on species and nerve size, resulting in a distance less than 5-mm in the mouse, 10-mm in the rat, while in humans it is considered 30-mm. [8] The limiting gap length is the main bottleneck of tubulization repair. This motivates a strong interest to investigate on novel NC designs. Another interesting challenge for NCs' fabrication is the tunability of their structural properties to cope with different nerve characteristics. Peripheral nerves are highly anisotropic and heterogeneous structures, whose mechanical properties can vary considerably between them, depending on their size and anatomical site. [9,10] Nerves may be subjected to a dynamic environment characterized by high compression and tension stresses, as it occurs for those located near articular joints or in deep regions of the body. [11] Hence, versatile NC fabrication procedures are particularly advantageous from both a clinical and an industrial point of view.
Several different designs and materials have been proposed for the fabrication of nerve conduits. [12,13] Among them silicone, poly(lactic acid-co-glycolic acid) (PLGA), poly--caprolactone (PCL), and chitosan (CH) tubes have been frequently used. Durable tubes have the disadvantage to remain in the body, where may induce a chronic inflammatory response, eventual compression of the regenerated nerve and require a second surgery to remove the conduit. NCs made of biodegradable and natural materials can overcome these problems, as they are gradually resorbed without significant deformation after guiding the growth of the nerve and possess similar physiochemical properties to the native tissue. CH is a linear polysaccharide obtained from the alkaline deacetylation of chitin, formed by d-glucosamine and Nacetyl-d-glucosamine units linked by -1,4-glycosidic bonds. [14] When the deacetylation degree (DA) reaches about 50%, chitosan becomes soluble in aqueous acidic media, which causes the amino groups of the chain to protonate and the polymer be-comes cationic. [15] This positive charge is thought to be responsible for its antimicrobial activity, via the interaction with negatively charged cell membranes of microorganisms, making it useful for making implantable biomedical devices. In the case of peripheral nerve regeneration, the positive charge of chitosan interacts with negatively charged axons and enhances functional recovery. [16] Furthermore, DA is a relevant factor able to influence the survival, proliferation, and cellular activity of SCs. [17,18] Nerve conduits made of CH have been tested in long gap nerve injuries with success [19][20][21] and have reached commercialization as Reaxon conduits. [18] However, chitosan and, in general, natural material-based NCs suffer from common limitations, such as poor mechanical strength that is even lower in physiological environments. [22,23] It is thus promising to combine natural substances with other materials or structures to obtain conduits with improved properties.
Here, we present a novel design of porous CH NCs with the incorporation of a cylindrical PCL mesh, with the aim to improve its mechanical properties over time and to provide a highly versatile fabrication procedure, with tunable structural characteristics. Extrusion-based 3D printing has been used to fabricate the PCL mesh, exploiting the versatility of additive manufacturing technologies to tune geometrical features. Moreover, freeze-dying was used to create a porous CH matrix embedding the PCL mesh, allowing nutrients exchange, and providing a favorable environment for cell proliferation. To the best of our knowledge, this is the first report that combines 3D printing with freeze-drying techniques to fabricate a nerve conduit with tunable properties. A patent application has been filed based on this NC design. [24] In this study, such hybrid conduits have been extensively characterized in vitro and tested in vivo to support axonal regeneration along a limiting gap of 15-mm in the sciatic nerve of rats.

Materials
High-purity medical-grade CH was purchased from Heppe Medical Chitosan GmbH (Halle, Germany) with DA ≥ 92.6% and viscosity in the range of 151-350 mPa s. PCL (M W = 80 kDa) was purchased from Merck (Darmstadt, Germany). Lysozyme powder from chicken egg white was purchased from Merck. All other reagents and chemicals were purchased from Merck.

PCL Mesh Fabrication
Planar rectangular PCL mesh with honeycomb cells was designed and fabricated with a 3D-Bioplotter Manufacturer Series (EnvisionTEC GmbH, Gladbeck, Germany). Planar dimensions were varied depending on the dimension of the target nerve. Thus, for small nerves as the sciatic nerve of the rat, planar mesh dimensions were set at 11 × 40 mm, whereas for bigger nerves were set at 20 × 40 mm. To print the PCL mesh, polymer beads were loaded inside a high-temperature printing cartridge and heated at 130°C for 15 min, to allow complete polymer melting. Metal extrusion needles with 0.4 mm diameter nozzle (Nordson, Westlake, USA) were used. Further polymer extrusion was operated by controlling of the 3D-Bioplotter with two units: the PerFactory software for ".STL" data import and slicing, and the VisualMachines software for material parameters and Bioplotter control. An extrusion pressure of 6.8 mbar and a Z-offset of 320 μm between the printing substrate and the extrusion needle were imposed for all the manufactured meshes. Printing speed was set at 0.4 and 1 mm s −1 , respectively to obtain a PCL mesh with different strand dimensions. Two different honeycomb patterns were chosen by setting printing parameters, one with larger cells by setting 1.2 mm distance between PCL strands and 2.2 mm as cell period. The mesh with this cell pattern was referred to as L_mesh. The other honeycomb pattern was chosen to have a smaller cell design and was manufactured by setting 0.7 mm as distance between PCL strands and 1 mm as cell period. The mesh with this cell pattern was referred to as S_mesh. PCL meshes were printed onto a Kapton substrate. Once the process was completed, meshes were left undisturbed for 15 min to allow complete polymer curing and washed with 70% ethanol (EtOH) to ease detachment. After printing, planar PCL meshes were treated to obtain a cylindrical shape by manually rolling them inside a dedicated cylindrical Teflon mold and incubating them in deionized (DI) water at 57°C for 1 min to allow polymer chains' rearrangement to give the desired shape to the mesh. After this step, the mold containing the heated polymeric structure was immediately incubated in DI water at room temperature for 1 min to cure the polymeric strands and to obtain the final cylindrical mesh. Optical microscopy (HRX-01, Hirox, Tokyo, Japan) was used to characterize PCL mesh morphology and dimensions.

Nerve Conduit Fabrication
A 2.5% (w/v) CH solution was prepared by dissolving the polymer in DI water with 1% (v/v) acetic acid, stirred for 2 h at 45°C and left overnight at room temperature. CH solution was then filtered with a metallic mesh (300 μm pore diameter), degassed to remove air bubbles, and stored at 4°C until further use. A schematization of the overall conduit fabrication procedure is reported in Figure 1.
The prepared CH solution was heated at 37°C in an incubator with an orbital shaker and gently injected inside a custom-made mold to avoid air bubble formation. The mold consists of various components, appropriately sized to obtain nerve conduits of different dimensions depending on the target nerve ( Figure S4, Supporting Information). A Teflon-coated stainless-steel mandrel was used to create the internal lumen of the conduit. For the rat sciatic nerve, a 2 mm diameter mandrel was employed. A hollow Delrin cylinder with a groove was fabricated to hold both the PCL mesh and the metallic mandrel in vertical position. In this way, the PCL mesh was fixed coaxial to the mandrel. These components were inserted into a cylindrical Teflon mold, and CH solution was then injected inside it. Then the mold was placed in a freezer at −20°C or −80°C for 12 h to allow complete CH solution freezing. After this time, the system was demolded by removing the Delrin cylinder with the frozen polymer from the Teflon mold by simply allowing gentle thawing of the external layer of CH solution in contact with the mold and extracting it. Then the frozen polymer still anchored to the metallic mandrel was placed in a glass vial and freeze-dried (Labconco, Kansas City, USA) for 48 h. After lyophilization, the polymeric structure was demolded from the mandrel and incubated at room temperature in a 1% (w/v) NaOH solution for 15 min, to neutralize any remaining acetic acid and to allow physical crosslinking via hydrogen bonding formation between the polymer chains. [49,50] After neutralization the structure was equilibrated at physiological pH by incubation in DI water for 30 min and then in phosphate buffered saline (PBS) for 24 h to yield the final conduits. Then, conduit samples were cut with a scalpel at various dimensions depending on the test and usage. NCs embedding PCL mesh printed with 1 mm s −1 and treated with the aforementioned freezing temperatures were used for morphological characterization and for cell experiments. These samples were referred to as Chi@PCL (T = −80°C) and Chi@PCL (T = −20°C) to distinguish between different CH network morphologies. For mechanical characterization, conduits with different PCL mesh design and strand dimensions were tested, as well as neat CH NCs without PCL mesh fabricated with the same process and used as control. Prior cell experiments and animal surgery, Chi@PCL NCs were sterilized by overnight incubation in a 70% EtOH solution and washed with sterile saline before implantation.

Nerve Conduits' Characterization
Morphological characterization of the conduits was carried out with optical and scanning electron microscopy (SEM). Samples were incubated in DI water for 3 h to remove any remnant salt and dried by freeze-drying to avoid CH network collapsing. Before SEM imaging, a thin layer of few nanometers of Pt was sputtered onto the samples that were further fixed on metal sample holders using carbon tape. SEM images were acquired using backscattered electron (BSE) and secondary emission electron (SED) modes for higher-resolution images. 5 and 15 kV were used to scan samples with BSE and SED modes, respectively. Overall samples' porosity was calculated using the ethanol displacement method. [51] Pore dimensions was estimated with Fiji (https://imagej.net/) approximating their shape to an ellipse and calculating the major and minor axes for each one and referring them to as longitudinal and transversal pore dimensions respectively. Pore size was measured in several regions of the conduits, namely within the wall, the internal and the external surface, and displayed using frequency distribution histograms. Furthermore, pore shape and alignment were evaluated using the directionality plug-in of Fiji software. Conduit dimensions (internal diameter and wall thickness) were calculated in dry state (freeze-dried after fabrication) and in hydrated state (after PBS incubation) with SEM imaging and optical microscopy. Water adsorption studies were also conducted to evaluate equilibrium swelling. [31] In vitro degradation experiments were carried out by incubating 5 mm long specimens (n = 5 for each freezing temperature) in PBS with 4 mg mL −1 lysozyme at 37°C under orbital shaking. Fresh lysozyme solution was replaced every week. After 2 and 4 months of incubation, samples were taken out from the solution, washed in DI water for 3 h, freeze-dried, and weighted. Polymer degradation over time was calculated as percentage of weight loss with the following equation (Equation (1)) where W i represents the initial weight of the samples prior to incubation with lysozyme solution and W f represents the weight after a certain time of incubation. Graphs and histograms of morphological characterization were plotted using GraphPad Prism 8. Mechanical characterization of the conduits was carried out to evaluate how the fabrication parameters of the Chi@PCL NCs (freezing temperature, v printing , and PCL mesh design; Table 1) could influence their mechanical properties, comparing them to neat CH NCs. Mechanical tests were performed using a tensile machine (Instron, USA) with custom setups. All the mechanical data were computed using MATLAB (Mathworks, USA). The conduits were cut in specific lengths depending on the test and incubated overnight in PBS at room temperature before performing the mechanical tests. Three-point bending tests were conducted fixing the NCs (15 mm long) on the lower grip between two supports spaced 10 mm apart and indenting the NC in the middle of the structure in the radial direction with a speed of 1 mm min −1 . The bending stiffness was computed as the slope of the strain/stress curve in the first linear region. [52] Radial compression tests were performed by placing the conduit (10 mm long) on a flat surface and then indented in the radial direction with a square shape over the longitudinal length of the NC, at a speed of 1 mm min −1 . Radial compression was evaluated from the strain/stress curve at 10%, 30%, and 50% of the compressive strain. The compressive stiffness (Young's modulus in compression) was computed from the theory of elasticity. [53] Furthermore, to evaluate the resistance of the Chi@PCL NCs to tensile stress of the suture thread during implantation, pull-out suture thread tests were performed, comparing Chi@PCL NCs fabricated with S_mesh (v printing = 1 mm s −1 ) to neat CH NCs. A 6-0 suture thread (PROLENE, Ethicon, USA) was clamped on the upper grips of the tensile machine, while it was secured on the NC with two knots at 2 mm from the top of the NC itself. The test was controlled in displacement with a speed of 1 mm min −1 and stopped when a breakage of the conduit wall or of the suture thread was noted.
For all the in vitro experiments, tubular Chi@PCL NCs were kept in EtOH 70% v/v in distilled water until use. Then, the conduits were washed three times with PBS without Ca 2+ and Mg 2+ (PBS w/o, Corning, 21-040-CMR) supplemented with 1% P/S and 2.5 μg mL −1 amphotericin B (Euroclone, ECM0009D). The NGCs were then cut open along the longitudinal axis, inserted in a Cell-Crown 12NX insert (Scaffdex, C00007S; inner open surface area: 1.07 cm 2 ) with the inner side facing upward, and the scaffold excess was removed with a sterile scalpel. The scaffolds in the CellCrown inserts were exposed to UV light for 30 min. The samples were air-dried in the biosafety cabinet for 1 h and kept for 30 min in the CO 2 incubator before cell seeding. The samples in the CellCrown12NX insert were kept in a 12-well plate (Thermo Sci-entific150628) for the duration of the subsequent experiments. The tests were performed by seeding cells on the inner side of the tubular scaffolds at a low concentration to avoid reaching over confluence at the end of the experiment.
Cell adhesion on the conduit samples was assessed with fluorescent staining using the LIVE/DEAD viability/cytotoxicity kit (L3224, Invitrogen) according to the manufacturer's protocol. The scaffolds (n = 3) were seeded with 100 μL of cell suspension (3500 cells cm −2 ) in GrM and incubated in the CO 2 incubator at 37°C for 1 h. Afterward, 1.5 mL of GM was added to each scaffold. The samples were kept in GM for 6 days, and the culture medium was changed every other day. On day 6 of culture, the samples were gently washed with PBS w/o, and a staining solution of PBS w/o with 4 μm EthD-1 and 2 μm calcein AM was added to each of them. Samples were then incubated for 20 min at 37°C in the dark. Hoechst 33342 staining was added to the staining solution (1:1000, H1399, Invitrogen) for an additional 10 min to counterstain the conduits. Representative fluorescence images were taken for each experimental condition with a Leica DMi8 microscope (Leica Microsystems, Wetzlar, Germany).
Cell viability on the scaffolds was evaluated by measuring the release of lactate dehydrogenase (LDH) in the culture medium and by normalizing it on the total DNA amount of each sample (n = 3). Cells growing on tissue culture polystyrene (TCPS) were used as a control (n = 3). The scaffolds were seeded as previously described. Control cells growing on TCPS were seeded at a concentration of 1000 cells cm −2 to avoid reaching over confluence at the end of the experiment. The scaffolds and controls were kept in GrM for 6 days, and the culture medium was changed every other day. Every time that the culture medium was renewed (GrM2, GrM4, and GrM6), the replaced medium was harvested and frozen at −80°C for LDH activity measurements. On day 6 of culture, each scaffold was gently washed with PBS w/o. Afterward, cells were lysed by adding 500 μL of nuclease-free water (Sigma-Aldrich, W4502) on each sample and subjecting the scaffolds to three freeze-thaw cycles at −20 and 37°C, respectively. For LDH activity quantification, the supernatants were thawed at 4°C, and the Lactate Dehydrogenase Activity Assay Kit (Sigma-Aldrich, MAK066) was used according to the manufacturer's protocol. LDH activity was detected by measuring the absorbance at 450 nm with a VICTOR X microplate reader (PerkinElmer). To quantify the total DNA amount of each sample, the cell lysates were thawed and the Quant-iTPicoGreendsDNA Assay Kit (In-vitrogenTM, P11496) was used according to the manufacturer's instructions. The DNA amount was proportional to the fluorescence intensity, measured with a VICTOR X microplate reader at the excitation/emission of 485 nm/535 nm. LDH activity measurements at the three time points (GrM2, GrM4, and GrM6) were summed for a measure of the cumulative LDH activity and then normalized to the amount of total DNA.
Animals were anesthetized with ketamine (75 mg kg −1 ) and xylazine (10 mg kg −1 ) intraperitoneally. The right hindlimb was shaved and disinfected with povidone-iodine solution. Then, the right sciatic nerve was surgically exposed at the midthigh. A 15 mm sciatic nerve segment was resected and excised. In the AG15 group, the nerve segment resected was sutured back to the proximal and distal stumps using three 10-0 nylon epineural sutures at each end, keeping its original orientation. In the conduit groups, an 18 mm Chi@PCL tube was sutured using one 10-0 epineural suture at each end leaving a gap of 15 mm between the nerve stumps. Finally, the incision was closed by layers and disinfected with povidone-iodine solution.
From 1 week prior to the surgery, amitriptyline (20 mL L −1 ) was administered in drinking water to prevent development of neuropathic pain. [54] Postoperative buprenorphine (0.03 mg kg −1 ) was given to treat surgical pain. All experimental procedures were approved by the Ethics Committee of the "Universitat Autònoma de Barcelona (UAB)" and "Generalitat de Catalunya" (reference #10 306) and followed the European Directive 2010/63/EU.

Functional Tests
The pinprick test was performed at 60, 90, and 120 days post injury (dpi) to assess the progression of nociceptive reinnervation in the hind paw. [55] For this purpose, the animals remained awake and slightly immobilized while the skin surface of the , D, and C footpads and fourth toe were pricked with a blunt-tipped needle to avoid damage to the skin. Depending on the response of the animal, a score was assigned. In case of no response, the value was 0. The total sum of these values was used to construct a global score of pinprick.
The walking track test was carried out at the same intervals to assess locomotor function. The plantar surface of the rat hindpaws was painted with black ink, and the rat was left to walk along a corridor with white paper on the base. [43] Footprints of the operated and intact paws were identified, and the print length (PrL), the distance between the tips of first and fifth toes (TS) and between the second and fourth toes (IT) were measured. Then the sciatic functional index (SFI) [56] was calculated for each rat.
Electrophysiological tests were performed at 60, 90, and 120 dpi to assess the reinnervation of target muscles. Animals were anesthetized as above. The sciatic nerve was electrically stimulated (Synergy, Viasys HealthCare) with transcutaneous needle electrodes placed at the sciatic notch. The ground electrode was placed at the knee and the compound muscle action potential (CMAP) of the tibialis anterior (TA), gastrocnemius (GM), and plantar interosseous (PL) muscles were recorded with needle electrodes. [43] Action potentials were amplified to measure the latency to the onset and the maximal amplitude. Control values were obtained from the contralateral paw during follow-up.

Histological Studies
At the end of the follow-up, set at 120 dpi, animals were euthanized by an overdose of sodium pentobarbital (200 mg kg −1 ). The sciatic nerve and footpads C and D were collected and fixed in 4% paraformaldehyde for 2 h at room temperature. After fixation, the harvested nerves were divided in three segments, proximal, medial, and distal. Medial segments of the nerve graft or conduit and the footpads were cryopreserved in sucrose 30% in PBS 0.1 m at 4°C for 24 h for immunohistochemical analyses. Proximal and distal segments of the sciatic nerve with the tube were postfixed in 3% paraformaldehyde + 3% glutaraldehyde in 0.1 m PB for light microscopy. Then, they were rinsed with 0.1 m PB, post fixed with 2% osmium tetroxide in 0.1 m PB for 2 h, dehydrated in a sequence of increasing concentrations of ethanol, and embedded in EPON resin.
The medial segment of the nerves was cross-sectioned in slices of 15 μm thickness with a cryotome (Leica). The samples were washed and incubated with blocking solution (PBST + 10% host serum) for 1 h at room temperature. Then, samples were incubated overnight with primary antibody at room temperature. After washing the samples, they were incubated in secondary antibody for 2 h at room temperature, washed and mounted with 4',6-diamidino-2-phenylindole (DAPI) (1:10 000; Sigma) for nuclear counterstain. Immunolabeling sections were viewed under an epifluorescence microscope (Olympus BX51). Primary antibodies used were for labeling axons (RT97; 1:200), Schwann cells (S100; 1:50), macrophages To determine the number of regenerated myelinated axons, the distal segment of the nerve graft/conduit including the suture level embedded in EPON was cut with an ultramicrotome of 0.5 μm thickness. Images were taken with a light microscope (Olympus BX40) at 40× magnification to measure the transverse area of the nerve and at 1000× magnification for selecting fields for myelinated axons counting. At least 30% of the nerve cross-sectional area was analyzed, and axon density was calculated. To calculate the number of regenerated axons per nerve, the area was multiplied by the density.
Sagittal sections of 60 μm thickness of the footpads were obtained with a cryotome. The slices were washed 10 min with PBS and incubated with blocking solution (PBST + host serum 1.5%) for 40 min. Then, they were incubated overnight with the primary antibody anti-PGP9.5 UCHL (1:500) + PBST-NDS at 4°C. After washes the slices were incubated with secondary antibody Cy3 donkey antirabbit (1:200). Finally, they were dried on gelatincoated slides, dehydrated, and mounted with mountant for histology (DPX) and viewed under an epifluorescence microscope (Olympus BX51).

Statistical Analysis
Data were expressed as mean ± standard error of the mean. The results of morphological characterization were analyzed by oneway Analysis of Varince (ANOVA) followed by the Tukey posthoc test, using GraphPad Prism 8 software. The results of three-point bending and radial compression resistance tests were analyzed by two-way ANOVA followed by the Tukey post-hoc test, using MATLAB R2021b software. The results of pull-out suture thread were analyzed by one-way ANOVA followed by the Tukey posthoc test, using MATLAB R2021b software. The results of in vitro experiments were analyzed using the Kruskal-Wallis test with Dunn's posthoc test. The box and whisker plots represented the median, maximum, and minimum values. The results of in vivo functional tests and histology were analyzed by two-way ANOVA followed by the Tukey posthoc test, using GraphPad Prism 8 software. Differences were considered statistically significant at p < 0.05.

PCL Mesh and Nerve Conduits' Characterization
Extrusion-based 3D printing allowed fine control over the PCL mesh geometry (Figure 2). Honeycomb cell design was chosen because of its ease of manufacturing (see the "Experimental Section"). Furthermore, a design composed of polymeric strands that intertwine to form hexagonal cells would allow us to obtain a double positive effect; they constitute the structural element of the conduit, while maintaining sufficient space to guarantee the exchange of nutrients and the obstruction of fibrotic tissue infiltration, both ensured by CH matrix. The PCL mesh has remarkable compliance and flexibility, resisting extensive manipulation, and it brakes only after extreme plastic deformation upon tensile strain (Video S1, Supporting Information). This behavior is due to the good mechanical properties and toughness of PCL. [25] Figure 2a shows a PCL mesh manufactured by setting nozzle printing velocity at 0.4 and 1 mm s −1 , respectively. These experiments were performed by printing the L_mesh design. The average strand width of PCL mesh printed with 0.4 mm s −1 nozzle velocity was 608 ± 45.8 μm, and significantly lower, 345 ± 23.1 μm (p < 0.0001) with 1 mm s −1 nozzle velocity (Figure 2d). PCL mesh thickness was measured and resulted in 261.5 ± 48.5 and 153.8 ± 25.2 μm for meshes printed at 0.4 and 1 mm s −1 , respectively (Figure 2e, p < 0.0001).
We also modified the honeycomb design while keeping constant nozzle printing velocity at 1 mm s −1 (Figure 2b). By printing S_mesh design it is possible to increase honeycomb cell number in the mesh (70 and 20 cells cm −2 for S_mesh and L_mesh, respectively). Importantly, both the width and the thickness of printed PCL strands did not differ when printing the two mesh designs at the same velocity (Figure 2d,e; average strand widths of 345 ± 45.1 and 356.6 ± 74.7 μm, and average thicknesses of 153.8 ± 25.2 and 142.7 ± 24.6 μm for L_mesh and S_mesh, respectively). These results demonstrated that a variation of honeycomb cell design at the same nozzle printing velocity does not affect PCL strand dimensions, allowing us to tune these parameters independently. After PCL mesh printing, a rolling procedure was used to obtain a cylindrical shape that was further anchored to the custom mold prior to CH matrix formation (Figure 2c). The time and temperature of the procedure were chosen to ensure gentle PCL chains' rearrangement, avoiding excessive heating that would cause mesh strands melting. The mold was designed to have the edges of the mesh touching during this procedure, which stick together as a result of the thermal rolling procedure ( Figure S1a, Supporting Information).
Cylindrical PCL meshes represent the structural element of the conduits around which the CH matrix is formed. By incubating the custom mold with the PCL mesh and the CH solution inside at the freezing temperature of −20°C, ice crystals nucleation occurred, and further lyophilization allowed us to obtain a dense sponge-like CH matrix with highly interconnected microstructured pores (Figure 3), consistent with previous reports. [22,26,27] Upon freeze-drying, the CH porous network completely surrounded the PCL mesh, which is placed in the middle of the conduit wall (Figure 3a), formed by a dense porous CH matrix (Figure 3b), present at the external and internal surfaces of the conduits (Figure 3c,d). Optical microscopy pictures confirmed correct embedment of both PCL mesh designs (L_mesh and S_mesh) within the CH matrix without any signs of matrix rupture or mesh misalignment (Figure 3e,f). On the contrary, the mesh is not visible either through the external or the internal NGC surface by SEM scan, confirming tight packaging of CH chains (Figure 3c,d). No differences in the CH porous network morphology or mesh packaging were noticed by embedding PCL meshes with different strands or honeycomb cell dimensions. Handling the Chi@PCL conduits showed remarkable flexibility and toughness, which allow them to withstand consistent compression and bending stresses without breaking or compromising their original structure (Video S2, Supporting Information). Importantly, the proposed fabrication procedure enabled obtaining conduits with different dimensions, depending on the nerve of interest ( Figure S1b, Supporting Information), that display the same pore morphology at different freezing temperatures ( Figure S2, Supporting Information). NC dimensions designed for the rat sciatic nerve are reported in Table 2, whereas those for larger nerves are reported in Table S1 (Supporting Information).
By using −80°C as freezing temperature, we obtained conduits with a different CH porous network, characterized by anisotropic polymeric lamellae that surround and embedded the PCL mesh (Figure 4a). This CH pores' morphology appeared different compared to that of conduits fabricated at −20°C, probably due to different ice crystal nucleation during CH solution freezing. In our fabrication procedure, the custom-made Teflon mold may have created a thermal gradient responsible for the oriented nucleation of the ice crystals. This phenomenon was previously reported when the freezing process was operated at very low temperatures. [27,28] Such occurrence gave rise to a preferential pore orientation of the CH matrix, which appeared as microstructured channels aligned in the longitudinal direction of the conduits (Figure 4c,d). PCL mesh resulted properly embedded within these NCs, as it occurred for Chi@PCL (T = −20°C ) tubes for both fabricated mesh designs (Figure 4e,f). Since modifying freezing temperature appeared to have an influence on CH matrix morphology, pore shape and dimensions were evaluated by SEM imaging, and the results of this analysis are displayed in Figure 5. Directionality analysis performed for Chi@PCL (T = −20°C) NCs showed no proper pores orientation, confirming the amorphous nature of their shape (Figure 5a-c).
On the contrary, Chi@PCL (T = −80°C) NCs displayed a preferential orientation in the longitudinal direction of the conduits (Figure 5d-f). Thresholding operation performed on SEM scans highlighted anisotropic pores of pseudoelliptical shape, resulting from the different freezing conditions. Both the −80 and −20°C conduits evidenced high overall porosity values of 90.2 ± 2.92% and 84.7 ± 2.44%, respectively. A statistically higher porosity (p < 0.001) resulted for Chi@PCL (T = −80°C) NCs, consistent with the different pores morphology of CH matrix evidenced by SEM imaging. Such a high value of overall porosity has been reported to be beneficial for tissue engineering scaffolds, as it allows proper nutrient diffusion and it offers a consistent space that promotes cells attachment and proliferation. [29,30] A highporosity value is associated with a consistent swelling index, as reported in Table 2 for both the types of conduits. Conduits water uptake reached equilibrium after 30 min of incubation in PBS at 37°C and remained stable for the whole duration of the  experiments (72 h, data are not shown), confirming good hydrophilicity and water retention of the NCs, essential conditions to favor nutrients' exchange. [31,32] Freezing temperature also influenced average pore dimension ( Figure 5; Table S2, Supporting Information). Longitudinal pores of −80°C conduits resulted remarkably larger (Table S2, Supporting Information) than in −20°C conduits. Frequency distribution showed marked difference between −80 and −20°C conduits in the longitudinal dimension (Figure 5h,i). Less difference was found in pores' size in the transversal pore dimension (Figure 5l,m). However, also in this case higher pore size was found for −80°C with respect to −20°C conduits (Table S2, Supporting Information).
Overall, morphological characterization of the conduits showed that the manufactured NCs possess a strongly interconnected CH porous network that completely embeds the PCL mesh. The mesh provides stability and elasticity to the structure. Furthermore, pores' morphology analysis demonstrated that by varying freeze-drying conditions it is possible to influence pores alignment and size.
The results of the mechanical characterization are expressed in terms of bending stiffness, radial compression, compressive stiffness, and suture thread breaking force. Figure 6a shows that the incorporation of the PCL mesh in the conduit structure remarkably increases the bending stiffness for all the v printing tested respect to the neat CH NCs. Comparing the Chi@PCL NCs to the neat CH NCs, there is a significant difference for both conduits embedding the 1 mm s −1 (p < 0.001) and the 0.4 mm s −1 L_mesh (p < 0.0001). Moreover, a statistically significant difference is also noted between groups of conduits embedding PCL mesh fabricated with different v printing (p < 0.05). These results demonstrated that, by varying the strand dimensions of the incorporated PCL mesh, it is possible to tune the bending stiffness of the NCs. However, no significant variations were found when comparing the different freezing temperatures within the same group for all the tested conduits (neat CH, Chi@PCL_v printing = 0.4 mm s −1 and Chi@PCL_v printing = 1 mm s −1 ) . This effect could be due to the predominant influence of the PCL mesh in the determination of the mechanical behavior of the NCs.
The same behavior displayed in Figure 6a can also be observed by analyzing the radial compression at three different strain percentages (10%, 30%, and 50%) comparing Chi@PCL and neat CH NCs (Figure 6b). By incorporating PCL mesh with higher strand dimensions (printed using v printing = 1 mm s −1 and v printing = 0.4 mm s −1 , respectively) within the conduits, there was an increase in the force per unit length values that rise nonlinearly with increasing strain. In particular, at a low strain percentage (10%), a significant difference in compression force between the neat CH NCs and the Chi@PCL NCs incorporating PCL L_mesh printed with 1 mm s −1 (p < 0.0001) and 0.4 mm s −1 (p < 0.0001) printing speeds, respectively, was found. Analyzing radial compression at 30% and 50% strain, this behavior was confirmed. A significant difference between the neat CH NCs and both the Chi@PCL NCs incorporating PCL mesh printed with 1 mm s −1 (p < 0.001) and 0.4 mm s −1 (p < 0.0001) printing speeds was observed, as well as between Chi@PCL conduits themselves (p < 0.01). High variability of radial compression resistance was observed for Chi@PCL NCs incorporating L_mesh printed with 0.4 mm s −1 speed. This is consistent with thickness variability of this PCL mesh design as shown in Figure 2e probably due to the printing process. However, such occurrence does not affect the overall mechanical behavior of the conduits, reporting significant increase of radial compression resistance with increasing printing velocity of PCL mesh. Interestingly, Figure 6. Conduits mechanical characterization. a) Bending stiffness of the neat CH NCs and Chi@PCL NCs incorporating L_mesh. Bending stiffness is significantly higher for the Chi@PCL NCs compared to neat NCs for conduits incorporating L_mesh printed at both 1 mm s −1 (****p < 0.0001) and 0.4 mm s −1 (**p < 0.01) printing velocity, and between the two different Chi@PCL NCs groups (*p < 0.05). b) Radial compression tests for neat CH NCs versus Chi@PCL NCs incorporating L_mesh fabricated with different printing velocities at various strain percentages. At 10% strain, there is a significant difference between the neat CH NCs and both the Chi@PCL NCs with 1 mm s −1 (****p < 0.0001) and 0.4 mm s −1 (****p < 0.0001), and between the neat NCs fabricated with different freezing temperatures (*p < 0.05). At 30% strain, in addition to the differences between the neat CH NCs and the Chi@PCL NCs with 1 mm s −1 (***p < 0.001) and 0.4 mm s −1 (****p < 0.0001), there is also a difference between the two groups of Chi@PCL NCs (**p < 0.01). At 50% strain, the same differences observed at 30% strain are reported. c) Bending stiffness for the Chi@PCL NCs with different mesh designs (L_mesh vs S_mesh, *p < 0.05) printed at the same printing velocity of 1 mm s −1 . d) Radial compression tests for Chi@PCL NCs incorporating PCL mesh fabricated with different designs (L_mesh vs S_mesh) printed at the same printing velocity of 1 mm s −1 . The differences are significant at 10% strain (*p < 0.05), at 30% strain (**p < 0.01), and at 50% strain (***p < 0.001).  a significant difference between the compressive force of neat CH NCs fabricated using different freezing temperatures was noted at 10% strain (p < 0.05). This effect could be due to the different pore morphology of conduits fabricated varying freezing temperature that becomes influential at low strain values when testing CH conduits without the incorporation of the PCL mesh. Nevertheless, this peculiarity was not found at the higher strain percentages. This occurrence proves that the freezing temperature does not significantly influence the overall mechanical properties of the conduits, whose stress resistance is mostly determined by the geometry of the PCL mesh.
In addition to the variation of printing velocity, we assessed whether modification of honeycomb cells design, by keeping constant nozzle printing velocity at 1 mm s −1 , could have an influence in the mechanical properties of our conduits. When testing conduits incorporating S_mesh, an increase of both bending stiffness (Figure 6c, p < 0.05), radial compression (Figure 6d, p < 0.05 at 10% strain, p < 0.01 at 30% strain, and p < 0.001 at 50% strain), and compressive stiffness ( Table 3) was evidenced with respect to conduits embedding L_mesh design. Results of radial compression tests are coherent with the computation of the compressive stiffness, whose value increases with increasing PCL strands' dimensions and honeycomb cell number (Table 3). It is worth remarking that the calculated compressive stiffness of our conduits is in perfect agreement with those of the native nerves of both rodents [33] and humans. [34] S_mesh design has been used for the pull-out suture thread tests, comparing the force at which the suture thread breaks the conduits wall. The results of this test are shown in Figure 7 and highlight the significant difference (p < 0.001) between neat CH NCs and Chi@PCL NCs. During tests performed on the neat CH NC, the suture thread finally breaks the wall of the conduits (Video S3, Supporting Information). On the contrary, the higher tensile strength of the PCL mesh strands compared to the CH porous matrix causes the suture thread to break instead of the Chi@PCL NCs wall (Video S4, Supporting Information). This suggests that the incorporation of PCL mesh allows the conduits to withstand higher traction forces caused by a suture thread without damaging their structure.
Overall, the results of the mechanical characterization indicated the possibility of finely tuning compression resistance and bending stiffness by simply varying PCL mesh geometry. Furthermore, computed compressive stiffness showed that our manufacturing procedure produces conduits with mechanical properties comparable to the native nerve.

In Vitro Cell Experiments
To assess the biocompatibility of the fabricated scaffold and its ability to sustain cell adhesion and proliferation, in vitro tests were performed comparing the two conduits, Chi@PCL (T = −20°C ) and Chi@PCL (T = −80°C) (Figure 8). After 6 days of proliferation, cells were adhered to the scaffolds forming small clusters and very few dead cells were present (Figure 8a,b). Interestingly, the presence of anisotropic lamellae in the Chi@PCL (T = −80°C) samples induced cell alignment along such structures (Figure 8b, inset). On the contrary, cell adhesion on Chi@PCL (T = −20°C) samples showed random cell alignment due to the absence of pores directionality for those samples. Further quantitative analyses showed no significant differences (p > 0.05) in cell viability between the sample Chi@PCL (T = −20°C), Chi@PCL (T = −80°C), and the control of cells on TCPS (Figure 8c). These studies demonstrated the biocompatibility of the conduit and its ability to support SCs viability and served as preliminary investigations prior to animal experiments.  the suturing needle though one small hole. The softness of the hydrogel depends on the degree of wetness, so the time of immersion in saline before implantation was controlled to 30 s.

In Vivo Experiments: Functional Recovery
At 60 dpi, all AG15 animals showed reinnervation in all muscles, as judged by the reappearance of recorded CMAP (Figure 9). In the CH15B group, 37% showed innervation in TA, 63% in GM, and none at PL level. Finally, among the CH15C, 75% showed innervation in the TA, all in the GM, and none in the PL. At 90 dpi, the amplitude of the CMAPs increased. All the animals with Chi@PCL conduits showed reinnervation in TA and GM muscles, but a lower proportion for the PL muscle (CH15B 12% and CH15C 0%). Finally, at 120 dpi, muscle reinnervation progressed, and in the PL muscle, all animals with AG15 but only 25% of the other groups had a recordable CMAP. At the end of follow-up, the mean amplitudes of TA (30.0 ± 3.6 mV), GM (41.9 ± 7.1 mV), and PL (2.5 ± 1.0 mV) CMAPs of the AG15 group were significantly higher than in the other groups, which were, respectively, 7.9 ± 3.5, 23.6 ± 3.3, and 0.07 ± 0.15 mV for CH15B and 13.4 ± 7.5, 15.7 ± 5.9, and 0.09 ± 0.2 mV for CH15C. In the conduit groups, the only significant difference occurred for the GM CMAP between CH15C and CH15B (p < 0.05).
Nociceptive sensitivity, assessed with the pinprick test, was completely abolished in the lateral side of the operated hindpaw after surgery. At 60 dpi, a slight response in proximal pads of the  paw was observed in all groups. The scores increased, progressing to more distal sites of the paw, although it was not complete at 120 dpi, when the AG15 group showed a significantly higher score than the conduit groups ( Figure 9).
The walking track test was performed at the same intervals. However, the incidence of autotomy reduced the number of animals with useful prints. As expected, after a complete transection of the sciatic nerve, the SFI showed minimal increase over time, without significant differences between the groups (data are not shown).

Histological Results
After 120 dpi, the repaired nerves were harvested. The conduits were slightly degraded but remained in place at the implantation site (Figure 10). This evidence is consistent with in vitro degradation studies that showed a mild degradation of the NCs, which underwent about 10% weight loss after 4 months of incubation in PBS with lysozyme ( Figure S3, Supporting Information).
Transverse sections stained with hematoxylin-eosin (HE) showed that the autografts had a well-demarcated perinerium, and the three branches of the nerve were visible. In the Chi@PCL conduits, the wall was preserved with the white areas corresponding to the PLC mesh and a trabecular structure corresponding to the CH matrix. In addition, there were cellular infiltrates crossing the polymer wall. At the lumen of the conduit, in all cases an amorphous structure was observed corresponding to the matrix generated to support nerve regeneration (Figure 11). Interestingly, the regenerative matrix filled the full space in the tube lumen (with a 2 mm internal diameter), thus providing a large newly formed tissue supporting axonal growth, in contrast to the usually thin regenerative cable seen centered in impermeable tubes.
The number of regenerating myelinated axons was quantified in semithin sections taken at the mid of the graft/conduit and at the distal end. Images in Figure 12 illustrate the presence of abundant myelinated axons and SCs at the midsegment, grouped in mini fascicles in the endoneurium of the AG, and in the lumen matrix within the conduits, with an increased amount of connective tissue. All groups showed many myelinated axons at the middle segment, but this number decreased at the distal end in the AG15 group, whereas in the conduit groups there were very few reaching this level. Regenerated fibers were found in the distal nerve, indicating successful axonal regeneration along the 15 mm gap, in all the rats of the AG15 group, in five of six rats of the CH15B group with available samples, and in four of eight rats of the CH15C group. The estimated values of myelinated axons are given in Table 4.
Immunohistochemical staining was performed against the relevant cell types in nerve regeneration, i.e., axons, SCs, macrophages, and fibroblasts (Figure 13). In the case of AG15 rats, the nerve structure was preserved, containing many axons and SCs throughout the graft. In the Chi@PCL conduits, the polymer structure of the wall has marked autofluorescence, making it difficult the appreciation of the cell infiltrates. Regenerated axons and SCs were observed in the intratubular matrix in all the cases with conduit repair. Although no quantification has been attempted, the density of neurofilament and of SC labeling appeared higher in group CH15C than in CH15B. Regarding macrophages, they were observed especially in the CH15C group inside the tube. Vimentin labeling was remarkable also in the CH15C group, suggesting higher entrance of non-neural cells across the wall in these conduits.
Pretein gene product (PGP) immunolabeling was performed to characterize the reinnervation of the paw skin. In control samples, there were bundles of nerves fibers forming a subepidermal nerve plexus from which single fibers entered the epidermis, and a central coil innervating the sweat glands (Figure 14). In rats of the AG15 group, the pad innervation followed the same pattern but with reduced density of nerve fibers. In the CH15B group, a few axons had managed to reach the limit of the epidermis, but without forming subepidermal plexuses. CH15C showed nerves that had formed subepidermal plexuses and a few fibers that had managed to enter the epidermis ( Figure 14). The estimated count of intraepidermal nerve fibers per mm length gave values of 18.5 ± 0.3 for group AG15, significantly more than the values of 2.90 ± 2.20 for group CH15B, and 3.08 ± 0.38 for group CH15C.

Discussion
Artificial nerve conduits are becoming an increasingly used alternative technique for the repair of nerve injuries resulting in a short gap. [18,35] However, failure to sustain regeneration in relatively long gaps hampers their clinical application. This limitation is forcing considerable efforts devoted to basic and Table 4. Histological analysis of the regenerated nerves. Density and number of myelinated axons counted in semithin sections taken at the midlength of the autograft or the conduit and at the distal end (* p < 0.005 vs AG15; ** p < 0.0001 vs AG15).  Figure 11. HE-stained nerves. Representative images at 40 ×, 100 ×, and 200 × magnification of medial segments of the different treatment groups at 120 dpi. Note the conduit wall, with holes corresponding to the PLC mesh, and the porous structure of the chitosan matrix. The regenerating cord is at the center of the conduit, in contact with the chitosan wall. Scale bar = 300 μm.
preclinical research in exploring new biomaterials and devices both made of natural or synthetics materials and manufactured by different techniques, which may allow us to optimize nerve conduits for regeneration in long nerve defects. Natural materials, such as CH, show excellent biocompatibility and biodegradability and help to prevent chronic inflammation and subsequent compression of regenerated nerves. In addition, a degradation product of CH, the chito-oligosaccharide, has been shown to promote cell proliferation and prevent apoptosis especially for SCs, the essential cells accompanying regenerating axons. [36,37] Following positive preclinical studies repairing a 15 mm limiting gap in the rat sciatic nerve, [19,20] CH tubes under the commercial name of Reaxon were approved for the repair of human nerve defects of a length up to 30 mm. However, CH presents drawbacks such as low mechanical strength, causing collapse of the conduit, and eventual wall rupture. The novelty of this study lies in the development of a fabrication procedure, which combines two different manufacturing technologies (freeze-drying and extrusion-based 3D printing) to obtain a hybrid conduit with tunable structural properties. Our process creates a highly interconnected porous matrix that surrounds a biodegradable PCL mesh, with different pore shape and dimensions obtained by varying the freezing temperature prior to lyophilization procedure (Figures 3 and 4). It is known that highly permeable conduits influence positively in comparison with impermeable or low-permeable tubes on nerve regeneration by allowing better metabolic exchange, diffusion into the lumen of growth factors generated in the external environment, retention of tropic factors secreted by the stumps, or even infiltration of extraneural cells that may contribute to form an intratubular regenerative cable. [38,39] The PCL mesh has been employed as a supporting structure that facilitates the surgical suture and allows us to sustain the CH matrix that suffers from mechanical instability and experience modifications of its structural integrity over time during implantation. Notably, sharp drops of ≈30% and ≈50% in tensile strength after, respectively, 1 and 2 months of incubation in a simulated physiological environment have been reported for neat CH NCs. [40] Additive manufacturing allowed us to obtain different PCL geometries by varying nozzle printing velocity and honeycomb mesh design (Figure 2). The incorporation of the PCL mesh within the CH matrix of the conduits remarkably enhances both the bending stiffness and the radial compression resistance of the NCs (Figure 6). Significant differences of those parameters were found when varying the mesh strand width or honeycomb cell design, confirming the possibility of precisely tuning mechanical properties of our conduits. On the contrary, modifying CH pore morphology did not produce significant differences in mechanical properties, confirming our hypothesis that PCL mesh represents the main structural element of our conduits. Furthermore, PCL mesh also increases the suture retention strength, providing more stability during implantation respect to neat CH NCs. It is worth noting that our fabrication procedure allowed us to tune the compressive stiffness of the conduits within a physiological range of 0.6-9 kPa ( Table 4) that is reported to support SCs' viability. [41,42] These values are consistent with previous studies on animal and human nerves and could be beneficial to provide stability to the two severed nerve stumps and support the regeneration process without causing additional damage due to mechanical mismatch between the tissue and the implant. [23] In vitro cell experiments performed prior to animal studies confirmed good biocompatibility of our conduits and their ability to support cell adhesion ( Figure 8).
In the present study, a 15 mm gap in the sciatic nerve of the rats was repaired with two variations of Chi@PCL conduits and compared with an ideal AG. We implanted tubes fabricated at Figure 13. Immunolabeling of the regenerated nerves. Representative images (10 ×) of nerves at 120 dpi showing myelinated axons labeled against Neurofilament RT97 (NFRT97-S), SCs labeled against S100, macrophages labeled against Iba1, fibroblasts labeled against vimentin, and cell nuclei labeled with DAPI in transverse sections of a control nerve, AG15, CH15B, and CH15C nerves. Scale bar = 300 μm. −80°C (CH15B) and −20°C (CH15C) freezing temperatures, respectively, and incorporating the S_mesh printed at 1 mm s −1 nozzle velocity, due to the reported good mechanical properties and ease of manipulation during implantation procedure conferred to the conduits by this PCL mesh design. Both types of NCs allowed successful axonal regeneration in animals when used to repair such a critical nerve defect in the rat sciatic nerve. As expected, slower reinnervation of the hindlimb muscles occurred compared to the AG. [19,43,44] This is to be expected as re-generation depends on the existence of the extracellular matrix in which blood vessels, fibroblasts, and SCs can form a new nerve structure. [45] In the AG, this extracellular matrix already exists and is full of pro-regenerative SCs. On the contrary, our NCs are hollow structures and therefore more time is needed for the formation of an initial fibrin cable on which the SCs will migrate, supporting the regenerative process. Despite this, there were many myelinated axons along the conduits in all the rats, indicating that the chosen NC design and the fabrication materials could support axonal regrowth and sprouting. Nevertheless, the number of myelinated axons that reached the distal nerve segment was quite low, as also indicated by the low amplitude of CMAPs in reinnervated muscles. Thus, longer follow-up time may be needed for the Chi@PCL NCs to reach distal regeneration. It is worth noting the study by Shapira and colleagues, [46] who compared CH Reaxon tubes with an AG in a 10 mm gap of the rat sciatic nerve, found similar numbers of myelinated fibers in the two groups. Such a result further supports the quality of nerve regeneration reported for the conduits presented in this study and tested to reinnervate longer distances (15 mm) that the limiting gap length.
Observations in the histological characterization showed that there had been infiltration of cells from the wall into the lumen of the Chi@PCL conduits (see Figures 11 and 13), consistent with previous studies on neat CH NCs. [47] All Chi@PCL conduits showed labeling for different cell types both in the wall and in the lumen of the conduit, which was fully occupied by an amorphous matrix. This is remarkable because previous studies had been performed with nonpermeable materials, such as silicone and PCL, in which the regenerated nerve was located in the center of the tube without adhering to the walls. [39,45,48] The porosity and characteristics of the conduit wall material seem to play a relevant role allowing infiltration by cells that contribute to secreting extracellular matrix components, which fill the lumen and support axonal regeneration. It is worth noting that in 15 mm limiting gap studies with silicone conduits, no animals had regenerated axons at the distal end of the tube, [5,19] highlighting that high porosity and permeability are important characteristics to enhance nerve regeneration over the limiting gap length. Although the conduits remained relatively intact at 120 dpi, it would be interesting to assess whether some degradation product of the CH matrix invaded the tube lumen and facilitated the formation of the regenerative matrix. A similar observation was previously documented with Polyactive tubes by Santos and colleagues. [44] Despite the described promising results, this study has also some limitations. Our fabrication process produced CH matrix with high values of pore dimensions for all the fabrication temperature tested. This could be responsible for the absence of appreciable differences in the regeneration performances between the two groups of conduits.

Conclusion
In conclusion, we described a novel design of porous/3D-printed Chi@PCL nerve conduits and the results of this study demonstrated high tunability of structural characteristics and their ability to support axonal regeneration. Our conduits allowed regeneration over the limiting gap and are therefore more effective than standard impermeable and semipermeable tubes made of silicone or other polymeric materials. Future experiments will be focused to further improve conduit morphology to match the AG regenerative performances, thus to speed up the clinical translation of our conduits.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.