Abstract

   Complicated peripheral nerve injuries or defects, especially at
   branching sites, remain a prominent clinical challenge after the
   application of different treatment strategies. Current nerve grafts
   fail to match the expected shape and size for delicate and precise
   branched nerve repair on a case-by-case basis, and there is a lack of
   geometrical and microscale regenerative navigation. In this study, we
   develop a sugar painting-inspired individualized multilevel
   epi-/peri-/endoneurium-mimetic device (SpinMed) to customize natural
   cues, featuring a selectively protective outer sheath and an
   instructive core, to support rapid vascular reconstruction and
   consequent efficient neurite extension along the defect area. The
   biomimetic perineurium dictates host-guest crosslinking in which new
   vessels secrete multimerin 1 binding to the fibroin filler surface as
   an anchor, contributing to the biological endoneurium that promotes
   Schwann cell homing and remyelination. SpinMed implantation into rat
   sciatic nerve defects yields a satisfactory outcome in terms of
   structural reconstruction, with sensory and locomotive function
   restoration. We further customize SpinMed grafts based on anatomy and
   digital imaging, achieving rapid repair of the nerve trunk and branches
   superior to that achieved by autografts and decellularized grafts in a
   specific beagle nerve defect model, with reliable biosafety. Overall,
   this intelligent art-inspired biomimetic design offers a facile way to
   customize sophisticated high-performance nerve grafts and holds great
   potential for application in translational regenerative medicine.

   Subject terms: Trauma, Translational research, Molecular modelling,
   Experimental models of disease
     __________________________________________________________________

   Nerve grafts often fail to match the expected shape and size for
   branched nerve repair. Here, the authors fabricate a sugar
   painting-inspired individualized multilevel
   epi-/peri-/endoneurium-mimetic device to achieve efficient regeneration
   of neural fibers.

Introduction

   Peripheral nerve injury is a common and destructive problem in clinical
   practice. It is primarily characterized by slow revascularization,
   irreversible axon damage and mismatching, and severe sarcopenia, all of
   which eventually lead to sensory or motor disability in
   patients^[40]1,[41]2. Currently, the most ideal clinical outcomes are
   expected to be achieved by simple end-to-end sutures for tension-free
   nerve gaps, whereas autografts remain the gold standard for critical
   defects (longer than 4 cm in humans)^[42]2–[43]4. Although autografts
   mostly meet the requirements for peripheral nerve regeneration (PNR),
   including high biocompatibility and natural structural composition,
   limitations such as limited donor availability, secondary damage, and
   size mismatching are still prominent and restrict their clinical
   application and therapeutic outcomes, especially in sophisticated nerve
   defect scenarios^[44]5,[45]6. Tissue engineering strategies in recent
   years have been fully developed and applied in the area of PNR, yet a
   majority of them fail to achieve delicate and individualized mimicry of
   the native structure and, therefore, lead to unsatisfactory
   outcomes^[46]7,[47]8. An ideal nerve graft is expected to guide
   geometrical and microscopic neurite outgrowth for efficient neural
   repair to alleviate the therapeutic dilemma.

   Previous studies have reported a number of biomimetic nerve guidance
   conduit (NGC) types fabricated by electrospinning, gas foaming, and
   solvent casting to drive neural regrowth following injury^[48]9,[49]10.
   However, these strategies are often difficult to translate and apply in
   complex clinical cases due to ambiguous bioactive components and
   difficulty in individualized manufacturing (e.g., producing specific
   architectures and mechanics). Thus, architectural simulation may be a
   promising direction for biomimetic nerve graft development. The natural
   nerve fiber is supported by multilevel neurium, epineurium,
   perineurium, and endoneurium, which contribute to a nerve repair
   platform^[50]11–[51]13. Therefore, the neurium-mimetic fabrication of
   grafts, similar to allogeneic acellular nerves, could be used to
   manufacture grafts with similar architecture. Irregular nerve defects
   need grafts with unique shapes and sizes including nonuniform
   morphologies or bifurcations, instead of common
   autografts^[52]14–[53]17. Regarding the difficulty in the accurate
   design and fabrication of nerve grafts with multiple branches, it is
   preferable to utilize advanced techniques for one-step
   three-dimensional (3D) fabrication, such as a form of additive
   manufacturing combined with digital imaging, for ideal graft
   production^[54]18,[55]19. To achieve “precision medicine”, another
   issue is integrating biomimetic layouts into grafts based on neural
   structures and the regenerative microenvironment to induce injured
   nerve fiber extending, for which detailed architectural designs and
   accessible fabrication techniques are needed. Precise nerve fiber
   guidance by biomimetic graft is undoubtedly helpful for promoting
   efficient neural regrowth. Therefore, individualized biomimetic
   strategies might hold promise for clinical translation and wide
   application.

   Achieving an optimized microscale design calls for a full understanding
   of the spatiotemporal patterns of various regenerative units and
   essential characteristics of the microenvironment. Rapid reconstruction
   of structural integrity and microenvironmental homeostasis is crucial
   to optimizing reparative outcomes, as delayed surgery will result in
   significant functional loss^[56]20. The microenvironment during PNR is
   a dynamic landscape formed by biophysical and biochemical cues that
   regulate tissue regrowth and cell behaviors through direct interactions
   and paracrine pathways^[57]9,[58]21. The clearance of myelin debris,
   vascular involvement, and Schwann cell (SC) remyelination together
   contribute to restoring the microenvironmental homeostasis of
   peripheral nerves, during which new vessel formation may be key for
   nourishing the microenvironment in a paracrine manner^[59]22–[60]24.
   For instance, newly formed vessels contribute to functional cell
   activities during tissue regeneration through nutrient delivery and
   microenvironmental regulation^[61]25. Although there have been numerous
   studies on vascularization in PNR focusing on vessel-related nutrient
   supply and waste discharge^[62]26–[63]28, the paracrine modulation of
   vessels remains an understudied field. Moreover, many studies have
   concentrated on regenerative vessels within the epineurium, neglecting
   the indispensable role of intraneural vessels^[64]29,[65]30. It is
   accepted that topological cue-induced vascular reconstruction guides
   regrowth direction (orientation or other well-organized cues) and
   provides microenvironmental homeostasis and a niche for neural tissue
   formation. Regarding the “nerve functional unit” composed of the
   axon/myelin and intraneural vessels, studies should focus on
   investigating the regrowth pattern of the intraneural vessel network
   within regenerative peripheral nerves and the paracrine modulation of
   myelin sheath formation and neurite outgrowth. These facts suggested
   that an intelligent nerve graft could be used to link biomaterials
   (host) and regenerative “nerve functional units” (guest) for maximal
   tissue regeneration efficiency, and after repair, allow the host-guest
   relationship to evolve such that fresh tissue (host) is integrated into
   the organism without biomaterial degradation products (guest) to
   achieve functional restoration.

   In the present study, we hypothesize that optimized multilevel
   neurium-mimetic architectures, achieved by additive manufacturing and
   phase separation, could provide appropriate physical chambers and guide
   the oriented growth of intraneural vessels and that the well-formed
   intraneural vasculature would further contribute to the endoneurium,
   maintaining the microenvironmental homeostasis of neural tissue through
   paracrine function. The “vessel-myelin-axon” could be regarded as a
   “functional unit” of the microstructure expected to promote efficient
   neural regeneration. Accordingly, a sugar painting-inspired
   individualized multilevel epi-/peri-/endoneurium-mimetic device
   (SpinMed) was designed and applied in various nerve defect models.

Results

Performance of SpinMed

   To ensure the adaptivity of the graft to the recipient site, sugar
   painting, a traditional art form in China, has inspired the biomimetic
   design of individualized grafts combined with phase separation, the raw
   material of which, sugar or caramel, confers plasticity and biosafety
   to the innovative nerve grafts (Fig. [66]1a). Regarding the
   indispensable roles of the neurium in microenvironment maintenance for
   neural growth, we constructed a multilevel neurium-mimetic architecture
   to mimic native epineurium/perineurium, offering structural support and
   a biomechanical microenvironment, and to induce the formation of an
   endoneurium-mimetic functional unit.

Fig. 1. Design and performance of the SpinMed graft.

   [67]Fig. 1
   [68]Open in a new tab

   a Schematic illustration of the design ideas, features, and
   applications of SpinMed. (i) The multilevel biomimetic strategy for
   mimicking the epineurium/perineurium to offer structural support and a
   biomechanical microenvironment and inducing the formation of an
   endoneurium-mimetic biological unit. (ii) The endoneurium-like
   regenerative unit is composed of fibroin and endothelial-derived
   molecules, which bind to SCs. (iii) Sugar painting-inspired additive
   manufacturing following digital imaging used for individualized nerve
   defect repair. b Gross image of the epineurium-like sheath obtained by
   the phase separation reverse-mold method after three-dimensional sugar
   printing. c Gross image of the SpinMed graft formed by filler perfusion
   of the outer shell. d Irregular SpinMed fabrication following MRI
   detection. e Scanning electron microscopy (SEM) images of natural
   nerve, SpinMed, and decellularized nerve with views of the outer
   surface, longitudinal section, and cross-section. f Illustration of
   compression tests, in which various grafts were restored after 40%
   compression for 10 cycles. g The compression stress-strain curves of
   the hollow construct, SpinMed graft, natural nerve, and decellularized
   nerve. h The compression stress-strain curves of the SpinMed filler,
   natural nerve, and decellularized nerve in a single compression. i
   Illustration of tensile tests, in which various grafts were stretched
   to rupture. j The tensile stress-strain curves of the hollow construct,
   SpinMed graft, natural nerve, and decellularized nerve. NN, natural
   nerve. DN decellularized nerve. The experiments in e, g, h, and j were
   independently repeated three times with similar results.

   In clinical practice, we found that areas of the posttraumatic PNR were
   often destroyed following cicatricial fibrosis, as identified in 23
   patients, due to the weak epineurium, making it difficult to keep out
   adjacent tissue (Supplementary Table [69]1). Single-cell RNA sequencing
   of human specimens revealed that smooth muscles actin α (αSMA)-positive
   muscle cell or pericyte invasion into the regenerative space primarily
   contributed to retarded neural regrowth (Supplementary Fig. [70]1;
   Supplementary Fig. [71]2a), and this result was similar to that of a
   previous study^[72]31. Linear regression analysis showed that
   intraneural fibrosis, especially in αSMA^+ areas, significantly
   compromised regenerative nerve tissue (Supplementary Fig. [73]2b, c).
   Further experiments using an αSMA-tk mice nerve crush model
   demonstrated that deletion of αSMA^+ cells restored, at least in part,
   neurite regrowth (Supplementary Fig. [74]2d–g). Consequently,
   epineurial design is expected to protect regenerative nerves from
   physical invasion and biological suppression of adjacent muscles and
   cells. The biomimetic epineurium was fabricated via phase separation
   following precise additive manufacturing (Supplementary Fig. [75]3a),
   where the epineurium with a hydrophobic surface and pores of ~1 μm in
   diameter, provided pericyte protection. The biomimetic perineurium
   filling the epineurial sheath was very similar to the natural
   perineurium, potentially providing physical interfaces for neural
   repair and supporting intraneural angiogenesis in an axial direction.
   Notably, the endoneurium, at the micrometer scale, is difficult to
   precisely mimic using current tissue engineering methods; thus, we
   employed a biological induction strategy involving host–guest
   crosslinking to facilitate spontaneous endoneurial formation.

   Based on the above designs, SpinMed grafts with various sizes and
   shapes were fabricated and submitted to examinations (Fig. [76]1b–d).
   Morphological detection showed differential outer and inner interfaces
   (Fig. [77]1e), along with a notable distinction in hydrophilicity, as
   observed through water contact angle measurements; this feature, at
   least in part, contributed to distinctive biodegradation behavior and
   cell adhesion affinities (Supplementary Fig. [78]3b–e). The
   function-enhanced epineurium-like sheath of the SpinMed grafts was
   dramatically superior to the natural structure, exhibiting high
   permeability to nanoscale proteins and lower permeability to
   muscle-derived pericytes with micrometer-scale dimensions
   (Supplementary Fig. [79]3f–i). These findings revealed that the
   optimized regeneration cues outperformed natural morphology simulation
   strategies, indicating that graft adaptivity to the regenerative
   microenvironment is a priority consideration in material development
   and preparation. On the other hand, the fibroin-based SpinMed filler
   was fabricated through freeze-drying with careful adjustments of the
   hyaluronic acid content to achieve the optimal filler architecture and
   rheological properties (Supplementary Fig. [80]4). Furthermore, we
   could precisely modulate the filler characteristics, such as by
   producing a regular topology or random morphology, which exhibited
   identical chemical properties but distinct physical features
   (Supplementary Fig. [81]5). The topology mimicking the perineurium
   showed aligning cues and effects similar to those of natural nerves and
   decellularized nerve grafts, achieved through a gradient freezing
   technique. The compression properties of the SpinMed grafts and other
   counterparts were investigated (Fig. [82]1f–h; Supplementary
   Fig. [83]3j, k). SpinMed displayed flexibility during repeated
   compressions, which facilitated the mitigation of mechanical
   disturbances from muscle contraction after implantation in vivo.
   Notably, the enhanced physical characteristics were primarily
   attributed to the outer sheath of SpinMed rather than the filler, while
   the strain-stress features of the filler were comparable to native
   nerve tissue during 0 to 20% compression. The tensile stress test
   revealed that the mechanical stress of the SpinMed graft was
   significantly superior to that of natural nerves and decellularized
   nerves, ensuring sufficient adaptability after implantation
   (Fig. [84]1i, j).

SpinMed accelerates axial extension of vessels and increases paracrine MMRN1
signaling

   To evaluate the angiogenic capacities of SpinMed, we first performed
   cell proliferation, migration, and tube formation assays using SpinMed
   filler in vitro. The EdU assay showed that the proliferative rates of
   human umbilical vein endothelial cells (HUVECs) in the SpinMed or
   random counterpart groups were slightly decreased compared to those in
   the tissue culture plate group, but the difference was not
   statistically significant (p = 0.1120) (Supplementary Fig. [85]6a, b).
   Scanning electron microscopy (SEM) observations revealed distinct
   migration patterns of cells on different interfaces. HUVECs adhered to
   the SpinMed filler exhibited faster migration than those guided by
   random cues (Fig. [86]2a, b). Moreover, cells cultured on SpinMed
   tended to form regular tubes on Matrigel (Fig. [87]2c, d). We performed
   in vivo implantation of SpinMed in rodent sciatic nerve defect models
   including a mouse 5-mm nerve defect model and a rat 15-mm nerve defect
   model, by which we evaluated the outcomes of axially vascular extension
   at 2 weeks after injury. The results indicated that vascular extension,
   along with the axial architecture of the SpinMed outperformed that of
   random architectures (Fig. [88]2e, f).

Fig. 2. SpinMed accelerates axial extension of vessels and increases
paracrine MMRN1 signaling.

   [89]Fig. 2
   [90]Open in a new tab

   a Detection of cell migration capacities after 24 h of culture on
   SpinMed filler or a random counterpart. Panel a created with
   BioRender.com released under CC BY-NC-ND. b Quantification of cells
   that migrated up to 3 mm under SEM (n = 8). c, d Representative images
   from the tube formation assay of HUVECs (c) and quantification of total
   tube length after culture with different interfaces (n = 6) (d). e, f
   Representative images of sciatic nerves harvested from mice or rats 2
   weeks after implantation (white full lines represent the initial stump
   terminal, and arrows indicate extended vessels) (e) followed by
   quantification of vascular extension rates (f) (n = 6). g Heatmap of
   differentially expressed genes among the SpinMed and random groups
   (n = 3). h KEGG analysis of differentially expressed genes between
   SpinMed and random interface cultures by Limma package. i, j Western
   blot images of MMRN1 expression in vascular endothelial cells (i) and
   quantification of the MMRN1/GAPDH ratio (n = 3) (j). k Simulated model
   of the MMRN1 protein and fibroin docking site, also considered an
   anchorage site. l Representative immunofluorescence images showing
   MMRN1 binding to random or SpinMed filler when cultured with HUVECs;
   the control images presented in the top right corner display the same
   without cultured cells. TCP tissue culture plate, Rand. random
   counterpart, Spin. SpinMed. Mean values are shown and error bars
   represent ± s.d., as analyzed by two-sided Student’s t-test in b, f and
   j, or one-way ANOVA with Tukey’s post hoc tests in (d). The experiments
   in e, k, and l were independently repeated at least three times with
   similar results.

   Proteomics was employed to explore the mechanisms of SpinMed treatment
   in nerve regeneration. Among all differentially expressed genes between
   SpinMed and random interface culture, multimerin 1 (MMRN1) was
   identified as a possible contributor to the angiogenic effects of
   SpinMed (Fig. [91]2g). Kyoto encyclopedia of genes and genomes (KEGG)
   analysis of differentially expressed genes indicated the potential
   involvement of the coagulation cascade and extracellular matrix
   (ECM)-receptor pathways (Fig. [92]2h). Western blot analysis identified
   increased MMRN1 expression in HUVECs cultured on the longitudinal axis
   of SpinMed filler (Fig. [93]2i, j). Complete failure of nerve
   regeneration was associated with misdirection of the blood vessels, and
   SC cords tended to follow the vascular regrowth direction towards the
   MMRN1 beads located in adjacent muscle and away from the bridge
   (Supplementary Figs. [94]6c–f and [95]7). In these cases, SCs migrated
   into the vascularized regions and deviated from their original
   directions, leading to the failure of axonal reconstruction. Moreover,
   MMRN1-blocking antibody application into the regenerative gap also
   compromised nerve healing in a rat nerve transection model, in which
   the number of differentiated SCs within the regenerative zone was
   positively associated with vessel volume (Supplementary Fig. [96]6g–j).
   Regarding the MMRN1-mediated molecular interaction, the structure
   simulated by the docking model revealed increased paracrine MMRN1
   signaling facilitated MMRN1-fibroin binding, and this interaction
   between the SpinMed filler and vessel paracrine factor established a
   stable anchorage for future endoneurial formation (Fig. [97]2k).
   Immunofluorescence assessment showed more MMRN1 bound to the SpinMed
   topological filler than to the random control when HUVECs were cultured
   on these scaffolds (Fig. [98]2l). Collectively, these results
   demonstrated that the topological cues of the SpinMed filler
   accelerated vascular extension, and helped to establish potential
   anchorages through increased paracrine MMRN1 signaling.

SpinMed-mediated host-guest crosslinking by high binding affinity between
MMRN1 or PECAM1 and fibroin

   To verify that fibroin used for fabricating the SpinMed filler had a
   high binding affinity for two major angiogenic proteins, MMRN1 and
   platelet endothelial cell adhesion molecule 1 (PECAM1), the docking
   models, intermolecular contact area, hydrogen bond (HB) number, and
   binding free energy (ΔG) values were dynamically simulated and
   measured. The molecular dynamics (MD) simulation displayed
   intermolecular docking models including hydrogen bond and van der Waals
   force binding sites (Fig. [99]3a–f). Compared with polycaprolactone
   (PCL) and collagen I, fibroin showed additional HB interactions, which
   was in agreement with its significantly higher affinity for MMRN1.
   Meanwhile, fibroin exhibited additional hydrophobic interactions with
   one or more hydrophobic residues of PECAM1 in comparison with PCL and
   collagen I and thus showed a significantly higher affinity for PECAM1.
   The ΔG determined by MD simulation also showed that the fibroin used in
   the SpinMed filler displayed a higher affinity for the two major
   angiogenic proteins (MMRN1 and PECAM1) than other commonly used
   materials (PCL and collagen I) (Supplementary Fig. [100]8;
   Supplementary Table [101]2). Quantification of the intermolecular
   contact area and HB number showed that fibroin was dramatically
   superior to PCL and collagen I during 0 to 100 ns, when, especially
   from 80 to 100 ns, the number of HBs between PECAM1 and fibroin was
   greater than that between PCL or collagen I and PECAM1
   (Fig. [102]3g–j). These results indicated that fibroin would act as an
   anchor by binding the regenerative vascular tissue with a higher
   affinity.

Fig. 3. Fibroin within SpinMed filler facilitates MMRN1 and PECAM1 binding.

   [103]Fig. 3
   [104]Open in a new tab

   a–c Detailed information on MMRN1 binding to fibroin (a), collagen I
   (b), and PCL (c) at docking sites. Red dashed line: hydrogen bond. d–f
   Detailed binding sites of PECAM1 docking to fibroin (d), collagen I (e)
   and PCL (f). Red dashed line: hydrogen bond. g, h Contact area of MMRN1
   (g) or PECAM1 (h) with fibroin, collagen I and PCL within 0–100 ns,
   calculated by molecular dynamics simulation. i, j The number of
   hydrogen bonds between MMRN1 (i) or PECAM1 (j) and fibroin, collagen I
   and PCL. PCL polycaprolactone, MMRN1 multimerin 1, PECAM1 platelet
   endothelial cell adhesion molecule 1. The computational simulations
   were independently repeated at least three times with similar results.

MMRN1 anchors contribute to the “endoneurium” for angiogenic-neurogenic
coupling

   To identify the processes of regenerative vessel-mediated neurogenesis
   and achieve direct angiogenic-neurogenic coupling, we performed a
   series of experiments, including proliferation, migration, and
   differentiation detection assays, in a coculture system of vascular
   endothelial cells and SCs (Fig. [105]4a; Supplementary Fig. [106]9a).
   The EdU assay showed an enhanced proliferative capacity for mouse
   primary SCs (mSCs) by topological induction but not for RSC96 cells
   (Supplementary Fig. [107]9b, c), and the migration rates of SCs
   determined by transwell assay were accelerated in topology-mediated
   vascular endothelial cell and SC coculture for both mSCs and RSC96
   cells (Supplementary Fig. [108]9d, e). SC homing and differentiation
   are necessary for remyelination. Vascular endothelial cells grown upon
   topological cues exerted significant effects on SC differentiation, as
   revealed by the branching SC ratio and branching length of SCs. Serum
   application was used as a positive control (Fig. [109]4b, c).

Fig. 4. MMRN1/HMMR/CDK1 complex contributes to SpinMed-induced
angiogenic-neurogenic coupling.

   [110]Fig. 4
   [111]Open in a new tab

   a Schematic illustration of the HUVEC and SC coculture system. Panel a
   created with BioRender.com released under CC BY-NC-ND. b, c
   Representative images of SC (mSC or RSC96 cell) differentiation when
   cocultured with HUVECs on various surfaces (b) and quantification of
   branching cell percentages and average branching lengths (n = 4) (c). d
   Mass spectrogram of identified HMMR and CDK1 binding to MMRN1 after
   immunoprecipitation (IP) (n = 3). e Molecular docking model of MMRN1,
   HMMR, and CDK1, as well as docking sites by hydrogen bonds. f IP
   analysis of binding among MMRN1, HMMR, and CDK1 using the HEK293T cell
   line. g, h Representative immunofluorescence images of DRG neurons
   cultured with Schwann cells after various treatments (g) and
   quantification of axon extension lengths (n = 5) (h). TCP, tissue
   culture plate. HMMR, hyaluronan mediated motility receptor. CDK1,
   cyclin-dependent kinase 1. Mean values are shown and error bars
   represent ± s.d., as analyzed by one-way ANOVA with Tukey’s post hoc
   tests in (c) and (h). The experiments in f were independently repeated
   three times with similar results.

   Immunoprecipitation mass spectrometry (IP-MS) and proteomics were
   cooperatively utilized to investigate the receptors of MMRN1 and the
   response elements located on SCs. We found that hyaluronan mediated
   motility receptor (HMMR) and cyclin-dependent kinase 1 (CDK1) were
   predicted to be the receptors of MMRN1 through direct interactions
   (Fig. [112]4d; Supplementary Fig. [113]9f, g; Supplementary
   Fig. [114]10). Subsequently, the intermolecular docking model was
   examined to illustrate their interactions (Fig. [115]4e; Supplementary
   Fig. [116]10). Several tissue or cell types were submitted to detect
   HMMR and CDK1 expression, high levels of which were predominant in SCs
   among potential responsive tissues of peripheral nerves (Supplementary
   Fig. [117]9h). We further employed the co-IP assay to confirm the
   interactions among MMRN1, HMMR and CDK1 in the HEK293T cell line, a
   transformed human embryonic kidney cell line, verifying the
   ligand-receptor relationships between MMRN1 and HMMR, MMRN1 and CDK1,
   as well as HMMR and CDK1 (Fig. [118]4f). In addition, we analyzed the
   involved candidate signal transduction pathways by gene set enrichment
   analysis (GSEA), and found that the coagulation cascade, ECM receptor
   interaction and oxidative phosphorylation primarily participated in SC
   activities associated with remyelination (Supplementary Fig. [119]9i),
   and this effect was likely to be regulated by the activation of the
   focal adhesion kinase (FAK) and extracellular-signal regulated kinase
   (Erk) that FAK/Erk1/2-MAPK signaling pathway (Supplementary
   Fig. [120]9j, k). We further evaluated the neurogenic effects driven by
   conditioned media (CM) collected from SC cultures. CM derived from the
   SpinMed group altered the axonal length of dorsal root ganglion (DRG)
   neurons (Fig. [121]4g, h). Therefore, the MMRN1/HMMR/CDK1 complex drove
   SC remyelination during SpinMed-mediated angiogenic-neurogenic
   coupling.

Accelerated nerve structural and functional restoration by SpinMed in rats

   The biosafety of SpinMed was first evaluated after implantation into
   rats, and the results showed that there was no significant cell
   toxicity or organ injury (Supplementary Fig. [122]11). During the
   period of 12-week follow-up, we performed gait analysis for locomotor
   activity examination at 8 and 12 weeks after implantation to evaluate
   the in vivo therapeutic effects of SpinMed compared with other grafts.
   The denervated paws were prone to shrinking, and the distance between
   digits was reduced. The rats that underwent SpinMed implantation
   exhibited improved toe spreading and gait than that from hollow and
   random groups at both 8 and 12 weeks (Fig. [123]5a; Supplementary
   Fig. [124]12a). The sciatic nerve function index (SFI) was employed to
   represent the global conditions of motor nerve restoration. Although
   the SFI evaluated in the SpinMed implantation group did not exceed that
   in the autograft group, the results showed that SpinMed dramatically
   restored the motor function of the sciatic nerve comparable to
   autograft (Fig. [125]5b; Supplementary Fig. [126]12b). We further
   evaluated the recovery of sensory function by assessing responses to
   thermal stimulation and the presence of mechanical allodynia
   (Supplementary Fig. [127]12c). In the hot plate test, rats receiving
   SpinMed implantation and autograft recovered more quickly from thermal
   nociception at 12 weeks, showing a reduced withdrawal response time
   compared with other controls, while no differences in mechanical
   allodyniano, revealed by Von Frey test, were observed among these
   groups (Supplementary Fig. [128]12d, e). We also assessed muscle
   responses to nerve-derived electrical pulses to reflect the status of
   the target gastrocnemius muscle. Overall, SpinMed improved
   electrophysiological recovery compared with its random counterparts and
   hollow control and achieved therapeutic outcomes similar to those of
   the “gold standard treatment” autograft at postoperative 12 weeks, as
   determined by the nerve conduction velocity (NCV) and recovery index of
   compound muscle action potential amplitude (CMAP) (Fig. [129]5c–e).
   These results suggest that SpinMed application is a promising nerve
   tissue engineering strategy.

Fig. 5. SpinMed restores peripheral nerve morphology and function in rodents.

   [130]Fig. 5
   [131]Open in a new tab

   a, b Schematic illustration of gait evaluation (a, top panel),
   representative images of footprints (a, bottom panel), and
   quantification of sciatic function index (n = 5) (b). Panel a and b
   were created with BioRender.com and released under CC BY-NC-ND. c–e
   Representative electromyographic patterns detected in each group
   (n = 5) (c) and quantitative analysis of nerve conduction velocity (d)
   and CMAP amplitude recovery index (e). f–h Representative TEM images of
   regenerative nerve tissue (transverse view, left panel) and zoom-in
   images (right panel) from various groups (n = 8) (f). g, h
   Quantification of parameters for the nerve ultrastructural analysis,
   including the axon diameter and myelin sheath thickness, shown by the
   g-ratio (g), and linear regression of the g-ratio and axon diameter
   (n = 8) (h). i Representative immunofluorescence images of Nestin and
   GFAP expression in the spinal cord (L5) (n = 5). j, k Gross view of
   gastrocnemius harvested from various groups (n = 5) (j) and
   quantification of gastrocnemius wet weight compared to contralateral
   counterpart (k). CMAP, compound muscle action potential. Mean values
   are shown and error bars represent ± s.d., as analyzed by one-way ANOVA
   with Tukey’s post hoc tests in b, d, e, g, h and k.

   Histological assessment was further carried out (Supplementary
   Fig. [132]13), along with transmission electron microscopy (TEM)
   examination of the cross-section as the “gold standard” observation
   method. We observed the regenerated sciatic nerve morphology by TEM
   (Fig. [133]5f) and quantified the myelination and axonal regeneration
   by the myelin sheath thickness and axonal diameter, and calculated
   g-ratio (Fig. [134]5g, h). The TEM evaluation revealed the presence of
   a higher density of myelinated fibers and a thicker myelin sheath
   within the lumen in the SpinMed group, comparable to those in the
   autograft group at postoperative 12 weeks. During the regenerative
   process, we examined the vessel, myelin, and axon at multiple endpoints
   (Supplementary Fig. [135]12f), the results showed that enhanced
   angiogenic activities and vessel volume were observed from SpinMed
   group samples at postoperative 4 weeks, as well as vessel-myelin-axon
   regrowth coupling at both 4 and 8 weeks (Supplementary Fig. [136]12g).
   Axonal regeneration and myelination were also assessed by
   immunofluorescence staining for β-tubulin III (Tuj1)/myelin basic
   protein (MBP) or neurofilament (NF) 200/S100β (Supplementary
   Fig. [137]14a–c). SpinMed implantation led to denser nerve fibers and
   myelinated nerve fibers near the autograft (Supplementary
   Fig. [138]14d). Similar results were also found throughout the entire
   length of regenerated nerves (Supplementary Fig. [139]14e, f).

   Axonal transport drives the signal relay and nutrient supply between
   the nervous system and the target tissue. FluoroGold (FG) tracer
   injection followed by histological detection of DRG neurons and spinal
   cord anterior horn (AH) tissue was employed to detect the transport
   capacities of regenerative axons (Supplementary Fig. [140]14g).
   Autofluorescence images showed that more positive sensory neurons
   within the DRG and motor neurons within AH regions after SpinMed
   implantation than those from hollow and random counterpart groups
   (Supplementary Fig. [141]14h), and furthermore, quantification analysis
   supported the comparable effects of the SpinMed graft and autografts,
   demonstrating favorable neurofiber reconnection and transport function
   restoration after 12-week SpinMed repair (Supplementary Fig. [142]14i).
   These injury signals and component alterations may also result in
   spinal cord remodeling. Nestin and glial fibrillary acidic protein
   (GFAP) were labeled to determine the number and distribution of neurons
   and glial cells within the entire spinal cord (Fig. [143]5i;
   Supplementary Fig. [144]15). Muscles are highly innerved to exert
   contract function, and long-term denervation leads to muscle atrophy
   and dysfunction. SpinMed application to a nerve defect mostly
   maintained the morphology and wet weight of the gastrocnemius examined
   at 12 weeks, but the effect was slightly inferior to that of autograft
   application (Fig. [145]5j, k; Supplementary Fig. [146]16).

Complicated nerve defect repair with reliable biosafety via an individualized
strategy in canines

   Commercial nerve grafts used in current clinical practice often fail to
   cure complicated nerve defects due to mismatched size and shape;
   however, the SpinMed graft designed in this study addresses, at least
   in part, these issues. The individualized SpinMed graft and other
   controls were fabricated and implanted into a beagle sciatic defect
   model for 12 months (Fig. [147]6a, b; Supplementary Fig. [148]17a). We
   monitored the biosafety of SpinMed by detecting the serum levels of
   aspartate aminotransferase (AST), alanine aminotransferase (ALT),
   lactate dehydrogenase (LDH) and blood urea nitrogen (BUN) at
   postoperative 4 weeks, 16 weeks and 12 months, and there was no
   significant alteration of the serum levels of these molecules
   (Supplementary Fig. [149]18). Furthermore, the SpinMed could achieve
   nearly complete degradation without obvious organ toxicity as revealed
   by long-term followup (Supplementary Fig. [150]18). Magnetic resonance
   imaging (MRI) detection revealed that the individualized SpinMed grafts
   restored branched nerve defect continuity well, as indicated by
   complete structural reconstruction, while grafts with straight
   structures only germinated nerve trunks but not clear branches
   (Supplementary Fig. [151]17b). Electrophysiological assessment showed
   that CMAP amplitude recovery index was improved by the individualized
   SpinMed grafts at postoperative 16 weeks and 12 months, indicating that
   precise structural reconstruction may also facilitate functional
   restoration (Fig. [152]6c; Supplementary Fig. [153]17c). Dynamic motor
   evaluation of the right hind limb joints of beagles showed greater
   flexibility in the SpinMed group, as indicated by the ankle joint and
   metatarsophalangeal joint flexion angle, and locomotor capacity grade
   (Supplementary Fig. [154]19; Supplementary Movie [155]1 to [156]4).

Fig. 6. SpinMed, as an individualized preclinical product repairs complicated
nerve defects in canines with long-term followup.

   [157]Fig. 6
   [158]Open in a new tab

   a Illustration of surgical procedures performed in complicated sciatic
   nerve defects in the “Single” “SpinMed” “Decellularized graft” and
   “Autograft” groups. b Surgical images of application methods in various
   groups, with labels for sciatic nerve (1), tibial nerve (2), and common
   peroneal nerve (3). c Analysis of nerve conduction velocity and CMAP
   amplitude recovery index by electrophysiology assessment at
   postoperative 8 and 12 months (n = 3). d, e Illustration of the sample
   collection positions at postoperative 12 months for TEM detection (d),
   where transverse ultrastructural views were obtained (n = 3) (e). Panel
   d created with BioRender.com released under CC BY-NC-ND. f, g
   Quantification of parameters for the nerve ultrastructural analysis at
   the P3 and P4, including the myelin sheath thickness (f) and the
   g-ratio (g) (n = 3). h, i Representative images of H&E (upper row) and
   toluidine blue (lower row) staining for regenerative nerves collected
   from the P2 at postoperative 12 months (n = 3) (h), and quantification
   of the number of myelinated nerve fibers (i). Sing. single counterpart,
   Spin. SpinMed, Decell. decellularized graft, Auto. autograft. Mean
   values are shown and error bars represent ±s.d., as analyzed by one-way
   ANOVA with Tukey’s post hoc tests in (c, f, g, and i).

   We selected 4 representative sites for TEM observation and toluidine
   blue (TB) staining of the transverse ultrastructural view at
   postoperative 16 weeks and 12 months (Fig. [159]6d; Supplementary
   Fig. [160]17d). At postoperative 16 weeks, the specimens collected from
   the individualized SpinMed group showed comparable ultrastructure to
   autografts in the proximal terminal (P1), with some even superior to
   autografts and the other counterparts at branching sites. The
   regenerative nerve collected from SpinMed group exhibited better myelin
   sheath thickness and g-ratio within P3 and P4 (Supplementary
   Fig. [161]17e, f; Supplementary Fig. [162]20). Moreover, neural
   integrity restoration by SpinMed also showed superiority to single
   counterpart, decellularized graft, and even autograft at postoperative
   12 months, as revealed by TEM parameter analyses, in which myelin
   sheath thickness displayed significant difference (Fig. [163]6e–g;
   Supplementary Fig. [164]21). Histology for regenerative nerve and
   innervated gastrocnemius muscle was performed to evaluate neural
   morphology and function. We found the number of myelinated nerve fibers
   was comparable to autograft, and targeted muscle atrophy was also
   significantly improved compared with decellularized graft application
   (Fig. [165]6h, i; Supplementary Fig. [166]22). These findings
   collectively demonstrated the practical value and translational
   potential of SpinMed.

Discussion

   We conducted an individualized design process without any loading
   components and synthesized a multilevel neurium-mimetic nerve graft
   named “SpinMed” composed of modified epineurium, topological
   perineurium, and spontaneously formed endoneurium to allow the most
   adaptation for nerve regeneration. The outer structure mimicking the
   epineurium serves as a selectively physical barrier for limiting access
   to cicatricial fibrosis following injury, as abnormal tissue invasion
   has been identified to inhibit neural repair^[167]32,[168]33. Enhanced
   mechanical properties conferred by the hollow PCL sheath beyond those
   of the native counterpart supported stump regrowth. A porous interface
   (pore size of approximately 1 μm), acquired by phase separation,
   permitted nanoparticles to cross but rejected cells in the radial
   direction. This method easily achieved porous architecture placement
   during solvent evaporation, and notably, this procedure saved a massive
   time and work compared to previous techniques^[169]34. Consequently,
   the enhanced epineurium allowed sufficient space for nerve growth,
   superior to allogeneic grafts, and importantly, this modified design
   based on regenerative preferences inspired us to revolutionize the