Bioinspired hierarchical fibre bundles from microfluidics for nerve regeneration

Abstract

Fibre-shaped scaffolds have demonstrated value in nerve tissue engineering because of their ability to direct cell alignment and remodel intrinsic fibrous tissues. Research in this aspect focuses on the creation of microfibres with nerve-mimicking morphology and spatial neurocyte guidance to improve the regeneration effects. Here, inspired by the anatomical structure of the nerve that is composed of parallelly assembled nerve fibres, we present novel hierarchical fibre bundles from microfluidics for peripheral nerve regeneration. The GelMA/alginate filaments encapsulated in hydrogel microfibre shells were consecutively generated from the coaxially assembled capillary microfluidic device, and their composition and morphology could be further tuned by changing the microchannel and fluid constructions. Based on the resultant microfibres, we have demonstrated that they could provide a biomimicking cell culture environment and promote the formation of nerve bundles. In addition, the hierarchical fibre bundles showed an improved regeneration capability of sciatic nerves when they were transplanted into sciatic nerve defect SD rat models. Thus, hierarchical fibre bundles are potentially valuable for clinical peripheral nerve regeneration.

Graphical abstract

Highlights

• •Fibrous scaffolds have shown potentials in nerve tissue engineering for their ability to guide cell alignment.
• •Inspired by the anatomical structure of the nerve, hierarchical fibre bundle is firstly fabricated from coaxial microfluidics.
• •The hierarchical microfibre bundles can provide a biomimicking cell culture environment and promote the formation of nerve bundles.
• •The proposed hierarchical microfibre bundle are valuable for clinical applications of nerve regeneration.

1. Introduction

Fibrous structures are one of the most common and representative constructions in human tissues, such as muscle fibres [1,2], blood vessels [3,4], nerves [5,6], and other similar anatomic architectures [[7], [8], [9], [10], [11]]. Intrigued by these assemblies, numerous fibre-shaped scaffolds have been explored as they can provide three-dimensional (3D) environments to physically mimic native tissues [[12], [13], [14], [15], [16]] and physiologically promote cell interactions [[17], [18], [19], [20], [21], [22]]. Some of these scaffolds are designed with hollow structures [[23], [24], [25], [26]] or special surface morphologies, like bulges, pillars, and grooved patterns, to facilitate the spreading, proliferation, and alignment of nerve cells [27,28], and guide and elongate the axons to form nerve pathways [[29], [30], [31]]. Although extensive progress has been achieved in nerve fibre rebuilding by using these existing fibres, the generally concave surface of the fibres cannot protect the cells from flow-induced shear stress during cell cultivation. In addition, most current scaffolds fail to reconstruct complex hierarchical nerve structures, which is essential to neurocyte culture and nerve fibre modelling. Therefore, the creation of microfibres with nerve fibre-mimicking morphology and spatial neurocyte guidance to realize nerve bundle regeneration is still highly expected.

In this paper, inspired by the anatomical structure of nerves, which consist of parallelly assembled nerve fibres, we present a novel hierarchical fibre bundle fabricated by microfluidics, aiming to improve peripheral nerve regeneration outcomes, as illustrated in Fig. 1. Microfluidics is promising in functional materials generation due to its ability to deal with multiple phases of fluids in a precise and controllable manner [[32], [33], [34], [35]]. Based on this fantastic technology, many kinds of microfibres, including those with complicated morphologies and compositions, have been achieved and served as drug delivery media or in vitro models in tissue engineering [[36], [37], [38]]. However, to our knowledge, microfibres with such hierarchical microstructures have never been generated from the microfluidic fabrication method, and their biomedical engineering applications, especially in nerve regeneration, remain unexplored.

Fig. 1.

Schemes illustrating the generation and application of the hierarchical microfibre bundle. (a) Inspired by the hierarchical anatomical structure of the nerve, the microfibre bundle can be prepared via a multi-injection microfluidic spinning method. (b) The microfibre bundle can be applied to the peripheral nerve repair, where the filaments of the bundle can guide the alignment of the nerve cells to facilitate the formation of cell bundle during nerve regeneration.

Thus, we herein employed hierarchically assembled microfluidic devices with multi-parallel glass capillaries as injection channels to create these biomimicking hierarchical fibre bundles for peripheral nerve regeneration (Fig. 1). The microfluidic devices were constructed by inserting a multi-barrel capillary array into three coaxially-aligned cylindrical capillaries. Alginate/Gelatin-Methacryloyl (GelMA)/Extracellular matrix (ECM) solutions were injected into the devices through surrounding barrels of the capillary array, the alginate solution flowed in the outer capillary, while calcium chloride (CaCl₂) solutions were from the central barrel of the capillary array as well as the middle and outermost cylindrical capillaries. These independent injections together with the rapid crosslinking of alginate made the simultaneous formation of encapsulated filaments and hydrogel microfibre shells. Benefitting from the flexible configuration of the assembled microfluidics, the number together with the compositions of encapsulated filaments could be well adjusted. Based on these resultant microfibres, we have demonstrated that the generated hierarchical fibre bundle could provide a biomimicking cell culture environment and promote the formation of cell bundles. The free-standing filaments offer physical guidance for cell seeding, migration, and alignment, facilitating the regeneration of nerve in a flexible way compared to traditional fibrous scaffolds. In addition, to evaluate the hierarchical fibre bundles, the transplantation into SD rats with 10 mm sciatic nerve defects was performed, revealing enhanced sciatic nerve regeneration capability compared to hollow microfibres. These results indicate that hierarchical fibre bundles are valuable for clinical applications of nerve regeneration.

2. Results and discussion

In a representative experiment, to achieve hierarchical structure, a microfibre containing one filament was attempted in a co-axially aligned multi-level microfluidic device (Fig. S1a). In this device, the encapsulated filament was generated from the inner seven-barrel capillary array, and the microfibre sheath was formed between the middle and the outer cylindrical capillary (Fig. 2a). To be specific, the central barrel of the capillary array was used for the introduction of the alginate/GelMA/ECM mixture (the second phase, P2); the middle phase (CaCl₂ solution, P3) flowed in the interval between the capillary array and the middle cylindrical capillary; the out phase of alginate (P4) was flowing inside the outer cylindrical capillary. The hierarchical fibre was then formed in two stages (Fig. 2b). First, fibre shell was formed based on fast gelation between laminar flows of P3 and P4; second, the inside filament was generated when the mixture (P2) jetted from the capillary array and encountered surrounding P3. Because the gelation between Ca ions and alginate was fast and the gelation took place simultaneously at both interfaces of the middle phase, the pre-gel of fibre shell was rapidly formed as well as the inner filament, which replicated the geometry of the microfluidic channel well. The later solidification of fibre shell was realized by immersing the outlet in the collection pool containing CaCl₂, and the GelMA composition was polymerized under exposure to ultraviolet light. The collected hierarchical fibre containing one filament inside was rinsed twice and observed under a microscope. It demonstrated that the hydrogel shell was uniform, suggesting a stable co-axial flow formed during the fibre fabrication and fast reaction took place to solidify the shell (Fig. 2c).

Fig. 2.

The generation of hierarchical fibre bundle. (a) Scheme illustrating the multi-level microfluidic device used for one filament-containing hierarchical fibre generation. (b) Real-time image of the in situ formation of the hierarchical fibre during microfluidic spinning. (c) Bright and fluorescence field microscopic images of the hierarchical fibre. The scale bar is 500 μm. (d) Diagram of microfluidic spinning of hierarchical fibre bundle with six filaments. (e) Bright and fluorescence field microscopic images of the hierarchical fibre bundle. The scale bar is 500 μm. (f-g) Relationships between filament diameter and (f) P1 flow rate, (g) P2 flow rate, respectively. (h-i) Relationships between shell thickness and (h) P3 flow rate, (i) P4 flow rate, respectively.

Benefitting from the flexible phase manipulation, the microfluidic device could also be utilized for the creation of hierarchical fibre bundle, which encapsulated six filaments in a similar manner. In this case, the first phase of CaCl₂, named P1, was introduced into the central barrel as the innermost phase; P2 was pumped simultaneously into the surrounding six barrels of the capillary array; P3 was injected through the region between the capillary array and the middle cylindrical capillary; P4 flowed inside the outer cylindrical channel; the last phase of CaCl₂ flowed in the outermost cylindrical capillary (Fig. 2d). The introduction of the innermost CaCl₂ solution separated the surrounding six streams and prevented their merging. From the microscopic images, the uniform fibre shell and six encapsulated filaments suggested the successful and stable generation of the hierarchical fibre bundle (Fig. 2e).

Fortunately, the diameter of encapsulated filaments and the fibre shell thickness could be precisely controlled by adjusting different phase flow rates. Specifically, when the flow rates of P2, P3, and P4 were kept constant, increasing the P1 flow rate could reduce the filament diameter (Fig. 2f). In contrast, increasing the P2 flow rate will enlarge it while maintaining the flow rates of P1, P3, and P4 (Fig. 2g). This could be attributed to the fact that the diameter of the co-flow of P1 and P2 would be mainly affected by P1, P2, and P3 flow rates. When P3 was fixed, the diameter of the co-flow, which could be approximately considered as the filaments diameter, was only related to the flow rates of P1 and P2. Specifically, the increase of P1 flow rates would cause the separation of multiple streams of P2, and thus decrease the stream diameter, resulting in decreased filaments diameters. Conversely, the increase of P2 flow rate would induce the increased filament diameters in this case.

Apart from the adjustable filament diameter, the shell thickness could also be changed by changing the flow rates of P3 and P4. It presented that the shell thickness would show a negative relationship with P3 flow rate, while it showed a positive relationship with P4 flow rate. As depicted in Fig. 2h, when the flow rates of P1, P2, and P4 were held constant, the increasing P3 flow rate led to a reduction in the shell thickness. Additionally, the increase in the shell thickness could result from the increasing P4 flow rate when the others remained fixed (Fig. 2i). This could be owing to that the shell thickness is mainly affected by the diameters of the outer alginate and the middle CaCl₂ streams. The increased flow rate of P3 would result in the increased diameter of CaCl₂ stream, and the increased flow of P4 would cause an increase in the diameter of alginate stream when fixing other microfluidic parameters, the shell thickness could be consequently adjusted by these two flow rates.

Based on biomimetic and practical considerations, we selected hierarchical fibre bundles with filaments diameters of about 50 μm to align with the scale of Schwann cells in the peripheral nervous system, and the cavity diameter of about 500 μm to facilitate cell infiltration and nutrient diffusion. This architecture is selected for its high fabrication reproducibility, structural stability during manipulation, and capability to promote cell adhesion and provide topographical guidance during culture and subsequent nerve repair. The morphology was further examined by scanning electron microscopy (SEM), which confirmed the presence of a single filament freely encapsulated within a hollow microfiber, both exhibiting desired dimensions (Fig. S1b). The mechanical property of the hierarchical fibre bundle, characterized by the tensile stress-strain curves, contribute to the overall structural stability. As shown in Fig. S2, the two-stage fracture process where the hollow microfibre ruptured first at approximately 60% strain followed by the encapsulated filament at approximately 150% strain suggested this structure can effectively bear and distribute deformation during implantation and provide stable physical guidance for cell seeding, migration, and alignment during nerve regeneration.

The multiple inlets of the capillary array posed opportunities to different numbers and configurations of encapsulated filaments in the resultant hierarchical fibre bundle. When two streams of P2 were introduced symmetrically into the capillary array, the hierarchical fibre containing two filaments could be obtained (Fig. 3a). For visualization, nanoparticles with red fluorescence were added to P2 when preparing the solutions. Similarly, to obtain hierarchical fibre containing three, four, and five filaments, corresponding numbers of P2 streams were introduced to the capillary array, respectively (Fig. 3b–3d). It is worth noting that the hierarchical fibre bundle containing 8 or 12 filaments could be achieved by changing the seven-barrel capillary array with nine-barrel or thirteen-barrel capillary array during the device preparation (Fig. 3e and 3f). The images indicated that the sheath of the hierarchical fibre was generally uniform, and the filaments were almost straight and separated. This should be attributed to the stable laminar flows formed inside the multi-level microfluidic device together with the fast gelation of alginate. Additionally, by introducing different kinds of inner mixtures, the composition of encapsulated fibre bundles could be modulated. The fluorescence images of these microfibres indicated that the encapsulated filaments could have their own chemical composition (using different fluorescence to distinguish them), which replicated that of the initial injecting flows (Fig. 3g–3l and Fig. S3–S4). By this means, hierarchical fibre bundles with complicated compositions and filament encapsulations could be generated, mimicking the anatomical structure of the nerve and showing great potential in nerve regeneration.

Fig. 3.

Hierarchical fibre bundles with different features. (a-f) Schematic illustration of hierarchical fibre bundles containing (a) two, (b) three, (c) four, (d) five, (e) eight, and (f) twelve filaments, respectively, the according injection patterns as well as their bright-field and fluorescence microscopy images. (g-i) Schematic illustration of hierarchical fibre bundles containing (g) one green and one red, (h) two green and two red, (i) three green and three red filaments, respectively, the corresponding injection patterns as well as their bright-field and fluorescence microscopic images. (j) Schematic diagram showing hierarchical fibre bundles containing red, green, and blue filaments, the injection mode, and corresponding microscopy images. (k) Diagram of hierarchical fibre bundles containing two red, two green, and one blue filaments, the injection mode, and corresponding microscopy images. (l) Diagram of hierarchical fibre bundles containing two red, two green, and two blue filaments, the injection mode, and corresponding microscopy images. The scale bar is 500 μm.

The in vitro cytocompatibility of the resultant fibre bundle was studied by coculturing them with Schwann cells for 24 h and 72 h. It indicated that both the alginate hollow microfibre and the hierarchical fibre bundle exhibited excellent biocompatibility and effectively promoted cell proliferation after 3 days of cultivation, as compared to the control group (Fig. 4a). To be specific, the hierarchical fibre bundle group even demonstrated a slight improvement in cell viability, likely due to the enhanced cell-adhesion properties of ECM. Additionally, the ability to induce the cell migration of these microfibres were also investigated by using a transwell assay. After co-culturing the microfibres with Schwann cells and performing crystal violet staining, the amount of migrated Schwann cells per field in the hierarchical fibre bundle group exceeded that in both the control and hollow microfibre groups (Fig. 4b and 4c). This suggested that the addition of ECM in hierarchical fibre bundle would promote the migration of Schwann cells during nerve cell cultivation.

Fig. 4.

Cell study of the hierarchical fibre bundle. (a) Cell viability after 24 h and 72 h of co-cultivation with microplate (control), hollow microfibre, and hierarchical fibre bundle. (b) Migrating Schwann cells count per field in each group. (c) The crystal violet staining images of migrated Schwann cells in each group. The scale bars are 100 μm (left column), and 50 μm (right column), respectively. (d) Fluorescence microscopic images of Schwann cells cultured in the hierarchical fibre bundle. The inner wall of the shell is marked by the white dashed line. The scale bars represent 150 μm. (n ≥ 3, ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, n.s.: not significant).

The hierarchical fibre bundle features with a microfibre sheath that can protect the cells from shear force-induced damage during cultivation, and several filaments to form bundles that can guide the nerve cell alignment. To prove this, the Schwann cells were introduced into the cavity of the hierarchical fibre bundle, seeded them on the containing filaments followed by cultivation for 5 days. In this process, the free-standing filaments were cut longer than the fibre sheath to ensure their encapsulation and facilitate stabilization on culture dishes. The calcein-green staining images showed that the cells could proliferate inside the hydrogel shell, aligned on the surface of the encapsulated filaments, and formed nerve bundles (Fig. 4d and S5). Specifically, compared to those located on the fibre shell and directly seeded in hollow microfibre, the cells adhered to the filaments displayed a relatively extended state and had orientation angles below 30°, demonstrating the guiding ability of the filaments. Moreover, the free-standing filaments would attach to each other during the cultivation, providing the opportunity for forming cell aggregates to facilitating the network formation. Taken together, these results indicated the nerve regeneration potential and the alignment of the nerve cells of this hierarchical fibre bundle.

The peripheral nerve regeneration capability of the resulting fibre bundles was investigated by implanting them into SD rats with 10 mm sciatic nerve defects (Fig. 5a). To keep the original fibre shell, cavity and encapsulated filaments of the hierarchical fibre bundle, a biocompatible silicone conduit was used to provide external support and align two fibres in parallel during the surgery. Although the silicone conduit may require surgical removal in clinical application compared to other biodegradable alternatives, the use of silicone here could also eliminate degradation-related variables. The rats were randomly assigned to five groups, including the sham, peripheral nerve injury (PNI), silicone conduit, hollow microfibre, and hierarchical fibre bundle group. Walking track analysis was performed over the 8 weeks following surgery (Fig. S6). Based on the footprints and formula, the sciatic function index (SFI) was thus calculated and used as a reliable score assessing the motor functional recovery (Fig. S7a). It indicated that during the 8 week of regeneration process, the SFI of each group increased, and the increase was the most obvious in the hierarchical fibre bundle group, showing the best recovery (Fig. S7c). Apart from this, electrophysiology study was also carried out 8 weeks post-surgery to explore the function recovery in each group (Fig. S7b). In this study, the compound action signals in response to sciatic nerve stimulation were measured and analyzed. It revealed that although compound action potential showed in each group, that in the hierarchical fibre bundle group was the closest to that in the sham group, indicating the best function recovery (Fig. S7d). This could also be learnt from the conduction velocity and latency analyses, where the hierarchical fibre bundle group showed the most increased conduction velocity together with the most reduced latency among experimental groups (Fig. S7e–S7f).

Fig. 5.

Application of fibres in peripheral nerve regeneration. (a) Diagram and digital images suggesting the nerve injury model and transplantation site of different groups of fibres. (b) Representative fluorescence microscopy images of NF-200, S100β, and DAPI-stained nerve longitudinal sections for (i) sham, (ii) PNI, (iii) Silicone conduit, (iv) Hollow microfibre, (v) Hierarchical fibre bundle groups, along with their corresponding merge images. (c) Representative fluorescence microscopy images of NF-200, MBP, and DAPI-stained nerve cross-sections for (i) sham, (ii) PNI, (iii) Silicone conduit, (iv) Hollow microfibre, (v) Hierarchical fibre bundle groups and their corresponding merge images. (d, e) Statistical analysis of the positive expression levels (%) of (d) NF-200, and (e) S100β reacted proteins for each group. (f) Analysis of myelin sheath thickness for each group. The scale bars represent (b) 50 μm, (c) 50 and 20 μm, respectively. (n ≥ 3, ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, n.s.: not significant, compared to PNI).

According to the above-mentioned results, the hierarchical fibre bundle showed the best nerve regeneration improving ability compared to other materials. This enhanced regenerative capacity is likely attributed to its spatial induction during nerve regeneration, which benefits from both the designed hierarchical architecture and its slow degradation in vivo. As shown in Fig. S8, the hierarchical fibre bundle retained approximately 80% of its original weight after two weeks of subcutaneous implantation, confirming a slow degradation rate during the early post-surgical period. Immunofluorescence staining further revealed robust axonal regeneration and remyelination, providing additional evidence of the improved nerve regeneration outcome. As shown in the fluorescence microscopy images of longitudinal section of the regenerated nerve, axons (NF-200 marked, green) and myelination (S100β marked, red) aligned directionally in the hierarchical fibre bundle group, in the contrast to other groups (Fig. 5b). This directional alignment could be attributed to the guiding capability of the filaments in the hierarchical fibre bundle, while the hollow microfibre, lacking filaments in its channel cannot provide the same directional guidance for the nerve cells during the repair process. Additionally, the fluorescence microscopy images of transversal section of the regenerated nerve also showed improved remyelination effect of hierarchical fibre bundle, where more intact myelination (MBP marked, red) were observed (Fig. 5c). Statistically, the upregulated expression of NF-200-positive and S100β-positive protein, as well as increased myelin sheath thickness in hierarchical fibre bundle group, indicated the facilitation ability of hierarchical fibre bundle during nerve regeneration application (Fig. 5d and 5e).

Given the critical role of the immune response in tissue regeneration, we also evaluated post-surgical inflammation and angiogenesis across different groups. Immunofluorescence staining of CD68, a marker of inflammation, revealed that no severe inflammatory response was observed in any group (Fig. S9a and S9c). Notably, the expression levels of CD68 were downregulated in both hollow microfibre and hierarchical fibre bundle groups, suggesting that these implants possessed the potential to modulate the inflammatory response. Such low-level inflammatory response along with low levels of hemolysis ratios of alginate, GelMA and their mixture suggested good short- to medium-term biocompatibility of the implanted hierarchical fibre bundles (Fig. S10). While comprehensive long-term organ toxicity studies were not performed, the biocompatible nature of alginate and GelMA supports the expectation of minimal systemic risk. Angiogenesis, essential for supplying oxygen and nutrients to regenerating nerve tissue, was assessed by CD31 and α-SMA staining (Fig. S9b and S9d). A marked increase in CD31 expression was observed in the hierarchical fibre bundle group, indicating that its promotion of nerve regeneration may be attributed, at least in part, to enhanced angiogenesis.

The treating performances for each group were also evaluated through gastrocnemius amyotrophy analysis, which serves as an indirect method to assess the nerve recovery. Gastrocnemius atrophy was first measured by the weight ratio (Fig. S11). It was observed that the weight loss decreased in hierarchical fibre bundle group compared to the PNI, silicone conduit, and hollow microfibre groups, indicating a superior muscular atrophy inhibition ability. This finding was further confirmed by examining the cross-sectional areas (CSAs) of muscle fibres after hematoxylin-eosin (H&E) staining (Fig. 6a and 6c). According to images and relative statistics, obvious muscle atrophy was observed in the PNI group, suggesting a successful model building. Whereas, the CSAs of muscle fibres distinctively increased in hierarchical fibre bundle group, demonstrating effective muscle atrophy prevention and nerve regeneration outcomes. In addition, evident signs of muscle atrophy could be seen in Masson's trichrome staining image of PNI group, including the disordered muscle fibres and collagen fibres deposited (Fig. 6b). It should be mentioned that these were attenuated in the hierarchical fibre bundle group, where reduced collagen volume fraction was observed, and the muscle fibres were significantly bulked up compared to PNI group (Fig. 6d). Taken together, the bioinspired hierarchical fibre bundle exhibited favourable nerve regeneration promotion performances and proved their application potentials in peripheral nerve injury treatment. In addition, the underlying molecular mechanisms, such as the activation of neurotrophic signalling pathways, the effects of environmental parameters, such as blood flow, pressure, and concentrations, remain to be elucidated. Future work could focus on the investigation on the effects of biological environment and the expression of nerve regeneration-related genes and to establish the mechanistic basis for the observed regenerative effects.

Fig. 6.

Muscle atrophy analysis. (a) Representative H&E staining images of cross-sectional gastrocnemius muscle from sham, PNI, Silicone conduit, Hollow microfibre, and Hierarchical fibre bundle groups. The black dashed circles highlight the cross-sectional areas of muscle fibres. (b) Representative Masson's trichrome staining images of cross-sectional gastrocnemius muscle in the corresponding five groups. The blue staining indicates collagen fibre deposition and muscle fibrosis. (c) Statistic analysis of the CSA of gastrocnemius muscles in each group. (d) Collagen volume fraction (%) of gastrocnemius muscles across the groups. The scale bars indicate 100 μm. (n ≥ 3, ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, n.s.: not significant, compared to PNI).

3. Conclusion

To conclude, inspired by the anatomical structure of nerve, we have generated biomimicking hierarchical fibre bundles by utilizing a multi-level microfluidic device with parallelly and hierarchically assembled capillary arrays. The formation of this fibre bundle is based on the fast ionic crosslinking of alginate at the interfaces of co-flows during the microfluidic spinning, and the photopolymerization of GelMA after collection. The materials used to generate such hierarchical fibre bundles could also be extended to polymers that could be rapidly crosslinked, such as photocrosslinkable hydrogels. Because of the independent injections and flexible fluids selection in microfluidics, hierarchical fibre bundles containing different amounts and compositions of filaments could be continuously obtained from this method. These fibre bundles then showed great cytocompatibility and cell migration induction, providing a biomimicking cell cultivation microenvironment and promoting the formation of nerve bundles. We also explored their application in 10 mm sciatic nerve defect SD rat models, which revealed that the spatial guidance provided by the hierarchical fibre bundles promoted the nerve regeneration and function recovery. These features convinced us that such a hierarchical fibre bundle is promising for clinical nerve regeneration and would hold a bright future in related tissue engineering applications.

4. Methods

Materials: Sodium alginate was from Aladdin. Calcium chloride (Anhydrous) was purchased from Sigma-Aldrich. Fluorescent polystyrene nanoparticles were bought from Invitrogen. GelMA was self-prepared according to the previously reported protocol. ECM proteins (containing collagens, fibronectin, tenascin, elastin, and several proteoglycans and glycosaminoglycans) were achieved from Beijing Zhengqingran Company. Lithium phenyl-2,4,6-trimethylbenzoylphosphinate (LAP) was bought from Macklin. All of the other chemical reagents were used as received, and all of the solutions were filtered before pumped into capillary microfluidic devices.

Microfluidic generation of hierarchical fibre bundle: The first level of injection channel was a seven-barrel capillary array with individual inlets, and its jetting orifice was tapered around 400 μm (total); the second level of injection channel was a cylindrical capillary with an internal diameter of 600 μm; the third level of injection channel was a cylindrical capillary with an internal diameter of 1200 μm; the collection channel was a cylindrical capillary with an internal diameter of 1500 μm. As for the fibre bundle containing eight or twelve filaments, the first level of injection channel could be replaced by a nine-barrel or a thirteen-barrel capillary with individual inlets. The four levels of injection channels were hierarchically and coaxially assembled on a glass slide and sealed by epoxy resin where necessary.

For the generation of hierarchical fibre containing one filament inside, the mixture of 1 wt% sodium alginate, 10 wt% GelMA, and 1 mg/mL ECM, regarded as the precursor of the filament, was the innermost phase flowing in the central barrel of the capillary array. The 1 wt% CaCl₂ solution flowed in the region between the capillary array and the second level of injection channel. The 1 wt% sodium alginate solution was taken as the outer phase that flowing inside the third level of injection channel. The 2 wt% CaCl₂ solution was taken as the outermost phase flowing inside the collection channel. A glass vial containing 2 wt% calcium chloride solution was used and covered the outlet of the microfluidic device during the spinning process. The fibres were then exposed to ultraviolet light to photopolymerize GelMA.

For the generation of hierarchical fibre bundles containing multiple filaments inside, the innermost phase was the 0.1 wt% calcium chloride solution flowing in the central barrel of the capillary array. The filaments precursors were introduced to the collection channel through the surrounding barrels of the capillary array. Similarly to the hierarchical fibre containing one filament, the 1 wt% calcium chloride solution and the 1 wt% sodium alginate solution were injected into the region between the capillary array and the second cylindrical capillary, and the region between the second and third cylindrical capillary, respectively. The resultant fibre bundles were collected and photopolymerized under exposure to ultraviolet light.

Rat Schwann cell cultivation: Schwann cells (RSC96) were cultured in Dulbecco's modified Eagle's medium, with 10% fetal bovine serum and 1% penicillin-streptomycin, and incubated at 37 °C with 5% CO₂. Cells directly seeded on the bottom of microplates were taken as the control group, and those co-cultured with hollow microfibre and hierarchical fibre bundle were named hollow microfibre, and hierarchical fibre bundle group, respectively. The cell viability of each group after 24 h and 72 h of cultivation was evaluated by using a Cell Counting Kit-8 assay.

Hematotoxicity analysis: According to the established protocol, the hematotoxicity evaluation of the materials used red blood cell (RBC) suspensions. Briefly, RBCs were obtained from rat blood through centrifugation and rinsed with PBS (pH 7.4) until the supernatant was clear and colourless. The diluted RBC solution was then mixed with deionized water (positive control, PC), PBS (negative control, NC), and material extracts, respectively, at a volume ratio of 1:4 at 37 °C for 4 h. Supernatant absorbance were recorded via a microplate reader. The hemolysis ratio was calculated according to the formula: [(A-A_nc)/(A_pc-A_nc)] × 100%, where A, A_pc, and A_nc represented the absorbance values of material extract group, PC, and NC groups, respectively.

In vivo degradation: Hollow microfibres made of 1 wt% alginate and hierarchical fibre bundles made of 1 wt% alginate with 10 wt% GelMA were prepared and weighed to record the initial weight (W₀). SD rats were anesthetized, the dorsal skin was shaved and disinfected, and incisions were made on both sides of the midline. Subcutaneous pockets were created by blunt dissection, sterilized samples were implanted, and the incisions were closed with surgical sutures and disinfected with iodine solution. At predetermined time points, rats were sacrificed and the samples were retrieved, rinsed, and weighed to obtain the remaining weight (W_t). The residual weight ratio was calculated as Wt/W₀ × 100%.

Animal surgery: All animal procedures were approved by the Animal Experimental Committee of Wenzhou Institute, University of Chinese Academy of Sciences, and were conducted in accordance with the established animal care guidelines. Five groups of 220-250 g healthy male SD rats were set and named as the Sham, the PNI, the silicone conduit, the hollow microfibre, and the hierarchical fibre bundle group. After the rats were anesthetized, the right sciatic nerve was carefully exposed, and a segment was removed to leave a 10 mm-defect after the nerve stumps retracted. Apart from the PNI group, the removed segment was implanted with silicone conduit, silicone conduit containing two hollow microfibres, and silicone conduit containing two hierarchical fibre bundles. After the surgery, all groups were intramuscularly injected penicillin (800,000 units) to avoid infection.

Walking track analysis: At 2, 4, and 8 weeks after surgery, the walking track analysis was conducted. The hind paws of rats were dipped the non-toxic inks, and the rats were allowed to walk in a narrow and straight space. The footprints were collected on the paper and measured to calculate the SFI based on the formula shown in Fig. S7a. The value reaching −100 suggested a total injury and a successful built of nerve defect model, while approaching 0 indicated a good recovery.

Electrophysiology test: Electrophysiological analyses were implemented on the 8th week. After the rats were anesthetized, the sciatic nerves were exposed. At the proximal site of the sciatic nerve, an electrode was attached to emit electrical signals. At the same time, another electrode was attached to the right gastrocnemius belly to form the electromyograph. The compound muscle action potentials (CMAPs) were measured real time, and the amplitude, latencies and conduction velocities of different groups were analyzed according to the real-time curves.

Histological analyses of the regenerated nerve and gastrocnemius muscle: Eight weeks post-surgery, the rats were sacrificed. The regenerated nerves and gastrocnemius muscles from both sides were collected and fixed in 4% paraformaldehyde for 24 h. After fixation, the tissue samples were embedded in OCT or paraffin, respectively, and sectioned. For nerve samples, longitudinal sections were stained using the antibodies including chicken anti-NF200 (1:1000), rabbit anti-S100β (1:300), while transverse sections were stained with chicken anti-NF200 (1:1000), rabbit anti-MBP (1:300), mouse anti-α-SMA (1:500) and rabbit anti-CD31 (1:100) or rabbit anti-CD68 (1:200), according to the protocol. After being incubated with fluorescent secondary antibodies (1:1000), all slices were treated with DAPI Fluoromount-GTM. For muscle samples, cross-sectional slices were stained with H&E, and Masson's trichrome to assess the CSA and pathological changes.

Characterization: The microfluidic spinning process was supervised through a microscope equipped with a fast camera (Pike, F032B). The collected hierarchical microfibre were characterized via the stereomicroscope (Olympus, SZX16). The tensile stress-strain curve of the hierarchical fibre bundle was measured by an electronic universal testing machine (Instron, 5944). All images of histological assessments were captured on Nikon Eclipse 80i, or a confocal laser microscope (Nikon, Ti-E&A1 plus).

Statistical analysis: All experimental data were presented as mean ± standard deviations. Statistical analysis was evaluated using Origin software. One-way ANOVA followed by Tukey's post hoc test was used for statistical evaluation, and differences were considered statistically significant at ∗p < 0.05, ∗∗p < 0.01, or ∗∗∗p < 0.001.

CRediT authorship contribution statement

Yunru Yu: Writing – review & editing, Writing – original draft, Methodology, Investigation, Funding acquisition. Binghui Jin: Methodology, Investigation, Data curation. Yi Chen: Methodology, Writing – review & editing. Jingjing Gan: Methodology, Writing – review & editing. Lu Fan: Writing – review & editing, Validation. Li Wang: Writing – review & editing, Methodology, Formal analysis. Yuanjin Zhao: Writing – review & editing, Supervision, Project administration, Investigation, Funding acquisition, Conceptualization.

Ethics approval and consent to participate

This work does not involve any human subjects.

All animal procedures were approved by the Animal Experimental Committee of Wenzhou Medical University （xmsq2024-0766）and were conducted in accordance with the established animal care guidelines.

Declaration of competing interest

Yuanjin Zhao is an editorial board member for Bioactive Materials and was not involved in the editorial review or the decision to publish this article. All authors declare that there are no competing interests.

Appendix A. Supplementary data

The following is the Supplementary data to this article: