We obtained bFGF lyophilized powder from the School of Pharmacy, Wenzhou Medical University (Wenzhou, China). Poloxamer 407 (P2443) was obtained from Sigma-Aldrich (Shanghai, China). Inhibitor targeting PAK1 activation-3 (IPA-3) and Bafilomycin A1 (Baf A1) were obtained from Selleck Chemicals (42521-82-4, 88899-55-2, Shanghai, China). Cy5.5-NHS was obtained from Aladdin (1469277-96-0, Shanghai, China). The primary and secondary antibodies used in this study are listed here. Rabbit monoclonal anti-PAK1 (ab223849), rabbit polyclonal anti-Bcl2 (ab196495), mouse monoclonal anti-P62 (ab56416), rabbit polyclonal anti-Beclin1 (ab62557), donkey anti-mouse IgG H&L Alexa Fluor 488-conjugated secondary antibody (ab150105), donkey anti-rabbit IgG H&L Alexa Fluor 647-conjugated secondary antibody (ab150075) were obtained from Abcam. Rabbit polyclonal anti-Myelin basic protein (MBP, 10458-1-AP), mouse monoclonal anti-S100 (66616-1-Ig), mouse monoclonal anti-GAPDH (60004-1-Ig), horseradish peroxidase (HRP)-conjugated Affinipure Goat Anti-Mouse IgG (H + L) (SA00001-1), and HRP-conjugated Affinipure Goat Anti-Rabbit IgG (H + L) (SA00001-2) were obtained from Proteintech. Rabbit polyclonal phospho-PAK1 (Thr423) (AF4463) was obtained from Affinity. Rabbit anti-BAX (#2772), and rabbit anti-Cleaved Caspase-3 (#9661) were obtained from Cell Signaling Technology. Rabbit polyclonal anti-LC3B (L7543) was purchased from Sigma-Aldrich. Rabbit polyclonal anti-ATG5 (autophagy-related protein 5, AP6026) was purchased from Bioworld.

Poloxamer 407 and bFGF powder were mixed in saline and stirred gently at 4°C. The final concentration of the P-bFGF solution was 200 mg/ml for poloxamer and 0.1 mg/ml (or 0.3 mg/ml) for bFGF. The P-bFGF solution was lyophilized, fixed to a copper sheet, and sprayed with gold to take shape. Its microstructure was observed using a scanning electron microscope (SEM, Regulus8230; HITACHI, Tokyo, Japan) and its elemental composition was determined by SEM-EDS (energy-dispersive X-ray spectroscopy).

The Schwann cell line RSC96 (ScienCell, Shanghai, China) was cultured in Dulbecco’s Modified Eagle Medium (DMEM)containing 10% fetal bovine serum and 1% penicillin-streptomycin solution in an incubator at 37°C and 5% CO₂. The cells were seeded in a six-well plate at an initial density of 2 × 10⁵ cells/mL. To determine the effect of P-bFGF, we divided the RSC96 cells into six groups: control group, H₂O₂ group, poloxamer group, bFGF group, P-bFGF group and IPA-3 group. The control group was cultured in normal DMEM, while the other groups were stimulated with 500 μmol H₂O₂ for 2 h. Four hours before H₂O₂ stimulation, the poloxamer group, bFGF group, and P-bFGF group were pretreated with poloxamer (200 mg/ml), bFGF (0.1 mg/ml), and P-bFGF (0.1 mg/ml) respectively. While the IPA-3 group was treated with P-bFGF (0.1 mg/ml) and IPA-3 (25 μM). To determine how bFGF and IPA-3 affect autolysosomal flux, we divided the RSC96 cells into seven groups: control group, H₂O₂ group, H₂O₂+bFGF group, H₂O₂+Baf group, H₂O₂+bFGF + Baf group, H₂O₂+bFGF + Baf + IPA-3 group, and H₂O₂+bFGF + IPA-3 group. We treated the cells with Baf A1 (1 nM) 4 h before H₂O₂ stimulation.

Male Sprague Dawley (SD) rats (200–220 g) were purchased from Charles River Laboratories (Beijing, China). The animal experiment in this study was approved by the Animal Experiment Ethics Committee of Wenzhou Medical University. The FNI model was estabilished following a previous study (Xia et al., 2020). Briefly, the rats were anesthetized by an intraperitoneal injection of 1% pentobarbital sodium (40 mg/kg). We exposed the main trunk of the left facial nerve 0.5 cm away from the stylomastoid foramen, and clamped a nerve (∼2 mm long) for 60 s with a hemostatic forceps. The rats were randomly divided into five groups (n = 10 per group): sham group, FNI group, poloxamer group (FNI treated with poloxamer), bFGF group (FNI treated with bFGF) and P-bFGF group (FNI treated with P-bFGF). The poloxamer group and P-bFGF group were injected with 15 μL poloxamer (200 mg/ml) and P-bFGF (200 mg/ml poloxamer and 0.3 mg/ml bFGF) on D0, respectively. The sham group, FNI group and bFGF group were injected with 15 μL of normal saline, normal saline, and bFGF (0.1 mg/ml) on D0, D3, and D7, respectively. On D7, five rats in each group were sacrificed. Facial nerves (1.5 cm long), including those from injured sites, were dissected and immediately stored in liquid nitrogen for western blotting. On D14, four to five rats in each group were sacrificed. The facial nerve was fixed in 4% paraformaldehyde (PFA) overnight, dehydrated with gradient absolute ethanol, and embedded in paraffin. The sample was cut into sections (5 μm thick) using a microtome (HistoCore AUTOCUT; Leica, Wetzlar, Germany).

To track the distribution and residence time of bFGF in the in vivo experiment, bFGF used in both P-bFGF and bFGF solutions was labeled with Cy5.5-NHS (Zhao et al., 2014). Briefly, after Cy5.5-NHS was mixed with bFGF (mass ratio 1:100), they were stirred overnight at 4°C in the dark. The unconjugated Cy5.5-NHS was eliminated through dialysis and gel chromatography, and the Cy5.5-bFGF solution was freeze-dried to obtain the Cy5.5-bFGF powder.

The Cy5.5-conjugated P-bFGF and bFGF were used to treat FNI rats. At 0, 6 h, 1, 3, and 7 day, the bioluminescence intensities of the two groups were recorded using an in vivo imaging instrument (IVIS Lumina XRMS, Shanghai, China). The luminous intensity was analyzed and quantified by the Living Image^® Software.

According to the method described by Wang et al. (2006), Blink reflex (BR), Vibrissae movement (VM) and tip position (TP) are the three main aspects to assess facial nerve function. For BR, the air was discharged 2 cm away from the injured eye quickly with a 5 ml syringe, and the eyelid movement of the rats was scored with 0 points for no difference between the two sides, 1 point for delayed action on the injured side, and 2 points for unclosed eyelid on the injured side. For VM, bilateral beard movement was counted for 30 s with 0 points for no difference between the two sides, 1 point for weakened movement on the injured side, and 2 points for the loss of beard movement. For TP, 0 points represented nose tip centered, and 1 point for nose tip not centered. The FNI model was considered to be successfully established when the score was 5. After modeling, the facial nerve function of rats was evaluated on D0, D3, D7 and D14, respectively. H&E and Masson’s staining were also performed to confirm that the modeling was successful.

According to the manufacturer’s protocol, the prepared sections were stained by H&E or Masson’s trichrome staining. Initially, the sections were dried at 56°C for 30 min, then immersed in xylene to dewax, and hydrated with gradient absolute ethanol and pure water. Then, different staining reagents were added to the sections. For H&E staining, the sample was immersed in hematoxylin for 7 min and eosin for 2 min. For Masson’s trichrome staining, the nucleus of the section was stained with hematoxylin for 5 min, and Ponceau/Acid Fuchsin solution was added to cover the entire tissue for 5 min. Then, the section was stained in 1% phosphomolybdic acid solution for 30 s and stained in aniline blue reagent for 1 min. Finally, the glass slide was mounted with neutral resin. The stained section was observed and imaged using an inverted microscope (DMILLED; Leica).

The tissue sections were subjected to antigen repair with high temperature and pressure. The cells were fixed in 4% PFA at room temperature for 30 min and then incubated with 0.5% Triton X-100 for 15 min. Then, the tissue sections or cells were blocked with 5% BSA for 30 min, and the tissues or cells were incubated with primary antibodies against MBP (1:100), S100 (1:100), LC3B (1:300), and Cleaved Caspase-3 (1:400) at 4°C overnight. After washing, these samples were incubated with Alexa Fluor conjugated secondary antibody (1:200) at room temperature for 1 h, and finally mounted with DAPI. Fluorescence images were captured using a laser confocal scanning fluorescence microscope (TCS SP8; Leica), and the fluorescence intensity was quantified by the ImageJ software.

The facial nerve was placed in the RIPA lysis buffer containing 1% PMSF and lysed in a multi-sample tissue homogenizer (Bionoon-24LD; BIONOON, Shanghai, China) for 10 min. The RSC96 cells were placed in the RIPA lysis buffer containing 1% PMSF. Subsequently, the homogenized facial nerve or cell suspension was centrifuged at 12,000 rpm at 4°C for 15 min, and the protein concentration in the supernatant was quantified by the BCA reagent. The same amount of total protein was separated by SDS (Sodium dodecyl sulfate)-polyacrylamide gel electrophoresis and transferred to a PVDF (polyvinylidene fluoride) membrane. The membrane was blocked using skim milk and incubated overnight with primary antibodies (including GAPDH, Bcl2, BAX, beclin1, LC3B, ATG5, p62, p-PAK1 and PAK1) at 4°C. On the following day, the membrane was incubated with a secondary antibody conjugated with HRP for 2 h at room temperature. An enhanced chemiluminescence kit was used to detect the protein band, and the band density was quantified using the ImageJ software.

The data were analyzed using Graphpad Prism9 statistical software. One-way ANOVA was performed for analyzing data with Figures 2D,E, 3B,D–G, 4B,D,E, 5B–G,I,J, 6B,D,E. For two variables, two-way ANOVA was performed (Figure 2A). The difference was considered to be statistically significant at p < 0.05.

To gelate at body temperature, we prepared a poloxamer solution with a concentration of 17% (w/w), which was found to be the most suitable in previous studies (Wang et al., 2017b). The poloxamer was in a liquid form at 4°C and turned into a hydrogel at 37°C (Figure 1A). Moreover, the ultrastructure of the poloxamer hydrogel was observed by SEM (Figure 1B). The results showed that the internal crosslinking of hydrogel formed a loose porous network, and the internal porous network was unchanged after bFGF was loaded. SEM-EDS was performed to analyze the elements of P-bFGF. The C and O elements represented poloxamer hydrogel, and the N and S elements represented bFGF. We found that bFGF was evenly distributed in the poloxamer hydrogel (Figure 1C).

To observe whether P-bFGF can continuously release bFGF in situ, Cy5.5-bFGF was used to treat FNI rats. Through in vivo imaging, we discovered that in the bFGF group, the bFGF solution diffused in the tissue, while in the P-bFGF group, bFGF was attached to the facial nerve, which allowed us to see the contour of the facial nerve clearly. With the passage of time, we found that the bioluminescence intensity of bFGF decreasd, while the bioluminescence intensity of the P-bFGF group rats became stronger than that of the bFGF group rats (Figure 1D). The quantification of bioluminescence intensity showed the same trend (Figure 1E), which proved that P-bFGF constantly released bFGF in the injured nerve in situ, and maximized the therapeutic effect of bFGF in vivo. The results showed that P-bFGF was suitable for local uniform delivery and slow-release after nerve injury in vivo.

To determine whether P-bFGF treatment promoted the recovery of facial nerve function, facial nerve function scoring was performed 0, 3, 7 and 14 days after surgery. On D0 after surgery, the score in each group was 5 points, suggesting that the model was successfully established. With time, the facial nerve function improved in all the groups, but the facial nerve function scoring revealed that the scores of the P-bFGF group and bFGF group were lower than those of the FNI group and poloxamer group. P-bFGF group was significantly lower than FNI group (****p < 0.0001) and poloxamer group (****p < 0.0001) and bFGF group was not significantly lower than FNI group and poloxamer group, which might be attributed to the long-lasting protection of the poloxamer hydrogel for bFGF bioactivity and slow drug release at the site of the facial lesion (Figure 2A). In summary, poloxamer alone had no therapeutic effect. This was also reflected in all the experiments that followed. We found that bFGF alone may not promote the recovery of the facial nerve function. P-bFGF, a mixture of bFGF and poloxamer, can significantly promoted the early functional recovery of facial nerve injury.

Hematoxylin & Eosin and Masson’s staining were used to evaluate the histological recovery of injured nerves in each group. Both H&E and Masson’s staining demonstrated that the nerve fibers in the FNI group and the poloxamer group were sparse and atrophic. The abnormal morphologies in the bFGF group and P-bFGF group improved. The nerve fibers in the P-bFGF group regenerated significantly and more regularly compared to those of the bFGF group (Figure 2B). This suggested that P-bFGF can promote the recovery of histological morphology of injured nerve fibers.

The S100 maker of Schwann cells can regulate the metabolism, movement and proliferation of cells. The presence of MBP (myelin basic protein) indicates myelin formation. Double IF staining of S100 and MBP showed that the functional protein expression and myelin density of the P-bFGF group and bFGF group were significantly higher than those of the FNI group and poloxamer group, and P-bFGF was higher than bFGF (Figure 2C). The qualitative fluorescence co-localization expression analysis and respective quantitative analysis of S100 and MBP showed the similar trend (Figures 2D,E). The results indicated that P-bFGF upregulated the functional protein S100 of Schwann cells and boosted the remyelination of the SCs.

To determine whether P-bFGF activated autophagy for facial nerve repair, we performed tissue IF staining and WB to detect the expression of autophagy-related proteins (including LC3B, Beclin1, P62, and ATG5). The IF staining showed that FNI slightly increased the fluorescence intensity of LC3B, and the bFGF and P-bFGF treatment further enhanced the fluorescence intensity of LC3B, the fluorescence intensity of the P-bFGF group was significantly higher than that of the bFGF group (Figure 3A). The quantitative analysis of fluorescence intensity showed a similar trend (Figure 3B). The WB results showed that the expression trends of LC3B-II, Beclin1, and ATG5 were consistent with the fluorescence intensity of LC3B, while the expression trend of P62 was the opposite (Figures 3C–G). These results indicated that P-bFGF activated autophagy in FNI.

To further determine whether P-bFGF treatment contributed to cell survival, we also measured the expression of apoptosis-related proteins by IF staining (Cleaved Caspase-3) and western blotting (BAX and Bcl2), and assessed the ability of P-bFGF to activate autophagy, which can help to reduce apoptosis after FNI. The pro-apoptosis protein Cleaved Caspase-3 and BAX were downregulated, while the anti-apoptosis protein Bcl2 was upregulated in the bFGF group and P-bFGF group, compared to their levels in the FNI group (Figures 4A,C). The quantitative analysis of fluorescence intensity and WB gray value showed a similar trend (Figures 4B,D,E). The above results confirmed that P-bFGF can promotes FNI repair by reducing apoptosis for the survival of nerve cells.

To study the molecular mechanism underlying the autophagy activation of P-bFGF, we performed the above treatment on different groups of RSC96 cells. The expressions of pathway-related proteins (including PAK1 and p-PAK1) and autophagy-related proteins (including LC3B, Beclin1, P62, and ATG5) were detected. Through the WB assay, we found that the phosphorylation level of PAK1 increased in the H₂O₂ group and the poloxamer group. The phosphorylation level of PAK1 further increased in the bFGF group and P-bFGF group. IPA-3 partially reversed the effect in the P-bFGF group. The expression of PAK1 was opposite to that of p-PAK1 (Figures 5A–C). This showed that bFGF activated the PAK1 pathway, but IPA-3 inhibited the activation of bFGF in the PAK1 pathway. The expressions of ATG5, Beclin1 and LC3B-II in the P-bFGF group significantly increased compared with the H₂O₂ group (**p < 0.01, ***p < 0.001 and ***p < 0.001, respectively), while the use of IPA-3 significantly reversed the treatment of P-bFGF (*p < 0.05, **p < 0.01 and *p < 0.05, respectively). The expression of P62 showed an opposite trend to the expression of ATG5, Beclin1 and LC3B-II (Figures 5A,D–G). To confirm the continuous autolysosomal function in the cells treated by bFGF and determine how IPA-3 affected autolysosomal flux. We used Baf A1, an inhibitor of late autophagy, to treat RSC96 cells after the administration of bFGF. Baf A1 further raised the expression of LC3B-II and P62, while IPA-3 reversed the expression of LC3B-II and further increased P62 (Figures 5H–J). This indicated that the autophagic flux was activated by bFGF and blocked by IPA-3, which suggested that P-bFGF activated autophagy through the PAK1 pathway.

To further confirm the molecular mechanism underlying the reduction of P-bFGF in H₂O₂-induced apoptosis of SCs, cell IF staining was performed to detect the expression of Cleaved Caspase-3, and WB was performed to detect the expressions of BAX and Bcl2. The fluorescence intensity of the pro-apoptosis protein Cleaved Caspase-3 in the P-bFGF group was significantly lower than that in the H₂O₂ group (****p < 0.0001), and IPA-3 treatment reversed the effect of the P-bFGF treatment (****p < 0.0001, Figures 6A,B). The grayscale value of the pro-apoptosis protein BAX was significantly lower than that in the H₂O₂ group (*p < 0.05), and IPA-3 treatment reversed the effect of the P-bFGF treatment (****p < 0.0001, Figures 6C,D). Moreover, the grayscale value trend of the anti-apoptosis protein Bcl2 was opposite to that of BAX. The value of the P-bFGF group was significantly higher than that of the H₂O₂ group (**p < 0.01), and IPA-3 treatment reversed the effect of the P-bFGF treatment (*p < 0.05, Figures 6C,E). These results indicated that P-bFGF protected SCs from being apoptotic via the PAK1 pathway.