A unidirectional drug‐release Janus membrane based on hydrogen bonding barrier effect for preventing postoperative adhesion and promoting tissue repair

Barrier membranes incorporated with anti‐adhesion drugs have been widely used for preventing postoperative adhesion. However, the bidirectional release of drugs may interfere with internal tissue healing. Herein, a unidirectional Janus membrane composed of two functional layers is prepared by sequential electrospinning. The outer layer is designed as random polycaprolactone fibers incorporated with tannic acid/Fe3+ particles to prevent tissue adhesion, while the inner layer is designed as oriented gelatin fibers to promote tissue repair. Hydrogen bonding was employed to act as a barrier for preventing the diffusion of the tannic acid to the inner side, and thus the release of tannic acid occurs in a unidirectional manner. As a result, asymmetric biological performances of the Janus membrane are achieved. The outer face can provide anti‐inflammatory capability to the membrane and inhibit the viability of fibroblasts, while the inner face can promote the proliferation and differentiation of tendon stem cells. Tendon injured model confirms the prominent inflammation modulation, adhesion prevention, and repair promotion capabilities of the Janus membrane in vivo. This study provides a new strategy to develop advanced functional barrier membranes for postoperative adhesion prevention and internal tissue repair promotion.


INTRODUCTION
Tissue adhesion is a common postoperative complication with high morbidity, which can cause chronic abdominal pain, dysfunction of adjacent tissue, and female secondary infertility. 1 Polymeric barrier membranes have emerged as tissue hyperproliferation, while the other side is required for proper adhesion to tissue and repair promotion capability. 6 However, it is of great difficulty for single polymeric membranes to simultaneously achieve these properties.
In general, adhesion prevention and repair promotion are two contradictory tissue behaviors. Janus system with asymmetric functional faces is an attractive candidate to fulfill different requirements for the barrier membranes. 7 Janus electrospun membranes allow sustained and programmed drug delivery, nutrient transport, and design of composition and structure, and can be employed for advanced barrier membranes. 2b, 8 Postoperative adhesion is mainly induced by uncontrolled inflammation and excessive proliferation of fibrous tissue. 9 Incorporation of anti-inflammatory or antifibrotic drugs such as ibuprofen, 10 celecoxib, 11 and mitomycin-C 12 into the membranes helps to prevent postoperative adhesion by inhibiting one of the above pathways for the formation of adhesion tissue. Our previous study demonstrated that metal/phenolic network complexes (MPNs) could not only intelligently manipulate inflammatory response but also prevent the infiltration of fibroblasts into the membranes, which was ascribed to the long-term controlled release of tannic acid (TA). 13 It is of great promise for MPNs to be applied in preventing postoperative adhesion since they can simultaneously inhibit both adhesion formation pathways. Nevertheless, bidirectional release of bioactive drugs may interfere with inner tissue healing. Therefore, an advanced strategy to fabricate unidirectional Janus electrospun membranes for preventing postoperative adhesion is strongly desired.
Hydrogen bonding, a primary participant in strong intermolecular interactions, is commonly employed for gel preparation and strength enhancement due to the formation of stable structures. 14 Herein, hydrogen bonding is adopted to provide barrier effect for the release of drugs to the inner tissue. We expect that the bioactive drugs diffused from the outer layer of the Janus membrane will be obstructed by the inner layer by virtue of the hydrogen bonding effect among the drugs and the inner layer. Thus, the drugs will be released in a unidirectional manner, and the healing of inner tissue is protected from their interference.
In the present study, we prepared a Janus electrospun membrane composed of double layers by sequential electrospinning. The outer layer (PTA) facing was composed of random PCL fibers and incorporated with TA/Fe 3+ particles to prevent adhesion formation and provide mechanical support; the inner layer facing injured tissue was composed of gelatin (GEL) fibers, which could provide great affinity to tissue and hydrogen bonding barrier effect to the released TA molecules. Besides, promotion of inner tissue repair is required for the inner layer of the Janus membrane. We have proved that the oriented topology of electrospun fibers can promote cell proliferation. 15 Hence, the inner layer of the Janus electrospun membrane was designed as an oriented structure to facilitate tissue healing. Peritendinous adhesion commonly occurs after tendon injury, which inevitably causes pain and impacts tendon healing. 16 Since the number of tendon cells in tendons is limited, the intrinsic healing driven by tendon cells is relatively slow. 5 Hence, it is necessary to take therapeutic measures to prevent postoperative adhesion and promote tendon healing. In this study, a tendon injury model was adopted, and the injured tendons were wrapped with the Janus membranes. The effect of the Janus membranes on adhesion prevention, repair promotion, and the underlying molecular mechanism were investigated. The Janus membrane with asymmetric composition and function holds great potential to be applied in internal tissue injury therapy.

Fabrication and characterization of the Janus membrane
The Janus membranes were fabricated by sequential electrospinning, as illustrated in Scheme 1. The membrane composed of PTA layer (random PCL fibers incorporated with 5 wt.% TA/Fe 3+ particles) and GEL layer (oriented gelatin fibers) was named as PTA-GEL, and the membrane composed of PCL layer and GEL layer was named as PCL-GEL.
The surface morphology of the Janus membranes was observed by scanning electron microscope (SEM). As shown in Figure 1A, the PTA face in PTA-GEL was composed of smooth fibers without obvious particle aggregation, presumably due to the homogeneous dispersion of TA/Fe 3+ particles in the electrospinning solution. EDS mapping exhibited a uniform distribution of Fe element (Supplementary Figure S1), demonstrating the uniform distribution of TA/Fe 3+ particles in the PTA face of PTA-GEL. In comparison with the PCL fibers in PCL-GEL (Supplementary Figure S2), the diameter of the PTA fibers in PTA-GEL was significantly decreased since the conductivity of the electrospinning solution was improved by TA/Fe 3+ particles. As shown in Figure 1B, the GEL face of PTA-GEL was composed of highly oriented fibers. The porosity of PTA-GEL was similar to that of PCL-GEL ( Figure 1D). The average pore size of PTA-GEL was 2.62 ± 0.39 μm, which was less than that of PCL-GEL (6.15 ± 0.46 μm, Figure 1C). Less pore size is conducive to limit the infiltration of cells into the inner tissue. 16

S C H E M E 1 (A)
Schematic showing the fabrication process of PTA-GEL Janus membrane. (B) Schematic of the mechanism by which PTA-GEL prevents peritendinous adhesion and promotes tendon repair. As shown in Figure 1E and Supplementary Figure S3, the water contact angle (WCA) of the PTA face for PTA-GEL was significantly lower than that of the PCL face for PCL-GEL, indicating that the addition of TA/Fe 3+ particles improved the hydrophilicity of the membranes. These results might be caused by the phenolic hydroxyl groups enriched on the fiber surface, confirming by the ATR-FTIR spectrum (Supplementary Figure S4). Additionally, the modulus of PTA-GEL was higher relative to that of PCL-GEL ( Figure 1F), indicating a better mechanical strength of PTA-GEL.
DPPH 17 and PTIO 18 were employed to evaluate the radical scavenging capability of the membranes by incubating the radical solutions with the membranes. The PCL electrospun membrane incorporated with FeCl 3 (PCL-FeCl 3 ) showed a low radical scavenging ratio, and the PCL-TA membrane incorporated with TA presented high scavenging capability (Supplementary Figure S5). These results indicated robust radical scavenging capability of TA, which was consistent with previous reports. 19 Thus, TA endowed the PTA-GEL membrane with great radical scavenging capability ( Figure 1G and H) Moreover, the radical scavenging capability of PTA-GEL could remain for several weeks, which was ascribed to the chelation effect of the Fe 3+ on the TA molecules in the TA/Fe 3+ nanoparticles. The sustained radical scavenging capability of the Janus membrane implied a promising potential for long-term inflammation manipulation. Besides, the release of TA from PTA-GEL presented a pH responsive manner (Supplementary Figure S6). The release rate of TA was higher in acidic environment than that in neutral environment, indicating that PTA-GEL could intelligently manipulate the inflammation response by releasing TA in response to the microenvironment.
The stable intermolecular interaction between TA and GEL has been widely reported, 20 which was formed by the hydrogen bonding among the functional groups of TA and GEL as illustrated in Figure 1I. 21 In this study, we speculated that PTA-GEL could unidirectionally release TA by virtue of the hydrogen bonding barrier effect of the GEL layer. To verify this conjecture, we first compared the diffusion of TA aqueous solution in water as well as in GEL aqueous solution. As shown in Supplementary Figure  S7, the TA solution dropped in GEL aqueous solution could not diffuse, while that dropped in water dispersed rapidly. Then, we investigated the relative retention ratio for the TA aqueous solution of GEL fibrous membrane and PCL fibrous membrane. The results in Supplementary Figure S8 revealed that most TA molecules were blocked by the GEL fibers, confirming the barrier effect of GEL fibers. Furthermore, the drug-release behavior of the Janus membranes was investigated in Transwell plates. As illustrated in Figure 1J, for pure PTA electrospun mem-brane, the concentration of TA in external and internal compartments showed no obvious difference, indicating that TA molecules were released from the pure PTA electrospun membrane in a bidirectional manner. However, for PTA-GEL, the concentration of external solution was much higher than that of internal solution, implying a unidirectional release behavior. This unidirectional release property makes it possible for PTA-GEL to prevent adhesion formation while avoiding side effects to tendon repair. Besides, PTA-GEL could be curled into a ringlike shape as exhibited in Supplementary Figure S9, demonstrating the capability to be wrapped around the injured tendons.

Regulation of inflammation
Inflammatory reactions induced by the injured tissue or implanted membranes play pivotal roles in the formation of postoperative adhesion. 22 It has been reported that in the early stage after injury, macrophages gather around the injured site and release a large number of proinflammatory cytokines, leading to the recruitment of fibroblasts for the formation of adhesion. 23 Hence, inhibiting the inflammatory response driven by macrophages has been recognized as a promising pathway to prevent tissue adhesion. 24 TA has been demonstrated with robust anti-inflammatory capability by scavenging inflammation-derived free radicals, inhibiting the activity of free radical producing enzymes, and regulating specific transcription factors to block the production of some inflammatory cytokines. 25 TA can coordinate with metal ions to form microspheres, which was reported to release TA molecules in a pH responsive manner. 26 Herein, TA molecules were coordinated with Fe 3+ to form TA/Fe 3+ microspheres, and the pH responsive release behavior of the TA/Fe 3+ particles has been confirmed. Thus, we speculated that the incorporation of TA/Fe 3+ particles would endow PTA-GEL with smart anti-inflammatory property. Systematic in vitro and in vivo experiments were carried out to investigate the inflammation regulation capability of the membranes. The membranes were cocultured with lipopolysaccharide (LPS) treated macrophages to investigate their regulatory capability to inflammatory cells. Reactive oxide species (ROS), key signaling molecules in inflammatory response, were detected by DCFH-DA (a fluorescent probe). As shown in Figure 2A, the macrophages treated with LPS (the control group) exhibited strong fluorescence intensity. The presence of the PCL-GEL membrane had little influence on the ROS fluorescence intensity, while the fluorescence in the cells cocultured with PTA-GEL was obviously decreased. This result revealed that PTA-GEL could downregulate the expression of intracellular ROS, which was further confirmed by the flow cytom-  Figure 2B. Then, the RNA of the macrophages was extracted to detect relative gene expression using RT-PCR, with the cells without any treatment (the blank group) as the reference group and GAPDH as the reference gene. As shown in Figure 2C, compared to the control group and the PCL-GEL group, the expression of M1 surface marker (CD86) as well as proinflammation-related genes (TNFα, iNOS, IL-1β, and IL-6) in the PTA-GEL group was significantly downregulated, while the expression of M2 surface marker (CD206) and anti-inflammation-related genes (IL-4 and IL-10) was upregulated. In addition, the expression of TGF-β1 and COL1a, key genes in the regulation of fibrosis, 27 was also suppressed after coculturing with PTA-GEL.
Subcutaneous implantation was performed to evaluate the in vivo anti-inflammatory property of the membranes. As shown in Figure 2D, a large number of immune cells were observed to aggregate around or infiltrate into PCL-GEL, while the aggregation and infiltration of the inflammatory cells were obviously decreased for the PTA-GEL group. Immunohistochemical (IHC) staining demonstrated that the expression level of TNFα was inhibited in the tissue around PTA-GEL (Supplementary Figure S10). Immunofluorescence (IF) staining images in Figure 2E exhibited that the cells infiltrated into PCL-GEL were mainly composed of proinflammatory macrophages (CD86 positive), while the proportion of proinflammatory macrophages was sharply reduced in the PTA-GEL group. Statistical analysis indicated that the number of the cells infiltrated in PTA-GEL was less than half of that in PCL-GEL (Supplementary Figure S11). These in vitro and in vivo results confirmed that PTA-GEL had excellent anti-inflammatory capability and could suppress the expression of fibrosis-related genes in macrophages, which was ascribed to the TA molecules that was released from the membrane in response to the pH of the inflammation microenvironment. The excellent anti-inflammatory function endowed PTA-GEL with great potential to prevent postoperative adhesion.

Regulation of fibroblasts
Since fibroblasts are the main component of the adhesion tissue, 28 L929 cells were selected to be seeded on the outer faces of the Janus membranes to further evaluate the anti-adhesion abislity in vitro. The CCK-8 tests revealed that after incubation of 1, 4, and 7 days, the L929 cells on the PCL face of PCL-GEL continuously proliferated with high viability, while the proliferation of the cells on the PTA face of PTA-GEL was notably reduced ( Figure 3A). The distribution and morphology of these cells culturing for 7 days were observed by SEM. The images in Figure 3B showed that there were a large number of spindle-like L929 cells on the PCL face of PCL-GEL, while much less cells could be observed on the PTA face of PTA-GEL. Then the adhesion and spreading of the seeded cells were evaluated. The cytoskeleton and the nuclei were respectively stained with FITC-phalloidin (red) and DAPI (blue), and the cells were observed by a confocal laser scanning microscope (CLSM). As shown in Figure 3C, the area of the L929 cells cultured on the PTA face of PTA-GEL was smaller than that on the PCL face of PCL-GEL, and still exhibited a rounded shape even after 7 days of incubation. The 3D merged images of the cells further confirmed the significant inhibition effect of the PTA face for PTA-GEL on the proliferation of L929 cells (Supplementary Figure S12). An Annexin V-FITC apoptosis detection kit was employed to investigate the apoptosis of L929 cells seeded on the membranes. As shown in Figure 3D and E, only few Annexin V positive cells could be observed on the PCL face of PCL-GEL, while the proportion of Annexin V positive cells was much higher on the PTA face of PTA-GEL. This result indicated that the incorporation of TA/Fe 3+ particles led to apoptosis of L929 cells, which was further confirmed by the Hoechst staining (Supplementary Figure S13). KEGG pathway analysis arising from RNA sequencing revealed that the cell cycle and RNA transport of the L929 cells incubated on the PTA face of PTA-GEL were notably downregulated compared with that cultured on the PCL face of PCL-GEL ( Figure 3F). On the basis of these results, it could be concluded that the PTA face of PTA-GEL could prevent adhesion of fibroblasts.

Regulation of tendon stem cells
Tendon stem cells (TSCs) were employed to investigate the effect of the Janus membranes on tendon healing in vitro. First, TSCs were respectively seeded on the PTA face and the GEL face of PTA-GEL to investigate the effect of the inner layer on cell viability. As shown in Figure 4A, the TSCs in the control group proliferated rapidly, while the proliferation rate of TSCs on the PTA face was sharply downregulated, demonstrating a proliferation inhibition effect of the PTA face for PTA-GEL. The proliferation rate of the TSCs on the GEL face of PTA-GEL was obviously improved compared with that of the TSCs on the PTA face of PTA-GEL ( Figure 4B). Additionally, most cells on the PTA face became apoptotic after incubation of 4 days, and the cells became dead after 7 days of culturing, while the cells on the GEL face showed high viability with little apoptosis or death for 7 days of incubation. These results indicated that the GEL layer of PTA-GEL could protect the viability and proliferation of the TSCs.
To evaluate the effect of oriented topology on the behavior of TSCs, PTA-rGEL composed of PTA fibers and random GEL fibers was prepared. The CCK-8 tests revealed that the absorbance of the CCK-8 testing solution for the cells on the oriented GEL face of PTA-GEL was significantly higher than that on the random GEL face of PTA-rGEL on day 7 ( Figure 4C), implying a promotion effect of the oriented structure on cell proliferation.Morphology of the cells culturing for 4 days was characterized by staining the cytoskeleton and nuclei. As shown in Figure 4D, the aspect ratio of the cells in the oriented group was larger than that in the random group, and the cytoskeleton of the cells in the oriented group aligned along the fiber longitudinal direction.
Then relative gene expression of the TSCs was detected by RT-PCR ( Figure 4E). The expression of SCX, the marker of precursor and differentiated tendon cells, 29 showed no significant difference in the random or oriented group to the control group. The regulation of TNC, an extracellular matrix protein encoding gene, is of great importance for tendon formation. 30 The PCR analysis demonstrated that the expression of TNC in the random or oriented group was similar to that in the control group. The expression of Tnmd, a marker of mature tendons, 31 was higher in the oriented group than that in the other groups, indicating that the oriented structure could promote tendon cell differentiation. Thus, it could be concluded that the GEL layer could not only protect the viability of TSCs, but also promote their proliferation and differentiation.

In vivo tendon injury model
It is necessary for anti-adhesion barrier membranes to be biodegradable. For tendon adhesion prevention, synthetic polymeric materials with a slow degradation rate are preferred, since the barrier function can be maintained for at least 6 weeks to minimize adhesion formation until the repair is completed. PCL has promising potential for the application in preventing tendon adhesion for its favorable mechanical properties, chemical stability, and flexibility. However, the degradation rate of PCL is so slow that residual PCL membranes may interfere with tendon slippage after complete tendon repair. In this study, improved hydrophilicity that was derived from the TA/Fe 3+ particles incorporated in the PCL nanofibers could accelerate the breakage of PCL chains by facilitating the invasion of body fluid into the nanofibers. Therefore, the degradation rate of the Janus membrane was increased (Supplementary Figure S14). Besides, the TA/Fe 3+ particles could effectively inhibit chronic inflammation triggered by the degradation products, leading to great in vivo biological safety. These results encouraged us to further investigate in vivo adhesion prevention and repair promotion effects of the Janus membrane. Herein, the Achilles tendons in rats were injured and wrapped with the electrospun membranes (Supplementary Figure S15). The group without membrane treatment was set as the control, and the PCL membrane (composed of pure PCL fibers) and PTA membrane (composed of pure PTA fibers) were utilized to be compared with PTA-GEL to respectively investigate the effect of TA/Fe 3+ particles and the GEL layer of PTA-GEL. In the PTA-GEL group, the PTA side faced the peritendinous tissue, and the GEL side faced the injured tendons. The inflammation response, adhesion formation, and tendon healing of these groups were evaluated.

Anti-inflammation
After tendon injury, danger signals including ATP, proteolytic enzymes, uric acid, and heat shock proteins are released by the injured cells and then recognized by immune cells to active the inflammatory response. 32 Proinflammatory cytokines released by the immune cells can amplify the inflammation response and cause adhesion formation. 33 Therefore, the concentration of proinflammatory cytokines in serum after surgery was detected to characterize the anti-inflammatory capability of the membranes. As shown in Figure 5A, compared with the normal group, the concentration of TNFα, IL-1β, and IL-6 in the control group was notably upregulated, which was ascribed to the surgical injury. Treatment of the PCL membrane further increased the expression of the inflammation cytokines. The expression of these cytokines was significantly downregulated in the PTA and PTA-GEL groups, with no significance to the normal group, indicating robust anti-inflammatory capability of the PTA and PTA-GEL membranes.
Macrophages play a pivotal role in the immune response and exert proinflammatory or anti-inflammatory functions through phenotype polarization. 34 On the 7th day postoperation, IF staining was performed to evaluate the expression of F4/80, a macrophage marker, 35 and nitric oxide synthase-2 (NOS2), a proinflammatory phenotype marker, 36 in the peritendinous tissue. The images in Figure 5B showed that after surgery, a large number of F4/80+ cells aggregated in the peritendinous tissue, and the treatment of the PCL membrane upregulated the expression of NOS2 in these cells. However, the expression of F4/80 and NOS2 in the PTA and PTA-GEL groups was significantly downregulated. Statistical analysis for the fluorescence intensity ratio of NOS2 to F4/80 indicated that the PCL membranes increased the proinflammatory polarization of macrophages, while that was significantly reduced in the PTA and PTA-GEL groups (Supplementary Figure S16A). Furthermore, the expression level of TNFα in the peritendinous tissue was characterized by the IHC staining ( Figure 5C). It could be observed that TNFα was highly expressed in the peritendinous tissue around the PCL membrane, while the expression level was obviously downregulated in the PTA and PTA-GEL groups. The mean optical density analysis supplied in Supplementary Figure S16B was consistent with this result. From the obtained results, it can be concluded that both PTA and PTA-GEL membranes could effectively inhibit the inflammatory response induced by tissue injury and the polymeric membrane, which was ascribed to the robust anti-inflammatory capability of the continuously released TA.

Prevention of postoperative adhesion
Peritendinous adhesion usually begins to form on the 7 th day postoperation, and persists for several weeks. 37 The regulation of adhesion at the early stage plays a key role in the successful intervention of postoperative adhesion. 38 Therefore, on the 7th day after surgery, the effect of the membranes on the prevention of postoperative adhesion was first evaluated. The H&E staining images in Figure 6A showed that most tendons in the control group were closely connected with peritendinous tissue without obvious gaps, where fibroblasts or immune cells invaded into the injured tendons. The tendons in the membrane-treated groups were isolated from surrounding fibrous tissue, remaining notable space between peritendinous tissue and the tendons. Statistical analysis of the isolated distance in Supplementary Figure S17 revealed that the space in the membrane-treated groups was significantly larger than that in the control group. Moreover, the penetration of cells into the membranes was assessed. The H&E staining images and analysis of invaded cell numbers in Supplementary Figure S18 suggested that compared with the PCL membrane, PTA and PTA-GEL markedly decreased the infiltration of cells into the membranes. These results indicated that the PTA and PTA-GEL membranes could not only provide physical gaps between the injured tendons and surrounding fibrous tissue but also prevent cell infil-tration into the injured site of the tendons, which was beneficial for the inhibition of adhesion formation.
Since TGF-β1 acts as a major mediator in the formation of adhesion by stimulating fibrotic factors in fibroblasts, 39 the level of TGF-β1 is a key indicator to evaluate postoperative adhesion. Herein, the expression of TGF-β1 in the peritendinous tissue was characterized by the IHC staining. The images in Figure 6B and analysis of average optical density in Supplementary Figure S19 exhibited TGF-β1 was highly expressed in the control and the PCL groups, while that was remarkably reduced by the PTA membrane, demonstrating an anti-adhesion effect of the TA/Fe 3+ particles. Moreover, the expression level of TGF-β1 in the PTA-GEL group was also downregulated with no significant difference to that in the PTA group, indicating that the combination of GEL layer as the inner side did not attenuate the effect of the outer PTA layer on the prevention of peritendinous adhesion.
On the 21 st day after operation, visual examine and gross scoring were performed to evaluate the long-term postoperative adhesion according to the standards for assessing peritendinous adhesion (Table S3, supplied in the Supporting Information). As shown in Figure 6C, dense adhesions were observed in the control group, hindering the penetration of a spatula between the peritendinous tissue and the tendon. The PCL membrane barely inhibited the formation of peritendinous adhesion, remaining more than 70% of the adhesion tissue to be sharply dissected. Whereas the PTA and PTA-GEL membranes could remarkably attenuate the formation of peritendinous adhesion, allowing the spatula to penetrate between the tendons and surrounding fibrous tissue easily. The scores of the PTA and PTA-GEL groups were 1.72 ± 0.26 and 1.70 ± 0.33, respectively, notably lower than the control and PCL groups (4.55 ± 0.35 and 4.25 ± 0.36, respectively), confirming an excellent anti-adhesion effect of the TA/Fe 3+ component. Histological analysis was also carried out to further evaluate postoperative adhesion. As shown in Supplementary Figure S20, no obvious boundary between the tendons and peritendinous tissue could be observed in the control group, indicating severe adhesions occurred after the surgery. In the PCL group, fibrous tissue formed between the injured tendon and the membrane, demonstrating that fibroblasts and immune cells penetrated across the PCL membrane and proliferated around or into the injured tendon. However, clear boundary between the tendons and surrounding tissue was observed in the PTA and PTA-GEL groups, confirming a long-term anti-adhesion effect of the PTA and PTA-GEL membranes.
As mentioned above, improper inflammation and fibrosis are the main causes of postoperative adhesion. To investigate the molecular mechanism of the PTA and PTA-GEL membranes to prevent peritendinous adhesion, inflammation-and fibrosis-related proteins in peritendinous tissue were detected by western blotting. PI3K/AKT signaling pathway is known to be an important regulator in inflammatory response via activating NF-κB signaling pathway, and the activation of PI3K/AKT is proved to be mediated by ROS. 40 In this study, we have demonstrated that the PTA and PTA-GEL membranes could significantly downregulate the level of ROS, so we speculated that PI3K/AKT and downstream NF-κB signaling pathways could be inhibited in these two groups. The western blotting bands exhibited in Figure 6D and corresponding statistical analysis supplied in Supplementary Figure S21 revealed that the phosphorylation of AKT was significantly downregulated in the PTA and PTA-GEL groups compared with the other groups. In resting cells, NF-κB exists in the cytoplasm in an inactive form bounding with its inhibitor IκB. Once activated, phosphorylated AKT promotes the phosphorylation of IκB, resulting in the release of NF-κB, and then allowing it to translocate into the nuclei to initiate transcription of inflammatory cytokines. 41 As a key protein in NF-κB signaling pathway, P65 plays an important role in inflammation regulation, overexpression of which would amplify the release of proinflammatory cytokines. 42 Figure 6B and Supplementary Figure S20 demonstrated that in the PTA and PTA-GEL groups, the level of P65 and the phosphorylation of IκB were notably downregulated, confirming the NF-κB signaling pathway was inhibited. These results indicated that the PTA and PTA-GEL membranes could inhibit the activation of PI3K/AKT and NF-κB signaling pathways, which might be caused by the robust ROS scavenging capability provided by the continuous released TA from the membranes.
TGF-β1 can stimulate extracellular matrix proteins by activating Smad2/3, and thus the phosphorylated Smad2/3 promotes the formation of postoperative adhesion. Whereas, Smad7 can hinder the phosphorylation of Smad2/3 by means of competitively binding TGF-β type I receptor (TβRI) or degrading TβRI by ubiquitin ligase. 43 The western blotting test suggested that the level of TGF-β1 was higher in the PCL group than that in the control group, while that was remarkably lower in the PTA and PTA-GEL groups. These results were consistent with the tendency in the IHC staining, which might be caused by the TA molecules released from the PTA and PTA-GEL membranes. The level of Smad2/3 showed no significant difference in the PCL and PTA groups to that in the control group, while that was reduced by PTA-GEL. Meanwhile, the phosphorylation of Smad2/3 was notably decreased by PTA and PTA-GEL, and the expression of Smad7 was increased. Therefore, the PTA and PTA-GEL membranes could inhibit fibrosis-related adhesion by restricting the TGF-β1/Smad2/3 and activating the TGF-β1/Smad7 signaling pathway. On the basis of these findings, it could be concluded that the TA/Fe 3+ -modified membranes could effectively prevent peritendinous adhesion by simultaneously inhibiting the inflammation-related PI3K/AKT/NF-κB signaling pathway and the fibrosis-related TGF-β1/Smad signaling pathway.

Promotion of tendon healing
To estimate whether tendon repair was promoted by the unidirectional design of Janus membrane, IF staining of CD44, a surface antigen of tendon stem cells, 44 was carried out on day 7. Figure 7A and B showed that after injury, CD44 positive cells aggregated around the injured sites of the tendons. The number of CD44 positive cells in the PCL group was similar to that in the control group, while the cell number in the PTA group was declined, indicating that released TA molecules might inhibit the recruitment and proliferation of the tendon stem cells. Whereas, the num-ber of CD44 positive cells was significantly improved in the PTA-GEL group, demonstrating excellent prohealing capability of the PTA-GEL membrane. On day 21, H&E staining was performed to the repaired tendons. As shown in Figure 7C, no obvious incision could be observed in the control, PTA, and PTA-GEL groups, but there remained an incision that had not been repaired in the PCL group. The production and reorientation of collagen are key steps in tendon healing. 45 Type I collagen (COL I) is the main composition of normal tendons with high alignment and promising mechanical strength. Type III collagen (COL III) is always produced at the early stage with poor arrangement and low mechanical strength, overexpression of which in injured tendons can damage collagen formation. [45][46] The types, distribution, and arrangement of the collagen at the repaired sites were characterized by Sirius red staining. As shown in Figure 7D and E, a large amount of COL III and a small amount of COL I were observed in the control group with high arrangement. The proportion of COL I in the PCL group was similar to that in the control group, but the arrangement of the collagen fibers was sharply declined, which might be caused by the inflammation reactions induced by the membrane. The ratio of COL I to COL III was the highest in the PTA-GEL group, confirming optimal tendon healing outcome of the PTA-GEL membrane.
At present, most research on barrier membranes just focuses on the enhancement of their adhesion resistance property by improving the physical and chemical properties or loading anti-fibrosis/anti-inflammatory drugs, ignoring the promotion of inner tissue healing. 47 Only a few barrier membranes are designed with both antiadhesion and prohealing functions. 2a,7a, c However, the bioactivities of these barrier membranes are limited since the membranes can only inertly perform above functions by tailoring their pore structure, surface charge, or surface hydrophilicity. Compared with these membranes, the Janus electrospun membrane prepared in this study had higher biological activities toward macrophages, fibroblasts, and tenocytes, and the regulation properties were rationally assigned to the two functional faces. TA/Fe 3+ particles endowed the outer face of the membrane with robust anti-inflammation and anti-fibrosis functions, leading to excellent adhesion prevention capability. The composition of the inner layer prevented the diffusion of the TA molecules to the inner tissue by virtue of the hydrogen bonding barrier effect, which could protect the healing of the inner tissue from the interference of TA molecules. In addition, since the oriented topology of the inner face was beneficial for cell proliferation and differentiation, the oriented inner face of the Janus membrane could facilitate the healing of inner tissue. The tendon injury model has confirmed an outstanding therapeutic effect of the Janus electrospun membrane on adhesion prevention and repair promotion. Besides tendon treatment, the Janus electrospinning membrane can be applied for other adhesion-related diseases. For example, it can be used to treat the abdominal wall defect, where the outer face can prevent the abdominal adhesions and the inner face can promote the abdominal wall to repair, as shown in Supplementary Figure S22. Moreover, the fabrication method of the Janus electrospun membrane is simple, and the composites of the membrane are of great biocompatibility and low cost. Therefore, considering the function, therapy efficacy, preparation method, biosafety, and cost of the materials, the Janus electrospun membrane prepared in this study holds great promise for clinical translation.

CONCLUSION
In this study, a Janus electrospun membrane was fabricated by combining random PCL fibers incorporated with TA/Fe 3+ particles as the outer layer, with oriented GEL fibers as the inner layer. The Janus membrane presented a unidirectional drug-release manner on account of the hydrogen bonding barrier effect provided by the inner layer, and asymmetric biological behaviors of the two functional faces were achieved. The outer face of the Janus membrane could regulate inflammatory response and inhibit the adhesion and proliferation of fibroblasts, while the inner face of the Janus membrane could protect the tendon stem cells from the interference of TA, and the oriented topology of which promoted the proliferation and differentiation of the tendon stem cells. By wrapping the injured tendons, the Janus membrane showed prominent in vivo performance. Peritendinous inflammation and adhesion were remarkably decreased by the Janus membrane via downregulating PI3K/AKT and TGF-β1/Smad signaling pathways. Specifically, on the 21 st day postsurgery, the adhesion score of the Janus membrane-treated group was 1.70 ± 0.33, notably lower than the control group (4.55 ± 0.35). Besides, the healing of the injured tendons was significantly improved by the Janus membrane. This unidirectional Janus strategy combining adhesion prevention and repair promotion capabilities holds promising potential not only in tendon injury treatment but also in internal tissue repair and anti-postoperative adhesion.

EXPERIMENTAL SECTION
Materials and experimental methods are provided in the Supporting Information.

C O N F L I C T O F I N T E R E S T S TAT E M E N T
The authors declare no conflict of interest.