GT type A (V900863, 300 Bloom from porcine skin), PCL (440744, average M _w = 80,000 Da), acetic acid (HAc; 338826, ≥99.8%), and 2, 2, 2-trifluoroethanol (TFE; T63002, ≥99.0%) were purchased from Sigma Aldrich (United States) and were used as received without further purification. Dulbecco’s modified Eagle medium nutrient mix F12 (DMEM/F12, 11330033), fetal bovine serum (FBS; 10099-141C), penicillin-streptomycin (15140-122), and trypsin (25200-056) were purchased from Thermo Fisher Scientific (United States). Anti-cardiac troponin T (ab45932) and anti-vimentin (ab92547) primary antibodies were purchased from Abcam (United Kingdom). Anti-rabbit IgG fluorescent secondary antibody (4412S) was purchased from Cell Signaling Technology (United States). 4′, 6-Diamidino-2-phenylindole (DAPI; C1002) was purchased from Beyotime Biotechnology (China). The neonatal rat/mouse cardiomyocyte isolation kit (nc-6031) was purchased from Cellutron Life Technologies (United States). Live/dead cell viability assay kits (L3224) were purchased from Thermo Fisher Scientific (United States). Cell Counting Kit-8 (CCK-8; CK04) was purchased from Dojindo Laboratories (Kumamoto, Japan). A modified hematoxylin-eosin (H-E) staining kit (G1121) was purchased from Solarbio Technology (China).

GT and PCL with mass ratios of 30:70, 50:50, and 70:30 were dissolved in TFE to produce a total concentration of 10% (w/v), and the solution was supplemented with a small amount of HAc (0.2 vol%) to obtain a transparent GT/PCL solution. In the case of pure-PCL membrane, PCL was dissolved in TFE at a concentration of 12% (w/v) without HAc. After 48-h stirring, the solutions were filled into a 10-ml syringe with a blunt needle (20G) and fed into the electrospinning device at 2 ml/h using a syringe pump (KDS100, KD Scientific, United States). A rotating stainless-steel drum (diameter = 5 cm, rotation speed = 100 rpm) was used as a collector to prepare membranes with uniform thickness. A high voltage (TXR1020N30-30, Teslaman, China) was applied between the needle tip and grounded collector, with the distance between tip and collector maintained at 13 cm. Other detailed electrospinning parameters in fabrication of different GT/PCL nanofibrous membranes were listed in Table 1. Aluminum foil was wrapped on the collector before electrospinning to facilitate membrane collection and tailoring. The obtained membranes were placed in a vacuum oven for at least 1 week to remove the residual solvent for subsequent use.

Membrane morphologies were observed by scanning electron microscopy (SEM; JSM-5600LV, JEOL, Japan) at an acceleration voltage of 10 kV. Prior to imaging, the membranes were sputter-coated with Pt for 60 s to increase their conductivity. Fiber diameters and pore sizes were determined from SEM images.

Attenuated total reflection Fourier transform infrared (FT-IR) spectra were recorded on an FT-IR spectrometer (Nicolet-Nexus 670, Thermo Fisher Scientific, United States) using a scan range of 500–4,000 cm⁻¹ and a resolution of 2 cm⁻¹.

Membrane surface hydrophilicity/hydrophobicity was evaluated by a video contact angle analyzer (Attension Theta, Biolin Scientific AB, Finland). Briefly, deionized water droplets (3 μL) were automatically dispensed onto the membrane surface, and the dynamic changes in their shape were recorded.

Given that the membranes would be applied in a moist environment at 37°C in vivo, their post-implantation shrinkage behavior was tested by immersing 1.5 cm × 1.5 cm membrane pieces with attached aluminum foil into physiological saline at 37°C. After 24 h, the membranes were removed and imaged using a high-resolution camera (EOS 6D, Canon, Japan). For post-implantation mechanical property evaluation, the membranes were cut into 5 cm × 1 cm pieces, immersed into physiological saline at 37°C for 24 h as well, loaded on a biomechanical testing machine (Instron-3343, Norwood, United States), and stretched as previously described (Feng et al., 2020).

All quantitative data were presented as means ± standard deviations. One-way analysis of variance followed by Tukey’s post hoc test was used to determine statistically significant differences between groups. Significant difference was considered at *p < 0.05; **p < 0.01; ***p < 0.001.

As shown in Figure 1A, all fibers exhibited a smooth surface without beads or bonding. The pure-PCL (0:100) membrane fibers were relatively thick (average diameter = 785.5 ± 401.5 nm), while thinner fibers were present in GT-containing membranes (average diameters = 304.7 ± 98.1 and 355.7 ± 81.0 nm for GT:PCL = 30:70 and 50:50, respectively). However, at a ratio of 70:30, the fiber diameter increased to 494.7 ± 166.6 nm. As for the pore size, which is an important parameter for selection of membranes as barriers to prevent cell penetration, increased with the increasing fiber diameter (Figure 1B), equaling 9.9 ± 2.3, 4.8 ± 1.2, 5.2 ± 1.4, and 7.8 ± 1.6 µm for different ratios of 0:100, 30:70, 50:50, and 70:30, respectively. Membrane composition was probed by FT-IR (Figure 1C). The characteristic peaks of PCL at 2,943 cm⁻¹ (asymmetric CH₂ stretch), 2,865 cm⁻¹ (symmetric CH₂ stretch), and 1,724 cm⁻¹ (C=O stretch) were observed for all membranes, while the characteristic peaks of GT at 1,652 cm⁻¹ (amide I) and 1,540 cm⁻¹ (amide II) were only observed for GT/PCL membranes. The amide I band was attributed to the random coil and α-helix conformations of GT. As expected, the areal fractions of PCL- and GT-specific peaks decreased and increased with increasing GT content, respectively.

The 0:100 membrane was highly hydrophobic, and the water droplets deposited thereon formed a large obtuse contact angle (137.0 ± 7.0°) and remained stable. The incorporation of GT greatly increased hydrophilicity, and all GT/PCL membranes presented hydrophilic surfaces with acute contact angles (Figure 2). In the case of the 30:70 membrane, water was rapidly absorbed to afford a contact angle of zero within 30 s. However, the rate of water absorption decreased at higher GT contents, in line with the fact that water absorption by pure-GT membranes is also insufficient (Feng et al., 2012).

As the membranes were implanted in vivo, they were immersed into saline held at 37°C to characterize shrinkage under simulated physiological conditions. Compared with their original size (i.e., the size of the underlying aluminum foil), the 0:100 membrane did not shrink, while 30:70, 50:50, and 70:30 membranes slightly contracted, which indicated that all membranes had good thermal stability (Figure 3A). However, these membranes showed different mechanical properties (Figure 3B; Table 2). The introduction of GT greatly improved the mechanical properties of membranes with the increased maximum tensile strength (Figure 3C) and strain at break (Figure 3E), while significant mechanical property deterioration was observed at a GT content of 70%, as exemplified by the decreased maximum tensile strength (Figure 3C) and Young’s modulus (Figure 3D). In particular, the maximum tensile strength decreased in the order of 4.4 ± 0.4 MPa (30:70) > 4.1 ± 0.6 MPa (50:50) > 1.4 ± 0.1 MPa (0:100) > 0.7 ± 0.2 MPa (70:30).

The viability of cardiomyocytes and cardiac fibroblasts on membranes was assessed using the live/dead cell assay. In the obtained images, green fluorescence represents live cells, while red fluorescence represents the dead. After 1 day of culturing, both cardiomyocytes and cardiac fibroblasts adhered well to all membranes, as exemplified by the numerous green dots (live cells) and few red dots (dead cells) (Figure 4A). The cells continued to grow with time, reaching ∼90% (cardiomyocytes) and 100% (fibroblasts) confluency on all membranes with bright green fluorescence on day seven. In addition, cell seeding efficiencies reached ∼70% (cardiomyocytes) and 80% (fibroblasts) on all membranes without significant differences (Figure 4B).

The morphology and proliferation of cells cultured on membranes were evaluated using immunofluorescence staining (Figure 5A). Troponin T (cardiomyocytes) and vimentin (fibroblasts) staining showed that both cells well adhered and spread on all membranes on day one, although a larger fibroblast spreading area was observed for the 70:30 group. With increasing cultivation time, cardiomyocytes and fibroblasts rapidly proliferated on all membranes at day five. In line with the results of immunofluorescence staining, those of the CCK-8 assay (Figure 5B) revealed that cardiomyocytes and fibroblasts exhibited steady and continuous proliferation without significant difference between groups, except for the downward trend in the 0:100 group at day seven for cardiomyocytes.

Collectively, these results indicated that all GT/PCL membranes had good biocompatibility with cardiomyocytes and cardiac fibroblasts and were suitable for applications in cardiac surgery.

One month after surgery, the heart/liver/kidney functions and CRP/immunoglobulin lever assessments were used to detect whether the membranes would cause adverse reactions in vivo implantation. Echocardiography (Figures 6A,B) showed that compared with the normal group (healthy rabbits that did not undergo surgery), the heart function indicators LVEF and LVFS in the positive control and 70:30 groups were significantly reduced, whereas no significant differences were observed among the 0:100, 30:70, and 50:50 groups. Conversely, there were no significant differences in liver (Figure 6C) or kidney (Figure 6D) function indicators among all groups, which suggested that the membranes were not toxic to the liver and kidneys. Additionally, the results of CRP and immunoglobulin level evaluation (Table 3) were not significantly different among all groups either, indicating that the membranes did not induce strong inflammation reaction and immune rejection response after implantation in vivo. Collectively, these results demonstrated that all the membranes exhibited sufficient biosafety for implantation in vivo, except for the 70:30 group, in which case a reduction in heart function (i.e., in LVEF and LVFS) was observed.

After echocardiography and blood examination, animals were anesthetized and sacrificed. The thoracic cavity was reopened, and adhesions between the heart and sternum were separated by dissection. In the positive control group, severe adhesions were formed at the injury site of the previous operation (Figures 7A,E), and more attention had to be paid during dissection to keep the heart intact. Strong individual differences were observed in the 0:100 group, i.e., some rabbits had no or slight adhesions that could be easily separated by blunt dissection (Figures 7B,F); whereas others had thick and heavy adhesions requiring continuous sharp dissection (Figures 7C,G). However, mild filamentous adhesions were observed in 30:70 (Figures 7D,H) and 50:50 (Figures 7I,K) groups, the membranes could be easily separated from the heart and maintained their original shape with a white appearance. In the 70:30 group, intense and solid adhesions were observed, and the original membrane shape was hardly recognizable due to the rapid degradation of GT (Figure 7J). Broken holes and bleeding were easily induced upon membrane detachment from the sternum due to the poor mechanical properties of this membrane (Figure 7L). Furthermore, according to the results of adhesion grade scoring (Figure 7M), the adhesions were severe in the positive control (score = 2.6 ± 0.9) and 70:30 (2.6 ± 0.5) groups. Much lower scores were observed for 30:70 (1.2 ± 0.4) and 50:50 (1.2 ± 0.8) groups, while a very large score fluctuation was observed for the 0:100 group (1.4 ± 1.5).

After overall observation, the membranes were carefully removed from surrounding tissues. Histological analysis demonstrated that though all the membranes had the same thickness before surgery, the thickness of 0:100 membrane significantly exceeded than others after implantation. Numerous inflammatory cells have entirely infiltrated into the 0:100 membrane regardless of whether adhesions were slight or strong (Figure 7N). High-magnification imaging showed that the membrane was divided into masses of small islands, many macrophages gathered and ranged along islands to form multinucleated giant cells thereby inducing membrane disintegration and phagocytosis (Figure 7R). In the 30:70 and 50:50 groups, obvious pink barriers were observed with several (30:70, Figures 7O,S) or no (50:50, Figures 7P,T) infiltrating cells and a small number of inflammatory cells and sporadic blood capillaries distributed on both sides of barriers. While in the 70:30 group, many monocytes, multinucleated giant cells, fibroblasts, blood vessels and muscular tissues have invaded into the membranes, and no barriers were formed (Figures 7Q,U).

Considering the 70:30 membrane was not only difficult to suture during surgery but also could not maintain its structural integrity to prevent cardiac postoperative adhesion after being implanted in vivo, and therefore, this membrane was not investigated further.

Echocardiography (Figures 8A,B) showed that compared with the normal group, the heart function (i.e., LVEF and LVFS values) was significantly reduced in the positive control group, while no significant differences were observed for other three groups (0:100, 30:70, and 50:50). Further blood examination showed that the liver/kidney function indicators (Figures 8C,D) of normal group were not significantly different from those of other groups, and CRP levels were still less than 1 mg/L for all groups. These results were consistent with those obtained 1-month after surgery, which illustrated the high biosafety of PCL and GT/PCL (30:70 and 50:50) membranes for long-term in vivo implantation.

Overall, dense and solid adhesions were observed between the pericardial defect area and sternum in the positive control group (Figures 9A,D). These adhesions were more serious than those formed 1-month after surgery and could only be separated by sharp dissection. And the adhesion conditions still showed large variations between individual animals in the 0:100 group, i.e., some rabbits had no or slight adhesions (Figures 9B,E), while the others had severe adhesions (Figures 9C,F). However, filamentous and sparse adhesions were observed between the membranes and epicardium in 30:70 (Figures 9G,I) and 50:50 (Figures 9H,J) groups, which could be easily separated by digital dissection. Specifically, the membrane became blurred and developed a yellow-like appearance similar to that of the native pericardium tissues. The heart was smooth and moist, and the coronary arteries were clearly visible. The related adhesion grade scores were shown in Figure 9K. The adhesion was strongest in the positive control group (score = 2.2 ± 1.1), with lower scores observed for 30:70 (1.0 ± 0.6) and 50:50 (0.8 ± 0.8) groups. As in the case of 1 month after surgery, a large standard deviation was observed for the 0:100 group (1.3 ± 1.4).

Histological staining showed that the 0:100 membrane gradually degraded, and most islands have disappeared (Figure 9L). Multinucleated giant cells were much bigger and more visible than 1 month after surgery, forming many dark purple (nucleus) bumps (Figure 9O). In the 30:70 group, some inflammatory cells together with small arteries and blood capillaries infiltrated the membrane, and the barrier became blurred (Figures 9M,P). However, an obvious pink barrier with no cell infiltration was observed in the 50:50 group (Figures 9N,Q), indicating that this membrane could act as a perfect barrier to prevent cell infiltration, which was in line with the results obtained 1 month after surgery.