Esophageal polymeric scaffolds were fabricated by electrospinning as described previously (Fig. 2A) [10]. Briefly, 30 wt% polyethylene oxide (Mw=100,000, Sigma) solution in distilled water was electrospun onto rotating stainless steel mandrels (diameter=2 cm) to 2 mm thickness. Then, 20 wt% PU (Pellethane, Lubrizol LifeSciences) solution in N,N-dimethylformamide (Junsei Chemical Co.) was also electrospun onto the collector at the same thickness. The inner layer was made at a linear speed of 0.31 m/s (1,200 rpm) so that random fiber arrangement could occur. On the other hand, the outer layer rotated the mandrel at a linear speed of 3.14 m/sec (12,000 rpm), which led to a circumferential orientation of the fibers. In both cases, the feeding rate of the solution was fixed to 0.5 mL/hr. The distance between the nozzle and flat sheet collector was set precisely at 25 cm, with 15 kV generated by a power supply. The prepared double-layered esophageal scaffold was dried overnight in a vacuum oven (38 °C) to remove residual organic solvent. The mechanical strengths of two-layered PU scaffolds were measured using a tensile test machine (AG-5000G; Shimadzu) which was equipped with a 50-kgf load cell at a crosshead speed of 10 mm/min. Specimens were prepared in the form of a standard dumbbell-shape (n=4).

For cell seeding, scaffolds were sterilized by soaking in 70% ethanol under ultraviolet (UV) irradiation. After 2 days of pre-wetting with basal medium, scaffold were coated with 10 μg/mL fibronectin (Sigma-Aldrich) for 1 hour to promote cell adhesion. Then, human adipose-derived mesenchymal stem cells (hMSCs, R7788-110; STEMPRO) at passage 4 were suspended at a density of 1×10⁶ cells/mL in Matrigel (354234; Corning) containing growth media (MesenPRO RS, Basal medium with growth supplement), and the suspension was evenly coated on the outer and inner layers of the polymer scaffolds. Cell seeded scaffolds were then incubated under a horizontal rotation at 1 rpm overnight for stable cell attachment. Subsequently, the esophageal scaffold was removed from the rotating system, and transferred into the chamber in a bioreactor device (Fig. 2B). Mechanical stimuli were applied with a hollow organ bioreactor along with same growth medium inside and outside of the scaffold as described in our previous study (Fig. 2C) [10]. A custom-designed perfusion-type bioreactor was designed to maintain steady pulsatile flow (Fig. 2D) [5]. Pulsatile flow can cause shear stresses in the esophageal grafts as well as intraluminal pressure and circumferential stretching for simultaneous differentiation into epithelial and muscle lineages. The differentiation behavior of these hMSCs into epithelial cell and smooth muscle cell lineages following the administration of chemical agents under bioreactor cultivation has been demonstrated in previous studies [10]. Prior to esophageal transplantation, the hMSC-seeded tubular scaffolds were incubated under a bioreactor system for 3 days. Cell cytotoxicity on the two-layered PU nanofiber scaffold (inner & outer surface) without bioreactor cultivation was analyzed after 5 days using a live and dead viability/cytotoxicity assay (Molecular Probes) according to the manufacturer’s instructions. hMSCs were grown on tissue culture plate to be used as controls. Images were captured via confocal microscopy (LSM510 META, Carl Zeiss). A CCK-8 assay kit (Dojindo) was used to determine cell proliferation on days 2 and 5. Each specimen was transferred to a new tissue culture plate (12-well) and 1ml of fresh medium was added to each well. Then, a 10% CCK-8 solution was filled and incubated at 37 °C for 2 hours. The reaction solution (100 μL) was transferred to a 96-well plate and by using a microplate reader (Thermo Scientific), the absorbance (OD value) was measured at the wavelength of 450 nm.

This study was carried out in strict accordance with the guidelines of the Animal Research Committee, Seoul National University Hospital. All protocols and experimental design parameters were reviewed and approved by the Institutional Animal Care and Use Committee of the Seoul National University Hospital (No. 17-0204-S1A0).

Four healthy male beagles aged 2 years and weighing 17.50–25.50 kg were used (OrientBio). The animals were divided into two groups as follows: (1) implantation of PU nanofiber scaffold (Graft group) (n=4), (2) for histological analysis, normal esophagus was collected at 5 cm away from the transplant site to replace the normal group (normal group). In addition, the animals were euthanized at 6 and 12 months after transplantation. All operative procedures were performed under general anesthesia following 24 hours of fasting. Next, a 10-mg subcutaneous injection of nalbuphine (Nalbuphine, Merck), an intramuscular injection of 0.3 mg/kg of midazolam (Hypnovel, Roche), and 5 mg/kg of ketamine (Ketamine Virbac) were administered. Endotracheal intubation was then achieved with 0.2 mL/kg/hr of propofol (Diprivan 1%, AstraZeneca). During surgery, all animals were perfused with a balanced crystalloid solution (10 mg/kg/hr), which was suppressed at the end of the operative procedure. Anesthesia was maintained by inhalation of 1% isoflurane and additional propofol. Mechanical ventilation were set to a total volume of 10 mL/kg, with a respiratory rate of 10–20 breaths/min and 60% of the fractional inspired oxygen. One intravenous ampoule of ampicillin (1 g)-sulbactam (0.5 g) was administrated before surgery as antibiotic therapy. Analgesia was provided by an intramuscular injection of 13 mg of nalbuphine (Nalbuphine1, Merck) before surgery. The detailed surgical procedure is as follows. The neck was shaved, and standard draping of the surgical site was performed for aseptic surgery. A ventral midline incision was made, and the strap muscle was divided to isolate the tracheoesophageal structure (Fig. 2E). A circumferential 5×3 cm-long resection of one-third of a partial full-thickness esophageal defect including esophageal muscle and mucosa was created (120 degrees of circumference). Reconstruction was supported by interrupted suture for the anastomoses of the 5×3 cm long PU graft (Fig. 2F). Esophageal graft was then covered with a sternocleidomastoid muscle flap for stable fixation and vascular supply of the implanted site (Fig. 2G). The muscle and skin tissues were closed with a 4-0 Vicryl suture.

The dogs recovered from anesthesia, were extubated and were monitored in the recovery room until they were resting comfortably in the sternal position. The dogs were kept in a cage overnight and returned to their larger run housing on postoperative day one. The dogs were given prophylactic antibiotics consisting of cefazolin (1 g/day intramuscularly) for 10 days. Analgesia was provided by intramuscular injection of 13 mg of nalbuphine 2/day (Nalbuphine) until postoperative day 5 and thereafter for analgesia as needed. NPO (nothing per oral) was continued for 6 d after surgery with intravenous total parenteral nutrition support (3,000 kcal/day, Combiflex Lipid Peri Inj) (Supplementary Fig. 1A). The oral liquid diet was started 7 d after surgery (Supplementary Fig. 1B, Supplementary Video 1), a half solid diet was started on the 10th day, and a normal diet (Oriental Yeast) was resumed 3 weeks after surgery.

Postoperative monitoring was performed daily of the animal’s survival and weight, assessing respiratory function and dietary conditions, behavior, and surgical complications such as signs of leakage. Animals were killed if the serious complication was not cured within 1 week. Intravenous administration of potassium chloride (KCL; 1–2 mmol/kg) was used to the euthanize the dogs under anesthesia at 6 and 12 months following surgery.

Esophagogastroduodenoscopy (EGD) was performed at 1, 6, and 12 months after implantation to evaluate esophageal structure and function. The animals were anesthetized (by 20 mg/kg ketamine alone) for EGD. EGD was systematically performed to evaluate esophageal patency and assess the inner aspect of the graft sites.

In order to evaluate the functional regeneration of the neoesophagus, a computed tomography (CT; GE Medical Systems) scan after the gastrografin swallow was performed at 1 month, 3 months and 6 months after implantation. The contrast medium, an amidotrizoic acid (Gastrografin; Schering) mixed with milk, was given 15 minutes before image scanning. The patency of the neoesophagus and food passage were evaluated.

The dogs were euthanized at 6 and 12 months after implantation. Immediately after euthanasia, the cervical esophagus, the grafts and the surrounding tissues were harvested. Scanning electron microscopy (SEM) analysis was performed to observe the morphological characteristics of the regenerated esophageal mucosa. Specimens were prefixed by immersion in 2% glutaraldehyde in 0.1 M phosphate buffer, pH 7.4, and were immediately postfixed for 2 hours in 1% osmium tetroxide (OsO₄; Sigma Aldrich) dissolved in phosphate buffered saline. Samples were treated in t-butyl alcohol after dehydration through a graded series of ethanol and dried in a freeze dryer. The fixed esophagus specimens were coated with platinum using a sputter coater (IB-5, Eiko) under argon atmosphere conditions. Samples were observed by an SEM (Model S-3000N, Hitachi) operated at an accelerating voltage of 25 kV.

The autopsy was carried out after the scheduled sacrifice. The gross finding of the grafts were accessed, and the graft sites and normal cervical esophagus were immediately fixed in 4% paraformaldehyde, embedded in paraffin, and sectioned serially into 5 μm-thick sections in the transverse direction. The tissue slides were deparaffinized and dehydrated in a graded series of alcohol (in phosphate-buffered saline [PBS]) for 5 minutes each. The tissue slides were then stained with hematoxylin and eosin (H&E) and Masson’s trichrome, following standard histological procedure. Elastin staining was also performed using the Elastin (Modified Verhoff’s) stain kit (ES4807) according to the manufacturer’s instructions. Histological images were captured in triplicate for each group using a light microscope (Olympus).

For immunohistochemistry, the specimens were immersed in 3% hydrogen peroxide (H₂O₂) in methanol for 30 minutes at room temperature to inactivate endogenous peroxidase. The slides were washed with PBS and then incubated with 3% bovine serum albumin to block nonspecific responses. The tissue sections were subsequently reacted with anti-keratin 5 (1:200 dilution, rabbit polyclonal antibody; Abcam) along with the anti-mouse Alexa Fluor 594 secondary antibodies (ab150116, Abcam). The thickness of the regenerative esophageal epithelium was calculated through the area indicated in red using the ImageJ tool bar (n=5). Tissue sections for alpha-smooth muscle actin (α-SMA; 1:50 dilution; mouse monoclonal antibody; ABIN126702) and von Willebrand factor (vWF; 1:200 dilution; rabbit monoclonal antibody; Abcam) were subsequently incubated using the horseradish peroxidase-conjugated kit (Vectastain), and they were visualized using the chromogenic substrate 3,3′ diaminobenzidine (Vector, pk-7800). Hematoxylin was used to counterstained cell nuclei. Images were captured by a fluorescence microscope (BX43-32FA1-S09; Olympus Optical, Japan). α-SMA+esophageal muscle was calculated from the brown-color area around the graft site using ImageJ (n=6). The number of vWF-positive vessels was calculated using ImageJ software. Five areas (high-power field; at ×400 magnification) were measured per slide (n=6 per group) by a blinded observer.

The data are expressed as median [interquartile range]. Statistical significance was determined via one-way analysis of variance with Kruskal-Wallis nonparametric test (GraphPad Prism 5, GraphPad Software).

The mechanical properties of two-layered PU nanofiber (dumbbell-shape, ASTM D412) were examined by the tensile test (Supplementary Table 1). The tensile strength was measured at 12.25± 1.86 MPa, the Young modulus at 1.57±0.25 MPa, and elongation at 48.4%±7.8%. The low Young modulus and high elongation value represented a suitable degree of flexibility for esophageal implants. The cytotoxicity of the two-layered PU nanofiber itself was also evaluated by a CCK-8 assay without bioreactor culture. After uniform cell adhesion onto the PU nanofiber, hMSCs were monitored for 5 days. To investigate cell cytotoxicity on the PU scaffold, live and dead staining was done on day 5 and showed that most hMSCs were viable and spread well on the nanofiber structure, indicating that the extracellular matrix (ECM)-like structure of the nanofibers provided a favorable environment (Supplementary Fig. 2A). In particular, hMSCs on aligned fibers (outer layer) showed greater elongation. The CCK-8 assay confirmed that hMSC proliferation increased over time on the inner and outer layers of the cylindrical PU scaffold (Supplementary Fig. 2B).

The surgical procedure was successfully completed in all dogs, and there was an uneventful immediate postoperative course in three of the four dogs. A cervical abscess from anastomosis leakage was identified after postoperative day 5 in one of the four dogs. This animal then had to be euthanized as it showed signs of uncontrollable neck swelling due to the persistence of the infection. It was assumed that the saliva leakage was caused by rupture of the implanted PU scaffold. The rest of the beagles showed no signs of esophageal stricture or infection at the final examination. They also continued to eat and drink normally until the scheduled period (12 months after surgery). The mean weight loss was 17.4%±7.7%, and all dogs regained their preoperative weight after 1 month (Supplementary Fig. 1C). The appearance and attitudes of the beagles were also very similar to the preoperative status, and they remained active (Supplementary Video 2). As the study progressed through its first 6 months, the weight of all three animals slowly increased from approximately 10–11 kg to 12–15 kg from their initial weight. There was a much greater increase in weight from the sixth month to the 12th month, with their weights at the time of sacrifice being nearly 1.5 times the initial weights.

The endoscopic examinations of the dogs showed the rapid formation of a normal-appearing mucosal surface by approximately 1 month postoperatively (Fig. 3A). No evidence of active inflammation, erosion, scarring, or necrosis was observed in the endoscopic examinations of the animals’ esophagus. Throughout the 12-month study, endoscopy was performed, and the images exhibited complete re-epithelization of the regenerated tissue with what appeared to be a complete epithelium at the 12-month point (Fig. 3A). At 1, 6, and 12 months after implantation, CT demonstrated a normal lumen with no signs of leakage or stricture (Fig. 3B).

The gross findings of the esophageal mucosa from the collected esophagus samples showed clear normal mucosal folds, whereas the graft group intermittently exhibited rough surface scars (left images in Fig. 3C). We conducted SEM analysis to interpret the morphology of these regenerated mucosal layers from a more microscopic perspective. Typical morphology of the wavy lumen was confirmed in the normal mucosa. The morphology of the graft group showed non-directional rough wrinkles.

Whole histology of the regenerated esophagus was examined by H&E staining (Fig. 4A). The regenerated esophageal epithelium was revealed to be highly complex in wrinkles and arranged atypically. The regenerated submucosa was considerably thick. However, the transplanted scaffold around the regenerated submucosal layer could not be found. It is presumed that the grafted scaffold was detached from the regenerated submucosal layer by continuous peristalsis of the esophagus. In particular, esophageal gland remodeling was observed sporadically at 6 months after transplantation (yellow arrows) (Fig. 4B). The morphology and thickness of the regenerative esophageal epithelium showed similar morphology to that of the normal epithelium through Masson’s trichrome staining (Fig. 4C).

Regeneration of the esophageal epithelium at 6 and 12 months after transplantation was identified by keratin-5 immunostaining. Fig. 5A represents the histological morphology of the regenerated esophageal epithelium, indicated in red. The newly regenerated epithelial layer was significantly hyperplastic compared to the normal epithelium at the implanted sites. In the next 12 months, the histological morphology of the stratified squamous epithelium became more similar to that of the normal epithelium (inset box). Quantitative analysis showed that the epithelial thickness tended to decrease slightly at 12 months compared with the thickness at 6 months (Fig. 5B). It was clear that the ECM-like properties of the PU nanofiber scaffolds had a structural effect on improving re-epithelialization.

Elastin staining confirmed that remodeling of elastin fibers had occurred in the newly formed submucosal layer (Fig. 5C). There was an abundant distribution of elastin fibers at 6 months, and morphology similar to that of normal tissue was confirmed at 12 months. Additionally, at 12 months after transplantation, the elastin distribution was not significantly different from that observed in normal tissue (Fig. 5D). The abundant presence of elastin components means that the elastic properties of the regenerated esophageal tissue were also similar to those of normal tissue.

α-SMA immunostaining was performed to investigate whether the tissue-engineered esophagus could affect esophageal muscle regeneration. Strong brown colors were observed around the implanted sites (Fig. 6A). α-SMA⁺ muscle was observed intermittently 6 months after transplantation. In particular, the alignment of the regenerated muscles was irregular and abnormal. In comparison, with Masson’s trichrome staining, the 6-month sample showed a dense penetration of collagenous tissues (blue color) between muscle layers (red color) (Fig. 6B). It was also difficult to distinguish between the circular and longitudinal muscles found in normal muscles. However, the amount of regenerated muscle increased over time (Fig. 6C). It is assumed that the hMSC-inoculated PU scaffold may have been effective for quantitatively increasing the amount of muscle regenerated to replace the lost esophageal muscle. However, it was not possible to accurately characterize the structural advantages of the inner and outer layers of the scaffold for muscle regeneration.

Immunostaining for vWF, an endothelial marker, was performed to confirm neovascularization in the esophageal submucosa, including the lamina propria, around the implanted site (Fig. 6D). New blood vessels were abundantly observed at both 6 and 12 months around the regenerated esophageal submucosa (black arrows). Quantitatively, both time points showed significantly higher neovascularization than in the normal esophageal layer (Fig. 6E).