The bioengineering of perfusable endocrine tissue with anastomosable blood vessels

Organ transplantation is a definitive treatment for endocrine disorders, but donor shortages limit the use of this technique. The development of regenerative therapies would revolutionize the treatment of endocrine disorders. As is the case for harvested organs, the ideal bioengineered graft would comprise vascularized endocrine tissue, contain blood vessels that could be anastomosed to host vessels, have stable blood flow, and be suitable for transplantation into various sites. Here, we describe a transplantable endocrine tissue graft that was fabricated by ex vivo perfusion of tricultured cell sheets (islet β-cells, vascular endothelial cells (vECs), and mesenchymal stem cells (MSCs)) on a vascularized tissue flap of in vivo origin. The present study has three key findings. First, mild hypothermic conditions enhanced the success of ex vivo perfusion culture. Specifically, graft construction failed at 37 °C but succeeded at 32 °C (mild hypothermia), and endocrine tissue fabricated under mild hypothermia contained aggregations of islet β-cells surrounded by dense vascular networks. Second, the construction of transplantable endocrine tissue by ex vivo perfusion culture was better achieved using a vascular flap (VF) than a muscle flap. Third, the endocrine tissue construct generated using a VF could be transplanted into the rat by anastomosis of the graft artery and vein to host blood vessels, and the graft secreted insulin into the host’s circulatory system for at least two weeks after transplantation. Endocrine tissues bioengineered using these techniques potentially could be used as novel endocrine therapies.


Introduction
The endocrine system comprises internal glands that secrete hormones into the vasculature to regulate distant target organs. Permanent hormone deficiency is usually treated by the administration of a hormonal preparation, but lifelong therapy is required. Although organ transplantation is a definitive treatment for hormone deficiency, donor shortages limit the availability of this strategy. Therefore, regenerative therapies are being developed to overcome these limitations.
Endocrine cells for transplantation can be readily obtained by the induction of pluripotent stem cells, but the development of an effective method of transplanting endocrine cells is critical. Transplantation of a cell suspension results in cell dispersion and poor angiogenesis [1,2], which compromise cell engraftment and functionality. Techniques based on cell sheets [3] and cell aggregates [4] can improve engraftment after transplantation. Several studies have demonstrated hormone secretion by bioengineered endocrine tissue, including islet β-cell sheets [5], islet β-cell spheroids [6] and islet-like organoids [4,7] in models of diabetes mellitus, and thyroid cell sheets [8,9] and thyroid gland organoids [10] in models of hypothyroidism.
Previously reported methods for the fabrication and transplantation of endocrine tissues rely on diffusion for oxygen/nutrient delivery to the transplanted graft until it becomes vascularized by blood vessels that grow from the recipient tissue. Subcutaneous tissue has been suggested as a suitable site for islet cell transplantation because the surgical procedure is not highly invasive and the graft can be easily monitored or removed in the event of a complication [3,11,12]. However, subcutaneous tissue is not rich in blood vessels, hence stable implantation is difficult to achieve using transplantation methods reliant on diffusion [11][12][13][14]. Furthermore, poor engraftment of islets can occur even in sites with a rich blood supply such as the renal capsule [15].
A free flap is often used during reconstructive surgery to cover a large tissue defect [16,17]. Engraftment of a tissue flap does not depend on oxygen/nutrient diffusion from surrounding tissues because the flap is perfused immediately after transplantation following the anastomosis of its blood vessels to those in the transplantation site. Therefore, we hypothesized that endocrine cell sheets and ex vivo perfusion methods could be used to fabricate vascularized endocrine tissue with an anastomosable artery and vein that would engraft well after transplantation.
There are no reports describing the ex vivo fabrication of endocrine tissue with anastomosable vessels as a free flap. One possible reason for this is the oxygen demand of endocrine cells [18][19][20]. In particular, pancreatic beta cells require a high oxygen supply to maintain their endocrine function and have been reported to undergo apoptosis in a hypoxic environment [18,21]. We hypothesized that the high cellular metabolism and oxygen demand under normothermic conditions (37 • C) might compromise the viability of endocrine cells when they are cultured at the high densities needed for organization into functional tissue. Since mild hypothermia reduces cellular metabolism and oxygen consumption [22], we evaluated whether lowering the culture temperature would facilitate the construction of three-dimensional (3D) endocrine tissue.
This proof-of-concept study aimed to establish whether a transplantable endocrine graft could be generated by ex vivo perfusion culture [23,24] of bioengineered cell sheets on a vascularized free flap. First, we determined whether endocrine tissue could be successfully constructed on a muscle flap (MF) [23], and the success of perfusion culture was examined at two different temperatures: normothermia (37 • C) and mild hypothermia (32 • C). Notably, we found that graft construction failed at 37 • C but succeeded at 32 • C. Next, we developed a vascular flap (VF) as an alternative to the MF, because the MF-derived construct exhibited leakage and necrosis that might increase the risk of hemorrhage or toxemia after transplantation. Our experiments indicated that the VF was more suitable for transplantation than the MF. Finally, we demonstrated that the VF-derived endocrine tissue construct could be transplanted into the rat by anastomosis of the graft artery and vein to host blood vessels. Importantly, the graft secreted insulin into the host's circulatory system for at least two weeks after transplantation (figure 1).

Ethics
All animal experiments were performed in accordance with protocols approved by the Ethics Committee for Animal Experiments of Tokyo Women's Medical University and complied with the ARRIVE guidelines for the care and use of laboratory animals. All animals were housed in individual cages and maintained at a constant temperature and humidity under a 12 h light cycle. Animals were given free access to food and water. Animal euthanasia was performed by exsanguination under 5% isoflurane in accordance with the American Veterinary Medical Association euthanasia guidelines.

Evaluation of the viability of cocultured cells and tricultured cells
35 mm cell culture dishes coated with fetal bovine serum (FBS) overnight were seeded with iGL cells and hASCs (3:2 ratio; 1.5 × 10 6 cells in total) for coculture or with iGL cells, GFP-HUVECs and hASCs (9:2:6 ratio; 1.7 × 10 6 cells in total) for triculture. The cells were incubated in Dulbecco's Modified Eagle Medium/Nutrient Mixture F-12 (DMEM/F12; #10565; Thermo Fisher Scientific, Waltham, MA, USA), which contains sodium bicarbonate as a buffering agent, supplemented with 10% FBS and 1% penicillin-streptomycin. The cells were cultured at 32 • C or 37 • C in a humidified atmosphere containing 5% CO 2 for one, two or three days. Subsequently, Figure 1. Study design and main findings. Left: perfusion culture was performed at 32 • C and 37 • C to establish the optimal temperature. Cell viability and cell sheet engraftment were improved at 32 • C vs. 37 • C. Middle: a VF-based endocrine tissue construct was developed and compared with a MF-based construct. The VF-based construct exhibited no leakage during perfusion, unlike the MF-based construct. Furthermore, the levels of tissue damage-related enzymes in the perfusion medium were lower for the VF-based construct than for the MF-based construct. In addition, the VF-based construct had higher insulin secretory capacity and was more suitable for transplantation as an endocrine tissue than the MF-based construct. Right: The VF-based endocrine tissue construct was successfully transplanted into a rat by anastomosis of its vessels with those of the host, and the graft secreted insulin for at least two weeks. the cells were stained with propidium iodide (PI; Thermo Fisher Scientific), which was used as a marker of dead cells, and Hoechst 33342 (NucBlue Live ReadyProbes Reagent; Thermo Fisher Scientific). Images of the cells were captured using an Eclipse Ts2-FL fluorescence microscope (Nikon, Tokyo, Japan) at a magnification of ×100. The numbers of Hoechstpositive cells and PI-positive cells were counted in each image using ImageJ 1.53 n (National Institutes of Health, Bethesda, MA, USA), and the ratio of the number of PI-positive cells to the number of Hoechstpositive cells was calculated as the non-viable cell rate.

Evaluation of cell metabolism and vascular network construction
35 mm cell culture dishes coated with FBS overnight were seeded with iGL cells, GFP-HUVECs and hASCs (9:2:6 ratio; 1.7 × 10 6 cells in total) for triculture. The cells were incubated in DMEM/F12 supplemented with 10% FBS and 1% penicillin-streptomycin.
The cells were cultured at 27 • C, 32 • C or 37 • C in a humidified atmosphere containing 5% CO 2 for one, two or three days. The levels of glucose (Stat Strip Express 900; Nova Biomedical, Waltham, MA, USA) and lactate (Stat Express Lactate; Nova Biomedical) in 50 µl samples of culture medium were measured daily. The molar ratio of lactate production to glucose consumption (L/G ratio) was calculated by dividing the lactate production values (mmol l −1 ) by the glucose consumption values (mmol l −1 ) [25]. Images of the cells were captured using an Eclipse Ts2-FL fluorescence microscope (Nikon) at a magnification of ×4. The average length of the endothelial network in each fluorescence image was evaluated using AngioTool v0.6a (National Institutes of Health).

Preparation of tricultured cell sheets for transfer onto MFs or VFs
Tricultured cell sheets were generated using a modification of a previously described method [12]. Tricultured cell sheets were fabricated on 60 mm temperature-responsive dishes (UpCell; CellSeed, Tokyo, Japan) using iGL cells, GFP-HUVECs and hASCs at a ratio of 9:2:6. Each dish was coated with FBS overnight, and 5.1 × 10 6 cells were seeded onto the dish and cultured for 24 h at 37 • C in a humidified atmosphere containing 5% CO 2 . DMEM/F12 (#10565; Thermo Fisher Scientific) containing 10% FBS, 1% penicillin-streptomycin and 0.007 mM Lascorbic acid (Fujifilm Wako Pure Chemical, Osaka, Japan) was used as the medium. Then, the cells were incubated at 20 • C for 1 h to detach them as an intact cell sheet. Three cell sheets were stacked into a triple-layered construct for transferal onto a MF. A single-layered cell sheet was transferred onto a VF.

MF fabrication
A MF was made from rat femoral muscle including the superficial femoral artery and vein (figure 3(a), upper). Male Sprague Dawley rats (350-530 g; Sankyo Labo Service Corporation, Tokyo, Japan) were anesthetized by inhalation of 2% isoflurane. A thermal cautery unit (TCU-150; Bio Research Center, Nagoya, Japan) was used to partially resect the femoral muscle from the surrounding tissue. Branches of the femoral artery and vein distal to the femoral MF were resected, but the proximal parts of the superficial femoral artery and vein were retained. The partially resected muscle tissue was returned to its original position, and the skin was sutured. One week later, the animal was anesthetized and heparinized (400 IU kg −1 , intravenous injection), and the muscle tissue was harvested together with the proximal parts of the superficial femoral artery and vein. The rat was then euthanized.

Transplantation of tricultured cell sheets onto the MF
Three tricultured cell sheets were stacked as a triple-layered construct for transplantation onto the MF using a previously described method [23]. Immediately after the initiation of ex vivo perfusion culture, a triple-layered tricultured cell sheet construct was transplanted onto the muscle using a piece of polyethylene sheet (supplementary video 1).

VF fabrication
A VF was made using the superficial femoral artery and vein of the rat (figure 3(a), lower). Male Sprague Dawley rats (420-550 g; Sankyo Labo Service Corporation) were anesthetized by inhalation of 2% isoflurane. The superficial femoral artery and vein were anastomosed with 10-0 nylon (BEAR Medic Corporation, Ibaraki, Japan) to form an arterio-venous (A-V) shunt. The blood vessels were returned to their original position, and the skin was sutured. The animal was anesthetized and heparinized (400 IU kg −1 , intravenous injection) 2-7 d later, and the branch vessels of the superficial femoral artery and vein were ligated. The superficial femoral artery and vein, including the A-V shunt, were harvested, and the rat was then euthanized.

Wrapping of a cell sheet around the VF
All procedures were carried out under sterile conditions. Tricultured cells were detached as a cell sheet from a temperature-responsive dish, and the cell sheet was inverted so that the adhesive side was uppermost. Next, the VF was placed on one edge of the cell sheet, and the cell sheet was wrapped around the VF by rotating the flap (supplementary video 3). The cell sheet adhered to the VF during the wrapping procedure.

Evaluation of blood perfusion in a partially resected MF or VF
A laser Doppler perfusion imaging system (MoorLDI2-IR; Moor Instruments, Devon, UK) was used to evaluate blood perfusion in partially resected muscle tissue before and seven days after in vivo incubation and in the femoral A-V shunt 2-7 d after in vivo incubation. The perfusion signal was visualized as a 16-color pseudo-color image, with blue indicating a low perfusion state and red indicating a high perfusion state. The average perfusion value for each imaged region was calculated based on the values for the pixels in that region.

Bioreactor set-up
The bioreactor (figure 3(e)) was a one-pass system consisting of a microprocessor-controlled delivery pump (IPC-N; Ismatec, Wertheim, Germany), a custom-made tissue culture chamber (Tokai Hit, Shizuoka, Japan), a pressure control unit (PCU-2000; Millar, Houston, TX, USA) and a data acquisition system (PowerLab 4/35; ADInstruments, Dunedin, New Zealand). The bioreactor system was maintained in an incubator (HERAcell 150; Thermo Fisher Scientific) at 37 • C or 32 • C under 4.0%-5.0% CO 2 . Polyvinyl chloride tubing (inner diameter, 3.18 mm; outer diameter, 4.76 mm; Tygon LMT-55; Saint-Gobain, Courbevoie, France) was used to deliver culture medium from a reservoir and drain medium into a waste bottle. PharMed Ismaprene tubing (inner diameter, 1.3 mm; Ismatec) was used for the delivery pump. The bioreactor system was connected to the artery and vein of the tissue flap with custom-made polyurethane tubing (inner diameter, 0.30 mm; outer diameter, 0.64 mm; Tokai Hit), and the lengths of the connecting tubes were adjusted according to the flap. The outlet fluid volume was measured by an electronic balance (EK-3000i; A&D, Tokyo, Japan), and the inlet delivery pressure was recorded by the data acquisition system.

Tissue perfusion culture
Tissues were perfused with DMEM/F-12 containing HEPES as the buffering agent (pH 7.3-7.4 under 4.0%-5.0% CO 2 ; #11330032; Thermo Fisher Scientific) and supplemented with 10% FBS, 1% penicillin-streptomycin, 20 ng ml −1 human basic fibroblast growth factor (PeproTech, Rocky Hill, NJ, USA) and 4 ng ml −1 human vascular endothelial growth factor-165 (PeproTech). The delivery pump perfused medium through the inlet tube to the arterial side of the MF or VF at a flow rate of 50 ml min −1 .

Imaging devices
The images shown in figures 2(a), 7(b) (d), (g) and (h) and supplementary video 7 were captured by a camera (iPhone 13 Pro; Apple, Cupertino, CA, USA). Images and videos showing the procedures used in the animal experiments (supplementary video 2) were obtained using a stereomicroscope (M651; Leica Microsystems, Wetzlar, Germany) equipped with a camera (3CCD camera; Toshiba, Tokyo, Japan) and an image capturing system (Video Capture Box; I-O Data Device, Kanazawa, Japan). Images of cell sheets before and after transplantation onto an MF or VF (figures 5(b)-(h), 6(b)-(h) and 7(j)) were captured using a fluorescence stereomicroscopy system (MVX10 and cellSens Dimension; Olympus, Tokyo, Japan).

Measurement of enzyme levels in the culture medium draining from the endocrine MF (EMF) or endocrine VF (EVF)
EMF and EVF viability were evaluated by quantifying the levels of enzymes in the culture medium that had perfused through the tissue. The culture medium was collected from the chamber's outlet tube into a bottle every 24 h. Lactate dehydrogenase (LDH) and creatine kinase (CK) levels were measured in medium collected from the EMF and EVF. The enzyme activities were analyzed by the Japan Society of Clinical Chemistry transferable method (L-Type LDH J and L-Type CK; Wako, Osaka, Japan) using an automatic biochemistry analyzer (7180; Hitachi, Tokyo, Japan).

Measurement of the luminescence of insulin-GLase secreted by the EMF or EVF
The culture medium leaving the chamber was collected into a bottle every 24 h. The luminescence activity of insulin-GLase was measured one, two, three and four days after the transplantation of tricultured cell sheets onto an MF or VF. A 20 µl sample of collected culture medium was mixed with 100 µl of the coelenterazine (CTZ)-containing buffer provided in the luciferase assay kit (Cosmo Bio). The maximal light intensity was measured at 1000 ms using a Nivo multimode plate reader (PerkinElmer, Waltham, MA, USA). The mean value of three independent measurements was used for each experimental condition.

Real-time bioluminescence imaging of insulin-GLase secreted by the EMF or EVF
A custom-made electron-multiplying chargecoupled device (CCD) camera system and AquaCosmos 2.6 software (ImageEM, Hamamatsu Photonics, Shizuoka, Japan) were used for real-time bioluminescence imaging after four days of perfusion culture [26]. A bioreactor system with an EMF or EVF was placed in a light-shielded box with an integrated CCD camera system. A grayscale surface image was taken under weak illumination. Then, the EMF or EVF was perfused with CTZ-containing buffer at a flow rate of 50 ml min −1 , and real-time bioluminescence imaging was performed for 30 min from the start of perfusion with CTZ-containing buffer. For anatomical localization, a pseudo-color image representing light intensity (blue, least intense; red, most intense) was generated during live imaging and superimposed over the grayscale reference image.

Histology and immunohistochemistry
The harvested tissue was fixed in 4% paraformaldehyde and routinely processed into 7 µm-thick paraffin-embedded sections. The sections were stained with hematoxylin-eosin (HE) and immunostained for insulin and CD31. The procedures used for immunostaining were as follows (table 1). Deparaffinized sections were incubated first with anti-insulin and anti-CD31 primary antibody for 2 h at room temperature and then with appropriate secondary antibody for 40 min at room temperature. The sections were embedded in a mountant containing the blue DNA stain, Hoechst 33342 (NucBlue; Thermo Fisher Scientific). A fluorescence microscopy system (BZ-X810; Keyence, Osaka, Japan) was used to capture immunofluorescence images and HE-stained images of the entire region of each EMF and EVF for evaluation of transplanted cell engraftment (supplementary figure 4) and to capture magnified HEstained images of the EMF and EVF (Eclipse E800; Nikon). A confocal microscopy system (FV1200 or FV10-ASW; Olympus) was used to capture detailed immunofluorescence images of the tricultured cell sheets, EMFs and EVFs.

Evaluation of the engrafted cell sheet area and insulin-positive area on the EMF
Images of the entire EMF stained with HE were used to calculate the area of tricultured cell sheet engraftment. The region of cell sheet engraftment was defined as the region with a high cell density lying on the connective tissue above the MF. Immunofluorescence images of the entire EMF stained with anti-insulin antibody were used to calculate the engrafted insulin-positive area. Each image was binarized into insulin-positive and insulinnegative regions, and the insulin-positive area was measured. ImageJ 1.53 n was used to binarize the images and measure the engraftment area. The average value obtained from three different sections of each EMF was used for the analysis.

Two-photon laser-scanning microscopy of culture medium circulation in a vascular network
To distinguish between VF-derived vessels and tricultured cell sheet-derived vessels consisting of GFP-HUVECs, the EVF was perfused with culture medium containing 1.0 × 10 −5 mM DyLight 594-conjugated Lycopersicon esculentum (tomato) lectin (Vector Laboratories, Newark, CA, USA) at a flow rate of 50 µl min −1 on days 3 and 4 of the four-day perfusion period (table 1). The EVF was harvested after perfusion for four days and fixed with 4% paraformaldehyde for three days at 4 • C. Then, the EVF was transparentized using the following procedure. First, the EVF was optically cleared with a clearing reagent (CUBIC Trial Kit; Wako). Next, the EVF was incubated in 50% ScaleCUBIC-1 Solution at room temperature for one day followed by 100% ScaleCUBIC-1 Solution at 37 • C for two days. Subsequently, the EVF was washed with phosphatebuffered saline at 37 • C for one day and incubated in 50% ScaleCUBIC-2 Solution at room temperature for one day. The tissue was then placed in 100% ScaleCUBIC-2 Solution at room temperature for two days. A two-photon laser-scanning microscopy system equipped with FluoView 3.1 software (FVMPE-RS; Olympus) and a ×10 objective lens (XLPLN10XSVMP; numerical aperture, 0.6; working distance, 8.0 mm) was used to evaluate the relationship between the 3D architecture of VF-derived vessels and tricultured cell sheet-derived vessels. 3D reconstruction (428.729 µm × 428.729 µm × 120 µm volume) was performed using a series of X-Y images.

In vivo transplantation of an EVF or a tricultured cell sheet
Six male Fischer 344 athymic rats (F344/NJclrnu/rnu; 270-400 g; CLEA Japan, Tokyo, Japan) were used as the recipients. Three rats underwent transplantation of an EVF with anastomosis of the graft vessels to those of the host, and three rats underwent transplantation of a single tricultured cell sheet. Each rat was anesthetized by inhalation of 2% isoflurane. A catheter (Surflo Flash 24G; Terumo, Tokyo, Japan) was placed in the superficial femoral vein on one side for blood sampling, and the EVF or tricultured cell sheet was transplanted onto the contralateral side of the animal.
For EVF transplantation, the EVF was removed from the perfusion chamber after resection of the regions where the arterial and venous catheters had been inserted, because it has been reported that endothelial cells are damaged and lost at the cannulation site [27]. The EVF was washed with saline solution, and its artery and vein were anastomosed end-to-end with the recipient's superficial femoral artery and superficial femoral vein, respectively, using 10-0 nylon (figure 7(a)). Successful anastomosis (i.e. blood flow through the graft vessels) was confirmed using a laser Doppler perfusion imaging system (moorLDI2-IR; Moor Instruments, Axminster, UK; supplementary figure 7). For transplantation of a tricultured cell sheet, the cell sheet was detached from a temperature-responsive dish and transplanted onto the femoral fascia overlying the superficial femoral artery and vein using a piece of polyethylene sheet (figures 7(f)-(h)).

Measurement of the luminescence of insulin-GLase secreted by a transplanted EVF or transplanted tricultured cell sheet in vivo
The secretion of insulin-GLase from a transplanted EVF or a transplanted tricultured cell sheet was evaluated by obtaining 1 ml samples of blood after transplantation. Blood samples obtained at 5 min to 3 h after transplantation were drawn from the sampling catheter inserted into the superficial femoral vein on the contralateral side (figure 7(i)). Blood samples taken one day after transplantation were obtained by cardiac puncture ( figure 8(j)). Blood samples obtained at later times (up to two weeks after transplantation) were drawn from the caudal artery ( figure 8(k)). Each blood sample was immediately centrifuged at ×800 g for 15 min to obtain serum, which was stored at −80 • . At the time of the analysis, the cryopreserved serum samples were thawed on ice in the dark. The GLase luminescence activity in each serum sample (20 µl) was measured by the addition of 100 µl of the CTZ-containing buffer provided in the luciferase assay kit (Cosmo Bio). The maximal light intensity was measured over 5 s using a Nivo multimode plate reader (PerkinElmer). Each serum sample was assayed three times, and the mean value of the three measurements was used for the analysis.

Real-time bioluminescence imaging of insulin-GLase secretion by a transplanted EVF in vivo
An arterial catheter (Surflo Flash 24G; Terumo) was introduced into the superficial femoral artery and advanced into the abdominal aorta. A custommade electron-multiplying CCD camera system (ImageEM, Hamamatsu Photonics, Shizuoka, Japan) and AquaCosmos 2.6 software (Hamamatsu Photonics) were used for real-time bioluminescence imaging after transplantation of the EVF. Each EVF-transplanted rat was anesthetized by inhalation of 2% isoflurane and placed in a light-shielded box. The CTZ-containing buffer (1 ml) was administered via the arterial catheter. Real-time bioluminescence imaging was started about 8 min before infusion of the CTZ-containing buffer. For anatomical localization, a grayscale image representing light intensity (white, most intense) was generated in AquaCosmos 2.6 software (Hamamatsu Photonics) and superimposed over the grayscale reference image.

Statistical analysis
All values are shown as the mean ± standard deviation of the mean (s.d.m.). Two groups were compared using the paired Student's t-test, and multiple groups were compared using one-way analysis of variance and Tukey's honestly significant difference posthoc test. P < 0.05 was considered significant.

Mild hypothermia improved cell viability in vitro
Since the construction of thick tissue comprising islet β-cells is particularly challenging given its high oxygen demand and vascular requirement [28][29][30], we utilized islet β-cells in this study so that the findings might be applicable to other endocrine tissue types. Specifically, we used iGL cells (a rat pancreatic β-cell line stably expressing a fusion protein of insulin and Gaussia luciferase) as islet β-cells, hASCs as MSCs, and GFP-HUVECs as vECs. First, we compared the viability of confluent cells in vitro between different culture conditions. Under coculture conditions (islet β-cells and MSCs), the proportion of PI-positive dead cells was significantly higher at 37 • C (normothermia) than at 32 • C (mild hypothermia) on days 2 and 3 with no significant difference on day 1 (supplementary figures 1(a) and (b)). Under triculture conditions (islet β-cells, MSCs and vECs), the proportion of dead cells was significantly higher at 37 • C than at 32 • C on day 3 with no significant differences on days 1 and 2 (supplementary figures 1(c) and (d)). Additionally, there was a trend toward the total number of adherent cells on day 1 being higher at 37 • C than at 32 • C under triculture conditions (supplementary figure 1(e)). Based on the above results, cell sheets for subsequent experiments were fabricated under triculture conditions at 37 • C.

Effects of temperature on cell metabolism and vascular network construction under triculture conditions in vitro
Glucose consumption during three days of in vitro culture under triculture conditions tended to increase as the temperature was elevated from 27 • C to 32 • C and 37 • C (supplementary figure 2(a)), implying higher cellular metabolism at higher temperatures. The lactate:glucose ratio also tended to increase as the temperature was elevated to 37 • C (supplementary figure 2(b)), indicating a shift from aerobic to anaerobic metabolism as the temperature was increased to 37 • C. The average vessel length during three days of in vitro culture was comparable between temperatures of 32 • C and 37 • C, although the average vessel length tended to be lower at 27 • C (supplementary figures 2(c) and (d)). The above findings suggest that lowering the temperature from 37 • C to 32 • C decreases cellular metabolism while maintaining angiogenesis. This raises the possibility that perfusion culture at mild hypothermia might be more suitable than perfusion culture at normothermia for the construction of endocrine tissue with a high oxygen demand (see below).

Preparation of tricultured cell sheets
Cell sheets containing islet β-cells, MSCs and vECs were fabricated in temperature-responsive culture dishes [31]. Lowering the temperature caused the cell sheet to detach from the dish and shrink to about one-third of its original diameter ( figure 2(a)). The three cell types in the cell sheet adhered to each other in a continuous manner (figures 2(b) and (c)). Immediately after detachment from the dish, the vECs in the cell sheet were uniformly distributed and unnetworked, and some vECs had pseudopodia (figures 2(d) and (e)).

Preparation of MFs and VFs
We used rat femoral MFs, which have been described previously [23], and VFs, which were developed in this study as an alternative to the MF that might be more suitable for transplantation. The MF was incubated in vivo for seven days (figure 3(a), upper panel) to increase intramuscular blood flow (figures 3(b) and (c)). Subsequently, the MF was resected and placed in the chamber of a bioreactor system, and a triple-layered tricultured cell sheet construct was transplanted onto the MF (supplementary video 1). The VF, which comprised blood vessels and surrounding connective tissue, was developed by forming an A-V anastomosis between the rat femoral artery and vein and then incubating the resulting structure in vivo for 2-7 d ( figure 3(a), lower panel). The patency of the A-V anastomosis at the end of incubation was confirmed by laser speckle contrast imaging ( figure 3(d)) and an empty-and-refill test (supplementary video 2). A single tricultured cell sheet was wrapped around the harvested VF (supplementary video 3), and the resulting construct was placed in the bioreactor chamber. Both flap types underwent perfusion culture in the bioreactor system (figure 3(e)). (e) Schematic diagram of the bioreactor system. The bioreactor was a one-pass system consisting of a delivery pump, a custom-made tissue culture chamber, a pressure transmitter and a data acquisition system. The MF or VF construct was placed in the tissue culture chamber, and the artery and vein were connected to two polyurethane tubes for perfusion of culture medium. Medium was pumped from an inlet bottle into the arterial side of the MF or VF and was collected from the venous side into an outlet bottle.

Lowering the temperature during ex vivo perfusion culture improved cell sheet engraftment onto the MF
The MF with three tricultured cell sheets was cultured in the bioreactor at 32 • C or 37 • C. The perfusion ratio (volume of fluid leaving the venous outflow tube/volume of fluid infused via the arterial inflow tube × 100%) was higher at 32 • C than at 37 • C on days 1-4 ( figure 4(a)), indicating better perfusion of the MF vasculature at 32 • C. Inlet pressure was numerically but non-significantly higher at 32 • C than at 37 • C (supplementary figure 3(a)).
HE-stained sections showed that the cell sheets engrafted onto the MF along their entire length when cultured at 32 • C but were thin with deficient areas when cultured at 37 • C (figures 4(b) and (c) and supplementary figure 4(a)). Cell sheet cross-sectional area ( figure 4(d)) and insulin-positive cell area in immunostained sections (figures 4(e)-(g) and supplementary figure 4(b)) were significantly larger at 32 • C than at 37 • C. These results indicate that lowering the perfusion culture temperature from 37 • C to 32 • C improves the construction of endocrine tissue.

Islet β-cells in cell sheets on the MF self-organized into 3D tissue with a dense vascular network after perfusion culture at 32 • C
The vessels in the connective tissue of the MF were distributed just below the cell sheet ( figure 5(a), left). After four days of perfusion culture at 32 • C, vECs in the engrafted cell sheets had generated a tubular network that connected with vessels in the MF (figures 5(a)-(e)). Observations made using HEstained tissue sections (figures 5(f) and (g)), stereomicroscopy (figure 5(h)) and sections immunostained for insulin (figures 5(i) and (j)) revealed that black ink perfused through the MF vasculature had entered the tubular structures formed by vECs in the engrafted cell sheets. Furthermore, the islet β-cells had self-organized into an islet-like pattern surrounded by a dense vascular network (figures 5(i) and (j)).

Islet-like tissue on the MF secreted insulin Insulin-GLase secretion by iGL cells (islet β-cells)
in the engrafted cell sheets was visualized by infusing a luciferase substrate (CTZ) into the MF four days after perfusion culture [32]. Luminescence imaging demonstrated insulin-GLase in the central part of the MF ( figure 5(k) and supplementary video 4), although leakage of luminescent fluid from the MF was also observed. Luminescence was detected in the fluid collected from the MF vein on day 1 of perfusion culture, and a notable increase in luminescence was observed on day 3 ( figure 5(l)). Since these findings confirmed insulin secretion by the engrafted tricultured cell sheets on the MF, the construct was termed an EMF.

Construction of cylindrical tissue containing islet β-cells by perfusion culture of a VF with a tricultured cell sheet under mild hypothermia
The VF weighed substantially less than the MF (<0.1 g vs. 1.2 ± 0.14 g). The VF was wrapped with a tricultured cell sheet before perfusion culture (supplementary video 3) to minimize the distance between its central blood vessels and the cell sheet ( figure 6(a), left). The perfusion ratio was 100% (n = 4) throughout the four days of perfusion culture at 32 • C, and the inlet pressure was lower in the VF than in the MF at 32 • C or 37 • C (supplementary figure 3(b)). The VF was visible through the translucent cell sheet immediately after wrapping of the tricultured cell sheet ( figure 6(b)), whereas the engrafted cell sheet was observed as white cylindrical tissue after ex vivo perfusion culture for four days ( figure 6(c)). The vECs in the tricultured cell sheet (green) generated a circumferential vascular network along the entire aspect of the cylindrical tissue (figures 6(a), (d), (e)).
HE-stained sections confirmed engraftment of the cell sheet onto the connective tissue surrounding the main artery and vein of the VF (figure 6(f)). Black ink that had been infused into the main artery of the VF was observed to have entered the lumens of the tubular structures within the cell sheet (figures 6(g) and (h)). Immunofluorescence imaging confirmed that vECs in the cell sheet had formed a network surrounding self-organized structures containing islet βcells (figures 6(i) and (j)). 3D imaging (figure 6(m) and supplementary video 5) demonstrated that the femoral artery and its branches in the VF (red; ECs) had formed connections with GFP-HUVECs in the cell sheet (green).

Enzyme levels in the venous outflow from the VF or MF
Tissue damage during perfusion culture was evaluated by measuring the levels of LDH and CK in the outflow fluid. LDH and CK levels in the MF outflow fluid were higher on day 1 of perfusion culture than on days 2-4 (supplementary figures 5(a) and (b)). Furthermore, the LDH level was significantly higher in the MF outflow fluid than in the VF outflow fluid on day 1 but was comparable between the MF and VF on days 2-4 (supplementary figure 5(b)).

Islet-like tissue on the VF secreted insulin
Luminescence imaging following CTZ administration demonstrated strong luminescence in the VF construct and outlet tube four days after perfusion culture. Unlike the EMF, leakage of luminescent fluid was not observed (figure 6(k) and supplementary video 6). Luminescence of the outflow fluid increased significantly between days 1 and 2 and was then maintained on days 3 and 4 ( figure 6(l)). These results confirmed that the engrafted tricultured cell sheet on the VF secreted insulin; therefore, the construct was termed an EVF.

The EVF secreted insulin after transplantation in vivo
The EVF was transplanted into an athymic rat by anastomosis of its central artery and vein with the host's femoral artery and vein (figures 7(a)-(c)). The EVF appeared white before transplantation (figure 7(b)) but turned red (figure 7(c)) and pulsated (supplementary video 7) immediately after transplantation, indicating perfusion with the recipient rat's blood. A stereomicroscopic image showing the GFP-HUVECs of the engrafted cell sheet is shown in figure 7(d).  Insulin-GLase luminescence was detected throughout the EVF after the administration of CTZ (figure 7(e) and supplementary video 8), confirming that the EVF secreted insulin in vivo.
In additional experiments, a single tricultured cell sheet containing the same number of islet β-cells used for the EVF was transplanted onto the superficial femoral artery and vein of a rat (figures 7(f)-(h)). Notably, insulin-GLase luminescence in blood samples was numerically higher for the EVF than for the tricultured cell sheet at 5 min, 1 h, 2 h and 3 h (figure 7(i)). Fluorescence stereomicroscopy revealed red blood cells within the tubular structures formed by vECs near the surface of the EVF ( figure 7(j)). Red  blood cells in vessel-like structures were also observed throughout the islet-like tissue in HE-stained sections (figures 7(k) and (l)).
Additional experiments were performed to evaluate the structure and function of the EVF at two weeks after transplantation. HE-stained sections confirmed the presence of red blood cells in vessel-like structures at one day and two weeks after transplantation (figures 8(a) and (b)). Immunofluorescence imaging revealed that self-organized structures composed of insulin-positive cells and vascular networks were still present at two weeks after transplantation (figures 8(c) and (d)). Perfusion imaging demonstrated patency of the EVF at two weeks (figures 8(e)-(i)). Notably, the EVF continued to secrete insulin into the host's bloodstream for at least two weeks (figures 8(j) and (k)).

Discussion
The present study describes a new method of constructing transplantable and functional endocrine tissue. Our study has three major findings. First, construction of perfusable endocrine tissue using tricultured cell sheets and an MF succeeded when perfusion culture was performed at 32 • C but failed at 37 • C. Second, the VF had advantages over the MF in the construction of transplantable endocrine tissue. Third, insulin secretion was greater for a transplanted EVF than for a transplanted tricultured cell sheet. Our findings provide proof-of-concept that transplantable and functional endocrine tissue can be engineered by ex vivo perfusion culture of a VF wrapped with a tricultured cell sheet.
Lowering the temperature decreases oxygen consumption in cultured cells [22]. However, a lower temperature downregulates the expression of cell adhesion factors [33]. We observed better cell viability at 32 • C than at 37 • C after triculture for three days (supplementary figures 1(c) and (d)), although the number of adherent cells tended to be lower at 32 • C than at 37 • C (supplementary figure 1(e)). Thus, our results are consistent with the reported advantages [22] and disadvantages [33] of cell culture under mild hypothermia. After considering the effects of temperature on cell adherence and viability, the cell sheets used in subsequent experiments were constructed by triculture for one day at 37 • C.
Previous reports have described the construction of 3D perfusable vascularized tissue from cardiomyocyte cell sheets under normothermic conditions [23,24]. However, our results (figures 4(c)-(e) and (g)) demonstrated that vascularized endocrine tissue could not be generated under normothermic conditions. Islet cells have abundant vasculature to meet their high demand for oxygen and nutrients [15,34]. In a previous study [23], the formation of connections between vECs in a cardiac cell sheet and vessels in a vascular bed required three days, but the cell sheets maintained their viability during this period. By contrast, the present study found that cell viability decreased after in vitro triculture for three days at 37 • C, which may explain why cell sheets engrafted poorly onto the MF under normothermic conditions. Based on the present in vitro experiments and prior reports describing the fabrication of two-dimensional vascular networks under mild hypothermia [35], a temperature of 32 • C was selected for perfusion culture. Self-organized aggregates of islet β-cells surrounded by a 3D vascular network were constructed by ex vivo culture at 32 • C (figures 5(i), (j), 6(i) and (j)). Such tissue is likely well suited for use as a transplantable graft, since 3D aggregation of islet cells increases their maturation [36,37]. To our knowledge, no previous studies have described 3D vascularized tissue fabricated under mild hypothermia. We believe that it may be difficult to construct tissues with high vascular and metabolic demands under normothermic conditions. Our finding that perfusion culture under mild hypothermia can overcome this issue may represent a breakthrough in tissue engineering.
Flaps used to construct vascularized tissues include MFs, intestinal tissues and engineered skin flaps [23,[38][39][40]. The mean perfusion ratio was 47% for skeletal muscle cultured for four days [23], 24.9% for skeletal MFs cultured for three days [39], 42.5% for intestinal tissue [39], and 55.4% for engineered skin flaps [40]. Although the perfusion ratio of the MF was improved to about 65% under mild hypothermia, leakage remained high at around 35% ( figure 4(a)). Luminescence imaging confirmed the leakage of insulin secreted by the EMF (figure 5(k) and supplementary video 4), which would reduce the efficiency of the graft as an endocrine organ. Additionally, LDH and CK were detected in the venous outflow of the MF (supplementary figures 5(a) and (b)), suggesting skeletal muscle damage or necrosis. Since transplantation of necrotic tissue may cause reperfusion syndrome [41], we developed a VF that lacks muscle tissue. The VF had a perfusion ratio of 100% and a lower level of LDH in the venous outflow than the MF (supplementary figure 5(a)). Additionally, the VF was smaller and weighed less than one-tenth of the MF, making it better suited to subcutaneous implantation. This study is the first to generate an insulin-secreting EVF (figure 6(k) and supplementary video 6). Interestingly, the EVF secreted more insulin-GLase than the EMF, even though it was constructed using a single tricultured cell sheet rather than a triple-layered construct (supplementary figure 6); possible reasons for this include less leakage and a smaller distance between the engrafted cell sheet and functional blood vessels (figures 5(a) and 6(a)). Taken together, our findings indicate that a VF is better suited to the construction of endocrine tissues than an MF.
Endocrine tissues such as organoids, spheroids and cell sheets have been reported to secrete hormones after transplantation into animal models [5][6][7][8]. However, data are limited regarding shortterm changes in hormone secretion after transplantation. Transplanted organoids, spheroids and cell sheets require several days for the construction of a vascular network, which depends on the angiogenic potential of the host environment [5,42]. Therefore, the endocrine function and viability of transplanted organoids, spheroids and cell sheets would likely be unpredictable. By contrast, the EVF had a functional blood vessel network that communicated with the host's circulatory system immediately after transplantation, allowing cell viability to be maintained and hormone to be secreted directly into the host's bloodstream. In the present study, we demonstrated insulin secretion by transplanted EVF (figure 8(k)) for two weeks. Then, the results of tissue sections transplanted for one day or two weeks showed the thick islet β-cell tissue without necrosis lesion. In our previous report, the upper thickness limit for layered cardiomyocyte sheets transplanted into subcutaneous tissue was 80 µm [43]. In the present study, the thickness of islet β-cell tissue was well over 100 µm in both day 1 and two weeks (figures 8(a)-(d)), and the results suggested the islet β-cell tissue was supported by the ex-vivo constructed vascularity. However, there are the following limitations to the in vivo transplantation experiments: as an insulinoma-derived islet βcell line, iGL cells [32], was used, the increase in luminescence intensity after two weeks (figure 8(k)) and the growth of islet β-cells on the sections shown in figures 8(b) and (d) might be related to the potential growth of the islet β-cell line.
To demonstrate the versatility of our method, the present study utilized islet β-cells, which have a high demand for oxygen/nutrients. Our technique might be suitable for the fabrication of various types of endocrine organ, which could be achieved by substituting islet β-cells for other endocrine organ-derived cells. Further development of our method, using endocrine cells differentiated from induced pluripotent stem cells or embryonic stem cells, might lead to the construction of endocrine organs that are suitable for transplantation into patients with endocrine disorders.

Conclusions
Functional vascularized endocrine tissue can be constructed by ex vivo perfusion culture of tricultured cell sheets on a VF under mild hypothermia, and the resulting graft exhibits endocrine function after transplantation in vivo. Further development of this technique may allow the fabrication of perfusable vascularized endocrine tissues suitable for use as novel regenerative therapies for endocrine disorders.

Data availability statement
All data that support the findings of this study are included within the article (and any supplementary files). The data that support the findings of this study are available from the corresponding author upon reasonable request.