Engineered mesenchymal stem-cell-sheets patches prevents postoperative pancreatic leakage in a rat model

Post-operative pancreatic fistula (POPF) following pancreatic resection is a life-threatening surgical complication. Cell sheets were prepared and harvested using temperature-responsive culture dishes and transplanted as patches to seal POPF. Two different mesenchymal stem cell (MSC) sheets were compared in terms of the preventative ability for pancreatic leakage in a rat model. Both rat adipose-derived stem cell (rADSC) and bone marrow-derived stem cell (rBMSC) sheets were transplanted. Those rADSC and rBMSC sheets are created without enzymes and thus maintained their cell-cell junctions and adhesion proteins with intact fibronectin on the basal side, as well as characteristics of MSCs. The rats with post-pancreatectomy rADSC- or rBMSC-sheet patches had significantly decreased abdominal fluid leakage compared with the control group, demonstrated by MR image analysis and measurement of the volume of abdominal fluid. Amylase level was significantly lower in the rats with rADSC-sheet and rBMSC-sheet patches compared with the control groups. The rADSC sheet patches had increased adhesive and immune-cytokine profiles (ICAM-1, L-selectin, TIMP-1), and the rBMSC sheets had reduced immune reactions compared to the control. This is first project looking at the feasibility of tissue engineering therapy using MSC-sheets as tissue patches preventing leakage of abdominal fluid caused by POPF.

Establishing rat model of pancreatic leakage after pancreatic surgery. Figure 3a and b show the resection sites in the rat pancreas. Three different models were developed: model I (common pancreatic duct division), model II (gastric and splenic duct division with distal pancreatectomy and splenectomy), and model III (splenic duct division with distal pancreatectomy and splenectomy). Each model's survival rate and accumulated abdominal fluid volume at 10 days after distal pancreatectomy (DP) were analyzed and are shown in Fig. 3c and d, respectively. In model I rats, 5.77 ± 1.29 fluid was observed due to post-operative pancreatic fistula (POPF) and all rats died on day 1 (n = 10). Model III rats survived for the whole experimental period (10 days), and no fluid was observed during or after surgery. Model II rats survived for the whole experimental period, with continuous secretion of pancreatic juice, with 4.85 ± 0.31 ml fluid observed on day 1 (n = 10). Thus, model II-hereafter referred to as distal pancreatectomy (DP)-was deemed as the adequate model for assessing the efficacy of materials for prevention of pancreatic leakage. Figure 4a shows the process of attaching cell sheets to the pancreas resection site. The pancreatic juice was secreted only where the duct was opened, which is the optimal location for attaching the cell sheet. After DP, the stem cell sheets were attached using CellShifter TM on the resection surface. After five minutes, which is a sufficient time for the cell sheet to attach to the resection site, the shifter was removed, leaving behind only the cell sheet. Figure 3b shows the GFP fluorescence of the cell sheets. To confirm the continued attachment of rADSC and rBMSC sheets on the pancreatic resection site, GFP was imaged one hour and one day after cell sheet transplantation. Strong green fluorescence of the GFP cells was detected on the pancreas resection site at one hour and one day post-operation. No significant difference in GFP fluorescence was observed between the transplanted rADSC-and rBMSC-sheet patches.

Attachment of cell sheets on the pancreas resection surface.
Efficacy of cell sheets for pancreatic leakage prevention. The DP model without cell sheet attachment (control), the DP model with rADSC sheet patches, and the DP model with rBMSC sheet patches were compared experimentally. To confirm the efficacy of stem cell sheet patches in preventing pancreatic juice leakage, abdominal fluid volume was collected and measured on days 1, 3 and 7. As shown in Fig. 5a, abdominal fluid retrieved from the control group was 4.99 ± 0.63 ml on day 1, 4.58 ± 1.39 ml on day 3, and 2.93 ± 1.76 ml on day 7 (n = 9). Abdominal fluid retrieved from the rADSC-sheet patch group was 1.50 ± 1.43 ml on day 1, 0.71 ± 0.16 ml on day 3, 0.31 ± 0.31 ml on day 7 (n = 9), and from the rBMSC-sheet patch attachment group was 0.55 ± 0.80 ml on day 1, 0.50 ± 0.10 ml on day 3, and 0.10 ± 0.00 ml on day 7 (n = 9). There was a significant reduction in fluid collection from both rADSC-and rBMSC-cell-sheet patch groups compared to the control group (Control vs. rADSC, p < 0.001 on day 1 and day 3; p = 0.005 on day 7. Control vs rBMSC; p < 0.001 on day 1 and day 3; p = 0.002 on day 7). There was no significant difference between rADSC and rBMSC-sheet patch groups on day 1 and 7 (p = 0.101 at day 1 and p = 0.108 at day 7). However, the rBMSC-sheet patch group had a smaller ascites volume than the rADSC sheet-patch group on day 3 (p = 0.004).
MR imaging showed a similar tendency regarding the ascites volume level. MR scans were taken on day 1, 3, and 7 for the control group and the cell-sheet patch groups. The fluid collection area is shown in ImageT2 which appeared bright (Fig. 5b). For the longitudinal section, the fluid collection area was measured in the kidney image. For the transverse image, the liver was sectioned from the top at 3 mm intervals and a total of 30 images were obtained. The control POPF rat showed abdominal fluid on day 1, 3 and 7 days after surgery, as shown in the white area, while the rats with rADSC and rBMSC-sheet patches showed considerably less abdominal fluid after surgery.

Measurement of amylase level from abdominal fluid collection.
To confirm the presence of pancreatic juice in the abdominal fluid collection, the amylase concentration and level were measured (Fig. 5c). Amylase concentration decreased over time in all groups. Amylase concentration tended to increase in the control group compared to the rADSC-sheet patch and rBMSC-sheet patch groups. The amylase concentration of the control SCIEnTIfIC REPORTS | (2018) 8:360 | DOI:10.1038/s41598-017-18490-9 group was significantly higher than that of both stem cell-sheet patch groups on day 3 (p = 0.015 the rADSC-sheet patch group and p = 0.042 for the rBMSC-sheet patch group). According to the reference, mean drain amylase level is elevated more than 3 times than serum level compatible in the definition of a chemical pancreatic leak in ISGPF rather than just pancreatitis 16 . The normal value of amylase concentration in serum was 748 ± 71 IU/L in our experiment. Amylase concentration in the abdominal fluid in all groups were elevated more than 10 times than normal level at day 1. While amylase concentration in sheets groups decreased with time and reached about 5 times of normal level at day 3 and 7, amylase concentration in DP group was still more than 10 times higher than normal. No significant difference in the amylase concentration and level was observed between the rADSC and rBMSC sheet patch groups. Body weight and hematological changes after distal pancreatectomy. At day 1 after distal pancreatectomy, weight loss occurred in all experimental groups. In the case of the sheet groups, the body weight tended to gradually increase until day 3 and 7, but the reduced body weight in DP group was not recovered during the first week (Fig. 6a). Similarly, inflammation-related white blood cells (Fig. 6b) and neutrophils ( Fig. 6c and d) increased in all animals immediately after surgery and then gradually decreased. The BMSC sheet group showed the fastest decline, while the DP group maintained high white blood cell and neutrophil counts, albeit without statistical significance.
Immunomodulatory effects of rADSC-sheet and rBMSC-sheet patches on the DP. Immune responses of rADSC-and rBMSC-sheet patches were examined with an immune array kit in order to determine inflammation and pancreatitis. On day 7 post-DP, the area 3 mm toward the pancreatic head from the pancreatic resection surface was removed from normal, DP control, DP with rADSC-sheet patch group, and DP with rBMSC-sheet patch rats. In the DP group, adhesion and immune markers, including soluble ICAM (CD54), LIX, L-selectin, Thymus chemokine, TIMP-1, showed increased levels compared to normal rats. The rADSC-sheet patch group showed higher levels of soluble ICAM-1 (CD54), L-selectin, thymus chemokine, and TIMP-1 compared with the other groups. The BMSC-sheet patch group showed no expression or significantly lower levels of these proteins compared with the control group and the rADSC-sheet-patch group (Fig. 7).
Immunohistochemistry and H&E staining of stem-cell-sheet patch groups. To confirm that the stem cell-sheet patches remain attached to the resection surface for prolonged periods, the transplanted cells were detected with GFP fluorescence. Many tissues and blood cells had autofluorescence; therefore, in order to remove autofluorescence and specifically identify the transplanted cells, GFP was identified by immunohistochemical (IHC) staining. When the tissues were examined on 1 and 3 days post-operation, adherence of the stem-cell sheet patches to the duct was observed (Fig. 8). All groups, including DP, rADSC-sheet patch with DP, and rBMSC-sheet patch with DP, showed inflammatory cell accumulation on the resection site. The presence of an extracellular matrix barrier on day 1 was confirmed. On day 3, the stem-cell-sheet groups had decreased inflammatory responses, and regeneration started to take place around the transplantation sites. A dense layer of regenerated tissue and adhesion of other organs to the exposed surface of the transplanted rADSC sheet patches were observed.

Discussion
While there has been significant improvement in pancreas surgical techniques, pancreatic fistula still occurs at a high rate. POPF is a life-threatening complication that significantly prolongs hospital stay and increases healthcare costs 17 . Substantial effort has been made to prevent pancreatic fistula, such as modifying the surgical procedure and drainage regime, and selecting materials that can minimize the occurrence of POPF [3][4][5]18,19 ; however, a definitive approach for preventing POPF is still lacking 6,7 . Non-resorbable adhesives such as neoprene or prolamine are commonly used in clinical practice to prevent the leakage of pancreatic juice; however, they are limited in that permanent closure can cause pancreatic atrophy and complete loss of exocrine function 20 . Fibrin glue sealant or other topical hemostatic agents have been used for pancreatic anastomosis, and polyglycolic acid felt may reduce pancreatic anastomotic leak. However, recent large-scale meta-analyses that assessed the effects of fibrin glue on pancreatic surgery reported that there were no significant effects [21][22][23][24][25] . Also, one report showed that fibrin glue followed by wrapping of the PGA mesh around the remnant pancreatic stump is associated with decrease in clinical POPF; however, other studies have reported that this procedure only reduces the severity, and not the incidence of POPF 26 . These controversial results indicate that the current materials in use are not sufficient for preventing pancreatic fistula. These materials are difficult to be applied onto wet and rough surfaces, and simply serve as a passive barrier for inflammation and fluid intake rather than inducing tissue regeneration. The characteristics of an ideal material for preventing POPF include flexibility, biocompatibility, adhesiveness, and ability to induce wound healing at the resection site. The physical and biological properties of materials should be carefully considered prior to clinical application. In this regard, the application of MSCs for preventing POPF is a promising approach due to their high regeneration ability and immune modulation function. Recently, many clinical studies have reported the successful clinical application of MSCs for wound healing [27][28][29] and immune diseases 30 . Despite the advantages of cell therapy using MSCs, traditional single cell injection method is difficult to be applied to local regions because of massive cell loss and low survival rate of cells. Hence, a novel technique for efficiently transplanting cells to target regions is needed. Cell sheet technology is advantageous since cell sheets are harvested without the use of damaging enzymes, which enables them to maintain their shape and characteristics. The sheet's adhesion proteins facilitate attachment to the tissue surface, and the cell-cell interaction within the sheet makes them respond similarly to a tissue 11,12 .
For the present study, it was very important to develop a clinically-relevant pancreatic leakage animal model. In order to do so, we injured the splenic and gastric duct (model II) at a position close to the portal vein and performed DP and splenectomy. We then injured the splenic and gastric duct (model II) at a position close to the portal vein and performed DP and splenectomy. This approach is different from that of a recent study which involved cutting the splenic duct and leaving the pancreas itself intact 31 . Neither recent animal model can fully represent clinical POPF, but model II is the most accurate model available at this time.
POPF includes not only pancreatic leakage of postoperative pancreatic juice, but also various life-threatening clinical consequences such as intraperitoneal inflammation, abscess with or without sepsis, and delayed gastric emptying 32 . For clinical application, these clinical symptoms must be verified. In the present study, we focused on the reduction of pancreatic leakage and amylase level considering the small size of the experimental animals. In our experiment, there was a significant difference in the amount of fluid and amylase level between the control group and the sheet application groups. Both kinds of cell sheets demonstrated potential in preventing pancreatic leakage. Although not fully representative of the clinical situation, we have found that the weight reduction and inflammatory processes such as white blood cell and neutrophil counts were less in the sheet groups than in DP group. Thus, the sheet groups showed less feeding problems and inflammatory reactions.
In the present study, we isolated MSCs from two different tissues; adipose tissue and bone marrow. Adipose derives stem cells, which are often used in regenerative therapies, can repeatedly be acquired and grow rapidly 33 . They are multipotent cells and secrete a wide range of regenerative factors. Bone marrow-derived stem cells are another viable source of mesenchymal stem cells for cellular therapy. They are reported to have exceptional potential in differentiation and immune modulation abilities [34][35][36] .
To monitor the attachment and survival of the stem cell-sheet, we isolated stem cells from GFP transgenic rats. The transplanted rADSC-and rBMSC-sheet patches remained attached to the resection site. After three days, rADSC and rBMSC were not only attached, but were also proliferating on the surface. The rACSC-sheet patches had a high degree of adherence to surrounding organs and tissues in the abdomen 37 . We suspect that these results are relevant to the studies that suggest ADSC recruit adhesion and regeneration factors such as FGF-2, VEGF, and HGF 38 . In contrast, rBMSC-sheet patches were not adhesive to other organs. This difference in adhesion of the two types of patches to other organs will be investigated in future studies. A similar tendency was demonstrated in the immune cytokine array as well.
The results of cytokine and chemokine array indicated an increase in soluble ICAM-1, a type of adhesion protein, in the rADSC-sheet patch group only. Also, the rADSC patch group showed an increase in the levels of other molecules related to wound healing and immune cytokines such as L-selectin, CXCL7, and TIMP-1. rBMSC sheet patches had fewer immune reactions compared to control and rADSC sheet patch group. MSCs actively respond to stress, apoptosis, and inflammatory response in damaged tissues; MSCs are also known to promote angiogenesis, regeneration, immune cell activation or inhibition, and cell recruitment 36,[39][40][41] . MSC-based tissue regeneration involves regulation of extracellular matrix precipitation, collagen synthesis, fibroblast proliferation, platelet activation, fibrinolysis, and angiogenesis 42 . The immune process often involves T cell inhibition, macrophage activation, and potential neutrophil replenishment. The results of the present study show that MSC-sheet patches are a promising approach for preventing pancreatic leakage. Also, MSC induces regeneration and immune modulation at the resection site. In this study, ADSCs were more likely to induce wound healing, and BMSCs tended to decrease inflammatory response. The results of the present study show that MSC-sheet patches are a promising approach to prevent pancreatic leakage. In addition, MSC induce regeneration and immune modulation at the resection site. However, as reported in previous studies, the characteristics of MSCs vary greatly depending on culture conditions, tissue source and donor characteristics, and inflammatory status of the graft site. More extensive research is needed to reveal the mechanism of MSC sheet action in vitro and in vivo, and long-term follow-up of the transplanted cells is necessary. We also plan to test our experimental design in a porcine model to further determine its clinical potential.

Materials and Methods
Fabrication of mesenchymal stem cell sheets. rADSC and rBMSC stem cells were isolated from transgenic green-fluorescent-protein (GFP)-expressing seven-week SD-type rats, SD-TG (CAG-EGFP) (Japan SLC, Hamamatsu, Japan). Adipose tissue was harvested from the inguinal region and digested in 0.075% collagenase type I solution (Worthington, NJ, USA). rBMSC were isolated by collecting attached cells after flushing the marrows of the femur and tibia. rADSC and rBMSC were cultured in Dulbecco's Modification of Eagle's Medium (DMEM) mixed with 10% fetal bovine serum and 1% anti-antibiotics (GIBCO, MD, USA) in a 37 °C, 5% CO 2 chamber.
The rADSC and rBMSC were cultured on temperature-responsive dishes (3.5 mm UpCell: ™ Thermo Fisher Scientific, MA, USA) to form rADSC and rBMSC sheets, respectively. rADSC or rBMSC (1.1 × 10 6 ) at passage 3 were seeded and cultured for two days. The sheets were transferred to a lower temperature chamber at 20 °C an hour before being transplanted to the rat pancreas. Figure 1a shows the scheme of rADSC and rBMSC isolation and formation of cell sheets (rADSC sheet: left panel and rBMSC sheet: right panel). The UpCell dishes are hydrophobic at 37 °C, but become hydrophilic at 20 °C which allows the cell sheets to detach from the dishes, while maintaining their cell-cell junctions and adhesion molecules (Fig. 1b). The cell sheets were attached to CellShifter ™ membranes (Thermo Fisher Scientific, MA, USA) that were then transplanted onto the pancreas-resection site. After five minutes, which is a sufficient time for the cell sheet to attach to the resection site, the shifter was removed, leaving behind only the cell sheet (Fig. 1c). The experimental procedure was as follows: Rats were anesthetized and placed in the supine position. A midline incision was performed on the abdomen. By holding the stomach with an atraumatic forceps, the duodenum and spleen were exteriorized. Through omentectomy, the stomach and spleen were mobilized and the short gastric vessel was separated after ligating with a black silk 4-0 tie. The portal vein was exposed once the area between the colon and pancreas was mobilized. After determining the resection surface of the pancreas left of the portal vein, the resection surface was fixed with forceps, and the pancreas parenchyma was held with different forceps. The vessels were then tied with black silk 7-0 and resected. The ducts were left divided. The survival rate for up to 10 days for all three models and the volume of postoperative fluid were measured (n = 10).  The osteogenic differentiation cultures were incubated for 28 days; the cells were then fixed and stained with 1% Alizarin Red solution pH 4.1 (Sigma-Aldrich, St Louis, MO, USA). The green fluorescence of the stem cells from SD-TG rats (CAG-EGFP) was confirmed by fluorescence microscopy (EVOS TM FL Auto Imaging System, ThermoFisher Scientific, MA, USA). To stain fibronectin of the cell sheet, the cell sheet was embedded in paraffin and sliced into 4 um-thick cross-sections. The sections were stained with primary fibronectin antibody (Santa Cruse Biotechnology, Inc., CA, USA) and secondary anti-mouse Goat Anti-Mouse Alexa Fluor ® 555 (Thermo Fisher Scientific, MA, USA) at 1:100 dilution.

Characterization of mesenchymal stem cells sheet.
Optical imaging of the GFP rADSC and rBMSC sheet. To observe the attached GFP-expressing rADSC-and rBMSC-sheet patches on the pancreatic resection site, GFP expression of rADSC and rBMSC was detected with an IVIS Spectrum system (Caliper Inc., Alameda, CA). For organ (ex vivo) imaging, fresh organs were placed on plates and analyzed. EGFP was excited at 488 nm (filter range 445 to 490 nm) and detected at 510 nm. The region of the interested (ROI) level was measured with radiance (photons/s/cm 2 /sr) using an analysis program, Living Image 4.4 (Caliper Life Sciences, PerkinElmer Inc.).
Magnetic resonance (MR) imaging. 9 Amylase assay in abdominal fluid. Following euthanasia, rat abdominal fluid was collected via a 10 ml syringe after sacrifice and transferred into a tube. Amylase concentration was measured with an amylase activity kit (Abcam, Cambridge, UK). The assay procedures were followed from the manufacturer's instructions. Amylase level was calculated by multiplying the amylase concentration by the abdominal fluid volume.
Cytokine array of tissues. Rat cytokines and chemokines array kit (R&D SYSTEMS, MN, USA; #ARY008) was used to carry out immune cytokine analysis. The tissue section 3 mm towards the pancreas head from the pancreas resection surface was used. The tissue was soaked in PBS with protease inhibitor (SIGMA-Aldrich) and homogenized. The supernatant was used for analysis. 400 µg of protein was used per membrane based on protein quantification. The subsequent steps were directed by the manufacturer's instructions. The experimental groups were as follows: normal rats that did not undergo any operation; a group that only underwent distal pancreatectomy (DP); and two experimental groups in which rADSCs sheets and rBMSCs sheets were attached to the resection surface after DP, respectively. Interleukins, activators of B lymphocytes, activations of natural killers, and multiple biological effectors were identified with this kit.
Immunohistochemistry and H&E staining. The rats were sacrificed either on day 1 or 3, and each pancreas was removed. Formalin-fixed, paraffin-embedded sections (4 μm in thickness) were deparaffinized, dehydrated through a graded alcohol series. To perform hematoxylin and eosin (H&E) staining, samples were deparaffinized and dehydrated, followed by applying hematoxylin (Sigma-Aldrich, MO, USA) and eosin staining (Sigma-Aldrich, MO, USA). Immunohistochemistry was performed using primary antibodies for GFP (dilution 1:1000, Abcam, Cambridge, UK). Dehydrated samples were blocked with hydrogen peroxide, and dried for 10 minutes at RT then for 20 minutes in an incubator at 65 °C. An automated slide preparation system (Benchmark XT; Ventana Medical Systems Inc., Tucson, AZ, USA) with OptiView DAB Detection Kit (Ventana Medical Systems) was used for immunohistochemistry. Statistical analysis. Statistical significance was determined using GraphPad Prism 5 (GraphPad Software, San Diego, CA, USA). Statistical significance of the differences between groups was analyzed with the Student's t-test and a two-way analysis of variance. P < 0.05 was used as the cut-off for determining statistical significance. The data are presented as the mean ± standard deviation, with the number of samples indicated in the figure legends.