Optimizing beta cell function through mesenchymal stromal cell mediated mitochondria transfer.

Pretransplant islet culture is associated with the loss of islet cell mass and insulin secretory function. Insulin secretion from islet β-cells is primarily controlled by mitochondrial ATP generation in response to elevations in extracellular glucose. Coculture of islets with mesenchymal stromal cells (MSCs) improves islet insulin secretory function in vitro, which correlates with superior islet graft function in vivo. This study aimed to determine whether the improved islet function is associated with mitochondrial transfer from MSCs to cocultured islets. We have demonstrated mitochondrial transfer from human adipose MSCs to human islet β-cells in coculture. Fluorescence imaging showed that mitochondrial transfer occurs, at least partially, through tunneling nanotube (TNT)-like structures. The extent of mitochondrial transfer to clinically relevant human islets was greater than that to experimental mouse islets. Human islets are subjected to more extreme cellular stressors than mouse islets, which may induce "danger signals" for MSCs, initiating the donation of MSC-derived mitochondria to human islet β-cells. Our observations of increased MSC-mediated mitochondria transfer to hypoxia-exposed mouse islets are consistent with this and suggest that MSCs are most effective in supporting the secretory function of compromised β-cells. Ensuring optimal MSC-derived mitochondria transfer in preculture and/or cotransplantation strategies could be used to maximize the therapeutic efficacy of MSCs, thus enabling the more widespread application of clinical islet transplantation.

Mouse mesenchymal stromal cells (MSCs) derived from multiple tissue sources, including kidney, adipose, and bone marrow (BM), have direct effects on donor islet β-cells to improve their survival and insulin secretory function during the in vitro culture period prior to transplantation. [3][4][5][6][7] These in vitro findings correlate with persistent improvements in subsequent islet post-transplantation function in vivo. Thus, we demonstrated improved graft curative capacity in streptozotocin-induced diabetic mice transplanted with islets cocultured with MSCs, whether grafted at the experimental renal subcapsular site 4 or at the clinically preferred intraportal route. 3 We, and others, have also demonstrated that these findings translate to clinically relevant human islets and human MSCs. 5,8,9 MSCs can influence islet function through a variety of mechanisms. We have identified MSC-derived soluble secretory products that mimic some of the beneficial effects of MSCs in vitro, 10,11 and shown that preculturing islets with a defined cocktail of MSCsecreted ligands also improved islet graft function in vivo, albeit not to the same extent seen with MSC coculture. 10 MSC-derived extracellular matrix further contributes to the beneficial effects of MSCs on islet function, 8 and a number of studies have highlighted the importance of direct MSC-islet cell-cell contact for islet functional survival. 4,6,12 Studies in other tissues have also demonstrated the capacity of MSCs to act as mitochondria donors by transferring functional mitochondria directly to adjacent cocultured cells in inflammatory and ischemic disease settings, resulting in the rescue of aerobic respiration. [13][14][15][16][17] Insulin secretion from β-cells is primarily controlled by mitochondrial ATP generation in response to elevations in extracellular glucose, and islet oxygen consumption rate (OCR) is a key predictor of islet transplantation outcome. [18][19][20] We have therefore addressed the hypothesis that the MSC-dependent enhancement of insulin secretory function [3][4][5][6]8,10,12,21,22 is associated with mitochondrial transfer from MSCs to neighboring islet β-cells.

| Human and mouse islet isolation
Human islets were isolated from six nondiabetic donors at the King's College Hospital Islet Transplantation Unit, with appropriate ethical approval (LREC 01-082). Islets were maintained in CMRL medium supplemented with 2% human albumin, 4 mM glutamine, 2 mM HEPES (pH 7.2-7.4), and 10 mM nicotinamide at 37 C, 5% CO 2 prior to establishing human MSC: human islet cocultures, which were maintained in RPMI-1640 (supplemented with 10% [vol/vol] FCS, 2 mmol/L-glutamine, and 100 U/mL penicillin/0.1 mg/mL streptomycin). Human islets were handpicked into groups of 80 for culture alone or with MSCs for 1-3 days, as specified. The characteristics of each donor (age, gender, body mass index [BMI], islet purity, and viability) are specified in Supplementary Table S1. Mouse islets were isolated from male CD1 mice (Charles River, Margate, Kent) aged 8-12 weeks, by collagenase digestion (1 mg/mL; type XI; Sigma-Aldrich, Poole, UK) followed by density gradient separation (Histopaque-1077; Sigma-Aldrich). After washing with RPMI-1640 medium, islets were handpicked into groups of 80 for culture alone or with MSCs for 1-3 days as specified.

| Direct contact coculture of islets and MSCs
We used a direct-contact monolayer configuration to coculture islets with MSCs, as previously described. [3][4][5]8   and analyzed semiquantitatively by a blinded investigator, using FIJI software (https://fiji.sc/). Specifically, images were stacked, a threshold set to remove background fluorescence, and the percentage of remaining GFP signal within the selected islet area was calculated.

| Islet mitochondrial bioenergetics
The Seahorse extracellular flux analyzer XF24 (Agilent, Cheshire, UK) was used to measure islet OCR, as per manufacturer's instructions. Briefly, islets which had been cultured alone or with MSCs [3][4][5]8 were washed in XA basal media (Agilent) supplemented with 2 mM glucose and 1% FBS, before hand-picking into groups of 100 islets/500 μL XA basal media, per XF24 well. Islet screens were carefully added to enclose islets in the depression of the islet microplate. OCR was measured under basal (2 mM) and maximal (20 mM) glucose concentrations, as well as with drugs acting on the respiratory chain: oligomycin (ATP synthase inhibitor; 10 μM, Sigma) and FCCP (uncoupler; 1 μM, Sigma). 23 Data were normalized to initial OCR under basal conditions to account for variations in islet size and are reported as percentage of basal OCR. Glucose-stimulated respiration was calculated by dividing the first OCR measurement after injection of 20 mM glucose by the last basal OCR measurement and multiplying by 100.

| Islet insulin secretory function
Insulin secretion in vitro was assessed in static incubations of isolated islets. Islets were preincubated for 2 hours in RPMI containing 2 mM glucose. Groups of three islets were transferred into 1.5 mL Eppendorf tubes and incubated at 37 C in a bicarbonate-buffered physiological salt solution, containing 2 mM CaCl 2 and 0.5 mg/mL BSA and either 2-or 20-mM glucose. Samples of the incubation medium were taken after 1 hour and stored at −20 C until assayed for insulin content using in-house radioimmunoassay. 11,24

| Statistical analysis
Results are expressed as means ± SEM. ANOVA with Bonferroni's multiple comparison post hoc test was used for comparisons among multiple groups. A Student's t test for comparisons between two groups was used. A P value of .05 was considered significant. All statistical analysis was performed using GraphPad Prism version 6.

| Islet mitochondrial bioenergetics after MSC coculture
The generation of ATP and other metabolic coupling factors by mitochondrial metabolism is essential for nutrient-induced insulin secretion 25 and glucose-stimulated OCR is an important predictor of islet transplantation outcomes. [18][19][20] To determine whether MSCs induce alterations in islet mitochondrial bioenergetics, we measured mouse islet OCR using the seahorse XF24 islet respirometry platform. Islets that had been cocultured with MSCs were separated from the MSC monolayer, by gentle pipetting, prior to measurements of islet OCR and glucose-stimulated insulin secretion (GSIS). Our measurements of islet oxygen consumption demonstrate improved islet mitochondrial bioenergetics in MSC cocultured islets ( Figure 1A). After a 2-hour preincubation in low glucose (2 mM), control islets stimulated with 20 mM glucose demonstrated a clear increase in OCR to approximately 1.6-fold their basal level. In MSC cocultured islets, glucose-stimulated OCR was increased to twofold of the basal level ( Figure 1B). Upon addition of 10 μM oligomycin (an ATP synthase inhibitor), respiration was reduced in both control and MSC cocultured islets. Addition of 1 μM FCCP, which induces maximal respiration by uncoupling oxidative phosphorylation from the electron transport chain, caused a sharp increase in OCR which was more pronounced in MSC cocultured islets than in control islets. The concentrations of glucose used for basal and glucose-stimulated OCR measurements mirror those used for our standard static islet insulin secretion assays ( Figure 1C). As shown in Figure 1C, we consistently observe an MSC-dependent potentiation of GSIS in both mouse 3,4 and human islets, 5,8 and using MSCs derived from multiple tissues including adipose, BM, and kidney. [3][4][5]8 We now demonstrate that the MSC-mediated improvements in islet insulin secretory function are associated with improved islet mitochondrial bioenergetics. F I G U R E 1 Islet mitochondrial bioenergetics after MSC coculture. A, Oxygen consumption rate (OCR) of mouse islets precultured alone (black circles) or with mouse adipose MSCs (blue circles), measured using the seahorse XF24 analyzer. OCR was measured under basal (2 mM) and maximal (20 mM) glucose concentrations, as well as with drugs acting on the respiratory chain: oligomycin (ATP synthase inhibitor; 1 μM, Sigma) and FCCP (uncoupler; 10 μM, Sigma). OCR was measured using 100 islets per well (n = 8 wells per group; results are representative of three separate coculture experiments). B, Glucose-stimulated OCR is increased in MSC cocultured islets, 100 islets per well (n = 8 wells per group), *P < .05 vs islets precultured alone, Student's t test. C, Insulin release at 2 and 20 mmol/L glucose of 10 replicates of triplicate islets cultured for 3 days with mouse adipose MSCs (white bars) or without MSCs (black bars), *P < .05 vs absence of MSCs at the same glucose concentration (two-way ANOVA with Bonferroni post hoc test). MSC, mesenchymal stromal cells, OCR, oxygen consumption rate β-cells that were mitochondria-GFP positive after 1 day of coculture was variable, ranging from 28.6% to 100%, with a mean of 68.2% ± 4.3% of β-cells containing MSC-derived fluorescent mitochondria.

| Mitochondrial transfer is more extensive to human islets than to mouse islets
We next sought to determine whether mitochondrial transfer occurs between mouse MSCs and cocultured mouse islets which are more accessible for experimental investigation. After 2 days, we observed diffuse MSC-derived mitochondrial-GFP labeling ( Figure 5A, arrowhead) within mouse islets cultured in direct contact with mouse adipose MSCs (Figure 5A,B). However, the defined intra-β-cell localization and abundance of vesicular mitochondrial-GFP labeling which we consistently observed in human adipose MSC cocultured human islets (Figures 2-4) was not evident in mouse islets. Heterogeneity in mitochondrial transfer capacity between different MSC tissue sources have been reported, 30 Figure 5F) did not reveal extensive mitochondrial transfer from human adipose MSCs to mouse islets. Thus, the lack of extensive mitochondria transfer to mouse β-cells is unlikely to be due to the mitochondrial donation capacity of MSCs, but most likely reflects differences between the ability of mouse and human β-cells to act as mitochondrial recipients.

| Enhanced mitochondrial transfer to hypoxiaexposed mouse islets
Reports of mitochondria transfer in other tissues are often in models of ischemic or inflammatory disease, 13-17 consistent with recipient cells responding to cellular stressors by signaling to MSCs to initiate mitochondrial transfer. Isolated human islets are less robust than mouse islets and express a hypoxic molecular signature during the in vitro culture period prior to transplantation, 31 which is not seen in isolated mouse islets. To determine whether hypoxia influences the extent of mitochondrial transfer from MSCs to islets, we exposed mouse islets to hypoxia (1% oxygen) for 16 hours prior to coculture with mouse or human MSCs. Control mouse islets cultured under normoxic conditions (20% oxygen) demonstrated diffuse mitochondrial-GFP labeling ( Figure 6A,B), as shown previously in Figure 5. In contrast, the extent of mitochondrial transfer to hypoxic mouse islets was more extensive, as shown in 3D Z-projection micrographs ( Figure 6C). The insulin immunostaining intensity was notably weaker in hypoxic mouse islets, as expected, but 0.88 μm islet slices ( Figure 6D) confirmed the intracellular localization of MSC-derived mitochondria within islet cells, including insulinpositive β-cells ( Figure 6D, arrowhead). Semiquantitative analysis of mitochondria-GFP labeling was assessed by calculating the percentage of the area within an islet containing GFP fluorescence. After 72 hours of coculture, 0.32% ± 0.06% of the islet area of control islets (normoxia) contained mitochondria-GFP labeling, which was increased to 1.22% ± 0.21% in hypoxia pre-exposed islets ( Figure 6E). Mitochondria-GFP was evident within phalloidinpositive structures localized to β-cells ( Figure 6F, arrowhead, and

| DISCUSSION
In other tissues, mitochondrial transfer from MSCs is associated with the rescue of metabolic viability in recipient cells which have been subjected to ischemic and inflammatory stresses 13-17 but, to our knowledge, this is the first report of mitochondria transfer into insulin-secreting β-cells in mouse and human islets. β-cells are metabolically active and use mitochondrial ATP generation to couple elevations in circulating glucose to β-cell depolarization and the exocytotic release of insulin. 32 Islet mitochondria are particularly vulnerable to hypoxic stresses during the isolation, purification, and in vitro culture of islets, and impaired mitochondrial mass and/or function results in defective insulin secretion and reduced β-cell survival. 33 Accordingly, islet mitochondrial OCR is a key predictor of islet transplantation outcome. [18][19][20] Numerous studies have demonstrated that MSCs improve β-cell function in vitro and in vivo and our observations suggest that the transfer of functional mitochondria may be an important mechanism underlying these beneficial effects.
Thus, we consistently observed MSC-derived mitochondria-GFP localized within >80% of human β-cells located in the outer 25 μm of each islet and at the region of direct contact with cocultured human MSCs. Mitochondrial-GFP was also observed penetrating as far as 40 μm into the 3D islet structure. The average islet diameter is approximately 150 μm, 34 so our observations suggest that up to 30% of β-cells are recipient to MSC-derived mitochondria, sufficient to induce a functional phenotype in the intact islets. The transfer of mitochondria to human β-cells increased during the first two days of coculture with no further increase thereafter, consistent with our previous reports that MSC coculture induces a significant potentiation of insulin secretion after 2 and 3 days, but not prior to this. 5,11 Our measurements of mitochondria transfer from mouse MSCs to cocultured mouse islets showed less extensive transfer than that seen in our studies using human cells. This is unlikely to reflect an inability of mouse MSCs to transfer mitochondria. MSCs are heterogeneous in their expression of soluble bioactive molecules, and their functional characteristics, including mitochondrial transfer capacity, 30 can vary depending upon tissue source, species, and passage number. 35 However, our measurements consistently demonstrated less extensive mitochondrial-GFP labeling in mouse islets cocultured with mouse BM-MSCs, mouse adipose MSCs, and human adipose MSCs when compared to human islets, suggesting that the species variation was not due to the superior functional capacity of human adipose MSCs to donate their mitochondria to neighboring islet cells. The most likely explanation for the more extensive mitochondrial transfer to human islets is that the level of cellular stress in isolated human islets is much greater than that of isolated mouse islets, partly because of the differences in the isolation processes and partly because of donor differences. Human islets experience a prolonged cold ischaemia time during the isolation process and they express characteristic hypoxia inducible  Figures 2-4). E,F, Magnification ×60, scale bar = 10 μm. These experiments were replicated with three separate mouse islet isolations factor-1α regulated genes, with a gene expression profile following culture under normoxic conditions (20% O 2 ) which resembles that of mouse islets exposed to hypoxia (1% O 2 ). 31 Thus, human islets are subjected to more extreme cellular stressors than mouse islets which may induce "danger signals" 36 for MSCs, initiating the donation of MSCderived mitochondria to human β-cells. Our observations of increased MSC-mediated mitochondria transfer to hypoxia-exposed mouse islets are consistent with this and suggest that MSCs are most effective in supporting the secretory function of compromised β-cells. Transfer of mitochondria from MSCs into β-cells may explain the observed effects of MSC coculture to increase islet OCR in response to elevated glucose.
The increased flux of oxidative phosphorylation may, in turn, explain the effects of MSCs to enhance GSIS which we have reported previously, [3][4][5]8,11 and confirmed in the current study.
Differences in islet donors may also influence experimental outcomes. Human islets are isolated from pancreases harvested from heart-beating, brain-dead donors and factors such as age, BMI, and duration of brain death have been shown to impact upon human islet isolation success and on islet function in vitro. 37,38 In contrast, mouse islets are isolated rapidly from healthy, lean, genetically homogenous, young animals. MSCs are reported to transfer mitochondria to cells deficient in mtDNA but not to otherwise healthy cells, 16,39 and previous studies have shown an age-related decline in mtDNA copy number in isolated human islets. 40 Most of our human islet donors were in middle age, in contrast to the relatively young 8-to 12-week-old mouse donors used in most islet studies. Differences in islet architecture 41,42 and differential expression of cell adhesion molecules may also contribute to differences in mitochondria transfer. For example, transfer of MSC-derived mitochondria to lipopolysaccharide-injured alveolar epithelial cells was dependent upon connexin 34 (Cx43)-mediated alveolar attachment 15 and human and mouse islets differ in their expression of adhesion molecules and gap junctional complex (GJC) components, including connexins. 43 Our imaging studies have demonstrated the presence of mitochondria-GFP microvesicles predominantly where MSCs are in direct contact with human islets, suggesting that islet-MSC contact is required for the transfer of mitochondria and subsequent improvements in islet insulin secretory function. In accordance, our previous studies have demonstrated that indirect transwell MSC-islet F I G U R E 6 Enhanced mitochondrial transfer to hypoxia-exposed mouse islets. Confocal micrographs showing representative 3-day cocultured mouse islets, which were exposed to hypoxia (1% O 2 ) for 16 hours before coculturing with mesenchymal stromal cells (MSCs). Green indicates MSC-derived BacMam mitochondria-GFP labeling, and red indicates insulin immunostaining of β-cells and blue represents DAPI (A-D and F-G). Cyan represents phalloidin (F,G) staining of F-actin, indicative of TNTs. A,C, Composite Z-projection micrographs of 25× 0.88 μm optical sections of a series of consecutive mouse islet slices starting at the MSC: islet interface and upward, in control mouse islets cultured under normoxia (A) and hypoxia pre-exposed islets (C), subsequently cocultured with mouse bone marrow (BM) MSCs. Magnification ×60, scale bars = 10 μm. B,D, Individual 0.88 μm insulin immunostained mouse islet slices, within the outer layer (first 10 μm) of cells of islets cultured under normoxia (B) and hypoxia pre-exposed islets (D). Magnification ×60, scale bars = 25 μm. E, Semiquantitative analysis of mitochondrial transfer determining the percentage of each islet area occupied by BacMam mitochondrial-GFP labeling in 24-33 separate islets, *P < .001 vs normoxia cultured islets, Student's t test. Data presented are representative of three independent experiments. F,G, Human adipose MSC cocultured mouse islets pre-exposed to hypoxia. Phalloidin staining of F-actin in a composite Z-projection of 25× 0.88 μm optical sections of a series of consecutive insulin immunostained mouse islet slices (F) and an individual 0.88 μm slice (G) demonstrating BacMam mitochondrial GFP-labeling within F-actin based TNT-like ultrastructures. Magnification ×60, scale bars = 5 μm coculture does not improve GSIS, in contrast to the robust improvements we have consistently observed using direct contact coculture of islets with MSCs. [3][4][5] Several mechanisms of mitochondria transfer from MSCs to injured tissues have been proposed, including microvesicles, TNTs, 14,26,44  TNTs, 29 which have lengths of several cell diameters. 45 We also detected TNT-like structures, composed of F-actin, spanning between the MSCs and neighboring β-cells. MSCs shed a diverse population of EVs, including mitochondria-containing microvesicles (0.1-1 μm in diameter). 28 MSC-derived EVs recapitulate the capacity of MSCs to transfer their mitochondria to recipient cells in models of lung injury 28 and the transfer of partially depolarized mitochondria from MSCs experiencing oxidative stress to cocultured macrophages enhances MSC survival whilst restoring mitochondrial bioenergetics in the recipient cells. 46 Together, the vesicular morphology of the mitochondrial-GFP labeling within human β-cells and the depth of penetration of the mitochondria transfer into the 3D islet structure are consistent with a mechanism involving both TNTs and microvesicles, as described previously. 15

| CONCLUSION
In conclusion, MSCs transfer mitochondria to islet β-cells during in vitro coculture, and this correlates with increased β-cell mitochondrial oxygen consumption and enhanced glucose-induced insulin secretion.
Mitochondrial transfer from human MSCs to human islets is more extensive than from mouse MSCs to mouse islets, most likely because isolated human islets are exposed to more extreme cellular stressors which initiate mitochondria transfer to human islets. Ensuring optimal β-cell mitochondrial mass and bioenergetics through MSC-mediated mitochondria transfer therefore offers a novel strategy for improving the outcomes of clinical islet transplantation as a therapy for T1D.