Insulin-Deficient Diabetic Condition Upregulates the Insulin-Secreting Capacity of Human Induced Pluripotent Stem Cell–Derived Pancreatic Endocrine Progenitor Cells After Implantation in Mice

Type 1 diabetes mellitus (T1DM), caused by insufficient insulin (INS) production due to the autoimmune destruction of pancreatic β-cells, includes a small subset of patients with uncontrollable glycemia despite current INS therapy, whose condition is described as brittle T1DM (1). In these patients, quality of life is dramatically threatened by impaired awareness of hypoglycemia and severe hypoglycemic events. Pancreatic islet transplantation is reported to be effective for these patients (2), though the donor shortage is a major obstacle for this promising therapeutic option. In recent years, alternative cell sources, such as islets derived from human stem cells and xenogeneic animals, have been intensively investigated to overcome the donor shortage (3,4).

Many studies have shown that the generation of pancreatic endocrine cells from human embryonic stem cells (hESCs)/human induced pluripotent stem cells (hiPSCs) has the potential to cure T1DM in small animal models (5–7). Most of the studies used in vitro stepwise differentiation protocols based on the knowledge from the developmental processes of pancreas, islets, and β-cells (8). Despite rapid progress, generating fully functional endocrine cell clusters that replicate all aspects of adult islet cells has remained elusive (9). This implies the significance of the in vivo differentiation/maturation process after implantation.

Implanted cells are generally exposed to multiple environmental factors in the host, like humoral factors and cell-to-cell interactions with other cells. These factors affect engraftability, the cell cycle, and function. In addition, these factors affect the differentiation/maturation process for immature cells. In fact, host sex (10) and hypothyroidism (11) have been shown to affect the differentiation/maturation process of implanted immature pancreatic cells generated in vitro. Although clinical trials have been conducted that used hESC-derived immature pancreatic cells for T1DM (12,13), the effect of diabetes in the recipients on the fate of the implanted cells has not been elucidated.

Streptozotocin (STZ)-induced diabetes has been reported to enhance the increase in plasma levels of human C-peptide (hC-peptide), a marker of INS secretion, after the implantation of hESC-derived pancreatic endoderm cells (PECs) in immunodeficient mice (14). After implantation, a part of the engrafted PECs is considered to differentiate into endocrine progenitor cells (EPCs), immature hormone-positive cells, and then to mature into adult-type endocrine cells, including β-cells or islets that secrete INS/C-peptide into the circulation. The study has attributed the increase in plasma hC-peptide levels to the acceleration of the differentiation of implanted PECs (14). However, it encompasses diverse aspects of cell fate, such as engraftability, the differentiation/maturation process, and proliferation ability. Additionally, STZ-induced diabetes is composed of hypoinsulinemia and hyperglycemia followed by metabolic changes in the whole body. The involvement of multiple factors necessitates the investigation of the detailed mechanism behind the phenomenon.

EPCs are the early-stage endocrine cell type that develop toward hormone-positive cells from PECs. Therefore, compared with PECs, EPCs can be more useful to understand the fate of immature endocrine cells under a host environment because the influence of the differentiation process into EPCs can be disregarded. Additionally, since EPCs theoretically contain less progenitors with the potential to differentiate into nonendocrine cells than PECs, EPCs may be preferable for transplantation. The purpose of this study is to investigate the effect of INS-deficient diabetes on the fate of immature endocrine cell grafts and the underlying mechanisms using EPCs generated in vitro from hiPSCs. We demonstrated, for the first time to our knowledge, that the INS-deficient diabetic condition upregulates the INS-secreting capacity of EPC grafts in immunodeficient mice. We further showed that the diabetic condition positively affected EPC grafts, resulting in an increase in the number of endocrine cells, including INS-producing cells, without changing the graft mass.

Two hiPSC lines for nonclinical use (Ff-I14s14 and Ff-WJ18) were provided by the Center for iPS Cell Research and Application of Kyoto University (Kyoto, Japan) and maintained and passaged in StemFit AK03N medium (Ajinomoto, Tokyo, Japan). The cells were directed into key stages of pancreatic development, including definitive endoderm (stage 1), primitive gut tube (stage 2), posterior foregut (stage 3), pancreatic endoderm (stage 4), and endocrine progenitor (stage 5) in vitro taking a total of 16 or 17 days. The differentiation protocols were based on those previously reported with some modifications: stages 1–4 (5,15–17); stages 3–5 (18); and stage 5 (19,20). The detail is provided as follows, and the brief schematic diagram is shown in Fig. 1A. The use of hiPSCs was approved by the ethical review committee of Takeda Pharmaceutical Company Limited (Fujisawa, Kanagawa, Japan) and Kyoto University (Kyoto, Japan).

Undifferentiated hiPSCs were resuspended with stage 1 medium containing RPMI medium (12633012; Thermo Fisher Scientific, Waltham, MA) supplemented with 2% (v/v) B27 (17504001; Thermo Fisher Scientific), 1% (v/v) penicillin/streptomycin (P/S; 168–23191; FUJIFILM Wako, Osaka, Japan), 5–100 ng/mL activin A (R&D Systems, Minneapolis, MN), 3 μmol/L CHIR99021 (Axon Medchem, Groningen, the Netherlands), 10 μmol/L Y-27632 (FUJIFILM Wako), and 1% (v/v) DMSO (045–24511; FUJIFILM Wako), seeded on iMatrix-511–coated plates (381–07363; Wako) at a density of 1–2 × 10⁵ cells/cm², and cultured for 1 day. For the next 2 days, the cells were cultured in RPMI medium with 2% B27, 1% P/S, 5–100 ng/mL activin A, and 1% DMSO. When using activin A at 100 ng/mL, 1% DMSO was not used to supplement in the medium.

Cells were exposed to Improved MEM Zinc Option (iMEM) medium (Thermo Fisher Scientific) supplemented with 1% B27, 1% P/S (iMEM-B27), and 50 ng/mL keratinocyte growth factor (KGF; R&D Systems) for 4 days.

Culturing was continued for 2 or 3 days in iMEM-B27 with 50 ng/mL KGF, 0.5 μmol/L 3-keto-N-aminoethyl-N′-aminocaproyldihydrocinnamoyl cyclopamine (KAAD-CYC; Toronto Research Chemicals, Toronto, Canada), 10 nmol/L 4-[(E)-2-(5,6,7,8-tetrahydro- 5,5,8,8-tetramethyl-2-naphthalenyl)-1-propenyl] benzoic acid (TTNPB; Santa Cruz Biotechnology, Dallas, TX), 100 ng/mL NOGGIN (Pepro-tech, Rocky Hill, NJ), and 250 μmol/L L-ascorbic acid (A4544; Sigma-Aldrich, St. Louis, MO).

Cells were dissociated into single cells by gentle pipetting after treatment with 0.25% trypsin–EDTA and seeded on iMatrix-511–coated plates at a density of 3–4 × 10⁵ cells/cm². Then, the cells were cultured for 4 days in iMEM-B27 with 100 ng/mL KGF, 50 ng/mL epidermal growth factor (R&D Systems), 10 mmol/L nicotinamide (VERITAS, Tokyo, Japan), 50 μmol/L Y-27632, and 250 μmol/L L-ascorbic acid.

For aggregation cultures, cells were dissociated into single cells, as described above, and seeded on a low-binding 96-well plate at a density of 30 × 10³ cells/well. Then, the cells were cultured for 3 or 4 days in iMEM-B27 with 250 nmol/L SANT-1 (S4572; Sigma-Aldrich), 50 nmol/L retinoic acid (Sigma-Aldrich), 10 μmol/L ALK5 inhibitor II (SC-221234A; Santa Cruz Biotechnology), 100 nmol/L LDN193189 hydrochloride (HY-12071A; Medchemexpress, Monmouth Junction, NJ), 1 μmol/L L-3,3ʹ,5-triiodothyronine (64245; Merck Millipore, Burlington, MA), 1 μmol/L XAV939 (X3004; Sigma-Aldrich), 50 ng/mL basic fibroblast growth factor (100–18B; Pepro Tech, London, United Kingdom), 1 μmol/L γ-secretase inhibitor XXI (Compound E, 565790; Merck Millipore), and 10 μmol/L Y-27632. The resulting aggregates were used for implantation studies.

Additional information on EPCs used in this study is provided in Supplementary Table 1. The differentiation quality at each stage, based on the developmental processes of pancreatic endocrine cells, was assessed by flow cytometry and immunostaining as described previously (5). The antibodies used are detailed in Supplementary Table 2. Differentiation of EPCs into INS-producing cells in vitro was performed based on reported protocols (18,20) with details provided in the Supplementary Data.

Male 7- to 15-week-old NOD.CB17-Prkdc^scid/J mice (NOD/SCID; Charles River Laboratories Japan, Kanagawa, Japan) were rendered as diabetic models by daily intraperitoneal injection with 50 mg/kg of STZ (Sigma-Aldrich) for 5 consecutive days. Age-matched male NOD/SCID mice were similarly injected with 0.05 mol/L citrate buffer (pH 4.5) and used as nondiabetic models. Diabetic Akita-NOD/SCID mice were generated by crossing of male AKITA/Slc mice (Akita; Japan SLC, Shizuoka, Japan) and female NOD/SCID mice. Mice showing hyperglycemia were selected from the offspring mice and backcrossed for >10 generations with NOD/SCID mice. Both SCID and Ins2^Akita mutations were confirmed by PCR-based genotyping. Age-matched NOD/SCID mice without Ins2^Akita mutation were used as wild type (WT) nondiabetic models. Animals had ad libitum access to a normal diet (CE-2; CLEA Japan, Inc., Tokyo, Japan) and tap water unless otherwise stated, and they were housed individually under controlled temperature (20–26°C), humidity (40–70%), and a 12-h light/12-h dark cycle. All animal experiments were conducted based on protocols reviewed by the Institutional Animal Care and Use Committee of Takeda Pharmaceutical Co., Ltd. For all implantation studies, mice were divided into groups based on data of nonfasting blood glucose level and body weight.

Mice were anesthetized with inhalable isoflurane, and then 192 hiPSC-derived EPC cell aggregates (30 × 10³ cells/aggregate) were implanted under the left kidney capsule. EPCs from the same batches were divided between implant recipients in individual studies. Sham-operated mice were also prepared in some experiments. All mice received a single subcutaneous injection of meloxicam (1 mg/kg, Metacam; Boehringer-Ingelheim, Ingelheim am Rhein, Germany) at the time of implantation. At the indicated weeks after implantation, blood samples were collected from the tail vein without anesthesia and placed in chilled tubes containing heparin and aprotinin. They were then centrifuged at 4°C to isolate the plasma. Plasma levels of hC-peptide were measured to assess the INS-secreting capacity and function of the grafts. In an oral glucose tolerance test, glucose was administered orally (2 g/kg) in overnight-fasted mice. After the glucose loading, blood samples were collected at the indicated times.

STZ-induced diabetic NOD/SCID mice were subcutaneously administered human INS (50 nmol/kg/day) (Peptide Institute, Osaka, Japan) by infusion from an implanted osmotic pump (Model 1002, Alzet; DURECT Corporation, Cupertino, CA) for 14 days, and only mice showing blood glucose lowering by the treatment received EPC implants. At the time of EPC implantation and at 4 and 8 weeks after the implantation, the osmotic pump was replaced with a new pump (Model 1004, Alzet; DURECT Corporation) to sustain the INS infusion up to 11 weeks after the implantation. To prevent hypoglycemic condition, the INS dosage was individually adjusted to 40–60 nmol/kg/day based on nonfasting blood glucose levels. Age-matched STZ-induced diabetic NOD/SCID mice without exogenous INS infusion and nondiabetic NOD/SCID mice also received EPC implants.

Nondiabetic 7-week-old NOD/SCID mice were implanted with EPCs and were thereafter fed with a high-fat diet (HFD; 60% calories from fat; D12492; Research Diets, Inc., New Brunswick, NJ) up to 12 weeks after the implantation. Age-matched nondiabetic NOD/SCID mice fed CE-2 (4.5% calories from fat) were used as the control diet (CD) group.

STZ-induced diabetic NOD/SCID mice were orally administered vehicle (0.5% methylcellulose, w/v) or a selective sodium-glucose cotransporter-2 (SGLT2) inhibitor dapagliflozin (10 mg/kg, suspended in the vehicle solution) once daily for 4 weeks. Each of the treatment groups was divided into two groups and received an EPC implant or the sham operation. After that, the treatment of vehicle or dapagliflozin was continued for 12 weeks. Age-matched nondiabetic NOD/SCID mice with a similar vehicle treatment were also prepared and received EPC implants. Dapagliflozin was purchased from Nard Institute, Ltd., Amagasaki City, Hyogo, Japan.

The cell aggregates used for implantation and engrafted kidneys harvested at 6 weeks after implant were fixed with 4% paraformaldehyde for 1–2 days at 4°C and processed for immunostaining as described previously (5). The antibodies used are provided in Supplementary Table 2. Fluorescence imaging was performed using a BZ-X710 fluorescence microscope (Keyence, Osaka, Japan). Areas of the whole graft and chromogranin A (CHGA), NK6 homeobox 1 (NKX6.1), INS, and glucagon (GCG)-immunoreactive cells were analyzed using cellSens Dimension software (Olympus, Tokyo, Japan). Quantification of the immunoreactive area was performed by the same method reported by Bruin et al. (21). Immunofluorescence staining was performed on three to four sections per graft sample, separated at least by 100 µm, and the average of all sections was used to represent each graft. Total graft area per section was set based on the border of renal parenchyma. Counting of INS⁺Ki67⁺ cells was manually performed under blind conditions using three to four sections per graft sample, separated at least by 100 µm, and the average of all sections was used to represent each graft.

Blood glucose was measured by ACCU-CHEK Aviva (Roche Diagnostics K.K, Tokyo, Japan). Plasma levels of hC-peptide and human INS were measured by respective ELISA kits (Mercodia, Uppsala, Sweden). Plasma mouse INS was also measured with Morinaga mouse INS ELISA kit (Morinaga Institute of Biological Science, Inc., Kanagawa, Japan).

Whole grafts carefully peeled off the recipient’s kidneys were measured for the weights and contents of human INS and GCG as follows. The whole graft samples were homogenized in an ethanol/HCl solution and then incubated overnight at 4°C. After removing the debris by centrifugation, concentrations of human INS and GCG were measured with respective ELISA kits (Mercodia). Similar preparations were performed on cell aggregates to measure hormone contents. INS content in the pancreas was measured similarly using a piece of the tissue, and with mouse INS ELISA kit (Morinaga Institute of Biological Science, Inc.).

Data were expressed as mean and SDs. Statistical differences between two animal groups were analyzed with Student or Aspin-Welch t tests, and adjustment for multiple testing was performed using the Bonferroni method. ANOVA was used to evaluate the differences in the levels of blood and plasma parameters among animal groups. Alternatively, statistical significance was first analyzed using Bartlett test for homogeneity of variances, followed by Bonferroni-adjusted Dunnett test for multiple comparisons.

The data sets generated during and/or analyzed during the current study are available from the corresponding authors upon reasonable request.

We differentiated hiPSC line Ff-I14s04 toward EPCs by protocols shown in Fig. 1A. After efficient differentiation of stage 1 (SRY-box [SOX]17⁺forkhead box [FOX]A2⁺) and stage 4 (PDX-1⁺NKX6.1⁺) (Supplementary Fig. 2A and B) as previously reported (5), cells were dissociated and aggregated for endocrine cell commitment at stage 5. We sequentially obtained definitive endoderm (SOX17⁺), pancreatic endoderm (PDX-1⁺NKX6.1⁺), and endocrine/endocrine precursor (CHGA⁺) (22) cells at the average induction rate of 98.6 ± 1.1%, 90.1 ± 0.6%, and 59.4 ± 4.2%, respectively. The majority (52.9 ± 4.0%) of the CHGA⁺ cells expressed NKX6.1, indicating the potential to be β-cells. On the other hand, the cells comprised very few INS- and GCG-expressing cells (Fig. 1B and C). Representative flow cytometry plots analyzing each stage marker regarding one sequential inducting experiment are shown in Fig. 1D and Supplementary Fig. 1. We did not detect an immunofluorescence signal of an endocrine progenitor marker neurogenin-3 (NGN3) in the large population of EPCs we generated (Supplementary Fig. 2C) because its expression pattern was transient during stage 5 (Supplementary Fig. 3) in agreement with a previous report (23). We thereby describe the cell population in which the timing of expression peak of NGN3 passes as EPCs. The EPC aggregates used for implantation studies contained a part of fetal pancreatic epithelial marker cytokeratin 19 (CK19)-expressing cells (24) and few exocrine marker trypsinogen- and multipotential progenitor marker SOX9-expressing cells (25) (Supplementary Fig. 2C).

To investigate whether an INS-deficient diabetic condition in the host regulates the fate of implanted EPCs, we implanted them under the kidney capsule of STZ-induced diabetic NOD/SCID mice and evaluated the graft function. In STZ treated–diabetic mice, nonfasting blood glucose maintained high levels (average, 504–576 mg/dL) compared with the mice without STZ-treatment. However, the EPC implantation decreased blood glucose levels after 12 weeks and normalized (average, 82–93 mg/dL) after 18 weeks (Fig. 2A). Plasma hC-peptide, an indicator of INS derived from the grafts, increased with time in both EPC-implanted groups but was higher in STZ-treated mice (STZ-EPC) than in mice without STZ-treatment (non–STZ-EPC) during the early postimplantation period (4–13 weeks) (Fig. 2B). The levels were on a plateau after reaching about 700 pmol/L in both groups. Notably, blood glucose was in the normal range during the plateau period of plasma hC-peptide in STZ-EPC mice. Endogenous INS deficiency in diabetic mice was not affected by the EPC implantation and dramatic decrease in plasma levels of mouse C-peptide during the early postimplantation period was confirmed in STZ-EPC mice compared with non–STZ-EPC mice (Supplementary Fig. 4). These results suggest that the diabetic environment upregulates the INS-secreting capacity of the EPC grafts.

To examine whether the diabetic condition promotes maturation of INS-producing/secreting cells in grafts as β-cells, we compared plasma hC-peptide levels in fasting and nonfasting states on 13 and 18 weeks after implantation. Plasma levels of hC-peptide decreased in response to an overnight fast in both STZ-EPC and non–STZ-EPC mice (Fig. 2C and D). Additionally, trends and levels of hC-peptide secretion were identical in both groups when mice were orally administered glucose, suggesting that β-cell maturation in grafts is not different in both groups at 18 weeks (Fig. 2E). These results suggest that upregulation of INS-secreting capacity of the EPC grafts in the diabetic environment is not due to facilitated β-cell maturation from EPCs in grafts. Since the host diabetic environment upregulates the INS-secreting capacity of the grafts during the early postimplantation period (4–13 weeks) (Fig. 2B), we focused on the early period for further analysis.

Next, to examine whether the positive effect of INS-deficient diabetic condition on INS-secreting capacity of EPC grafts was dependent on the iPSC line, we used EPCs derived from another human iPSC cell line Ff-WJ18. Ff-WJ18–derived EPCs, which contained 97.8 ± 1.0% of SOX17⁺ cells at definitive endoderm stage, 74.4 ± 13.7% of PDX-1⁺NKX6.1⁺ cells at pancreatic endoderm stage, and 58.3 ± 14.4% of CHGA⁺ cells and very few INS-expressing cells at endocrine progenitor stage, were implanted under the kidney capsule of STZ-treated diabetic NOD/SCID mice. During the period of hyperglycemia, plasma hC-peptide levels were sustainably increased up to 10 weeks after implantation in EPC-implanted mice while the increase were not evident in mice without STZ-treatment (Fig. 3A and B). Additionally, we used Akita-NOD/SCID mice, which carry an Ins2^Akita mutation resulting in INS-dependent diabetes, to investigate the host effect in another diabetic model. The levels of endogenous mouse INS in plasma and pancreas were almost depleted in both diabetic models (Supplementary Fig. 5). Consistently, the intensity of hyperglycemia was similar between STZ-treated and Akita mice (Fig. 3C). By EPC implantation, plasma hC-peptide levels were sustainably increased up to 8 weeks postimplantation in all groups, while the levels were higher in diabetic mice than in corresponding nondiabetic control mice at each time point (Fig. 3D). These results suggest that the host effect is cell-line independent and diabetic model independent.

To determine the causative regulator of the host effect, we investigated whether INS deficiency in diabetes upregulates INS-secreting capacity of the grafts. First, we subcutaneously implanted an osmotic pump to infuse human INS at 40–60 nmol/kg/day before EPC implantation (Fig. 4A). The INS infusion ameliorated STZ-induced hyperglycemia below average 300 mg/dL before and up to 11 weeks after EPC implantation. Plasma human INS levels were increased by the INS infusion while plasma hC-peptide levels were lowered to the levels in nondiabetic mice (Fig. 4B and C).

Next, we used a hyperinsulinemia model that has a compensatory increase in endogenous INS induced by feeding an HFD. NOD/SCID mice were implanted with EPCs and thereafter fed with CD or HFD for 12 weeks. Nonfasting blood glucose levels were similar in mice fed with CD and HFD 12 weeks after implantation, except that the HFD feeding induced hyperinsulinemia (Fig. 4D and E). In contrast to plasma INS, plasma hC-peptide was downregulated up to 12 weeks after implantation in HFD-fed mice (Fig. 4F). On the basis of the downregulation, the hyperinsulinemia in HFD-fed mice was confirmed to be due to INS resistance and compensatory upregulation of endogenous INS secretion (Fig. 4E). Collectively, an increase in plasma INS attenuated human C-peptide secretion from the EPC grafts, suggesting that an INS deficiency in the host plays a positive role in the upregulation of the INS-secreting capacity of the grafts.

To investigate the involvement of plasma glucose levels for the host effect, hyperglycemia was attenuated without INS supplementation in STZ-induced diabetic NOD/SCID mice before and after EPC implantation. Dapagliflozin, a selective SGLT2 inhibitor that inhibits renal glucose reabsorption, was administered orally once daily at a dose of 10 mg/kg, which is the dose estimated to lower blood glucose levels (26). The dapagliflozin treatment stably improved hyperglycemia, yet nonfasting blood glucose levels were over ∼300 mg/dL, even with EPC implantation (Fig. 5A). After the EPC implantation, plasma hC-peptide levels in mice with dapagliflozin were comparable to those in mice with vehicle treatment (Fig. 5B), which were higher than those in nondiabetic mice. These results suggest that partial attenuation of hyperglycemia via INS-independent mechanisms has little influence on the host effect for EPC grafts.

To clarify how the INS-secreting capacity is improved in the grafts, we analyzed cell mass and composition of the explanted EPC grafts at 6 weeks after implantation. The appearance and weight of the whole graft were not different between diabetic and nondiabetic mice despite different plasma hC-peptide levels at 6 weeks after implantation (Fig. 6A and B and Supplementary Fig. 6A). In contrast, whole graft contents of both INS and GCG were higher in the grafts from diabetic mice compared with that from nondiabetic mice (Fig. 6C). When analyzed individually, a strong positive correlation was shown between plasma hC-peptide levels on 6 weeks and INS content of the explanted grafts (Fig. 6D). In addition, grafts from diabetic mice contained approximately three times more immunoreactivity areas of CHGA and INS, whereas that of NKX6.1 and GCG did not show significant differences (Fig. 6E–H). Moreover, the INS-immunoreactivity area in grafts from diabetic mice contained more INS⁺Ki67⁺ proliferative cells (Fig. 6I and Supplementary Fig. 6B). The grafts from both diabetic and nondiabetic mice contained PDX-1⁺CK19⁺ duct-like structures in which SOX9⁺ progenitor cells were almost exclusively observed (Supplementary Fig. 6C). These structures included few CHGA⁺ cells and similarly contained Ki67⁺ cells between grafts from diabetic and nondiabetic mice, suggesting that these are less likely to be related to the INS-secreting capacity of grafts. Regarding β-cell maturation markers, immunofluorescence signals of urocortin-3 and MAFA in INS⁺ cells were not observed in either group (data not shown). Taken together, these results suggest that an INS-deficient diabetic condition in the host increases endocrine cells, including INS-producing cells, without affecting graft mass at least via proliferation. We preliminarily analyzed the grafts at 24 weeks after implantation in another implantation study (Fig. 2), and we found that the grafts from diabetic mice contained more INS and GCG compared with that from nondiabetic mice without affecting graft mass (INS: 5.27 ± 1.66 and 1.43 ± 0.52 nmol/whole graft in diabetic and nondiabetic mice, respectively; GCG: 1.59 ± 0.27 and 0.37 ± 0.02 nmol/whole graft in diabetic and nondiabetic mice, respectively). This suggests that the increase of INS-producing cells in diabetic hosts is maintained after reaching a plateau level in plasma hC-peptide.

To examine the influence of glucose and INS on the differentiation propensity of EPCs toward INS-producing cells, EPCs were cultured for 10 days with previous report-based differentiation medium (18,20). Under the condition of both low and high glucose, the differentiation induction efficiencies of INS⁺NKX6.1⁺ cells were slightly but significantly increased by adding an INS receptor antagonist S961 (Supplementary Fig. 7A and B). In addition, the proportion of INS⁺NKX6.1⁺ cells was also increased by changing the medium to be INS-free in the low-glucose medium. The increase by changing to the INS-free medium was not reached to statistical difference in the high-glucose medium (P = 0.059). Regarding the influence of glucose, condition of high glucose did not increase the values of INS⁺NKX6.1⁺ cell proportion compared with that of low glucose. The proportion of INS⁺Ki67⁺ cells and the aggregate cell number did not differ between groups (Supplementary Fig. 7C–E). These results suggest that, at least for an in vitro differentiation induction system, lowering the INS action directly promotes the differentiation of EPCs into INS⁺NKX6.1⁺ cells.

Intensified research efforts are ongoing to establish cell therapy using stem cell–derived pancreatic cells against T1DM (4). One of the unsolved issues for their clinical application is an influence of the host environment on the fate of cells after implantation. We examined whether a host INS-deficient diabetic condition affects the fate of hiPSC-derived EPCs in immunodeficient mice and made two major findings. First, the host diabetic environment accelerates an increase in INS secretion capacity of the EPC graft via mechanisms that involve INS deficiency at least as a part. Second, the grafts from diabetic hosts contained more endocrine cells, including INS-producing cells, without increasing the graft mass compared with that from nondiabetic hosts. On the basis of highly positive correlation between the INS secretion capacity and INS content of the EPC grafts, these two findings are likely to be related.

The mechanism underlying the accelerated increase in plasma human INS/C-peptide levels in diabetic conditions (Figs. 2B, 3B and D, 4C, 5B, and 6A) could be explained by two intragraft events. One is an increase in the number of INS⁺ cells in the grafts. The other potential explanation is an increase in INS-secretory capacity from individual cells, which indicates accelerated maturation toward adult-type β-cells. Immunohistochemical analyses suggest the former model by showing that EPC grafts contained more INS⁺ cells and proliferating INS⁺ cells in diabetic hosts (Fig. 6G–I). Additionally, despite findings in the in vitro model, lowering INS action promoted the differentiation of EPCs into INS⁺NKX6.1⁺ cells without increasing the proportion of INS⁺Ki67⁺ cells (Supplementary Fig. 7A–D), suggesting that the shift in the differentiation propensity toward INS⁺ cells is also possibly involved in the increased number of INS⁺ cells in diabetic hosts. On the other hand, plasma hC-peptide derived from EPC grafts responded to fed/fast status in both diabetic and nondiabetic mice even when the absolute values were different between the two groups (Fig. 2C). Considering the increase in INS⁺ cell number, there is no clear evidence to support the latter model. Therefore, we reason that the increased number of INS⁺ cells mainly contribute to the accelerated increase in INS secretion, at least in part via promoted proliferation of INS⁺ cells and possibly via the shifted differentiation propensity toward INS⁺ cells.

Numerous alterations occur in the diabetic condition, such as abnormal INS levels, hyperglycemia, and resulting complications (2). On the basis of our findings from intervention of blood INS levels (Fig. 4B and E) and in vitro experiments (Supplementary Fig. 7), we are proposing that INS deficiency in the host plays a positive role in the upregulation of the INS-secreting capacity of EPC grafts. While there are no reports using human fetal pancreas to investigate the host effect after implantation as in this study, reports using hESC-derived PECs support this idea. During the early period after PECs implantation, diabetic mice showed higher plasma hC-peptide levels compared with nondiabetic mice, and the increase was of a lesser degree upon exogenous INS supplementation (14). The EPCs we used in this study are developmentally considered to contain less progenitors that possess the potential to differentiate into nonendocrine cells compared with PECs. The observation of the implanted PECs reported previously (14) might involve not only the differentiation propensity of nonendocrine progenitors but also the proliferation of hormone-expressing cells in the grafts. Collectively, immature cell types, such as PEC, EPC, and immature hormone-positive cells, are likely to compensate for a decrease in blood INS levels. On the other hand, the involvement of blood glucose levels is not evident. In our study, the upregulation of the INS-secreting capacity of EPC grafts was not affected when nonfasting blood glucose levels were stably lowered by repeated dosing of dapagliflozin (Fig. 5B). Given that dapagliflozin-treated mice still maintained around 300 mg/dL of nonfasting blood glucose levels even with EPC implantation, the question of whether blood glucose levels of <300 mg/dL could have direct effect on EPC graft remains to be solved.

Our finding from in vitro experiments that lowering INS action facilitated the differentiation of EPCs into INS⁺ cells suggests the involvement of INS signaling as a molecular mechanism (Supplementary Fig. 7). Indeed, some reports have indicated the relevance of INS-signaling molecules to the differentiation of immature pancreatic cells. Phosphatidylinositol 3-kinase (PI3K) is known to be activated by INS, and its inhibition by wortmannin or Ly294002 facilitated differentiation of human fetal-derived islet-like cell clusters in vitro (27). Considering that PI3K functions as a negative regulator of islet development, this could be a mediator of our in vitro finding (Supplementary Fig. 7A and B). One of the downstream molecules of PI3K is FOXO1 of which transcriptional activity is inhibited by activation of INS signaling via phosphorylation at Ser²⁵⁶. Unlike PI3K inhibition, treatment with AS1842856, a chemical FOXO1 inhibitor, promoted differentiation from pancreatic progenitors to INS⁺ cells in hESCs (28). Because FOXO1 was inhibited from pancreatic progenitor to INS⁺ cell differentiation in the report, the INS action via FOXO1 inhibition seems to facilitate INS⁺ cell differentiation. Thus, FOXO1 is less likely to be involved in a mechanism of our in vitro finding. Therefore, although the detail remains unknown, our proposed mechanism is that the activation of INS signaling via PI3K through downstream molecules other than FOXO1 negatively regulates differentiation into INS⁺ cells, while the attenuation of the signaling such as in the INS-deficient diabetic conditions results in facilitated differentiation. Regarding the increase in proliferating INS⁺ cells found in vivo (Fig. 6I), we assume that some cell types other than implanted EPCs are involved, since an increase in INS⁺Ki67⁺ cells was not observed in vitro (Supplementary Fig. 7C and D). Further studies with genetically modified mice in terms of INS signaling may contribute to elucidating the direct target cells of INS action in vivo.

We found that HFD feeding after implantation attenuates the INS-secreting capacity of the implanted cells (Fig. 4D–F), which is apparently in conflict with the previous report by Bruin et al. (29) that showed no impact on the implanted cells. However, the timing of HFD feeding is different between the two studies because we started immediately after implantation while cells were implanted following a 7-week HFD feeding, at which time the onset of INS resistance was confirmed in the report. We speculate that the timing of the onset of INS resistance explains the major difference between the report and our findings since the onset of INS resistance may alter the whole-body signaling state (30). In addition, the detailed experimental designs were different, such as the implantation site (subcutaneous space and kidney subcapsule), encapsulation (an immunoisolative device and naked), and even the cell type (PECs and EPCs). It is important to elucidate the mechanism underling the different behavior of immature pancreatic cells in future studies.

In nondiabetic mice, plasma mouse INS/C-peptide was decreased following EPC implantation (Supplementary Fig. 4). A similar phenomenon was observed in previous reports using PECs (31) and human islets (32). Mouse normoglycemia is considered a mild hyperglycemic state for human islets, and xenoislet grafts are reported to shift the glycemic levels to typical of the islet donor species in mice (32). Therefore, the lower threshold of glucose-stimulated INS secretion from EPC grafts compared with endogenous islets is considered to cause the replacement of plasma mouse INS with human INS derived from EPC grafts in nondiabetic mice.

Posttransplant care for the recipients potentiates islet graft function (2). For example, the type of immunosuppressant affects the islet graft function in the clinical setting (33). Another case is that the reduction of adult islet graft workload by a combination of fasting and INS treatment during the early posttransplant period improves the long-term engraftment in rats (34). In the current study, we demonstrated that exogenous INS treatment to control blood glucose levels lowered the INS-secreting capacity of EPC grafts (Fig. 4C), which contained few INS-expressing cells as of implantation (Fig. 1B and C and Supplementary Fig. 1). Therefore, we speculate that INS treatment for posttransplant care might lower the therapeutic potential of the grafts for cell therapy against T1DM when using incompletely matured pancreatic endocrine cells. Recently, as a noninsulin adjunct pharmacological therapy against T1DM, SGLT2 inhibitors including dapagliflozin were approved and were reported to improve glycemic control with reducing the total daily INS dose (35). Because INS deficiency is considered to potentiate the function of immature grafts, selection or combinational use of hypoglycemic agents other than INS might be a good option to achieve postimplantation glycemic control during graft maturation. Taken together, results in this study would contribute to gain insights for future cell therapy using alternative sources of adult islets in terms of what therapy would serve as the most suitable postimplantation care for T1DM patients.

Acknowledgments. The authors thank Drs. Shinya Yamanaka and Kenji Osafune (CiRA, Kyoto, Japan) and Drs. Seigo Izumo, Yasushi Kajii, and Atsushi Nakanishi (Takeda Pharmaceutical Company Limited, Kanagawa, Japan) for supporting the collaboration research between Takeda Pharmaceutical Company Limited and CiRA (T-CiRA, Kanagawa, Japan). The authors are grateful for the technical assistance of Shika Inoue, Aki Kuwano, Ayako Makita, Miho Ohra, Aika Takahashi, Atsuno Kaneto, Akina Shima, Keiko Nishijima, Marii Ise, Mayumi Katsumata, Rie Tezuka, and Dr. Jun Yasumoto (all from T-CiRA). The authors thank Drs. Midori Yamasaki, Kensuke Sakuma, and Shuhei Konagaya for helpful discussions; Stephanie Napier for proofreading the manuscript; and Junji Yamaura, Miyuki Yamatani, and Goshi Nakamura for supporting the in vitro experiments (all from T-CiRA). The authors also thank Takeshi Yamamura for breeding and supply of Akita-NOD/SCID mice (Axcelead Drug Discovery Partners Inc., Kanagawa, Japan).

Duality of Interest. This work was supported by Takeda Pharmaceutical Company Limited. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. T.M. and T.T. conceptualized the study. T.M., H.U., N.T.-Y., and H.H. created the methodology for the study. T.M., H.U., N.T.-Y., and H.H. conducted the investigation. T.M. wrote the original draft. R.I., H.M., and T.T. reviewed and edited the manuscript. R.I., H.M., and T.T. supervised the study. T.M. and T.T. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented at the 55th Annual Meeting of the European Association for the Study of Diabetes, Barcelona, Spain, 16–20 September 2019.