Introduction

Induced pluripotent stem cells (iPSCs) can be generated directly from somatic cells [1–4], which makes it possible to obtain patient-specific multipotent cells and differentiate these cells into particular lineages for cell replacement therapy and avoid the problems of donor cell shortage, immune rejection and ethical debates related to procedures using human embryos. Disease-specific iPSCs have been established from individual patients [5–9], and differentiation of human iPSCs into lineage-specific cells has been achieved [10–13]. However, the efficacy and safety of iPSC-derived cells in clinical applications in human disease therapies must first be evaluated using animal models.

So far, in mouse models, lineage-specific cells have been differentiated from iPSCs to cure several types of disease including sickle cell anaemia [14], Parkinson’s disease [15], haemophilia A [16] and diabetes [17]. However, rodent models cannot fully recapitulate human diseases and there have not yet been reports of iPSC application research in large animal models.

The rhesus monkey (Macaca mulatta, rhesus macaque), one of the most frequently and thoroughly studied model species of all non-human primates, is a suitable candidate. It shares approximately 93% of its DNA sequence with humans and exhibits greater similarity to human physiology and susceptibility to infectious and metabolic diseases than rodents do. Moreover, we demonstrated in our previous study that rhesus monkey (RM) iPSCs can be generated by transduction of the same four transcription factors used in humans [18]. Given these similarities, the rhesus monkey is a very relevant primate model and would help in the investigation of iPSC therapeutic application in certain diseases, such as type 1 diabetes.

Type 1 diabetes is caused by the autoimmune destruction of beta cells in the pancreas. Pancreatic islet transplantation is a promising therapy for type 1 diabetes, because it has been demonstrated to allow insulin independence with glycaemic control in some patients for over 2 years [19]. However, the extreme lack of donor beta cells and the threat of immune rejection impede its widespread application. To solve these problems, several studies have attempted the pancreatic differentiation of human iPSCs [9, 12, 20] based on protocols used in human embryonic stem cell (ESC) pancreatic differentiation [20–22], and obtained insulin-producing cells with some characteristics of islet beta cells. However, no functional studies of these iPSC-derived insulin-producing cells using in vivo transplantation were performed in these studies.

In the present study, we established a new four-stage protocol to differentiate RM iPSCs to generate definitive endoderm (DE), pancreatic progenitors (PP) and endocrine precursors (EP), and then to further induce these EP cells into insulin-producing cells by both in vivo maturation and in vitro differentiation. Moreover, the RM iPSC in vitro-derived insulin-producing cells can secrete insulin in response to glucose stimulation, and rescue hyperglycaemia in approximately 50% of streptozotocin (STZ)-treated diabetic mice. To our knowledge, this is the first report detailing the generation of functional insulin-producing cells from primate iPSCs, and will provide the basis for achieving proof of principle of iPSC-based therapy for type 1 diabetes in large animal models.

Methods

Animals

Six- to 8-week old male NOD/SCID mice (Vital River Laboratories, Beijing, China) were used for transplantation, and all animal procedures were approved by the Institutional Animal Care and Use Committee of Peking University.

Generation and maintenance of RM iPSCs

The detailed protocol for establishing RM iPSCs has been described previously [18].

Directed differentiation of RM iPSCs into pancreatic insulin-producing cells

The differentiation protocol is described in Fig. 1a. The media used here, RPMI1640, DF12 and DMEM, were all supplemented with 2 mmol/l Glutamax, 0.1 mmol/l non-essential amino acids, 55 μmol/l beta-mercaptoethanol and penicillin/streptomycin. DF12 and DMEM were also supplemented with 1% B27 (all from Gibco, Carlsbad, CA, USA).

Fig. 1
figure 1

The scheme of RM iPSC differentiation into insulin-producing cells. a The differentiation protocol is divided into four stages, and the growth factors and duration for each stage of differentiation are shown. EX4, exendin-4; Nico, nicotinamide. b–j Quantitative PCR analysis showed the dynamic expression of several key genes during pancreatic beta cell development. For each gene, the sample with the highest expression level was set to 1, and the other samples were normalised to this level. Data represent mean ± SD, and error bars indicate SD. Undif, undifferentiated. Similar results were obtained in at least three independent experiments

Full size image

In step 1, the undifferentiated iPSCs were treated with 100 ng/ml activin A (PeproTech, Rocky Hill, NJ, USA) and 25 ng/ml wingless-related MMTV integration site 3A (WNT3A) (R&D Systems, Minneapolis, MN, USA) in RPMI1640 for 1 day. On the following 2 days, 100 ng/ml activin A was added, supplemented with 0.2% FBS. Cells were then digested with 0.25% trypsin (Gibco) into single cells and seeded at a density of 0.5 × 105 to 1 × 105 cells per well of a 48-well plate on mouse embryonic fibroblast (MEF) feeders. In step 2, the cells were cultured in DF12 supplemented with 2 μmol/l all-trans retinoic acid (RA; Sigma, St Louis, MO, USA) and 50 ng/ml keratinocyte growth factor (KGF; PeproTech) for 6 days. Following this, in step 3, cells were treated with 50 ng/ml EGF (PeproTech) and 1 μmol/l SB431542 (Tocris Bioscience, Ellisville, MO, USA) in DMEM for 5 days and, in the final step, the cells were incubated in DMEM with 50 ng/ml hepatocyte growth factor (HGF; PeproTech), 50 ng/ml IGF1 (PeproTech), 50 ng/ml exendin-4 (Sigma) and 10 mmol/l nicotinamide (Sigma) for 7 days.

The in vitro differentiation experiments were carried out using R14 and R19 cell lines and repeated more than five times. Similar results were obtained and data presented are from R14, unless otherwise indicated.

Immunofluorescence

The differentiated RM iPSCs were briefly washed with PBS, fixed in 4% (wt/vol.) paraformaldehyde for 20–30 min, washed three times in PBS, and blocked with PBST (PBS/0.1% Triton X-100 [Sigma]) containing 3% normal donkey serum (Jackson ImmunoResearch Laboratory, West Grove, PA, USA) for 60 min. Cells were incubated with primary antibodies in PBST overnight at 4°C and then washed three times. They were then incubated with secondary antibodies in PBS containing 0.1% BSA (Jackson ImmunoResearch Laboratory) for 1 h at room temperature, avoiding exposure to light. Afterwards, cells were washed with PBS. The nuclei were stained with 1 μg/ml DAPI (Roche Applied Science, Indianapolis, IN, USA) for 5 min. Images were captured by a Nikon Eclipse TE2000-U fluorescence microscope (Nikon, Tokyo, Japan) or a Leica confocal microscope TCS-SP2 (Leica, Solms, Germany).

The primary and secondary antibodies are described in the Electronic supplementary material (ESM) Tables 1 and 2, respectively.

Flow cytometry

For surface marker staining, cells were digested with 0.25% trypsin into single cells. After being washed, the cells were incubated in 200 μl PBS containing 0.5% BSA for 15 min and then stained with phycoerythrin-conjugated mouse anti-human chemokine (C-X-C motif) receptor 4 (CXCR4) antibody (BD Biosciences, Bedford, MA, USA) and allophycocyanin (APC)-conjugated mouse anti-human kinase insert domain receptor (KDR) (R&D Systems), CD117 (BD Biosciences) or CD140a (BD Biosciences) antibodies at 5μl/2 × 105 cells for 30 min at 4°C. Cells were then washed three times and resuspended in 300 μl PBS for analysis.

The intracellular staining was performed using Cytofix/Cytoperm and Perm/Wash buffer (BD Biosciences) according to the manufacturer’s instructions. The primary and secondary antibodies are described in ESM Tables 1 and 2, respectively.

Data were acquired on a FACS Calibur (BD Biosciences) and were analysed using CellQuest software (BD Biosciences). At least three independent experiments were carried out for each test.

Real-time quantitative PCR

RNA was isolated using the RNeasy Plus Mini Kit (Qiagen, Valencia, CA, USA) following the manufacturer’s instructions. Following this, 2 μg RNA was reverse transcribed using the EasyScript First-Strand cDNA Synthesis SuperMix (Beijing Transgen Biotech, Beijing, China). Real-time PCR was performed using the ABI Prism 7300 Sequence Detection System. The primer sequences were designed according to the corresponding human genes and are shown in ESM Table 3. PCR was performed in duplicate for each sample, and three independent experiments were carried out.

Glucose-stimulated insulin secretion and intracellular insulin content analysis

To analyse glucose-stimulated insulin secretion, cells were washed twice with PBS and first incubated with Krebs–Ringer buffer free of glucose for 60 min at 37°C. The cells were then incubated with Krebs–Ringer buffer containing low (2.5 mmol/l) or high (25 mmol/l) concentrations of glucose for 60 min at 37°C and supernatant fractions were collected. To analyse the intracellular content of insulin, the final stage cells ware treated with lysis buffer. The assays for the insulin content were performed using an Insulin ELISA kit (Mercodia, Uppsala, Sweden).

Transplantation

Cells were transplanted into NOD/SCID mice. Differentiated cells were digested with 1 mg/ml dispase (Gibco) and resuspended in ultra-low dishes (Corning, Corning, NY, USA) the day before transplantation.

Diabetes was induced by intraperitoneal injection of 120 mg/kg body weight STZ (Sigma) in NOD/SCID mice. The experimental group had about 0.5 × 107 to 1 × 107 differentiated cells implanted under the left kidney capsule, and the control group experienced a similar operation with transplantations of PBS. The blood glucose levels were monitored weekly with glucometer strips (Accu-check, Roche) using blood from snipped tails.

The grafts were examined after 8–10 weeks. The kidneys bearing the grafts were embedded in OCT (optimal cutting temperature) compound (Sakura, Torrance, CA, USA) and cut into cryostat sections. Immunofluorescence analyses were performed to detect the expression of pancreatic endocrine hormones and transcription factors. The antibody information can be found in ESM Tables 1 and 2.

Statistical analysis

For insulin release data, a two-tailed unpaired Student’s t test was used for statistical evaluation. For gene expression comparison among different treatments, statistical analysis was performed using a one-way ANOVA, with Duncan’s multiple range post-hoc tests. In all statistical analyses, p < 0.05 was considered statistically significant.

Results

Screening of RM iPSCs for pancreatic differentiation

We generated RM iPSCs by retrovirus-mediated transduction of the monkey transcription factors octamer binding protein (OCT)3/4, SRY-box containing gene (SOX)2, Kruppel-like factor 4 (KLF4) and c-MYC into the adult fibroblasts as previously described [18].

Because it has been reported that different human ESC and iPSC lines vary in their capacity for differentiation [10, 23], we performed a screen of the RM iPSCs to select cell lines for pancreatic differentiation. Nineteen RM iPSC lines from the same rhesus monkey were induced to generate DE cells and analysed by flow cytometry for chemokine (C-X-C motif) receptor 4 (CXCR4) abundance. As shown in ESM Fig. 1, these RM iPSCs exhibited various levels of CXCR4 and approximately four- to five-fold differences were observed. Among these cells, R14 and R19 showed the highest CXCR4 levels of over 90% and also exhibited the greatest ability to generate pancreatic and duodenal homeobox 1 (PDX1)+ PP cells upon further differentiation (data not shown). Therefore, R14 and R19 were chosen for the pancreatic differentiation study.

Establishment of a new four-stage differentiation protocol

Using current knowledge of signalling regulation in embryonic pancreatic development, we established a four-stage protocol for RM iPSC differentiation into pancreatic insulin-producing cells (Fig. 1a), with steps 1 and 4 similar to the ViaCyte protocol [24]. The following four stages, DE induction, PP generation, EP specification and insulin-producing cell maturation, recapitulate the embryonic pancreatic beta cell development very well [25]. Gene expression analyses showed that, during this differentiation, RM iPSCs rapidly lost expression of the pluripotent gene OCT3/4 (also known as POU5F1) and stage-specific marker genes exhibited the highest expression levels in their corresponding stages, such as SOX17 for the DE stage, PDX1 for the PP stage, NGN3 (also known as NEUROG3) for the EP stage and INS for the final stage (Fig. 1b–j).

Efficient generation of DE and PP cells from RM iPSCs

In stage 1, RM iPSCs were treated with activin A and WNT3A for DE induction, and to monitor this process, we followed gene expression changes by immunochemistry during this 3 day period. Brachyury, the mesendoderm marker, was produced in a large portion of cells after 1 day of differentiation and the number of cells producing this marker gradually decreased, while the number of SOX17+ (protein encoded by SRY-box containing gene 17) cells increased at day 2. At the end of day 3, over 90% of the differentiated RM iPSCs were SOX17+ and no Brachyury+ cells remained (ESM Fig. 2a). Moreover, the SOX17+ cells were positive for forkhead box A2 (FOXA2), but negative for α-fetoprotein (AFP) (ESM Fig. 2b,c), excluding the presence of extraembryonic endoderm cells. Flow cytometry analysis showed that there were about 94.4% (92.2% ± 4.1%, n = 5) CXCR4+CD117+ cells. These CXCR4+CD117+ cells did not express the mesoderm markers KDR or CD140a (ESM Fig. 2d), suggesting that no mesoderm cells were induced. Therefore, RM iPSCs can be differentiated into DE cells efficiently.

In step 2, we used the method established in our laboratory, which added a dissociation treatment after step 1 to obtain high efficiency of PDX1+ cells [26]. The induced DE cells from RM iPSCs were trypsinised into single cells, replated onto MEF feeders, and subjected to RA and KGF treatment for 6 days. At the end of this stage, the replated cells expanded and formed clusters, which exhibited much higher efficiency of generating PDX1+ cells (Fig. 2a) than those cells without dissociation treatment (ESM Fig. 3a). Further experiments showed that most of these PDX1+ cells produced SOX9, FOXA2, hepatic nuclear factor (HNF) 4 alpha, HNF6 and GATA binding protein 6 (GATA6) (Fig. 2a–e), but did not produce AFP or causal type homeobox 2 (CDX2) (ESM Fig. 3b,c), indicating that these PDX1+ cells are PP cells. In addition, flow cytometry analysis showed that about 87.5% (82.5% ± 6.0%, n = 5) of the cells became PDX1 positive (Fig. 2f). These results suggest that RM iPSCs can be efficiently differentiated from DE cells to PP cells.

Fig. 2
figure 2

RM iPSC-derived DE efficiently differentiates into PP cells. a–e Most of the PDX1-producing PP cells produce SOX9, FOXA2, HNF6, HNF4A and GATA6. f Flow cytometry analysis showed that, at the end of stage 2, approximately 87.5% of the cells are PDX1+ PP cells. Scale bars, 30 μm. Undif, undifferentiated; Dif, differentiated; FL1-H, height of fluorescence intensity. Similar results were obtained in at least three independent experiments

Full size image

SB431542 promotes the generation of NGN3+ EP cells

We have previously identified EGF as being able to expand the PDX1+ PP cells [20]. In addition, activation of the TGF-β pathway could stimulate PP cells to experience exocrine differentiation, while decreasing the generation of EP cells [27]. Therefore, we treated the PDX1+ PP cells with both EGF and the TGF-β inhibitor SB431542 to promote the generation of NGN3+ (NEUROG3 protein) EP cells. Real-time PCR analysis indicated that this combination could increase the expression levels of NGN3, PDX1, HNF6 (also known as ONECUT1), NKX6-1 and PAX6, compared with single-factor treatment (ESM Fig. 4a–e).

We then tested the production of important endocrine transcription factors by immunocytochemistry. As previously reported, cells co-producing HNF6 and PDX1 are considered the progenitors of the NGN3-producing cells [28], and NGN3 is used as an EP marker [29, 30]. In this stage, cells positive for PDX1 and HNF6 were detected (Fig. 3a). NGN3-producing cells appeared in the domains of the HNF6+ cells and the PDX1+ cells (Fig. 3b,c). However, not all HNF6+ or PDX1+ cells produced NGN3, and there were also a few NGN3+ cells that did not produce PDX1 or HNF6. In addition, NK6 homeobox 1 (NKX6-1) and NK2 homeobox 2 (NKX2-2) were detected in the clusters of PDX1+ cells (Fig. 3d,e). Therefore, these results suggested that endocrine specification of RM iPSCs can be achieved.

Fig. 3
figure 3

RM iPSC-derived PPs differentiate into EPs. a–c Immunocytochemistry analysis showed that, at the end of stage 3, cells co-producing PDX1 and HNF6 could be detected, most of which were also positive for NGN3. d, e Many PDX1+ cells also produce NKX6-1 and NKX2-2. Scale bars, 50 μm (b), and 100 μm (a, c–e). Similar results were obtained in at least three independent experiments

Full size image

We then questioned whether RM iPSC-derived EP cells can differentiate into insulin-producing cells in vivo. Stage 3 cells were implanted into the kidney capsules of NOD/SCID mice. Approximately 2 months later, these implanted cells formed compact grafts in most of the mice, and the insulin-producing cells were detected with cluster morphology and produced NKX6-1 and C-peptide (Fig. 4a,b). Meanwhile, these insulin-positive (INS+) cells did not produce glucagon. However, it seems that a few INS+ cells produced somatostatin. In some areas, cells producing glucagon or somatostatin, but not insulin, were observed (Fig. 4c–e). We also found a few cells producing proinsulin (ESM Fig. 5a). In duct-structure areas, cells producing CK19 were observed (ESM Fig. 5b). These results showed that EP cells are competent to generate insulin-producing cells in vivo.

Fig. 4
figure 4

Generation of mature insulin-producing cells in vivo; the grafts were removed from mice 8–10 weeks after transplantation of stage 3 differentiated cells. a,b Immunocytochemistry analysis showed that the INS+ cells produce NKX6-1 and C-peptide (C-PEP). c–e INS+ cells did not produce glucagon (GCG) or somatostatin (SST), although GCG+ and SST+ cells could be detected. Scale bars, 50 μm (b, e), 100 μm (a, c–d)

Full size image

Generation and characterisation of insulin-producing cells in vitro

To obtain insulin-producing cells in vitro, the stage 3 EP cells were incubated with exendin-4, nicotinamide, HGF and IGF1 for 7 days. Following this, insulin-producing cells were observed, which produced C-peptide, indicating the de novo synthesis of insulin, as opposed to uptake from the culture medium (Fig. 5a). These cells co-produced PDX1 and NKX6-1, but did not produce other endocrine hormones, such as glucagon or somatostatin (Fig. 5b,c and data not shown). A small number of cells producing glucagon or somatostatin could be detected (ESM Fig. 6a,b). There were also a few amylase+ exocrine cells (ESM Fig. 6c), and we could also detect the production of proprotein convertase subtilisin/kexin type 2 (PCSK2) (ESM Fig. 6d), an enzyme involved in processing proinsulin into insulin and C-peptide. FACS analyses showed that approximately 23.0% (21.1% ± 5.8%, n = 3) of cells were insulin positive (Fig. 5d). Taken together, these results indicate that RM iPSCs can be differentiated into insulin-producing cells through a four-stage protocol in vitro.

Fig. 5
figure 5

Generation and characterisation of insulin-producing cells in vitro. a–c Immunocytochemistry analysis showed that, at the end of stage 4, PDX1+ cells produce NKX6-1 and C-peptide (C-PEP), and the INS+ cells also produce C-peptide. d FACS analysis showed that approximately 23.0% INS+ cells are generated. e Glucose-stimulated insulin secretion analysis showed that the insulin secretion of the stage 4 differentiated cells in a high concentration glucose environment is about 2.7-fold (p < 0.05) of that secreted in a low concentration glucose environment. Scale bars, 50 μm. Data represent mean ± SD, and error bars indicate SD. Undif, undifferentiated; Dif, differentiated; FSC, forward scatter; Similar results were obtained in at least three independent experiments

Full size image

We performed insulin content analysis of the insulin-producing cells and found it was approximately 8 × 10−19 to 10 × 10−19 mol/cell. As it is currently difficult for us to obtain primary adult rhesus monkey beta cells, we could not evaluate how close the insulin content of RM iPSC-derived insulin-producing cells is to primary adult beta cells. To demonstrate the insulin release of insulin-producing cells, we incubated these cells with glucose at low (2.5 mmol/l) and high (25 mmol/l) concentrations. Results showed that there was about a 2.7-fold (p < 0.05, n = 3) increase in insulin secretion at the high glucose condition, suggesting that these insulin-producing cells have the ability to release insulin in response to glucose stimulation (Fig. 5e).

In vitro-generated insulin-producing cells could rescue STZ-treated diabetic mice

We transplanted the final stage cells into the kidney capsules of NOD/SCID mice and examined the grafts after 2 months to assess the survival of these cells. We detected the presence of insulin-producing cells, which also produced C-peptide but did not produce glucagon or somatostatin (Fig. 6a–c). These results indicate that the insulin-producing cells survived during this period.

Fig. 6
figure 6

Transplantation assay of RM iPSC in vitro-derived insulin-producing cells in a mouse model for diabetes. a–c Immunocytochemistry analysis for grafts from mice transplanted with stage 4 differentiated RM iPSCs showed that the INS+ cells produce C-peptide (C-PEP), but do not produce glucagon (GCG) or somatostatin (SST). d Some diabetic mouse models (group 1, n = 8; black line) implanted with final stage cells could regulate blood glucose to a normal level. The control diabetic mouse models (n = 8; dotted line) implanted with PBS maintained high levels of blood glucose at all times. Scale bar, 100 μm. Data represent mean ± SD, and error bars indicate SD

Full size image

We then tried to determine the ability of these insulin-producing cells to regulate blood glucose levels in diabetic mouse models. We first generated diabetes models in NOD/SCID mice by STZ treatment (120 mg/kg of body weight) [31–33]; this dose of STZ was intended to result in a relatively lower mortality and higher rate of diabetic induction. Only STZ-treated mice with blood glucose above 13.9 mmol/l for 10–12 consecutive days were used for further experiments. This 13.9 mmol/l criterion has been used in previous studies in our laboratory and by others [31, 34–36]. We stained insulin in the pancreas of these mice 12 days after STZ treatment, and found that there were only a few INS+ cells remaining in these diabetic mouse models compared with normal untreated mice (ESM Fig. 7).

The final stage cells were implanted into the kidney capsules of diabetic NOD/SCID mouse models, with a control group implanted with PBS, and the glucose concentrations were observed for at least 45 days. In the control group (n = 8), the blood glucose concentration remained high, and did not decrease below 13.9 mmol/l in any of these diabetic mouse models. In the cell-implanted mice, the blood glucose exhibited different trends and mice were divided into two groups accordingly (n = 8 for each group). In both groups, the blood glucose level decreased in the first week. Over the following 5 weeks, the levels of blood glucose in group 1 mice further decreased to below 13.9 mmol/l (Fig. 6d), while in group 2, the blood glucose subsequently increased and remained hyperglycaemic (data not shown). The proportion of euglycaemic mice after cell transplantation was 50%. Therefore, these results demonstrate that insulin-producing cells differentiated from RM iPSCs can regulate glucose metabolism to rescue hyperglycaemia in some diabetic mouse models.

Discussion

In this paper we have described a new four-stage protocol for the pancreatic differentiation of RM iPSCs, and each stage of differentiation was characterised by specific gene expression analysis (Figs 2, 3 and 5, ESM Fig. 2). We demonstrate for the first time that insulin-producing cells can be generated from non-human primate iPSCs.

Several previous studies have attempted the pancreatic differentiation of human iPSCs [9, 12, 20]. Tateishi et al. [12] first reported the generation of insulin-producing islet-like clusters containing both C-peptide-positive and glucagon-positive cells from human iPSCs by using Jiang’s protocol [21]. In our laboratory, we obtained PDX1 and C-peptide double positive cells from human iPSCs based on our optimal strategy for human ESCs [20]. Maehr et al. demonstrated that iPSCs can be generated from patients with type 1 diabetes and could be subsequently differentiated into insulin-producing cells [9]. However, none of these studies examined the in vivo function of iPSC-derived insulin-producing cells [9, 12, 20]. In the present study, we transplanted RM iPSC-derived insulin-producing cells into immunodeficient diabetic mouse models to assess their function. We found that these cells could survive for at least 2 months in vivo, co-producing PDX1, NKX6-1 and C-peptide. Moreover, these insulin-producing cells could restore approximately 50% of diabetic mouse models to euglycaemia.

In this study, we established a quantitative system using flow cytometry analysis to evaluate the differentiation efficiency. We chose CXCR4 as a marker for DE cells [37], PDX1 as a marker for PP cells [38, 39] and insulin as a marker for insulin-producing cells. Our FACS data presented here demonstrate that the generation of DE and PP cells can be achieved at very high efficiency (Fig. 2 and ESM Fig. 2), while further differentiation into insulin-producing cells is inefficient (Fig. 5). Immunostaining experiments further confirmed these results. This indicates that, in future work, identification of the critical signals and optimisation of the differentiation conditions for PP cells to generate EP cells, and then insulin-producing cells, from iPSCs need to be further investigated.

In our protocol, the high efficiency of over 85% for generating PDX1+ PP cells from RM iPSCs was due to the optimisation of the procedure following the DE stage, according to a recent report in our laboratory [26]. However, our previous work did not use in vivo assays to examine whether the disruption and re-establishment of cell–cell contact would affect the maturation and functioning of insulin-producing cells. In the present work, we transplanted the differentiated cells in stages 3 and 4 from RM iPSCs into immunodeficient mice and detected the presence of insulin-producing cells in the grafts (Figs 4 and 6). Therefore, our results provide evidence that cell dissociation at the end of the DE stage is crucial for efficient PP differentiation and does not prevent the generation of insulin-producing cells.

During the process of directed differentiation from RM iPSCs into insulin-producing cells, we obtained NGN3+ EP cells at the end of stage 3. In pancreatic development, beta cells are generated from a transient population of NGN3+ EP cells [40]. However, in previous reports on pancreatic differentiation from iPSCs, generation of NGN3+ EP cells has not been demonstrated [9, 12, 20]. In the present study, we obtained NGN3+ cells, most of which co-produced PDX1 and HNF6 (Fig. 3). Moreover, we demonstrated for the first time the competence of primate iPSC-derived EP cells to produce insulin-producing cells in vivo. Similarly, a previous study demonstrated that transplantation of hESC-derived pancreatic endoderm in mice could generate insulin-producing cells in vivo [41]. The transplanted cells in this study produce PDX1, FOXA2 and HNF6, and some produce NKX6-1, indicating a cell fate as PP cells. Our transplanted RM iPSC-derived stage 3 cells produced PDX1, NGN3, NKX6-1 and NKX2-2 (Fig. 3), and were referred to as pancreatic EP cells. Therefore, the in vivo maturation competence of cells was demonstrated at a later differentiation stage than in the previous report, which also indicates the validity of our differentiation protocol for the generation of insulin-producing cells.

We identified the chemical compound SB431542 that, in the presence of EGF, promoted the generation of NGN3+ EP cells. SB431542 specifically antagonises TGF-β signalling by inhibiting type 1 activin receptor-like kinase (ALK) receptors ALK4, ALK5 and ALK7. The roles of TGF-β in pancreatic development have been previously reported [27, 42–44]. In several studies, inactivation of TGF-β signalling in mice either by generation of activin receptor mutants [42, 43] or conditional generation of SMAD family member 7 (SMAD7) [44] decreased endocrine cell differentiation and resulted in islet hyperplasia. In another study, mice deficient in growth differentiation factor (GDF) 11 harboured increased numbers of NGN3+ EP cells, although their beta cell number was reduced [27]. Our results suggest that inhibition of TGF-β signalling could promote NGN3+ endocrine specification, which is consistent with a recent report regarding pancreatic differentiation of human ESCs [45].

In summary, our results demonstrate that insulin-producing cells can be generated by this stepwise differentiation method from RM iPSCs. In the future, we will transplant these insulin-producing cells into the corresponding rhesus monkey model for diabetes to further investigate the efficacy and safety of autologous iPSC-derived beta-like cells for type 1 diabetes therapy in rhesus monkey models.