Imeglimin Ameliorates β-Cell Apoptosis by Modulating the Endoplasmic Reticulum Homeostasis Pathway

Patients with type 2 diabetes have a complex phenotype that includes the impairment of insulin secretion and resistance to the action of insulin, both of which contribute to the development of hyperglycemia (1,2). Loss of functional β-cell mass is one mechanism by which diabetes can develop (3). Diabetes conditions include hyperglycemia, oxidative stress, and inflammation that perturbs correct protein folding, leading to the accumulation of misfolded proteins in the endoplasmic reticulum (ER) lumen and inducing ER stress in β-cells. ER stress plays an important role in the progressive reductions in insulin secretion that are associated with β-cell failure and apoptotic β-cell death (4,5).

Imeglimin, a tetrahydrotriazine-containing class of oral glucose-lowering agents, is synthesized from metformin as a precursor through a single-step chemical reaction (6,7). Imeglimin reportedly affects the liver, skeletal muscle, and pancreas, thereby improving glycemic control (7); reduces glucose and lipid biosynthesis in the liver by suppressing anabolic processes; and increases glucose uptake in skeletal muscle, which contributes to its antihyperglycemic action (7,8). In patients with type 2 diabetes, fasting blood glucose levels, postprandial glycemia, and insulin secretion in response to glucose during a hyperglycemic clamp were improved after treatment with imeglimin for 1 week (9). Imeglimin increased insulin release in response to glucose in isolated islets from rats via the augmentation of mitochondrial metabolism and enhanced glucose-stimulated insulin secretion (GSIS) in vivo in rats fed a high-fat diet (10). In an animal model of metabolic syndrome–related cardiomyopathy, imeglimin mitigated left ventricular diastolic, vascular endothelial, and renal dysfunction at least partly through a reduction in oxidative stress (11). Imeglimin prevented endothelial cell apoptosis induced by oxidative stress through target mitochondrial permeability transition pore opening and cytochrome C release (12). Imeglimin was also shown to normalize glucose tolerance and improve insulin sensitivity by protecting mitochondrial function from oxidative stress and favoring lipid oxidation in the livers of mice fed a high-fat, high-sucrose diet (8). However, the effects of imeglimin on cell survival in human β-cells remain unclear.

In this research, we investigated the effects of imeglimin on β-cell apoptosis in multiple models, including immortalized cells, mouse islets, human islets, human pluripotent stem cells (hPSCs), and in vivo treatment in mice with β-cell failure. We clarified that the modulation of the ER homeostasis pathways by imeglimin shifts the translational status from repression to the recovery phase and improves the survival of β-cells by protecting them from apoptosis.

MIN6K8 cells were used and maintained in high-glucose (25 mmol/L) DMEM supplemented with 10% FBS, 0.05% β-mercaptoethanol (99% cell culture tested), and 1% penicillin/streptomycin. MIN6K8 cells were seeded in six-well plates and treated with 1 μmol/L thapsigargin (33637-31; Nacalai Tesque), 1 μmol/L integrated small-molecule stress response inhibitor (ISRIB) (SML-0843; Sigma-Aldrich), or 1 mmol/L imeglimin (imeglimin hydrochloride, HY-14771A/CS-2121; MedChemExpress) and used for Western blotting.

Akita mice with a C57BL/6N background were obtained from Japan SLC. Wild-type (WT) and Akita mice with similar body weight were weaned at 21 days after birth and fed a standard diet (MF; Oriental Yeast, Tokyo, Japan) containing or not containing 0.13% imeglimin from days 21 to 35 after birth. On the 35th day, blood samples were taken from the tail vein before dissection. C57BL/6J mice were obtained from Japan SLC and CLEA Japan. C/EBP homologous protein (CHOP) knockout (CHOP^−/−) mice were described previously (13). All the experiments were conducted on male littermates. All animal procedures were performed in accordance with institutional animal care guidelines and the guidelines of the animal care committees of the Gunma University and Yokohama City University. The animal housing rooms were maintained at a constant room temperature (25°C) and on a 12-h light (7:00 a.m.)/dark (7:00 p.m.) cycle.

Human islets were obtained from the Alberta Islet Distribution Program or Prodo Laboratories. Further details of the human islets are described in Checklist for Human Islets (14). All studies and protocols using human islets were approved by the Clinical Islet Laboratory, University of Alberta/Alberta Health Services, Canada. Upon receipt, human islets were cultured in Miami Media #1A (Cellgro/Mediatech) overnight. The islets were treated with or without 1 μmol/L thapsigargin, 1 mmol/L imeglimin, or 1 μmol/L ISRIB for 24 h in Final Wash/Culture Medium (Cellgro), embedded in agarose, and used for the immunostaining studies.

H9 hPSCs were differentiated into β-like cells following a published protocol (15). All studies were approved by the A*STAR Institutional Review Board 2020-096, Singapore. Upon receipt, the pseudoislets were cultured in Miami Media #1A overnight, embedded in agarose, and used for the immunostaining studies.

After preincubation of MIN6K8 cells at 2.8 mmol/L glucose for 12 h, cells were cultured at 20 mmol/L glucose for 2, 4, 8, or 24 h. Puromycin (dihydrochloride, A1113803; Invitrogen) then was added to the culture medium at 10 μg/mL for 10 min, and the cells were washed twice with cold PBS and lysed as described previously (16). The cells were treated with 1 μmol/L thapsigargin and/or 1 mmol/L imeglimin as described above. Western blotting was performed as described below. Puromycin incorporation was also assessed using O-propargyl-puromycin (OPP) probe by Protein Synthesis Assay Kit (601100; Cayman Chemical) according to the manufacturer’s instructions (3 × 10⁴ MIN6K8 cells/well).

Mouse islets or cells were solubilized in lysis buffer with protease inhibitors (03969-21; Nacalai Tesque) and phosphatase inhibitors (07575-51; Nacalai Tesque). The protein concentration was measured using a BCA Protein Assay Kit (06385-00; Nacalai Tesque). After SDS-PAGE at 30 mA for 1 h, proteins were transferred to polyvinylidene fluoride membrane (Immobilon-P; Millipore) using Trans-Blot Turbo (25V-1.0A; Bio-Rad) for 30 min. After blocking with Blocking One (L1F6030; Nacalai Tesque) for 30 min at room temperature, immunoblotting was performed with antibodies described in Supplementary Table 1. Densitometry was performed using ImageJ software.

Total RNA was isolated from pancreatic islets using an RNase-Free DNase and RNeasy Kit (QIAGEN, Valencia, CA). cDNA was prepared using High-Capacity cDNA Reverse Transcription Kits (Applied Biosystems) and was subjected to quantitative PCR using SYBR Green Gene Expression Assays (7900 Real-Time PCR System; Applied Biosystems) with THUNDERBIRD qPCR Master Mix (TOYOBO). Each quantitative reaction was performed in duplicate. Data were normalized according to the β-actin or GAPDH level. The primers described in Supplementary Table 2 were used for amplification.

Blood glucose levels and serum proinsulin or insulin levels in mice were determined using Glutest Neo Super (Sanwa Kagaku Kenkyusho, Nagoya, Japan) and proinsulin ELISA kit (10-1232-01; Mercodia) or insulin ELISA kit (M1102; Morinaga, Yokohama, Japan), respectively. For evaluating GSIS from the islets of the C57BL/6J mice after treatment with or without 1 mmol/L imeglimin and 100 nmol/L liraglutide (Novo Nordisk), 10 isolated islets were incubated in Krebs-Ringer bicarbonate buffer (pH 7.4) with 3.9 mmol/L, 11.1 mmol/L, or 16.7 mmol/L glucose for 60 min (preincubation with 3.9 mmol/L glucose with or without imeglimin for 60 min). The islets were extracted with acid ethanol, and the insulin/proinsulin concentration in the assay buffer and the insulin/proinsulin content in the islets were measured using an insulin ELISA kit.

Mouse isolated islets, human islets, or hPSC-derived β-like cell were fixed with 4% paraformaldehyde solution for 15 min, embedded in low-melting-temperature agarose, and fixed again with 10% formalin solution or 4% paraformaldehyde, as described previously (17,18). More than five pancreatic tissue sections from each animal were analyzed after fixation with 10% formalin overnight and paraffin embedding (Tissue-Tek VIP 5 Jr, 19-h program; Sakura Finetek Japan). The sections were immunostained with antibodies to insulin, insulin B (C-12), proinsulin, glucagon, or CHOP (F-168) (antibodies described in Supplementary Table 1). Biotinylated secondary antibodies, a VECTASTAIN Elite ABC Kit, and a DAB Substrate Kit (#PK-6200; Vector Laboratories) were used to examine the sections using bright-field microscopy to determine the β-cell mass. TUNEL staining was performed in vivo or in vitro using the ApopTag In Situ Detection Kit (Chemicon). For the 5-ethynyl-2′-deoxyuridine (EdU) incorporation assay, the islets were stained using the Click-iT Plus EdU Alexa Fluor 488 Imaging Kit (C10637; Thermo Fisher Scientific), and Alexa Fluor 488–, 555–, and 647–conjugated secondary antibodies (Invitrogen) were used for fluorescence microscopy. All the images were acquired using a BZ-9000 microscope (Keyence) or FluoView FV1000-D confocal laser scanning microscope (Olympus) following a published protocol (19). At least 50 islets per mouse (1,100–3,500 nuclei) (in vitro), 30 islets per isolated islet group (1,000–1,500 nuclei) (in vivo), or 25–30 per isolated islet group in human islets or hPSC-derived cells (800–1,500 nuclei) were analyzed for TUNEL or EdU staining.

Islets were isolated from mice as described elsewhere (18). Isolated islets were cultured overnight in RPMI medium (Wako Pure Chemical Industries) containing 3.9, 5.6, 11.1, or 16.7 mmol/L glucose supplemented with 10% FBS and 1% penicillin/streptomycin. Islets were treated with 1 μmol/L thapsigargin, 100 nmol/L liraglutide, 1 μmol/L ISRIB, 0.1 mmol/L imeglimin, 1 mmol/L imeglimin, or 5 μmol/L GADD34 inhibitor (guanabenz acetate, G0456; Tokyo Chemical Industry, Tokyo, Japan).

Twenty isolated islets from C57BL/6J mice were treated with or without 1 mmol/L imeglimin and were incubated using a culture insert for 24 h in RPMI medium containing 5.6 mmol/L glucose, 1 mmol/L pyruvate, and 10% FBS. The islets were then washed with PBS (−) and seeded onto poly-L-lysine–coated (#25988-63-0; Sigma-Aldrich) XF96 cell culture microplates (Agilent Technologies) containing 160 μL/well assay medium. The culture microplates were then centrifuged at 500 rpm for 7 min at room temperature and incubated for 1–2 h at 37°C in a non-CO₂ incubator. XF RPMI medium (Agilent Technologies) containing 5.6 mmol/L glucose (Sigma-Aldrich), 1 mmol/L pyruvate (Nacalai Tesque), and 2 mmol/L L-glutamine (Nacalai Tesque) was used as the assay medium. The oxygen consumption rate (OCR) and the extracellular acidification rate were measured using a Seahorse XF96 Analyzer (Agilent Technologies). Basal respiration was measured for 18 min. Islets then were sequentially exposed to glucose (11.1 mmol/L final concentration), oligomycin (4 μmol/L final concentration), carbonyl cyanide-p-trifluoromethoxyphenyl-hydrazon (FCCP) (1 μmol/L final concentration) and rotenone/antimycin (2.5 μmol/L final concentration) for 30 min. Wave 2.6.0 software (Agilent Technologies) was used to analyze the basal respiration, maximal respiration, ATP production, and proton leak.

Isolated islets (n = 4) from 8-week-old C57BL/6J mice were treated for 24 h with 1 mmol/L imeglimin or vehicle alone (DMSO) in the presence of 11.1 mmol/L glucose in RPMI medium (10% FBS). The cDNA microarray analysis was performed by following a published protocol (19).

Data analysis was performed with GraphPad Prism 8 (GraphPad Software). All data are reported as the mean ± SEM and were analyzed using a Student t test or ANOVA. Differences were considered significant if P < 0.05 or < 0.01.

This study was conducted with the approval of the animal care committees of the Yokohama City University (F-A-19-053) and Gunma University (21-035). All animal procedures were performed in accordance with institutional animal care guidelines and the guidelines of the Animal Care Committee of the Yokohama City University. All the studies and protocols of human islets were approved by the Yokohama City University (B171100025) and Gunma University (HS2020-174) ethics boards.

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

First, we investigated the effects of imeglimin on insulin secretion and β-cell proliferation. The treatment of mouse islets with 1.0 mmol/L imeglimin, but not 0.1 mmol/L imeglimin, significantly increased insulin release in response to glucose and by 2.64-fold under 16.7 mmol/L glucose compared with the no-treatment group, although the increase was lower than that by the glucagon-like peptide 1 receptor agonist liraglutide (Fig. 1A and Supplementary Fig. 1A). Imeglimin had no further effect on insulin secretion under low-glucose conditions. These results were consistent with the finding that imeglimin increased GSIS in vivo and in vitro in rats (10). Both liraglutide and imeglimin had no significant effects on the insulin content in the islets compared with the vehicle control (Supplementary Fig. 1B). Treatment with a combination of liraglutide and 1.0 mmol/L imeglimin did not produce an additional increase in insulin secretion compared with the liraglutide treatment alone (Supplementary Fig. 1C). The expression of complex I of oxidative phosphorylation (OXPHOS) protein was significantly increased after treatment with imeglimin under high-glucose conditions (Fig. 1B). An analysis of mitochondrial respiration revealed that mitochondrial basal respiration, maximal respiration, and ATP production increased significantly, whereas no change in proton leak was observed after treatment with imeglimin in mouse isolated islets under high-glucose conditions (Fig. 1C–F). Treatment with imeglimin significantly increased EdU-incorporated proliferating β-cells in islets under high-glucose (11.1 mmol/L) conditions (Fig. 1G). Polo-like kinase-1 (Plk1), centromere protein-A (Cenpa), and cyclin A2 (Ccna2) are mitogenic genes that affect the cell cycle, and these genes are regulated by forkhead box M1 (FoxM1) (20). Imeglimin upregulated the gene expression of Plk1, Cenpa, and Ccna2 in islets, compared with the effects of the vehicle control, under high-glucose conditions (Fig. 1H). Treatment with imeglimin did not alter the gene expression of cyclin D2 (Ccnd2) significantly, compared with the vehicle control, under high-glucose conditions (Fig. 1H).

Next, we investigated the effects of imeglimin on β-cell apoptosis induced by high-glucose conditions. Both liraglutide and 1.0 mmol/L imeglimin significantly reduced the proportion of TUNEL⁺ apoptotic β-cells in mouse islets subjected to high-glucose conditions for 24 h (Fig. 2A). A gene expression microarray analysis of the islets demonstrated 245 significantly upregulated (greater than twofold) genes and 451 significantly downregulated (<0.5-fold) genes after treatment with imeglimin under high-glucose conditions (Fig. 2B–D). The gene expression microarray analysis also showed that imeglimin increased the expression levels of Hspa1a (8.66-fold, P < 0.01), Hspa1b (6.06-fold, P < 0.01), Stc2 (3.99-fold, P < 0.01), Ddit3 (Chop, 3.14-fold, P < 0.01), activating transcription factor 3 (Atf3) (3.12-fold, P < 0.01), Ern1 (Ire1a, 2.38-fold, P < 0.05), Nr4a2 (2.38-fold, P < 0.01), Ero1l (2.13-fold, P < 0.05), and Ppp1r15a (Gadd34, 2.05-fold, P < 0.01) (Supplementary Table 3), all of which have been identified as ER stress–related genes (21). The gene expression of binding immunoglobulin protein (Bip), protein kinase R-like ER kinase (Perk), Atf4, X-box binding protein 1 (Xbp1), or Atf6 were unaffected by imeglimin in the microarray analysis. The gene expression of semaphorin 6a (Sema6a), a membrane-bound protein (22), was upregulated by imeglimin in the microarray data (3.24-fold, P < 0.01) (Supplementary Table 3). A pathway analysis of the upregulated and downregulated genes indicated that imeglimin possibly modulated the mitogen-activated protein kinase, Kit receptor, insulin signaling, epidermal growth factor receptor 1, or G-protein-coupled receptor pathways (Supplementary Table 4). We confirmed that imeglimin upregulated the gene expression of Chop (Ddit3), Sdf2l1, Atf3, and Sema6a under high-glucose conditions (Fig. 2E and Supplementary Fig. 2). The gene expression of Gadd34 (Ppp1r15a), transmembrane p24 trafficking protein 10 (Tmed10), Ern1, Atf4, Atf6, Stbd1, pancreatic and duodenal homeobox 1 (Pdx1), and nuclear erythroid 2 p45-related factor 2 (Nrf2) were not altered after treatment with imeglimin (Fig. 2E and Supplementary Fig. 2). The protein level of CHOP was significantly increased in islets after imeglimin treatment under 11.1 mmol/L glucose, but not under 3.9 mmol/L glucose (Fig. 2F). We explored the proteins that were likely to interact with imeglimin based on a combination of two- or three-dimensional similarities using the SwissTargetPrediction webtool (Supplementary Table 5) (www.swisstargetprediction.ch) (23,24). We found that sigma-1 receptor (SIGMA1R), a molecular chaperone regulating nuclear signaling and cell survival at the ER-mitochondria interface, was a predicted target of imeglimin (25,26). However, the gene expression of Sigma1r did not differ significantly between the two groups under high-glucose conditions (Supplementary Fig. 2).

Since imeglimin altered the expression of ER stress–related genes in the islets, we focused on the effect of imeglimin on the ER stress–mediated pathway in the context of β-cell survival. Imeglimin significantly protected β-cells from apoptosis induced by thapsigargin, which induces ER stress by inhibiting sarcoplasmic/ER Ca²⁺-ATPase proteins (Fig. 3A). The expression levels of ER-related genes, such as Chop (Ddit3), Gadd34 (Ppp1r15a), Sdf2l1, Atf3, and Hspa1a, were further increased by imeglimin, while the expression of Ern1, Sigma1r, Atf4, and Atf6 tended to be increased under ER stress at 11.1 mmol/L glucose in isolated islets (Fig. 3B and Supplementary Fig. 3A). Our results also showed that imeglimin significantly increased the expression of Nrf2, but not the expression of Ppp1r1a, Eif2a, Sema6a, and Tmed10, under ER stress (Supplementary Fig. 3A). The thioredoxin-interacting protein (Txnip) was reported as a mediator of ER stress–induced cell death through the activation of inflammasome (27). B-cell lymphoma 2 (Bcl2) and B-cell lymphoma-extralarge (Bcl-XL) act as antiapoptotic genes, while Bcl-2–associated X protein (Bax) and Bcl-2 homologous antagonist/killer (Bak) are proapoptotic genes in islet apoptosis (28,29). The ratio of Bax/Bcl2 is thought to be an indicator of cell apoptosis. Imeglimin did not alter the expression of Bcl-XL, Bax, Bak, and Bax/Bcl2 ratio but significantly decreased Txnip expression and significantly increased Bcl2 expression under high glucose (Fig. 3B and Supplementary Fig. 3A). Under 3.9 mmol/L glucose level, CHOP tended to be increased either by thapsigargin or by a combination of thapsigargin and imeglimin (Supplementary Fig. 3B). Notably, imeglimin increased CHOP and decreased the phosphorylation of eIF2α under high glucose in the presence of thapsigargin (Supplementary Fig. 3B). Active 50-kDa ATF6α was not detected under high glucose, and imeglimin did not alter uncleaved 90-kDa ATF6α, SDF2L1, or GADD34 and IRE1α proteins under high glucose (Supplementary Fig. 3B). We previously reported that glucose-mediated signaling suppressed the expression of CHOP (16). Therefore, we also confirmed that imeglimin increased the protein level of CHOP, and this was associated with a decrease in the phosphorylation level of eIF2α under ER stress at a physiological concentration of 5.6 mmol/L glucose (Fig. 3C). The protein level of SDF2L1 showed no significant imeglimin-induced changes under ER stress (Fig. 3C). Since the expression of ATF4, an upstream protein of CHOP, was not altered after imeglimin treatment (Supplementary Fig. 3A), the increased ATF4 level might be returned to the basal level by eIF2α dephosphorylation. Because eIF2α controls protein synthesis through a conserved mechanism, we used ISRIB, which rescues translation by targeting eIF2B, to investigate the effects of decreased phosphorylation of eIF2α (30,31). Under ER stress, ISRIB significantly protected the β-cells from apoptosis, and treatment with a combination of ISRIB and imeglimin did not show a further reduction in the proportion of TUNEL⁺ apoptotic β-cells compared with the imeglimin treatment (Fig. 3D).

Next we investigated the necessity of CHOP for antiapoptotic action of imeglimin in β-cells by using CHOP^−/− mice. Blood glucose levels were similar between WT and CHOP^−/− mice (data not shown). Imeglimin failed to ameliorate the β-cell apoptosis induced by thapsigargin in CHOP-deficient islets (Fig. 4A). We also treated islets with imeglimin in the presence of thapsigargin and guanabenz, a GADD34 inhibitor. Imeglimin failed to protect β-cells under ER stress against apoptosis when cells were treated with GADD34 inhibitor (Fig. 4B).

Akita mice, which carry a heterozygous conformation-altering mutation (Cys96Tyr) in the Ins2 gene, manifest enhanced ER stress and apoptosis of β-cells (32,33). Control WT or Akita 21-day-old mice were fed a standard diet or a diet containing 0.13% imeglimin for 2 weeks (Supplementary Fig. 4A). Body weight gain and both β-cell and α-cell proportions in the islets were similar in all groups (Fig. 5A and Supplementary Fig. 4B). Imeglimin improved hyperglycemia and recovered serum insulin levels in Akita mice with no significant changes compared with control mice (Fig. 5B and C). We also treated the 5-week-old control or Akita mice with or without imeglimin for 3 days. Both serum proinsulin and insulin levels were significantly increased by imeglimin in control mice (Supplementary Fig. 4C). Although the serum proinsulin and insulin levels almost disappeared in Akita mice, the proinsulin/insulin ratio was augmented by imeglimin (Supplementary Fig. 4C). Insulin and proinsulin content in islets were not changed by imeglimin in both genotypes (Supplementary Fig. 4C). Proinsulin gene expression was increased by imeglimin in control mice but not in Akita mice (Supplementary Fig. 4D). Immunohistochemical analyses of proinsulin also showed decreased intensity in Akita mice, and imeglimin seemed to recover the fluorescence intensity (Supplementary Fig. 4E). In the Akita mice, the β-cell mass was significantly decreased, and the ratio of TUNEL⁺ apoptotic β-cells was significantly higher than the corresponding values in WT mice (Fig. 5D and E). Imeglimin restored the β-cell mass and ameliorated β-cell apoptosis in the Akita mice (Fig. 5D and E). Imeglimin significantly increased the mRNA expression of Chop (Ddit3) and tended to increase the expression of Sigma1r and Hspa1a but did not alter the expression of Gadd34 (Ppp1r15a), Sdf2l1, Atf3, Sema6a, Nrf2, Ern1, Mafa, Mafb, Atf4, Bax, Bcl2, Bax/Bcl2, Bak, or Txnip in islets from Akita mice in vivo (Fig. 5F and Supplementary Fig. 5A). On the other hand, the expression of Chop (Ddit3), Sigma1r, Gadd34 (Ppp1r15a), Sdf2l1, Atf3, and Sema6a were significantly increased in isolated islets from Akita mice after treatment with imeglimin compared with those from control WT mice or untreated Akita mice in vitro (Fig. 5G and Supplementary Fig. 5B). Imeglimin did not alter the expressions of Nrf2, Ern1, Mafa, Mafb, Atf4, Atf6, Hspa1a, Bax, Bcl2, Bax/Bcl2 ratio, Bak, or Txnip in Akita-isolated islets under high-glucose conditions (Supplementary Fig. 5B). Immunohistochemical analyses of endocrine pancreas also demonstrated that signal intensity of CHOP was increased in Akita mice treated with imeglimin compared with that in control WT mice or Akita mice without treatment (Fig. 5H).

We next assessed whether imeglimin could protect human islets from apoptosis under high-glucose conditions and found that imeglimin protected β-cells from apoptosis in human islets from donors without diabetes (Fig. 6A). Imeglimin also protected the β-cells from apoptosis under ER stress in human islets from donors without diabetes (Fig. 6B). A combination with ISRIB and imeglimin did not further protect against ER stress–induced β-cell apoptosis in human islets (Fig. 6B). Importantly, the CHOP (DDIT3), GADD34 (PPP1R15A), ATF3, and NRF2 gene expression levels, but not that of SDF2L1 and HSPA1A, were significantly increased by imeglimin treatment, and ATF4 tended to be increased by imeglimin in human islets from donors without diabetes under ER stress at 5.6 mmol/L glucose (Fig. 6C). Imeglimin also seemed to decrease the phosphorylation of eIF2α and BAX protein levels (Supplementary Fig. 6). We further examined the protective effects of imeglimin on β-cell survival in hPSC-derived insulin-positive pseudoislet organoids. Indeed, imeglimin also promoted the survival of insulin-positive β-like cells in hPSC-derived pseudoislets under ER stress (Fig. 6D).

Finally, we examined protein synthesis because imeglimin attenuated eIF2α phosphorylation under ER stress. Protein synthesis assessed by puromycin incorporation was decreased by thapsigargin at 4, 8, and 24 h, but not at 2 h, compared with control (Fig. 7A and B). Imeglimin restored the protein synthesis at all time points, but statistical significance was not achieved at 2 h (Fig. 7A and B). ISRIB also recovered protein synthesis under conditions of ER stress at a level comparable to that by imeglimin (Supplementary Fig. 7A). At the 8-h time point, we also assessed protein synthesis by fluorescence intensity of OPP probe incorporation into MIN6K8 cells. Imeglimin also restored the protein synthesis under ER stress as assessed by OPP probe (Supplementary Fig. 7B). The CHOP level was further increased by imeglimin at 2, 4, and 24 h compared with the thapsigargin-alone group (Fig. 7A and B). The phosphorylation of eIF2α was decreased by imeglimin at the 3- and 8-h time points under ER stress in the MIN6K8 cell line (Supplementary Fig. 7C).

In this research, we showed that imeglimin enhanced glucose-induced insulin secretion, increased β-cell proliferation, and prevented β-cell apoptosis in a glucose-dependent manner. We also demonstrated that imeglimin prevented ER stress–mediated β-cell apoptosis, at least in part, through the modulation of ER homeostasis, including negative feedback of the CHOP/GADD34 pathway, ATF3, or SDF2L1-mediated signaling.

Our results showed that imeglimin upregulated complex I protein, which establishes the hydrogen ion gradient by pumping hydrogen ions into the intermembrane space. Imeglimin also increased ATP production, mitochondrial basal respiration, and maximal respiration under high-glucose conditions in isolated islets. Imeglimin increased GSIS possibly by improving the mitochondrial function and increasing ATP production under high-glucose conditions in isolated islets. In human endothelial cells, imeglimin can decrease reactive oxygen species production and inhibit reverse electron transfer through complex I (12). Furthermore, a previous study demonstrated that complex I activity is required for changes in phospholipid metabolism and mitochondrial-associated membrane (MAM) formation, which are important for autophagy (34).

The MAM integrates many signaling pathways, such as those required for cellular survival, by modulating lipid transport and Ca²⁺ signaling between the ER and mitochondria (35,36). Sigma1r has been implicated in neuroprotection, carcinogenesis, and neuroplasticity and is a Ca²⁺-sensitive and ligand-operated receptor chaperone at MAM; Sigma1r knockdown also reportedly causes apoptosis by compromising IRE1-XBP1 signaling in CHO cells (25). In Akita-isolated islets, imeglimin treatment tended to increase Ern1 mRNA expression under high-glucose conditions. Previous studies have also indicated that Sigma1r plays a key role in modulating retinal stress in a diabetes mouse model (37). In our report, imeglimin significantly upregulated the gene expression of Sigma1r in isolated islets from Akita mice under high-glucose conditions, indicating that imeglimin might regulate ER homeostasis through Sigma1r. Since imeglimin upregulated Ccna2, Plk1, and Cenpa gene expression, an adaptive proliferation pathway through CENP-A/PLK1/FoxM1 signaling might be involved in the β-cell proliferation induced by imeglimin (18). A previous report showed that unfolded protein response (UPR) and subthreshold ER stress evoked the proliferation of β-cells through activation of ATF6 (38). In contrast, acute deletion of insulin enhanced β-cell replication by relieving ER stress (39). Although glucose signaling attenuates ER stress–induced CHOP production under high glucose (19), imeglimin restored the attenuated CHOP expression. It might be possible that imeglimin increased β-cell proliferation through the induction of mild ER stress. Our pathway analysis indicated that imeglimin possibly regulates the MAPK/EGFR1/G-protein–coupled receptor signaling pathway. Further investigation of the effects of imeglimin on these pathways is warranted.

Interestingly, imeglimin significantly upregulated the expression of CHOP but ameliorated β-cell apoptosis. When ER stress is severe and chronic, such as in the presence of persistent hyperglycemia, and cellular homeostasis is not restored, UPR-mediated efforts to correct the protein folding defect fail, and the apoptotic pathway is preferentially activated via the phosphorylated eIF2α/ATF4/CHOP signaling pathway (40,41). Previous studies have indicated that the β-cell protective effects of metformin on lipotoxicity can be partly attributed to the suppression of ER stress–induced CHOP expression (42). We showed that glucokinase activators suppress the expression of CHOP and BAX and protected against β-cell apoptosis under ER stress in an extracellular signal–regulated kinase 1/2–dependent and IRS-2–independent manner (19). Several studies have also shown that the suppression of CHOP evidently averted apoptosis induced by ER stress (32,43,44). In contrast, some reports demonstrated that CHOP-induced GADD34 mediates the translational recovery against ER stress in β-cells by recruiting and activating the catalytic subunit of protein phosphatase 1 (PP1) to promote eIF2α dephosphorylation (45–47). Exendin-4, a glucagon-like peptide 1 receptor agonist, reportedly targeted ATF4 expression to modulate ER stress responses, resulting in the enhancement of β-cell function and insulin biosynthesis (48). In fact, increased Bcl-2, decreased Txnip, and unaltered Bax expression were observed in imeglimin-treated islets, although CHOP expression was upregulated. These results also support our hypothesis that increased CHOP by imeglimin does not induce β-cell apoptosis.

Imeglimin treatment also significantly upregulated the expression of GADD34 in isolated islets under high-glucose conditions or in the presence of ER stress and restored the global protein translation evaluated by incorporation of puromycin into β-cells under ER stress. These results suggest amelioration of translation. The phosphorylation of eIF2α mostly depends on PERK in β-cells, but the rapid dephosphorylation seen upon glucose treatment is mediated by PP1 and GADD34 (49). PP1 activity is negatively regulated by Ppp1r1a (I-1) (49) and positively regulated by GADD34 and CReP (50), which specifically recruit PP1 to eIF2α to affect its dephosphorylation. In our microarray gene expression analysis, we also found that the expression of Ppp1r1a (2.53-fold, P < 0.002) was significantly decreased after treatment with imeglimin under high-glucose conditions. PP1 and GADD34 might synergistically provide more beneficial effects via translational recovery by treatment with imeglimin.

CHOP is also activated in response to severe ER stress in β-cells from Akita mice, which carry the C96Y mutation in the insulin 2 gene (32). We confirmed that CHOP expression was higher in islets from Akita mice compared with islets from the WT control used in this study. Notably, imeglimin further increased the gene expression of CHOP in the islets from the Akita mice and prevented the loss of β-cell mass by protecting against apoptosis, resulting in better glycemic control. GADD34 was also increased after treatment with imeglimin in isolated islets from Akita mice under high-glucose conditions. Thus, imeglimin may facilitate the CHOP/GADD34/PP1 feedback pathway to dephosphorylate eIF2α and restore translational inhibition in vivo. However, translational suppression ought to be unchanged in Akita islets because ER stress is caused by dominant-negative mutant insulin (51). Therefore, besides regulating the translation, imeglimin may directly modulate genes related to the apoptotic pathway. Imeglimin lost the ability to protect β-cells from ER stress in CHOP-deficient islets. GADD34 inhibition blunted amelioration of β-cell apoptosis and an increase in CHOP expression by imeglimin under ER stress. Those results also imply that CHOP and GADD34 are required for the protective effects of imeglimin. Activation of the PERK/eIF2α/ATF4 axis causes the cell adaptation signaling pathway. However, prolonged ER stress oppositely induces apoptotic cell death via the ATF4/CHOP pathway when adaptive signaling fails. In the intestinal epithelial cell, QRICH1 has been demonstrated to be a central mediator of the PERK/eIF2α axis in parallel with ATF4 (52). In our results, imeglimin increased CHOP but not ATF4 expression under ER stress at 24 h. However, ATF expression was transiently upregulated by imeglimin under ER stress at 2 h (data not shown). These data indicate that imeglimin may upregulate ATF4 to induce CHOP in β-cells, but ATF4 expression might be modulated by eIF2α phosphorylation state.

Besides ElF2α, PERK also phosphorylates NRF2, preventing oxidative stress through the induction of heme oxygenase 1 (HO-1), an antioxidant gene (53,54). Previous studies also indicated that NRF2 activity reduced the expression of Sema6a in ischemic neurons (55). In our results, imeglimin treatment increased the gene expression of both Sema6a and Nrf2 in isolated islets under high-glucose conditions after treatment with thapsigargin. Imeglimin might provide more beneficial effects in response to the specific mechanism of ER stress induced by oxidative stress under high-glucose conditions.

In obese and diabetic mice, the restoration of hepatic SDF2L1 expression reportedly ameliorates glucose intolerance and fatty liver by decreasing ER stress (56). In our results, imeglimin increased SDF2L1 expression in islets and tended to increase the expression of TMED10, a major counterpart of SDF2L1. Since imeglimin prevented β-cell apoptosis in ISRIB-treated islets, another pathway in addition to eIF2α, such as SDF2L1, might be involved in the modulation of ER stress by imeglimin. Imeglimin has also been shown to prevent endothelial cell apoptosis induced by oxidative stress through target mitochondrial permeability transition pore opening and cytochrome C release (12). Further research is required for a comprehensive understanding. However, we clearly demonstrated, for the first time, that imeglimin influences β-cell ER homeostasis both in vivo and in vitro.

Since there is substantial evidence for the presence of ER stress in β-cells in both type 1 and type 2 diabetes, the UPR is important for human β-cell function and survival (57). Transplantation of insulin-producing cells is an alternative therapy when endogenous β-cells have already been nearly abolished because of prolonged ER stress. hPSC-derived pancreatic pseudoislets potentially become an inexhaustible source of insulin-producing cells for the treatment of diabetes. We found that imeglimin regulated the β-cell survival not only in human islets but also in hPSC-derived β-like cells under ER stress. Interestingly, the effect of imeglimin to protect against ER stress–mediated cell death showed a more pronounced effect in hPSC-derived pancreatic pseudoislets compared with that in human islets. Pancreatic β-cells are highly sensitive to excessive ER stress and dysregulated eIF2α phosphorylation; imeglimin may have a more beneficial function for cell survival in hPSC-derived islets.

In this study, the effects of imeglimin on ER stress were examined in isolated mouse islets, insulin gene–mutated Akita mice, human cadaveric islets, and hPSC-derived pseudoislets. On the basis of the results, we propose that imeglimin may have a beneficial effect on ER homeostasis, particularly the restoration of protein synthesis (Fig. 7C). These effects are mediated, at least in part, through a negative feedback of the CHOP-GADD34 pathway. However, some limitations should be noted. The puromycin incorporation was normalized to GAPDH, which itself might be altered by translational modifications. In the context of obesity and type 2 diabetes, ER stress is activated in various tissues, such as hypothalamus, liver, muscle, adipose, and β-cell, in both humans and mice (58,59). Under conditions related to insulin resistance, ER stress in the liver acts as a key homeostatic regulator of protein, lipid, and glucose metabolism (60,61). Therefore, future studies are needed to determine whether a similar mechanism involving imeglimin may occur directly in skeletal muscle or liver tissue, and additional physiological or pathophysiological models should be examined to determine the possible effects of imeglimin on ER stress. Most importantly, the efficacy and feasibility of imeglimin in clinical management or regenerative medicine for diabetes should be assessed to open new avenues for improving β-cell survival.

Acknowledgments. The authors thank Professor Patrick MacDonald (Alberta Diabetes Institute, University of Alberta) for providing the human islets from the Alberta Diabetes Institute IsletCore. The authors thank Professor Tetsuro Izumi (Institute for Molecular and Cellular Regulation, Gunma University) for discussion. The authors also thank Fuyumi Murai (Institute for Molecular and Cellular Regulation, Gunma University), Misa Katayama, Mitsuyo Kaji, and Eri Sakamoto (Yokohama City University) for secretarial and technical assistance. J.L. is a fellow of the Rotary Yoneyama Memorial Foundation.

Funding. This work was supported by a MEXT of Japan Grant-in-Aid for Young Scientists (B) (18K16240), a Novo Nordisk Pharma Ltd. Junior Scientist Development Grant provided by the Kanae Foundation for the Promotion of Medical Science, the Suzuken Memorial Foundation, and the Japan Foundation for Applied Enzymology. The Ono Medical Research Foundation, the Kamome Memorial Foundation of Yokohama City University, the Japan IDDM Network, the Takeda Science Foundation, and the Mochida Memorial Foundation for Medical and Pharmaceutical Research (to J.S.). A.K.K.T. and J.S. are supported by the AMED-A*STAR SCICOP Joint Grant Call 192B9002.

Duality of Interest. A.K.K.T. is a cofounder of BetaLife Pte Ltd. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. J.L., R.I., Y.To, T.O., A.S., M.K., K.N., T.T., D.M., and J.S. performed the experiments. J.L., R.I., Y.To, T.O, A.S., M.K., K.N., T.T., D.M., T.K., A.M.J.S., R.S.E.C., A.K.K.T., S.O., Y.Te., and J.S. revised the manuscript. J.L. and J.S. contributed to the discussion and wrote, reviewed, and edited the manuscript. T.K. and A.M.J.S. contributed to the human islet preparation. R.S.E.C. and A.K.K.T. contributed to preparation of the hPSC-derived insulin-positive β-like cells. J.L., Y.To, T.O., M.K., Y.Te., and J.S. analyzed the data. S.O. contributed to CHOP^−/− mice preparation and discussion. J.S. contributed to the experimental design and conception of the study. All authors read and approved the final manuscript. J.S. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented in a plenary session at the 79th Scientific Sessions of the American Diabetes Association, San Francisco, CA, 7–11 June 2019.