MEK/ERK Signaling in β-Cells Bifunctionally Regulates β-Cell Mass and Glucose-Stimulated Insulin Secretion Response to Maintain Glucose Homeostasis

The insulin secretion capacity of pancreas is determined by the amount of β-cells and the function of individual β-cell in secretion of insulin, both of which could be affected by various factors in the pathogenesis of diabetes (1,2). Mitogen-activated protein kinase (MAPK) kinase (MEK)/extracellular signal–regulated kinase (ERK) signaling is involved in many cellular functions including cell growth and survival and cellular differentiation (3). In mammals, the MEK/ERK pathway involves the kinases ERK1 (Mapk3) and ERK2 (Mapk1) and the MAPK kinases MEK1 (Map2k1) and MEK2 (Map2k2). At present, ERK1 and ERK2 are the only known substrates of MEK1 and MEK2 (4), whose functions are known to be redundant, consistent with the high homology in their amino acid sequences (5).

The MEK/ERK signaling pathway is activated by multiple upstream stimuli including insulin (6), and our group, as well as others, has demonstrated a significant role that the insulin signaling pathway plays in β-cells in the pathophysiological regulation of glucose metabolism (7). The MEK/ERK signaling pathway could also be activated by other hormones such as glucagon and growth factors including insulin-like growth factor 1 and platelet-derived growth factor—or by environmental stress as well (6,8,9). Besides, glucose and other nutrients may provoke MEK/ERK signaling activation in cultured β-cell lines, where glucose is proposed to stimulate MEK/ERK signaling in both insulin receptor–dependent and –independent fashions (8,10). While MEK/ERK signaling has been extensively studied in numerous models and its significant roles for β-cell physiology have been proposed (9,11), a systematic in vivo assessment of the roles that MEK/ERK signaling plays in β-cells for glucose metabolism has not been conducted.

Here we have performed in vivo assessments of the significance of MEK/ERK signaling in β-cells, which is downregulated in diabetic mice and in humans with diabetes. Studies using β-cell–specific Mek1/2-deficient mice and phosphoproteomics analyses of their islets and cultured β-cells have revealed that MEK/ERK signaling is required for β-cell mass regulation and plays a role in exocytosis of insulin granules, suggesting that MEK/ERK signaling plays roles in both the quantitative and the qualitative control of pancreatic β-cells.

Autopsied pancreatic tissues from 13 subjects without diabetes and those from 5 subjects with type 2 diabetes were evaluated. Clinical data are briefly summarized in Supplementary Table 1. All patients were autopsied within 5 h of death for avoidance of autolytic changes in pancreas. Excised body of the pancreas was fixed with 10% neutral buffered formalin and embedded in paraffin (FFPE). Snap-frozen pancreas was also prepared for phosphorylated (phospho)-ERK immunofluorescence. The use of pancreatic tissue was approved by the ethics committee of the Hirosaki University Graduate School of Medicine (approval no. 2017-121).

Male db/db mice, misty/misty mice, and C57BL/6J mice were purchased from CLEA Japan. For preparation of DIO mice, C57BL/6J mice were assigned either a high-fat diet (HFD) (HFD32; CLEA Japan) or a low-fat diet (a normal chow diet [NCD], CE-2; CLEA Japan) at 5 weeks of age. Genetically engineered mice used in this study are presented in Table 1. β-Cell–specific Mek1 deletion was conducted by crossing of Mek1-floxed mice with MIP-Cre/ERT mice and intraperitoneal injection of tamoxifen (Tmx) (100 mg/kg body wt for five consecutive days) (Sigma-Aldrich) at 6 weeks of age. For Mek1 and Mek2 abrogation in β-cells, MIP-Cre/ERT mice were crossed with mice homozygous for Mek1-floxed alleles on a systemic Mek2-knockout background, with Cre induction by Tmx at 6 weeks of age (MIP-Cre/ERT⁺/Mek1^f/f/Mek2^ko/ko [Tmx⁺]: hereafter referred to as βMek1/2DKO mice). The offspring were obtained by crossing of MIP-Cre/ERT⁺/Mek1^f/f mice with Mek1^f/f mice or by crossing of MIP-Cre/ERT⁺/Mek1^f/f/Mek2^ko/ko mice with Mek1^f/f/Mek2^ko/ko mice, and these mice were maintained in their respective closed colonies on a 129/SvEv/C57BL/6JJcl mixed background. As it has already been suggested that the expression of human growth hormone (hGH) in MIP-Cre/ERT construct itself may affect β-cell homeostasis (12,13), we used littermates of the same genotype injected with vehicle (corn oil) (MIP-CreERT⁺/Mek1^f/f/Mek2^ko/ko [Tmx⁻]: hereafter referred to as Tmx⁻ controls) as the experimental controls. All the experiments were conducted after >4 weeks’ washout period following Tmx administration. HFD feeding was started 2 weeks after the last intraperitoneal injection of Tmx. Animal care and experimentations were approved by the Animal Care and Use Committee of the National Center for Global Health and Medicine.

MIN6 clone 4 (MINcl4) β-cell line (14) was maintained in DMEM containing 25 mmol/L glucose, 15% FBS, 0.07 mmol/L 2-mercaptoethanol, 100 units/mL penicillin, and 0.05 mg/mL streptomycin in humidified 5% CO₂ at 37°C.

Insulin tolerance tests, glucose tolerance tests (GTTs), and the measurements of plasma insulin levels on a glucose challenge were conducted as previously described (15). Blood glucose levels were measured with Glutest Neo alpha or Glutest Mint (Sanwa Kagaku Kenkyusho). Plasma insulin levels were measured with Morinaga Mouse Insulin ELISA kits and Morinaga Ultra Sensitive Mouse Insulin ELISA kits (Morinaga Institute of Biological Science, Inc.). For estimation of insulin clearance, mice were intraperitoneally injected with human insulin (Eli Lilly) (1.0 units/kg body wt) and blood samples were collected at the indicated time points. Plasma human insulin levels were measured with a human insulin–specific ELISA kit (Mercodia).

Pancreas was perfused with Liberase TL (Roche) solution and subsequently digested for 24 min at 37°C as previously described (15). Islets were picked manually, washed with PBS, and immediately used for further experiments or maintained in recovery media (16).

MINcl4 cells and freshly isolated islets were lysed by sonication in buffer A (17). Other tissues were homogenized in homogenization buffer (17). The samples were prepared by heating at 100°C with Laemmli buffer for 5 min and subjected to immunoblotting with use of antibodies listed in Table 1. Primary antibodies were diluted in 2.5% BSA in Tris-buffered saline with Tween, except for MEK1, MEK2, phospho-MEK1/2, and phospho-AKT antibodies, where Can Get Signal Solution 1 (TOYOBO) was used for dilution. The blots were developed with use of Chemi-Lumi One L (Nacalai Tesque) or BM Chemiluminescence Western Blotting Substrate (POD) (Roche). The band intensity was quantified with Image Lab (Bio-Rad Laboratories).

Several consecutive 4-μm-thick sections were obtained from FFPE specimens. The proportion of β-cell area relative to total pancreatic parenchymal area (β-cell volume density [V_β]) was determined by insulin immunostaining (Abcam) with the point-counting method on at least 2,000 islets in each subject captured with Axio Imager A1 (Carl Zeiss) as previously described (18,19). For phospho-ERK immunofluorescence, 7-μm-thick frozen sections were incubated with primary antibodies for insulin (Abcam) and phospho-ERK (Santa Cruz Biotechnology), followed by incubation with Alexa Fluor 594– and Alexa Fluor 488–conjugated secondary antibodies (Invitrogen). Immunofluorescence images were captured at ×40 magnification with the BZ-X700 (KEYENCE). All morphometric analyses were performed in a blinded manner.

The adult mice were perfused with PBS and 4% paraformaldehyde (PFA) under anesthesia. Resected mouse pancreases were further fixed in 4% PFA overnight and embedded in paraffin. Embryonic pancreases from mouse embryos (embryonic day 14.5) were fixed in 4% PFA and embedded in paraffin. Paraffin sections (3 μm) were deparaffined and autoclaved in 0.01 mol/L citrate buffer (pH 6.0) for antigen retrieval. For MEK1, NGN3, and SOX9 staining, sections were additionally incubated in 0.1% Triton X-100 for permeabilization after antigen retrieval. Sections were stained with antibodies listed in Table 1 and subsequently incubated with Alexa Fluor 594– and Alexa Fluor 488–conjugated secondary antibodies (Invitrogen), except for the MEK1 staining, where EnVision+ System- HRP Labeled Polymer Anti-rabbit (Dako) was used as a secondary antibody and DAB Peroxidase Substrate Kit ImmPACT (Vector Labs) was used for visualization. Counterstain was conducted with hematoxylin-eosin. The TUNEL staining was performed with an In Situ Cell Death Detection Kit (Roche). Images were obtained with the BZ-X700 (KEYENCE) and analyzed with Analysis Application Hybrid Cell Count (KEYENCE).

Total RNA was isolated from isolated islets with TRIZOL reagent (Invitrogen). Sample preparation and RNA sequencing were performed as previously described (20). FASTQ files were imported to CLC Genomics Workbench (CLC-GW) (version 10.1.1; QIAGEN). Reads were mapped to the mouse reference genome (mm10) and quantified for the annotated 49,585-gene set provided by CLC-GW.

Total RNA was extracted by RNeasy Mini Kit (QIAGEN), followed by reverse transcription with a high-capacity cDNA reverse transcription kit (Applied Biosystems [ABI]). Quantitative RT-PCR was performed by ABI StepOnePlus with use of KAPA PROBE FAST qPCR Master Mix (2X) ABI Prism (Kapa Biosystems). Gene expression levels were normalized to the expression of cyclophilin A (CypA) in each sample. The primer sequences for CypA are as follows: F, GGTCCTGGCATCTTGTCCAT, and R, CAGTCTTGGCAGTGCAGATAAAA. Other primers and probes, purchased from ABI and Integrated DNA Technologies, are listed in Table 1.

Two-photon imaging of the islets was performed with an inverted laser-scanning microscope (IX81; Olympus) with 60X oil immersion objective lens (1.2 numerical aperture) at the wavelength of 830 nm as previously described (21). Glucose-stimulated exocytotic events of insulin granules were detected with extracellular polar tracer, sulforhodamine B (Thermo Fisher Scientific), counted in a region of interest (typically 3,200–4,500 μm²) and normalized to an area of 800 μm². The acetoxymethyl esters of the Ca²⁺ indicators, Fura-2 (Thermo Fisher Scientific), as well as those of the caged-Ca²⁺ compound NP-EGTA (Thermo Fisher Scientific) were loaded to the islets as previously described (22). The intensity of Fura-2 poststimulation fluorescence (F) was analyzed in the cytoplasm and normalized by the resting fluorescence (F₀). Photolysis of NP-EGTA was induced with a brief flash (1.0 s) of a mercury lamp (model IX-RFC; Olympus).

Isolated islets were placed in recovery media for 2 h in humidified 5% CO₂ at 37°C (16). The islets were lysed by sonication in buffer A. Islets (500–1,000) were pooled to make one sample. MINcl4 cells were maintained in Krebs-Ringer bicarbonate buffer (KRBB) (15) with 0.2% BSA supplemented with 2.8 mmol/L glucose for 2 h at 37°C. Thirty minutes before glucose stimulation, U0126 (10 μmol/L) or vehicle (0.1% DMSO) was added to the buffer. The cells were then stimulated for 5 min by 2.8 mmol/L or 22.2 mmol/L glucose with U0126 (10 μmol/L) or vehicle (0.1% DMSO). After glucose stimulation, MINcl4 cells were lysed in buffer A. Lysates of islets or MINcl4 cells were boiled, sonicated, and subjected to reduction, alkylation, and digestion with trypsin and Lys-C (23). Phosphopeptides were enriched by a Fe-IMAC/C18 StageTip and labeled by TMT 10plex reagent. Labeled phosphopeptides were fractionated into seven fractions (24). The phosphoproteomics analysis was conducted with Q Exactive Plus (Thermo Fisher Scientific). Phosphopeptide identification was carried out with MaxQuant. The settings of liquid chromatography–tandem mass spectrometry and database search were performed according to the methodology of a previous phosphoproteomics study (23).

Statistical significance was determined by an unpaired two-tailed Student t test, one-way ANOVA with the Tukey multiple comparisons test, and two-way ANOVA with the Sidak multiple comparisons test and with the Tukey multiple comparisons test, with use of GraphPad Prism, version 7 (GraphPad).

RNA-sequencing data generated and analyzed during the current study are available in DDBJ repository with accession code E-GEAD-385. Phosphoproteomics data generated and analyzed during the current study are available in jPOSTrepo and can be accessed with accession code JPST001021. No applicable resources were generated or analyzed during the current study.

We first investigated the two different obese animal models, namely, db/db mice (25) and DIO mice (26). The db/db mice showed marked hyperinsulinemia at 6 weeks of age compared with the lean misty/misty control mice, and thereafter their circulating insulin levels declined and hyperglycemia developed (27) (Fig. 1A–C), while the plasma insulin levels continuously increased in the DIO mice, with their hyperglycemia being only mild for up to 20 weeks of HFD feeding (Fig. 1F–H). Interestingly, unlike the levels of Akt phosphorylation, which were significantly lower in the islets of both of the obese models compared with their controls (15) (Fig. 1D and I and Supplementary Fig. 1A and B), the levels of MEK1/2 and ERK1/2 phosphorylation were significantly lower in the db/db mouse islets (Fig. 1D and E and Supplementary Fig. 1A), while those in the DIO mouse islets were maintained or even higher compared with their controls (Fig. 1I and J and Supplementary Fig. 1B). These results prompted us to hypothesize that maintained MEK/ERK signaling would be necessary for the compensatory β-cell adaptation in insulin resistance.

On the other hand, a cohort of patients with nonobese type 2 diabetes (Supplementary Table 1), showing a smaller proportion of β-cell area relative to the total pancreatic parenchymal area (V_β) (19) (Fig. 1K), exhibited a significantly lower proportion of phospho-ERK1/2⁺ cells in insulin⁺ cells compared with the subjects without diabetes (Fig. 1L and Supplementary Fig. 1C), with a strong correlation observed between V_β and the proportion of phospho-ERK1/2⁺ cells in insulin⁺ cells (Fig. 1M). These data suggested a possibility that MEK/ERK signaling could be involved in the regulation of β-cell mass under nondiabetic conditions and also in the pathogenesis of a certain type of diabetes in humans.

To assess the role of MEK/ERK signaling in vivo, we created a mouse model with defective pancreatic β-cell MEK/ERK signaling by targeting MEK1 and MEK2, the upstream kinases responsible for ERK1 and ERK2 phosphorylation (28). While the mice with β-cell–specific Mek1 deletion or those with systemic Mek2 deletion did not exhibit complete abrogation of ERK1/2 phosphorylation in the islets (Supplementary Fig. 2A and B), consistent with the functional redundancy of MEK1 and MEK2 as previously reported (5), βMek1/2DKO mice showed robust reductions in the levels of MEK1 protein expression as well as those of MEK1/2 and ERK1/2 phosphorylation in the islets compared with the Tmx⁻ control mice (Fig. 2A and B and Supplementary Fig. 2C). The levels of MEK1 protein expression remained unchanged in the liver, muscle, epididymal white adipose tissue, and hypothalamus (Supplementary Fig. 2D).

To clarify the significance of MEK/ERK signaling in β-cells in diet-induced obesity, we fed βMek1/2DKO mice with HFD and investigated their metabolic phenotypes. While the body weight gain was similar between the DIO βMek1/2DKO mice and the Tmx⁻ control DIO mice, the DIO βMek1/2DKO mice showed higher blood glucose levels with ad libitum status (Fig. 2C and D) with significantly lower plasma insulin levels (Fig. 2E) than the Tmx⁻ controls at 12 weeks or 18 weeks of HFD feeding. The DIO βMek1/2DKO mice showed higher plasma glucose levels during a GTT than the Tmx⁻ controls, as they developed obesity after 19 weeks of HFD feeding (Fig. 2F–H). The DIO βMek1/2DKO mice also showed a moderate but significant decrease in plasma insulin levels on a glucose challenge than the Tmx⁻ controls at 22–23 weeks of HFD feeding (Fig. 2I), while there was no difference between the groups in their blood glucose levels during insulin tolerance tests (Supplementary Fig. 2E–G). We found no metabolic differences among Mek2^ko/ko, Mek2^wt/ko, and Mek2^wt/wt (Supplementary Fig. 3A–I) or between the Mek1^f/f/Mek2^ko/ko mice injected with Tmx and those injected with vehicle (corn oil) (Supplementary Fig. 4A–I) under HFD conditions. Taken together, these data indicated that Mek1/2 deletion in β-cells exacerbated glucose intolerance with lower insulin secretion under HFD feeding conditions.

Next, we histologically assessed the islets of DIO βMek1/2DKO mice. In parallel with the results of GTTs, we observed a significantly smaller proportion of insulin⁺ islet area in the total pancreatic area in the βMek1/2DKO mice at 18 and 23 weeks of HFD feeding than in the DIO Tmx⁻ control mice (Fig. 3A), which was accounted for by the lower number of large islets their size >5 × 10⁴ μm² when normalized to the total pancreatic area (Fig. 3B). Furthermore, the proportion of Ki67⁺ cells in insulin⁺ cells was lower, while the proportion of TUNEL⁺ cells in insulin⁺ cells also tended to be lower, in the islets of DIO βMek1/2DKO mice than in those of DIO Tmx⁻ control animals (Fig. 3C and D). We found neither Ngn3⁺ cells (Ngn3^hi) in the islets as an indicator of de-differentiation of β-cells (29) in DIO βMek1/2DKO mice at 18 weeks of HFD feeding nor significant difference in Ngn3 mRNA expression levels between the DIO βMek1/2DKO and DIO Tmx⁻ control mouse islets (Fig. 3E and F and Supplementary Fig. 5A). The islets of DIO βMek1/2DKO mice showed a significantly higher ratio of glucagon (Gcg)⁺ area to insulin⁺ area and tended to show higher levels of Gcg and Aristaless-related homeobox (Arx) mRNA expression than those of Tmx⁻ control islets (Supplementary Fig. 5B and C). However, the extent of increase in these α-cell markers was reciprocal to the extent of reduction in the insulin⁺ islet area, suggesting that transdifferentiation of β-cells into α-cells was unlikely to account for the mechanism of lower insulin⁺ islet area observed in the DIO βMek1/2DKO mouse islets. Thus, these data indicated that the lower insulin secretion in the DIO βMek1/2DKO mice was associated with smaller β-cell mass accompanied by decreased β-cell proliferation compared with that in the Tmx⁻ control mice. Consistently with the decreased β-cell proliferation in the DIO βMek1/2DKO mice, RNA-sequencing and real-time PCR analyses revealed that the expression of cyclin D1 (Ccnd1) was lower, and that of cyclin-dependent kinase inhibitor 1a (Cdkn1a, p21) higher, in DIO βMek1/2DKO islets than in those of Tmx⁻ controls after as early as 3 weeks of HFD feeding (Fig. 3G and H). These data suggested that Mek1/2 abrogation in β-cells led to dysregulated expressions of cell cycle regulatory genes.

The levels of insulin secretion are determined by both the total amount of pancreatic β-cells and the insulin-secretion capacity of individual pancreatic β-cells (1,2). To assess the insulin-secretion capacity of β-cells in DIO βMek1/2DKO mice, we conducted two-photon excitation imaging analyses of the islets, which allowed for assessment of insulin granule exocytotic events and Ca²⁺ influx on a single-cell basis (21,30). There was a trend toward lower number of exocytotic events associated with high-glucose stimulation in the DIO βMek1/2DKO mouse islets within 10 min (Fig. 4A), suggesting that the β-cells of the DIO βMek1/2DKO mice may have impairment in their glucose-stimulated insulin secretion (GSIS) response. The insulin content normalized by the number of islets or islet DNA contents were comparable between the DIO βMek1/2DKO islets and Tmx⁻ control islets (Supplementary Fig. 6A–C). The assessments of individual β-cells for Ca²⁺ influx revealed that the levels of Ca²⁺ influx on glucose stimulation were comparable between the β-cells of DIO βMek1/2DKO mice and those of Tmx⁻ controls (Fig. 4B), while the same extent of a large and rapid increase in Ca²⁺ influx induced by photolysis of a caged-Ca²⁺ compound (Fig. 4C) (22) led to a markedly lower number of exocytotic events in the βMek1/2DKO mouse islets in comparison with Tmx⁻ controls (Fig. 4D).

Interestingly, the lower number of exocytotic events was clearly observed after glucose stimulation in the β-cells of lean βMek1/2DKO mice fed NCD (Fig. 5A). The insulin contents normalized by the number of islets or islet DNA contents were comparable between the DIO βMek1/2DKO islets and Tmx⁻ control islets (Supplementary Fig. 6D–F). Also, the βMek1/2DKO and Tmx⁻ control mouse islets were comparable in their ATP production on high-glucose stimulation (Fig. 5B). The glucose-stimulated Ca²⁺ influx was slightly lower in the β-cells of βMek1/2DKO mice than that in Tmx⁻ control mice (Fig. 5C), while the provoked Ca²⁺ influx again led to markedly fewer events of exocytosis in βMek1/2DKO mouse islets than in those of Tmx⁻ controls (Fig. 5D and E). The NCD-fed βMek1/2DKO mice also showed a moderate but significant decrease in plasma insulin levels on a glucose challenge at 12 weeks of age than the Tmx⁻ control mice, while otherwise no metabolic differences were observed between the groups (Fig. 5F–K and Supplementary Fig. 7A). Plasma human insulin levels after human insulin injection were comparable between the groups, suggesting that insulin clearance was not affected in βMek1/2DKO mice (Supplementary Fig. 7B). Immunohistochemistry analyses showed that the islet areas were comparable between βMek1/2DKO mice and Tmx⁻ controls at a younger age and were different only at 23 weeks of age, with a lower distribution of larger islets seen in βMek1/2DKO mice than in Tmx⁻ controls (Supplementary Fig. 7C and D). We again ruled out the possible impact of Mek2 deletion alone or Tmx injection per se on the difference in plasma insulin levels on a glucose challenge (Supplementary Fig. 7E–O). Collectively these data suggested that MEK/ERK signaling could play significant roles in the regulation of insulin exocytosis in pancreatic β-cells.

We next explored the molecular mechanism(s) whereby Mek1/2 abrogation led to a lower number of exocytotic events in β-cells. First, in order to assess the downstream targets of MEK/ERK signaling in β-cells, we compared phosphoproteome of βMek1/2DKO mouse islets versus those of Tmx⁻ control mice fed NCD (Fig. 6A). We identified 10,886 phosphosites in 3,718 proteins (Fig. 6B and C). In gene ontology (GO) analyses (31), the genes with the GO term “insulin secretion” were highly enriched in the 190 molecules harboring the 241 sites that were phosphorylated at significantly lower levels in βMek1/2DKO mouse islets than in the Tmx⁻ control mouse islets (P value <0.05, fold change < −1.5) (Fig. 6D), with the phosphorylation sites of exocytosis-regulating proteins SNAP25, PCLO, ITPR3, RAB37, SYTL4, and RIMS2 being significantly less phosphorylated (Fig. 6E). Gene expression analysis showed no differences between the groups in these molecules and those established to be involved in exocytosis of insulin granules such as STX1A, VAMP2, or a dock protein, RAB27A (32) (Supplementary Fig. 8A). In addition, the genes with GO terms related to cytoskeleton were highly enriched in the less phosphorylated proteins in βMek1/2DKO mouse islets compared with Tmx⁻ controls (Fig. 6D and Supplementary Fig. 8B), among which molecules, such as MARCKS, MAPT, and MAP2, have been reported to be involved in exocytosis of insulin or other secretory granules (33–35). Thus, these results suggested the possibility that the MEK/ERK pathway regulated insulin secretion through dynamic changes in phosphorylation status of exocytosis-regulating molecules including cytoskeleton–related molecules in pancreatic β-cells.

Finally, we attempted to elucidate the possible upstream signaling responsible for MEK/ERK activation in β-cells. According to previous reports, glucose stimulation of β-cells causes phosphorylation of ERK1/2 in minutes (10,36), as we confirmed in the MINcl4 β-cell line (Fig. 7A). Moreover, pretreatment of MINcl4 cells with MEK inhibitor U0126 led to a decrease in their GSIS response (Fig. 7B), as already reported (37). Collectively with the results from our in vivo models, these data prompted us to hypothesize that glucose stimulation activated MEK/ERK signaling, which in turn caused phosphorylation of GSIS-related molecules and thereby impinged on GSIS response. To test the hypothesis, we compared phosphoproteomic profiles of MINcl4 cells cultured in low concentration of glucose (LG) with those stimulated with 22.2 mmol/L glucose for 5 min in absence or presence of U0126 (HG or HGi) (Fig. 7C). We identified 12,168 phosphosites in 3,589 proteins in total (Fig. 7D). GO analyses of the 282 molecules harboring the 373 sites that were phosphorylated at significantly higher levels in HG condition than in LG condition (P value <0.05, fold change >1.5) demonstrated that the MAPK signaling pathway was a main signaling pathway regulated by acute glucose stimulation (Fig. 7E and Supplementary Fig. 9A). A total of 155 phosphosites were phosphorylated at significantly higher levels in HG condition in comparison with both LG and HGi conditions (P value <0.05) (Fig. 7F). We considered them as the phosphosites regulated by glucose stimulation in a MEK/ERK signaling–dependent manner, which included ERK1^T203, ERK1^Y205, ERK2^T183, and ERK2^Y185 residues (Fig. 7G). GO analyses of the 112 molecules harboring these 155 phosphosites showed high enrichment of GO terms related to cytoskeleton (Fig. 7F and Supplementary Fig. 9B). Between the phosphosites showing less phosphorylation in the βMek1/2DKO islets than in the Tmx⁻ control mouse islets (P value < 0.05) and those showing higher phosphorylation in HG condition in comparison with LG and HGi conditions in MINcl4 cell (P value < 0.05), we found in common 21 sites of 15 proteins (Fig. 7H and Supplementary Table 2). Collectively, these data suggested that glucose was a potential stimulant causing MEK/ERK-mediated phospholyrations in β-cells, which could regulate GSIS response.

To date, several genetic factors have been suggested to cause maladaptation of β-cells leading to insufficient insulin production in the pathogenesis of diabetes, while its exact molecular mechanisms remain unclear (2).

We began by investigating the islets of murine type 2 diabetic models and found reductions in the levels of ERK1/2 phosphorylation in the islets of db/db mice, which were associated with a progressive decline in their plasma insulin levels as they grew. In contrast, the levels of ERK1/2 phosphorylation were not reduced, or remained even higher, in the islets of DIO mice, where only mild hyperglycemia was observed with relatively maintained plasma insulin levels. Of note, the mechanism causing the observed differences in phosphorylation levels of ERK1/2 and MEK1/2 in the islets of different mouse models is unclear, while there could be composite mechanisms reflected in the differences in the levels of responsible humoral factors, potentially including insulin, or in the cellular sensitivity to these stimuli, leading to the altered activation status of the signaling cascade.

While we found that MEK/ERK signaling is necessary for adaptive hyperplasia of β-cells in obesity, it is reported that obesity in humans only leads to a mild increase in β-cell mass compared with mouse models (1). Interestingly, however, a cohort of nonobese patients with type 2 diabetes showed lower proportion of phospho-ERK1/2⁺ cells in insulin⁺ cells with lower V_β compared with subjects without diabetes. As βMek1/2DKO mice on NCD showed significantly lower islet area at 23 weeks of age, it is plausible that MEK/ERK signaling is required not only for adaptive β-cell hyperplasia in obesity but also for maintenance of β-cell mass under lean conditions, where the defects of MEK/ERK signaling could lead to hyperglycemia due to insulin insufficiency. In support of the MEK/ERK involvements in human diabetes pathogenesis, several genome-wide association studies showed an association between type 2 diabetes and a single nucleotide polymorphism (SNP) (rs5945326) near dual-specificity phosphatase 9 (DUSP9), an ERK-selective cytoplasmic phosphatase, in different ancestries (38–40). Other SNPs, SNP rs1894299 on DUSP9 and SNP rs9648716 on BRAF, a kinase upstream for MEK1/2, are also reported to be associated with type 2 diabetes (41,42).

To determine the functional significance of MEK/ERK signaling in β-cells, we created a murine model with Mek1/2 abrogation and investigated its metabolic phenotypes. Our analyses collectively underscored the importance of MEK/ERK signaling–mediated cell proliferation in compensatory β-cell expansion under insulin-resistant conditions, which is plausible in light of the well-established roles that MEK/ERK signaling plays in cell proliferation (3). On the other hand, our data also indicate a significant role that MEK/ERK signaling plays in GSIS response, as revealed by the lower plasma insulin levels on a glucose challenge with comparable insulin sensitivity and the lower numbers of exocytotic events in two-photon excitation microscopy analyses in the islets of βMek1/2DKO mice. Of note, our analyses indicated impairment in the regulation of exocytosis machinery downstream of the Ca²⁺ influx in the β-cells of βMek1/2DKO mice. It has been suggested that reorganization of the actin cytoskeletal network is required for insulin granules to properly access cellular membrane in both the first and second phases of GSIS response (43,44). Our phosphoproteomics analyses of the isolated islets from βMek1/2DKO mice showed that the levels of phosphorylation in the proteins related to exocytosis and cytoskeleton were altered on Mek1/2 abrogation, suggesting that MEK/ERK signaling regulates these molecules by altering their phosphorylation levels and thus is involved in proper GSIS response. Additionally, as dramatic changes in cytoskeletal organization are also known to occur during mitotic processes (45), the MEK/ERK signaling–mediated cytoskeletal modifications could also be a part of mechanisms for cell proliferation regulated by MEK/ERK signaling.

We noticed that 21 sites with significantly higher phosphorylation levels in HG condition than in LG and HGi conditions were also identified in the less phosphorylated phosphosites in βMek1/2DKO islets, whereas the overlap between the phosphosites identified in these two models was not large. These data could be interpreted to indicate that upstream inputs other than glucose could have been involved in the regulation of altered phosphosites detected in the islets, including those derived from the non–β-cell components producing humoral factors in a paracrine manner. Further analyses are needed to fully elucidate the roles that MEK/ERK signaling plays in exocytosis regulation, which will include determining the upstream stimulants as well as the exact substrates of MEK/ERK signaling in β-cells during GSIS response and the significance of such posttranscriptional modifications for the functions of the exocytosis machineries.

Of note, while our experimental design properly controlled the effects of hGH expression from the MIP-Cre/ERT construct (12,13), we could not strictly rule out the possibility that the transient nuclear translocation of Cre/ERT protein could have modified the phenotypes of βMek1/2DKO mice, which is one of the limitations in our study.

Finally, from a therapeutic perspective, our results suggest that activation of the MEK/ERK pathway in pancreatic β-cells may have therapeutic potential for diabetes, in that it is expected to increase β-cell mass and enhance GSIS response at the same time. However, it has been reported that pharmacological inhibition, rather than activation, of MEKs improves glucose homeostasis in diabetic mouse models through control of peroxisome proliferator–activated receptor γ (PPARγ) in adipose tissues, thereby ameliorating insulin resistance (46). Besides, the RAS/RAF/MEK/ERK signaling is one of the major pathways known to be activated through mutations in cancer (47,48). Thus, simply activating MEK/ERK signaling in a systemic manner may have deleterious consequences. Instead, identifying and targeting the responsible mechanisms whereby ERK becomes less phosphorylated in the islets under diabetic conditions may represent a more promising strategy for the development of antidiabetes therapeutics.

Taken together, our data provide evidence that MEK/ERK signaling has bifunctional roles in β-cells in maintaining glucose homeostasis, suggesting that defects in MEK/ERK signaling could contribute to β-cell maladaptation in the pathogenesis of diabetes. Further research targeting the MEK/ERK pathway is required to formulate optimal therapeutic strategies for diabetes.

Acknowledgments. The authors thank Dr. H. Ishihara (Nihon University School of Medicine) for providing MIN6 clone 4 cell lines. The authors also greatly appreciate the technical support of Dr. W. Nishimura (International University of Health and Welfare Graduate School) in delivering embryonic pancreases. The authors’ thanks are also due to M. Nakano and T. Oyama for assistance in immunohistochemistry as well as F. Takahashi, Y. Masaki, and R. Honma for technical assistance.

Funding. This work was supported by Japan Society for the Promotion of Science KAKENHI grants JP16K19067 (to Y.M.I.) and JP19K16547 (to Y.M.I.) and National Center for Global Health and Medicine grants 29-1021 (to Y.M.I.) and 20A1011 (to Y.M.I.).

Duality of Interest. M.M. is supported by LSI Medience Corporation through the cross-appointment system of University of Tsukuba. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. Y.M.I. organized and performed the experiments, wrote the manuscript, and secured funding. K.U. conceptualized, organized, and wrote the manuscript and secured funding. M.A. organized and wrote the manuscript. N.K., S.O., S.T., H.S., and K.S. performed the experiments. H.K. provided expertise and analyzed the study data. J.A. performed phosphoproteomics analyses. M.M. performed transcriptomic analyses. H.M. provided human data. Y.M. and N.T. performed two-photon excitation imaging analyses. J.C. provided mouse strains. All authors reviewed, edited, and approved the manuscript. K.U. 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.