X-Box Binding Protein 1 Is Essential for Insulin Regulation of Pancreatic α-Cell Function

We used male mice for all experiments. Mice were housed in pathogen-free facilities and maintained on a 12-h light/dark cycle at the Foster Biomedical Research Laboratory of Brandeis University in Waltham, Massachusetts. All protocols were approved by the Brandeis University Institutional Animal Care and Use Committee and were in accordance with National Institutes of Health (NIH) guidelines. Blood glucose was monitored with a Glucometer (Elite, Bayer), plasma insulin by ELISA (Crystal Chem, Downers Grove, IL), plasma glucagon by radioimmunoprecipitation assay (RIA; Linco, St. Charles, MO), and plasma glucagon-like peptide 1 (GLP-1) by ELISA (Linco). Glucose and insulin tolerance tests were performed as described previously (26). For the pyruvate challenge test, blood glucose was monitored at 15, 30, 60, and 120 min after an intraperitoneal pyruvate injection (2 g/kg body weight).

Islets were isolated from 6-month-old mice, as described previously (26). After 24-h culture in 7 mmol/L glucose, islets were used in secretion assays, as reported earlier (27). Islets were preincubated at 37°C for 30 min in Krebs-Ringer buffer (KRB) supplemented with 5.5 mmol/L glucose, transferred to 1.5-mL tubes (15 islets per sample), and incubated in 500 μL KRB with 2.2 or 16.7 mmol/L glucose for 1 h at 37°C. At the end of the incubation period, aliquots from each sample were stored at −20°C for glucagon and insulin assays. Islet DNA, glucagon, and insulin were extracted from aliquots of islets, as reported earlier (27). Glucagon and insulin were measured by RIA or ELISA, as described above.

Mice were anesthetized, and the pancreas was rapidly dissected and processed, as described previously (26), and immunostained for insulin (Abcam), glucagon (Sigma-Aldrich), and somatostatin (Abcam). Photomicrographs were obtained with a charge-coupled device camera, and the β-cell, α-cell, δ-cell, and total pancreatic area were estimated using ImageJ software. For transmission electron microscopy, islets were processed by the Joslin Advanced Microscopy Core (DRC) Facility, as reported previously (27).

Glucagon-secreting αTC6 cells were cultured as reported previously (6), and experiments were performed using 80–90% confluent cells.

Lentiviral vector plasmid for murine XBP1 short hairpin RNA (shRNA) and control nonsilencing (NS) shRNA were purchased from Open Biosystems. Mouse complete spliced XBP1 open reading frame was amplified from αTC6 cDNA, after thapsigargin treatment, using PCR and cloned into pCDH-CMV-MCS-EF1-Puro lentiviral vector (System Biosciences) for overexpression, and SparQ dual promoter lentiviral vector (System Biosciences) for inducible mouse spliced XBP1. We used SparQTM Cumate Switch Inducible Expression System (System Biosciences) for induction of spliced XBP1. Lentivirus particles were generated by following the manufacturer’s recommendation (Open Biosystems). αTC6 cells were plated 24 h before viral infection and then incubated with lentivirus for 24 h. For generating stable cell lines, we began selection with 4 μg/mL of puromycin 48 h after infection and maintained selection for ∼14 days before experiments. We generated three separate stable cell lines in each group.

Cells were lysed in RIPA buffer, as reported earlier, and total protein concentration was determined using a bicinchoninic acid assay (Pierce) (27). Samples were resuspended in reducing SDS-PAGE sample buffer, boiled, and resolved by SDS-PAGE. Proteins were subsequently transferred onto polyvinylidene difluoride membranes, blocked in blocking buffer (Thermo Scientific), and incubated with primary antibodies overnight at 4°C. XBP1, ATF6, TATA binding protein (TBP), β-actin, C/EBP homologous protein (CHOP), and IR antibodies were from Santa Cruz. Binding immunoglobulin protein (BiP), IRE1α, pPERK, pJNK, JNK, p-insulin receptor substrate 1 (IRS1) (pS307), pAkt (pT308), pAkt (pS473), total Akt, forkhead box O1 (FoxO1), and cleaved caspase 3 antibodies were from Cell Signaling. pIRE1α antibody is from Novus Biologicals. PERK antibody was from Rockland. pIR (pY972) and pIRS1 (pY896) antibodies are from Invitrogen. IRS1 antibody was from Millipore. Relative protein amounts were estimated by ImageJ software.

Cells were preincubated for 30 min at 37°C in KRB with 25 mmol/L glucose. Subsequently, the cells were treated with 1,000 μL KRB containing 0.5, 5.5, or 25 mmol/L glucose in a 12-well plate for 1 h. At the end of the incubation, aliquots from each sample were used for measurement of glucagon by RIA. Then cells were lysed for DNA extraction.

Cells on coverslips were incubated for 40 min with 2 μmol/L Fura-2 acetoxymethyl ester at 37°C and washed by KRB with 0.5 mmol/L glucose. Experiments were performed as described earlier (28).

RNA was extracted from cells using RNeasy Mini Kit (QIAGEN), and 1 μg RNA was used for a reverse transcription step using the high-capacity cDNA Archive Kit (Applied Biosystems). cDNA was analyzed and amplified using the ABI 7900HT system (Applied Biosystems). TBP was used as an internal control. Primers for glucagon: 5′-TGAATTTGAGAGGCATGCTG-3′ (forward) and 5′-TGGTGCTCATCTCGTCAGAG-3′ (reverse) and for TBP: 5′-ACCCTTCACCAATGACTCCTATG-3′ (forward) and 5′-ATGATGACTGCAGCAAATCGC-3′ (reverse).

All data are expressed as means ± SE and were analyzed with an unpaired two-tailed Student t test or ANOVA and post hoc tests as appropriate. Differences were considered significant at P < 0.05.

α-Cell–specific XBP1 knockout (αXBPKO) mice were created by crossing mice carrying the Xbp1^flox allele (25,29) with mice expressing Cre recombinase driven by the glucagon promoter (6,30). The efficiency of recombination by Glu-Cre was consistent with our previous report (6), with ∼80% of α-cells expressing β-gal (Fig. 1A). αXBPKO mice were born normally and did not exhibit abnormalities in the postnatal period. Body weight, nonfasting blood glucose, insulin, and glucagon levels did not show significant differences between groups until age 6 months (Fig. 1B–E), when αXBPKO mice exhibited glucose intolerance (Fig. 1F) and insulin resistance (Fig. 1G). An intraperitoneal injection of glucose failed to suppress glucagon secretion in αXBPKO mice compared with a significant suppression in controls (Fig. 1H). Hepatic glucose production after pyruvate injection was also higher in αXBPKO mice compared with controls (Fig. 1I). Insulin and plasma GLP-1 levels were not significantly different between groups (Fig. 1J and K).

To investigate the potential defects in secretory function, independent of systemic signals, we examined hormone secretion from freshly isolated islets. As expected, exposure of control islets to high glucose (16.7 mmol/L) suppressed glucagon release compared with low glucose (2.2 mmol/L), whereas glucagon secretion from αXBPKO islets was unaffected by high glucose levels (Fig. 1L). However, both groups exhibited a similar insulin secretory response to high glucose stimulation (Fig. 1M), and the glucagon and insulin content of isolated islets were not significantly different (Fig. 1N and O). Taken together, these data suggest that XBP1 deficiency in α-cells leads to dysregulation of glucagon secretion and altered glucose homeostasis.

Histomorphometric analyses of pancreatic sections in 6-month-old αXBPKO mice showed normal islet architecture (Fig. 2A and B), and no significant differences were evident in the relative areas of insulin, glucagon, or somatostatin-positive cells between groups (Fig. 2C–E). Evaluation of apoptosis by TUNEL staining and BrdU incorporation into α-cells revealed no differences between groups (data not shown).

To examine whether absence of XBP1 promotes ER stress, we performed ultrastructural analyses of pancreas sections using electron microscopy. α-Cells could be distinguished from β- and δ-cells by their distinct secretory granules (31). In the αXBPKO mouse, the ER was severely dilated in ∼70% of α-cells compared with normal architecture in controls (Fig. 2F and G). These data indicate that XBP1 deficiency promotes ER stress in α-cells.

To begin to investigate the mechanisms underlying the phenotype, we generated three independent stable αTC6 cell lines, each with a XBP1 knockdown (αXBPKD) or overexpressing mouse spliced XBP1 (mXBP1s; αXBPOE). Stable expression of shRNA of XBP1 or mXBP1s was achieved using a lentiviral vector to transduce αTC6 cells. Nonsilencing shRNA and empty vector were used as controls for knockdown and overexpression cells, respectively. The effectiveness of knockdown or overexpression of the target protein was confirmed by Western blotting (Fig. 3A and D). In αXBPKD cells, the expression of XBP1s was reduced by ∼85% after thapsigargin treatment compared with controls, and the expression of XBP1s in αXBPOE cells was ∼10-fold higher than control cells in the basal state.

Next, we examined the proteins associated with the UPR by Western blotting. As expected, the expression of BiP, a downstream target of XBP1s, was decreased in αXBPKD cells (Fig. 3B and C). However, pIRE1α, a protein upstream of XBP1, was increased in αXBPKD cells, even in the basal condition (Fig. 3B and C). These results confirm the feedback regulation of IRE1α by XBP1 in α-cells, a feature that has been reported in other cell types, including β-cells (25). Examination of other UPR markers revealed increased active ATF6 and reduced pPERK in αXBPKD cells compared with control cells (Fig. 3A–C). To explore the effects of ER stress on activation of JNK (32), we examined the expression of pJNK in the αXBPKD cells. As expected, pJNK was upregulated in αXBPKD cells compared with controls (Fig. 3B and C). The expression levels of BiP, pIRE1α, and pJNK in αXBPOE cells were virtually opposite to that observed in αXBPKD cells. However, active ATF6 and pPERK showed a trend that was similar in αXBPKD and αXBPOE cells compared with their respective controls (Fig. 3D–F). Taken together, these data suggest feedback mechanisms between XBP1 and IRE1α and that XBP1 deficiency activates IRE1α and phosphorylation of JNK. The alterations in ATF6 and PERK warrant further investigation.

We next examined the effects of activation of JNK on phosphorylation of IRS1. As expected, IRS1 serine phosphorylation, which has been shown to have a negative effect on IRS1 signaling in some studies (33), was significantly increased in the basal state in αXBPKD cells compared with controls (Fig. 4A and B). In contrast, IRS1 tyrosine phosphorylation was significantly lower, and consequently, Akt Thr³⁰⁸ and Akt Ser⁴⁷³ phosphorylation were both markedly decreased in αXBPKD cells after insulin stimulation (Fig. 4A and B). In αXBPOE cells, however, we observed decreased IRS1 serine phosphorylation after insulin stimulation, although IRS1 tyrosine phosphorylation, Akt Thr³⁰⁸ phosphorylation, and Akt Ser⁴⁷³ phosphorylation were not significantly different compared with control cells (Fig. 4D and E).

On the basis of our previous report that impaired insulin signaling in α-cells leads to enhanced glucagon secretion (6), we hypothesized that αXBPKD cells would exhibit an inability to suppress glucagon secretion by insulin. Examination of control cells revealed that glucagon secretion was maximally suppressed at 5.5 mmol/L glucose compared with 0.5 or 25 mmol/L glucose (Fig. 4C and F). These results are consistent with previous reports showing maximal suppression of glucagon secretion at ∼7 mmol/L glucose concentration but paradoxically increased at higher glucose (>10 mmol/L) (34). In addition, insulin in the concentration range between 1 pmol/L and 1 μmol/L has been reported to suppress glucagon secretion from αTC6 cells in the presence of high glucose (25 mmol/L) (35). However, the failure of 100 nmol/L insulin to suppress glucagon secretion in the presence of high glucose (25 mmol/L) suggests impaired insulin signaling in αXBPKD cells (Fig. 4C). Glucagon secretion was not significantly different between control and αXBPOE cells (Fig. 4F).

Consistent with the data in αXBPKO mouse α-cells, our observation in αXBPKD cells suggests that XBP1 deficiency impairs insulin signaling and induces the inability of insulin to suppress glucagon secretion.

Next, we measured cytoplasmic Ca²⁺ concentrations ([Ca²⁺]) to assess the mechanism of dysregulation of glucagon secretion in αXBPKD cells. Consistent with previous reports (4,34), cytoplasmic [Ca²⁺] oscillations in control cells were dampened by application of high glucose and insulin (Fig. 5A, C, and F). A similar trend was evident in αXBPKD cells, although the response was blunted compared with control cells (Fig. 5A, B, and F). However, the average [Ca²⁺] level in αXBPKD cells when incubated in 0.5 or 25 mmol/L glucose or in 25 mmol/L glucose with 100 nmol/L insulin, was not significantly different from control cells (Fig. 5A, B, and E). On the other hand, the basal [Ca²⁺] was significantly higher in αXBPOE cells compared with controls, although the [Ca²⁺] oscillations were similar in the two groups (Fig. 5C–F). These data suggest that alterations in [Ca²⁺] are unlikely to underlie the dysregulation of glucagon secretion in XBP1 deficient α-cells. Further studies are warranted to carefully dissect the alterations in [Ca²⁺] when the expression levels of XBP1 are manipulated.

Because our experiments in αXBPKD cells suggested a role for XBP1 in the regulation of glucagon secretion, we next examined potential effects at the transcriptional level. Consistent with earlier reports (36,37), glucagon gene expression in control cells was significantly suppressed by insulin at a glucose concentration of 10 mmol/L but not at 2.8 mmol/L (Fig. 6A and B). In contrast, glucagon gene expression in αXBPKD cells was not suppressed by insulin, although basal glucagon gene expression was lower (Fig. 6A). On the other hand, interestingly, the ability of exogenous insulin to inhibit glucagon gene expression was obvious in αXBPOE cells (Fig. 6B).

We next examined the expression of nuclear FoxO1 because it has been reported to directly bind the preproglucagon gene promoter and regulate glucagon gene expression (37). In αXBPKD cells, nuclear FoxO1 was already decreased compared with control cells in the basal state, and insulin stimulation did not promote further exclusion of FoxO1 from the nucleus (Fig. 6C and E). In contrast, stimulation of αXBPOE cells with exogenous insulin had a significant effect on FoxO1 exclusion from the nucleus (Fig. 6D and E). Together, these data suggest that XBP1 regulates glucagon gene expression via the FoxO1 pathway in α-cells.

Histological analyses of pancreas sections from αXBPKO mice showed a normal complement of islet cells, indicating that absence of XBP1 does not enhance α-cell death in the basal state in vivo (Fig. 2). Therefore, we next sought to study the effects of XBP1 on cell growth and apoptosis in our in vitro models. First, examination of growth in αXBPKD and αXBPOE cells did not show significant differences between groups when compared with their respective controls (Fig. 7A and B). CHOP, a proapoptotic protein induced by ER stress, and active caspase 3 were significantly lower in αXBPKD cells in the basal state and after thapsigargin treatment (Fig. 7C and D). Surprisingly, αXBPOE cells also exhibited a decrease in CHOP expression after thapsigargin treatment, whereas expression of active caspase 3 was not significantly altered (Fig. 7E and F). One potential explanation for this conflicting result is that αXBPOE is a stable cell line that expresses high levels of XBP1s. This prompted us to design an experiment to examine the effects of “acute” overexpression of XBP1s by using cumate to induce XBP1s (38). After confirmation of induction of XBP1s by Western blotting (Fig. 8A), and consistent with an increased expression of XBP1s, the downstream target BiP was upregulated in the basal state (Fig. 8B). Interestingly, CHOP and active caspase 3 expression were both significantly increased after acute induction of XBP1s (Fig. 8B). Although there was no difference in the expression of pIRE1α, phosphorylation of JNK was significantly upregulated by acute induction of XBP1s (Fig. 8B). The activation of JNK is possibly related to apoptosis induced by acute induction of XBP1s, because JNK is generally activated by various cellular stress pathways (39). The expression of active ATF6 and pPERK after induction of XBP1s showed a trend that was similar to that observed in αXBPOE cells (Fig. 8A and B). These results from complementary models suggest that acute induction of XBP1s promotes, whereas a knockdown of XBP1 attenuates, apoptosis in α-cells.

To directly examine the role of XBP1, a transcription factor that plays a crucial role in the unfolded protein response in pancreatic α-cells, we created and phenotyped mice with a conditional knockout of XBP1 (αXBPKO mice). Here, we report that αXBPKO mice exhibit dysregulation of glucagon secretion and impaired glucose tolerance that is secondary to activation of the IRE1α-JNK pathway and consequent inhibition of IRS1 signaling leading to altered insulin action.

Although impaired glucose utilization due to insulin resistance contributes to the hyperglycemia in patients with T2D, an increase in hepatic glucose production due to altered gluconeogenesis also contributes (40). In normal subjects, the plasma glucagon concentration falls after oral glucose ingestion, whereas in subjects with T2D, the levels remain high (41), and some patients also exhibit a paradoxical increase in glucagon levels (42). Thus, an inability to suppress glucose production from the liver due to increased glucagon levels is associated with postprandial hyperglycemia in T2D patients. One possible mechanism underlying the hyperglucagonemia in T2D is impaired insulin signaling in α-cells that prevents the appropriate suppression of glucagon secretion. Indeed, mice with functional disruption of insulin receptors in α-cells (αIRKO) exhibit enhanced glucagon secretion and a diabetic phenotype (6). The similar phenotypes of the αXBPKO and αIRKO models support the notion that ER stress in α-cells is linked to altered insulin signaling and contributes to the postprandial hyperglucagonemia in T2D.

Among secretory cells, the β-cell is highly susceptible to ER stress (12,13), which may partly contribute to a significant reduction in β-cell mass in T2D, whereas α-cell mass is relatively normal (10), despite being exposed to high glucose and/or high free fatty acid levels. Our observation that absence of XBP1 did not affect α-cell mass in the αXBPKO mice or cell growth in the αXBPKD cells indicates that α-cells are relatively resistant to ER stress. One possible reason that α-cell development and/or survival are unaffected by lack of XBP1 is a relatively lower demand for glucagon secretion compared with insulin. Nevertheless, the selective regulatory effect of XBP1 on hormone secretion but not on cell growth in the α-cell requires further investigation. It is also worth noting that XBP1 plays diverse roles in other cell types. For example, XBP1-deficient lymphoid chimeric mice are completely deficient in normal plasma B cells, although the number of activated B lymphocytes are normal (23), suggesting that XBP1 is required for the terminal differentiation of B lymphocytes to plasma cells. On the other hand, the deletion of XBP1 in pancreatic acinar cells (24) and intestinal Paneth cells (43) promoted death by apoptosis during development. In hepatocytes and salivary gland acinar cells, XBP1 deficiency led to a modest reduction of secretory protein (24,29) but did not result in cell death or any other overt pathology in these cells, suggesting that XBP1 is essential in a small subset of highly secretory cells.

Hepatocytes and adipocytes have both been reported to be susceptible to ER stress in response to obesity and exhibit inhibition of insulin signaling (44), whereas neuronal cells in the brain manifest altered leptin signaling (45). Further, XBP1 deficiency in the liver or adipose tissue induces the activation of IRE1α and of JNK, with a consequent inhibition of insulin signaling (44), a finding that is also evident in the αXBPKD cells. Taken together, our findings suggest that downregulation of PI3K-Akt signaling is associated with unsuppressed glucagon secretion in XBP1-deficient α-cells. However, the recent report that IRS1 serine phosphorylation is beneficial to insulin signaling (46) suggests the effects mediated by the IRS1/JNK pathway are dependent on context and/or cell type. The reduction in nuclear translocation of XBP1s in the livers of obese mice (47) suggests that obesity promotes XBP1s deficiency in hepatocytes. Whether obesity similarly affects other tissues, including α-cells, with consequent defects in dysregulation of glucagon secretion and hyperglucagonemia is not understood and requires investigation.

The lack of a significant alteration in [Ca²⁺] in αXBPKD cells, despite an increase in glucagon secretion, suggests cAMP is a dominant second messenger in the regulation of glucagon secretion in α-cells (48). Consistent with this possibility, other studies have reported that glucagon secretion at high glucose concentrations does not require elevation of [Ca²⁺] (34).

The complementary alteration in glucagon gene expression in αXBPKD and αXBPOE cells indicates that XBP1 regulates glucagon gene expression via FoxO1 and is consistent with previous reports in hepatocytes that XBP1 directly interacts with FoxO1 (49). However, the significantly lower basal glucagon gene expression in αXBPKD cells, despite a relatively normal glucagon secretion, suggests the regulation of glucagon translation is independent of mechanisms that modulate gene expression. It is possible that the mammalian target of rapamycin complex 1 pathway, a downstream target of Akt that regulates protein synthesis (50), is involved in this mechanism.

In summary, we report that XBP1 interacts with proteins in the insulin-signaling pathway to regulate glucagon secretion in α-cells. We propose that ER stress in α-cells contributes to hyperglucagomemia and altered glucose homeostasis and that ER chaperones could be used in therapeutic approaches to modify α-cell function in T2D.

Some reagents used in this research work were supported by NIH RO1-DK-67536 (R.N.K.), NIH R37-DK046960 (R.T.K.), and NIH K99-DK090210 (C.W.). M.A. is the recipient of a Research Fellowship from the Hiroo Kaneda Research Aid Fund of the Sunstar Foundation of Osaka, Japan.

No potential conflicts of interest relevant to this article were reported.

M.A. designed and performed research, analyzed data, and wrote the manuscript. C.W.L., J.H., R.M., and B.H. performed research. S.L. and R.T.K. performed the experiment on Ca²⁺ measurements. R.N.K. supervised the project, designed the research, and wrote the manuscript. M.A. and R.N.K. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

The authors thank Jonathon N. Winnay, PhD (Joslin), for discussions, Christopher Cahill (Joslin), for assistance with electron microscopy, Laurie Glimcher, PhD (Weill Cornell Medical School, NY), for XBP1 floxed mice, and Pedro Herrera, PhD (Geneva, Switzerland), for sharing the glucagon-Cre mice.