Pancreatic β-Cell Rest Replenishes Insulin Secretory Capacity and Attenuates Diabetes in an Extreme Model of Obese Type 2 Diabetes

Chronic hyperglycemia resulting from the failure of pancreatic β-cell function and mass to compensate for insulin resistance marks the onset of obesity-linked type 2 diabetes (T2D). Early in the pathogenesis of obese T2D, a compensatory increase in β-cell function and/or mass can cater to increased metabolic load and insulin resistance. With time, however, the chronic demand on the β-cell becomes exhaustive, leading to inadequate functional β-cell mass through apoptosis or loss of identity and ultimately to frank T2D (1–4). This progressive β-cell exhaustion in the pathogenesis of T2D has led to the concept that transient β-cell rest might be beneficial to the treatment of T2D by decreasing β-cell demand and subsequently improving function and/or promoting the survival of remaining endogenous β-cells (5). There is some precedence for this by improving insulin sensitivity in specific situations, such as gestational diabetes (6), marked calorie restriction (7), or short-term direct inhibition of endogenous insulin secretion in obese T2D (8). However, attenuating metabolic demand to improve β-cell function has not been examined in detail at the level of the β-cell using current T2D therapeutic approaches.

Whether failure of β-cell function or reduced β-cell mass is key to the onset of T2D often has been debated, but in humans, disease pathogenesis is likely to be variable and contributed by both (9). For β-cell dysfunction in T2D, there is a characteristic loss of glucose sensing, blunted first-phase glucose-induced insulin secretion, and presumed inadequate insulin production (10). In contrast, obese T2D mouse models have indicated that β-cell insulin production is markedly elevated despite severely depleted intracellular insulin stores (11). Of note, when pancreatic islets isolated from these obese T2D mice were incubated at normal glycemia (5.6 mmol/L) overnight, the rate of insulin production normalized, intracellular insulin secretory granule stores replenished, and glucose-induced biphasic insulin secretion was restored (11).

We examined whether a glucose-lowering therapeutic strategy (by the glucagon-like peptide 1 receptor [GLP-1R] agonist liraglutide and/or sodium/glucose cotransporter 2 inhibition [SGLT-2i] with dapagliflozin) or an insulin sensitizing approach (using the thiazolidinedione [TZD] rosiglitazone alone or in combination with dapagliflozin) alleviates the demand on the β-cell in vivo in BKS.Cg-Dock7^m +/+ Lebr^db/J (hereafter referred to as KS db/db) mice (an extreme model of obese T2D because of failed β-cell compensation) to restore endogenous β-cell function. We chose this model (rather than polygenic or diet-induced models of diabetes) because it presents with a catastrophic and early loss of β-cell function. Thus, if therapeutic efficacy can be observed in such an extreme model, we anticipate that it would extend to less deleterious pathophysiologies of the same disease. In addition, we used lean C57BLKS/J mice as comparative controls. Our findings highlight that glycemic lowering with established T2D therapies, especially in combination, also can instigate β-cell rest in vivo by using the inherent adaptive flexibility of β-cells to improve their secretory capacity and function, thereby slowing progression of the disease.

Four cohorts of male KS db/db mice were used. C57BLKS/J animals were used as lean controls. Animals were group housed with ad libitum access to water and rodent chow in a controlled environment. Cohorts 1 and 2 were used to study the combination of dapagliflozin/liraglutide, whereas cohorts 3 and 4 were used to study the combination of rosiglitazone/dapagliflozin. Cohorts 1 and 3 were used to demonstrate the efficacy of combination therapies, pancreatic insulin content, and immunohistochemistry, whereas cohorts 2 and 4 were used for islet experiments. Cohorts 2, 3, and 4 were purchased from The Jackson Laboratory (Farmington, CT) and studied at MedImmune LLC, whereas cohort 1 was purchased from Charles River Laboratories Italia (Calco, Italy) and studied at Gubra ApS. All cohorts were purchased at 7 weeks of age, acclimated for 1 week, randomized according to body weight and percent HbA_1c (%HbA_1c), and studied for 4 weeks. Animals were dosed daily on the basis of body weight. Rosiglitazone (U.S. Pharmacopeia, Rockville, MD) and dapagliflozin (AstraZeneca UK) were dosed orally at 3 mg/kg and 1 mg/kg, respectively using hydroxypropyl methylcellulose (HPMC) vehicle. Liraglutide (Novo Nordisk, Bagsværd, Denmark) was dosed subcutaneously (0.2 mg/kg) using PBS vehicle. Control animals in cohorts 3 and 4 received oral HPMC daily; control animals in cohorts 1 and 2 received both oral HPMC and subcutaneous PBS daily. At the end of the 3rd study week, intraperitoneal glucose tolerance test (ipGTT), fasting plasma glucose (FPG), fasting plasma insulin, and %HbA_1c were determined. At the end of the 4th study week, body composition was determined, and animals were sacrificed for the collection of pancreas, islets, plasma, or liver as appropriate. Animal care, use, and experimental protocols were approved by the institutional animal care and use committees of MedImmune and Gubra.

Six-hour fasted mice were injected intraperitoneally with 1.5 g/kg glucose in saline. Blood glucose was determined at 0, 5, 10, 15, 60, and 180 min for cohort 1 and 0, 15, 30, 60, 120, and 240 min for cohort 2. The circulating plasma glucose levels of the mice were determined colorimetrically, whereas the plasma insulin level was determined through ELISA at 15 min.

Pancreatic insulin content was determined from whole pancreas through acid-ethanol extraction and insulin ELISA. Plasma glucose was determined through colorimetric glucose oxidase kit (Cayman Chemical, Ann Arbor, MI). Plasma insulin was determined through ELISA (Meso Scale Diagnostics, Rockville, MD). The %HbA_1c was determined colorimetrically from whole blood (Crystal Chem, Elk Grove Village, IL). Plasma cholesterol, triglycerides, AST, and ALT were determined using the cobas c111 autoanalyzer (Roche, Basel, Switzerland). Liver fat was determined using the Bruker minispec mq (Bruker, Billerica, MA). Body composition was determined using the Bruker LF90II analyzer.

For immunohistochemistry, pancreata were fixed, embedded, and cut into 5-μm sections. Sections were stained on the BOND RX system (Leica Microsystems, Buffalo Grove, IL) with rabbit anti-MafA (Bethyl Laboratories, Montgomery, TX) or rabbit anti-Ki67 (Abcam, Cambridge, MA) followed by antigen retrieval, 3,3′-diaminobenzidine staining, and glucagon/insulin detection as previously described (12). β-Cell and α-cell mass were quantified from insulin/glucagon (brown/pink) stained sections (n ≥ 3) using HALO software (Indica Laboratories, Corrales, NM). Here, insulin/glucagon/hematoxylin chromogenically stained slides were scanned and uploaded to the HALO software suite. HALO discriminates between stained/unstained area and thus can quantitate the area of the pancreas occupied by either insulin or glucagon. From these values, we can determine the area ratios of pancreas to β-cell or α-cell (i.e., β-cell mass, α-cell mass, respectively).

Freshly isolated islets were fixed in 0.1 mol/L cacodylate buffer containing 4% paraformaldehyde/2% glutaraldehyde. Samples were resin embedded, sectioned, and stained as previously described (11). Micrographs were imaged and quantified as previously described (12). Mature insulin granules and immature insulin granules were quantified per total cytoplasmic area from ≥20 electron micrographs per group (n ≥ 3 biological replicates, ≥1.0 mm² total cytoplasmic area).

Pancreatic islets from C57BL/6J, KS db/db, or C57BLKS/J mice were isolated by collagenase digestion, as previously described (13). Human islets were obtained from Prodo Laboratories (Aliso Viejo, CA). Glucose-stimulated insulin secretion experiments were in C57BL/6J islets cultured overnight in RPMI media (10% FBS, 5.6 mmol/L glucose) (Thermo Fisher Scientific, Waltham, MA). Insulin secretion was assessed by 1-h static incubation in the presence of rosiglitazone (10 μmol/L), dapagliflozin (10 μmol/L), or both dissolved in 0.2% DMSO, as previously described (11), after preincubation at 37°C for 4 h with the same study compounds. For quantitative RT-PCR (qRT-PCR) and perifusion, freshly isolated islets from KS db/db and C57BLKS/J mice were used. Islet RNA was isolated using TRIzol reagent (Thermo Fisher Scientific). cDNA was synthesized using the Superscript III First-Strand Synthesis System (Thermo Fisher Scientific). qRT-PCR analysis was performed using Taqman Gene Expression Assay probe/primer sets (Thermo Fisher Scientific) for Ins1, Ins2, Gck, Slca2a, Pdx1, and Rna18s; PrimePCR Probe Assay (Bio-Rad, Hercules, CA) was used for Mafa. Results are shown as the target gene expression relative to Rna18s expression and normalized to vehicle expression using the 2^(−ΔΔCt) method. For perifusion, an equivalent number of medium-sized islets were placed into individual perifusion columns connected to a 12-channel perifusion apparatus (Biorep Technologies, Miami Lakes, FL). After 90 min of preperifusion with 2.8 mmol/L Krebs-Ringer bicarbonate HEPES buffer (KRBH), biphasic islet insulin secretion was assessed. KRBH (2.8 mmol/L glucose) was pumped through the chambers followed by KRBH (16.7 mmol/L glucose), as previously described (14). The perfusate was collected at 2-min intervals in 96-well plates and maintained at 4°C. Islets were lysed for protein content, and perfusate insulin concentrations were determined through ELISA.

Data normality was determined by the D’Agostino and Pearson test. Parametric data were analyzed by one-way or two-way ANOVA followed by Tukey test. Nonparametric data were analyzed by Kruskal-Wallis test followed by Dunn multiple comparisons test. For islet experiments (Figs. 3 and 4), data for individual cohorts were analyzed separately to confirm that no statistical differences existed between combined cohort data and individual cohort data (data not shown). Data are presented as mean ± SD or box and whisker plots ± minimum/maximum, as appropriate. All data were analyzed using GraphPad Prism 7.02 (GraphPad Software, San Diego, CA). Statistical significance was set at P ≤ 0.05.

Eight-week-old male KS db/db mice were treated with select diabetes therapies: peroxisome proliferator–activated receptor-γ (PPARγ agonism (rosiglitazone), GLP-1R agonism (liraglutide), SGLT-2i (dapagliflozin), or the combination of SGLT-2i plus GLP-1R or PPARγ agonism over 4 weeks. Study animals displayed markedly different weight profiles (Figs. 1A–D and 2A–D), depending on GLP-1 or PPARγ therapy. Vehicle-treated animals had increased body weight by ∼10% (Figs. 1E and 2E), whereas liraglutide monotherapy led to a 7.5% increase, similar to lean animals (Fig. 1E); dapagliflozin monotherapy led to an ∼15% increase (Figs. 1E and 2E); and rosiglitazone monotherapy led to a 29% increase (Fig. 2E). The combination of dapagliflozin/rosiglitazone led to a 34% increase in body weight in contrast to the only 8% increase observed for dapagliflozin/liraglutide, a fourfold difference (Figs. 1E and 2E). Body composition analysis revealed reductions in fat mass after treatment with dapagliflozin (Fig. 2F), liraglutide, and dapagliflozin/liraglutide (Fig. 1F). Lean mass correspondingly was increased by liraglutide and the combination of dapagliflozin/liraglutide but unaffected by dapagliflozin or rosiglitazone (Figs. 1G and 2G). Rosiglitazone administration increased liver fat, liver weight, and plasma ALT, AST, and cholesterol but sharply reduced plasma triglycerides (Supplementary Table 1). Despite not influencing liver fat, dapagliflozin monotherapy increased liver weight (Supplementary Table 1).

Vehicle-treated animals displayed a 2-point increase in %HbA_1c, which was significantly mitigated by monotherapy treatments (Figs. 1H and 2H). However, both combination strategies were more effective than the respective monotherapies; dapagliflozin/rosiglitazone led to a 1-point reduction in %HbA_1c (Fig. 2H), whereas dapagliflozin/liraglutide maintained %HbA_1c, similar to lean animals (Fig. 1H). The observed improvements in %HbA_1c were paralleled by improved glucose tolerance. Combination-treated animals displayed improved glucose excursions compared with vehicle controls (Figs. 1I and 2I). The total glucose area under the curve (AUC) values for both combination-treated groups were significantly reduced, nearly to the level of the lean animals (Figs. 1J and 2J). In vivo insulin secretion during the ipGTT was significantly improved by both liraglutide and rosiglitazone, although this effect was far more pronounced by liraglutide in line with its insulinotropic action (Figs. 1K and 2K). All monotherapies reduced FPG levels versus vehicle-treated animals, but combination strategies elicited a more robust decrease comparable to lean animals (Figs. 1L and 2L). The restoration of plasma glucose to near-normal levels by both combination strategies was facilitated by correspondingly greater plasma insulin levels capable of overcoming the inherent insulin resistance in KS db/db animals (Figs. 1M and 2M). Although clear differences exist in the absolute levels of both fasting glucose and insulin between cohorts, the relative changes in glucose confirm the largely similar therapeutic efficacies of the glucose-lowering strategy (liraglutide ± dapagliflozin) or by improving insulin sensitization (rosiglitazone ± dapagliflozin) (Supplementary Fig. 1A and B, respectively). Finally, HOMA of insulin resistance (HOMA-IR), a relative measure of insulin sensitivity, was unaffected by liraglutide therapies but had a tendency to improve with rosiglitazone therapies, although not statistically significantly (Figs. 1N and 2N).

Despite stark differences in body weight, combinations of dapagliflozin and liraglutide or rosiglitazone increased pancreatic insulin content more than threefold (Fig. 3A). Of note, no significant change in β-cell mass was observed in combination-treated animals (Fig. 3B and Supplementary Fig. 2A). Correspondingly, no difference in islet β-cell replication was observed (Supplementary Fig. 3 and Supplementary Table 2). However, combination-treated mice displayed a greater number of mature insulin granules (Fig. 3C), although the content of immature granules was unchanged (Fig. 3D). Representative electron micrographs of β-cells from untreated KS db/db mice (Fig. 3E) paralleled the morphology observed in previous studies of obesity-linked T2D (11): marked mature insulin secretory granule degranulation, a greater proportion of immature secretory granules relative to mature secretory granules, and an expansion of the Golgi apparatus and endoplasmic reticulum. Rosiglitazone (Fig. 3F), dapagliflozin (Fig. 3G), and liraglutide (Fig. 3H) all demonstrated some ability to increase insulin granule number. However, the combination of dapagliflozin/rosiglitazone (Fig. 3I) or dapagliflozin/liraglutide (Fig. 3J) induced marked insulin regranulation and reduced endoplasmic reticulum area. Staining for MafA, a key pancreatic transcription factor exclusive to β-cells (15), increased proportionally to the intensity of insulin in lean control and combination-treated islets (Fig. 3P–R) versus monotherapy-treated (Fig. 3M–O) and vehicle control islets (Fig. 3L), suggesting that β-cell recovery of endogenous insulin secretory capacity underlies significant antidiabetic efficacy (Supplementary Table 3).

Expression of key genes related to specific β-cell function were examined from isolated islets after 4-week treatment. Ins1 transcript levels were significantly increased from both combination-treated animals by approximately eightfold (Fig. 4A), whereas Ins2 transcript levels were significantly increased in islets from liraglutide monotherapy (fivefold) and liraglutide/dapagliflozin combination-treated animals (sixfold) (Fig. 4B). The mRNA levels of Slc2a2 and Gck were increased in combination-treated animals versus vehicle controls (Fig. 4C and F) as were Mafa transcript levels, particularly rosiglitazone/dapagliflozin (12-fold) (Fig. 4D), complementary to MafA immunostaining (Fig. 3). Pdx1 expression also was significantly increased in islets from animals treated with liraglutide monotherapy and combination therapies (Fig. 4E).

Because loss of first-phase insulin secretion is a hallmark of β-cell secretory dysfunction (16), we assessed the insulin secretory response in islets from vehicle controls, treated mice, and lean controls. Immediately after isolation, islets were preperifused with basal 2.8 mmol/L glucose KRBH for 60 min followed by perifusion with 2.8 mmol/L glucose KRBH and then stimulatory 16.7 mmol/L glucose KRBH to assess biphasic insulin secretion. Islets from monotherapy-treated animals demonstrated some recovery of first-phase insulin secretion and improved second-phase insulin secretion versus vehicle control islets (Fig. 4G). However, islets from combination-treated animals showed robust first-phase insulin secretion and sustained second-phase insulin secretion, even compared with islets from monotherapy-treated animals. Islets from combination-treated animals elicited significantly greater first-phase (Fig. 4H) and second-phase (Fig. 4I) insulin secretion relative to vehicle controls, indicating a synergistic improvement in recovery of functional insulin secretory capacity over monotherapy treatments.

The onset of common obesity-linked T2D is marked by loss of functional β-cell mass that no longer maintains normal glucose homeostasis. However, whether β-cell function and/or mass drive T2D disease pathogenesis remains unclear (3). Here, use of the KS db/db mouse, an extreme model of hyperglycemic/hyperinsulinemic obesity-induced insulin resistance (17), has revealed a bias toward β-cell dysfunction as a major driver in T2D. In the same mouse model, it was shown that endogenous insulin stores in remaining β-cells are markedly depleted, and these animals were profoundly hyperinsulinemic. However, insulin production was markedly upregulated, and the newly synthesized (pro)insulin rapidly secreted in efforts to compensate for increased metabolic demand (11). In this instance, the whole β-cell ultrastructure adapted to one of insulin production rather than to one of storage, implicating that apparent β-cell dysfunction is a consequence of a hard-working β-cell rather than of an inherent functional defect (11). Moreover, because of the degranulation of β-cells in such obese diabetic pancreatic islets, the use of insulin as an immunohistochemical marker likely underestimated measures of β-cell mass. Of note, parallel ex vivo studies in isolated islets from KS db/db mice incubated at euglycemia overnight led to normalized β-cell ultrastructure, restoration of β-granule insulin stores and secretory capacity, and consequent normalization of glucose-regulated insulin secretion (11).

This study supports the hypothesis that metabolic demand on the β-cell can be reduced in vivo using common T2D therapeutics (especially in combination) to yield beneficial consequences for the treatment of obese T2D by indirectly restoring endogenous β-cell function and insulin secretory capacity independent of changes in body weight. However, the disease pathophysiology in the KS db/db mice versus human obese/T2D warrants mention. This animal model was chosen as one of extreme obesity/insulin resistance with significantly compromised functional β-cell mass to better reveal that approaches to instigate β-cell rest in vivo could improve endogenous β-cell secretory capacity and function. But there should be reservation in that loss of leptin signaling is not a primary contributor to common human obese/T2D, and the severe hyperinsulinemia present in these mice contrasts with a relatively moderate hyperinsulinemia observed in the progression of human T2D (18). However, despite these concerns, this study complements previous clinical work of instigating β-cell rest by using the less orthodox therapeutic strategies of diazoxide to inhibit endogenous insulin secretion in humans (8) or insulin therapy in mice (19). Dietary restriction using sharp reductions in caloric intake can result in improved β-cell function (7). In this case, it is likely that improvement in insulin sensitivity as a consequence of weight loss was the primary driver of this functional reversal. A recent study of physician-assisted caloric restriction also found that the rate of diabetic remission correlated with the degree of weight lost (20), as does bariatric surgery (21). Currently, the ongoing Restoring Insulin Secretion (RISE) clinical trial is investigating the preservation of β-cell function after combinatorial pharmaceutical intervention (metformin ± liraglutide) in adult and pediatric patients with prediabetes and early T2D (22). Here, the more conventional therapeutic strategies were chosen to lower glucose homeostasis in vivo by distinct mechanisms of action that did not necessarily have direct effects on pancreatic β-cells. SGLT-2i lowers circulating glucose by preventing glucose reuptake in the proximal kidney (23), without effects on β-cell insulin secretion (Supplementary Fig. 4). TZDs are partial PPARγ agonists and primarily act as insulin sensitizers, with little effect on insulin secretion (Supplementary Fig. 4), and have shown promise in delaying the net loss of β-cell mass during the pathogenesis of rodent T2D (24). Although GLP-1R agonism, with GLP-1 analogs such as liraglutide, acts as a glucose-dependent insulin secretagogue (25), in obese T2D, it is also quite effective at inducing satiety and weight loss that, together with effects on delaying gastric emptying, improve glucose homeostasis (26,27). As monotherapies, these three distinct therapeutic strategies demonstrated some degree of improving glucose homeostasis on the severe obese-diabetic KS db/db mouse model, but they were synergistically more effective in combination. Indeed, emerging clinical data support these results; for instance, SGLT-2i was found to improve incretin sensitivity in individuals with uncontrolled T2D (28). Here, SGLT-2i in combination with either GLP-1R agonism or TZD therapy led to additive improvements in HbA_1c, FPG, and glucose tolerance, despite a dramatic difference in body weight between the two combination-treated groups (Figs. 1 and 2). Both SGLT-2i/GLP-1R agonism or SGLT-2i/TZD combination strategies markedly improved endogenous β-cell function. Fasting insulin levels improved as FPG decreased; there was a marked increase in pancreatic insulin content that was paralleled by a substantial regranulation of endogenous β-granule insulin stores, which in turn replenished insulin secretory capacity and restored normal glucose-induced biphasic insulin secretion. Furthermore, both combination therapies increased the expression of key β-cell transcription factors Mafa and Pdx1, which control the expression of other β-cell functional genes (29) that also were increased by these treatments (Fig. 4). Of note, β-cell mass did not appreciably change in these KS db/db mice with either combination therapy, emphasizing that compensation to insulin resistance occurs predominantly through alterations in β-cell function (30).

Pancreatic β-cells are remarkably adaptable in response to metabolic status. For example, β-cell insulin secretory capacity essentially is shut down during prolonged fasting as a protective measure against hypoglycemia, yet the fasted β-cells remain poised to rapidly restore effective insulin production hours after refeeding (12). Indeed, in patients who underwent gastric bypass, β-cell sensitivity to glucose was observed to be reduced, even many years after the procedure, implicating an adaptation to prevent hypoglycemia from excess postprandial secretagogue-induced insulin secretion (31). The current findings suggest that therapeutic alleviation of insulin resistance or improvement in glucose homeostasis reduces the demand on endogenous β-cells for increased insulin production in T2D, allowing β-cell adaptation to rapidly restore insulin secretory capacity and normalize insulin secretion. The adaptive consequence of reduced glycemic burden is improved β-cell function, which ultimately contributes to attenuating T2D. These findings likely have translational implications for the treatment of common obese T2D in humans. First, an early diagnosis of T2D and subsequent effective combination treatment should preserve endogenous β-cell function and, as such, delay disease progression. Second, the β-cell itself should not necessarily be a direct therapeutic target; rather, pathophysiologies of T2D that increase demand on the β-cell should be primarily targeted to then allow the β-cell to rest and consequently restore normal β-cell function and glucose homeostasis. Indeed, direct insulinotropic therapies targeting an already overworked β-cell in T2D, in the absence of enhanced secretory function, could be detrimental and accelerate disease pathogenesis.

Acknowledgments. The authors thank Dr. Jotham Austin II and Yimei Chen of the University of Chicago Advanced Microscopy Core. The authors would also like to thank the in vivo pharmacology group at Gubra ApS.

Duality of Interest. This work was funded by MedImmune. B.B.B. and M.L.B. are previous employees of MedImmune. C.B., J.C., M.S., W.K., C.R., J.T., C.J.R., and J.S.G. are current employees of MedImmune. B.B.B., M.L.B., G.H., and R.V.G. are current employees of Gubra ApS. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. B.B.B. designed the study, wrote the manuscript, and performed animal studies and all islet work. C.B., J.C., M.S., and W.K. performed immunohistochemistry. M.L.B. performed qRT-PCR. G.H. and R.V.G. performed animal studies. C.R., J.T., C.J.R., and J.S.G. aided in the experimental design. B.B.B. 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 at the 77th Scientific Sessions of the American Diabetes Association, San Diego, CA, 9–13 June 2017.