Dehydration-induced AVP stimulates glucagon release and ketogenesis

Abstract

Gliflozins, such as dapagliflozin, belong to a class of drugs that inhibit the sodium-glucose cotransporter 2. Gliflozins have been found to raise glucagon levels, a hormone secreted from pancreatic islet alpha-cells, which can trigger ketosis. However, the precise mechanisms through which gliflozins increase glucagon secretion remain poorly understood. Additionally, gliflozins induce osmotic diuresis, resulting in increased urine volume and plasma osmolality. In this study, we investigated the hypothesis that a compensatory increase in arginine-vasopressin (AVP) mediates dapagliflozin-induced increases in glucagon in vivo. We show that dapagliflozin does not increase glucagon secretion in the perfused mouse pancreas, neither at clinical nor at supra-clinical doses. In contrast, AVP potently increases glucagon secretion. In vivo, dapagliflozin increased plasma glucagon, osmolality, and AVP. An oral load with hypertonic saline amplified dapagliflozin-induced glucagon secretion. Notably, a similar increase in glucagon could also be elicited by dehydration, evoked by 24-hour water restriction. Conversely, blockade of vasopressin 1b receptor signaling, with either pharmacological antagonism or knockout of the receptor, resulted in reduced dapagliflozin-induced glucagon secretion in response to both dapagliflozin and dehydration. Lastly, blocking vasopressin 1b receptor signaling in a mouse model of type 1 diabetes diminished the glucagon-promoting and ketogenic effects of dapagliflozin. Collectively, our data suggest that AVP is an important regulator of glucagon release during both drug-induced and physiological dehydration.

Graphical Abstract

New & Noteworthy:

Gliflozin-induced ketogenic effects partly result from increased glucagon levels. This study shows that dapagliflozin-triggered glucagon secretion is not directly mediated by the pancreas but rather linked to arginine-vasopressin (AVP). Dehydration, common in diabetic ketoacidosis, elevates AVP, potentially explaining the increased ketoacidosis risk in gliflozin-treated patients. Thus, our results highlight AVP as a potential therapeutic target to mitigate the risk of ketoacidosis associated with gliflozin treatments in patients with diabetes.

Introduction

Glucagon, a hormone synthesized in the alpha-cells of the pancreatic islets, plays a pivotal role in regulating glucose metabolism. It potently stimulates hepatic glucose output and is secreted in response to declining blood glucose levels (‘counter-regulation’). In type-2 diabetes, over-secretion of glucagon exacerbates the hyperglycemic impact of insufficient insulin, making therapeutic strategies targeting glucagon signaling a means to improve glycemic control (1, 2). Conversely, in type 1 diabetes, glucagon release during states of hypoglycemia is severely impaired (3–6).

Beyond its counter-regulatory function, glucagon is a key hormone in the body’s response to fast and starvation, promoting gluconeogenesis and ketogenesis. The ketogenic effect of glucagon has been long recognized and is most pronounced during conditions of low circulating insulin, such as during prolonged fasting or in cases of type 1 diabetes (7–9). Glucagon stimulates ketogenesis in many animal species, both in in vitro experiments (10–19) and when administered in vivo (11, 19–22). In humans, plasma glucagon levels correlate with ketone body concentrations (23–25), and glucagon administration has been shown to increase ketogenesis in both healthy individuals (26–29) and those with diabetes (25, 26, 28, 30–33). Furthermore, a more recent study involving human participants demonstrated that a single dose of a glucagon analog led to an increase in circulating ketones (34).

Diabetic ketoacidosis is a potentially life-threatening complication of diabetes characterized by metabolic acidosis due to the excessive production of ketones. A key clinical feature of diabetic ketoacidosis is severe dehydration (35). Effective treatment requires intravenous infusion of fluids, potassium, and insulin therapy to normalize the dehydration, electrolyte imbalance, hyperglycemia, and ketosis. Importantly, elevated glucagon is considered a hallmark of this condition (25, 30–32, 35, 36).

Gliflozins, such as dapagliflozin, are a class of oral anti-diabetic medicines that act by inhibiting the sodium-glucose cotransporter 2 (SGLT2). This transporter is responsible for ~90% of glucose reabsorption by the kidney. Inhibition of SGLT2 decreases the renal glucose reabsorption capacity in the early proximal tubules (37), resulting in an osmotic diuretic effect (38). This effect subsequently leads to an increase in urine volume and plasma osmolality (39–41). This osmotic diuresis, in turn, triggers the release of arginine-vasopressin (AVP, also known as anti-diuretic hormone) and thirst to increase water intake and reabsorption, to maintain body fluid volume (39, 41–43). While these drugs are generally well tolerated and have beneficial effects on cardiac function, blood pressure, and body weight, they do increase the risk of developing diabetic ketoacidosis by ~3-fold (44, 45). Notably, 70% of these cases arise under euglycemic conditions. This increase in plasma ketones from gliflozins results from increased ketone body production, rather than reduced clearance (46). Consequently, these side effects have stalled the widespread use of gliflozins for treating hyperglycemia, particularly in individuals with type 1 diabetes. The observation that gliflozins increase circulating levels of glucagon in both mouse (47–52) and human (53–58), may explain the cause of this ketogenic side effect. However, the mechanisms by which gliflozins increase circulating glucagon are not well understood.

Glucagon secretion is subject to various regulatory factors, including hypoglycemia (59–63), amino acids (64, 65), and somatostatin (66–68). Notably, vasopressin 1b receptor expression is high in both mouse and human alpha-cells (69–72), restricted to alpha-cells (73, 74), and upregulated when glucagon receptor signaling is interrupted (75). The ability of exogenous AVP to increase glucagon secretion has been known for some time (76). Recent studies demonstrate the crucial role of endogenous AVP as a regulator of in vivo glucagon release (63, 77). Here we explore the hypothesis that dapagliflozin increases circulating glucagon via the involvement of AVP. We also considered a possible physiological explanation for the presence of a receptor typically associated with water retention in islet alpha-cells and whether this mechanism becomes disrupted in type 1 diabetes.

Materials and methods

Ethics

All animal experiments were carried out in compliance with the ARRIVE guidelines and conducted in strict accordance with regulations enforced by the research institution. Experiments conducted in the UK were done in accordance with the UK Animals Scientific Procedures Act (1986) and University of Oxford. Animal experiments conducted in Sweden were approved by the Animal Ethics Committee of the University of Gothenburg, and were performed in accordance with EU guidelines for the care and use of laboratory animals (2010/63/EU).

Animals

All animals were kept in a specific pathogen-free (SPF) facility under a 12:12 hour light:dark cycle at 22 °C, with unrestricted access to standard rodent chow and water. C57BL/6J mice used in this study are referred to as wild-type mice. Avpr1b^−/− and littermate controls (Avpr1b^+/+) were bred, genotyped, and maintained as previously described (78). Wild-type, Avpr1b^−/− and controls of both sexes were used. Only female NOD mice were used because they develop type I diabetes earlier and with a higher incidence than males (79). These mice were purchased directly from JAX (001976, The Jackson Laboratory, Bar Harbor, Maine, USA). There were no exclusions of animals. The groups were not randomized, but animals were randomly divided into treatment groups. The investigators were not blinded to the treatment or genotype.

Hormone secretion measurements in the perfused mouse pancreas

Dynamic measurements of glucagon were performed using the in situ perfused mouse pancreas at physiological perfusion rates as previously described (67). C57BL6/J mice, 8–9 weeks of age, were culled by cervical dislocation, and the aorta was quickly cannulated by ligating above the coeliac artery and below the superior mesenteric artery. The pancreas was immediately perfused with Krebs-Ringer bicarbonate (KRB) buffer at a rate of ~0.25ml/min using an Ismatec Reglo Digital MS2/12 peristaltic pump to maintain cell viability and function. The KRB solution was maintained at 37 °C with a Warner Instruments temperature control unit TC-32 4B in conjunction with a tube heater (Warner Instruments P/N 64–0102) and a Harvard Apparatus heated rodent operating table. The pancreas was first perfused for 20 min with 8 mM glucose before commencing the experiment to establish the basal rate of secretion. Thereafter, dapagliflozin (12.5 µM or 100 nM) and AVP (100 nM) were added to the 4mM glucose perfusion. The effluent was collected by cannulating the portal vein and using a Teledyne ISCO Foxy R1 fraction collector. The effluent was collected in intervals of 1 min for up to 50 minutes into 96-well plates containing aprotinin and kept on ice. The plates were immediately frozen at −80 °C until analysis of glucagon and insulin. Glucagon from the perfusate was measured by ELISA using a Human U-plex glucagon system (Meso Scale Discovery, MD, U.S.A.) according to the protocol provided. Insulin was measured by using an insulin ELISA (10-1281-01; Mercodia, Sweden).

Plasma measurements

Plasma samples for AVP and osmolality measurements require large (>50 µL) sample volumes and were therefore collected after cervical dislocation. Samples for glucagon and glucose were collected under restraint via tail vein bleeds. All samples were collected at room temperature. Tail vein bleeds were aided by briefly (45 sec) warming the tail by gentle compression with a latex glove filled with lukewarm water (~30 °C). All plasma glucagon measurements were conducted with the 10 µL Glucagon ELISA (10-1281-01; Mercodia, Sweden). Insulin was measured using the 10 µL Insulin ELISA (10-1247-01; Mercodia, Sweden). Plasma osmolality was measured with a freezing point osmometer (Osmomat D30). Plasma AVP was measured with an ELISA (E09272m; Cusabio, China). Beta-hydroxybutyrate was measured using a FreeStyle Ketone Monitor (Abbott Diabetes Care, UK), and blood glucose measurements were taken with a Contour Next blood glucose meter (Ascensia, UK), both allowing repeated measurements due to the small test volume. All plasma samples were stored at −80°C and measured within 2 weeks of sampling.

Dapagliflozin injections in vivo

Samples for blood glucose and plasma glucagon measurements were taken from mice in response to dapagliflozin. Both sexes were used for these experiments. At ~08:00 am, mice were individually cage without food but with ad lib access to water for 1 hour. After this period, the mice were restrained, blood glucose measured and a tail vein sample of blood was taken into EDTA-coated tubes for glucagon measurements.

In the first set of experiments, dapagliflozin (10 mg/kg, Cat no 11574, Cayman Chemical, Michigan, USA) or vehicle (5% DMSO in PBS) was then administered i.p. and mice were re-caged without food but with ad lib access to water for 6 hours, when a final blood sample was taken.

In some experiments, mice were re-caged at this point with the addition of water restriction. After 6 hours following i.p. injection of dapagliflozin, a second blood sample was taken. These mice were fasted for the 6-hour study period.

In other experiments, a blood sample was also taken at 5 hours post dapagliflozin, followed immediately by an i.p. injection of the vasopressin 1b receptor antagonist SSR149415 (30 mg/kg, Cat. No. 6195, Tocris Bioscience, UK) or vehicle. A final blood sample was then taken 1 hour later (at 6 hours post-dapagliflozin). These mice were fasted for the 6-hour study period.

In another set of experiments, 4 hours after dapagliflozin administration (i.p.), mice were given an oral gavage (0.9% NaCl, 5% NaCl or sham gavage; 10 µL/g body weight) and re-caged without access to water for 2 hours. Blood samples (25–30 µL each) were collected at baseline and 4 and 5 hours after dapagliflozin administration. These mice were fasted for the 6-hour study period. During this period, mice were monitored for reflux.

Aprotinin (1:5, 4 TIU/ml; Sigma-Aldrich, UK) was added to all blood samples. At the end of all experiments, blood samples were centrifuged at 2700 rpm for 10 min at 4 °C to obtain plasma. The plasma was then removed and stored at −80 °C. Plasma was stored < 2 weeks before measurement.

Water restriction (dehydration) experiments

Wild-type mice (8–9 weeks of age) were single housed one week prior to experimental manipulation. During this time, 24h food consumption was measured. Next, mice were used for two consecutive experimental trials. For the first trial, the water restriction trial, blood was taken at 16:00 on day 1 and 2 for blood glucose and glucagon measurements (as above). During this trial, mice were completely water restricted for 24h (water was removed), but had ad lib access to food. The amount of food consumed during this 24h period was measured at the end of this trial and used for the subsequent food restriction trial. For the food restriction trial, blood was taken at 16:00 on day 1 and 2 for blood glucose and glucagon measurements. During the food restriction trial, mice had ad lib access to water between day 1 and 2, but were given the same amount of food consumed in the dehydration trial. The two trials were separated by 2 days without any manipulation.

NOD mouse experiments

Female NOD mice (NOD/ShiLtJ) were purchased directly from Jackson Labs at 6 weeks of age. Female NOD mice start developing type I diabetes when they are >11 weeks old, but up until 18 weeks of age, 30–50% of the mice remain normoglycaemic (79). At 11–14 weeks of age, diabetic and non-diabetic mice were randomly assigned to either the vasopressin 1b receptor antagonist SSR149415 or vehicle treatment. Diabetes was defined as a fed (AM) blood glucose measurement exceeding 10 mM on the day prior to the scheduled experiment. At 07:15 AM on the day of the experiment, mice were transferred to a cage without food but with ad lib access to water. At 07:30 AM (0 hours) a blood sample was taken and dapagliflozin (10 mg/kg, Cat no 11574, Cayman chemical, USA) was administrated to all mice. Eight hours later, at 03:30 PM, a blood sample was again taken. Mice were then randomly assigned to either vehicle (veh) or vasopressin 1b receptor antagonist SSR149415 (30 mg/kg, Cat. No. 6195, Tocris Bioscience, UK) treatment, which was administered 8 hours after dapagliflozin. A final blood sample was then taken 3 hours later (At 06:30 PM, 11 hours after Dapagliflozin injection). Blood samples were used to measure blood glucose, beta-hydroxybutyrate, and plasma glucagon.

Statistical tests of data

All data are displayed as box and whisker plots with median and quartiles indicated. Unless otherwise stated, n refers to the number of mice. Statistical significance was defined as p < 0.05. All statistical tests were conducted in Prism 9.2 (GraphPad Software, San Diego, CA, USA). For the two groups in Fig. 2 and Fig. 6, a t-test was conducted. For more than two groupings, a one-way ANOVA was conducted. A repeated measures one-way ANOVA was used in Fig. 1, for glucagon and insulin secretion over time, and in Fig 5; food intake over time. If data were separated by two treatments/factors, then a two-way ANOVA was conducted. A repeated measures two-way ANOVA was used if appropriate.

Figure 2: Dapagliflozin increases plasma osmolality, glucagon and AVP.

a. Plasma glucagon levels 6 h after dapagliflozin (10 mg/kg) or saline vehicle injection, n=8 WT mice/treatment.

b. Blood glucose levels 6 h after dapagliflozin or vehicle injection, n=10 WT mice/treatment.

c. Plasma osmolality 6 h after dapagliflozin or vehicle injection, n=9 WT mice/treatment.

d. Change in plasma AVP 6 h after dapagliflozin or vehicle injection, n=9 WT mice/treatment.

Differences are based on unpaired t-test and data are represented as boxplots with median and quartiles indicated. p<0.05=*, p<0.01=**, p<0.001=***

Figure 6: Basal blood glucose regulation status in NOD mice.

a. Basal blood glucose in diabetic (n=17) and non-diabetic (n=20) mice.

b. Basal plasma insulin in non-diabetic (n=20) and diabetic (n=7) mice.

c. Basal plasma glucagon levels in diabetic (n=11) and non-diabetic (n=9) mice.

d. Basal blood beta-hydroxybutyrate (BHB) levels, an indicator of ketosis, in diabetic (n=13) and non-diabetic (n=11) mice. All are female NOD mice in the basal state (fed, 07:30 AM). Differences are based on an unpaired t-test between the diabetic and non-diabetic mice, p<0.05=*, p<0.001=***. All data are represented as boxplots with median and quartiles indicated.

Figure 1: Dapagliflozin does not increase glucagon secretion in situ.

a. Glucagon secretion in response to 12.5 µM dapagliflozin (Dapa). 4G = 4 mM glucose, 8G = 8 mM glucose (n=4 WT mice).

b. Average secretion of glucagon during steady state for a. (n=4 WT mice).

c. Glucagon secretion in response to 100 nM Dapa and 100 nM AVP (n=4 WT mice).

d. Average secretion of glucagon during steady state for c. (n=4 WT mice).

Differences in b. and d. are based on paired sample t-test and data are represented as boxplots with median and quartiles indicated. p<0.05=*

Figure 5: AVP-induced glucagon secretion is increased during physiological states of dehydration and helps to maintain blood glucose.

a. Wild-type mice were first subjected to a 24-hour water restriction trial, wherein food consumption was monitored but water was completely restricted. Food consumption during this trial decreased by ~32%. During the second trial (the food restriction trial), the same mice were given unrestricted access to water, but had the same quantity of food available consumed in the first (water restriction) trial. One-way repeated measurement ANOVA; p<0.01=**. Tukey’s multiple comparison tests. n=8 wild-type mice.

b. Blood samples were taken during the water and food restriction trials for glucose and glucagon measurements. n=8 wild-type mice.

c. Plasma glucagon in the water and food restriction trial.

d. Blood glucose before and after 24-hour water restriction in Avpr1b^+/+ and Avpr1b^−/− mice, n=5/genotype.

e. Blood glucagon before and after 24-hour water restriction in Avpr1b^+/+ and Avpr1b^−/− mice.

Differences in b.-d. are based on two-way repeated measurement ANOVA, within trails/genotype; p<0.05=*, p<0.01=**, and between trails/genotype; p<0.05=†. Tukey’s multiple comparison tests. All data are represented as boxplots with median and quartiles indicated.

Results

Dapagliflozin does not increase glucagon secretion in situ

The direct effect of dapagliflozin on stimulating glucagon release from islet alpha-cells is disputed (47, 48, 50–52, 80–87). We employed the in situ perfused mouse pancreas to investigate whether dapagliflozin can directly increase glucagon output. When administrated at high doses (12.5 µM), there was no stimulation of glucagon output (Fig. 1a–b). We also show that clinically relevant doses of dapagliflozin (100 nM, (88)) did not quite attain statistical significance (p=0.07) when it comes to increasing glucagon output from the perfused mouse pancreas, in keeping with previous results (48)(Fig. 1c–d). In contrast, AVP (100 nM) potently stimulated glucagon secretion from the perfused mouse pancreas (Fig. 1c–d). Insulin secretion was not altered by dapagliflozin (0.1 µM or 12.5 µM) or AVP (Supplemental Fig. S1).

Dapagliflozin increases plasma osmolality, glucagon and AVP

Despite dapagliflozin failing to increase glucagon secretion from the perfused mouse pancreas (Fig. 1), the SGLT2 inhibitor has consistently been reported to increase circulating glucagon in vivo, in both rodents (47–52, 86, 87) and humans (53–56). Administration of 10 mg/kg dapagliflozin i.p. increased circulating glucagon 2.3-fold compared to the saline control group (Fig. 2a). As expected, dapagliflozin also decreased blood glucose, from ~7–8 mM to ~4–5 mM (Fig. 2b, 3a, 4a). In the perfused mouse pancreas, a comparable decrease in glucose only increased glucagon secretion 1.5-fold (Fig. 1b), making it unlikely that dapagliflozin acts solely by lowering glucose. Gliflozins are known to cause dehydration and increase circulating AVP (42, 43), therefore we also measured plasma osmolality and AVP in response to dapagliflozin. Dapagliflozin increased both plasma osmolality (Fig. 2c) and AVP (Fig. 2d).

Figure 3: Hypertonic saline exacerbates dapagliflozin-induced glucagon release.

a. At 0 hours, dapagliflozin was given (10 mg/kg). Four hours later, mice were given an oral gavage (sham gavage, 0.9% NaCl or 5% NaCl; 10 µL/g). Blood samples were then taken 5 and 6 hours after dapagliflozin injection. Blood glucose measurements., n=6–8 WT mice/treatment group.

b. Plasma glucagon measurements for a. n=6–8 WT mice/treatment group.

Differences are based on two-way repeated measurement ANOVA, within treatment; p<0.05=*, p<0.01=**, p<0.001=***, and between treatments; p<0.05=†, p<0.01=††. All data are represented as boxplots with median and quartiles indicated.

Figure 4: Blocking vasopressin 1b receptor signaling abolishes dapagliflozin-induced glucagon secretion.

a. Blood glucose at 0, 5 and 6 h after administration of dapagliflozin (Dapa; 10 mg/kg). At 5 h either saline vehicle or the vasopressin 1b receptor antagonist SSR149415 (30 mg/kg) was administered. n=7–8 WT mice/treatment group.

b. Plasma glucagon for experiments in a.

c. Blood glucose at 0 and 6 h after administration of Dapa (10 mg/kg) or saline vehicle in Avpr1b^−/− mice or littermate controls (Avpr1b^+/+ mice). n=7–9 mice/treatment.

d. Plasma glucagon for experiments in c.

Differences are based on two-way repeated measurement ANOVA, within treatment; p<0.05=*, p<0.01=**, p<0.001=***, and between treatments; p<0.05=†. All data are represented as boxplots with median and quartiles indicated.

Dapagliflozin does not increase plasma glucagon in Avpr1b knockout mice

Based on these data, we hypothesized that the dapagliflozin-induced increase in circulating glucagon (Fig. 2a) is principally a consequence of the concurrent increase in circulating AVP (Fig. 2d). We reasoned that if dapagliflozin principally acts via increasing circulating AVP, then its glucagon-elevating (glucagonotrophic) effects in vivo should be modulated by altering the hydration status of the mouse. To explore this, we injected dapagliflozin and then (4 hours later) administrated mice with hypertonic saline (5% NaCl) or isotonic saline (0.9% NaCl) via oral gavage, or sham gavage (Fig. 3a–b). Blood glucose levels were not different across the treatment groups (Fig. 3a). The glucagon response to dapagliflozin was greater in mice treated with hypertonic saline compared to isotonic saline (Fig. 3b). Gavage itself did not affect blood glucose and glucagon and data were the same as with isotonic saline.

To investigate the involvement of vasopressin we pretreated mice with the vasopressin receptor 1b antagonist SSR149415 (89). Vasopressin receptor 1b is the most abundant subtype expressed in α-cells (90). Pre-treatment with SSR149415 completely abolished the dapagliflozin-induced elevation of plasma glucagon despite a comparable decrease in glucose levels between groups (Fig. 4a–b). Next, we injected dapagliflozin into whole-body Avpr1b^−/− mice and littermate controls (78). Unlike what was observed in littermate control mice, dapagliflozin did not increase circulating glucagon in Avpr1b^−/− mice while the reduction in glucose was similar between genotypes (Fig. 4c–d).

AVP-induced glucagon secretion maintains glucose homeostasis during dehydration

Collectively, these data point to AVP being an important regulator of glucagon secretion during dapagliflozin-induced dehydration with resultant hyperosmolarity. We proceeded to determine whether AVP can induce glucagon release during water restriction as a more physiological model of dehydration.

We first measured 24-hour food intake in mice with unrestricted access to both food and water (Fig. 5a). Next, we subjected these mice to two different trials, each lasting 24 hours. First, we completely deprived the mice of water but gave them unrestricted access to food. The mice reduced their food intake by 32% but their plasma glucose levels remained stable (Fig. 5a–b). Next, we subjected the same mice (after a 2-day recovery period) to unrestricted access to water but food intake was adjusted to match that in the water restriction experiment. Using this experimental paradigm, the blood glucose fell (Fig. 5b). Glucagon was increased following water restriction, but not in the food restriction experiment (Fig. 5c). The maintenance of plasma glucose during water restriction is due to vasopressin 1b receptor signaling and elevation of plasma glucagon (63, 77, 91) (Fig. 5b, c). Accordingly, water restriction led to hypoglycemia and curtailment of the glucagon response in Avpr1b^−/− mice (Fig. 5d–e). These findings are from completely water-restricted mice with access to food. A detailed characterization of food intake in future studies is needed to determine if findings are due to different feeding patterns.

AVP-induced glucagon release contributes to diabetic ketoacidosis

We finally tested whether AVP-induced glucagon release contributes to diabetic ketoacidosis. We used the non-obese diabetic (NOD) mouse as a model of human type 1 diabetes. Both diabetic and non-diabetic female mice were injected with dapagliflozin, followed by an injection of the vasopressin receptor 1b antagonist SSR149415 or vehicle. As expected, blood glucose levels were higher in diabetic mice, particularly in the fed state (07:30 AM; Fig. 6a). Basal plasma glucagon was also higher in diabetic compared to non-diabetic mice (Fig. 6b) while diabetic mice had lower basal insulin release (Fig. 6c). The impact of diabetes is a lot more dramatic than the insulin levels indicate as the increase in blood glucose should have resulted in a massive stimulation of insulin. Both hormones displayed considerable variability in the fed state. Although diabetic mice exhibited hyperglucagonemia, there was no difference in basal beta-hydroxybutyrate, an indicator of ketosis, between the phenotypes (Fig. 6d).

We tested the effects of SSR149415 on dapagliflozin-induced changes in plasma glucose, glucagon, and beta-hydroxybutyrate. Delta changes were calculated to adjust for the variation at baseline in the diabetic group. In Fig. 7a–c, data are displayed as the change relative to baseline with the raw data shown in Fig. 7d–f. Dapagliflozin decreased blood glucose levels in both healthy control and diabetic mice and this effect was not influenced by SSR149415 (Fig. 7a, d). The fall in plasma glucose induced by dapagliflozin was associated with elevation of glucagon in control mice, this effect was not seen in mice with diabetes (Fig. 7b, e). In control mice, this stimulatory effect was abolished by SSR149415 and in mice with diabetes the vasopressin receptor antagonist reduced plasma glucagon (Fig. 7b, e). Beta-hydroxybutyrate levels were increased by dapagliflozin in both control and diabetic mice (Fig. 7f). The effect was greater in the diabetic mice despite unaltered glucagon, an effect that could be antagonized by SSR149415 (Fig. 7c, f), suggesting that some of AVP’s ketogenic effects are glucagon-independent in diabetic mice.

Figure 7: The ketogenic and glucagonotrophic effect of dapagliflozin is reduced by vasopressin 1b receptor blockade in NOD mice.

a. Change in blood glucose from basal in response to dapagliflozin (Dapa)+veh or Dapa+SSR149415 treatment (ΔGlucose). n=4–7 mice/group.

b. Change in plasma glucagon from basal in response to Dapa+veh or Dapa+SSR149415 treatment (ΔGlucagon). n=4–6 mice/group.

c. Change in beta-hydroxybutyrate from basal in response to Dapa+veh or Dapa+SSR149415 treatment (ΔBHB). n=4–6 mice/group.

Panel d.-f. present the values used to calculate the delta change presented in panel a.-c..

d. Blood glucose in response to Dapa+veh or Dapa+SSR149415 treatment. n=4–7 mice/group.

e. Plasma glucagon in response to Dapa+veh or Dapa+SSR149415 treatment. n=4–6 mice/group.

f. Plasma glucagon in response to Dapa+veh or Dapa+SSR149415 treatment. n=4–6 mice/group.

All are female diabetic and non-diabetic NOD mice. Differences in a.-c. are based on an unpaired t-test, SSR vs. veh within groups, p<0.05=*, p<0.01=**. Differences in d.-f. are based on one-way repeated measurement ANOVA, Sidak’s multiple comparison tests. 0 hour vs. 11 hour: p<0.05=*, p<0.01=**, p<0.001=***, and SSR vs. veh: p<0.05=†. All data are represented as boxplots with median and quartiles indicated.

Discussion

In this study we show that the SGLT2 inhibitor dapagliflozin increases circulating glucagon and AVP levels in mice. The increase in glucagon was not due to relative hypoglycemia, because mirroring the dapagliflozin-mediated fall in blood glucose did not markedly elevate glucagon secretion in the perfused mouse pancreas. Additionally, dapagliflozin failed to robustly stimulate glucagon secretion when vasopressin 1b receptor signaling was either pharmacologically or genetically blocked. Therefore, we propose that AVP secretion contributes to the observed elevation in circulating glucagon in response to the pharmacological SGLT2 inhibition (Fig. 8).

Figure 8: Schematic of how SGLT2 inhibitors increase circulating glucagon.

The ability of gliflozins to directly increase glucagon release from pancreatic islet alpha-cells is contested (1). Gliflozins block glucose reuptake in the early proximal tubule of the nephron (2). This results in increased glucose excretion and therefore a decrease in blood glucose (3). The decrease in blood glucose likely has a direct (but small) stimulatory effect on glucagon secretion. The increased glucose concentration in the collecting duct causes osmotic diuresis and therefore an increase in urine volume and plasma osmolality (4). This results in the release of AVP from magnocellular neurons (due to central osmosensing) into the circulation via the posterior pituitary (5). Circulating AVP stimulates islet glucagon release from the pancreas (6), via a process involving vasopressin 1b receptors. Therefore, gliflozins increase circulating glucagon via AVP, which in turn increases the risk of diabetic ketoacidosis (7).

We observed that dapagliflozin fails to robustly stimulate glucagon secretion from the perfused mouse pancreas, even at doses far exceeding those seen clinically. These findings are in keeping with other experiments conducted in isolated mouse islets (50, 87), the perfused mouse (87) or rat (85) pancreas, isolated human islets (51, 86, 87), and islets from whole-body SGLT2 knockout mice (92). Our previous study demonstrated the importance of SGLT2 for the regulation of somatostatin secretion in the context of high insulin (47), an experimental situation not investigated here. It should be noted that other groups have reported alterations in glucagon output from isolated mouse and human islets in response to SGLT inhibitors (48, 52, 80–84). Although we detect a small non-significant increase in glucagon output at both low and high concentrations of dapagliflozin, it seems clear that it is not enough to explain the 2.3-fold increase in circulating glucagon seen in vivo and this study was undertaken to address this aspect.

We found that dapagliflozin produces a small decrease in blood glucose, in itself is insufficient to account for the increase in circulating glucagon by mechanisms intrinsic to the alpha cells, but that it is associated with a significant increase in plasma osmolality. This glucagonotrophic effect of dapagliflozin was 2.5-fold higher in combination with dehydration. We have previously reported that moderate hypoglycemia increases plasma glucagon via AVP and that this effect becomes impaired in type 1 diabetes (63). By analogy we now hypothesized that the dapagliflozin-induced elevation of glucagon in vivo is also mediated by AVP. This hypothesis is supported by two key observations. First, the effects of dapagliflozin on plasma glucagon are antagonized by pharmacological or genetic inhibition of AVP receptors. Second, that 24-h water restriction increases plasma glucagon and that this effect is lost in Avpr1^−/−.

Whilst the ability of exogenous AVP and AVP analogs to potently stimulate glucagon secretion ex vivo and in situ has been known for some time (91), the precise physiological role of this regulation is not understood. We recently investigated the role of AVP in regulating glucagon secretion in vivo in response to 2-deoxyglucose and exogenous insulin (63). However, these are non-physiological challenges and do not explain why AVP stimulates glucagon release. Dehydration, on the other hand, is a homeostatic challenge that triggers a cascade of physiological responses due to the critical importance of maintaining adequate levels of hydration. A classic behavioral response to extended periods of dehydration is a reduction in food intake, known as dehydration anorexia (93). This physiological adaptation is important in the context of water balance by limiting the intake of osmolytes from food (which would exacerbate plasma hyperosmolality), reducing the amount of water used during digestion, and helps maintaining the integrity of the body’s fluid compartments. In rodent models of dehydration, where drinking water is replaced with hypertonic saline, food intake can decrease to as little as 20% of baseline intake, and body weight can decrease by 20% (93, 94). We found that mice subjected to 24 hours of water restriction displayed dehydration anorexia, resulting in a 32% reduction in food intake. Moreover, despite this reduction in food intake, we observed that blood glucose was maintained. Water-restricted mice also displayed increased circulating glucagon levels, consistent with an elevation in circulating AVP due to dehydration. In contrast, mice exposed to food restriction but with unrestricted access to water showed a decrease of ~ 1 mM in blood glucose. Circulating glucagon was not elevated in food-restricted mice. While it could be argued that the maintenance of blood glucose during dehydration is simply due to the blood being concentrated, we observed that blood glucose was not maintained in water restricted Avpr1b^−/− mice. The conclusion we draw from these data is that during states of water restriction/dehydration, elevated circulating AVP stimulates glucagon release, which helps to maintain blood glucose in dehydration anorexia.

Dehydration is a major feature of patients with diabetic ketoacidosis and is due largely to the osmotic diuresis associated with glycosuria. Rehydration therapy is key to treat the condition and often involves intravenous transfusion of as much as ~4L of isotonic saline within the first 12 hours. It is well known that plasma osmolality is elevated in diabetic ketoacidosis. Consequently, plasma AVP increases to >30 times their baseline values (95–98). Indeed, clinical trials are presently exploring the potential of copeptin, a stable surrogate for AVP (99), for monitoring ketoacidosis (100, 101). This substantial increase in AVP may explain why elevated glucagon is considered a hallmark in diabetic ketoacidosis. It also provides an explanation as to why the risk of diabetic ketoacidosis is elevated in gliflozin-treated patients (44, 45). These medications promote ketogenesis by stimulating glucagon release through AVP. Our findings in non-obese diabetic mice suggest that this pathway contributes to the development of ketogenesis in response to dapagliflozin.

We acknowledge some limitations to this study. First, diabetic mice did not exhibit ketosis at baseline, despite elevated glucagon levels. However, these measurements were obtained from fed mice in the early stages of their diabetic phenotype, occurring within four days following a blood glucose reading exceeding 20 mM. It is therefore plausible that beta-cell destruction that is a hallmark of diabetes in NOD mice was not complete, and the residual circulating insulin might have limited the ketogenic potential of glucagon. Finally, while blocking AVP signaling completely prevented the dapagliflozin-induced increase in ketones in non-diabetic mice, there was only a partial suppression of circulating ketones in diabetic mice. This clearly suggests that processes other than AVP-induced glucagon secretion are important for the development of ketoacidosis. Regardless of this, our data suggest that AVP-induced glucagon secretion contributes to diabetic ketoacidosis in response to gliflozins.

Supplementary Material

Supplemental Figure 1: Dapagliflozin does not alter insulin secretion in situ

a. Insulin secretion in response to 12.5 µM Dapagliflozin (Dapa). 4G = 4 mM glucose, 8G = 8 mM glucose (n=3 WT mice).

b. Average secretion of insulin during steady state for a. (n=3 WT mice).

c. Insulin secretion in response to 100 nM Dapa and 100 nM AVP (n=4 WT mice).

d. Average secretion of insulin during steady state for c. (n=4 WT mice).

Differences in b. and d. are based on paired sample t-test and data are represented as boxplots with median and quartiles indicated. p<0.05=*

Data availability statement

The authors declare that all data supporting this study’s findings are available from the lead author upon reasonable request.