Effects of 13-Hour Hyperglucagonemia on Energy Expenditure and Hepatic Glucose Production in Humans

More than 80% of patients with type 2 diabetes are overweight or obese and would benefit from weight loss. Glucagon-like peptide 1 receptor (GLP-1R) agonists are currently the only available medications indicated for the treatment of type 2 diabetes that can also reduce body weight in a clinically meaningful manner. However, weight loss efficacy of GLP-1R agonists is restricted as a result of dose-limiting nausea and vomiting (1). Glucagon (GCG), derived from proglucagon as is GLP-1, has primarily been characterized as a counterregulatory hormone that responds to hypoglycemia and fasting by stimulating hepatic glucose production (HGP), fatty acid β-oxidation, and ketogenesis (2). In addition, GCG is known to reduce appetite (3) and increase energy expenditure (EE) acutely. For example, intravenous infusion of GCG and somatostatin for 2–3.5 h, which raised plasma GCG levels approximately three- to fourfold without concomitant hyperinsulinemia, increased EE ∼7–14% in young healthy individuals (4,5). Moreover, intravenous GCG for 45–55 min, which raised plasma GCG ∼20-fold without prevention of hyperinsulinemia, increased EE ∼10–15% (6,7). To put that into clinical perspective, such an increase in EE, if sustained and not offset by increased food intake or behavioral changes, would cause weight loss of ∼7–10% within 6 months (8). GCG receptor (GCGR) agonists have thus emerged as a possible suitable therapeutic partner for GLP-1R agonists for increased weight loss efficacy, in which GLP-1R agonism will counterbalance the hyperglycemic effects of GCGR activation, and GCGR agonism will promote additional weight loss through increased EE and reduced food intake. However, hitherto human studies that have demonstrated GCG-induced stimulation of EE have infused GCG for only 45–210 min (4–7). Cloned rat hepatocytes become refractory to further stimulation of cAMP synthesis by GCG after 4-h exposure to this hormone (9). Further, the rise in glucose production in response to sustained hyperglucagonemia was found to merely last ∼1 h in healthy and insulin-deficient humans (10), suggesting that the effect of GCG to stimulate EE may also be relatively short-lived. In the current study, we therefore investigated the effects of 13-h hyperglucagonemia versus 3-h hyperglucagonemia on EE and glucose metabolism in humans.

First, 10 healthy individuals (3 males and 7 females) were studied to determine the intraindividual variation of measuring resting EE (REE) and the respiratory quotient (RQ) in unique small-room metabolic chambers. These participants were 40 ± 4 (SEM) years of age and had a BMI of 25.8 ± 1.6 kg/m². In the subsequent main study, six healthy lean males completed the experiments to determine the effects of 13-h hyperglucagonemia versus 3-h hyperglucagonemia on EE and glucose metabolism using a prospective, randomized, single-blind, three-period crossover design. Participants of this study were 33 ± 3 years of age and had a BMI of 23.5 ± 0.6 kg/m². All participants were screened and determined to be healthy by medical history, physical examination, hematological and biochemical testing, and 12-lead electrocardiogram; none had unstable body weight (>3-kg weight gain/loss over the past 3 months), fasting plasma glucose ≤3.61 or ≥5.56 mmol/L, excessive alcohol or caffeine consumption, or used nicotine, illicit drugs, or medications known to influence substrate or energy metabolism within the last 3 months. The studies were approved by the Florida Hospital Institutional Review Board, and all participants provided informed written consent.

To determine the intraindividual variation of REE measurements in our small-room metabolic chamber, participants underwent two study visits 2–7 days apart. Participants arrived at the Translational Research Institute for Metabolism and Diabetes (TRI-MD) of Florida Hospital at ∼7:30 a.m. following an overnight 10-h fast. After vitals and metabolic weight were taken, participants entered one of two identical respiratory chambers, measuring 1.2 × 2 × 1.3 m. The chambers were in the resting bed configuration, which includes a raised platform, 30° angled wedge, and mattress resulting in a chamber volume of 3,200 L (Fig. 1). The chambers have a glass front with ports for intravenous infusions and blood sampling and are equipped with an intercom, a video camera, and a radar motion detector for monitoring the participant’s physical activity. Participants were placed in a Semi-Fowler position on the platform bed and instructed to stay as still as possible without falling asleep. Following a 30-min stabilization period, REE was measured for 60 min while participants were verified to be awake.

O₂ consumption and CO₂ production are calculated based on the amount of O₂ and CO₂ flowing into the chamber and the changes in O₂ and CO₂ concentrations in the calorimeter measured by Siemens analyzers (Siemens Ultramat/Oxymat 6; Siemens, Bartlesville OK). The flow rate into each chamber is controlled by a mass flow controller calibrated by MEI Research, Ltd. (Edina, MN) using flow metrology equipment that is traceable to standards set by the National Institute of Standards and Technology. For validation of the system, four low-flow mass flow controllers (also National Institute of Standards and Technology traceable through MEI) inject known quantities of N₂ (which displaces O₂) and CO₂ into a sealed chamber to simulate O₂ consumption and CO₂ production, respectively, demonstrating a sensitivity of 2%. Gas infusions over 1,335–2,001 min demonstrated continued reproducibility with coefficients of variation (CV) of 1.63 ± 3.97 and 1.55 ± 3.43% for VO₂ and VCO₂, respectively.

To investigate the effects of 13-h versus 3-h hyperglucagonemia on EE and glucose metabolism, participants underwent three study visits separated by 7 to 8 days. Prior to each of these visits, participants underwent a diet stabilization period of 2 days, in which they consumed standard weight–maintaining meals, consisting of 35% fat, 15% protein, and 50% carbohydrates. All meals, including snacks, were packed at TRI-MD for takeout. Participants were admitted to the Clinical Research Unit for an overnight stay in the evening prior to the metabolic assessments. Participants consumed a standardized dinner at 7:00 p.m. and subsequently fasted until the end of the study. At ∼7:30 p.m., an intravenous antecubital line was inserted in both arms for intravenous infusion and blood sampling. At ∼9:00 p.m., participants entered a respiratory chamber configured as described above but without the angle wedge for measurement of sleeping EE (SEE), REE, RQ, and substrate oxidation. Starting at 10:00 p.m., participants received a continuous infusion of 6,6-²H₂ glucose (12.9 ± 1.2 µmol/min [mean ± SD]) for measurement of glucose turnover in addition to one of the following three infusion protocols: GCG (Novo Nordisk, Bagsvaerd, Denmark) from −600 min (10:00 p.m.) to 0 min (8:00 a.m.) followed by GCG, octreotide (OCT; Fresenius Kabi, Lake Zurich, IL), and insulin (INS; Eli Lilly & Co., Indianapolis, IN) from 0 to 180 min (IP-A), saline (SAL) from −600 to 0 min followed by GCG/OCT/INS from 0 to 180 min (IP-B), or SAL from −600 to 0 min followed by SAL/OCT/INS from 0 to 180 min (IP-C) (Fig. 2). All GCG infusions were given at 6 ng/(kg × min), OCT was given at 30 ng/(kg × min), and INS was given at 0.07 mU/(kg × min) for baseline insulin replacement. The dose of the GCG infusion was selected based on published data indicating 7–14% increases in REE with doses of 3.3–6.0 ng/(kg × min) without adverse events (4,5) but risk of nausea and vomiting at ∼10 ng/(kg × min) (11). The infusion protocols A–C were given in a randomized, single-blind, balanced, crossover fashion. For a given participant, the same respiratory chamber was used throughout the entire study. Blood samples were obtained at −605, −600, −585, −570, −555, −540, −510, −480, −300, −60, −30, 0, 15, 30, 45, 60, 90, 120, 150, and 180 min (IP-A and IP-C). In IP-B, blood samples at −585 to −300 min were omitted to reduce the volume of blood draws. Blood samples were analyzed for plasma or serum concentrations of GCG, insulin, glucose, and fibroblast growth factor 21 (FGF21) as well as for 6,6-²H₂ glucose enrichments. Blood pressure and heart rate were measured at −605, −480, −300, −60, 0, 60, 120, and 180 min using a vital signs monitor. Urine was collected from −600 to 0 min and from 0 to 180 min for measurement of urea excretion rates.

Plasma GCG was measured by immunoaffinity liquid chromatography–mass spectrometry with a lower limit of quantitation of 0.78 pmol/L (12). Plasma insulin was measured using a multiarray human insulin kit (Meso Scale Discovery, Gaithersburg, MD). Plasma glucose was measured by a YSI 2300 STAT PLUS glucose analyzer (YSI Incorporated, Yellow Springs, OH). 6,6-²H₂ glucose enrichments were determined following conversion of glucose to its aldonitrile pentaacetate derivative and subsequent analyses by gas chromatography–mass spectrometry under ammonia chemical ionization conditions using a modified version of a previously published method (13). Serum FGF21 concentrations were measured by the Quantikine Human FGF-21 Immunoassay (R&D Systems, Minneapolis, MN) with a lower limit of quantitation of 15.6 pg/mL.

Rates of appearance (RA) and disappearance (RD) of plasma glucose were calculated by the non–steady-state equation of DeBodo et al. (14).

EE at zero physical activity was calculated as the y-intercept of the linear regression analysis between EE and the activity score measured by the chamber’s radar motion detector (15). Because of considerable variability in sleep and awake phases overnight among the infusion protocols, data from 4:00–5:00 a.m. were used for analysis of SEE when participants were most frequently in deep sleep and activity was the lowest (no activity in 51–53 out of 60 min in all three infusion protocols); for REE, data from 8:01–9:00 a.m., 9:01–10:00 a.m., and 10:01–11:00 a.m. were used, during which participants were always awake. Rates of protein oxidation were determined from the rates of urinary urea excretion, and the rates of carbohydrate and lipid oxidation were calculated according to Frayn (16).

Plasma GCG, insulin, glucose, and serum FGF21 concentrations during the overnight infusion were compared between IP-A and IP-C by paired Student t tests.

Repeated-measures ANOVA models were used to compare data among IP-A, IP-B, and IP-C. For comparison of SEE, treatment and period were the two main effects, whereas the participant was treated as random effect. For comparisons of the hourly REE, three main effects (treatment, period, and time) and two of their interactions (treatment × period and treatment × time) were included, whereas the participant was treated as a random effect, and three repeated measures nested in each visit were with autoregressive covariance structures. Post hoc pairwise comparisons were conducted with adjustment for multiple comparisons using the Tukey test for SEE and the step-down Šidák procedure for REE. Similar models were used for other variables.

All data are given as means ± SE unless otherwise noted. P values <0.05 were considered statistically significant. All statistics were performed using SAS (version 9.4; SAS Institute Inc., Cary, NC).

In the test-retest experiments in 10 healthy participants, REE averaged 1,618 ± 252 and 1,605 ± 238 kcal/24 h (SD) on the first and second measurement, respectively; RQ averaged 0.875 ± 0.075 and 0.879 ± 0.075, respectively. The difference between the first and the second measurement was −14 ± 53 kcal/24 h (CV 2.11 ± 1.08%) for REE and 0.004 ± 0.024 (CV 1.50 ± 1.09%) for RQ. The correlation coefficients were 0.978 and 0.949 for the duplicate measurement of REE and RQ, respectively (Fig. 3A). Power calculations indicate that studying six participants allows the detection of an ∼4% difference in REE and an ∼3% difference in RQ with 80% power in paired comparisons (Fig. 3B).

In response to the overnight GCG infusion in IP-A, plasma GCG increased to 80–90 pmol/L within 30 min and remained at similarly increased concentrations until the end of the infusion. In response to the SAL infusion in IP-C, plasma GCG concentrations did not change from baseline. Mean plasma GCG concentrations during the overnight infusion were 86 ± 6 pmol/L in IP-A and 5.2 ± 1.1 pmol/L in IP-C (P < 0.001). Compared with IP-C, IP-A increased plasma glucose ∼30% within 30 min of the GCG infusion; plasma glucose subsequently decreased such that mean concentrations during the last hour of the overnight infusion were ∼8% increased (P < 0.023; P = 0.08 after Tukey adjustment for multiple comparisons including IP-B) (Fig. 4 and Table 1). Plasma insulin concentrations increased to a peak at 60 min and subsequently decreased in IP-A, but decreased progressively in IP-C. During the last 60 min of the overnight infusion, mean plasma insulin concentrations were ∼70% greater in IP-A than in IP-C (28.3 ± 9.7 vs. 17.2 ± 6.6 pmol/L, P < 0.05; NS after Tukey adjustment) (Fig. 4). During that period, mean glucose RA was ∼35% greater in IP-A than in IP-C (P < 0.009 after Tukey adjustment), as was mean glucose RD (P < 0.011 after Tukey adjustment) (Fig. 4 and Table 1). Values of plasma GCG, glucose, insulin, glucose RA, and glucose RD from −60 to 0 min were similar in IP-B and IP-C.

During the following 3-h infusion of GCG/OCT/INS, plasma GCG remained at similarly increased levels as during the overnight GCG infusion in IP-A and increased within 45 min to levels comparable to those in IP-A in IP-B. Plasma GCG decreased to levels near or below limits of quantitation during the 3-h infusion of SAL/OCT/INS in IP-C (Fig. 4). In IP-A and IP-B, plasma glucose concentrations peaked at 60 min and subsequently decreased to a similar degree. However, peak and mean plasma glucose concentrations as well as glucose area under the curve (AUC) were significantly reduced in IP-A compared with IP-B. In IP-C, plasma glucose decreased to a nadir of 4.2 ± 0.2 mmol/L at 60 min and subsequently increased to levels similar to those at baseline. Accordingly, mean glucose levels and glucose AUC during the 3-h infusion were significantly greater in IP-A and IP-B compared with IP-C (Fig. 4 and Table 1). In contrast to plasma glucose levels, glucose RA was increased similarly in IP-A and IP-B, such that mean glucose RA and AUC glucose RA were comparable in both infusion protocols. In IP-C, glucose RA decreased to a nadir at 60 min and subsequently increased to rates similar to baseline consistent with the responses in plasma glucose concentrations. Thus, mean glucose RA and AUC glucose RA during the 3-h infusion were significantly greater in IP-A and IP-B than in IP-C (Fig. 4 and Table 1). Overall, similar results were found for glucose RD (Fig. 4 and Table 1). In IP-A and IP-B, glucose RD increased with a delay compared with glucose RA but equaled glucose RA at ∼60 min. From ∼90–180 min of the infusion, glucose RD slightly exceeded glucose RA. Plasma insulin concentrations were comparable during the 3-h infusions in five participants but were significantly increased IP-A and IP-B compared with IP-C in one participant, such that mean levels were ∼60% greater in IP-A and IP-B than in IP-C but without statistical significance (Fig. 4).

EE and RQ in response to the infusion protocols are shown in Fig. 5 and Table 2. For SEE and REE, there was a significant treatment effect but no significant period effect. SEE was increased in IP-A compared with IP-B and IP-C by ∼65–70 kcal/24 h (P < 0.023 and P < 0.017, respectively). The difference in SEE remained statistically significant after adjustment for multiple comparisons for IP-A versus IP-C (P < 0.041) but became marginally significant for IP-A versus IP-B (P = 0.053). SEE was similar in IP-B and IP-C (P = 0.91). Mean REE during the 3-h infusions was increased in IP-A compared with the SAL infusion in IP-C by ∼50 kcal/24 h (P < 0.007; P < 0.015 after adjustment for multiple comparisons). This was because of increased REE during the first 2 h but not the last hour of the 3-h infusion, as REE tended to progressively increase in IP-C. REE increased during the GCG/OCT/INS infusion in IP-B with a statistical trend (P < 0.069), but mean REE in IP-B and IP-C were not significantly different.

There were no differences in RQ, carbohydrate, or fat oxidation at 4 to 5 a.m. or during the 3-h infusions among IP-A, IP-B, and IP-C. However, overnight protein oxidation calculated from urinary nitrogen excretion was greater in IP-A than in IP-B (0.069 ± 0.003 vs 0.053 ± 0.005 g/min; P < 0.023) and IP-C (0.051± 0.004 g/min; P < 0.013); no differences were detected during the 3-h infusions among the three infusion protocols.

Serum FGF21 concentrations tended to decrease during the overnight GCG infusion but tended to increase during the overnight SAL infusion in IP-C such that mean FGF21 levels were approximately twofold lower in IP-A compared with IP-C (48 ± 16 vs. 95 ± 28 pg/mL; P < 0.02). During the 3-h infusions, serum FGF21 responses continued to be suppressed in IP-A, but did not differ significantly between the three infusion protocols (Fig. 6).

Systolic blood pressure and heart rate did not differ among the three infusion protocols during the overnight or the 3-h morning infusion. However, mean diastolic blood pressure was significantly lower during the overnight infusion in IP-A (64.0 ± 2.4 mmHg) compared with IP-C (71.6 ± 3.2 mmHg; P < 0.006) and significantly lower during the 3-h morning infusion in IP-A (67.1 ± 2.0 mm Hg) compared with IP-B (72.6 ± 2.3; P < 0.031) and IP-C (74.0 ± 4.2 mm Hg; P < 0.018). No adverse events were recorded at any time of the experiments.

Using a combination of tracer techniques and room indirect calorimetry, the current study demonstrates for the first time that GCG persistently stimulates not only HGP but also EE for several hours.

Consistent with previous observations, overnight infusion of GCG in IP-A resulted in transient hyperglycemia, such that during the last hour of the 10-h infusion plasma, glucose concentrations were only slightly elevated compared with the SAL infusion in IP-B and IP-C. Nevertheless, during that period, glucose RA was increased ∼35–40%. Glucose RA subsequently increased during the following GCG/OCT/INS infusion in IP-A to similar rates as in IP-B, in which the GCG/OCT/INS infusion was preceded by an overnight SAL infusion. Because systemic glucose RA consists of hepatic and renal glucose production but GCG stimulates glucose production only by the liver (17), these data demonstrate comparable hepatic GCGR activation and downstream effects in response to 13- and 3-h hyperglucagonemia, when the normal increase in insulin secretion is inhibited. Accordingly, the current study provides further evidence against an adaptation of the liver to GCG as proposed by some earlier studies (10,18,19).

We found that SEE, measured at 6 to 7 h of the overnight infusion, was increased by ∼65–70 kcal/24 h in response to GCG compared with SAL. Mean REE, measured at 10–13 h of the infusions, was increased by ∼50 kcal/24 h in response to the GCG/OCT/INS that followed the overnight GCG infusion in IP-A compared with the SAL/OCT/INS infusion in IP-C. Although this was a result of increased REE in the first 2 h but not the last hour of the infusion, these data indicate that GCG-induced stimulation of EE is sustained for a minimum of 12 h at least under the present experimental conditions. The increase in SEE and REE by ∼50–70 kcal/24 h (or 3 to 4%) in response to our GCG dose of 6 ng/(kg × min) is indisputably of marginal clinical significance. Higher GCG doses of 50 ng/(kg × min) increased REE acutely by ∼10–14% (6,7), signifying a potential dose-response relationship. However, such doses are associated with greater risk of nausea and whether such doses unabatedly stimulate EE to a greater degree than in the current study is unclear. Therefore, further research is needed to identify the optimal dose of GCG for sustained stimulation of EE at acceptable safety and tolerability, including its use in combination with GLP-1 analogs or other weight loss drugs to boost weight loss efficacy.

In previous studies, infusion of GCG at 3.3–6 ng/(kg × min) with somatostatin for 2 to 3.5 h was found to increase REE ∼7–14% compared with the somatostatin control experiment by canopy indirect calorimetry (4,5). In the current study, using participants of similar characteristics and number, infusion of GCG/OCT/INS for 3 h after overnight SAL infusion increased REE progressively, but values were not significantly increased compared with the SAL/OCT/INS control experiment. A potential explanation for this finding (and lack of increased REE in the third hour of GCG/OCT/INS infusion in IP-A) compared with our control experiment may be the progressive rise in REE in the latter. Contrary to this rise, REE tended to decrease during the somatostatin infusion in the control experiment of previous studies (4,5), but whether differences between the selective somatostatin 2 receptor agonist OCT used in the current study and somatostatin explain these divergent results is unclear.

We found that overnight GCG infusion, which raised plasma GCG to ∼80–90 pmol/L, suppressed serum FGF21 levels and that serum FGF21 was if anything lower during the GLUC/OCT/INS infusion in IP-A than during the SAL/OCT/INS infusion, indicating that circulating FGF21 was not involved in the increased EE. This finding conflicts with reports by Arafat et al. (20) and Habegger et al. (21) that GCG 1 mg administered intramuscularly led to significantly increased plasma FGF21 concentrations in healthy lean individuals and obese individuals, respectively. However, in the latter study and in contrast to the current study, plasma FGF21 levels were markedly below the lower limit of quantitation of the assay used and GCG levels were not reported. In former study (20), plasma FGF21 levels were just above the detection limit after GCG administration but below it at baseline; moreover, the reported increase in plasma GCG to ∼100 pmol/L is ∼20-fold lower than described for 1 mg intramuscular GCG in the Food and Drug Administration label (22). Therefore, whether GCG truly increases circulating FGF21 and at what dose in humans appears to remain uncertain. Additional possible explanations for the inconsistent findings between the present and previous studies (20,21) are differences in the route of GCG administration leading to different pharmacokinetics and differences in the study populations, including an exclusively male population in the current study versus a mixed-sex population (20). We would like to emphasize that our findings do not preclude an involvement of FGF21 in GCG-stimulated EE. GCG has been shown to not only increase glucose cycling (23) but also stimulate thermogenesis in brown adipose tissue (24), which may be mediated through locally increased FGF21 production (20). Furthermore, it is possible that the greater efficacy of GCG to stimulate EE in previous studies (6,7) was mediated through increased FGF21 and that its absence explains the relatively low efficacy in the current study. Further research is hence needed to examine the regulation of FGF21 by GCG and its role in GCG-induced EE in humans.

A strength of the current study is the randomized, crossover design, but a limitation is that relatively few participants were studied. However, as a result of the high reproducibility of our chamber measurements, the study was adequately powered to detect changes in EE of ∼4% or 60 kcal/24 h, which is of borderline clinical significance. Another potential limitation is that we studied only nonobese healthy males to avoid the influence of medical conditions associated with obesity, concomitant medications, and the menstrual cycle on EE. To our knowledge, though, there is no evidence suggesting that the comparison of 13- to 3-h hyperglucagonemia, the primary outcome of the study, was affected by the study population. Indeed, short-term infusion of GCG at 50 ng/(kg × min) increased EE virtually equally in overweight/obese (mean BMI 29.3 kg/m²) and lean (mean BMI 22.5 kg/m²) young individuals (6,7), indicating that adiposity does not substantially influence the magnitude of GCG-stimulated EE. Nevertheless, additional studies may be needed to further examine the effect of GCG on EE in other populations, particularly obese and female individuals.

In summary, the current study provides evidence that in healthy nonobese individuals, stimulation of HGP and stimulation of EE are sustained during hyperglucagonemia of longer duration when insulin secretion is inhibited. The increase in EE at the GCG dose of 6 ng/(kg × min) was of marginal clinical significance and not mediated by increased circulating FGF21 levels.

Acknowledgments. The authors thank the TRI-MD staff for assistance and especially thank the study participants.

Funding. The study was supported by Merck & Co., Inc.

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

Author Contributions. M.C., M.E.L., K.B., and S.R.S. designed the study, researched the data, and reviewed and edited the manuscript. S.Pa. and S.Pr. researched the data and reviewed and edited the manuscript. A.Y.H.L., Y.C., J.S., S.Y., C.B., F.Y., J.M., and E.W.-K. researched the data. C.M. designed the study, researched the data, and wrote the manuscript. C.M. 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.