High Doses of Exogenous Glucagon Stimulate Insulin Secretion and Reduce Insulin Clearance in Healthy Humans

Glucagon, discovered originally as a hyperglycemia-causing contaminant in the process of insulin purification, is generally considered to be a counterregulatory hormone that opposes the actions of insulin (1–3). Glucagon stimulates glycogenolysis and gluconeogenesis in hepatocytes, and the rise in plasma glucagon in response to a fall in blood glucose supports a role in hypoglycemic counterregulation. Circulating glucagon is derived almost entirely from pancreatic α-cells and promotes hepatic glucose production (HGP) through the glucagon receptor (GCGR) that is expressed most densely in the liver. The current view of metabolic regulation holds that glucose homeostasis is maintained through the balanced actions of insulin and glucagon on HGP, as well as the actions of insulin to mediate disposal of glucose into peripheral tissues. However, this model assumes that glucagon is secreted primarily at fasting or lower levels of glycemia and does not account for glucagon release after meals (4). Although frequently overlooked because experimental hyperglycemia alone suppresses plasma glucagon, there is abundant evidence that glucagon is released in response to protein-containing meals (5–8), e.g., most meals consumed by humans.

Insulinotropic actions of glucagon were described >50 years ago (9,10), but after some initial interest, this effect lay fallow until recently. In the 1970s, glucagon was used clinically to test β-cell function in people with diabetes, and in those with insulinopenia there was disproportionate HGP that overwhelmed the minimal insulin response and caused hyperglycemia (11). These observations, combined with greater understanding of glucagon signaling in the liver, focused attention on the effects of glucagon to raise blood glucose. Thus, for most of the past 5 decades, the predominant action ascribed to glucagon has been hypoglycemic counterregulation, and the focus of glucagon-related drug development has been antagonism of the GCGR to lower blood glucose (12). While some GCGR antagonists showed promising efficacy, none progressed to clinical use because of adverse effects (13,14).

In recent years, mechanistic studies in animal models and isolated rodent and human islets (15,16) have rekindled interest in the insulinotropic actions of glucagon. Studies in mice have supported a model whereby two divergent actions of glucagon are dependent on glycemic status: During fasting, glucagon acts predominantly at the liver to raise blood glucose, whereas in the fed state when glycemia is elevated, the insulinotropic effects predominate and lower blood glucose (17). These mechanistic findings have not yet been extended to human physiology. In the studies described herein, we tested the hypothesis that in healthy humans, glucagon given at fasting euglycemia (∼90 mg/dL) would have a predominant action to increase endogenous glucose production (EGP), while during experimental hyperglycemia (∼150 mg/dL), the effect of glucagon would be more pronounced to stimulate insulin secretion. We used a glucose tracer to calculate EGP, and simultaneous measurements of C-peptide and insulin to calculate insulin secretion rates (ISRs). The study used healthy volunteers in a crossover design to evaluate these outcomes.

Healthy men and women aged 18–65 years were recruited from the Durham, North Carolina, area. Inclusion criteria included fasting plasma glucose ≤95 mg/dL or hemoglobin A_1c (HbA_1c) ≤5.9% (41 mmol/mol) at the screening visit. Subjects with diabetes; diabetes among first-degree family members; significant, chronic gastrointestinal (GI), hepatic, renal, vascular, or inflammatory disease; and medications that alter glucose metabolism or GI function were exclusion criteria. These studies were approved by the institutional review board at Duke University, and all participants provided written informed consent prior to the studies. Study procedures were conducted at the Duke University Center for Living.

Glucagon was prepared using commercially available glucagon rescue pens (Eli Lilly). Briefly, diluent was injected into 1 mg lyophilized glucagon per the manufacturer’s recommendations at bedside. Then, the glucagon was diluted in saline (with whole blood added to 2% volume to prevent the peptide adhering to plastic) ∼15 min prior to infusion. [6,6-²H₂]Glucose (Cambridge Isotopes) had >95% heavy isotope incorporation and was prepared as a 3.5% solution in saline for infusion. Dextrose infusion during the hyperglycemic clamp study was 20% dextrose labeled to 2% [6,6-²H₂]glucose.

Subjects were instructed to maintain their normal diet and exercise routine prior to study visit. On study days, subjects arrived at the research site between 0730-0800 h after an overnight fast (no food after 2100 h). Intravenous catheters were placed in each forearm for blood sampling and infusions. The arm used for blood sampling was warmed to maintain blood flow and arteriovenous admixture.

The study design is depicted in Fig. 1A. After a blood sample was taken to measure isotopic background, a primed-continuous infusion of [6,6-²H₂]glucose was initiated and maintained throughout the course of the study (a 4 mg/kg over 5 min prime followed by continuous 0.04 mg/kg/min). Ninety minutes after the tracer bolus, fasting samples were taken at −30, −20, −10, and 0 min. At time 0, subjects either remained on tracer infusion alone (saline study) or had additional infusion of deuterated glucose–labeled 20% dextrose to achieve and maintain a glycemic target of 150 ± 10 mg/dL (hyperglycemic clamp study); glycemia was monitored at the bedside using a glucose analyzer (YSI 2300 STAT Plus; Yellow Springs Instruments, Yellow Springs, OH) to monitor plasma glucose and maintain the hyperglycemic target during clamp admission. Glucometers (StatStrip Xpress 2; Nova Biomedical, Waltham, MA) were used in parallel on whole blood for comparison. Baseline sampling during both days of study occurred at 60, 70, 80, 90, 100, 110, and 120 min. After the 120-min sample, an infusion of glucagon (100 ng/kg/min) was initiated, with sampling continuing every 10 min until the end of the study. The first four subjects received glucagon for 60 min for both saline and hyperglycemia clamp admission, all reported GI side effects after completing the infusions, and one did not complete the saline admission. The protocol was adapted, and the next five subjects received a shorter, 30-min glucagon infusion. Data from all nine subjects were analyzed during the period where glucagon duration was the same (i.e., the first 30 min of glucagon infusion) for both their saline and hyperglycemic clamp admission (121–150 min). The data from the last 30 min of the studies (151–180 min) are presented but were considered exploratory and were not formally analyzed. Subjects were randomly assigned and counterbalanced to receive either saline or glucose clamp on their first visit, with the opposite treatment on their second visit. Blood samples were collected in EDTA tubes containing an antiprotease cocktail with aprotonin and diprotin A. Samples held on ice were spun down for plasma within 60 min of collection, and plasma samples were stored at −80°C until assay.

Plasma enrichment of [6,6-²H₂]glucose was measured by liquid chromatography–tandem mass spectrometry (LC-MS/MS) as previously reported (18). Plasma immunoreactive insulin and C-peptide were measured by a multiplexed LC-MS/MS method (19) and compared with commercially available ELISA kits (Alpco, Salem, NH). Plasma glucagon was measured by ELISA (Ansh Labs, Webster, TX) (20) and by LC-MS/MS. The latter used monoclonal antibodies produced at the Fred Hutchinson Cancer Center Antibody Technology Facility (21) to immuno-enrich glucagon from 200 μL of plasma that had been protein depleted with ethanol on ice (22).

EGP and R_d were calculated using Steele equations for nonsteady state (23,24), with a distribution volume of 200 mL/kg body weight and a correction of 0.65. Within each period, the slope of the change of tracer:tracee ratio over time was calculated and used in nonsteady-state equations. Integrated area under the curve (iAUC) was calculated using GraphPad Prism software, with the fasting values for each day used as the baseline. Baseline iAUC used the last 30 min of baseline (90–120 min) to allow for direct comparison with the 30-min glucagon infusion. We applied a combined modeling strategy using a two-compartment model for C-peptide and a one-compartment model for insulin (25), with incorporation of kinetic constants of C-peptide (26) and implementation in WinSAAM (27) to estimate the ISR, whole-body insulin clearance, and first-pass hepatic extraction of insulin. Insulin half-life (t_1/2) was calculated from the fractional clearance rate. Metabolic clearance rate of insulin (MCR_i) was calculated by first binning ISR into 10-min intervals and then dividing by insulin concentrations over those periods to estimate MCR_i at each blood sampling point.

Given the shortened (30-min) glucagon infusion in five subjects, data were compared from −30 to 150 min (including 30 min of glucagon infusion). Fasting values included −30-, −20-, −10-, and 0-min samples; baseline values were from 60-, 70-, 80-, 90-, 100-, 110-, and 120-min samples; and glucagon values included 130-, 140-, and 150-min samples. Data collected between 150 and 180 min, i.e., the last 30 min of glucagon infusions, are only available for four subjects. These data are shown as temporal depictions but were not used in statistical analyses given the small sample size. To detect a fourfold increase in EGP and ISR, we used data from previous studies in healthy subjects (28,29) to predict that 10 subjects were needed (α = 0.05, β = 0.95). Nonnormally distributed data were log-transformed prior to analysis. Data were analyzed using a mixed linear model with repeated measures with Tukey adjustment for post hoc comparisons (SAS, University Edition). Preplanned comparisons were fasting, baseline, and glucagon periods within and between conditions of saline and hyperglycemic clamp. Data are reported in tables using least squares means with SEs.

To test agreement between the YSI glucose analyzer and StatStrip glucometer, which have been validated in pediatric subjects (30), single blood samples were simultaneously measured on two YSI analyzers (two YSI measurements) and duplicated on two glucometers (four glucometer measurements) (Supplementary Methods, Supplementary Fig. 1, and Supplementary Table 1). Correlation between the measurements and agreement by Bland-Altman analysis were tested using GraphPad Prism (version 9.0). This comparison is included here as methodological information for use in glucose clamps; glucose determinations from whole blood measured with StatStrips can be done within seconds, while YSI measurements from plasma require minutes. Despite a slight positive bias, the reduced time to measure glycemia with the StatStrip glucometers (10–20 s vs. 90 s) allowed for rapid adjustments in glucose infusion rate (GIR) to maintain glycemic targets.

The data sets generated and/or analyzed during the current study are available from the corresponding author upon reasonable request.

Subjects included nine adults (one male, eight females) aged 28.7 ± 10.5 years without diabetes (fasting glucose 89 ± 8 mg/dL, HbA_1c 5.1 ± 0.3%) (Table 1). One subject did not complete the saline study, leaving a final set of eight saline and nine clamp studies for analysis. Subjects participating had variable GI side effects, including anorexia, nausea, vomiting, and/or GI urgency. None of these occurred during the glucagon administration but, rather, were reported 30–90 min following cessation of the infusions. This range of side effects has been described for clinical and research use of glucagon and was expected for some people on this basis and discussed in the consent process. The GI side effects were transient and self-limited. Nevertheless, these necessitated an adaptation from a 60- to a 30-min glucagon infusion to attempt to manage these events.

Fasting glucagon was similar between the saline and hyperglycemic clamp admissions at ∼15 pmol/L, and these did not change during the saline baseline period. The hyperglycemic clamp caused a small, but detectable, decrease in fasting glucagon (10.6 pmol/L, P < 0.05). Glucagon infusion resulted in concentrations of ∼1.1 nmol/L during both the saline and hyperglycemic clamp conditions (Supplementary Fig. 2). To confirm these concentrations, a small subset of samples were measured by both ELISA and LC-MS/MS (Supplementary Table 2); results from both assays were compatible and indicated a reduction in glucagon during the hyperglycemic clamp, although the ELISA did not read as low as the LC-MS/MS assay; both methods detected nanomolar glucagon concentrations during glucagon infusions.

Fasting blood glucose was comparable on the 2 study days (86.7 vs. 88.4 mg/dL) (Table 2). During the saline study, blood glucose decreased slightly to 84.4 mg/dL during the baseline period. Glucagon infusion increased blood glucose linearly and achieved a maximum concentration of 138.6 ± 6.6 mg/dL (range 106–164 mg/dL), for an average concentration of 120.2 mg/dL across the 30-min glucagon infusion (Fig. 1B and Table 1). Those who continued to receive glucagon from 150–180 min continued to maintain high glucose concentrations (150.7 ± 6.3 mg/dL), while those who had glucagon ceased at 150 min dropped to 126.2 ± 11.8 mg/dL (Supplementary Fig. 3A).

During the hyperglycemic clamp study, blood glucose was maintained at the target of 150 ± 10 mg/dL with a coefficient of variation of 4.6 ± 0.84% (Table 2 and Fig. 1C). This target was maintained with an average GIR of 4.6 ± 0.6 mg/kg/min during the baseline period, which was decreased to 3.4 ± 0.7 during glucagon infusion (P = 0.07 vs. baseline, paired t test). In the four subjects who continued receiving glucagon from 150–180 min, a GIR of ∼8 mg/kg/min was needed to maintain the glycemic target (Supplementary Fig. 3B). The five subjects who stopped receiving glucagon at 150 min also required an increase in GIR to maintain target glycemia, reaching ∼20 mg/kg/min at the end of the procedure (Supplementary Fig. 3B). These changes in GIR likely reflect the distinct dynamics of glucagon and insulin action, with the more rapid onset of glucagon-stimulated EGP reducing the need for exogenous glucose to maintain the clamp before the effects of hyperinsulinemia drove the glucose requirement upward ∼15 min later. Subjects who stopped receiving glucagon required a higher GIR from 150–180 min to maintain the glycemic target compared with those with continuous infusion of glucagon (Supplementary Fig. 3B), suggesting that cessation of glucagon quickly lowered EGP.

Primed-continuous infusion of [6,6-²H₂]glucose achieved ∼2% enrichment during both saline and clamp studies at fasting and baseline (Supplementary Fig. 4A and B). Fasting EGP was comparable on both study days (∼2.3 mg/kg/min) (Table 2 and Fig. 2A). During the saline study, EGP remained consistent from fasting throughout the baseline period, and glucagon administration rapidly and steadily increased EGP (Fig. 2A and Supplementary Fig. 4C). Hyperglycemic clamp conditions decreased EGP to 1.0 mg/kg/min (P < 0.05 vs. fasting, P < 0.05 vs. saline baseline) (Table 2 and Fig. 2A and B), but similar to the saline study glucagon administration rapidly and steadily increased EGP (Fig. 2A). Glucagon given during the saline study, increased EGP to ∼5.1 mg/kg/min (P < 0.0001 vs. baseline) (Table 2 and Fig. 2B), while glucagon given during hyperglycemic clamp increased EGP to ∼4.0 mg/kg/min (P < 0.0001 vs. baseline, P < 0.05 vs. saline + glucagon) (Table 2 and Fig. 2B). Under both saline and hyperglycemic clamp conditions, for subjects who had continued glucagon infusion from 150 to 180 min, EGP remained high, while it decreased in subjects who had glucagon infusion stopped (Supplementary Fig. 4C). When EGP was adjusted for baseline on each admission, glucagon infusion caused a 2.5-fold increase in EGP during saline and an almost fourfold increase during hyperglycemic clamp.

R_d was comparable at fasting between saline and clamp admissions (Fig. 2C). During the hyperglycemic clamp study, R_d increased to ∼5.1 mg/kg/min during the baseline period (P < 0.001 vs. saline baseline) (Fig. 2C and D). Under both conditions, glucagon administration decreased R_d initially, with a subsequent rise through the end of the study (Fig. 2C). Averaging R_d over the baseline and glucagon periods did not reveal glucagon-induced changes within the studies; instead, R_d was significantly impacted by the glycemic condition (i.e., saline vs. hyperglycemic clamp) (Fig. 4D). R_d continued to increase from 150 to 180 min, regardless of whether glucagon administration was stopped (Supplementary Fig. 4D). R_d was influenced by glycemia (saline vs. hyperglycemic clamp), regardless of glucagon infusion, whereas EGP was influenced by glucagon regardless of starting glycemia.

Initial measurement of insulin and C-peptide used a commercially available ELISA; however, we observed near-equimolar concentrations of C-peptide and insulin in several subjects during hyperglycemic clamp conditions (Supplementary Fig. 5A, specifically at times 130–150 min) despite the known differences in clearance of these proinsulin products (31–33). This finding persisted despite diluting samples, suggesting that one of the assays had systematic bias. To rectify this concern, an alternative assay method using an LC-MS/MS multiplex approach was used (19). Insulin concentrations were comparable between the ELISA and LC-MS/MS assays (Supplementary Fig. 5B); however, C-peptide concentrations were approximately two times higher when measured by LC-MS/MS (Supplementary Fig. 5C). The LC-MS/MS assays were therefore used to generate the insulin and C-peptide results presented in this study.

Fasting C-peptide and insulin concentrations were ∼460 pmol/L and 58 pmol/L, respectively, in both the saline and hyperglycemic clamp studies (Fig. 3 and Table 3). With continued fasting during the saline study, there was a slight decrease in C-peptide to 405 pmol/L (P < 0.01 compared with fasting) (Fig. 3A) without a significant change in plasma insulin (57.6 vs. 54.0 pmol/L, P = 0.4) (Fig. 3C) or ISR (27.1 vs. 24.4 pmol/min) (Fig. 3E). Glucagon infusion at euglycemia elicited a steady increase in insulin, C-peptide, and ISR (Fig. 3A, C, and E, respectively, and Table 3) that paralleled the rise in blood glucose (Fig. 1B). Continued glucagon infusion from 150 to 180 min continued to increase C-peptide and insulin, while cessation of glucagon infusion led to gradual decreases in β-cell output (Supplementary Fig. 6A, C, and E).

The hyperglycemic clamp caused rapid three- to fourfold elevations in C-peptide, insulin, and ISR compared with fasting (Fig. 3B, D, and F, respectively). Glucagon administration elicited an immediate, robust increase in C-peptide (from 1,698 ± 191 at 120 min to 3,748 ± 415 pmol/L at 130 min) and insulin (from 278 ± 144 at 120 min to 1,497 ± 797 pmol/L at 130 min) concentrations, and these levels remained elevated throughout the infusion (Fig. 3 and Table 3). ISR increased rapidly during glucagon administration, with a plateau after ∼10 min of glucagon (Fig. 3F). In the subjects who had continued glucagon infusion from 150 to 180 min, high concentrations of C-peptide and insulin were maintained, while cessation of glucagon at 150 min caused a gradual diminution of β-cell secretions (Supplementary Fig. 6B, D, and F).

When analyzing β-cell secretion during these studies, it was noted that the C-peptide: insulin ratio decreased in response to hyperglycemia and glucagon administration. The reduced ratio was driven by an increase in insulin rather than decreased C-peptide (Table 3), suggesting a change in the rate of insulin removal from the circulation. Despite interindividual variability in the C-peptide:insulin ratio throughout the fasting and baseline periods (Fig. 4A and C), the decrease observed during glucagon administration was immediate (Fig. 4A and B) and occurred in all subjects (Fig. 4C). To examine this result, we used several approaches to assess insulin clearance, including computing the C-peptide:insulin ratio, calculating MCR_i from ISR/insulin concentrations, and using a mathematical model to calculate insulin t_1/2.

In examining the C-peptide:insulin ratio throughout our studies, both hyperglycemia and glucagon appeared to decrease the ratio (Fig. 4A and B). Hyperglycemia alone reduced the C-peptide:insulin ratio from 10.1 ± 1.1 during fasting to 7.3 ± 0.8 (Fig. 4C and Table 4). Glucagon administration, beginning both at fasting euglycemia and hyperglycemia, decreased the ratio to 4.0 ± 0.5 and 2.8 ± 0.4, respectively (P < 0.0001 for comparisons with respective baselines) (Table 4). The effect of glucagon on the ratio of β-cell peptides is depicted (Fig. 4A and B) in the responses of subjects with longer and shorter infusions. In the subjects with extended glucagon infusions (time 150–180 min), glucagon maintained a suppressed C-peptide:insulin ratio, while for those who had glucagon stopped at 150 min, the ratio began to increase toward baseline values under both saline and hyperglycemic clamp conditions.

Calculation of MCR_i supported a decrease in insulin clearance in response to glucagon at hyperglycemia, with a 50% decrease from baseline (0.53 ± 0.05 at hyperglycemia alone to 0.27 ± 0.05 during hyperglycemia with glucagon). This effect of glucagon to decrease in MCR_i was greater when the infusion was started at hyperglycemia compared with euglycemia (0.44 ± 0.05 during saline study, P = 0.003). In the saline study, glucagon infusion did not significantly decrease MCR_i compared with saline baseline (Fig. 4D and Table 4). Insulin t_1/2 and the fraction of insulin reaching the periphery were calculated with a mathematical model (25). During fasting glycemia or hyperglycemia during the glucose clamp, insulin t_1/2 was ∼3 min, and glucagon infusion during both conditions increased insulin t_1/2 to 7.9 ± 0.9 min during the saline study and to 7.4 ± 0.8 min during the hyperglycemic clamp (Fig. 4E and Table 4). Taken together, these results suggest that glucagon alters the kinetics of insulin clearance.

The GCGR is homologous to the receptors for the incretins glucagon-like peptide 1 (GLP-1) and glucose-dependent insulinotropic polypeptide (GIP) and like the incretin receptors, is expressed by pancreatic β-cells. In mice, we observed that the systemic effect of glucagon on blood glucose was dependent on whether it was given in the fed or fasting state (17). In the current study, we sought to determine whether the insulinotropic properties of glucagon were dependent on ambient glycemia in humans, testing the hypothesis that glucagon shared insulinotropic characteristics with GLP-1 and GIP. Recent studies indicated that β-cell glucagon signaling is important for nutrient-stimulated insulin secretion in murine and human islets and that this effect may be a normal component of the regulatory response to feeding (15–17,34). In this proof-of-concept study, we found that glucagon has significant actions to raise peripheral insulin concentrations both by augmenting glucose-stimulated insulin secretion and reducing first-pass hepatic clearance of insulin. Taken together with a large body of evidence that α-cell secretion is stimulated during the absorption of protein or mixed-nutrient meals (5–7), these findings support a role for glucagon in postprandial glucose disposition.

Several features of our experimental design merit comment. Our protocol established two distinct glycemic settings to test the hepatic and β-cell effects of glucagon in our healthy human volunteers. Each subject had both a saline and a clamp study on separate days and served as their own control. During each study day, subjects were at near-steady state for glucose, ISR, EGP, and glucagon during baseline periods (90–120 min) within the saline and clamp protocols; this provided distinct euglycemic and hyperglycemic conditions from which to assess glucagon actions. We used a dose of glucagon that was substantially greater than those typically used to mimic plasma levels in the physiologic range (100 compared with 0.5–50.0 ng/kg/min) (35–38) to try to mimic the nanomolar concentrations we previously estimated were operative in the islet (15). Finally, we measured insulin, C-peptide, and glucagon using novel, validated LC-MS/MS assays to optimize specificity (19). Results from these measures suggest that beyond stimulation of their release, glucagon affected the relative concentrations of insulin and C-peptide, suggesting differential insulin clearance, which was tested using a robust array of calculations of insulin dynamics.

The primary goal of this study was to compare the relative effects of glucagon on insulin secretion and glucose production at euglycemia and moderate hyperglycemia meant to approximate the prandial state. During the saline study, small, yet significant decreases in β-cell secretion were observed throughout the baseline period, likely because of extended fasting. Glucagon infusion at fasting glucose increased glycemia linearly, and C-peptide, insulin, and ISR rose concurrently, reaching levels after 30 min (time = 150 min) that exceeded β-cell secretion in response to the hyperglycemic clamp alone. In contrast, starting the glucagon infusion during moderate experimental hyperglycemia caused an abrupt, rapid rise of C-peptide, insulin, and ISR to peaks three- to fourfold greater than what was achieved during the saline study. These findings indicate an interaction of elevated glucose and glucagon to stimulate insulin secretion, with the differences in the shapes of the curves for insulin, C-peptide, and ISR in the hyperglycemia and saline studies suggesting distinct β-cell responses at lower and higher levels of increased glycemia. However, because glucagon infusion at euglycemia stimulates EGP and raises blood glucose, it is not possible to determine whether the insulinotropic effects are as strictly glucose dependent as those of GIP and GLP-1 (39).

A paracrine model of glucagon regulation of β-cell function raises the possibility for tonic as well as episodic mediation of insulin secretion. Indeed, studies with ex vivo islets support α-cell products as important for β-cell tone (15,40,41), suggesting that local glucagon signaling in the fasting state has a role in islet function. In isolated mouse and human islets, most of the stimulatory action of glucagon is mediated by the GLP-1 receptor (15), a finding now supported by data from the human study of Farahani et al. (42), who also reported that modest elevation of ambient glucose promotes glucagon-stimulated insulin release. Of note, the counterregulatory rise in glucagon does not increase C-peptide during hypoglycemia (43,44). Thus, it is reasonable to infer that the distinct patterns of β-cell secretion during the protocols in the current study indicate a strong interaction between glucagon and glycemia.

Before glucagon was used clinically to counteract insulin-induced hyperglycemia, it was used to test β-cell secretory capacity in patients with diabetes (45). This application stemmed from pioneering work by Samols et al. (9), who first noted enhanced insulin secretion in a subset of patients with insulinoma whom they treated with glucagon, and then replicated this effect in healthy subjects. In their study, using large (0.25- and 1-mg) boluses of glucagon and 1-min blood sampling, they were able to demonstrate increases in plasma insulin that were distinct from the effects of hyperglycemia due to HGP. These investigators demonstrated direct insulinotropic effects of glucagon that were dose-dependent and proposed that intraislet stimulation of β-cells by α-cells had physiologic importance during meals (10). In retrospect, the latter conjecture was prescient given recent findings supporting important α-cell-to-β-cell communication for insulin secretion (15,16,40). Other groups, also working 30–50 years ago, reported findings compatible with our recent observations, including enhanced glucagon-stimulated insulin release at higher blood glucose levels and relatively high glucagon infusion rates necessary to activate β-cell secretion (46–48). Thus, the results presented here confirm and extend work that has been forgotten for the most part, buried by the many subsequent developments in incretin biology (49) and a consensus view of glucagon framed primarily as a hormone opposing the actions of insulin (50).

When embarking on these studies, we planned to use glucagon action at the liver as a positive control since under fasting conditions, it potently stimulates EGP (29). We hypothesized that glucagon action to raise blood glucose would predominate when given during fasting euglycemia and did not anticipate the glycogenolytic actions during concomitant experimental hyperglycemia and hyperinsulinemia. Previous work has demonstrated that at fasting glycemia (5 mmol/L), insulin acts as a brake on EGP in a dose-dependent manner, with insulin concentrations as low as 90 pmol/L reducing EGP (51). In fact, during our hyperglycemic clamp conditions, we achieved insulin concentrations of ∼250 pmol/L and noted a modest but significant decrease in EGP. However, at neither the clamp level of insulinemia nor the massive concentrations (>1,300 pmol/L) achieved immediately after glucagon plus hyperglycemia was glucagon-stimulated EGP significantly suppressed (Fig. 2 and Table 2). Interestingly, the peak glucagon levels during our studies approximated those of circulating insulin (∼1.1 nmol/L glucagon vs. 1.3–1.8 nmol/L insulin), suggesting that at near-equimolar concentrations, the effects of glucagon to regulate HGP predominate over those of insulin. While equimolar portal levels of glucagon and insulin rarely occur in normal physiology, it seems likely that one:one ratios may be approached in some diabetic states.

During our assessment of insulin secretion, we noticed large, consistent changes in the molar ratios of C-peptide to insulin coincident with starting and stopping the glucagon infusion, prompting us to consider whether insulin clearance was being altered. While we did not design this study to test insulin clearance, our analyses indicate that glucagon causes hyperinsulinemia by reducing insulin removal as well as stimulating insulin secretion. In this study, the hyperglycemic clamp alone caused a small decrease in the C-peptide:insulin ratio compared with fasting baseline, in keeping with a previous report (52), but did not alter the MCRi or circulating insulin t_1/2. In the euglycemic saline study, glucagon infusion increased C-peptide to levels that were slightly higher than those measured under clamp baseline conditions, while insulin concentrations were nearly twofold greater (∼488 pmol/L during saline + glucagon vs. ∼250 pmol/L during clamp baseline) (Table 3). There was a similar degree of change in the C-peptide:insulin ratio before and after glucagon administration during hyperglycemia. Thus, glucagon, given at either euglycemia or hyperglycemia, caused an ∼60% reduction in the C-peptide:insulin ratio. Because the circulatory kinetics of insulin and C-peptide differ, the C-peptide:insulin ratio provides only suggestive evidence of variable insulin clearance (53). However, we also noted that during glucagon administration, there was a significant decrease in the MCRi as well as a two- to threefold prolongation of insulin t_1/2. Early studies of glucagon stimulation tests reported decreases in the C-peptide:insulin ratio (45,48), presaging the findings reported here. Moreover, a reduced C-peptide:insulin ratio has been reported in case studies of individuals with glucagonomas under fasting (54) and arginine stimulation (55). Thus, the findings reported herein are consistent with data collected previously in human studies, and support an effect of glucagon to promote insulin clearance. Given that the major site of glucagon action is in the liver, it is likely that hepatic removal accounts for this effect.

The results of the current study bring together two discrete bodies of the published literature to support a new component of glycemic regulation. The first set of data includes the recent findings from several groups indicating that α-cell products exert paracrine control of the β-cell, augmenting nutrient-stimulated insulin secretion and contributing to glucose tolerance (6,15,16,40). The second body of evidence comes from older studies using large glucagon boluses to test β-cell function in humans with diabetes, work that established an insulinotropic effect of glucagon and raised the possibility that it also affected insulin clearance (9,45,46,48). Our results add to these by demonstrating an effect of glucagon to enhance glucose-stimulated insulin secretion in amounts that are compatible with concentrations in the islet and add to previous findings that glucagon reduces the removal of insulin from the circulation. Taken together, there is now evidence to support a model whereby nutrient-induced α-cell secretion promotes glucose disposition by increasing peripheral insulin levels through both increased secretion and reduced clearance. This model has potential application in the physiologic distribution of dietary carbohydrate between the liver and skeletal muscle, the two principal organs of glycogen synthesis. Moreover, this model is likely to have important implications in the pathophysiology of diabetes, where elevated α-cell activity is well established. Finally, recent advances in drug development now include GCGR agonism as part of the antidiabetic mechanism of action in multireceptor-activating, incretin-based molecules (56–58); the findings reported here may actually have more relevance in pharmacology where supraphysiologic glucagon activity is operative.

This study has several important limitations that must be considered when interpreting our results. First and foremost, our estimate of a glucagon dose to mimic paracrine regulation is rough at best and based on a study of perifused mouse islets (59). However, the circulating levels reached in this study are in the nanomolar range, well above plasma levels measured in the portal vein (60,61) or peripherally during hypoglycemia in humans (43,44), and thus compatible with the higher concentrations that would be active in the confined space of the islet. Second, both hyperglycemia and exogenous glucagon have been reported to enhance glomerular filtration rate (62,63) and potentially increase clearance of C-peptide, effects we did not account for in our modeling of insulin secretion and clearance. However, an increase in C-peptide clearance with either hyperglycemia or glucagon would have tended to reduce estimates of ISR and clearance parameters, thus underestimating differences that we noted to be significant anyway. Therefore, we do not think this potential difficulty would have a meaningful impact on the interpretation of this study. Third, we did not anticipate the severity and uniformity of GI side effects when we initiated this study and had to modify our protocol to minimize discomfort in our subjects. Of note, there was a marked difference in the temporal pattern of these side effects, which occurred ∼60 min after cessation of glucagon, compared with the almost immediate nausea we have previously noted in humans when we reached threshold rates of GLP-1 infusion (64,65) and the nausea reported after ≥6 h of glucagon administration in recent work (66). Regardless, we truncated the time of glucagon exposure from our initial design of 60 min to 30 min, and this reduced the number of blood samples during this experimental period and, therefore, some of the resolution on our estimates during glucagon administration. Fourth, our observations on insulin clearance were post hoc and made at supraphysiologic concentrations of glucagon; these will need to be confirmed in studies with more direct assessments of insulin clearance and at a range of glucagon doses. Finally, our sample size of nine subjects is relatively small and was composed almost entirely of women. While the effect sizes of our major outcomes were large and consistent across the group, and it seems unlikely that they would not be seen in a more heterogeneous cohort, the findings will need to be extended to other groups.

In summary, the results described here add to a growing body of work suggesting that glucagon has a broader role in glucose metabolism than simply as a counterregulatory hormone. In healthy subjects, glucagon is a potent insulin secretagogue, with characteristics similar to the incretins. Moreover, this insulinotropic activity may be complemented by an action to reduce insulin clearance, a combination well suited to accentuate the systemic rise in prandial insulinemia. We propose from these results a model whereby α-cells act in concert with β-cells during meal absorption to regulate the amount and distribution of secreted insulin and facilitate appropriate disposition of glucose among the tissues.

Acknowledgments. The authors thank Lorraine Elliott-Penry, Florence Briones, and Georgianne Gedon-Lipscomb, members of the Duke Molecular Physiology clinical research unit, for excellent care of the research subjects; Johanna Johnson for study coordination; Liezl Fos for recruitment efforts; the Duke Investigational Drug Services Pharmacy for material preparation; Alyssa Zidek, Jonathan Kenyon, William Bennett, Megan Reaves, Cris Slentz, and Andrew Hoselton for operational assistance; and Derek Nunez for thoughtful discussion and guidance on this project.

Funding. This study was funded by National Institute of Diabetes and Digestive and Kidney Diseases grants R01DK101991 (D.A.D.), F32DK12142 (S.M.G.), and U01DK121289 (A.N.H.).

Duality of Interest. D.A.D. consults for Eli Lilly, Sun Pharma, and MBX Biosciences. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. S.M.G. contributed to the study conception and design; data collection and analysis, modeling, and interpretation; and preparation of the manuscript. E.G., M.A.E., J.O.B., A.N.H., W.H., and G.Z. contributed to the data analysis. D.S. contributed to the modeling and data interpretation. J.T. contributed to the data interpretation. J.C. and D.A.D. contributed to the study conception and design, data interpretation, and editing of the manuscript. All authors reviewed the manuscript prior to submission. D.A.D. 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 in abstract form at the 81st Scientific Sessions of the American Diabetes Association, Virtual Meeting, 25–29 June 2021.