Intranasal Insulin Suppresses Endogenous Glucose Production in Humans Compared With Placebo in the Presence of Similar Venous Insulin Concentrations

Dysregulation of insulin-mediated suppression of hepatic glucose production (HGP) is a hallmark of type 2 diabetes (1). It is well established that activation of hepatic insulin receptors with ensuing activation of downstream insulin signaling pathways lowers hepatic glucose output by decreasing gluconeogenesis and increasing glycogen synthesis (1). In addition to the direct action of insulin on hepatocytes, insulin can indirectly affect HGP by altering free fatty acid (FFA) flux and suppressing glucagon secretion (2–4). Animal studies have indicated that insulin may also indirectly regulate hepatic glucose output via effects on the central nervous system (CNS)—the so-called brain–liver axis (1,5,6).

Injection of insulin into the third cerebral ventricle in mice activates K_ATP channels in the mediobasal hypothalamus (via activation of the insulin receptor–phosphoinositide 3-kinase pathway), which activates second-order neurons in the brain stem. This in turn lowers expression of gluconeogenic enzymes and glucose production by the liver, an effect that is abrogated by surgical resection of the hepatic branch of the vagus nerve (6). CNS insulin action in the dorsal vagal complex has more recently been shown to regulate HGP via activation of the insulin–insulin receptor–extracellular signal–related kinase pathway (7). These brain–liver axis effects are observed in the absence of changes in plasma insulin concentration (5–7). A brain–liver axis also has been demonstrated in dogs. Insulin delivery into the head arteries augments hepatic glycogen synthesis and reduces mRNA expression of gluconeogenic enzymes but causes no acute change in HGP (8).

In recent human studies a single dose of intranasal insulin (INI; 160 IU) modulated food-related activity in the CNS (9), reduced overall food intake in males (10), improved cognitive performance in females (10), and reduced intake of palatable food and increased satiety in females (11). These effects are likely mediated by direct action of insulin on CNS, although other non-CNS effects of nasally administered insulin cannot be definitively excluded. At a lower dose of 40 IU, INI increased the cerebrospinal fluid insulin concentration, with no significant change in measured serum insulin concentration sampled at 10-min intervals for the first 40 min and at 20-min intervals for a further 40 min (12). Small lipophilic peptides such as insulin can potentially enter the CNS directly by diffusing across the olfactory epithelia and intercellular spaces into the subarachnoid space (12,13). Intranasally administered peptides can also enter the CNS indirectly via uptake into the olfactory bulb and axonal transport (12,13). Recent studies in humans showed that INI administered at a higher dose of 160 IU can acutely lower plasma glucose and alter peripheral insulin sensitivity (11,14,15), an effect postulated to occur via insulin delivery to the CNS. However, the higher dose of INI (160 IU) transiently increased peripheral insulin concentration (9,14,15), which may contribute to the acute effects of INI on peripheral glucose metabolism and insulin sensitivity (11,14,15).

In this single-blind, placebo-controlled, crossover study, we aimed to investigate whether INI action regulates endogenous glucose production (EGP) in humans. We assessed EGP following the administration of 40 IU of INI or intranasal placebo (INP) with primed, constant infusion of d-[6,6′-²H₂]glucose (D2-glucose) (Fig. 1). Participants were studied under conditions of an arterial pancreatic clamp in which systemic venous insulin and glucagon concentrations are clamped at basal concentrations to prevent fluctuations in peripheral arterial insulin and glucagon concentrations. Lower portal insulin and glucagon concentrations than would be expected in the basal state—that is, this is a state of hepatic insulin and glucagon deficiency—are noted with this clamp technique. Under these conditions we demonstrated for the first time in humans that INI lowers EGP. This effect was seen in the presence of similar venous insulin concentrations.

Eight healthy men with no medical illnesses and taking no medications were recruited by advertisements in the local press. Their demographic and biochemical parameters are shown in Table 1. They underwent a 75-g oral glucose tolerance test, routine screening blood tests, and urinalysis. Those with abnormal tests were excluded. Each participant was studied on two occasions, 4 weeks apart, in a single-blind, placebo-controlled, crossover trial. Each participant received 40 IU of INI lispro (Humalog; Eli Lilly Canada, Toronto, Ontario, Canada) during one visit and placebo during the other. The order of the visits was randomized.

Pilot studies indicated that peripheral insulin spillover was inevitable, with doses ranging from 80 IU to as low as 10 IU (Supplementary Fig. 1). A dose of 40 IU INI was chosen because in previous studies it increased cerebrospinal fluid (CSF) insulin concentration in humans (12), and in our pilot studies a higher dose of INI (80 IU) resulted in a greater spillover of insulin lispro into the peripheral circulation (Supplementary Fig. 1). To ensure similar venous plasma insulin concentrations between treatments, 0.005 IU/kg insulin lispro was infused intravenously over 30 min, starting at the time of INP administration. The intravenous insulin lispro dose was identified in a pilot study using varied doses (data not shown).

Participants were admitted to the Metabolic Test Centre of Toronto General Hospital the morning before the study (Fig. 1). Volunteers had a standardized mixed meal at 5 p.m. and were not permitted any food or drink orally except water until the conclusion of the study. The following day at 7 a.m. (t = −120 min), an arterial pancreatic clamp was started and continued for the remainder of the study (until 3 p.m.; t = 360 min) to neutralize any potential effects of arterial pancreatic hormone fluctuations on EGP. The clamp comprised the following infusions: somatostatin 30 µg/h (Sandostatin; Novartis Pharmaceuticals Canada, Dorval, Quebec, Canada) to inhibit pancreatic insulin and glucagon secretion with concomitant replacement at basal concentrations of insulin (Humulin R; Eli Lilly Canada) at 0.05 mU/kg/min; human recombinant growth hormone (Humatrope; Eli Lilly Canada) at 3 ng/kg/min; and glucagon (Eli Lilly Canada) at 0.325 ng/kg/min. Autologous serum (3 mL), freshly prepared from the subject’s blood, was added to the saline as a carrier before hormone dilution.

Also at 7 a.m. (t = −120 min), a primed, constant infusion (22.5 µmol/kg bolus followed by 0.25 μmol/kg/min) of D2-glucose (Cambridge Isotope Laboratories, Tewksbury, MA) was started and continued until the conclusion of the study at 3 p.m. (t = 360 min). At 9 a.m. (t = 0 min), participants received either 40 IU of insulin lispro (100 IU/mL Humalog; Eli Lilly Canada) or placebo (insulin diluent; Eli Lilly Canada) via a metered nasal dispenser (Pharmasystems, Markham, Ontario, Canada). A single spray dispenses 0.1 mL (10 IU) of insulin. A single spray was administered in each nostril while inhaling. Another spray was similarly administered in each nostril 1 min later to give a total dose of 40 IU of insulin lispro. Because of spillover of 40 IU insulin lispro into the peripheral circulation (Supplementary Fig. 1), volunteers received an intravenous infusion of insulin lispro during the placebo visit to ensure similar venous insulin concentration in both treatment groups, as described above.

Blood samples (10 mL) were collected every 30 min for the first 120 min (t = −120 to 0 min) after starting the arterial pancreatic clamp, every 5 min for 30 min after administration of INI/INP (t = 0 to 30 min), and every 10 min thereafter until the conclusion of the study (t = 30 to 360 min). A 20% dextrose solution was administered as necessary to maintain euglycemia.

Plasma was separated from blood samples in a refrigerated centrifuge at 3,000 rpm for 15 min at 4°C. Sodium azide (70 mg/L blood; Sigma-Aldrich, Oakville, Canada) and aprotinin (1.94 mg/L blood; Sigma-Aldrich) were added to the plasma to prevent hydrolysis and protein degradation. Plasma was dried and derivatized, and stable isotope enrichments were determined (16). Derivatized samples were analyzed with gas chromatography/mass spectrometry (Agilent 5975/6890N; Agilent Technologies Canada Inc., Mississauga, Ontario, Canada) with electron impact ionization using helium as the carrier gas. Selective ion monitoring with a charge-to-mass ratio of 242 and 244 was performed. Atom percentage of excess fraction (APE) was calculated for each sample as APE = tracer/(tracer + tracee).

Commercial kits were used to measure total insulin (Millipore, Billerica, MA), growth hormone (Abcam Inc., Toronto, Ontario, Canada), FFA (Wako Industrials, Osaka, Japan), triglyceride (TG) (Roche Diagnostics), and glucagon (Millipore). An insulin lispro kit (Millipore) (specificity for lispro 100%, specificity for human insulin ∼0.05%) was used to measure lispro concentration in our pilot studies (Supplementary Fig. 1).

EGP was calculated as described previously (17). During a steady state, the rate of glucose appearance (R_a) was equivalent to the rate of glucose disappearance (R_d), where R_a = tracer infusion rate/APE fraction and EGP rate = R_a – glucose infusion rate (17).

Results are presented as mean ± SEM. Paired t tests were used to compare plasma glucose, venous insulin, plasma glucagon, glucose infusion rates, and EGP. A P value <0.05 was considered significant. Post hoc analysis revealed a power of 98% to detect a change in EGP as well as plasma insulin in the final 180 min of the study (t = 180–360 min). The power to detect a change in insulin concentration during the entire study was 94%.

The study was carried out according to the principles of the Declaration of Helsinki and was approved by University Health Network Research Ethics Board, Toronto, Ontario, Canada. All participants gave written informed consent.

Mean glucose concentration over time during the study is depicted in Fig. 2A. INI treatment transiently lowered plasma glucose concentrations, with a mean nadir concentration at 180 min (INP 6.1 ± 0.3 vs. INI 5.1 ± 0.1 mmol/L; P = 0.01) (Fig. 2A). With infusion of 20% dextrose, blood glucose concentrations were not significantly different in the final 120 min of the study. There was no significant difference in mean venous plasma insulin concentration (Fig.2B) between treatments (INP 71 ± 7 vs. nasal insulin 77 ± 11 pmol/L; P = 0.6). As expected, based on pilot data, there was a transient increase in venous insulin after INI administration (Supplementary Fig. 1). Administration of a 30-min intravenous insulin lispro infusion along with INP administration ensured venous insulin concentrations were similar between the groups.

Intravenous glucose in the form of 20% dextrose was infused, as required, to maintain euglycemia. The mean dextrose infusion rate over time is illustrated in Fig. 2C. The mean dextrose infusion rate in the final 180 min of the study (t = 180–360 min) was significantly higher with INI treatment (placebo 1.1 ± 0.9 vs. INI 6.6 ± 1.6 μmol/min/kg; P = 0.015) (Fig.2D); the maximal difference occurred at 250 min (placebo 1 ± 1 vs. INI 8.4 ± 1.1 μmol/min/kg; P = 0.006) (Fig. 2C). These changes occurred despite similar venous insulin concentrations.

Mean EGP rate over time is shown in Fig. 3A. INI lowered EGP, with the nadir value at 250 min (placebo 12.3 ± 1.3 vs. INI 6 ± 0.9 µmol/min/kg; P = 0.01). Mean EGP in the final 180 min of the study (t = 180–360 min) was significantly lower with INI treatment (placebo 11.7 ± 1 vs. INI 7.6 ± 0.6 µmol/min/kg; P = 0.02) (Fig. 3B). The correlation coefficient between peak insulin after INI or placebo administration and decline in EGP from baseline in the final 180 min of the study was 0.38.

The rate of glucose disposal (R_d) over time is depicted in Fig. 3C. There was no significant difference in glucose disposal. Mean R_d in the final 180 min of the study was not different between the groups (placebo 12.9 ± 1.2 vs. INI 13.7 ± 1 μmol/min/kg; P = 0.4) (Fig. 3D). The specific activity of D2-glucose over time is shown in Supplementary Fig. 3.

Plasma FFA concentrations are shown in Fig. 4A. FFA concentration was significantly lower at 240 min with INI (placebo 0.2 ± 0.05 vs. INI 0.1 ± 0.03 mmol/L; P = 0.03). Mean FFA concentration in the final 180 min (t = 180–360 min) was not significantly different between treatments (placebo 0.20 ± 0.05 vs. INI 0.13 ± 0.03 mmol/L; P = 0.12) (Fig. 4B).

Plasma TG concentrations are shown in Fig. 4C. There was no significant difference in mean TG concentration from t = 180 to t = 360 min (placebo 0.7 ± 0.1 vs. INI 0.6 ± 0.1 mmol/L; P = 0.23) (Fig. 4D).

Rodent studies have demonstrated that insulin action in the CNS can reduce HGP (5,6). Previous human studies that deployed INI at a dose that increases CSF insulin concentration (12) reported changes in peripheral glucose concentration and insulin sensitivity, suggesting CNS insulin may regulate peripheral glucose metabolism (11,14,15). In these studies, however, there was a transient increase in venous insulin concentration after INI administration. The current study is the first human study to definitively demonstrate that INI (40 IU) suppresses EGP compared with INP. Importantly, the experimental design of our study ensured that venous insulin concentrations were similar between treatments.

A 40-IU dose of INI previously caused a rapid increase in CSF insulin concentration (12) without increasing serum insulin. In this study, under conditions of an arterial pancreatic clamp (during which endogenous insulin secretion cannot be modulated), 40 IU of INI (insulin lispro; Eli Lilly Canada) also transiently increased venous insulin concentration, as measured by a specific lispro assay. We also detected peripheral spillover with a dose as low as 10 IU (Supplementary Fig. 1). In this study we administered the lowest dose of nasal insulin that was previously shown to increase CSF insulin concentration (12) while minimizing systemic spillover by not using a higher dose. Subjects treated with INP were given an infusion of insulin lispro to try and mimic the increase in venous insulin concentrations after spillover of INI (Fig. 1). This experimental approach ensured similar venous insulin concentrations between the treatment groups.

The peripheral spillover of INI, detected with frequent blood sampling (every 5–10 min), may have implications in the interpretation of certain findings from previous studies using higher doses of INI. Three studies used higher doses of INI (160 IU) and reported relatively modest lowering of plasma glucose concentration within 30–45 min of administration (11,14), as well as improved peripheral insulin sensitivity (15). These studies did not measure plasma insulin as frequently as we did in the current study. In one study (14), plasma insulin (measured every 30 min) was significantly higher at 30 min, with a decline in C-peptide and subsequent insulin concentration, suggesting peripheral spillover of insulin with a decline in endogenous insulin secretion. Plasma insulin concentration (measured every 15 min) was transiently higher at 15 min after administration of 160 IU of INI in a recent study by the same group (15). Insulin concentration was not reported for the first 45 min in the study by Hallschmid et al. (11). In these studies of INI, excluding a contribution of peripheral insulin action to the rapid decrease in plasma glucose concentration is not possible. It is worth noting that other studies using 160 IU of INI did not report changes in plasma glucose concentrations (9,18). This may be because of the relative infrequency of blood sampling (every 30 min compared with every 5–10 min in the current study) (11). In addition, in the absence of an arterial pancreatic clamp, endogenous insulin and glucagon secretion can be modulated to prevent major fluctuations in plasma glucose (9,14). With an arterial pancreatic clamp, the normal portal to peripheral insulin gradient is lost, resulting in relative hepatic insulin deficiency, which may have permitted INI to lower EGP. In the aforementioned studies (9,14) the physiological portal peripheral gradients of insulin and glucagon were maintained; this may have rendered the liver less sensitive to INI. Finally, these studies were of a shorter duration (≤180 min) (9,11,14,18) than the current study (360 min), in which EGP declined after 180 min, and therefore a late glucose-lowering effect would have gone undetected.

Unlike the relatively rapid decrease of plasma glucose seen with INI in some studies (11,14,15), extra-pancreatic K_ATP channel activation (likely CNS K_ATP channel activation, a downstream target of insulin action in the CNS) previously decreased EGP in humans over the course of ∼6–7 h; parallel studies using rats demonstrated that an equivalent dose of diazoxide is detectable in CSF after 1 h and plateaus at ∼4 h (16). We therefore speculate that the rapid decrease in plasma glucose that occurred with 160 IU in previously published studies (11,14) was likely due to transient systemic insulin absorption from the INI administration, which in turn induces peripheral insulin action, and that any potential CNS regulation of EGP by insulin would be a slower process. Consistent with our hypothesis, despite similar venous insulin concentrations in both treatment groups throughout the study, INI lowered EGP after ∼180 min, and EGP remained lower at the conclusion of the study (at 360 min). This time scale of INI action is similar to that reported with intracerebroventricular injection of insulin in rodents (5,6) and suggests that INI, potentially via CNS insulin action, does not rapidly regulate EGP in the acute setting. It is not known whether prior exposure to INI, as occurs with longer-term administration of INI, affects EGP, but 8 weeks of treatment with INI did not affect fasting insulin and glucose (19).

Although the exact mechanism by which INI lowers EGP remains to be determined, reduced expression of hepatic gluconeogenic enzymes such as PEPCK and G6Pase via hepatic vagal efferents secondary to CNS insulin action is a plausible explanation. This is based on the previous findings that 1) INI increases CSF insulin concentration (12) and 2) intracerebroventricular insulin injection reduces the expression of gluconeogenic enzymes and HGP in rodents, an effect abrogated by resection of hepatic vagal efferents (5,6). A previous study demonstrated a reduction in FFA with 160 IU of INI treatment in humans with no reported change in plasma insulin concentration (18), although as discussed above at this dose peripheral insulin spillover may have contributed. In the current study, mean FFA between 180 and 360 min tended to be lower with INI but did not reach statistical significance. In view of the relatively small sample size of the current study, we cannot exclude a significant difference in FFA had we studied more individuals, and therefore reduced FFA flux to the liver could potentially contribute to the lowering of HGP. There were no significant differences in the concentration of growth hormone and glucagon between treatments. Previous studies of dogs demonstrated that pulmonary insulin delivery can regulate insulin sensitivity independent of plasma insulin concentration (20). Nasally inhaled aerosols are deposited in the lungs (21). INI can potentially deliver insulin to the lungs with secondary effects on glucose metabolism. It is not possible to rule out as yet unidentified CNS- and/or non-CNS-mediated effects as contributors to EGP reduction.

Although this study suggests that INI, possibly acting via the CNS, can regulate EGP, the contribution of this pathway to glucose homeostasis in normal human physiology remains to be determined. Under our arterial pancreatic clamp conditions, the pancreas is unable to modulate endogenous insulin and glucagon secretion with changing glucose concentration. In addition, with an arterial pancreatic clamp, as was the case in previous rodent studies (5,6), portal and peripheral insulin concentrations are likely to be identical because the normal portal-peripheral insulin concentration gradient is abolished when insulin delivery does not occur via the portal circulation. In normal physiology, portal insulin concentration is nearly three times greater than its peripheral concentration (8,22). Hence during an arterial pancreatic clamp there is relative hepatic insulin deficiency (8,22). Under these conditions, CNS-mediated effects of insulin are dominant and lower EGP. With an arterial pancreatic clamp, however, the portal-to-peripheral glucagon gradient (23) is also lost, which may have minimized the effects of relative hepatic insulin deficiency on EGP. Consistent with the hypothesis that only under conditions of relative hepatic insulin deficiency can CNS insulin lower EGP, in experiments carried out using dogs, when a pancreatic clamp was instituted with a normal portal-peripheral insulin concentration gradient, CNS insulin delivery augmented glycogen synthesis but had no effect on EGP 4 h after administration (8). In this study, however, the CNS–to–non-CNS insulin gradient was not maintained. Blockade of hypothalamic insulin action in the setting of physiological hyperinsulinemia and portal-peripheral insulin concentration gradient blunted induction of glucokinase gene transcription and abrogated the inhibition of glycogen synthase 3β transcription but caused no net change in EGP (24). This effect was seen with a similar physiological increase in CNS, liver, and peripheral circulation. Although it is possible some differences might be ascribed to interspecies effects, the presence of a portal-peripheral insulin gradient likely abrogated the effect of CNS insulin action. Intriguingly, there was a significant reduction in mRNA levels without significant changes in protein expression of gluconeogenic enzymes (8). Whether a study of longer duration, such as our 6-h study, would have affected EGP remains unclear.

Whether INI would lower plasma glucose in a more physiological setting, in the absence of an arterial pancreatic clamp and presence of a normal portal-peripheral insulin concentration gradient, over the time course of the current study (between 180 and 360 min after administration) is currently not known. Previous studies using INI in the absence of a pancreatic clamp reported a very modest (∼5%) reduction in plasma glucose concentration within 2 h of administration, which may reflect insulin action due to peripheral spillover (11,14). Notwithstanding interspecies differences and differing assays, based on hyperinsulinemic clamp studies of dogs with measurement of plasma and CSF insulin (25), as well as CSF insulin concentration after administration of 40 IU of INI in humans (12), it is likely that 40 IU of insulin causes a supraphysiological increase in CSF insulin concentration. In a previous study assessing the effects of activation of extra-pancreatic K_ATP channels with an arterial pancreatic clamp, a physiological increase in insulin concentrations in the placebo group did not alter EGP, although pharmacological activation of K_ATP channels did reduce EGP (16). Despite the nonphysiological aspects of the study discussed above, we demonstrated for the first time that nasally administered insulin decreases EGP in humans compared with INP in the presence of similar venous insulin concentrations.

Resistance to direct insulin action in the liver and increased HGP are hallmarks of type 2 diabetes (1). Current treatment modalities for type 2 diabetes have potential side effects, including weight gain with subcutaneous insulin and sulphonylureas, which can potentially exacerbate hepatic insulin resistance (26). Whether INI can acutely lower EGP in individuals with insulin resistance and type 2 diabetes and whether chronic treatment with INI affects glycemic control remain to be determined. An additional potential advantage of INI is that, unlike subcutaneous insulin, it is less likely to cause weight gain. Animal studies have suggested that CNS insulin action reduces appetite and results in weight loss (27). In human studies of INI, an acute reduction in appetite was reported after a single dose in men (10), with a reduction in postprandial satiety and intake of palatable food in women (11). In addition, 8 weeks of INI treatment in slim healthy volunteers modestly reduced fat mass and body weight in men; modest weight gain occurred in women, which is thought to be due to an increase in extracellular water, but, importantly, there was no change in body fat (19). It must be noted, however, that in rodent models of insulin resistance induced by a high-fat diet, CNS-mediated effects of insulin are blunted (28), suggesting the presence of hypothalamic insulin resistance. Whether a similar phenomenon occurs in obese insulin resistant humans remains to be determined.

In conclusion, we showed that INI, at a dose that is known to increase CSF insulin concentration, lowers EGP under conditions of experimental portal hypoinsulinemia compared with INP. This effect occurs despite similar venous insulin concentrations between treatments. Additional studies are needed to evaluate whether this pathway is amenable to therapeutic manipulation in insulin resistance and type 2 diabetes.

Acknowledgments. The authors are indebted to Brenda Hughes and Patricia Harley (University Health Network) for their nursing assistance in conducting the clinical protocol and Linda Szeto (University Health Network) for technical assistance.

Funding. S.D. and C.M. are recipients of postdoctoral fellowship awards from the Banting and Best Diabetes Centre, University of Toronto, and S.D. is the recipient of a Focus on Stroke 12 Fellowship Award from the Heart and Stroke Foundation of Canada. K.K. is funded by a Juvenile Diabetes Research Foundation-Canadian Clinical Trials Network postdoctoral fellowship. G.F.L. holds the Sun Life Financial Chair in Diabetes and the Drucker Family Chair in Diabetes Research.

Duality of Interest. This study was funded by Eli Lillyhttp://dx.doi.org/10.13039/100004312 Canada (grant no. F3Z-CA–0082 to G.F.L.). G.F.L. has served on advisory boards to Eli Lilly Canada. No other potential conflicts of interest relevant to this article have been reported.

Author Contributions. S.D., C.X., C.M., and G.F.L. designed the study and interpreted data. S.D., C.X., C.M., and K.K. acquired and analyzed data. S.D. and G.F.L. wrote the manuscript. G.F.L. obtained funding and supervised the study. All authors edited the manuscript. G.F.L. 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.