Effects of Intranasal Insulin on Hepatic Fat Accumulation and Energy Metabolism in Humans Free

Evidence from rat models indicates that insulin signaling in the central nervous system (CNS) contributes to the regulation of hepatic glucose metabolism (1). Intracerebroventricular application of insulin has been shown to decrease endogenous glucose production (EGP) through activation of hypothalamic ATP-dependent potassium (K_ATP) channels (2). However, findings in dogs indicate that, in contrast to rodents, brain insulin action does not acutely regulate glucose production and gluconeogenesis in larger mammals (3).

Because EGP regulation differs among species (4) and its impairment is a metabolic hallmark of type 2 diabetes, examining whether brain insulin signaling also modulates rates of EGP in humans is important. Of note, activation of K_ATP channels with diazoxide may suppress EGP in healthy humans (5), supporting an insulin-mediated brain periphery crosstalk in humans. Intranasal insulin administration raises cerebrospinal fluid insulin levels (6) and could therefore mimic brain insulin delivery in humans. Most recently, two studies using intranasal insulin revealed conflicting results, reporting either a reduction of EGP (7) or no acute changes in glucose metabolism (8).

Intranasal insulin may also lower plasma free fatty acids (FFAs) and rates of labeled glycerol appearance (9), suggesting central insulin regulation of lipolysis in humans as demonstrated before in rodents (10). Intranasally administered insulin may further affect body weight and fat content (11). Because insulin resistance tightly associates with hepatic fat accumulation (12), examining whether central insulin may improve hepatic glucose and lipid metabolism is of interest. So far, no studies have addressed the effects of intranasal insulin on hepatic energy metabolism in patients with insulin resistant type 2 diabetes compared with healthy individuals under physiological fasting conditions.

This randomized controlled crossover clinical study investigated intranasal insulin effects on hepatic insulin sensitivity (HIS), lipids, and energy homeostasis. To examine extreme states of normal and impaired glucose metabolism, we enrolled young healthy lean control subjects (CON) and elderly overweight patients with type 2 diabetes. In a subgroup of CON, we further studied the metabolic effects of an intravenous insulin injection mimicking the transient increase in serum insulin levels observed upon intranasal insulin application.

Ten insulin-naive patients with type 2 diabetes taking oral glucose-lowering medication were enrolled in this randomized controlled single-blind crossover single-center trial. Patients taking thiazolidinediones were excluded from the study because of the drug’s possible long-term CNS effects (13). Patients withdrew their oral glucose-lowering medication for 3 days before the experiments to exclude possible effects on metabolic tests (14,15). Two patients with type 2 diabetes were taking atorvastatin and one gemfibrozil; neither patient was withdrawn from the study. Ten young lean and healthy volunteers not taking any medications, without a family history of diabetes, and with normal glucose tolerance based on a standard 75-g oral glucose tolerance test were also enrolled. Before inclusion, all participants gave written informed consent. This trial was approved by the ethics board of Heinrich Heine University Düsseldorf. The participants underwent screening, including medical history, clinical examination, and blood tests. None had clinical or laboratory signs of infection or hepatic, vascular, renal, or endocrine diseases. All participants were sedentary and refrained from any exercising for 3 days before the study. Female participants were either postmenopausal or examined between days 5 and 8 of their regular menstrual cycle.

The participants arrived at 7:00 a.m. at the German Diabetes Center after a 10-h overnight fast and remained fasted until the end of the study day. Intravenous catheters were inserted in both antecubital veins for blood sampling and infusions. At time point −180 min, participants received a continuous infusion (0.036 mg ⋅ min⁻¹ ⋅ kg body weight⁻¹) of d-[6,6-²H₂]glucose (99% enriched in ²H glucose; Cambridge Isotope Laboratories, Andover, MA) after a priming bolus of 3.6 mg ⋅ kg body weight⁻¹ ⋅ fasting plasma glucose [mg/dL]/90 [mg/dL] for 5 min (16). The tracer infusion lasted until 180 min in all participants and to study delayed insulin effects (17) for an extended period until 360 min in a subgroup of 12 participants (6 CON and 6 patients with type 2 diabetes). Before and during the infusions, blood samples were drawn to measure tracer enrichments, metabolites, and hormones. All participants were studied using identical protocols on 2 study days spaced by at least 7 days, except for the intranasal administration of insulin or placebo at time 0 (18). One puff of the 0.1-mL spray solution contained either 10 IU human insulin (100 IU/mL Actrapid; Novo Nordisk, Copenhagen, Denmark) or 0.1 mL vehicle as placebo. Eight puffs were administered in each nostril, resulting in a total dose of 160 IU insulin or 1.6 mL vehicle on the respective study days. The participants were blinded to the order of spray application. On a third day, a subgroup of eight CON received 0.1 IU human insulin (Actrapid) i.v. and otherwise underwent the identical study protocol.

All measurements were performed in a 3-T magnetic resonance scanner (Achieva 3T; Philips Healthcare, Best, the Netherlands) using a 14-cm circular ³¹P surface transmit-receive coil (Philips Healthcare) for ³¹P magnetic resonance spectroscopy (MRS) and the built-in ¹H whole-body coil for localization and proton spectroscopy. Participants were scanned in the supine position at baseline and at 180 min, and the long-duration subgroup was scanned at 360 min. For the acquisition of ³¹P spectra, a volume of interest of 6 × 6 × 6 cm³ was positioned within the liver, and three-dimensional localized spectra were obtained using image-selected MRS (19) (time of repetition 4 s, number of signal averages 192, acquisition time 13 min, spectral width 3,000 Hz, data points 2,048). Absolute quantification of phosphorus metabolites (γ-ATP and Pi) was performed as previously described (20) using the AMARES (advanced method for accurate, robust, and efficient spectral fitting of MRS data) algorithm (21) in jMRUI (Java-based magnetic resonance user interface) software (22).

For assessment of liver fat content, a set of nonwater-suppressed and water-suppressed ¹H spectra were acquired using stimulated echo acquisition mode (repetition time/echo time/mixing time 4,000/10/16 ms, number of signal averages 32, volume of interest 3 × 3 × 2 cm³). Data from localized ¹H-MRS were analyzed to assess fat content as previously described (23), and absolute concentrations were expressed as percent hepatocellular lipids (HCLs) relative to water content using the equations by Longo et al. (24). Concentrations of phosphorus metabolites were corrected for the volume captured by lipid droplets within hepatocytes (25). Reproducibility in acquisition and intra- and interobserver variability in spectral processing of ³¹P-MRS was reported previously (20).

Blood samples were immediately chilled and centrifuged, and supernatants were stored at −80°C until analysis. Venous blood glucose concentrations were measured by a glucose oxidase method (EKF Biosen C-Line glucose analyzer; EKF Diagnostic GmbH, Barleben, Germany) (26). Serum triglycerides (TGs) were enzymatically analyzed on a Roche cobas c 311 analyzer (Roche Diagnostics, Mannheim, Germany). FFAs were quantified enzymatically (intra-assay coefficient of variation [CV] 1%, interassay CV 2.4%) (Wako, Neuss, Germany) in samples containing orlistat to prevent ex vivo lipolysis (27). Serum C-peptide, serum insulin, and plasma glucagon levels were measured by radioimmunoassay (intra-assay CV for all 4–6%; interassay CV 6–7, 5–9, and 5–10%, respectively) (Millipore, St. Charles, MS).

After plasma deproteinization using Ba(OH)₂-ZnSO₄, atom percent enrichments (APEs) of ²H were measured on a Hewlett-Packard 6890 gas chromatograph equipped with a 25-m CP-Sil 5 CB capillary column (inner diameter 0.2 mm, film thickness 0.12 μm; Chrompack/Varian, Middelburg, the Netherlands), interfaced with a Hewlett-Packard 5975 mass selective detector as previously described (28). APEs of the fragments C3–C6 with average mass units of 187 for endogenous glucose and 189 for [6,6-²H₂]glucose were determined by using selected ion monitoring. APE was calculated as mass ratio corresponding to the tracer enrichment in plasma glucose. Intra- and interassay CVs were 0.6 and 1.0%, respectively.

Rates of EGP were calculated by dividing the tracer ([6,6-²H₂]glucose) infusion rate times the tracer enrichment by the tracer enrichment in plasma glucose and subtracting the tracer infusion rate (25). Fasting HIS was estimated as the HIS index by multiplying 100 times the inverse of the product of EGP and serum insulin (29). HOMA indices of fasting insulin resistance and β-cell function and QUICKI were calculated as previously described (30,31).

Data are presented as mean ± SEM and were subjected to two-way ANOVA with repeated-measures factors time and treatment. Areas under concentration-time curves (AUCs) were calculated according to the trapezoidal rule and compared with two-sided paired and unpaired t tests for within- and between-group comparisons, respectively. Comparisons at baseline between CON and patients with type 2 diabetes were performed with the two-sided unpaired t test. P < 0.05 indicates significant differences.

As expected, patients with type 2 diabetes were older and had greater BMI; waist circumference; and HbA_1c, fasting blood glucose, serum C-peptide, and cholesterol levels than CON (Table 1). Mean fasting insulin levels trended higher but were not significantly different from CON likely due to larger within-group variability in patients with type 2 diabetes. Indices of insulin resistance, insulin sensitivity, and β-cell function were different between groups, whereas fasting plasma TG and FFA levels were comparable. Within each group, no differences in baseline EGP, hormones, and metabolites among placebo, intranasal, and intravenous insulin studies were noted.

After intranasal insulin application, there was a trend for a decrease in blood glucose only in CON (P = 0.060_ANOVA treatment × time), with a maximal decrease of ∼5% at 30–40 min compared with placebo (Fig. 1A and B). Likewise, the AUC_{0–180 min} for glucose was lower after insulin application only in CON (Fig. 2A). Similar differences were noted at 40 min after insulin in the subgroups studied for 360 min.

Intranasal insulin resulted in lower FFA concentrations in CON (P < 0.01_ANOVA treatment × time) (Fig. 1I) and a greater maximal decrease (138 ± 41 μmol/L, 33.3 ± 8.5%) at 60 min versus baseline than after placebo administration (23.4 ± 27.3 μmol/L, −11.7 ± 9.7%). No changes were observed in patients with type 2 diabetes (P = 0.456_ANOVA time × treatment) (Fig. 1J). The AUC_{0–180 min} and AUC_{0–60 min} for FFA were not different between experimental conditions in both groups (Fig. 2B). TG and glycerol levels remained unchanged, and their respective AUC_{0–180 min} were similar after both interventions in CON and patients with type 2 diabetes (Fig. 2C and D).

In CON, serum insulin levels transiently increased by 3.7 ± 0.7 μU/mL and nearly doubled (92%) at 10 min after insulin versus placebo (P < 0.001_ANOVA time × treatment) (Fig. 1C). In patients with type 2 diabetes, the trend toward an increase in serum insulin by 59% at 10 min was not significant (4.9 ± 2.1 μU/mL, P = 0.3_ANOVA time × treatment) (Fig. 1D). In subgroups studied for 360 min after intranasal insulin/placebo, similar changes in the time course were found at 10 min in CON and at 20 min in patients with type 2 diabetes. Overall, AUC_{0–180 min} for insulin was comparable after insulin versus placebo administration in both groups (Fig. 2E). After intranasal insulin, serum C-peptide levels decreased only in CON and remained lower compared with placebo from 30 to 180 min (P < 0.01_ANOVA time × treatment) (Fig. 1E and F). In CON, AUC_{0–180 min} for C-peptide was lower after intranasal insulin versus placebo but not in patients with type 2 diabetes (Fig. 2F). Lower C-peptide levels were also found at 40, 100, 120, and 140 min after insulin in the subgroup of CON studied for 360 min but not in patients with type 2 diabetes. Neither intranasal insulin nor placebo affected glucagon concentrations in CON and patients with type 2 diabetes (Fig. 1G and H).

Baseline EGP was similar under insulin and placebo conditions in both groups (P = 0.93 CON insulin vs. placebo, P = 0.98 patients with type 2 diabetes insulin vs. placebo). Neither insulin nor placebo application affected rates of EGP, which remained comparable over 180 min and 360 min. Between-group analysis revealed higher HIS index values in CON than in patients with type 2 diabetes at baseline. In CON, the HIS increased by 4.0 ± 1.1 and 3.7 ± 1.4 kg ⋅ min ⋅ dL/(mg ⋅ μU) after both insulin and placebo applications, respectively (P < 0.05 at 180 min vs. baseline for insulin and placebo) but was not altered in patients with type 2 diabetes (Fig. 3A and B).

At baseline, HCL was lower in CON and did not differ between the experimental conditions (Table 1). In CON, HCL decreased by 35% at 180 min only after insulin (P = 0.04 at 180 min vs. baseline) but not after placebo. In patients with type 2 diabetes, HCL did not change under either experimental condition (Fig. 4A and B).

Baseline hepatic ATP and Pi concentrations were not different between conditions in CON and patients with type 2 diabetes. In CON, ATP increased at 180 min after insulin by 18% (0.5 ± 0.2 mmol/L, P = 0.03 vs. baseline) but was 23% lower in patients with type 2 diabetes at 180 min after intranasal insulin than after placebo (2.8 ± 0.2 and 3.4 ± 0.2 mmol/L, respectively; P = 0.03 vs. placebo) (Fig. 4C and D). Intranasal insulin did not affect hepatic Pi in either group, whereas it increased only in patients with type 2 diabetes by 0.4 ± 0.1 mmol/L after placebo administration (P = 0.02 at 180 min vs. baseline).

Insulinemia was comparable between intravenous and intranasal insulin studies as demonstrated by similar AUC_{0–20 min} (Fig. 5A), and a trend for higher insulin levels under both conditions compared with placebo was observed (P = 0.110 intranasal insulin vs. placebo, P = 0.088 intravenous insulin vs. placebo). Additionally, the absolute maximal decrease in blood glucose (Fig. 5B) and plasma FFA levels was comparable between intravenous and intranasal insulin studies and similarly differed from placebo (P = 0.02 ∆max glucose intravenous insulin vs. placebo, P < 0.0001 ∆max FFA intravenous insulin vs. placebo). No differences were observed in EGP after intravenous insulin administration. The relative percent increase in HCL was 34% (P = 0.02 baseline vs. 180 min) (Fig. 5C), whereas hepatic ATP concentrations remained unaltered (Fig. 5D).

This study shows that intranasal insulin administration does not affect glucose production over up to 6 h in humans with or without type 2 diabetes. However, intranasal insulin resulted in increased absolute hepatic ATP and decreased hepatic TG concentrations in glucose-tolerant CON but not in patients with type 2 diabetes. This was observed in the presence of a transient increase in serum insulin along with a minor transient decline in circulating glucose and FFA levels. Intravenous application of insulin unexpectedly led to comparable changes in circulating insulin, glucose, and FFA and did not affect hepatic ATP but increased hepatic TG concentrations.

In CON, blood glucose levels transiently and slightly decreased after intranasal insulin administration. Comparable insulin doses have been reported to lead to significant changes in blood glucose levels in some (32,33) but not all previous studies (18,34). Of note, the fall in blood glucose levels was preceded by a temporary rise in serum insulin concentration along with a prolonged reduction of C-peptide levels. Again, this increase in serum insulin is consistent with some (32–34) but not all previous reports (18). The absence of detection of these ultra-short-term changes in circulating insulin and glucose concentrations is likely due to a lower frequency of measurements in previous studies. Even the current study with its 10-min blood sampling interval might still underestimate the possible maximal changes in circulating insulin and subsequent peripheral metabolites and hormones. Nevertheless, short-term doubling of peripheral insulin levels followed by lower levels and AUC_{0–180 min} for glucose and C-peptide indicates that some amounts of intranasal insulin can be absorbed and suppress endogenous insulin secretion. However, the sustained reduction of C-peptide in CON after application of single-dose intranasal insulin suggests the operation of other mechanisms inhibiting insulin secretion, such as neural mechanisms initiated by cerebral insulin, lower FFA levels, and improved insulin action. The transient decrease in glucose concentration in CON could be due to greater glucose disposal and/or HIS (35). In elderly patients with type 2 diabetes, the response of peripheral insulin levels to intranasal insulin varied considerably, which might be a result of altered intranasal mucosal function. This as well as the impaired β-cell function and ambient hyperglycemia likely contributed to the absence of changes in the measured metabolites and hormones.

We assessed EGP in the fasting physiological state and related it to ambient insulinemia to provide a direct measure of fasting HIS (29), which takes into account the sensitivity of the liver to small changes in serum insulin levels (36). This approach, therefore, reflects physiological basal conditions of overnight fasting, whereas the hyperinsulinemic-euglycemic clamp creates a state of continuous submaximal stimulation by insulin. The finding of comparable EGP and HIS after intranasal insulin contributes human data to the ongoing discussion on effects of brain insulin on peripheral, mainly hepatocyte and adipocyte metabolism observed in mice (1,10) but not in dogs (37). Differences in the experimental settings may explain variable results obtained in other species. In mouse models, effects of brain insulin on EGP were demonstrated under hyperinsulinemic clamp conditions during peripheral and CNS insulin infusion (1,2). Additionally, stimulation of hypothalamic insulin signaling was initiated in advance and continued throughout the clamps. Moreover, the nonphysiological high-dose cerebrospinal fluid insulin infusion, hepatic hypoinsulinemia, and hypoglucagonemia in rodents could be the reason for the observed suppression of EGP. When brain and portal hormone levels were adequately adjusted to correspond to physiological ratios of glucoregulatory hormones, no changes in EGP were detected (37). Nevertheless, brain hyperinsulinemia decreased glycogen synthase phosphorylation in dogs (37), which may partly explain the increase in hepatic ATP in the current study.

Although central regulation of glucose metabolism by insulin in humans cannot be excluded, HIS is predominantly modulated by direct insulin action (38). Furthermore, we detected no changes in EGP for up to 6 h after nasal insulin application, suggesting that brain insulin signaling is not primarily responsible for acute effects on liver glucose metabolism. On the other hand, elegant experiments in humans demonstrated suppression of hepatic glucose output appearing 6 h after diazoxide consumption (5), an intervention that activates hypothalamic K_ATP channels, thereby mimicking brain insulin action. Most recently, intranasal application of insulin lispro also decreased EGP at 3–6 h, again during pancreatic clamps (7). These studies point to interesting effects of central insulin signaling observed in the somatostatin-mediated absence of physiological endocrine counterregulation. In addition, not only diazoxide but also several nutritional and hormonal signals can activate hypothalamic channels (17), and insulin lispro seems to act more potently in the brain than human insulin (39).

We also found no effect of insulin applied through the intranasal route on HIS. This seems to differ from recent findings suggesting that intranasal insulin might modulate insulin sensitivity of glucose metabolism through brain regions known to affect the regulation of the autonomous nervous system outflow (33). Nevertheless, the current study primarily monitored fasting glucose production but not peripheral insulin sensitivity during hyperinsulinemic clamps. The observed increase in the HIS index during both placebo and insulin exposure in CON only probably relates to a decline in serum insulin levels resulting from a prolonged fasting state after placebo administration and a compensatory reduction of insulin secretion at 3 h after the preceding insulin peak after intranasal insulin administration.

Of note, the current study demonstrated that HCLs transiently decrease upon intranasal insulin exclusively in CON. These findings suggest novel peripheral metabolic effects of intranasal insulin but require scrutiny and careful interpretation because both brain insulin signaling and increased serum insulin could be responsible. To address this question, we applied insulin intravenously to match the spillover of intranasal insulin in the peripheral circulation in a subgroup of CON. Of note, intravenous insulin application resulted in increased HCLs at 180 min, which is in line with the known effect of insulin on lipogenesis and the observed increase in HCLs by short-term insulin infusion (14). Thus, the contrasting results observed with intranasal and intravenous insulin administration under otherwise matched metabolic conditions suggest the operation of a central mechanism of insulin action for lowering liver fat content. Of note, CON exhibited very low baseline HCL levels (0.5 ± 0.1%), and the slight decrease in HCLs observed at 180 min disappeared at 360 min. However, a previous study reported catabolic effects of long-term intranasal insulin administration in men (11), which could also reduce liver fat storage. Moreover, the contrasting increase in HCLs observed with intravenous insulin rather suggests that the intranasal route exerts unique HCL-lowering effects. The moderate reduction of FFA levels after intranasal insulin in CON is likely a result of the known direct insulin effect on adipose tissue lipolysis, which is usually blunted in patients with insulin resistant type 2 diabetes. Of note, previous studies described diminished lipolysis and enhanced lipogenesis by brain insulin signaling in rodents (10) and by intranasal insulin in humans (9). Thus, it cannot be excluded that reduced adipose tissue lipolysis contributed to the reduction of HCLs in CON. On the other hand, the lack of any effect of intranasal insulin on HCLs in the current patients with type 2 diabetes could result from not only alterations of brain insulin signaling and central insulin resistance in obesity and type 2 diabetes (40,41) but also other factors such as age, sex, visceral fat distribution, and adipose tissue insulin resistance. Of note, lipid-lowering medication in 3 of 10 patients with type 2 diabetes was not withdrawn before testing, which limits the interpretation of possible effects on lipogenesis in the current study.

Another novel finding of the current study was the rise in hepatic ATP after intranasal insulin administration to CON but not to patients with type 2 diabetes. The intravenous insulin administration failed to affect hepatic ATP, again suggesting a central effect of insulin independent of peripheral insulinemia. Reduced mitochondrial activity has been described in insulin resistant skeletal muscle and liver and plays an important role in the development of nonalcoholic fatty liver disease (25,42). In skeletal muscle, insulin can increase flux through ATP synthase in healthy humans but not in patients with type 2 diabetes (43), indicating impaired mitochondrial plasticity in insulin resistant states (42). We recently reported that lower muscle ATP synthase flux associates with higher HCLs in humans (44). Impaired mitochondrial plasticity, therefore, could have also been present in the liver of the current patients with type 2 diabetes. However, we cannot exclude that brain insulin signaling contributes to the regulation of liver energy homeostasis because intranasal insulin can enhance energy levels in brain (45). It remains unclear whether the diminished response to intranasal insulin in the current patients with type 2 diabetes is due to participants’ age, insulin resistance, or being overweight, but evidence from human studies suggests that peripheral insulin sensitivity and body weight are important factors for adequate insulin action in the brain (41,46).

This study benefits from the close monitoring of key metabolites and EGP under physiological fasting conditions as well as from use of localized measures of intrahepatic energy metabolism independent of EGP for the first time. In addition, this study confirmed the spillover of intranasal insulin into the systemic circulation (32,33). The established dose of 160 IU insulin administered intranasally in previous studies has been clearly shown to influence CNS functions (33,34,47). Because the spillover of insulin might limit the use of this approach when assessing peripheral metabolism, we established a protocol for intravenous insulin application to match the effects of systemic spillover on circulating insulin and metabolites. Nevertheless, this approach does not allow detection of the intracerebral mechanisms leading to the observed hepatic effects. Finally, the current study cannot discriminate whether the different response of the patients with type 2 diabetes to intranasal insulin was due to other factors than insulin resistance known to influence brain insulin action, such as age, sex, or obesity.

In conclusion, intranasal insulin application does not affect fasting HIS but can stimulate hepatic energy metabolism and reduce lipid storage in healthy humans. The changes proved to be independent of concurrent transient increases in serum insulin levels and unique for the intranasal route of administration. Intranasal insulin effects are blunted in patients with type 2 diabetes, which may result from lower ambient insulinemia and/or impairment of the indirect effects of insulin on peripheral metabolism.

Acknowledgments. The authors thank Andrea Nagel, Nicole Achterath, Kai Tinnes, and Myrko Esser (Institute for Clinical Diabetology, German Diabetes Center, Düsseldorf, Germany) for excellent help with the experiments.

Funding. This work was supported by the Ministry of Science and Research of the State of North Rhine-Westphalia (MIWF NRW) and the German Federal Ministry of Health (BMG). This study was supported in part by a grant of the Federal Ministry for Research (BMBF) to the German Center for Diabetes Research (DZD), by a grant of the Helmholtz Alliance Imaging and Curing Environmental Metabolic Diseases (ICEMED), and by the Schmutzler-Stiftung.

The funding sources had no input in the design and conduct of this study; in the collection, analysis, and interpretation of the data; or in the preparation, review, or approval of the article.

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

Author Contributions. S.G. contributed to the research and collection of data and writing of the manuscript. C.K. and A.B. contributed to the data collection and manuscript. P.N. contributed to the data collection. M.H., A.F., and H.-U.H. contributed to the manuscript. J.S. and M.R. contributed to the study design, research of data, and manuscript. M.R. 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 74th Scientific Sessions of the American Diabetes Association, San Francisco, CA, 13–17 June 2014.