Glucagon Resistance at the Level of Amino Acid Turnover in Obese Subjects With Hepatic Steatosis

Patients with type 2 diabetes exhibit elevated fasting levels of glucagon and inadequate suppression of glucagon after eating carbohydrate-rich meals (1,2). Glucagon is a well-known stimulator of hepatic glucose production (3,4), and hyperglucagonemia is recognized as an important component of the pathophysiology of type 2 diabetes, contributing to the hyperglycemic state (5,6). However, the mechanisms underlying hyperglucagonemia remain unclear. In a previous study, we found fasting hyperglucagonemia in obese individuals without diabetes and in obese patients with type 2 diabetes, but not in their lean counterparts; that is, fasting hyperglucagonemia was present independently of type 2 diabetes (7). In light of these findings, we speculated that hyperglucagonemia could be related to obesity, rather than being an issue related to type 2 diabetes per se (8). Importantly, obesity is strongly associated with fat accumulation in the main target organ of glucagon, namely the liver (9), and findings from a study by Longuet et al. (10) suggested that liver-specific disruption of the glucagon receptor gives rise to an α-cell–stimulating signal originating from the liver. In line with these considerations, we recently showed that individuals without diabetes and patients with type 2 diabetes with nonalcoholic fatty liver disease (NAFLD) exhibited significantly higher levels of fasting glucagon compared with individuals without diabetes and patients with type 2 diabetes without NAFLD—again, independently of the diabetic state (11). Interestingly, we also found significantly elevated fasting levels of circulating amino acids in the groups with NAFLD.

Recent new evidence supports the concept that hepatic amino acid metabolism and glucagon secretion are linked in a feedback cycle, the liver–α-cell axis (12,13), in which circulating amino acids stimulate glucagon secretion and α-cell proliferation, while glucagon regulates amino acid levels by stimulating hepatic amino acid turnover (14–17). Combining our findings of obesity- and NAFLD-associated hyperglucagonemia (7,11) with the new understanding of glucagon’s role in the liver–α-cell axis, we hypothesized that hepatic steatosis leads to glucagon resistance at the level of hepatic amino acid turnover, entailing elevated concentrations of circulating amino acids and hyperglucagonemia (8). The aim of this study, therefore, was to evaluate differences in the hepatic effects of exogenously administered glucagon between healthy lean subjects (nonsteatotic) and healthy obese subjects with hepatic steatosis (i.e., NAFLD). By applying a pancreatic clamp with somatostatin, we were able to avoid glucagon-induced changes in insulin levels and show glucagon resistance at the level of amino acid turnover in obese steatotic subjects. In addition, we carried out a full transcriptomic analysis of liver biopsy specimens, which showed significant changes in pathways controlling amino acid metabolism in the obese steatotic subjects compared with healthy lean individuals. Thus, we provide new data supporting the idea that steatosis-associated glucagon resistance may lead to reduced glucagon-induced amino acid turnover, elevated circulating levels of glucagonotropic amino acids, and, consequently, hyperglucagonemia in humans.

The study was approved by the Research Ethics Committee of the Capital Region of Denmark (Regionsgården, Hillerød; reg. no. H-6-2014-097) and was conducted in accordance with the latest revision of the Declaration of Helsinki.

The study included 30 healthy male participants between 25 and 80 years of age; 15 of these subjects were lean (mean ± SD age, 40.9 ± 12.8 years; BMI, 23.2 ± 1.60 kg/m²; HbA_1c, 30.2 ± 2.9 mmol/mol; fasting serum C-peptide, 336 ± 125 pmol/L; HOMA of insulin resistance, 0.82 ± 0.34 [all data here are presented as mean ± SD]) and 15 were obese (age, 35.8 ± 9.3 years; BMI, 33.6 ± 2.12 kg/m²; HbA_1c, 31.4 ± 3.3 mmol/mol; fasting serum C-peptide, 716 ± 249 pmol/L; HOMA of insulin resistance, 2.45 ± 0.96). Exclusion criteria included liver disease, consumption of over 14 units of alcohol per week, prediabetes (HbA_1c >42 mmol/mol), diabetes, and a first-degree relative with diabetes.

For evaluation of hepatic steatosis, fibrosis, and inflammation, an ultrasound-guided liver biopsy was performed on each participant by a radiologist (C.S. or M.J.K.) in the Department of Radiology, Gentofte Hospital, University of Copenhagen. Biopsies were taken between 1 and 2 weeks before the experimental day (see experimental procedures). The biopsy was performed by using a BARD MONOPTY Disposable Core Biopsy Instrument; the specimen obtained was immediately divided and distributed into formalin, RNAlater, or liquid nitrogen. Tubes with formalin and RNAlater were stored at −20°C, and snap-frozen samples were stored at −80°C. Histological examination of all biopsy specimens was performed by the same pathologist (J.L.L.) at the Department of Pathology, Herlev Hospital, University of Copenhagen.

Each participant was examined during one experimental day at Gentofte Hospital, after a 10-h overnight fast that included avoiding water, coffee, and tobacco. Initially, the participants underwent a DEXA scan for evaluation of body composition. The participants were then placed in a semirecumbent position in a hospital bed, and a cannula was inserted into each cubital vein, one from which to collect blood samples and one (in the contralateral arm) for administering infusions. The forearm from which the blood samples were drawn was wrapped in a heating pad (50°C) for collection of arterialized blood. At time −120 min, an infusion of stable isotopes was initiated; this infusion consisted of [6,6-²H₂]-glucose (prime: 17.6 µmol × kg⁻¹; continuous infusion: 0.6 µmol × kg⁻¹ × min⁻¹), [1,1,2,3,3-D₅]-glycerol (prime: 2 µmol × kg⁻¹; continuous infusion: 0.1 µmol × kg⁻¹ × min⁻¹), and [¹⁵N₂]-urea (prime: 84 µmol × kg⁻¹; continuous infusion: 0.15 µmol × kg⁻¹ × min⁻¹). After 2 h (when the tracer had achieved a steady state), a 3-h pancreatic clamp with somatostatin (450 µg × h⁻¹) was applied (time 0 min). At the same time, infusions of insulin (0.1 mU × kg⁻¹ × min⁻¹) and glucagon (0.6 ng × kg⁻¹ × min⁻¹) were initiated in accordance with previous studies (18,19). At time 90 min, the infusion rate of glucagon was increased fivefold (to 3.0 ng × kg⁻¹ × min⁻¹), producing high physiological plasma levels of glucagon (20). Urine was collected during the experiment for assessment of urea synthesis. To maintain diuresis, we applied an isotonic saline infusion, which was adjusted in order to maintain a constant total infusion rate of 300 mL × h⁻¹ (i.e., it was reduced when infusions were added or increased at times 0 and 90 min). Because the pancreatic clamp lasted only 3 h, we decided not to substitute growth hormone (19,21). Blood samples were drawn at times 120, 90, 30, 15, and 0 min before the initiation of the pancreatic clamp, and then samples were drawn every 15 min. Blood samples were collected in chilled tubes containing EDTA and a dipeptidyl peptidase 4 inhibitor (valine-pyrrolidide; final concentration, 0.01 mmol/L) for analysis of glucagon. For analysis of amino acids, blood was collected in chilled tubes containing EDTA. For analyses of insulin and C-peptide, blood was collected in dry tubes and left to coagulate at room temperature for 20 min. All tubes were centrifuged for 20 min at 1,200g and 4°C. Plasma samples for glucagon and amino acids were stored at −20°C, and serum samples for insulin and C-peptide at −80°C until analysis.

Synthetic somatostatin was demonstrated to be ≥98% pure and identical to the natural human peptide through high-performance liquid chromatography and sequence analysis. Glucagon and somatostatin were dissolved in sterilized water containing 2% human albumin, dispensed into glass vials under sterile conditions, and stored frozen at −20°C until the experimental day. On the experimental day, insulin, glucagon, and somatostatin were diluted in saline containing 0.5% human albumin.

Plasma glucose was measured at the bedside by using the glucose oxidase method (YSI 2300 STAT PLUS Analyzer). Plasma glucagon concentrations were analyzed by using a C-terminal–specific radioimmunoassay (antibody code 4305) with a sensitivity of ∼1 pmol/L (22). Serum insulin and C-peptide were analyzed by using two-site assays: an electrochemiluminescence immunoassay and a Roche/Hitachi Modular Analytics system. Plasma enrichment of [6,6-²H₂]-glucose, [1,1,2,3,3-D₅]-glycerol, and [¹⁵N₂]-urea was determined by using liquid chromatography–tandem mass spectrometry, as described previously (23). Amino acids were analyzed as previously described (11).

Liver mRNA was extracted by using a NucleoSpin mRNA Plus kit. RNA sequencing libraries were prepared with NeoPrep by using an Illumina TruSeq Stranded mRNA Library Kit and sequenced on the NextSeq 500 system with an NSQ 500 High-Output kit (version 2). STAR version 2.5.2a was used to map the reads to the human genome from GRCh38 Ensembl version 89 (24), and differential expression analysis was done with DESeq2 version 1.18.1 (25). Genes differentially expressed at a 5% false discovery rate were regarded as regulated. Pathway analysis was performed by using the “piano” package (version 1.18.1) for R (26) and the Reactome Pathway Database (27).

Baseline statistics are presented as standard descriptive statistics and were compared by using the Student t test or the Mann-Whitney U test, as appropriate. Data from the experimental day were evaluated on the basis of the area under the curve (AUC) during steady state within the three periods of the experimental day (i.e., the preclamp period and the basal and high glucagon infusion periods). AUCs were calculated by using the trapezoid rule and were adjusted for time. Normally distributed data were compared by means of ANOVA in a mixed model that included subject identification as a random factor and covariance structures aiming at best fit (through a top-down modeling strategy). ANOVA was carried out with the “nlme” package in R statistical software. Adjusting for steatosis in the linear regression model did not affect the results. Therefore, separation of the individuals into lean and obese groups was maintained. Results are presented as the mean ± SEM unless otherwise stated. The metabolic clearance rate (MCR) of glucagon was calculated using the following formula.

The mean of the glucagon concentrations from time 120 to 180 min was used for the variable [Glucagon]_{steady‐state}.

Data sets from the study are available from the corresponding author (F.K.K.) upon reasonable request. The results of the global differential expression analysis are included in Supplementary Table 1. Parts of the RNA sequencing data were previously reported (28,29).

Fasting serum insulin was significantly higher in the obese group (131.9 ± 13.7 pmol/L) than in the lean group (43.9 ± 4.7 pmol/L) (P < 0.0001) (Fig. 1A), as was C-peptide (802.1 ± 52.9 pmol/L in the obese group vs. 397.1 ± 44.0 pmol/L in the lean group; P < 0.0001) (Fig. 1B). When somatostatin infusion was initiated at time 0 min, C-peptide levels dropped significantly (P < 0.0001) in both lean and obese subjects, reflecting the shutdown of pancreatic insulin secretion. The insulin infusion resulted in stable serum insulin concentrations in both groups throughout the pancreatic clamp, though the concentrations were slightly higher in the obese group (Fig. 1A). As illustrated in Fig. 1C, fasting plasma glucagon concentrations were significantly higher in the obese group (median, 11 pmol/L; range, 5–17 pmol/L) than in the lean group (median, 6 pmol/L; range, 4–11 pmol/L) (P = 0.013). Basal infusion of glucagon during the pancreatic clamp resulted in similar plasma concentrations of glucagon in the two groups (P = 0.48). When the infusion rate of glucagon was increased, plasma concentrations of glucagon increased significantly (P < 0.0001) by ∼4.5-fold in the lean group and by 5-fold in the obese group (Fig. 1C). During the high-dose glucagon infusion, we observed a slightly but significantly lower MCR in the obese group (36.5 ± 2.09 mL × kg⁻¹ × min⁻¹) than in the lean group (29.1 ± 1.30 mL × kg⁻¹ × min⁻¹) (P = 0.0079).

Fasting plasma glucose concentrations were similar in the two groups (5.4 ± 0.1 vs. 5.6 ± 0.1 mmol/L for lean and obese, respectively; P = 0.24). In the lean group, plasma glucose levels did not change in response to the pancreatic clamp (P = 0.79), whereas plasma glucose slightly increased in the obese group, although this increase was not significant (P = 0.10) (Fig. 2A). During the period with the high rate of glucagon infusion, plasma glucose levels increased significantly in both groups (P < 0.0001 for both); this increase was significantly higher in the obese indi`iduals than in the lean (P < 0.001). We did not administer any exogenous glucose, and therefore any increase in the R_a of glucose (R_aglucose) reflects endogenous glucose production (EGP). Despite their lower fasting plasma glucagon concentrations, the baseline level of EGP was higher in the lean group than in the obese (P = 0.032) (Fig. 2B). During the basal glucagon infusion, the difference in EGP between the groups was not significant (P = 0.085). EGP increased significantly (P < 0.0001) in both lean and obese individuals in response to the high infusion rate of glucagon, with higher levels obtained in the lean group (P = 0.034). Although EGP tended to be higher in the lean group throughout the entire experiment, the graphs (Fig. 2B) appeared to show a parallel shift; thus, the groups seemed to respond similarly to the glucagon infusions. To examine any differences in the effect of glucagon on EGP between the groups, we calculated the incremental AUC during the high glucagon infusion using as a baseline the mean of the R_aglucose values during the basal glucagon infusion. Calculated this way, there was no difference in the effect of glucagon on EGP (498 ± 43 vs. 435 ± 61 μmol/L × min for lean and obese, respectively; P = 0.31) (Fig. 2D). The R_d of glucose (R_dglucose) was higher in the lean group during both basal and high infusion rates of glucagon (P = 0.01 for both time periods (reflecting higher insulin sensitivity) and increased significantly in both groups when the infusion rate of glucagon was increased (P < 0.0001 for both groups) (Fig. 2C).

Plasma glycerol was significantly reduced in the lean group after initiation of the pancreatic clamp (P < 0.0001); this reduction coincided with a substantial decrease in the R_a of glycerol (R_aglycerol) (i.e., lipolysis) (P < 0.001). The R_d of glycerol decreased in the lean group upon initiation of the pancreatic clamp but remained stable throughout the remainder of the experiment (P = 0.38). There were no changes in glycerol concentrations or kinetics in the obese group throughout the entire experiment. Data are provided in Supplementary Fig. 1.

The baseline concentration of total amino acids was significantly higher in the obese group (3,466 ± 87 μmol/L) than in the lean group (2,925 ± 88 μmol/L) (P < 0.0001) (Fig. 3). The basal infusion of glucagon resulted in stable levels of amino acids in both groups, but when the infusion rate of glucagon was increased, total amino acid levels dropped significantly in the lean group (P = 0.0001) but increased slightly but significantly in the obese group (P = 0.0042). The ratios between the fasting concentrations of specific amino acids in lean and obese individuals are presented in Fig. 4. Of special interest is that fasting concentrations of tyrosine and alanine (potential drivers of the liver–α-cell axis [11,14,15]) were increased by 1.13- and 1.11-fold, respectively, in the obese compared with the lean group. Plasma levels of urea did not change during the experiment, and the urea tracer did not reveal any changes in urea synthesis (Supplementary Fig. 2).

The liver biopsies were histologically evaluated with a focus on the degree of steatosis (i.e., percentage of hepatocytes containing cytoplasmatic lipid droplets). In the lean group, 2 of 15 subjects had steatosis (7% and 40%), whereas steatosis was found in 14 of 15 subjects in the obese group, covering a wide spectrum ranging from less than 5% of hepatocytes to levels as high as 40%. A prerequisite for the diagnosis NAFLD is the presence of at least 5% steatosis (30). Thus, by definition, only seven of the obese subjects had NAFLD. None of the subjects had inflammation, hepatocyte ballooning, or fibrosis. Differences in hepatic gene expression between the groups were evaluated by using full RNA sequencing, which unveiled 352 differentially regulated genes in the obese group (211 downregulated and 141 upregulated) (Fig. 5A). To get an overview of biological pathways affected by obesity, we performed an enrichment analysis, which revealed that most of the downregulated pathways were associated with amino acid metabolism, although differentially expressed genes were found across a broad spectrum of metabolic pathways (Fig. 5B). Furthermore, upon analyzing the expression of genes involved in amino acid metabolism, we found downregulation of several genes that encode amino acid transporters (e.g., SLC1A7, SLC7A2, SLC1A2, and SLC25A18), whereas none were upregulated (Fig. 5C). As previously reported, CPS1, which encodes the urea cycle enzyme carbamoylphosphate synthetase, was also downregulated in the obese group (29). The results of the global differential expression analysis are included in Supplementary Table 1.

The main findings of this study include fasting hyperaminoacidemia and hyperglucagonemia and a complete lack of glucagon-induced reductions of circulating amino acid concentrations among obese individuals with a high prevalence of biopsy-verified hepatic steatosis. Concomitantly, lower hepatic expression of several genes involved in amino acid metabolism was found in the obese individuals. We interpret these findings as an indication of steatosis-induced hepatic glucagon resistance at the level of amino acid turnover, and we suggest that perturbed hepatic amino acid metabolism may represent an important determinant of fasting hyperaminoacidemia, potentially causing hyperglucagonemia in obese and/or steatotic individuals.

We previously identified fasting hyperglucagonemia in obese individuals and in patients with NAFLD, independently of the presence of diabetes (7,11), and speculated that this occurred as a consequence of steatosis-induced hepatic glucagon resistance (8). Almost all the obese individuals in our current study displayed hepatic steatosis, although, surprisingly, only 7 of 15 participants met the definition of NAFLD with ≥5% steatosis. Nevertheless, the downregulated amino acid metabolism pathways in the obese group suggest that normal amino acid metabolism and ureagenesis were affected, probably as a result of hepatic steatosis, as recently demonstrated by De Chiara et al. (31).

Insulin and glucagon respond dynamically to changes in plasma glucose levels (32). We wanted to assess the isolated effect of glucagon on hepatic metabolism. To avoid counterregulation by insulin during glucagon infusion, we applied a pancreatic clamp with infusion of somatostatin. This shuts down the secretion of pancreatic insulin, as reflected by the immediate reduction of C-peptide levels in response to the clamp (Fig. 1B). By applying intravenous infusions of glucagon and insulin, we mimicked basal and high physiological levels of glucagon while maintaining stable basal concentrations of insulin (Fig. 1A and C). Glucagon concentrations increased 4.5-fold in the lean group and 5-fold in the obese group in response to the high infusion rate. The higher levels in the obese group are most likely a result of the lower MCR of glucagon we found in that group. Given these circumstances, our study provides a conservative estimate of hepatic glucagon resistance in the obese group.

Assessing potential disturbances in glucose metabolism seems to be an appropriate first step in the evaluation of hepatic glucagon resistance. Both lean and obese individuals had normal glucose tolerance, but plasma glucose levels increased significantly in the obese group compared with the lean in response to the high infusion rate of glucagon. This difference may be explained by differences in peripheral insulin sensitivity between obese and lean individuals, which is reflected by a lower R_dglucose (i.e., a lower rate of glucose removal from the bloodstream) in the obese group. Before the pancreatic clamp was applied, EGP was higher in the lean group. This could be interpreted as a greater hepatic sensitivity to glucagon, be a consequence of the lower serum insulin concentrations among the lean subjects, or both. Notably, we found no difference in the effect of glucagon between the groups during the basal or the high infusion of glucagon (when insulin levels were kept constant), as reflected by similar increments in EGP (Fig. 2D). Cirrhosis has been associated with reduced effects of glucagon on glucose metabolism, most likely as a result of reduced functional liver tissue mass (33,34). Considering this, a reduced effect of glucagon at the level of glucose metabolism might constitute a late development in liver disease, whereas other parts of hepatic metabolism may be affected at earlier stages of NAFLD (e.g., simple steatosis).

Glucagon infusion did not seem to exert any effect on glycerol metabolism in lean or obese individuals. However, the effect of insulin was evident in the lean group after initiation of the clamp, when R_aglycerol (i.e., a measure of lipolysis, expressed per kilogram of fat mass) decreased when insulin levels increased before reaching a steady state. This was reflected by a concomitant decrease in plasma glycerol levels and, in turn, a decrease in the R_d of glycerol, to which the liver is a main contributor. Despite decreased insulin levels in the obese group during the clamp, glycerol kinetics were unchanged, most likely because of adipose tissue insulin resistance.

As mentioned, evidence is accumulating for a specific interplay between circulating amino acids, glucagon secretion, and glucagon-induced hepatic amino acid turnover, a liver–amino acid–α-cell axis (10–14,16,35). It has been proposed that hepatic steatosis leads to fasting hyperglucagonemia by disturbing this axis (13,35), and the current study supports this proposal. Consistent with this, a mouse model of complete glucagon resistance displayed reduced amino acid turnover, increased plasma levels of glucagon, and proliferation of pancreatic α-cells (15). This effect was most likely mediated by a circulating factor arising from the liver (10), and later studies revealed that one or more amino acids were involved (14,16), emphasizing the feedback loop between the liver and α-cells (17). Additionally, the relationship between hepatic amino acid turnover, circulating amino acids, and pancreatic glucagon secretion is supported by cases of glucagon abundance and lack of glucagon signaling in humans. In patients with glucagonoma (i.e., glucagon-producing tumors), overt diabetes is observed in only ∼30% (35), whereas hypoaminoacidemia is reported as a main feature of the condition (36). On the other hand, humans with inactivating mutations in the glucagon receptor gene are characterized by α-cell hyperplasia, hyperaminoacidemia, and glucagon hypersecretion but normal glycemic control (37). Thus, during these extreme conditions, disturbances in glucose metabolism are not central, whereas disruption of normal glucagon signaling leads to elevated levels of amino acids, resulting in hyperglucagonemia and α-cell hyperplasia. Several amino acids have been proposed as potential drivers of the liver–α-cell axis. Tyrosine and alanine have been highlighted (11,14,15), and, consistent with this, plasma concentrations of these particular amino acids were higher in the obese group than in the lean group (Fig. 4). We were not able to measure glutamine, which has also been implicated. Of interest, a recent study of the effects of amino acids on the isolated perfused mouse pancreas showed that alanine, arginine, and proline in particular, but not glutamine, stimulated glucagon secretion (38).

In this study, the obese individuals were characterized by both fasting hyperglucagonemia and hyperaminoacidemia (Fig. 1C and Fig. 3A), as previously described (11). This could potentially be a result of disruption of the liver–α-cell axis. A main finding is the significant reduction of plasma amino acids in response to elevated glucagon levels in the lean group, but not in the obese group, despite higher glucagon concentrations among the latter. Several mechanisms for this should be considered. First, could changes in peripheral uptake of amino acids be involved? The peripheral tissue uptake of amino acids and hence protein synthesis are increased by insulin rather than glucagon (39,40), but because levels of insulin did not change during the clamp, increased peripheral uptake is an unlikely explanation. However, hepatic insulin resistance could be involved, as it has been associated with attenuation of the amino acid–lowering effect of glucagon and thus could be a causal factor in the development of impaired hepatic glucagon signaling (12). Second, could the decrease in amino acids among our lean subjects relate to renal excretion? Because healthy individuals excrete only minimal amounts of amino acids in the urine (99% of filtered amino acids are reabsorbed [41])—and this is thought to occur in a glucagon-independent manner (20)—renal effects seem unlikely to explain our results. The reduction is more likely to reflect a direct effect of glucagon on the liver. Here, glucagon is known to promote amino acid metabolism associated with gluconeogenesis and ureagenesis (20,39). In our experiments, glucose production did increase, probably reflecting glycogenolysis rather than gluconeogenesis. By measuring urea turnover we could not detect an effect on ureagenesis, but it is possible that smaller changes in urea production, as possibly elicited here by a short-lived elevation of glucagon levels, might not be detected because of the size of the urea pool and the effective renal excretion of urea, a process that might be stimulated by glucagon (42).

It is a limitation to our study that we did not apply amino acid tracers, and their use is clearly warranted in future studies. Contrary to our findings in the lean individuals, there was a slight but significant increase in amino acid levels in our obese cohort in response to the high-dose glucagon infusion (Fig. 3A and B). This finding seems paradoxical, and as mentioned above, we hypothesize that it may reflect hepatic glucagon resistance due to hepatic steatosis. On the basis of these findings, however, we are not able to establish causality, and other obesity-related (patho)physiological changes may well be involved in the observed disturbance of amino acid metabolism. Our comparison of liver biopsies from healthy lean and obese steatotic individuals provides potential mechanistic insights into the physiology underlying the observed differences between the groups. RNA sequencing revealed that several genes involved in amino acid metabolism were downregulated in the obese individuals (Fig. 5C). In comparison, only four genes involved in glucose metabolism (ENO3, PC, GOT1, and PCK1) were differentially regulated between the groups (Supplementary Table 1). This may explain why amino acid metabolism was impaired in the obese group but glucose metabolism was not. Interestingly, two of the downregulated genes encoding amino acid transporters, SLC38A3 and SLC7A2, have also been shown to be downregulated in mouse models of glucagon resistance (14,16). Furthermore, CPS1, which encodes the urea cycle enzyme carbamoylphosphate synthetase and has recently been tied to reduced ureagenesis in patients with nonalcoholic steatohepatitis (31), was also downregulated in obese subjects compared with lean (29). This suggests that the attenuated amino acid metabolism in response to glucagon in obese individuals may be a consequence of alterations in mechanisms controlling both substrate availability and the rate of hepatic ureagenesis and gluconeogenesis.

In conclusion, our study shows that elevated glucagon levels during controlled circumstances rapidly reduce plasma amino acid levels in lean individuals, whereas this effect was lost in an obese cohort with hepatic steatosis, who also had high fasting plasma levels of amino acids and glucagon. We speculate that this is a consequence of hepatic glucagon resistance due to steatosis and that downregulation of genes involved in hepatic amino acid metabolism could constitute the underlying mechanism.

Funding. This study was funded by the Novo Nordisk Foundation, the A.P. Møller Foundation, and Jacob and Olga Madsen’s Foundation.

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

Author Contributions. F.K.K. conceived the study. M.P.S., J.I.B., A.L., and F.K.K. designed the study. M.P.S., M.D., C.S., and M.J.K. collected the clinical data. G.v.H., K.R., J.L.L., N.J.W.A., and J.J.H. provided analyses. M.P.S. and J.I.B. performed statistical analyses. M.P.S. drafted the manuscript. J.I.B., A.L., M.D., G.v.H., C.S., M.J.K., K.R., J.L.L., N.J.W.A., J.J.H., T.V., and F.K.K. reviewed and edited the manuscript. M.P.S. and F.K.K. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented at the 78th Scientific Sessions of the American Diabetes Association, Orlando, FL, 22–26 June 2018.