Hypoglycemia and hyperglycemia are both predictors for adverse outcome in critically ill patients. Hyperinsulinemia is induced by inflammatory stimuli as a relevant mechanism for glucose lowering in the critically ill. The incretine hormone GLP-1 was currently found to be induced by endotoxin, leading to insulin secretion and glucose lowering under inflammatory conditions in mice. Here, we describe GLP-1 secretion to be increased by a variety of inflammatory stimuli, including endotoxin, interleukin-1β (IL-1β), and IL-6. Although abrogation of IL-1 signaling proved insufficient to prevent endotoxin-dependent GLP-1 induction, this was abolished in the absence of IL-6 in respective knockout animals. Hence, we found endotoxin-dependent GLP-1 secretion to be mediated by an inflammatory cascade, with IL-6 being necessary and sufficient for GLP-1 induction. Functionally, augmentation of the GLP-1 system by pharmacological inhibition of DPP-4 caused hyperinsulinemia, suppression of glucagon release, and glucose lowering under endotoxic conditions, whereas inhibition of the GLP-1 receptor led to the opposite effect. Furthermore, total GLP-1 plasma levels were profoundly increased in 155 critically ill patients presenting to the intensive care unit (ICU) in comparison with 134 healthy control subjects. In the ICU cohort, GLP-1 plasma levels correlated with markers of inflammation and disease severity. Consequently, GLP-1 provides a novel link between the immune system and the gut with strong relevance for metabolic regulation in context of inflammation.
Hypoglycemia and hyperglycemia are both indicators for adverse outcome in critically ill patients (1–5). Efforts to reduce mortality by tight glucose control in the intensive care setting have provided mixed results. Whereas initial single center trials suggested survival benefit of intensive glucose control (6), larger multicenter studies reported even increased mortality of the intensive glucose therapy arm (7). This has been attributed to a 10-fold increase of hypoglycemic episodes in patients with tight blood glucose control, which predicted mortality in a dose- and intensity-dependent manner (8,9). Overly tight glucose control has therefore been banned from the intensive care unit (ICU), with current recommendations aiming for blood glucose values of 140–180 mg/dL (10). Yet, the therapeutic potential of tight glucose control in a hypoglycemia avoiding modus remains under debate for ICU patients.
Inflammation causes a biphasic response of glucose metabolism. An early drop of blood glucose is followed by a secondary raise thereafter. Whereas this secondary raise can be attributed to the occurrence of insulin resistance, a variety of mechanisms contribute to the early drop of blood glucose (11,12). These include hyperinsulinemia, enhanced systemic glucose consumption, depletion of glycogen stores, and suppression of gluconeogenesis (13–15). Relevance of inflammation-dependent hyperinsulinemia has consistently been found in mice, rats, dogs, cows, and men (15–19), with inhibition of insulin secretion being able to prevent hypoglycemia in the inflammatory setting (20). Interestingly, lipopolysaccharide (LPS)-dependent insulin secretion was found to occur in a glucose-dependent manner, with pronounced insulin secretion only occurring under hyperglycemic but not euglycemic conditions (21). This suggests a protective mechanism to be in place that regulates inflammation-dependent insulin secretion.
GLP-1 is an incretin hormone, which is released from the gut in response to nutritional stimuli (22). GLP-1 binds to its receptor on endocrine pancreatic cells, leading to insulin secretion and suppression of glucagon release in a glucose-dependent manner (23). This mode of action has made GLP-1 an attractive therapeutic target for the treatment of diabetes (24). In addition, the clinical utility of GLP-1–based therapies has been evaluated for blood glucose control in critically ill patients (25). Surprisingly, GLP-1 was currently found to be induced by endotoxic conditions in mice, leading to insulin secretion and glucose lowering (26). Following up on this investigation, we here study the mechanisms of inflammatory GLP-1 secretion and explore its relevance for human biology.
Research Design and Methods
C57BL/6 mice were purchased from Charles River Laboratory. Interleukin-6 (IL-6) knockout B6.129S2-IL6tm1Kopf/J mice (IL6−/−) and IL-1 receptor knockout B6.129S7-IL1r1tm1Imx/J mice (IL1R−/−) were obtained from The Jackson Laboratory. Six-week-old male mice were housed under specific pathogen–free conditions in either individually ventilated or filter top cages with a 12-h light/12-h dark cycle with free access to autoclaved water and regular chow diet ad libitum. Mice were given intraperitoneal injections of LPS from Escherichia coli 0111:B4 (Sigma-Aldrich) (100 ng/g body weight, unless otherwise indicated), IL-1β, IL-6, and tumor necrosis factor-α (TNF-α) (all PeproTech) (4 ng/g body weight), respectively. NaCl (0.9%) was used as control. In some experiments, a single oral gavage of sitagliptin (Januvia; Merck Sharp & Dohme) (40 μg/g) was performed 1 h prior to LPS injection, or a single dose of exendin (9–39) (Bachem) (100 nmol/L/kg) was administered by intraperitoneal injection 15 min prior to LPS injections. Blood samples were collected in EDTA-containing tubes supplemented with the DPP-4 inhibitor Diprotin A (Calbiochem) and frozen at −80°C. All mice (endotoxin-injected and controls) were food deprived at the time point of LPS injections or in some experiments when sitagliptin or exendin (9–39) was applied, unless otherwise stated. All animal experiments were approved by local authorities and complied with German animal protection law.
GLUTag L cells (27) were grown and maintained in DMEM + GlutaMAX-I with glucose 1 g/L (Invitrogen) supplemented with 10% FBS and 1% penicillin/streptomycin. Medium was exchanged every 3 days, and the cells were trypsinized and reseeded at a 1:3 dilution when ∼80% confluence was reached. For GLP-1 secretion experiments, GLUTag cells were stimulated with GIP (glucose-dependent insulinotropic polypeptide) (1 μmol/L), IL-6, IL-1β, and LPS (all 100 ng/mL) for 15 min after 24-h incubation in serum-deprived medium with 0.1 mmol/L glucose. GLP-1 concentrations were measured by using a total GLP-1 ELISA Kit (Millipore).
The design of the clinical study has been reported previously (28). In brief, we enrolled 155 patients (92 male and 63 female with a median age of 64 years; range 18–90) who were admitted to the General Internal Medicine ICU at the RWTH Aachen University Hospital Aachen. Written informed consent was obtained from the patient, his or her spouse, or the appointed legal guardian. Not included in this study were patients who were expected to have a short-term (<72 h) intensive care treatment due to postinterventional observation or acute intoxication. Patients who met the criteria proposed by the American College of Chest Physicians and the Society of Critical Care Medicine Consensus Conference Committee for severe sepsis and septic shock were categorized as sepsis patients, the others as nonsepsis patients (29). As a control population, we analyzed blood samples from 134 patients with angiographic exclusion of cardiovascular disease and without diabetes from our cardiovascular biobank (103 male and 31 female with a median age of 58 years [range 19–90]; BMI 27.8 kg/m2 [range 18.2–67.8]). Blood was collected in a random nonfasting manner. Study protocols and biosampling were approved by the local ethics committee and conducted in accordance with the ethical standards laid down in the 1964 Declaration of Helsinki (ethics committee of the University Hospital Aachen, RWTH Aachen University).
Blood glucose was measured by using a glucometer (Contour; Bayer). Serum concentrations of insulin (ALPCO), glucagon (ALPCO), active GLP-1 (7–36) (Millipore), and total GLP-1 (Millipore) were determined by ELISA kits according to the manufacturers' protocol. DPP-4 enzyme activity was determined as described elsewhere (30).
Blood samples were collected upon admission to the ICU (prior to therapeutic interventions). After centrifugation at 2,000g at 4°C for 10 min, serum and plasma aliquots of 1 mL were frozen immediately at −80°C. Total GLP-1 plasma levels were determined by using a commercial ELISA kit (Millipore). IL-6 was measured by commercial chemiluminescence assays (Siemens Healthcare and R&D Systems), according to the manufacturers' instructions.
For animal studies, results are presented as mean ± SEM. Statistical significance of differences was analyzed by using unpaired Student t test.
Data are given as median and range. Differences between two groups were assessed by Mann-Whitney U test for post hoc analysis. Box plot graphics illustrate comparisons between subgroups, and they display a statistical summary of the median, quartiles, range, and extreme values. The whiskers extend from the minimum to the maximum value excluding outside and far out values, which are displayed as separate points. An outside value (indicated by an open circle) was defined as a value that is smaller than the lower quartile minus 1.5 times the interquartile range or larger than the upper quartile plus 1.5 times the interquartile range. A far out value was defined as a value that is smaller than the lower quartile minus three times the interquartile range or larger than the upper quartile plus three times the interquartile range (31).
All values, including “outliers,” have been included for statistical analyses. Correlations between variables have been analyzed using the Spearman correlation tests, where values of P < 0.05 were considered statistically significant. All statistical analyses were performed with SPSS version 12.0 (SPSS).
Consistent with a recent study, we have found LPS to dose- and time-dependently increase serum concentrations of total GLP-1 in C57BL/6 mice (26). This became apparent 120 min post–LPS injection with a maximal raise of 3.4-fold (Fig. 1A and B). Similar results were detected for active GLP-1, which was significantly elevated 120 min post–LPS injection (Fig. 1C). The GLP-1 increase was preceded by the raise of IL-6, which was already elevated 30 min post–LPS injection (Fig. 1D). No change in serum activity of the GLP-1–degrading enzyme DPP-4 was detected in response to endotoxin treatment (data not shown). This was paralleled by a significant LPS-dependent increase in serum insulin concentrations (Fig. 1E and F) and a reduction of blood glucose (Fig. 1G and H), which followed a similar time course as found for GLP-1.
Since GLP-1 induction occurred relatively delayed to early inflammatory cytokines, including IL-6, we hypothesized one of these early LPS-dependent cytokines to be responsible for the later induction of GLP-1. Indeed we found IL-1β, which is rapidly increased in response to LPS, to induce GLP-1 secretion in a similar manner as endotoxin when injected into C57BL/6 mice (Fig. 2A). This was accompanied by a 4.5-fold increase of circulating insulin concentrations (Fig. 2B). Furthermore, IL-6 was found to significantly, although less robustly, increase circulating GLP-1 levels (Fig. 2C) and increase serum insulin concentrations (Fig. 2D). No relevant effect of TNF-α was found on GLP-1 or insulin concentrations in C57BL/6 mice (Fig. 2E and F).
To evaluate whether LPS increases GLP-1 in an IL-1–dependent manner, we injected IL1R−/− mice in comparison with wild-type controls with endotoxin. Abrogation of IL-1 signaling did not, however, affect LPS-dependent GLP-1 secretion (Fig. 2G), leading to a similar raise of serum insulin and drop of blood glucose (Fig. 2H and I), whereas the glucose-lowering potential of IL-1β was lost in IL1R−/− mice (data not shown).
As an alternative mechanism, we evaluated the relevance of IL-6 for LPS-dependent GLP-1 secretion. Importantly, abrogation of IL-6 secretion in respective knockout animals markedly blunted the LPS-dependent GLP-1 induction (Fig. 2J) and abolished the increase of serum insulin (Fig. 2K) while partially preventing glucose lowering (Fig. 2L). We next evaluated the relevance of IL-6 for IL-1–dependent GLP-1 induction and found GLP-1 secretion to be similarly impaired in IL6−/− knockout mice when injected with IL-1β (Fig. 2M). Consistently, only IL-6, but not LPS or IL-1β, proved able to increase GLP-1 secretion from GLUTag cells under in vitro conditions (Fig. 2N). These data demonstrate endotoxin-dependent GLP-1 induction to be mediated by a cascade of primarily secreted inflammatory cytokines, with IL-6 being necessary and sufficient for GLP-1 secretion.
To assess the functional relevance of GLP-1 for glucose homeostasis in response to endotoxemia, we pharmacologically inhibited the GLP-1–degrading enzyme DPP-4. Single oral application of sitagliptin 1 h before LPS injection proved sufficient to lower serum DPP-4 activity of C57BL/6 mice independently of LPS (Fig. 3A). Importantly, DPP-4 inhibition strongly augmented serum concentrations of active GLP-1 in response to LPS (Fig. 3B), which was paralleled by a pronounced increase of serum insulin (Fig. 3C) and led to an additional drop of blood glucose values (Fig. 3D). Relevance of GLP-1 for LPS-dependent glucose lowering was further demonstrated by use of the GLP-1 receptor antagonist exendin (9–39), which markedly blunted LPS-dependent increase of serum insulin (Fig. 3E) and protected mice from early endotoxin-induced blood glucose lowering (Fig. 3F). These observations suggest GLP-1 to be a necessary stimulus for LPS-dependent hyperinsulinemia. Since GLP-1 is known to induce insulin secretion in a glucose-dependent manner, we evaluated whether this mode of action remained intact under endotoxemic conditions. We therefore injected mice under fed and fasted conditions with LPS with respective differences in their basal blood glucose values. Although LPS caused a similar raise of GLP-1 in both states (Fig. 3G), insulin induction was markedly blunted in the fasted state (Fig. 3H), leading to no further decrease of serum glucose (Fig. 3I). This observation confirms GLP-1 to increase insulin secretion in a glucose-dependent manner, which remained intact under inflammatory conditions.
GLP-1 is further known to reduce pancreatic α-cell glucagon secretion, which contributes to its glucose-lowering potential. To evaluate whether this regulatory mechanism remains present under inflammatory conditions, we assessed glucagon serum concentrations of mice with augmentation or inhibition of the GLP-1 system by treatment with a DPP-4 inhibitor or exendin (9–39) before endotoxin challenge. As expected, LPS increased serum glucagon concentrations, which were blunted by DPP-4 inhibition and augmented by exendin (9–39) treatment, contributing to the glucose-lowering potential of GLP-1 during inflammation and mimicking the postprandial state (Fig. 3J) (32).
We next asked whether circulating GLP-1 concentrations would also be increased in humans under inflammatory conditions. Indeed mean plasma concentrations of total GLP-1 were found to be 6.9 times higher in 155 critically ill patients admitted to the ICU in comparison with 134 healthy control subjects (97.0 vs. 13.9 pmol/L, P < 0.001) (Fig. 4A and Table 1). Among the ICU cohort, higher GLP-1 plasma concentrations were found in patients presenting with sepsis (101.1 vs. 67.5 pmol/L in nonseptic patients, P = 0.009) (Fig. 4B) and more severe disease (109.0 pmol/L in patients with APACHE II >10 vs. 73.4 pmol/L in patients with APACHE II ≤10, P = 0.002) (Fig. 4C). Positive associations of GLP-1 were found with IL-6 levels in the critically ill patients (r = 0.380, P < 0.001) (Table 2) but not in the normal control cohort (r = 0.09, P = NS). Similarly GLP-1 levels were found to be associated with the inflammatory biomarker C-reactive protein (CRP) (r = 0.384, P < 0.001) in the ICU cohort (Table 2). Significant association of GLP-1 with insulin was only found in nonseptic (r = 0.45, P < 0.05) but not septic patients (Table 2). No association of GLP-1 was found with blood glucose concentrations. Similar results were obtained when only investigating nondiabetic patients.
In this study, we found circulating GLP-1 concentrations to be markedly increased in critically ill patients, and a variety of inflammatory stimuli, including endotoxin, IL-1, and IL-6, proved sufficient to induce GLP-1 secretion in mice. This observation seems surprising given the intestinal location of GLP-1–producing L cells and their strong functional relevance in the physiology of feeding. GLP-1 is known to be secreted in response to nutritional stimuli, leading to pancreatic insulin secretion and suppression of glucagon release (22,23). We found a similar GLP-1 response to be in place under inflammatory conditions, demonstrating a profound cross-talk between the immune system and the gut. Inflammatory stimuli thereby mimic the postprandial state featuring GLP-1–dependent insulin secretion and glucagon suppression at high blood glucose levels (Fig. 5). The physiological relevance of this cross-talk remains incompletely understood but suggests evolutionary benefit of tight glucose control during acute infections.
While Nguyen et al. (26) have recently reported GLP-1 secretion to be increased in response to endotoxin, we extend this observation to other inflammatory stimuli, including IL-1β and IL-6. Interestingly, we found endotoxin- or IL-1β–dependent GLP-1 secretion to require an inflammatory cascade, leading to IL-6 as the relevant GLP-1 secreting stimulus. This observation suggests a broad cross-talk between the GLP-1 system and other IL-6–driven disease processes. Ellingsgaard et al. (33) recently reported IL-6 to cause GLP-1 induction, which they found to be relevant for insulin secretion and glucose metabolism in response to exercise but also in models of obesity. IL-6–dependent GLP-1 secretion was thereby not limited to the gut but also occurred in pancreatic α-cells, suggesting an inflammation-dependent intraislet cross-talk between α- and β-cells (33). Importantly, we found GLP-1 secretion to hold functional relevance for insulin secretion and blood glucose lowering in mice treated with inflammatory stimuli. These results are consistent with the improved GLP-1–dependent glucose tolerance of endotoxin-treated mice as reported by Nguyen et al. (26). We extend these observations to the preserved GLP-1–dependent suppression of glucagon release under inflammatory conditions. In addition, we found GLP-1–dependent insulin secretion to occur only in fed but not in the fasted mice, making the GLP-1 system an unlikely cause for inflammation-dependent hypoglycemia. Any blood glucose lowering, also if done to the normal range, might however increase the vulnerability of the organism to other hypoglycemic stimuli. Endotoxin is thereby known to suppress hepatic gluconeogenesis via activation of TLR4, MyD88, and NF-κB, enhance systemic glucose consumption, and deplete glycogen stores, which in conjunction can lead to hypoglycemia (13–15). Indeed, blockade of the GLP-1 system by use of GLP-1 receptor antagonists or IL-6 knockouts markedly blunted endotoxin-dependent hyperinsulinemia in our experiments but only modestly and temporarily counteracted the glucose-lowering effects of LPS. This suggests a predominant role of additional glucose-lowering stimuli under inflammatory conditions.
Clinical relevance of the GLP-1 system during acute inflammation is suggested by the marked increase of circulating GLP-1 levels found in critically ill patients in this study. Importantly, the observed difference of GLP-1 between the ICU cohort and normal control subjects greatly extended the physiological raise of GLP-1 expected to occur in response to feeding or after bariatric surgery (34,35). As limitations of our study, blood samples of both cohorts were taken in a random, nonfasting manner, making the time since last food ingestion a possible confounding factor. As critical illness reduces appetite, this is however unlikely to explain the observed differences of GLP-1 levels. In addition, critical illness does cause multiple metabolic alterations with impairment of different organ functions. This might hold relevance for the observed increase of circulating GLP-1. Importantly, we found higher GLP-1 concentrations in patients with more severe disease and sepsis in comparison with those with less severe disease and without sepsis within the ICU cohort. In addition, significant associations of GLP-1 were found with markers of inflammation, including IL-6 and CRP, within the ICU cohort. This does suggest an inflammation-dependent regulation of GLP-1 secretion to be present in humans. Consistently, others have reported circulating GLP-1 to be increased in states of chronic inflammatory disease, including the metabolic syndrome, coronary artery disease, or heart failure (36–38).
No associations of GLP-1 levels were found with glucose concentrations in the ICU cohort, and only modest associations were found with insulin concentrations in nonseptic but not septic patients. This might be explainable by the heterogeneous characteristics of the critically ill patient cohort. In addition, associations with metabolic parameters in a cross-sectional analysis might require assessment of “active GLP-1” as the more temporary, insulin-stimulating form of the peptide. Longitudinal assessment of circulating GLP-1 in direct comparison with insulin secretion and glucose metabolism will help to evaluate the glucose regulatory relevance of the GLP-1 system in critically ill patients.
In addition, it will be of major relevance to evaluate the prognostic implications of an activated GLP-1 system in critically ill patients. Experimental evidence suggests beneficial effects of GLP-1 during sepsis, with reduced mortality being found in endotoxin-challenged DPP-4 knockout rats and activation of the GLP-1 receptor improving prognosis in the same model (39). Mechanistically this might be mediated by anti-inflammatory capacities of GLP-1, leading to reduced oxidative stress and tissue protection (40,41).
In conclusion, we found GLP-1 concentrations to be strongly increased in critically ill patients. These results were recapitulated in mice in which endotoxin increased GLP-1 secretion by an inflammatory cascade depending on IL-6. This proved to be relevant for inflammation-dependent insulin secretion, which we found to happen in a glucose-dependent manner. Additional studies are required to further characterize the glucose regulatory capacity of GLP-1 in critically ill patients and investigate its prognostic relevance.
Acknowledgments. The authors thank Dan Drucker (Department of Medicine, Mt. Sinai Hospital, Lunenfeld-Tanenbaum Research Institute, University of Toronto, Toronto, Ontario, Canada) for the kind permission to use GLUTag cells, Katharina Reising (Department of Internal Medicine I, University Hospital Aachen, Germany) for experimental support, and the Clinical Trial Center Aachen and the Biobank of the RWTH Aachen University for sample collection and storage of the normal control population.
Funding. This study is supported by grants from the Marga und Walter Boll-Stiftung, EASD/Amylin Paul Langerhans Grant and German Scientific Foundation (LE1350/3-1 to M.L.), and the META OBIHEP BMBF (360365 to C.T.).
Duality of Interest. No potential conflicts of interest relevant to this article were reported.
Author Contributions. F.K. was involved in experimental design and procedure, data acquisition and analysis, and writing the manuscript. C.M. was involved in experimental design and procedure and data acquisition and analysis. J.M., S.D., H.M.F., and C.L. were involved in experimental procedure and data acquisition. C.T. and A.K. performed the ICU study. F.T. performed the ICU study and statistical analysis of the clinical study and was involved in writing the manuscript. N.M. was involved in experimental design and writing the manuscript. M.L. performed the experimental design, was involved in data analysis, and wrote the manuscript. M.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.
Prior Presentation. This study was submitted to the 74th Scientific Sessions of the American Diabetes Association, San Francisco, CA, 13–17 June 2014.