Lipopolysaccharides-Mediated Increase in Glucose-Stimulated Insulin Secretion: Involvement of the GLP-1 Pathway

Lipopolysaccharides (LPS) or endotoxins are components of the cell wall of gram–negative bacteria. When LPS enters the bloodstream, it activates Toll-like receptor 4 (TLR4), which is located at the surface of immune cells, leading to the release of proinflammatory cytokines and inflammation (1). Excessive inflammatory response to LPS can be harmful and may lead to endotoxic shock and ultimately death. Because gram–negative bacteria are present in large amounts in the gut lumen and the environment, bacterial translocation and the resulting endotoxemia can occur as a result of inflammatory bowel disease (2), a high-fat diet (3), or cigarette smoking (4). Whatever the severity of the insult, endotoxemia is accompanied by marked metabolic alterations, including changes in glucose metabolism, which is of particular importance. Early hyperglycemia is a common metabolic feature following exposure to bacterial endotoxins (5). It is known to contribute to poor immune function, oxidative stress, increased susceptibility to infectious complications, impaired recovery of organ failure, and endothelial and myocardial dysfunction (6,7). In critically ill patients, hyperglycemia is not only a marker of illness severity but also a predictor of a poor clinical outcome (7). Although effective glucose control through aggressive insulin therapy has been reported to improve clinical outcome, it remains unclear to what extent insulin infusion may in fact be beneficial (8). Hyperglycemia, moderate-to-severe hypoglycemia, and hyperinsulinemia have all been found to be associated with an increased risk of death in critically ill patients admitted to the intensive care unit (9,10). In LPS-treated mice, a decrease in both inflammation and mortality was reported when the plasma glucose level was strictly maintained at normal values, suggesting that appropriate/tight glucose control is a major determinant of outcome (11). Whether abnormalities in glucose control may vary according to the endotoxemic insult is unknown, and the molecular mechanisms involved are poorly understood.

In the current study, the glucose and insulin responses to experimentally induced hyperglycemia were investigated in LPS-treated mouse models. Endotoxins were administered through either the acute intraperitoneal injection of a single dose of LPS, a continuous intraperitoneal infusion of a low dose of LPS using mini-osmotic pumps, or the use of the phospholipid transfer protein knockout (PLTP-KO) mouse model, which is known to display an impaired capacity to inactivate and detoxify LPS in the body (12). In all of the models used in the current study, we found that LPS increased glucose-stimulated insulin secretion (GSIS), which has been shown to be due, at least in part, to an increase in the level and activity of glucagon-like peptide 1 (GLP-1).

Five- to six-month-old WT mice (Charles River, L'Arbresle, France), GLP-1 receptor knockout (GLP-1R-KO) mice (13), and PLTP-KO mice (a kind gift of Dr. Jiang [14]) from a homogeneous C57BL6/J background were housed in a controlled environment. All of the animal experimental procedures were validated by the local ethics committee of the University of Burgundy (protocol number 1209). Mice had free access to water and standard diet—A03-10 Rats/Mice Breeder Diet (SAFE, Augy, France). The diet contained 21.4% protein, 5.7% mineral, 5.1% fat, 4% cellulose, and mixture of vitamins and micronutrients.

LPS were from Escherichia coli 055:B5 (Sigma, St. Louis, MO). The mice were either given intraperitoneal injections of LPS (0.5, 1, or 2 mg/kg) or had a mini-osmotic pump implanted subcutaneously (Alzet Model 2004; Alza, Palo Alto, CA) and linked to the peritoneal cavity by a catheter. The pumps were filled with either NaCl (0.9%) or LPS to infuse 300 or 1,000 μg ⋅ kg⁻¹ ⋅ day⁻¹ for 4 weeks. After 4 weeks, plasma samples were drawn and frozen at −20°C. The mice were then killed and muscle tissue and the liver were immediately excised, immersed in liquid N₂, and stored at −80°C for further mRNA analyses.

After fasting overnight, the mice were fed with glucose by oral gavage (1.5–2 g/kg). Blood was drawn from the tail vein at 0, 15, 30, 60, 90, and 120 min following the glucose load, and glucose concentration was determined with a glucose meter (OneTouch Ultra, Milpitas, CA). Oral glucose tolerance tests (OGTTs) were performed either 6 h after the acute intraperitoneal injection of LPS or at the end of the 28-day infusion period. In some experiments, sitagliptin (Januvia, Merck Sharp and Dohme-Chibret, France) (5 mg/kg) or exendin(9–39) (Ex-9) (5 μg/mouse) was administered via oral gavage or intraperitoneal injection, respectively, 30 min prior to the glucose gavage. NaCl (0.9%) was administered as a control. The area under the curve (AUC) (0–120) was calculated for each group of mice, and the plasma insulin concentration was measured 30 min before and 15 min after the glucose load. The insulinogenic index (IGI) (i.e., an index of β-cell function) was calculated as the Δ insulin to the Δ glucose ratio (ΔI_0–15/ΔG_0–15, pmol/mmol)

After 6 h of fasting, mice were given an intraperitoneal injection of 0.5 U/kg insulin (Humalog), and blood glucose was measured as mentioned above.

After fasting overnight, mice were given an injection of the dipeptidyl peptidase 4 (DPP4) inhibitor sitagliptin (5 mg/kg). One hour later, the mice were given an intraperitoneal injection of LPS (2 mg/kg) or NaCl (0.9%) as the control. Six hours after the LPS or NaCl injection, the mice were given 2 g/kg of glucose through oral gavage. Blood samples were collected in EDTA-containing tubes supplemented with the DPP4 inhibitor (Millipore, France) 30 min before and 15 min after the glucose load to assess the plasma GLP-1 concentration.

Total RNA was isolated from tissues using Trizol reagent (Invitrogen, France). Total RNA (0.5 µg) was reverse transcribed using MMLV reverse transcriptase (Invitrogen) and random primers at 42°C for 1 h. Quantitative PCRs were performed using ABI Prism 7900 (Applied Biosystems, France). Primer sequences for the targeted mouse genes are available upon request. The mRNA expression level of target genes was normalized to levels of β-actin mRNA and the results were expressed as relative expression levels. The data were quantified by the method of 2^-ΔΔCt.

Muscles were isolated and preincubated for 10 min in Krebs-Henseleit buffer (pH 7.4) containing 2 mg/mL BSA, 2 mmol/L sodium pyruvate, and 20 mmol/L HEPES. The muscles were then incubated for 45 min in the absence or presence of 100 nmol/L insulin as previously described (15).

LPS concentrations were measured in total plasma, HDL, and the lipoprotein-free fraction (LFF) through the quantitation of 3-β-hydroxymyristate concentration according to the general procedure previously described (12). Plasma insulin and GLP-1 concentration were determined by ELISA (ALPCO, Salem, NH), in accordance with the manufacturer’s instructions. Muscle glycogen content was determined with the glycogen synthase assay kit (BioVision, Mountain View, CA), in accordance with the manufacturer’s instructions. The concentrations of interleukin-6 (IL-6) and tumor necrosis factor-α (TNFα,) were measured in mouse plasma by Milliplex MAP 5-Plex kit (mouse cytokine/chemokine magnetic bead panel) from Millipore (Millipore, Billerica, MA), according to the manufacturer's protocol and using a LuminexR apparatus (Bio-Plex 200, Bio-Rad).

Results are presented as mean ± SEM. Statistical significance of differences was analyzed by the unpaired t test.

To evaluate the glucose and insulin responses during LPS-mediated inflammation, wild-type (WT) mice were given a single acute intraperitoneal injection of 2 mg/kg of LPS. Plasma β-hydroxymyristate content, reflecting plasma LPS level, was markedly increased upon LPS injection (603.2 ± 12.2 ng/mL in LPS-injected mice [n = 8] vs. 68.4 ± 16.2 ng/mL in NaCl controls [n = 8]; P < 0.001). Six hours after the LPS injection, hyperglycemia was induced by gavage with a glucose solution (2 g/kg). As shown in Fig. 1A and B, the glucose tolerance curve was improved with the LPS injection and GSIS increased (Fig. 1C). Consistent results were obtained by using two lower doses of LPS. Whereas in fasting conditions, no significant differences were observed in insulin plasma levels upon injection of either 0.5 mg/kg or 1 mg/kg of LPS (data not shown). Again, with either dose of LPS, GSIS after the glucose load was significantly enhanced as revealed by the insulin plasma levels at t = 15 min (766.9 ± 125.5 pmol/L in LPS-injected mice [n = 10] vs. 434 ± 37.49 pmol/L in NaCl controls [n = 9], P < 0.05 with the 0.5 mg/kg dose; 666.5 ± 89.39 pmol/L in LPS-injected mice [n = 10] vs. 373.8 ± 31.16 pmol/L in NaCl controls [n = 10], P < 0.01 with the 1 mg/kg dose). The function of pancreatic β-cells was evaluated by calculating the IGI at the 15-min time point (i.e., in the very early phase of insulin secretion). As shown in Fig. 1D, the acute intraperitoneal injection of LPS increased the IGI.

We next assessed the consequences of sustained LPS administration by implanting mice with mini-osmotic pumps that continuously delivered LPS or NaCl for 28 days (300 µg/kg/day). This LPS dose was reported to mimic the metabolic endotoxemia induced by a high-fat diet (3). Plasma β-hydroxymyristate content was modestly but significantly increased upon LPS infusion (LPS-infused mice [n = 7], 51.9 ± 3.9 ng/mL; NaCl controls [n = 4], 35.8 ± 1.3 ng/mL; P < 0.05). In contrast to the acute intraperitoneal injection of LPS (Fig. 1A), the continuous intraperitoneal infusion of LPS did not improve the glycemic response during the OGTT (Fig. 2A), as indicated by similar values for the AUC (Fig. 2B). However, GSIS and the IGI were increased with LPS treatment, although to a lesser extent than with acute LPS injection (Fig. 2C and D vs. Fig. 1C and D). Using a higher dose of LPS (1 mg/kg/day) during continuous infusion led to the same enhancement of GSIS after a glucose load as shown by the insulin plasma levels at t = 15 min (LPS-infused mice [n = 4], 892.5 ± 131.4 pmol/L; NaCl controls [n = 4], 431.8 ± 10.1 pmol/L; P < 0.05).

To further evaluate the impact of LPS on glucose and insulin responses, we took advantage of the PLTP-KO mouse model, which displays a reduced capacity to detoxify LPS (12). After acute intraperitoneal LPS injection, plasma β-hydroxymyristate content was higher in PLTP-KO mice as compared with WT mice (371.5 ± 24.3 ng/mL in PLTP-KO mice [n = 6] vs. 298.7 ± 23.8 ng/mL in WT controls [n = 6], P < 0.05). In the absence of LPS, PLTP expression per se did not modify the glucose response to oral glucose challenge (Fig. 3A and B, NaCl). In LPS-treated mice, glucose clearance was increased to a significantly greater extent in PLTP-KO mice than in WT mice (Fig. 3A and B, LPS). Whereas GSIS was not dependent on PLTP expression in the absence of the LPS challenge (Fig. 3C, NaCl), insulin levels were significantly greater in PLTP-KO mice after acute LPS injection (Fig. 3C, LPS). Similarly, the IGI was significantly increased after acute intraperitoneal injection of LPS in WT and PLTP-KO mice (Fig. 3D).

We next assessed the consequences of chronic LPS infusion in WT and PLTP-KO mice. Continuous infusion of a low dose of LPS did not lead to significant differences in LPS plasma levels between WT and PLTP-KO mice after 28 days of LPS infusion (52.3 ± 2.6 ng/mL in PLTP-KO mice [n = 8] vs. 56.6 ± 3.5 ng/mL in WT mice [n = 7], P = NS). However, and as PLTP is involved in LPS transfer between the plasma LFF and plasma lipoproteins, we next determined whether this experimental set up was effective to modulate LPS distribution between HDL and LFF as reported earlier with acute intraperitoneal injection (12). In agreement with earlier observations, PLTP-KO mice that received a continuous infusion of LPS through the implanted mini-osmotic pump displayed a lower 3-β-hydroxymyristate acid content of the plasma HDL fraction (Fig. 4A), a higher LPS content of the LFF fraction (Fig. 4B), and a lower HDL-to-LFF LPS ratio (Fig. 4C). In further support of the greater proinflammatory effect of LPS when infused in mice with a PLTP-deficient trait, liver Tnf-α mRNA levels were significantly higher in PLTP-KO than in WT mice (Fig. 4D). Compared with data before pump implantation (Fig. 5A and B, −LPS), glucose clearance was increased in LPS-infused WT mice, and the increase was even greater in PLTP-KO mice (Fig. 5A and B, +LPS). Insulin levels and the IGI both increased with LPS infusion, independently of PLTP expression (Fig. 5C and D). In contrast, insulin sensitivity was higher in PLTP-KO mice (Fig. 5E and F).

Because glucose clearance and insulin sensitivity in LPS-infused PLTP-KO mice were greater than those in LPS-infused WT mice, the ability of muscle to use glucose was investigated. As shown in Fig. 6A, insulin-mediated glucose uptake was more efficient in PLTP-KO than in WT mice. Consistently, muscle Glut 4 and Glycogen synthase-1 mRNA levels were significantly higher in PLTP-KO than in WT mice (Fig. 6B and C, respectively). The glycogen content of muscle was also higher in PLTP-KO mice (Fig. 6D).

As GLP-1 was previously identified as a downstream target of inflammatory cytokines that also regulates the magnitude of GSIS (16,17), we examined the GLP-1 axis during an LPS challenge. Under fasting conditions, intraperitoneal injection of LPS significantly increased plasma GLP-1 levels in WT mice (Fig. 7A, left panel: 6.32 ± 0.65 pmol/L in LPS-treated vs. 1.16 ± 0.17 pmol/L in untreated WT mice; P < 0.01). Furthermore, plasma GLP-1 levels increased further after oral glucose loading in control WT mice, and the increase was significantly greater in LPS-treated WT mice (Fig. 7A, right panel: 17.8 ± 2.98 pmol/L in LPS-treated vs. 4.07 ± 0.17 pmol/L in control mice; P < 0.001). Finally, LPS-induced increases in plasma insulin levels were observed both in fasting and after the glucose load (Fig. 7B), and a significant, positive correlation was found between plasma GLP-1 and insulin levels (Fig. 7C).

Next, we assessed whether the acute LPS-mediated attenuation of glycemic excursion could be further enhanced by increased levels of active GLP-1 achieved through treatment with sitagliptin, a DPP4 inhibitor (18). Glycemic excursion in response to LPS was further reduced by the combination of LPS and sitagliptin (Fig. 7D and E).

To assess the importance of GLP-1 as a transducer of the LPS-mediated glycemic response, we assessed whether the administration of the GLP-1R antagonist Ex-9 could modify the glucoregulatory actions of LPS. Administration of Ex-9 substantially attenuated the LPS-mediated change in glucose disposal (Fig. 7F and G).

To further investigate the role of the GLP-1/GLP-1R pathway in the LPS-mediated changes in glucose homeostasis, we assessed LPS action in GLP-1R-KO mice (19). As compared with NaCl-injected mice, which displayed low levels of LPS (WT mice [n = 11], 33.6 ± 3.5 ng/mL; GLP-1R-KO mice [n = 11], 43.9 ± 6.2 ng/mL; P = NS), plasma LPS concentrations were markedly increased after acute intraperitoneal injection of LPS and to a similar extent in WT and GLP-1R-KO mice (WT mice [n = 11], 362.0 ± 17.5 ng/mL; GLP-1R-KO mice [n = 11], 352.1 ± 25.2 ng/mL; P = NS). Interestingly, whereas acute intraperitoneal injection of LPS led to the same LPS plasma level in WT and GLP-1R-KO mice, the plasma levels of IL-6 and TNFα (i.e., two main proinflammatory cytokine targets of LPS) displayed a more pronounced increase in GLP-1R-KO as compared with WT mice (Fig. 8A and B). The LPS-mediated change in glucose response in WT mice was completely absent in GLP-1R-KO mice (Fig. 8C–E). Furthermore, the LPS-mediated increase in GSIS observed in WT mice was not observed in GLP-1R-KO mice (Fig. 8F and G), and enhanced β-cell function in response to LPS was completely lost in GLP-1R-KO mice (Fig. 8H).

In the current study, the insulin response to experimental hyperglycemia was studied in mice with LPS-mediated inflammation. Our data support the hypothesis that LPS increases GSIS, resulting in increased glucose clearance. Consistent observations were made with acute versus continuous LPS infusion and, to a greater extent, when the PLTP-mediated detoxification pathway of LPS did not operate. The LPS-mediated changes in glucose homeostasis were found to require a functional GLP-1R–dependent pathway.

Sepsis is known to impair whole body glucose disposal (20), resulting in hyperglycemia, often associated with metabolic adaptation and increased insulin secretion. However, bacterial infection or LPS administration does not always lead to consistent changes in glucose disposal or insulin secretion (21–28). For instance, the prevention of hyperglycemia by depleting glycogen storage did not abrogate the increase in serum insulin levels in endotoxemic animals (29). This indicates that the relationship between endotoxemia and glucose homeostasis is rather complex, and one hypothesis holds that the effect of LPS might be dependent on the metabolic context and the experimental conditions used (including in particular the magnitude and duration of the endotoxemic challenge). In the current study, glucose and insulin responses were first investigated in mice given a single intraperitoneal injection of LPS. Six hours after the LPS injection, GSIS significantly increased and consistent observations were made with distinct LPS doses. GSIS increase was associated with increased glucose disposal. In contrast, in an earlier study, a decrease in glucose disposal was reported 48 h after the LPS injection (30). These contrasting observations might be partly related to temporal differences. In support of this hypothesis, plasma LPS concentrations were reported to peak 9 h after a single intraperitoneal injection of LPS in mice, but returned to barely detectable levels after 48 h (12). In keeping with the above, increased insulin sensitivity was observed 2 h after LPS injection in humans (31,32), whereas insulin resistance was observed 24 h after the initial LPS challenge (33).

The hypothesis of a differential impact of transient versus continuous LPS on glucose metabolism was addressed in the current study by comparing a single injection of LPS with a continuous infusion achieved using mini-osmotic pumps. After 28 days of continuous infusion of LPS, GSIS was found to be significantly increased and again consistent observations were made with distinct LPS doses. However, there were no changes in glucose disposal. Finally, to further explore the effect of experimental endotoxemia on glucose and insulin responses in vivo, we studied PLTP-deficient mice in which the binding of LPS to plasma HDL is delayed. LPS-lipoprotein complexes are known to be less pyrogenic and less active in inducing the release of proinflammatory cytokines than is the case for LPS in the plasma LFF (34–36). The PLTP-deficient trait is therefore associated with impaired neutralization and clearance of LPS (12). Interestingly, using both acute injection or continuous intraperitoneal infusion, endotoxemia was consistently associated with greatly increased GSIS and glucose disposal in PLTP-deficient mice. These observations support the hypothesis that GSIS is an adaptive response to endotoxemia.

The underlying mechanism of the LPS-mediated changes in glucose metabolism has been a matter of intense research and speculation. On the one hand, LPS itself might exert a direct insulin-like activity (37–39). Alternatively, LPS might act indirectly through the stimulation of pancreatic insulin secretion, and a pancreas excised from a LPS-treated animal displays increases in both basal insulin secretion (40,41) and GSIS (40–42). However, neither perfusion of excised pancreas with LPS (41), nor LPS treatment of cultured pancreatic β-cells (expressing TLR4 [43]) were able to directly increase insulin secretion (44), suggesting a more complex pathway with intermediate mediators. IL-1, which has been shown to increase insulin secretion in endotoxemic rats (45), and IL-6, which has been shown to promote hypoglycemia during acute LPS-induced inflammation (46) and to enhance insulin secretion by increasing GLP-1 secretion from L cells and α-cells (17), are possible glucoregulatory targets of LPS. Although IL-6 secretion is known to be highly inducible by LPS, a LPS challenge was not used in the studies cited above. This led us to investigate whether increased production of GLP-1 (a potent inducer of insulin secretion [16]) might underlie the LPS-mediated increase in GSIS. Our data clearly show that LPS injection increases plasma levels of GLP-1 under both fasting and glucose-stimulated conditions. Moreover, increased levels of GLP-1 correlated positively with increased GSIS and were associated with higher glucose disposal during a OGTT. The glucose response after treatment with LPS was of lower magnitude when GLP-1 degradation was prevented with sitagliptin (18). Ex-9, an antagonist of the GLP-1 receptor, prevented the LPS-mediated change in glucose homeostasis. Finally, the LPS-increased GSIS and glucose disposal was totally blunted in GLP-1R-KO mice. Collectively, these findings argue for an important role for GLP-1 as a downstream metabolic target of a subset of LPS actions. Indeed, GLP-1 has been shown to enhance glucose uptake and to increase the conversion of glucose into glycogen (47) in muscle, thus contributing to an overall increase in insulin sensitivity (48). Similar findings were observed in the current study when PLTP-KO mice were continuously infused with LPS. In humans, GLP-1 administration has been associated with increased β-cell function as revealed by the increase in the IGI (49). In the current study, the higher IGI after LPS administration was completely abrogated in GLP-1R-KO mice that, as compared with WT mice, displayed a more pronounced increase in plasma levels of IL-6 and TNFα after acute LPS injection.

The current study adds to the recently recognized enteroendocrine pathway through which the proinflammatory mediator IL-6 (well known as an LPS target) can contribute to increased insulin secretion. Because IL-6 increases responsiveness to glucose through the increased production of GLP-1 by intestinal L cells, IL-6 results in increased insulin secretion and production by pancreatic β-cells (17). It is possible that the same molecular mechanism may apply to both the IL-6–mediated (45) and the LPS-mediated effects (present study). It was suggested that IL-6 increased GLP-1 production through increased proglucagon transcription and prohormone convertase (PC)1/3 expression (17). In keeping with this hypothesis, LPS sensitizes adenyl cyclase activity (50) and increases the production of cAMP (51), which is known to be involved in the regulation of proglucagon gene expression (52). In addition, the proglucagon protein precursor is known to undergo posttranslational processing through the action of two main PCs, PC1/3 and PC2 (53), Given that both PC1/3 and PC2 are induced when immune cells are exposed to LPS (54,55), a role for enhanced PC1/3 activity cannot be excluded. Alternatively, neural pathways may also be involved, as sensory nerves have been shown to be involved in the glucose metabolic response to endotoxins (56) and the expression of the neuropeptide calcitonin gene-related peptide, which is thought to play a role in the regulation of GLP-1 secretion (57), is upregulated by LPS (58). Finally, a direct effect of the LPS challenge on TLR4 signaling might make a significant contribution to the LPS-induced production of GLP-1. Indeed, it has recently been shown that enteroendocrine cells expressed functional TLR4 (59), and hypoglycemia and GSIS were reported to be blunted in TLR4-deficient mice (30).

Cumulative evidence in favor of a link between endotoxemia, enteroendocrine cells, and glucose metabolism was recently brought in patients with inflammatory bowel diseases. In Crohn disease with abnormally elevated endotoxemia (2), enteroendocrine cell activity was reported to be enhanced (60). It was associated with higher expression and circulating levels of GLP-1 (60,61). In addition, DPP4 expression and activity are reduced in patients with Crohn disease, thus making a putative contribution to increased GLP-1 expression levels and hyperinsulinemia of inflammatory bowel disease (62,63). Observations of the current study come in direct support of sustained endotoxemia as a significant contributor to enhanced insulin secretion, which constitute one main trait of inflammatory bowel diseases, and a protective factor on relapse rate (61,64).

Acknowledgments. The authors thank Philip Bastable, of Centre Hospitalier Universitaire, Dijon, France, for manuscript editing.

Funding. This work was supported by INSERM U866, the Conseil Régional de Bourgogne, The Fonds Européen de Développement Régional, the Université de Bourgogne, and a French government grant managed by the French National Research Agency under the program “Investissements d’Avenir” (ANR-11-LABX-0021).

Duality of Interest. A.T.N. is supported mostly by the Popular Committee of Dong Thap province (Vietnam) and by AgroSup Dijon (France). D.J.D. is supported by the Canada Research Chairs program and the Banting & Best Diabetes Centre-Novo Nordisk Chair in Incretin Biology. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. A.T.N. and C.D. researched data. S.M. and V.D. researched data and reviewed and edited the manuscript. P.V. and P.B. reviewed and edited the manuscript. D.J.D. contributed to discussion and reviewed and edited the manuscript. L.L. contributed to discussion and wrote, reviewed, and edited the manuscript. J.G. designed research, researched data, and wrote the manuscript. J.G. 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.