Acute Hyperinsulinemia Restrains Endotoxin-induced Systemic Inflammatory Response: An Experimental Study in a Porcine Model

The National Committee on Animal Research Ethics (Copenhagen, Denmark) approved the protocol, and the work was performed according to the guidelines in the Guide for the Care and Use of Laboratory Animals . 10Thirty-eight female Landrace pigs (weight, 35–40 kg) were fasted overnight but were allowed free access to water. They were premedicated with intramuscular ketamine (10 mg/kg) and midazolam (0.25 mg/kg). Anesthesia was induced with intravenous ketamine (5 mg/kg). The pigs were orally intubated, and anesthesia was maintained with a continuous intravenous infusion of fentanyl (60 μg · kg⁻¹· h⁻¹) and midazolam (6 mg · kg⁻¹· h⁻¹). Neuromuscular blocking drugs were not used. The animals were ventilated with a volume-controlled ventilator (Servo 900 ventilator; Siemens Elema, Solna, Sweden) with a positive end-expiratory pressure of 5 cm H²O. Tidal volume was kept at 10–15 ml/kg, and the respiratory rate was adjusted (20–25 breaths/min) to maintain normocapnia (arterial carbon dioxide tension [Paco²] in the range of 34–45 mmHg). Ventilation was performed with oxygen in air aiming at an arterial oxygen tension (Pao²) greater than 105 mmHg. After dissection of the neck vessels, a Swan-Ganz catheter (Edwards Lifescience Corp., Irvine, CA) was inserted in the pulmonary artery via  the right cava superior vein. Localization of the balloon-tipped catheter was determined by observing the characteristic pressure trace on the monitor as it was advanced through the right side of the heart into the pulmonary artery. Another catheter (5 French; St. Jude Medical Company, St. Paul, MN) was inserted into the left carotid artery for continuous blood pressure monitoring and blood sampling. A urine catheter was inserted for urine collection.

Expired minute volume, tidal volume, respiratory rate, peak and mean inspiratory pressure, and inspired oxygen fraction (Fio²) were monitored continuously throughout the study. Blood samples were collected from the arterial catheter every 30 min and more frequently in relation to lipopolysaccharide infusion for analysis of blood gases and acid–base status.

Continuous observations were performed of arterial blood pressure, heart rate (from the electrocardiogram), and pulmonary artery pressure (PAP). Pulmonary capillary wedge pressure was measured intermittently. Cardiac output was continuously monitored using an Edwards Vigilance Monitor (Edwards Lifescience Corp.).

Animals were allocated into four groups in random order as shown in figure 1. After induction of anesthesia and tracheal intubation, the animals were left to stabilize for 60 min. The animals in group 1 (anesthesia, n = 10) were subjected to general anesthesia for 570 min, solely. In group 2 (HEC, n = 9), the animals were in addition exposed to HEC for 570 min. In group 3 (lipopolysaccharide, n = 10), the animals were given a 180-min infusion of lipopolysaccharide. Finally, the animals in group 4 (lipopolysaccharide–HEC, n = 9) were subjected to a combination of HEC and lipopolysaccharide infusion. During the 570-min study period, the animals in group 1 (anesthesia) and group 2 (HEC) received 10 ml · kg⁻¹· h⁻¹isotonic saline, whereas the animals in the two other groups received 20 ml · kg⁻¹· h⁻¹to maintain blood pressure during and after lipopolysaccharide infusion. At study termination, the animals were killed by evisceration of the heart preceded by a fentanyl bolus of 1 mg.

Escherichia coli  lipopolysaccharide endotoxin, (E. coli  026:β6, Bacto Lipopolysaccharides; Difco Laboratories, Detroit, MI) was dissolved in saline 120 min before each experiment to dissolve any precipitate. After a stabilization period, lipopolysaccharide infusion was started at baseline at a rate of 2.5 μg · kg⁻¹· h⁻¹, and it was increased stepwise to 15 μg · kg⁻¹· min⁻¹during 30 min. After this, the fusion was kept at a rate of 2.5 μg · kg⁻¹· h⁻¹during 150 min and was thereafter discontinued.

Insulin (Insulin Actrapid; Novo-Nordisk, Copenhagen, Denmark) was infused intravenously at a constant rate of 0.6 mU · kg⁻¹· min⁻¹for 570 min. Plasma glucose was clamped at 5 mm by infusion of 20% glucose.

Blood for determination of insulin, glucagon, and FFAs was drawn hourly throughout the study, and blood for cytokines was drawn at 0 (baseline), 240, 480, and 570 min. Blood gases, plasma lactate, and base excess were analyzed by an ABL 700 (Radiometer, Copenhagen, Denmark) hourly.

Plasma glucose concentrations were measured every 5–10 min. They were determined in duplicate immediately after sampling (Beckman Instruments, Palo Alto, CA). Serum insulin concentrations were measured in duplicate by a two-site immunospecific insulin enzyme-linked immunosorbent assay (DAKO Diagnostics, Cambridgeshire, United Kingdom). Plasma glucagon was determined by an in-house radioimmunoassay, and serum FFA was assayed by a colorimetric method using a commercial kit (Wako Chemicals, Neuss, Germany).

Fresh frozen plasma samples (−80°C) obtained from EDTA-stabilized blood were used for cytokine measurements with enzyme-linked immunosorbent assays. Commercially available kits specific for porcine tumor necrosis factor (TNF) α and interleukin (IL) 10 (Biosource International, Camarillo, CA) were used according to the manufacturer’s instructions. Determinations were performed in duplicate. The detection levels for the assays were TNF-α less than 6 pg/ml and IL-10 less than 3 pg/ml.

Concentrations of IL-6 and IL-8 were determined in EDTA plasma from arterial blood samples in a sandwich immunoassay. Porcine specific matched pairs of anticytokine antibodies along with recombinant cytokine standards were obtained from R&D Systems (Abingdon, United Kingdom). All detecting antibodies were biotinylated. A time-resolved fluorometric assay using an Eu³⁺-streptavidin complex for detection in a Wallac Victor 2 Fluorometer (Wallac, Turku, Finland) was used. After appropriate titration of antibody concentrations, control parameters regarding assay sensitivity, precision, recovery from spiking, and dilutional linearity were produced. All standard curves, controls, and samples were analyzed in a dilution of 50% EDTA plasma. Standardized pooled cytokine-free sterile filtered porcine EDTA plasma was used for standards and controls. A dilution buffer of phosphate-buffered saline, 2% bovine serum albumin, and 0.05% Tween-20 was used. All samples were run in triplicates and analyzed in a blinded, randomized manner. Standard curves were created using a four-parameter logistic curve fit (Microcal Origin; Universal Imaging Corporation, Downingtown, PA). Median sample concentrations were calculated from this.

Microtiter plates (Nunc, Copenhagen, Denmark) were coated with capture antibody in phosphate-buffered saline and incubated for 18 h at 40°C. Blocking unspecific binding sites was done with phosphate-buffered saline, pH 7.4, containing 5% nonfat dried milk and 0.1% Tween-20 for 2 h. Plates were washed three times in 50 mm Tham, pH 7.3, containing 0.2% Tween-20 between all steps hereafter. Standards, samples, and controls were loaded and incubated for 2 h at ambient temperature. Plates were incubated with secondary antibody in dilution buffer for 2 h at ambient temperature. Eu³⁺-streptavidin (1:2,000) in assay buffer (Wallac) was added and incubated for 30 min. One hour of signal enhancement (Enhancement Solution; Wallac) was performed, followed by analysis in the fluorometer and registration of light intensities by the assay software (Multicalc; Wallac).

The protocol yielded assay sensitivities of 6 pg/ml (IL-8) and 12 pg/ml (IL-6). An acceptable precision was achieved with both interrun and intrarun variations ranging between 5 and 10% (percent coefficient of variation).

Neutrophils were harvested for chemotaxis in EDTA-Vacutainers (Becton, Dickinson, Franklin Lakes, NJ) at baseline, 240 min, and 480 min, and their ability to perform chemotaxis ex vivo  was determined by an under-agarose assay as described by Kongstad et al.  11Briefly, polymorphonuclear neutrophils were isolated immediately after blood collection by erythrocyte sedimentation in 5% dextran (T-500; Pharmacia, Uppsala, Sweden), centrifuged over Lymphoprep® (Nycomed, Oslo, Norway) and hypotonic lysed and washed. An agarose medium consisting of 1.2% agarose and 0.25% gelatin in RPMI was added to specially constructed chemotaxis chambers with a gelatin-coated microscope slide at the bottom of each chamber. Six series of three wells were cut in each gel, with a well diameter of 2 mm and an interwell distance of 3 mm. Ten microliters suspension with approximately 5 × 10⁵cells was added to the middle well in each series. To the outermost wells, 10 μl RPMI was added as control, and to the opposite outermost wells, 10 μl chemotactic factor, zymosan-activated serum (10 mg/ml), was added. The chambers were immediately sealed, incubated at 37°C for 135 min, and terminated by flooding the agarose with methanol overnight followed by formaldehyde for 60 min. The gel was removed, and the cells were stained. The chemotactic migration of the polymorphonuclear neutrophils was measured as the linear distance they had moved from the margin of the well containing cells toward the chemoattractant well. The spontaneous migration was measured as the linear distance from the well containing cells and the control well. Measurements were performed in duplicate in two different chambers.

All statistical analyses were performed using Intercooled Stata 8.1 (Stata Corporation, College Station, TX).

Data were log transformed (the natural logarithm, base e) when appropriate. If raw data were used in the analysis, the arithmetic mean and the SD were stated for each group at each time point, and if the log-transformed data were used, the geometric mean and the SD on the log scale (an estimate of the coefficient of variation) were stated.

A two-way repeated-measurement analysis (three measurements for each pig: 60, 240, and 570 min) of each variable was performed using lipopolysaccharide (±) and HEC (±) as explaining factors. Generally, this design allowed us to test the interacting effects of lipopolysaccharide and HEC, in both cases with and without dependency of time. The assumptions of the models were verified by inspecting scatter plots of the residuals versus  fitted values and normal quantile plots of the residuals.

The general ANOVA model was as follows:

Comparison of changes over time were first performed. If the interaction among lipopolysaccharide, HEC, and time was statistically significant, we restricted the comparison to pairwise testing of selected groups. If the interaction among lipopolysaccharide, HEC, and time was not statistically significant, we continued to compare the effect of lipopolysaccharide and HEC (with regard to changes over time) independent of each other. If there was no statistical significance between lipopolysaccharide and time (with regard to changes over time), we tested whether there was a general difference (with regard to independent of time) between lipopolysaccharide (±) animals. The same was done for HEC. In addition, we tested whether there was an interaction between lipopolysaccharide and HEC independent of time.

In terms of insulin, glucagon, and FFA, the assumptions of the model were not fulfilled. Consequently, all four groups were compared, with relation to the relative rate of change between 60 and 240 min, by Kruskal-Wallis test, and if a statistical significant difference was found, the groups were compared pairwise with multiple testing.

The same general ANOVA model as above was used. Only the time points changed.

Data regarding TNF-α and IL-10 were tested with t  tests between lipopolysaccharide and lipopolysaccharide–HEC groups at the 240-min time point because these were the only groups with measurable TNF-α and IL-10 signals. Normality was assumed on the basis of normal quantile plots. Data regarding IL-6 and IL-8 were not assumed to be normally distributed on the basis of normal quantile plots. The relative rates of change in cytokine concentration between 240 and 480 min and 240 and 570 min, respectively, were tested with use of the Mann–Whitney two-sample rank sum test.

Data of the effects on cytokine concentrations are shown in figure 2. In the lipopolysaccharide pigs (group 3) and the lipopolysaccharide–HEC pigs (group 4), TNF-α and IL-10 plasma concentrations increased during the lipopolysaccharide infusion and peaked at 240 min after the start of infusion. The TNF-α concentrations in the lipopolysaccharide–HEC group were significantly lower than in the lipopolysaccharide group (P = 0.04). Although there was a trend toward a lesser increase in IL-10 concentrations in the lipopolysaccharide–HEC group at 240 min, there was no significant difference between the two groups. In the anesthesia group (group 1) and the HEC group (group 2), TNF-α and IL-10 were not detectable in plasma.

Interleukin 6 and IL-8 were detectable in plasma only after lipopolysaccharide infusion (groups 3 and 4). The peak concentrations of both cytokines were observed 240 min after the start of lipopolysaccharide infusion. The decrease in plasma IL-6 from 240 to 480 min after lipopolysaccharide infusion was significantly higher in lipopolysaccharide–HEC pigs (group 4) compared with lipopolysaccharide pigs (group 3) (P = 0.04). There was no statistically difference in plasma IL-8 at any time between lipopolysaccharide pigs and lipopolysaccharide–HEC pigs.

Circulating glucose, insulin, FFA, and glucagon data are shown in figure 3. Plasma glucose was comparable in all four groups. We found a statistically significant difference between the four groups with relation to relative change from 60 to 240 min for insulin (P = 0.001), FFA (P = 0.032), and glucagon (P = 0.009). The pairwise comparisons yielded the following results:

1. The HEC group showed a higher serum insulin response compared with the anesthesia group (P < 0.05), and the lipopolysaccharide–HEC group showed a higher serum insulin response than the HEC group (P < 0.05).

2. Serum FFA was higher in the lipopolysaccharide group compared with the HEC group (P < 0.05).

3. In the lipopolysaccharide–HEC group, the glucagon response was diminished compared with the lipopolysaccharide group (P < 0.05), and the lipopolysaccharide group showed a marked plasma glucagon response compared with the HEC group (P < 0.05).

Base excess and lactate data are shown in figure 4. Lipopolysaccharide infusion resulted in decreased levels in base excess over time (P < 0.001) and increased levels of lactate over time (P = 0.001).

Chemotaxis data are shown in figure 5. Lipopolysaccharide infusion resulted in decreased chemotaxis over time (P = 0.01). Furthermore, there was an effect of the HEC independent of time (P = 0.03).

Hemodynamic parameter data are shown in table 1. Pigs remained hemodynamically stable throughout the study period in the HEC and anesthesia groups. Pigs in the lipopolysaccharide and lipopolysaccharide–HEC groups showed a steep increase in PAP over time (P = 0.00).

Neither lipopolysaccharide infusion nor the combination of HEC and lipopolysaccharide infusion had any major effects on mean blood pressure, pulmonary capillary wedge pressure, or cardiac output. Lipopolysaccharide infusion decreased the ratio of Pao²:Fio²over time (P = 0.00).

The lipopolysaccharide-induced inflammatory condition was treated with increased rate of fluid infusion (saline), increased ventilation (tidal volume and respiratory rate), and increased Fio². In the lipopolysaccharide group, four animals developed arrhythmias/ventricular fibrillation/cardiac arrest. These conditions were successfully treated with defibrillation.

In this experimental study, we demonstrated that short-term hyperinsulinemia, together with normoglycemia, vastly reduced the systemic inflammatory and metabolic responses to endotoxemia in a porcine model. It was clearly demonstrated that maintenance of hyperinsulinemia reduced plasma concentrations of glucagon and TNF-α, the latter supporting the hypothesis that insulin has antiinflammatory effects.

The infused amounts of insulin far exceeded the amounts necessary to maintain normoglycemia. This is a major difference from the clinical situation where normoglycemia usually is established with much lower insulin doses. 3However, the beneficial effects of hyperinsulinemia in sepsis seem to be analogs to the effect of glucose–insulin–potassium solutions used for improvement of cardiac performance in patients with acute myocardial infarction, 12where insulin and glucose are administered in large doses without targeting normoglycemia. The aim of insulin–glucose–potassium is to stimulate myocardial metabolism of glucose instead of fatty acids when oxygen supply is compromised in the ischemic myocardium. Furthermore, the ischemic myocardium produces high amounts of TNF-α. 13This production is suppressed by insulin, 4probably through decreased concentration of nuclear factor κB, which is an important regulator of cytokine-inducible gene expression. 14 

High concentrations of plasma TNF-α and the ratio TNF-α:IL-10 15have previously been reported to correlate with poor outcome in humans with severe sepsis and to link with the presence of insulin resistance. However, TNF-α is only one of an orchestra of proinflammatory cytokines that initially can be found in plasma during sepsis. TNF-α is believed to play an important role as a mediator of multiorgan dysfunction and as an inducer of the secretion of other proinflammatory cytokines, such as IL-1β, IL-6, and IL-8. However, a certain amount of TNF-α seems to be necessary for optimal function of the immune system. Blockade of the TNF-α response to sepsis was associated with reduced survival in animals, just as a clinical trial has shown increased mortality in patients treated with a TNF-α antagonist. 16 

Hyperinsulinemia was also associated with a diminished response of IL-6, another proinflammatory molecule. It has previously been shown that the proinflammatory cytokines TNF-α, 17IL-6, 18and IL-8 increase blood lipids, enhance gluconeogenesis, stimulate the production of catabolic hormones, and decrease insulin sensitivity as part of the initial inflammatory stress response. The modified proinflammatory cytokine response may thus have contributed to the diminished glucagon and FFA response.

In the current study, there was a already marked decrease in plasma FFA concentrations in the stabilization period, which illustrates how easily the animals were stressed although they were heavily premedicated before induction of anesthesia and intubation. The FFA concentrations in the animals exposed to anesthesia alone were rather stable throughout the study. Animals receiving lipopolysaccharide had plasma FFA concentrations similar to those of pigs in the anesthetic group. One would have expected that circulating FFAs would be higher during lipopolysaccharide exposure than anesthesia per se ; however, tiny and transient fluctuations in serum insulin may have obscured this finding because of the high efficacy of insulin on antilipolysis. 19However, the key issue is that hyperinsulinemia in combination with lipopolysaccharide administration is characterized by suppressed FFA concentrations. In large amounts, FFAs contribute to metabolic acidosis and cellular and organ dysfunction. This goes along with the significant reduction in base excess observed in pigs receiving lipopolysaccharide in this study and the impaired neutrophils chemotaxis. It remains to be clarified whether lower FFA concentrations per se  contribute to the beneficial effect of insulin treatment in critically ill surgical patients.

Free fatty acids are toxic for the ischemic myocardium and may lead to damaged cardiac cell membranes, calcium overload, and arrhythmias, particularly in the case of insulin deficiency. 20The increase in lactate concentration and the subsequent metabolic acidosis are also advocated to explain the deleterious effect of hyperglycemia in cerebral tissue. Moreover, the ability of insulin to reduce the catabolic glucagon concentrations may also be part of the antiinflammatory effect of insulin because plasma concentrations of glucagon are normally enhanced during sepsis, 21although the effect of glucagon itself on the antiinflammatory response is not well elucidated.

Through the observation period, parallel impairments in pulmonary and hemodynamic function were observed in pigs receiving lipopolysaccharide. The infusion of lipopolysaccharide had a major effect on PAP, manifested as an abrupt increase in PAP occurring 20–25 min after the start of the infusion. Within 5–10 min, the mean PAP increased from a range of 15–20 mmHg to a range of 45–50 mmHg, with peak levels during the period of infusion of lipopolysaccharide at an increasing rate. The PAP remained elevated in the lipopolysaccharide and lipopolysaccharide–HEC groups, although at a lower level throughout the study period. The increase in PAP was accompanied by discrete increases in partial pressure of carbon dioxide (Pco²) and plasma lactate, whereas Pao²showed a decreasing tendency. Insulin did not, however, have a significantly beneficial effect on the respiratory and cardiovascular systems in the current study.

The function of neutrophils is impaired in patients with diabetes, proportionally according to the degree of hyperglycemia. 22,23In the current short-term study, the hyperinsulinemia could not restore the impaired chemotactic response induced by endotoxin. Neutrophils play a key role in the innate immune system. Circulating neutrophils are activated by proinflammatory cytokines. 24When activated, they transmigrate from the circulation to the tissue/organs (chemotaxis). Neutrophil infiltration with release of lysosomal enzymes and free oxygen radicals may cause tissue damage. The proinflammatory cytokine IL-8 plays a pivotal role for recruitment and activation of neutrophils. There was no difference in plasma IL-8 between the groups receiving lipopolysaccharide. However, this does not rule out any difference at organ level. 25 

Endotoxin (lipopolysaccharide) is a membrane component of gram-negative bacteria and produces multiple endocrine and metabolic effects that mimic those seen in acute gram-negative sepsis. The animals received an infusion of lipopolysaccharide with gradual increase and a total infusion time of 180 min. This does not, however, mimic the clinical situation but allows us to monitor the acute inflammatory response to endotoxin. In contrast to most clinical courses of severe sepsis, the injury was well defined with respect to time and lipopolysaccharide dose and was controlled in this setup.

It seems that the physiologic response to acute endotoxemia in pigs is analogs to the systemic inflammatory response in humans and is composed by a transitory proinflammatory and antiinflammatory cytokine response and impaired neutrophil function. The exogenous insulin acutely modulates the innate immune system by decreasing an inappropriate proinflammatory response, restoring normal glucose concentrations and decreasing FFAs and glucagons, playing an antiinflammatory role. Longer term effects of hyperinsulinemia on the proinflammatory cytokine response remain to be determined.

The authors thank Lene Vestergaard (Technical Assistant, Department of Anesthesiology and Intensive Care, Aarhus University Hospital, Denmark), Henrik Sørensen (Technical Assistant, Institute of Experimental Clinical Research, Aarhus University Hospital), and Rolf Dall, M.D. (Research Fellow, Department of Anesthesiology and Intensive Care, Institute of Experimental Clinical Research, Aarhus University), for their skilled technical assistance.