After cannulation and the start of the sedative infusions cecal ligation and double intestinal puncture (CLIP) was performed as previously described 3334 under additional pentobarbital anesthesia. The procedure was performed under sterile conditions with the abdominal skin disinfected with 70% alcohol. Laparotomy was conducted through a 2 cm lower-midline incision. The cecum was exposed and ligated immediately distal to the ileocecal valve to avoid intestinal obstruction and then punctured twice with an 18-gauge needle, squeezed gently to force out a small amount of feces, and then returned to the abdominal cavity. The abdomen is closed with 3-0 silk sutures in two layers. Following completion of CLIP the sedative (or saline) infusions were continued without bolus administration.

Venous blood samples (1 ml) were drawn for the measurement of plasma cytokine (TNF-α and IL-6) concentrations at two, four, and six hours after CLIP (n = four to six per group). Double the volume of saline was injected to replace blood lost after each sampling. A total amount of 4 mL of blood was drawn from each animal over eight hours. Samples were centrifuged at 3500 rpm for 10 minutes at 4°C and plasma was collected and stored frozen at -80°C until assaying. IL-6 and TNF-α were measured in duplicate using a commercially available ELISA kit (Biosource, CA, USA). The sensitivities of the assays were 3 pg/ml for IL-6 and 3 pg/ml for TNF-α and 3 pg/ml for IL-6.

At the end of the experimental period (nine hours for western blot experiments) spleens were harvested (n = four per group). The samples of spleen were then homogenized (Polytron homogenizer by Kinematica, Bethlehem, PA, USA) in ice-cooled lysis buffer (20 mm Tris-HCl, 150 mm NaCl, 1 mm Na₂DTA, 1 mm EGTA, 1% Triton, 2.5 mm sodium pyrophosphate, 1 mm β-glycerophosphate, 1 mm Na₃VO₄, 2 mm dl-dithiothreitol, 1 mm phenylmethanesulfonyl, and 1 μg/ml leupeptin; pH 7.5) and centrifuged at 3000 g for 10 minutes at 4°C. The supernatant was further centrifuged twice, initially at 12,000 g for 15 minutes at 4°C and a second time at 20,000 g for 45 minutes at 4°C. The protein concentration of supernatant was determined with the Bradford protein assay (Bio-Rad, Herts, UK). The supernatant (10 μg protein per sample) were denaturated in NuPAGE LDS Sample buffer (Invitrogen, Paisley, UK) at 70°C for 10 minutes and then were loaded on a NuPAGE 4 to 12% Bis-Tris Gel (Invitrogen, Paisley, UK). After electrophoresis, the proteins were electrotransferred to a nitrocellulose membrane (Hybond ECL; Amersham Biosciences, Buckinghamshire, UK) and incubated with a blocking solution composed of 5% fat dry milk in Tween-containing Tris-buffered saline (pH 8.0, 10 mm Tris, 150 mm NaCl, 0.1% Tween). The blocked membrane was incubated overnight at 4°C with the cleaved caspase-3 antibody (New England Biolab, Hitchin, United Kingdom). After washing with Tween-containing Tris-buffered saline for four times, the membrane was incubated for one hour at room temperature with the appropriate horseradish peroxidase-conjugated secondary antibody directed at the primary antibody. The bands were then visualized with enhanced chemiluminescence (New England Biolab, Hitchin, United Kingdom) and exposed onto Hyperfilm ECL film (Amersham Biosciences, Buckinghamshire, United Kingdom). Subsequently, the membrane was re-probed with caspase 3 and beta-action primary antibody respectively and the rest procedures were repeated again as above. The band density was analyzed densitometrically and normalized with the housekeeping protein beta-actin and then presented as percentage of control.

Animals were monitored every two hours via video recording of the animal in its cage following the initial eight-hour sedative infusion period and animal mortality was noted (n = 10 per group). After 16 hours of follow up (i.e., 24 hours post CLIP) all animals were sacrificed by lethal sodium pentobarbital injection.

The results are presented as mean ± standard error of the mean. Statistical analysis was performed by analysis of variance followed by post-hoc Newman Keuls testing using the instat program. Twenty-four hour mortality was analyzed by Chi squared test. A P < 0.05 was set as significant.

The CLIP model employed induced severe sepsis with lethargy and sickness behavior observable in the saline-infused animals. Nine of the 10 animals died within 24 hours (90%) indicating that very severe sepsis was provoked (Figure 1). Sedation with either drug significantly decreased mortality at 24 hours after CLIP compared with saline (P < 0.01; midazolam 30% and dexmedetomidine 20% mortality, respectively). However, no difference was noted between dexmedetomidine and midazolam (P = 0.6).

Both midazolam and dexmedetomidine reduced TNF-α levels compared with saline-treated controls. At two hours the saline group had a significantly higher level (166 ± 37 pg/ml) than either midazolam (51 ± 12 pg/ml) or dexmedetomidine (50 ± 11 pg/ml); this pattern was also present at four hours (saline 130 ± 54 pg/ml; midazolam 55 ± 8 pg/ml; dexmedetomidine 62 ± 39 pg/ml) and five hours (saline 141 ± 30 pg/ml; midazolam 62 ± 20 pg/ml; dexmedetomidine 73 ± 40 pg/ml). Integrated over time revealed an area under the curve of 626 ± 137 in the saline group, 232 ± 40 in the midazolam group, and 244 ± 93 in the dexmedetomidine group. Thus the reduction in mortality effect in the sedative group was associated with a reduction in TNF-α levels in both sedated groups (Figure 2).

In contrast to sedation with midazolam, dexmedetomidine reduced IL-6 levels relative to the saline group (P < 0.05; Figure 3). At two hours the saline (188 ± 37 pg/ml), midazolam (176 ± 40 pg/ml), and dexmedetomidine groups were similar (50 ± 11 pg/ml). At four hours the IL-6 levels in the dexmedetomidine group (181 ± 15 pg/ml) were significantly lower than midazolam (312 ± 39 pg/ml) and saline (282 ± 70 pg/ml) groups. At six hours the IL-6 levels in the dexmedetomidine group (262 ± 38 pg/ml) were again lower than midazolam (371 ± 14 pg/ml) and saline (455 ± 96 pg/ml) groups. The mean area under the curve was 1135 ± 187 in the saline group, 1132 ± 90 in the midazolam group, and 771 ± 100 in the dexmedetomidine group.

At death or at eight hours after CLIP splenic caspase-3 expression was reduced in the dexmedetomidine group relative to both midazolam and controls (P < 0.05; Figure 4) with similar effects on both the 17 and 19 KDa caspase-3 fractions. Interestingly midazolam reduced expression of the 17 KDa but not the 19 KDa fractions relative to saline.

This model of sepsis in healthy rats does not necessarily replicate vulnerable patients with sepsis. Although attempts were made to fluid resuscitate the animals this was in a protocol driven manner and thus was not necessarily analogous to the clinical situation where resuscitation is titrated to patient's needs determined by invasive hemodynamic monitoring. We chose not to administer antibiotics, a departure from clinical practice, because we wanted to observe the consequences of acute polymicrobial sepsis. Our model is analogous with acute sepsis that is severe enough to require provision of sedation for mechanical ventilation and can lead to death within hours in the absence of appropriate management. We chose a limited sedative period as continuous sedation cannot be provided for more than 12 hours in animals according to the institutional license and all animals received further pentobarbital boluses to allow blood sampling in the animals randomized to saline. Although we scaled the dexmedetomidine and midazolam drug doses using established methodology and there were no observable differences in the level of animal sedation, it is possible that the level of sedation did differ between the groups. Future studies looking at electroencephalogram-guided sedation are planned to overcome this caveat to our experiment. We have previously used caspase-3 expression as a marker of apoptosis for which it is well validated 2; however, our approach of using splenic western blotting lacks specificity for vulnerable cell types such as lymphocytes, although the CLIP model does induce apoptosis in these cells. Therefore, apoptosis of other cell types (including endothelial cells and macrophages) may have contributed to the caspase-3 expression. These cells appear to have less relevance to clinical sepsis 23 and thus may have skewed our data.

Previous preclinical studies had shown that sedation with dexmedetomidine does improve mortality from endotoxic shock in rats compared with a non-sedated group 22. Based upon the inflammatory and apoptosis biomarkers we would anticipate superior benefits of sedation with dexmedetomidine vs midazolam in the acute phase of sepsis; possible reasons why this putative benefit was not borne out by the mortality data may relate to the 'hyper-aggressive' septic state that appears primarily to be TNF-α dependent (as mortality benefits were associated with reduced TNF-α levels). It is noteworthy that midazolam and dexmedetomidine reduced TNF-α levels by a similar amount although previous clinical trials have suggested that dexmedetomidine was superior to midazolam in this regard 24. Dexmedetomidine has also been shown to improve mortality and reduce inflammatory cytokine levels induced by CLIP in mice when dexmedetomidine was started prior to the sepsis 23 though the dosing schedule in this study was irregular. In our study the sedatives were commenced by infusion shortly before provoking sepsis and therefore the levels were unlikely to be therapeutic as sepsis was induced.

The anti-inflammatory effects of dexmedetomidine have now been shown against endotoxin (compared with saline) 22, in single CLIP 23, in double CLIP (compared with midazolam; Figures 2, 3) and in critically ill humans (compared with propofol 25 or midazolam 24). How dexmedetomidine induces its anti-inflammatory effect is currently unclear though it may be related to its central sympatholytic effects 2330 and relative stimulation of the cholinergic anti-inflammatory pathway 1617. Inflammation also appears to alter the effects of α₂ adrenoceptor stimulation shifting them from a pro- to an anti-inflammatory effect 35.

The effect of the sedatives on IL-6 require further consideration as IL-6 levels are predictive of mortality in septic humans 36 and animals 37. Therefore, the reduction of IL-6 levels by dexmedetomidine relative to midazolam and saline may prove crucial in future studies. The achieved significance value of P = 0.05 means the results are of borderline significance though we suspect this is due to a reduced sample size in the midazolam group. Power analysis based on our results suggests that six animals per group are required to achieve power to find a statistical difference of P < 0.05. Therefore our study was designed with appropriate power but a loss of two animal samples in the midazolam group, leaving a sample size of four animals in that group, may have been responsible for our result that is of borderline significance. The superiority of dexmedetomidine's ability to reduce IL-6 levels has already been shown in humans 2425; however, it should be noted that dexmedetomidine was administered immediately after the septic insult in this study. This is important as the timing of anti-IL-6 therapy is critical; delays greater four hours after CLIP show no benefit in septic animals 38.

How midazolam induces an anti-inflammatory effect is unclear but immune cells express both the peripheral benzodiazepine receptor 39 and gamma-amino butyric acid receptors 40 and thus at least two local targets exist for benzodiazepines. For example, midazolam suppressed lipopolysaccharide-induced TNF-α activity in macrophages, an effect that was blocked by the peripheral benzodiazepine receptor antagonist PK 11195 39. Midazolam also inhibits lipopolysaccharide-induced up-regulation of cyclooxygenase 2 and inducible nitric oxide synthase in a macrophage cell line. Other markers of immune cell activation (induced by lipopolysaccharide) such as IκB-α degradation, nuclear factor-κB transcriptional activity, phosphorylation of p38 mitogen-activated protein kinase and superoxide production were also suppressed by the midazolam 41.

Interestingly dexmedetomidine and midazolam appear to exert opposite effects on innate immunity. Dexmedetomidine appears to potentiate macrophage function and phagocytosis 272829, while, as described above, midazolam inhibits it 394142. This may be related to opposing effects on p38 mitogen-activated protein kinase signaling in these cells 4143. Thus although both sedatives suppressed circulating cytokines, at a local level the effects on macrophages may have been very different. Benzodiazepine induced suppression of immunity has been noted against Salmonella typhimurium with 15 days of diazepam treatment 19 and Klebsiella pneumoniae with three days of diazepam treatment in vivo 20. In these settings of infection, diazepam treatment increased animal mortality. Thus longer treatment times may be needed to show impairment of immune responses by midazolam than used in this study. We consider that differing effects on innate immunity may explain why critically ill patients sedated with dexmedetomidine experienced fewer infections than those patient sedated with midazolam in a recent randomized controlled trial of 366 critically ill patients 44. Further studies addressing the relative effects of longer dosing schedules and different doses of the two sedatives on innate immune responses are in progress. It is interesting to note that daily interruption of sedative infusions appear to be associated with fewer infective complications 45; this may be related to the reduced dose of sedatives resulting in less inhibition of the immune system. Recently, deep sedation has been associated with increased mortality in the critically ill 46 although it is unclear whether this affected immune responses. In this study we did not measure depth of sedation with electroencephalogram monitoring; however, based on recently published clinical data 46, future studies should consider this. Nonetheless our data suggests that the sedatives are equally able to reduce mortality during the acute phase of sepsis and therefore that choice of sedative in this acute phase may not matter.

Apoptotic (or programmed) cell death occurs in physiological conditions; for example, it is an important mechanism by which immune responses are controlled via activated cell death of lymphocytes. Sepsis induces apoptosis in lymphocytes, dendritic cells and enterocytes and death of these cells appear pivotal to the pathogenesis of the hypo-inflammatory phase of the condition 23. Prevention of this apoptotic injury with inhibitors of the caspase enzymes 47, regarded as the final executioners in apoptosis or of over expression of anti-apoptotic proteins, has been shown to improve survival in animal models of less acute sepsis.23 Critical mediators of this septic apoptotic injury include pro-apoptotic proteins such as BAX and activated caspase-3 23.

Both midazolam and dexmedetomidine reduced the burden of splenic caspase-3 expression indicating that they may exert some anti-apoptotic effects in the presence of severe sepsis. It is possible that in the present model, TNF-α binding stimulated the extrinsic apoptotic cascade. Thus the observed inhibition of apoptotic markers may be, in part, due to suppression of the inflammatory response. This would account for why both sedatives showed some anti-apoptotic ability. Interestingly, midazolam was only capable of reducing the 19 KDa fragment of cleaved caspase-3; why it had such an effect is currently unclear. Nonetheless, dexmedetomidine exhibited significantly superior anti-apoptotic effects, consistent with previous reports demonstrating that dexmedetomidine could prevent apoptotic injury from hypoxia and isoflurane in neurons 2648. α₂ adrenoceptor stimulation reduces pro-apoptotic proteins such as BAX and increases anti-apoptotic Bcl-2 signaling 49, indicating activity against the intrinsic apoptotic cascade. As apoptotic mechanisms are highly conserved and therefore anti-apoptotic agents are likely to work in different tissue types we hypothesized that stimulation of α₂ adrenoceptors by dexmedetomidine may inhibit septic apoptosis. Indeed activation of AKT/protein kinase B, extracellular regulated signalling kinase and Bcl-2 improves survival in sepsis 23 and these effectors are upregulated by dexmedetomidine 4950. Therefore, the reduction in sepsis-induced splenic apoptosis is plausible (Figure 3).

The consequences of apoptosis may be more relevant in clinical sepsis and in the less acute phase of sepsis in animal models. Also, in acute severe sepsis apoptosis of cells may have a protective effect by dampening the immune response; improved mortality has been noted from endotoxic shock in animals treated with apoptotic cells 51. This suggests a complex and dynamic set of circumstances pertain during sepsis expressed in apoptotic and inflammatory responses that are observed at different times. Indeed corticosteroids show anti-inflammatory effects (that have correlated with increased speed of reversal of septic shock in the CORTICUS trial 10) but exacerbate lipopolysaccharide-induced apoptosis 52. However an agent, such as dexmedetomidine, that can combat both inflammation (in the early phase of sepsis) and apoptosis (in the later phase of sepsis) could have particular utility in septic patients. These data also help explain the remarkable mortality benefit we have seen in septic patients from the MENDS study 32. This hypothesis will need evaluation in further preclinical studies.