Introduction

Severity of sepsis and the risk of death associated with it may influence the efficacy of anti-inflammatory therapies in sepsis [1, 2]. Prior analysis of preclinical and clinical trials suggested that mediator-specific anti-inflammatory agents, while beneficial when the risk of death was high, were less beneficial or had effects that were harmful when the risk was low [3, 4]. Consistent with this, based on the results of a phase III trial, recombinant human activated protein C (rhAPC), which has anti-inflammatory effects, was recommended for use in patients with a high, but not in those with a low risk of death [5].

Meta-analyses of clinical trials have suggested that low-dose corticosteroids that have anti-inflammatory as well as other actions, improve hemodynamic function and survival in some septic patients [6]. While analysis of these trials did not demonstrate a relationship between risk of death and the effects of corticosteroids, this treatment was employed selectively in patients with persistent vasopressor-dependent shock who generally have a relatively high mortality [79]. However, in a recently published study patients with community-acquired pneumonia who demonstrated evidence of sepsis, but in whom overall mortality was low, also appeared to benefit from low-dose corticosteroids [10]. On the other hand, in findings presented at recent meetings from the Corticosteroid Therapy of Septic Shock trial (CORTICUS), corticosteroids, although decreasing the time to shock reversal, did not improve survival [11]. Control mortality rate in this trial was 31% and lower than in most earlier studies. Since corticosteroids are now widely recommended for use in septic patients, further defining whether risk of death alters their effects is important.

Pneumonia is a common cause of sepsis in critically ill patients today [12]. In order to study the potential influence of risk of death on the effects of corticosteroids, we first developed a model of E. coli pneumonia in C57BL/6J mice in which a regimen of antibiotic and fluid treatment was shown to increase survival. While not the most common etiologic agent, E. coli is recognized to be a cause of severe community- and ventilator-acquired pneumonia, resulting in sepsis [13, 14]. Then, we asked whether a regimen of hydrocortisone first demonstrated to be beneficial when the risk of death was high would also be beneficial as this risk was reduced. Since it was unclear which hydrocortisone dose might be beneficial, three different ones were investigated, one high one similar to that in reports from the literature, and two lower ones [1518].

Materials and methods

Animal care

The protocol used in this study was approved by the Animal Care and Use Committee of the Clinical Center of the National Institutes of Health. Animals were maintained four to a cage in rooms with light cycling to simulate day- and night-time conditions. Animals were observed as outlined below.

Study design

In initial experiments testing the effects of antibiotics and fluids, male C57BL/6J mice (n = 111) were briefly anesthetized with isoflurane and challenged with intratracheal E. coli in doses producing low or high lethalities (0.625 to 7.5 × 109 CFU/kg). Bacteria preparation is outlined in the Electronic Supplementary Material. Animals were then treated with daily antibiotics starting at 4 h after challenge (ceftriaxone, 100 mg/kg in 0.1 ml; SC, Roche Laboratories, Nutley, NJ, USA) for 4 days and fluids (normal saline, 0.5 ml, SC) for 1 day or neither (control).

In subsequent experiments testing the influence of risk of death on hydrocortisone, male C57BL/6J mice were challenged with intratracheal E. coli and then treated with the regimen of antibiotics and fluids tested above. The doses of E. coli employed varied from high ones in earlier experiments (15 × 109 CFU/kg) to medium (5 × 109 CFU/kg) and then low (2.5 × 109 CFU/kg) doses in later experiments (Table E1 in the Electronic Supplementary Material). Following E. coli inoculation mice (n = 476) were randomized to be treated with one of three doses of hydrocortisone (5, 25 or 125 mg/kg) in 0.1 ml of normal saline (NS), or NS only (control) i. p., bid for 2 days and then qd for 2 days, starting either immediately or 12 h after bacteria inoculation. The volume of fluid each animal received was similar. The hydrocortisone regimens chosen for study were based on a review of the literature (Table 1) [1518]. The full rationale for this is outlined in the Electronic Supplementary Material (ESM).

Table 1 Summary of the steroid regimen (type, initial treatment time, dose, and frequency) and effect on survival in published mouse studies with E. coli and two other enteric gram-negative bacterial intraperitoneal challenges

Following inoculation in the present study, animals were observed every 2 h for the initial 24 h, every 4 h for the second 24 h and then qd for up to 168 h. Animals alive at 168 h were considered survivors. In another set of experiments testing the effects of hydrocortisone on laboratory measures, mice (n = 161) were challenged with nothing or intratracheal NS (non-infected control) or E. coli 15 × 109 CFU/kg. Animals receiving intratracheal NS or E. coli were also randomized to be treated with hydrocortisone (125 mg/kg) or NS (control) started immediately after inoculation along with antibiotics and fluid as above. Animals were randomly selected at 24 and 48 h for circulating blood cell counts, plasma nitrate/nitrite and cytokine levels, quantitative blood counts and lung lavage with cell counts, protein levels, and quantitative bacteria counts as outlined in the ESM.

Bacteria inoculation

In brief, E. coli 0111:B4 was stored in 1-ml aliquots of bactopeptone broth (Difico, Detroit, MI, USA) and glycerol at –70 °C [19]. Preparation and inoculation of the bacteria is outlined in the ESM.

Study agents

A stock concentration of hydrocortisone sodium succinate (100 mg ACT-O-VIAL® System; Pharmacia & Upjohn, Kalamazoo, MI, USA) was dissolved in 2 ml diluent daily and then further diluted in normal saline to the necessary concentrations to be administered as described in the study design.

Statistics

A SAS 9.1 program was used for analysis in this study. To minimize animal usage, where statistical analysis justified averaging over hydrocortisone or E. coli doses, data were pooled to increase the sensitivity of analysis, as has been described previously [2022]. A Wilcoxon test was used to compare survival curves of hydrocortisone and placebo. Cox proportional hazards model or the χ2 test were used to analyze and compare the effects of differing doses of hydrocortisone on the odds ratio of survival (95% confidence interval) at 168 h with differing doses of E. coli. For all other laboratory parameters, a three-way ANOVA, accounting for challenge (E. coli versus normal saline) and dose of hydrocortisone and measurement time, was used. With E. coli challenge, a two-way ANOVA accounting for hydrocortisone dose and measurement time was used for laboratory parameters. Data were log-transformed where appropriate for analysis. All results were expressed as mean ± SEM and p-values ≤ 0.05 were considered significant. The relationship between control mortality rate and the efficacy of hydrocortisone in individual experiments was analyzed as described previously [3] and in the ESM.

Results

Effects of antibiotics and fluids with E. coli doses producing high or low mortality rates

In animals receiving no treatment, decreasing doses of intratracheal E. coli decreased mortality rates (p ≤ 0.006 for the effect of decreasing bacteria dose on mortality rate; Table 2). Treatment with antibiotics and fluid (see Materials and methods) decreased mortality rates with either high or low E. coli doses. This regimen of antibiotics and fluid was used in all subsequent experiments comparing hydrocortisone with placebo.

Table 2 Total and surviving number of animals challenged intratracheally with differing E. coli doses (× 109 CFU/kg) and treated with either nothing (control) or antibiotics and fluids a

Effect of hydrocortisone on survival with E. coli doses producing high or low mortality rates

In control animals receiving placebo treatment (normal saline), decreasing doses of intratracheal E. coli (15, 5 or 2.5 × 109 CFU/kg) decreased final mortality rates from 94 to 12% (p < 0.0001 for the effect of increasing E. coli dose on mortality rate in animals treated with antibiotics and fluid; Fig. 1, Table E1). Time of initiation of hydrocortisone did not alter the effects of treatment on survival and these data were combined in analysis (not significant for the effects of hydrocortisone started at 0 versus 12 h). All doses of hydrocortisone (5, 25, and 125 mg/kg) resulted in survival curves above those of controls with all doses of E. coli, starting as early as 72 h with some combinations (Fig. 1, Table E1). With a hydrocortisone dose of 25 mg/kg and an E. coli dose of 5 × 109 CFU/kg, the rate of mortality initially appeared faster with treatment than placebo. However, this was not significant based on a Wilcoxon test nor was it observed with higher or lower doses of treatment or bacteria (Fig. 1, Table E1). Importantly, treatment with each of the three doses of hydrocortisone resulted in an odds ratio of survival on the side of benefit with each of the three doses of E. coli (Fig. 2). These effects comparing the three doses of treatment and bacteria challenge were similar (p = 0.85, Breslow–Day test) and highly consistent (I2 = 0% [95% CI, 0%, 9.7%], p = 0.92; Fig. 2). When averaged across treatment and E. coli doses, hydrocortisone increased the odds ratio of survival in a highly significant pattern (2.04 [1.37, 3.03], p = 0.0004).

Fig. 1
figure 1

Proportion of animals surviving following challenge with one of three doses of intratracheal E. coli (15, 5, or 2.5 × 109 CFU/kg) and treatment with one of three doses of hydrocortisone (5, 25, or 125 mg/kg administered as described in Materials and methods) or normal saline (control)

Fig. 2
figure 2

The mean effect of three doses of hydrocortisone (5, 25 or 125 mg/kg administered as described in Materials and methods) on the odds ratio of survival (95% CI) with each of the three doses of intratracheal E. coli (15, 5, or 2.5 × 109 CFU/kg) and averaged over the E. coli doses

The consistent benefit of hydrocortisone across a wide range of risks of death can also be demonstrated with the results from individual experiments (Fig. 3). In the majority of these, whether the control mortality rates were high or low, hydrocortisone treatment was beneficial and increased the odds ratio of survival. As a result, the weighted regression line for this relationship remained above the no effect line and had a slope that was not significant (p = 0.29; Fig. 3). If one analyzes the effects of hydrocortisone on the odds ratio of survival in experiments with control mortality rates > 50% (higher risk, mean control mortality rate 82 ± 2%) and control mortality rates ≤ 50% (lower risk, mean control mortality rate 28 ± 3%, p < 0.0001 for high vs. low control mortality rates), both effects are individually significant (2.14 [1.09, 4.21], p = 0.03, and 2.25 [1.06, 4.76], p = 0.03 respectively), but not different when compared (p = 0.92).

Fig. 3
figure 3

The relationship between the control odds of dying (i. e., control mortality rate) and the odds ratio of survival from experiments in mice testing three different doses of hydrocortisone (5 mg/kg, circles; 25 mg/kg, squares; or 125 mg/kg, diamonds). Each symbol represents an individual experiment comparing hydrocortisone with placebo in 10–16 animals challenged with high, medium or low doses of E. coli. The relationships between odds ratio and control odds were similar among the three hydrocortisone doses and the overall relationship for all doses is shown by the weighted regression line

Effect of hydrocortisone on blood and lung measures with E. coli challenge

Compared with no challenge, animals receiving intratracheal NS did not have significant differences in any blood measure at any time point (not significant for all, data not shown). In blood, compared with non-infected animals receiving intratracheal normal saline, E. coli challenge resulted in significant reductions in total white blood cells, lymphocytes (percentage and concentration), platelets, and IL-4 levels, and significant increases in neutrophils (percentage), monocytes (percentage), IL-6 and -10, TNFα, JE, MIP1 and 2α, RANTES, GM-CSF, and nitric oxide levels (p ≤ 0.05 for the effect of E. coli at 24 or 48 h or both; Table 3). In blood in infected animals, compared with placebo, hydrocortisone treatment significantly decreased lymphocyte percentage and IL-6, INFγ, and NO levels, and increased RANTES levels at 48 h (p ≤ 0.05 for each; Fig. 4). In trends approaching significance hydrocortisone decreased nitric oxide levels at 24 h and the concentration of lymphocytes, TNF and MIP2 at 48 h (p = 0.08 to 0.14; Fig. 4). In infected animals, compared with placebo, hydrocortisone did not alter any lung lavage parameters or blood or lung bacteria counts (not significant for all, data not shown).

Fig. 4
figure 4

Mean (± SEM) a percentage of circulating lymphocytes, b IL-6, c INFγ, and d NO levels 24 and 48 h after intratracheal normal saline or E. coli challenge (15 × 109 CFU/kg) among animals treated with hydrocortisone (125 mg/kg) or placebo. Compared with animals receiving intratracheal normal saline, E. coli challenge resulted in a decreased percentage of circulating lymphocytes and increases in IL-6, and NO at both time points and increases in INFγ at 24 h (p ≤ 0.05 for all). The p-values in the figure are for the effects of hydrocortisone versus placebo in E. coli-challenged animals

Table 3 Mean (± SEM) circulating platelet counts (PLT, × 106/mm3), lymphocyte percentage (PLYM, %) and counts (CLYM, × 103/mm3), plasma cytokines [log(pg/ml)] and nitric oxide (NO, μM) levels at 24 and 48 h after intratracheal E. coli challenge in animals treated with hydrocortisone or placebo or non-infected animals a

Discussion

In this study, hydrocortisone treatment added to the beneficial effects of antibiotics and fluids in mice challenged with doses of intratracheal E. coli, producing both high and low control mortality rates. When averaged over dose of treatment and E. coli challenge, the beneficial effects of hydrocortisone were highly significant. Consistent with its known anti-inflammatory effects, hydrocortisone reduced vascular lymphocytes, inflammatory cytokines (IL-6, IFNγ, TNF, and MIP2), and NO levels significantly or with values approaching significance. These effects of hydrocortisone in the present study differ from other agents with anti-inflammatory actions that were noted either in preclinical or clinical studies to be less beneficial or even harmful as the risk of death due to sepsis decreased [35]. The relationship between the effect of hydrocortisone on the odds ratio of survival and the control odds of dying in these experiments, while not significant, was also different from those noted in similarly designed experiments in rats testing either inhibitors of TNF (i. e., P-80 TNF soluble receptor and tyrphostin AG-556) or superoxide (i. e., superoxide dismutase mimetic M40401) [34].

It has been proposed that modulation of the host mediator response during sepsis may only be effective when this response is excessive and associated with a high risk of death [1, 3]. Conversely, when this response is appropriate and associated with low risk, such agents may only interrupt protective host defense mechanisms or produce other deleterious effects that worsen outcome. In this context, there are several reasons why doses of corticosteroids that were beneficial with a high risk of death, may still have improved outcome when the risk was low. Clinically, protective doses of corticosteroids both inhibit inflammatory mediator release and improve hemodynamic function, two actions that are likely to be beneficial with severe sepsis and a high risk of death [710, 2327]. With less severe sepsis, however, the potential deleterious immunosuppressive effects of corticosteroids may be offset by persistent improvements in cardiovascular function. The benefit of providing hemodynamic support (fluid support) during sepsis, independent of risk of death, is well accepted clinically and has been documented in animal studies [3, 28]. In prior published randomized clinical studies with relatively low placebo mortality rates (i. e., 30%) low-dose corticosteroids reversed or decreased the incidence of sepsis-associated shock [10, 27]. In another study employing a cross-over design in septic patients with an overall mortality rate of 30%, corticosteroids, while inhibiting inflammatory mediator release, also significantly increased blood pressure and reduced the need for vasopressor support [26]. In rodent sepsis models, steroid treatment has also been shown to improve hemodynamic function [29]. Several different mechanisms provide a basis for such improvements including among others augmented catecholamine receptor function and reduced nitric oxide production [26, 29, 30]. Consistent with this latter mechanism, in the present study in mice, hydrocortisone decreased circulating nitric oxide levels.

It is also possible, however, that the anti-inflammatory effects of corticosteroids at the doses examined here in this mouse model did not substantially interfere with essential host defense functions. Although hydrocortisone reduced cytokine levels with E. coli challenge, blood and lung bacteria counts were not increased. In several clinical trials, in contrast to much higher doses of corticosteroids studied clinically in the past, the lower doses studied more recently have not been reported to increase the risk of infection [6]. In one study low-dose corticosteroid treatment, while inhibiting IL-6, IL-8, and other mediators, was not associated with evidence of monocyte or granulocyte dysfunction or impaired bacterial clearance [26].

The findings in this mouse model raise the possibility that differences in control mortality rate may not be the basis for the differing effects corticosteroids had, comparing the recent CORTICUS trial with earlier trials [6, 11, 31]. Consistent with these prior trials, in the CORTICUS trial treatment did have beneficial hemodynamic effects. Corticosteroids significantly decreased the time to the reversal of shock (i. e., p < 0.003) and resulted in a trend toward an increase in the percentage of patients with shock reversal (i. e., p = 0.17). Thus, whether other differences between patients enrolled in CORTICUS and those of earlier trials negated the potential beneficial survival effects of improved hemodynamics with corticosteroids will have to be examined. For example, there was an increased trend (p = 0.10) toward superinfections in patients receiving corticosteroids in the CORTICUS trial that was not evident in prior trials testing similar treatment doses. Factors leading to a risk such as the sight or type of the original infection causing sepsis or the presence of underlying conditions (e. g., diabetes mellitus, alcohol disease) should be considered.

In contrast to a prior analysis of clinical trials testing either high- or low-dose corticosteroids, dose of treatment did not influence the beneficial effects of hydrocortisone in this mouse model [6]. However, based on body surface area, even the largest dose of hydrocortisone employed here in mice was not greater than doses that have been used beneficially clinically [32]. It has also been demonstrated that the rate of metabolism of corticosteroids in mice is higher than in humans [33]. Based on this last point, it is possible that more frequent dosing would have augmented the effects of hydrocortisone in this model. Early mortality in the present study did appear greater with a hydrocortisone dose of 25 mg/kg in animals challenged with an intermediate dose of E. coli. However, this initial pattern of mortality was not observed with any other combination of treatment (higher or lower) and bacterial challenge (higher or lower) and likely represents variability in the model. Importantly, consistent with all other groups, by 168 h, mortality was lower with treatment.

It must be noted that corticosterone plays a more central role in the endogenous steroid response in the mouse than hydrocortisone. Furthermore, hydrocortisone may have greater immunosuppressive effects in this species than corticosterone [34]. Therefore, it is possible that administration of corticosterone in these studies may have had different effects from hydrocortisone, especially with regard to cytokine and NO release. It should also be noted that doses of corticosteroid treatment that were beneficial in the present E. coli model and in published mouse studies with gram-negative enteric peritonitis have been immunosuppressive in mice challenged with Listeria monocytogenes [35]. As facultative intracellular bacteria, the clearance of L. monocytogenes may differ from E. coli. Furthermore, even with other enteric bacteria, hydrocortisone was less beneficial or lacked benefit in the absence of antibiotic treatment (Table 1). Thus, type of bacteria and the absence or presence of antibiotics may alter the effects of corticosteroid treatment.

In conclusion, despite its potential anti-inflammatory effects, hydrocortisone added to the beneficial effects of antibiotics and fluids in this mouse model of pneumonia with E. coli doses produced both high and low control mortality rates. These findings raise the possibility that regimens of corticosteroids that are beneficial in septic patients with a high risk of death may be applicable in those with a lower risk as well. Consistent with this, “physiologic” dose corticosteroids have had beneficial hemodynamic effects in trials in which control mortality rates have been both higher and lower. While this is promising, in light of the lack of a consistent beneficial survival effect with corticosteroid treatment clinically, further investigation is necessary to clearly define the role of this treatment in sepsis.