INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Despite advances during the last three decades in antimicrobial therapies and in surgical and diagnostic procedures, mortality rates from sepsis and septic shock have remained high (35 to 40%) (1, 2). Moreover, the incidence of reported cases of sepsis and septic shock in hospitals has been dramatically increasing. Clinical trials of antiinflammatory agents have shown these therapies to produce minimal, if any, benefit (tumor necrosis factor antibodies, interleukin-1 receptor antagonists, platelet activity factor antagonist, and antiprostaglandins), or to be harmful to patients with septic shock (high dose glucocorticoids and soluble TNF receptors) (2). New therapeutic agents and approaches are needed.

One such agent is recombinant granulocyte colony-stimulating factor (G-CSF), a potent enhancer of circulating neutrophil number and function (3). In non-neutropenic animal models of bacterial pneumonia, G-CSF has been shown to improve survival rates (4). This has led to clinical trials of G-CSF in non-neutropenic patients at risk for or with bacterial pneumonia. The hope in these trials has been that G-CSF would stimulate host defenses and improve microbial clearance, thereby decreasing inflammatory injury and improving survival rates.

However, in non-neutropenic patients with community- acquired pneumonia and hospitalized patients with pneumonia, G-CSF treatment did not produce convincing benefit (7, 8). Further, in patients with head trauma receiving mechanical ventilation, G-CSF prophylaxis did not improve outcome or lower the risk for pneumonia (9). The reason for the disparate results between initial human clinical trials and animal models is unknown. Animal models by design, however, are controlled and commonly use only a single bacterial isolate to produce pneumonia, whereas patients in a clinical trial with pneumonia, vary with respect to the type of infecting bacteria. Therefore, if the effects of G-CSF were to differ depending on the type of infecting organism, this might make it more difficult to show benefit in a population of patients with different types of pneumonia than in an animal model using a single bacterial isolate. To evaluate the influence of bacteria type on the effects of G-CSF with pneumonia, we pretreated non-neutropenic rats with G-CSF prior to intrabronchial inoculation with Escherichia coli or Staphylococcus aureus and measured neutrophil and blood bacteria counts, cytokine levels, and survival rates.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Study Design

Male Sprague-Dawley rats weighing 200 to 250 g were randomized to receive G-CSF (100 µg/kg) or placebo subcutaneously daily for 6 d. Immediately before the third dose, animals were further randomized to receive intrabronchial inoculation with either S. aureus or E. coli. Cefotiam was given 4 h after bacterial inoculation and continued twice daily (100 mg/kg/d total) for 4 d. In some animals (n = 87, 21 or 22 animals per group), blood was obtained serially and survival was recorded at 4, 6, 12, 24, 36, 72, and 96 h (survivors) after bacterial inoculation. Other animals (n = 70, 17 or 18 animals per group) were killed at 6 or 24 h for blood and lung cultures (6-h only) and lung lavage.

Serial Blood Analysis

Animals before G-CSF and immediately before bacterial inoculation, and 4, 6, and 24 h after, received 1.5% isoflurane and 100% oxygen for 3 to 5 min via a face-shoulder mask. Tail artery blood was collected during each brief anesthetic in heparinized syringes for complete blood count (CBC) and arterial blood gas analysis (ABG). Blood was collected only at 4 h for tumor necrosis factor levels (Rat TNF-ELISA; Biosource International, Camarillo, CA). Total blood leukocytes were counted using a hemocytometer and a differential was determined on a blood smear using a Wright stain. Alveolar-to-arterial oxygen difference was calculated using standard formulas (4).

Blood and Lung Culture and Lung Lavage at 6 or 24 h

Selected animals received 1.5% isoflurane and 100% oxygen via mask at 6 or 24 h. After anesthesia, using sterile technique, 5 ml of blood were obtained via cardiac puncture and inoculated in blood culture tubes (Liqod, TSB; Roche, Grenzach-Wyhlen, Germany). The animals were then killed by cervical dislocation. Using sterile technique, the right lower lobe of the lung was removed and placed in 50 ml of brain heart infusion broth (BHI; Gibco, Paisley, UK). The remaining lung was lavaged with 9 ml PBS in 3-ml aliquots. The lavage fluid was then collected and centrifuged, and the cell pellet was resuspended in PBS. The number of cells in the PBS were counted with a hemocytometer, and a cell differential was determined on a smear using a Wright stain. Inoculated blood and lung tissue with bacterial growth at 24 h were further analyzed to determine bacterial type.

Bacterial Inoculation

Frozen aliquots of either E. coli or S. aureus were thawed and inoculated into 250 ml of brain heart infusion (Gibco). After incubation for 5 h, suspensions were centrifuged at 4° C, washed twice in PBS, and resuspended. The concentration of bacteria was determined turbidometrically. The suspension was then adjusted with PBS to produce a concentration of 4 × 10⁹ CFU/ml of E. coli or 8 × 10⁸ CFU/ml of S. aureus. The concentrations were confirmed by plating serial dilutions on the appropriate culture medium and counting colonies. Animals were then given ketamine intramuscularly (100 mg/kg, Ketanest; Parke-Davis, Berlin, Germany) and xylazine (8 mg/kg, Rompun; Bayer, Leverkusen, Germany) followed by 100% oxygen for 2 min via face-shoulder mask. The trachea was then visualized with a size 0 laryngoscope. An 18G arterial catheter was inserted into the trachea just beyond the carina and 0.5 ml of bacteria (8 × 10⁹ CFU/kg body weight of E. coli and 1.6 × 10⁹ CFU/kg body weight of S. aureus) was instilled and the catheter was withdrawn. The animals were then placed in a 100% oxygen chamber for 5 min and returned to their cages.

Reagents

Recombinant human G-CSF (Neupogen 30; AMGEN, Munich, Germany) was prepared as previously described (4). Preliminary microbiologic testing showed that E. coli and S. aureus were sensitive to cefotiam (Spizef; Takeda Pharma, Aachen, Germany), the antibiotic used in the study.

Animal Care and Use

The experimental protocol for this study was approved by the Animal Care Committee of the city of Freiburg, Germany. A maximum of three animals was maintained in a cage. Animals had access to food and water throughout the study.

Statistics

Survival was assessed using a Cox regression model. We tested whether the two treatment groups (G-CSF versus placebo-treated) produced different outcomes in the two pneumonia groups (E. coli and S. aureus). In addition, the effects of treatment groups for each pneumonia type was also calculated. Intragroup and intergroup differences for all parameters were assessed at each time point for all four groups using a two-way ANOVA. A p value less than 0.05 was considered significant. Blood culture data were analyzed with a Breslow data test of difference between E. coli and S. aureus.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Clinical Manifestations and Survival

At 6 h after intrabronchial inoculation with S. aureus and E. coli, animals appeared weak and lethargic. Without G-CSF, S. aureus and E. coli challenges produced similar mortality rates (p = NS) (Figure 1). With G-CSF, the effect of S. aureus and E. coli on mortality rates was significantly different (p = 0.001) and opposite. G-CSF with S. aureus pneumonia improved survival rates (relative risk: 0.16; 95% confidence interval: 0.04 to 0.73), whereas G-CSF with E. coli pneumonia decreased survival rates (relative risk: 18.17; 95% confidence interval: 7.5 to 44).

View larger version (12K): [in this window] [in a new window]   Figure 1.   Proportion of animals surviving after inoculation with E. coli (A) or S. aureus (B) pneumonia.

Blood and Lung Analysis

After three doses of G-CSF (immediately prior to bacterial inoculation) in all animals alike, there were significant increases in circulating neutrophil counts (Figure 2) and total white blood cell counts (data not shown) compared with placebo-treated control animals (both, p = 0.001). Without G-CSF from 0 to 6 h after S. aureus and E. coli challenges, there were, in all animals alike, decreases in circulating neutrophil counts. With G-CSF, the reduction in circulating neutrophil counts from 0 to 6 h were significantly greater with E. coli than with S. aureus pneumonia (G-CSF effect on the decrease in CNC, E. coli versus S. aureus, p = 0.001). Importantly, circulating neutrophil counts with G-CSF during S. aureus pneumonia remained increased at 6 h compared with placebo. In contrast, with G-CSF in E. coli-challenged animals, circulating neutrophil counts at 6 h were decreased compared with placebo.

View larger version (22K): [in this window] [in a new window]   Figure 2.   Serial mean (± SEM) changes in circulating neutrophil numbers in control E. coli and S. aureus animals (solid lines with circles and squares, respectively) and G-CSF E. coli and S. aureus animals (dashed lines with circles and squares, respectively).

With G-CSF after S. aureus, the pneumonia-associated increases in alveolar-to-arterial oxygen difference were unchanged, whereas with G-CSF after E. coli pneumonia, the alveolar-to-arterial oxygen difference was increased (G-CSF effect with E. coli versus S. aureus, p = 0.03). With G-CSF at 6 and 24 h, after S. aureus challenges the percentage of positive blood cultures was decreased compared with placebo-treated control animals (9 and 24%, respectively), whereas with G-CSF at 6 and 24 h after E. coli challenges, the percentage of positive blood cultures was increased compared with placebo-treated control animals (31 and 15%, respectively) (G-CSF effect with E. coli versus S. aureus, p = 0.05).

With and without G-CSF at 4 h, serum TNF levels (mean ± SEM, pg/ml) were increased (p = 0.01) with E. coli pneumonia (259 ± 104, G-CSF and 279 ± 98, placebo), compared with S. aureus pneumonia (51 ± 17, G-CSF, and 48 ± 20, placebo). Lung cultures were positive for the infecting bacteria in all animals tested 6 h after inoculation. With and without G-CSF, there were at 6 h similar increases (p = 0.01) in lung lavage neutrophils in E. coli- and S. aureus-treated animals. With or without G-CSF, there were no other significant differences in any other parameter measured as outlined in the methods (p = NS) between E. coli- and S. aureus-treated animals.

Additional Studies

To assess if differences in the number of bacteria or site of infection were factors explaining the differences in the effect of G-CSF on survival rates, a less virulent S. aureus was studied, and E. coli challenges were given intraperitoneally instead of intrabronchially. The control mortality in these studies was similar for the two challenges (38 versus 37%, p = NS). However, as in our studies above, the effects of G-CSF pretreatment on survival were significantly different and opposite comparing intrabronchial inoculation with a less virulent S. aureus (60 × 10⁹ CFU/kg body weight) versus intraperitoneal E. coli (10 × 10⁹ CFU/kg body weight) inoculation (relative risk and 95% confidence intervals: 0.47, 0.20 to 1.10 versus 1.76, 0.69 to 4.53, p = 0.04).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

In our study, prophylactic treatment with G-CSF had profoundly different and opposite effects in rats with E. coli versus S. aureus pneumonia. E. coli pneumonia was associated with high TNF levels. In this setting, G-CSF was associated with marked reductions in circulating neutrophil numbers to levels that were paradoxically lower than controls. This relative neutropenia was associated with decreased bacterial clearance, increased lung injury, and worsened survival rates. S. aureus pneumonia was associated with low TNF levels. Circulating neutrophil numbers with G-CSF in this setting were significantly increased after infection. These increases in neutrophil numbers were associated with augmented bacterial clearance and improvements in lung injury and survival rates.

The beneficial effects of G-CSF in both animal models and patients with infection are believed to be at least in part due to increases in circulating neutrophil number and function (3). In our study, prophylactic G-CSF's detrimental effects on survival with E. coli and beneficial effects with S. aureus were correlated to levels of circulating neutrophils. Several mechanisms related either to augmented host defense or to increased inflammatory tissue injury might provide a basis for such an association.

First, marked reductions in circulating neutrophils with G-CSF during E. coli pneumonia may have impaired host defenses and worsened infection. Evidence for this is found in the relative neutropenia and increased rate of positive bacterial cultures in the G-CSF-treated E. coli group. Decreased circulating neutrophil counts such as observed in this study have been associated with reduced bacterial killing and clearance and worsened outcome in other animal models as well as in human infection (10). The ability for G-CSF during S. aureus pneumonia to increase circulating neutrophil counts, and thereby reduce bacteremia, may have decreased inflammatory mediator release and lung injury and improved survival rates.

Second, in association with G-CSF the far greater reduction in circulating neutrophils during E. coli versus S. aureus pneumonia may in large part reflect greater numbers of circulating cells adhering to and degranulating on endothelial surfaces. Such adherence and degranulation by neutrophils results in endothelial injury (13). The lung is especially susceptible to such injury, which is frequently manifested by arterial hypoxemia (16). Consistent with this, G-CSF-treated animals with E. coli pneumonia had significantly greater arterial hypoxemia than did S. aureus animals.

Lastly, it is possible that with E. coli but not with S. aureus inoculation, extravascular injury related to G-CSF-stimulated neutrophils may have disrupted a compartmentalized host defense response in the lung. This lung injury with G-CSF in E. coli animals may have resulted in greater amounts of bacteria and bacterial products entering the bloodstream. Maintenance of this compartmentalization in S. aureus animals, however, may have promoted the positive host defense effects of G-CSF-stimulated neutrophils and improved bacterial clearance in the lung itself.

Several lines of evidence suggest why G-CSF might have had differing effects on circulating neutrophil numbers and injury comparing animals with E. coli versus S. aureus pneumonia. TNF levels are high during E. coli infection, a bacterial organism with endotoxin in its outer cell membrane. In vitro, both TNF and G-CSF have been shown to increase CD14 expression in response to endotoxin, resulting in upregulation of adherence receptors on neutrophils (19). Further, in in vitro studies TNF has also been shown to augment G-CSF-stimulated neutrophil adhesion, activation, and mediator release (20). In vivo, treatment with G-CSF increases reduction in neutrophil number after intravenously administered endotoxin and worsens lung injury and mortality rates after intrabronchial endotoxin or E. coli challenge (25). Additionally, endotoxin has been shown to work synergistically with other activators of neutrophil function to increase neutrophil sequestration and injury in the lung (29). Thus, there is ample data to suggest that G-CSF therapy in the presence of high TNF levels associated with gram-negative infection and endotoxemia may result in increased neutrophil activation, sequestration, and relative neutropenia.

It is possible that use of a cephalosporin antibiotic may have caused endotoxin release from E. coli. This could aggravate the detrimental effects of E. coli but not S. aureus pneumonia. It is not known if G-CSF has any effect, beneficial or harmful, during antibiotic-induced endotoxin release.

During S. aureus infection TNF levels are low and potentially less likely to stimulate neutrophil activation and sequestration. Further, G-CSF does not augment activation of other known mediators and mechanisms associated with the pathogenesis of S. aureus infection such as staphylococcal toxic shock syndrome toxin-1-stimulated T-cell reactivity (30). Further, in vivo treatment with G-CSF enhances circulating neutrophil numbers and survival during superantigen or S. aureus challenges (30, 31).

In other studies, dose and route of bacterial and inflammatory challenges have been shown to alter the effects of treatment with G-CSF (28, 32, 33). To exclude the potential effects of these variables in the present study, we showed that using a large dose of a less virulent S. aureus and altering the route of administration of E. coli did not influence the beneficial and harmful effects of G-CSF with these two organisms. Our study provides direct evidence that type of organism in addition to dose and route of challenge may influence the effects of prophylactic G-CSF.

The harmful effects of prophylactic G-CSF with E. coli in the present study differ from studies with other gram-negative bacteria challenges. G-CSF was beneficial in the setting of intrabronchial P. aeruginosa challenge (34, 35). These differences may also relate in part to differing patterns of microbial toxin and host inflammatory mediator release with these two different bacteria types. However, P. aeruginosa was administered in these other studies to animals that had been previously weakened either with hemorrhage or alcohol administration, conditions that may have blunted the inflammatory effects and amplified the host defense effects of G-CSF. The effects of prophylactic G-CSF with E. coli in the present study also differ from ones we have noted in a canine model of peritonitis and pneumonia (4, 5). In addition to being a larger animal, this canine model permits administration of volume support, an important standard therapy in sepsis. Thus, species and use of differing concurrent therapies may also alter the effects of G-CSF. Finally, the effects of G-CSF with E. coli in the present study differed from its effects in another rat model using an infected intraperitoneal fibrin clot (36). It is possible that the use of such a clot may slow the release of bacteria into the bloodstream and maximize the beneficial effects of G-CSF stimulated neutrophils in the extravascular space.

Our data suggest the type of bacterial infection and associated mediators released may be one of the critical factors that determine the effects of prophylactic G-CSF on circulating neutrophils and outcome during sepsis and septic shock. Whether such factors would alter the effects of G-CSF administered at the onset of infection, rather than prophylactically, requires further study. However, the present findings suggest that G-CSF may not be beneficial as a prophylactic therapy in clinical trials of non-neutropenic patients at risk for pneumonia produced from a variety of bacterial isolates, but still show benefit in certain animal models of pneumonia studying a single bacterial isolate.