Protection with Antibody to Tumor Necrosis Factor Differs with Similarly Lethal Escherichia coli  versus Staphylococcus aureus  Pneumonia in Rats

The Animal Care Committee (Weimar, Germany) approved the experimental protocol for this study. Up to four animals were maintained in a cage. Animals had free access to food and water throughout the study.

Male Wistar rats (n = 231) weighing 250–300 g were randomized to receive an intratracheal inoculation with either E. coli  or S. aureus . At the time of bacterial inoculation, animals were further randomized to receive in 0.3-ml volumes of antibody to TNF intratracheally and control serum intraperitoneally, control serum intratracheally and antibody to TNF intraperitoneally, or control serum in both compartments (table 1). Cefotiam (daily dose, 100 mg/kg) (Spizef; Takeda Pharma, Aachen, Germany) was given 4 h after bacterial inoculation and continued twice daily for 4 days. Animals were observed up to 96 h after bacterial inoculation. Either before or after inoculation, some animals were randomly selected to have either lung analysis (bacteria, leukocyte count, or TNF concentration) after death (6, 24, or 96 h after inoculation; 18–20 rats per study group) or serial circulating blood analysis (bacteria, leukocyte count, oxygen concentration, and TNF concentration) (before and 2, 6, 24, and 96 h after inoculation; 20–22 rats per study group).

Frozen aliquots of either E. coli  (American Type Culture Collection 25922) or S. aureus  (American Type Culture Collection 25923) were thawed and inoculated into 250 ml brain–heart broth (Gibco, Paisley, Great Britain). After an incubation period of 5 h, the suspensions were centrifuged at 4°C and washed twice in phosphate-buffered saline. The final suspensions, as estimated by turbidimetry and compared with a predetermined standard curve, were prepared to produce a concentration of 2 × 10⁹colony-forming unit (CFU)/ml E. coli  or 6 × 10⁹CFU/ml S. aureus . The concentrations were confirmed by plating serial dilutions on the appropriate culture medium and counting colonies. Animals were then given ketamine (100 mg/kg) (Ketanest; Parke–Davis, Berlin, Germany) and xylazine (8 mg/kg) (Rompun; Bayer, Leverkusen, Germany) intramuscularly. Following preoxygenation with 100% oxygen for 2 min via  a face–shoulder mask, the trachea was visualized with a modified size 0 laryngoscope. An 18-gauge arterial catheter was inserted into the trachea, and 0.2 ml bacterial suspension (1.6 × 10⁹CFU/kg body weight E. coli  and 4.8 × 10⁹CFU/kg body weight S. aureus ) mixed with 0.3 ml treatment serum (antibody to TNF or control serum) was instilled, and the catheter was withdrawn. The animals were then placed in an oxygen chamber for 5 min and returned to their cages. Microbiologic testing before the study showed that both E. coli  and S. aureus  strains were susceptible to cefotiam, the antibiotic used in the study. The doses of E. coli  and S. aureus  and of antibiotic administered to animals in these studies had been shown previously to produce similar lethality rates. 42 

The antibody to TNF and control sera used in this study were provided by Thomas Hartung, M.D., Ph.D. (Faculty of Biology, University of Konstanz, Germany). The antibody to TNF serum contained a polyclonal antibody raised in sheep to murine TNF-α. The activity was such that 1 ml antibody to TNF serum neutralized 1 μg murine TNF-α. Control serum was obtained from sheep cared for under similar conditions but not immunized.

For lung analysis, tissue bacteria cultures and lavage cell, bacteria, and TNF concentrations were determined. Animals received isoflurane, 1.5%, and 100% oxygen via  a mask at 6, 24, or 96 h. After blood sampling (see below), anesthetized animals were killed by cervical dislocation, and using sterile technique, the right lower lobe of the lung was removed and placed in 5 ml phosphate-buffered saline. The lung lobe was weighed and then homogenized, and 100 μl homogenate was serially diluted in phosphate-buffered saline and then inoculated on agar plates for bacterial counts. The remaining lung was lavaged with 9 ml phosphate-buffered saline in 3-ml aliquots. The lavage fluid was collected. Of this fluid, a portion was used for quantitative bacteria cultures or determinations of TNF concentrations. The remainder was centrifuged, and the cell pellet was resuspended in phosphate-buffered saline. The number of cells in phosphate-buffered saline was counted with a hemocytometer, and a differential cell count was determined on a smear using Wright stain.

For circulating blood analysis, animals were anesthetized with isoflurane, 1.5–2%, in 100% oxygen via  a face–shoulder mask, and tail artery blood was collected in heparinized syringes for determination of TNF-α serum concentrations (2, 6, 24, and 96 h) and leukocyte counts and arterial blood gas analysis (both before and at 6, 24, and 96 h). Total blood leukocytes were counted using a hemocytometer, and a differential cell count was determined on a blood smear using Wright stain. After arterial blood gas analysis, differences in alveolar-to-arterial oxygen tension were calculated using a standard formula. For quantitative bacteria culture of circulating blood from animals selected for lung analysis, after anesthesia and before sacrifice, blood was obtained using sterile technique via  puncture of the abdominal aorta, and 100 μl was inoculated on blood agar plates.

To measure rat TNF-α concentrations in blood or lung lavage samples, a solution, 0.5%, of rabbit antibody to rat TNF-α serum was incubated in 96-well plates. After 24 h of incubation at 4°C and washing, standard and test serum (or lavage) fluids were added, and the plates were incubated for 4 h at room temperature. After washing, a 1 μg/ml concentrated biotin-conjugated antibody to rat TNF (Pharmingen, San Diego, CA) was added. After repeated washing, 0.5 μg/ml concentrated streptavidin–peroxidase conjugate solution (Dianova, Hamburg, Germany) was added. After a further 30 min, the wells were washed, and the peroxidase substrate 3,3′,5,5′-tetramethylbenzidine (Sigma, Deisenhofen, Germany) was added. The reaction was stopped after 15 min with 1 m H²SO⁴, and the absorption of the solution was read at 450 nm. In some animals from each group, TNF-α bioactivity concentrations were determined in lung lavage samples at 6 h only for comparison with the immunoassay. All reported results are those of the immunoassay unless specified. This was done with the WEHI 164 cell bioactivity assay as previously described. 44Serum interleukin-6 (IL-6) concentrations were determined using a commercial assay (Biosource, Camarrilo, CA).

Survival was assessed using a Cox proportional hazards model with terms for type of bacteria (E. coli  or S. aureus ), treatment (antibody to TNF or placebo), route of treatment (intratracheal or intraperitoneal), and the various interactions. 45Treatment effects are presented as relative risk ± SEM of the estimate. ANOVA was performed on all other parameters in an attempt to explain the difference in treatment effect between the two bacterial challenges. 46This model included terms for type of bacteria, treatment, route of treatment, time, and the various interactions. Data are presented as mean ± SEM. P < 0.05 was considered statistically significant;P < 0.10 was considered a trend. Data were transformed where appropriate for analysis.

At 6 h after intratracheal inoculation with either E. coli  or S. aureus , animals appeared weak and lethargic. In controls without antibody to TNF treatment, E. coli  and S. aureus  challenges produced similar mortality rates (P = not significant [NS]) (fig. 1). Compared with control treatment, antibody to TNF had similar effects on mortality rates associated with intratracheal and intraperitoneal routes of treatment (P = NS), and we combined over this variable. Antibody to TNF reduced mortality rates with both E. coli  and S. aureus  (fig. 1). However, these reductions (relative risk of death ± SEM) were greater with E. coli  (−1.6 ± 0.6;P = 0.006) than with S. aureus  (−0.5 ± 0.04;P = 0.185) in a trend that approached being statistically significant (P = 0.09 for the differing effects of antibody to TNF with E. coli vs. S. aureus ) (fig. 2).

The effects of antibody to TNF on both lung and blood parameters, with the exception of lung lavage TNF concentrations, were not significantly different when comparing the differing time points or the two routes of antibody to TNF treatment studied (P = NS for all). We therefore averaged across time and route of treatment for all parameters except lung lavage TNF concentrations to increase our ability to determine a cause for the beneficial effect of antibody to TNF on overall survival as well as for the possible advantage it showed with E. coli  compared with S. aureus .

Compared with S. aureus , E. coli  was associated with increased lung lavage TNF concentrations (5.9 ± 0.2 vs.  4.8 ± 0.3.9 log pg/ml, respectively;P = 0.001) and neutrophil counts (11.9 ± 1.0 vs.  5.0 ± 0.7 cells × 10⁴/ml, respectively;P = 0.0001) and with fewer lung lavage bacteria (6.8 ± 0.8 vs.  9.2 ± 1.0 log CFU/ml, respectively;P = 0.003) and lung tissue bacteria (7.3 ± 0.5 vs.  8.8 ± 0.6 log CFU/g, respectively;P = 0.05) (table 2).

Similarly (P = NS) with both E. coli  and S. aureus , compared with controls, intratracheal but not intraperitoneal antibody to TNF reduced lung lavage TNF concentrations, but these decreases were smaller over time (P = 0.0001) (table 2;fig. 3). At 6 h for the two bacteria combined, lung lavage TNF concentrations measured by immunoassay (9.1 ± 0.2 log pg/ml with control treatment vs.  8.8 ± 0.1 log pg/ml with intraperitoneal antibody to TNF vs.  5.6 ± 0.3 log pg/ml with intratracheal antibody to TNF) were of the same magnitude and showed the same ordering by route of administration as TNF bioactivity concentrations (9.9 ± 0.1 log pg/ml with control treatment vs.  9.6 ± 0.1 log pg/ml with intraperitoneal antibody to TNF vs.  0 with intratracheal antibody to TNF) except with intratracheal antibody to TNF (bioactivity concentrations were so reduced as to be nondetectable). Compared with control treatment, similarly (P = NS) with both bacteria types and routes of treatment, antibody to TNF was associated with reductions in bacteria in both lung lavage (9.6 ± 1.2 vs.  6.6 ± 0.7 log CFU/ml, respectively;P = 0.002) and lung tissue (9.2 ± 0.8 vs.  7.7 ± 0.5 log CFU/g, respectively;P = 0.05) samples (table 2;fig. 4). Other lung analysis measures did not differ (P = NS) when comparing bacteria types or treatments during the study.

Compared with S. aureus , E. coli  increased serum TNF (5.8 ± 0.1 vs.  5.0 ± 0.1 log pg/ml, respectively;P = 0.001) and IL-6 (5.5 ± 0.4 vs.  3.9 ± 0.7 log pg/ml, respectively;P = 0.02) concentrations and alveolar-to-oxygen gradients (409 ± 8 vs.  373 ± 7 mmHg, respectively;P = 0.001) and decreased circulating neutrophils (1.3 ± 0.1 vs.  1.8 ± 0.1 cells × 10³/mm³, respectively;P = 0.0001) and lymphocytes (2.9 ± 0.3 vs.  3.6 ± 0.2 cells × 10³/mm³, respectively;P = 0.0001) (tables 3 and 4).

Compared with control treatment, similarly (P = NS) with both bacteria types and routes of treatment, antibody to TNF was associated with a trend toward reduction in serum TNF concentrations (5.5 ± 0.1 vs.  5.2 ± 0.1 log pg/ml, respectively;P = 0.09) and significant increases in circulating neutrophils (1.2 ± 0.1 vs.  1.7 ± 0.1 cells × 10³/ml, respectively;P = 0.01) and lymphocytes (2.7 ± 0.2 vs.  3.6 ± 0.2 cells × 10³/ml, respectively;P = 0.001) (tables 3 and 4;fig. 5). Differently for the two bacteria types, compared with controls, antibody to TNF reduced alveolar-to-arterial oxygen gradients more with E. coli  than with S. aureus  (P = 0.04) and increased serum IL-6 levels with E. coli  but reduced them with S. aureus  (P = 0.003) (table 3;fig. 6). Other parameters in blood analysis did not differ (P = NS) when comparing bacteria type or treatment during the study.

Compared with S. aureus , similarly lethal E. coli  doses in this rat model increased serum and lung lavage TNF concentrations, lung lavage neutrophil counts, and serum IL-6 concentrations and decreased arterial oxygenation (i.e. , increased alveolar-to-oxygen gradients) and circulating neutrophil and lymphocyte counts. With the exception of lung lavage TNF concentrations, antibody to TNF had similar effects whether it was administered via  intratracheal or intraperitoneal routes. Overall with both bacteria types, antibody to TNF was associated with decreased lung bacteria and serum and lung lavage TNF concentrations, increased circulating neutrophils and lymphocytes, and improved survival rates. Although not reaching statistical significance, there was a strong trend for improved survival rates to be greater with E. coli  than with S. aureus . This difference in survival rates with antibody to TNF with E. coli  compared with S. aureus  was associated with bigger improvements in oxygenation and with increases in serum IL-6 concentrations.

Increased circulating neutrophils and lymphocytes and reduced lung bacteria counts with treatment during both E. coli  and S. aureus  pneumonia suggest that improved survival rates with antibody to TNF may have been related to better host defense. Decreased serum TNF concentrations with antibody to TNF may have reduced endothelial activation and leukocyte adhesion molecule expression in nonpulmonary tissues, increasing the availability of these host effector cells at the lung for microbial clearance. 9,47,48Although increases in leukocyte counts were not evident in lung lavage samples, increased numbers of effector cells may have been present in pulmonary interstitial spaces. Alternatively, however, the effects of antibody to TNF on circulating leukocytes and lung bacteria may represent independent protective mechanisms. Decreases in serum TNF concentrations and increased circulating leukocytes with antibody to TNF may reflect reductions in systemic vascular inflammatory injury and improved systemic hemodynamic function as has been shown in other models. 49,50In addition, decreases in lung TNF concentrations with antibody to TNF may have reduced local inflammatory responses that have been suggested to interfere with host defense. 51It is also possible that reductions in lung TNF concentrations removed a direct stimulus for bacterial growth as has been demonstrated in in vitro  systems. 52–54This latter mechanism appears less likely since intratracheal treatment, which resulted in the greatest reductions in lung lavage TNF concentrations, did not have differing effects on lung bacteria counts or survival rates when compared with intraperitoneal treatment.

Our results differ from those of other studies, which demonstrated that inhibition of TNF interferes with local host defense function and outcome with extravascular infection. 4,55These differences however may in part reflect variation in the underlying severity of infection studied. We have shown that independent of its type and route, the severity of an infectious challenge has an important influence on the effect of antiinflammatory agents in animal models of sepsis. 56With mild infections (control mortality rates of less than 25%), antiinflammatory agents may interfere with local protective host mechanisms and worse outcome. However, at higher mortality rates, antiinflammatory agents become progressively more beneficial as the host loses its ability to contain an extravascular nidus of infection and the ensuing systemic inflammatory response results in injury in noninfected tissues. This influence of disease severity is demonstrable in clinical trials testing antiinflammatory agents in sepsis as well. 37,57,58 

We hypothesized that antibody to TNF would be more efficacious with E. coli  than with S. aureus  in this pneumonia model based on prior studies showing that TNF levels were greater with the former than with the latter. 42Once again in this study, TNF concentrations with similarly lethal challenges were greater with E. coli  than with S. aureus . With this, although not reaching a statistically significant level, there was a strong trend (P = 0.09) for antibody to TNF to improve survival rates more with E. coli  than with S. aureus . The physiologic basis for this advantage may be related to the improved oxygenation observed with treatment during E. coli  but not S. aureus  infection. Alternatively, suppression of serum IL-6 concentrations with antibody to TNF during S. aureus  but not E. coli  infection may represent a maladaptive response, although not one sufficient to totally negate the beneficial survival effect associated with treatment. Of note, although not significant, reductions in lung bacteria counts were greater with E. coli  than with S. aureus , suggesting that there may have been differential improvements in host defense with antibody to TNF when comparing the two infection types. Overall, however, these findings are consistent with other work demonstrating that the role of TNF in the pathogenesis of organ injury and outcome during lethal infection with differing bacterial types may not be the same. 38–41Therefore, as this study suggests, type of bacterial infection may be an important determinant altering the effectiveness of mediator-specific antiinflammatory agents such as TNF antagonists in sepsis. We showed previously in this same model that neutrophil stimulation with granulocyte colony-stimulating factor greatly improved survival with S. aureus  but worsened it with E. coli . 42 

Studies suggest that injury to the alveolar–capillary membrane resulting in the intravascular uptake of TNF produced locally at a site of pneumonia may be an important pathogenetic step in the development of systemic inflammation and death. 59,60Injury related to the production of TNF itself has been implicated in the loss of this compartmentalized response. 61On the basis of this, we reasoned that inhibiting alveolar TNF directly would result in the greatest improvements in outcome in our model. However, despite significant decreases in lung lavage TNF concentrations with intratracheal but not with intraperitoneal antibody to TNF, route of treatment did not alter lung leukocyte counts, any parameter of outcome, or systemic TNF concentrations with either bacteria type. This finding is consistent with those of other studies suggesting that the local pulmonary production of TNF even during lethal pneumonia may not be necessary alone to cause either the lung injury or the systemic inflammatory response observed. 62Other mediators (e.g. , IL-1 or IL-8) produced in the lung in addition to TNF may be sufficient for the extravascular recruitment of inflammatory cells and the lung injury that occurs. Furthermore, loss of compartmentalization with the release of either microbe and microbial toxins or host mediators other than TNF may be sufficient to produce the systemic inflammatory response associated with infection in this model.

In conclusion, although antibody to TNF improved survival with both lethal E. coli  and S. aureus  pneumonia in this rat model, this protective effect was stronger with E. coli  than with S. aureus . These findings suggest that the efficacy of mediator-specific antiinflammatory agents in sepsis may relate in part to the type of underlying infection. Because Gram-positive bacterial infections are increasingly frequent in patients presenting with sepsis, our findings suggest that accounting for the type of bacterial infection may be important to improve the effectiveness of antiinflammatory therapies for these patients.