Using a murine model of sepsis, we found that the balance of tissue pro- to anti-inflammatory cytokines directly correlated with severity of infection and mortality. Sepsis was induced in C57BL/6 mice by cecal ligation and puncture (CLP). Liver tissue was analyzed for levels of IL-1β, IL-1 receptor antagonist (IL-1ra), tumor necrosis factor (TNF)-α, and soluble TNF receptor 1 by ELISA. Bacterial DNA was measured using quantitative real-time PCR. After CLP, early predominance of proinflammatory cytokines (6 h) transitioned to anti-inflammatory predominance at 24 h. The elevated anti-inflammatory cytokines were mirrored by increased tissue bacterial levels. The degree of anti-inflammatory response compared with proinflammatory response correlated with the bacterial concentration. To modulate the timing of the anti-inflammatory response, mice were treated with IL-1ra before CLP. This resulted in decreased proinflammatory cytokines, earlier bacterial load, and increased mortality. These studies show that the initial tissue proinflammatory response to sepsis is followed by an anti-inflammatory response. The anti-inflammatory phase is associated with increased bacterial load and mortality. These data suggest that it is the timing and magnitude of the anti-inflammatory response that predicts severity of infection in a murine model of sepsis.
- bacterial load
the systemic inflammatory response syndrome (SIRS) represents an intense host inflammatory response to an infectious or noninfectious insult (6). The SIRS response is mediated by the production of proinflammatory cytokines such as tumor necrosis factor (TNF)-α, IL-1β, and IL-12. However, this early response is followed by a transition to a sustained production of anti-inflammatory mediators such as IL-10, IL-1 receptor antagonist (IL-1ra), and soluble TNF receptor 1 (sTNFR1) (5). Sepsis represents SIRS in response to infection. Early animal studies of sepsis suggested that a largely unopposed proinflammatory response to severe sepsis resulted in increased organ injury and mortality, whereas a greater anti-inflammatory response resulted in less severe sepsis (38). However, more recent findings have suggested that the unopposed proinflammatory response in septic patients occurs less frequently than originally anticipated and that higher levels of anti-inflammatory mediators may in fact be more predictive of mortality (14, 26, 31, 32). It has been shown that this intense anti-inflammatory response may induce a state of immunosuppression in patients with sepsis and SIRS (34). It is of interest that signs of organ dysfunction and mortality often occur during this period of immunosuppression.
The murine cecal ligation and puncture (CLP) model has been shown to induce a polymicrobial sepsis that closely mimics human sepsis (40). By monitoring the gauge of the puncture needle, one can modulate the outcome of the response. We chose to use a 19-gauge needle, which induces sepsis and death at 48–96 h post-CLP in C57BL/6 mice. The mice were killed at various time points that have previously been shown to correlate with organ injury and mortality (11). Liver injury plays a central role in sepsis, so we determined levels of tissue inflammatory cytokines and their specific antagonists in homogenized liver samples. Bacterial concentration was measured by quantitative real-time PCR with primers specific for the bacterial 16s ribosome gene. We found that after an initial proinflammatory response, there was a transition to a predominantly anti-inflammatory response in liver tissue by 24 h. We found that the magnitude of this local anti-inflammatory response was associated with local severity of infection (bacterial load) and mortality. To evaluate the possible link between anti-inflammatory cytokines and mortality, we increased the early IL-1ra load by pretreating mice with IL-1ra before CLP. This resulted in earlier and more severe infection (bacterial load) as well as increased mortality. These data suggest that it is the timing and magnitude of the anti-inflammatory response rather than the proinflammatory response that predicts severity of infection in sepsis.
MATERIALS AND METHODS
CLP and IL-1ra administration.
C57BL/6 mice (female, age 6 wk; Harlan Laboratories, Indianapolis, IN) underwent CLP with a 19-gauge needle as previously described (38). Ebong et al. (13) have shown that increasing the size of the needle results in increased mortality. In their model, an 18-gauge needle was associated with 100% mortality, whereas a 21-gauge needle was associated with 50% mortality. Because laparotomy itself can cause inflammation (41), one group of control mice did not undergo laparotomy. To assess the degree of inflammatory response from CLP relative to surgery alone, sham laparotomy without CLP was performed on a separate group of animals. For the immunomodulation experiments, C57BL/6 mice were given 1 μg of intraperitoneal recombinant murine IL-1ra (R&D Systems, Minneapolis, MN) 1 h before CLP and at the time of CLP. This is consistent with doses given in a previous study (30). Liver homogenates from these animals were used to determine cytokine levels and bacterial load and are represented in Figs. 3, A and B, and 4B. To assess the difference between bacterial load at 6 h in CLP animals and animals treated with IL-1ra before CLP, a separate study was performed. The data from this study are represented in Fig. 4A. Survival studies were performed on separate animals. For all studies, numbers of mice are indicated in the figure legends. University of Iowa Animal Care Unit postprocedural care and monitoring guidelines were followed.
Liver harvest and homogenization.
At predetermined time points of 6, 16, and 24 after CLP, mice were anesthetized with ketamine before death. Livers were harvested for assessment of cytokine and bacterial concentrations. Before homogenization, livers were perfused free of intravascular cells. Livers were homogenized in 1 ml of complete antiprotease cocktail (Roche Pharmaceuticals, Basel, Switzerland), sonicated on ice, and then centrifuged at 4,000 rpm for 10 min at 4°C. Supernatants were collected and stored at −70°C for cytokine analysis.
For measurement of IL-1β, TNF-α, IL-1ra, and sTNFR1 in homogenized liver, ELISA kits from R&D Systems were used according to the manufacturer's instructions.
Quantitative real-time PCR.
DNA was isolated from liver homogenates using the Easy-DNA kit (Invitrogen, Carlsbad, CA) following the manufacturer's instructions. DNA was quantitated using the PicoGreen kit (Molecular Probes, Eugene, OR). In each well of a 96-well, optical quality PCR plate (Bio-Rad, Hercules, CA), 5 ng of experimental sample DNA was added to 48 μl of PCR reaction mixture containing 180 μM each dNTP (Invitrogen), 3 mM MgCl2 (Invitrogen), 1/10,000 SYBR Green I DNA Dye (Molecular Probes), 0.2 μM of each sense and antisense primers (IDT, Coralville, IA), and 2.5 units of Platinum Taq DNA Polymerase (Invitrogen). Amplification was then performed in an iCycler iQ Fluorescence Thermocyler (Bio-Rad) as follows: 3 min at 95°C, followed by 45 cycles of 20 s at 95°C, 20 s at 58.5°C, 20 s at 72°C, and 10 s at 81°C. Fluorescence data was captured during the 10-s dwell to ensure that primer-dimers were not contributing to the fluorescence signal generated with SYBR Green I DNA Dye. Specificity of the amplification was confirmed using melting curve analysis. Data were collected and recorded by iCycler iQ software (Bio-Rad) and expressed as a function of threshold cycle (Ct), the cycle at which the fluorescence intensity in a given reaction tube rises above background (calculated as 10× the mean standard deviation of fluorescence in all wells over the baseline cycles). Primers used to amplify bacterial 16s ribosomal RNA (rRNA) gene are as follows (5′ to 3′): forward, GAGGAAGGIGIGGAIGACGT; reverse, AGGAGGTGATCCAACCGCA. Similar primers were used in a study to detect bacteremia in critically ill patients (10). Quantitation of 16s rRNA gene copy number in experimental samples was determined by comparing Ct values with a standard curve generated by spiking mouse genomic DNA with serially diluted Escherichia coli genomic DNA. Sensitivity of the assay was determined to be 200 copies (∼30 genomes) of E. coli DNA per 5-ng sample.
We used quantitative real-time PCR for determining bacterial load. Although standard plating techniques detect increased bacterial concentration in tissue following CLP (11), this method is less sensitive than PCR in that it cannot quantify exact numbers of organisms and may not be able to detect low levels of bacteria (10). Furthermore, sepsis is associated with an active cellular and humoral response, resulting in bacterial killing. Standard culture techniques, which rely on bacterial viability, may not represent the true bacterial load in the setting of a brisk serum bactericidal response (1, 39). In an animal model, PCR was shown to be more sensitive than standard culture in detecting bacteremia in animals previously treated with antibiotics (18).
Statistical analyses were performed using the Prism program from GraphPad Software (San Diego, CA) and are specifically stated in the figure legends.
CLP induces first proinflammatory and then anti-inflammatory mediators.
We chose to analyze TNF-α and IL-1β as important sepsis-related proinflammatory cytokines. For these studies, TNF-α and IL-1β levels in liver homogenates were measured by ELISA. We found that the tissue levels of TNF-α and IL-1β peaked at 6 h after CLP and then trended down toward baseline by 24 h (Fig. 1A). In our CLP model, ∼50% of the animals are dead by 48 h, and all the animals are dead by 4 days post-CLP. TNF-α concentrations in the liver were relatively low at baseline (7,790 pg/g liver ± 1,907, means ± SE) and increased to a peak of 46,259 pg/g liver ± 1,650 at 6 h (P < 0.001). Similarly, IL-1β levels were low at baseline (1,319 pg/g liver ± 172) and peaked at 12,268 pg/g liver ± 3,961 at 6 h after CLP (P < 0.001). These findings suggest that tissue inflammation begins early after the onset of sepsis. Sham surgery resulted in a mild increase in hepatic proinflammatory cytokines at 6 h after surgery (TNF-α: 9,419.2 pg/g liver ± 219.2; IL-1β: 3,252.4 pg/g liver ± 738.1). This represents a small fraction of the total proinflammatory response observed following CLP.
We measured levels of sTNFR1 and IL-1ra by ELISA. We found a peak in tissue levels of these anti-inflammatory mediators at 24 h (Fig. 1A). Tissue levels of sTNFR1 and IL-1ra were low at baseline (540 pg/g liver ± 150 and 2,303 pg/g liver ± 194, respectively) and peaked at 24 h after CLP with mean values of 6,128 pg/g liver ± 603 and 16,422 pg/g liver ± 673, respectively (P < 0.001). In contrast, we found no increase in anti-inflammatory mediators after sham surgery compared with nonsurgical controls at all time points (data not shown). These findings are consistent with previous studies that found the initial serum proinflammatory response was followed by an anti-inflammatory response (8).
Analysis of the ratio of anti- to proinflammatory cytokines in CLP mice demonstrates a predominantly anti-inflammatory response at 24 h.
In a study of patients with sepsis, it has been shown that a sustained high ratio of serum anti-inflammatory to proinflammatory mediators is associated with more severe disease and poor outcome (14). We calculated the ratios of tissue sTNFR1 to TNF-α and IL-1ra to IL-1β after CLP (Fig. 1B). In the control mice, the ratio of IL-1ra to IL-1β showed a slight predominance of anti-inflammatory mediators. At 6 h after CLP, the ratio of IL-1ra to IL-1β changed significantly to one of inflammatory predominance (P < 0.01) consistent with the early peak in IL-1β. At 24 h after CLP, there was a significant change to anti-inflammatory predominance that far exceeded the baseline relationship in both the IL-1ra/IL-1β and sTNFR1/TNF-α analyses (P < 0.001). This transition to anti-inflammatory predominance occurred concomitantly with increased mortality 24 h after CLP.
Bacterial load correlates with the anti- vs. proinflammatory ratio and not the magnitude of the individual cytokines.
Bacterial concentration was measured by quantitative real-time PCR in liver homogenates (Fig. 2A). We found that bacterial concentration in liver was barely detectable at baseline (0.896 organisms/g liver ± 0.104, geometric mean ± SE) and peaked 24 h after CLP at 473 organisms/g liver ± 306 (P < 0.001). This corresponds to a 1.5 log increase in bacterial load. This is consistent with tissue levels of bacteria measured by standard plating (11). Sham surgery resulted in no increase in hepatic bacterial load (data not shown).
We compared bacterial concentration with ratios of anti- to proinflammatory mediators at 24 h after CLP (Fig. 2B). We found that the degree of anti-inflammatory predominance correlated with bacterial load (IL-1ra:IL-1β, r2 = 0.97; sTNFR1:TNF-α, r2 = 0.83). We also compared bacterial concentration at 24 h with individual levels of TNF-α, IL-1β, sTNFR1, and IL-1ra and found no correlation, with r2 values of 0.2, 0.1, 0.6, and 0.1, respectively (data not shown). These findings suggest that the relative concentrations of pro- and anti-inflammatory mediators may be more predictive of severity of infection than the absolute amounts of each individual mediator.
Pretreatment with IL-1ra blunts the initial proinflammatory response and alters the timing of the anti-inflammatory response after CLP.
To evaluate whether we could alter the initial proinflammatory response, we treated mice with 1 μg of recombinant murine IL-1ra 1 h before and at the time of CLP. IL-1ra is known to inhibit the function of IL-1β. We found that pretreatment with IL-1ra suppressed the initial proinflammatory response of both IL-1β and TNF-α following CLP (Fig. 3A). The exact mechanism of TNF-α suppression is unclear; however, a previous study has shown that IL-1ra suppresses TNF-α release from macrophages (22). At 6 h following CLP, IL-1β and TNF-α levels in liver homogenates were not significantly different from baseline. Comparison of the mean values of proinflammatory cytokines at 6 h following CLP revealed a significant difference between the CLP group (Fig. 3A, solid line) and the IL-1ra+CLP group (Fig. 3A, dashed line) (P < 0.01).
Interestingly, the baseline levels of IL-1β were the same in the group that was pretreated with IL-1ra and the group without pretreatment. Potential explanations for this finding include a lack of effect of IL-1ra on baseline IL-1β levels or, perhaps, a 1-h exposure to IL-1ra is not sufficient to alter IL-1β levels in the liver.
We calculated the ratios of tissue sTNFR1 to TNF-α and IL-1ra to IL-1β after CLP in animals pretreated with IL-1ra (Fig. 3B). At 6 h following CLP, the ratio of IL-1ra to IL-1β showed a significant anti-inflammatory predominance (P < 0.001) consistent with the blunted proinflammatory response. This ratio returned to baseline by 16 h. The ratio of sTNFR1 to TNF-α transitioned to anti-inflammatory predominance by 16 h following CLP (P < 0.001).
Pretreatment with IL-1ra results in earlier and increased severity of infection following CLP.
To investigate the relationship between a predominantly anti-inflammatory state (high levels of IL-1ra) and severity of infection, we treated mice with 1 μg of recombinant murine IL-1ra 1 h before and at the time of CLP. To evaluate the local severity of infection, we measured bacterial load in liver homogenates of IL-1ra treated compared with untreated CLP mice using quantitative real-time PCR (Fig. 4A). In the IL-1ra-pretreated mice, we found that the bacterial concentration in liver homogenates was barely detectable at baseline and increased to 347,600 ± 72,350 organisms/g liver at 6 h after CLP (P < 0.001). This corresponds to a 5 log increase in bacterial load. The increase in bacterial load at 6 h after CLP was greater in the mice pretreated with IL-1ra (P < 0.001) compared with CLP alone.
To determine whether this earlier severity of infection correlated with a state of anti-inflammatory predominance, we evaluated the ratio of IL-1ra:IL-1β in liver homogenates from mice pretreated with IL-1ra. For this experiment, we used a different group of mice than those represented in Fig. 4A. At 6 h after CLP, the ratio of IL-1ra:IL-1β showed an anti-inflammatory predominance that far exceeded the ratio seen after CLP alone (Fig. 3B). We then compared bacterial concentration with ratios of anti- to proinflammatory mediators (Fig. 4B). We found that the degree of anti-inflammatory predominance correlated with bacterial load (r2 = 0.75). These data indicate that a predominantly anti-inflammatory state is associated with increased bacterial load. Although we used separate animals for the experiments in Fig. 4, A and B, the mean bacterial load in the two groups (5.41 ± 0.132 logs and 4.62 ± 0.374 logs, respectively) was not significantly different based on a Student's t-test analysis.
Pretreatment with IL-1ra increases mortality after CLP.
Finally, we asked if the increased anti- vs. proinflammatory ratio and increased bacterial load in IL-1ra-treated mice altered mortality after CLP. The mortality in the group receiving IL-1ra with CLP occurred significantly earlier than in the non-IL-1ra-treated mice (P < 0.001; Fig. 5). The increase in mortality correlates with increased anti-inflammatory predominance and increased bacterial load. In the untreated mice, mortality after CLP increased at 24 h (Fig. 5, solid line). There was 100% survival at 6 and 16 h, whereas only 70 and 40% survived at 24 and 48 h, respectively. The mortality after CLP was similar to that found in a previous study from our laboratory (11). In the mice pretreated with IL-1ra, there was earlier and increased mortality (Fig. 5, dashed line). We found 37.5% mortality at 6 h in the IL-1ra-treated mice compared with 0% mortality in the untreated group. In addition, the IL-1ra-treated mice had 67% mortality at 16 h and 100% mortality at 24 h. As a composite, these data suggest that the timing and magnitude of anti-inflammatory mediators play an important role in the outcome of CLP-induced sepsis.
We found that the early proinflammatory response to sepsis in the liver is followed by a marked production of anti-inflammatory mediators and a transition to an anti-inflammatory state. This observation is consistent with prior studies that documented the same transition in response to sepsis (34). During this state of anti-inflammatory predominance, monocytes are deactivated, resulting in reduced antigen presentation and decreased production of proinflammatory cytokines (3, 36). If significant numbers of viable bacteria are still present at the time of transition to anti-inflammatory predominance, they are more free to replicate and cause a severe infection. Using a CLP model of sepsis, we found that production of proinflammatory cytokines after CLP is followed by a period of anti-inflammatory cytokines. This is in contrast to sham surgery, which resulted in a very minor proinflammatory response with no subsequent anti-inflammatory response. Surprisingly, the bacterial load in the liver generated by CLP correlated with the ratio between anti- and proinflammatory cytokines rather than the magnitude of each cytokine production. We were able to modulate the timing and severity of the anti-inflammatory predominance by changing IL-1ra levels. Adding exogenous IL-1ra to mice undergoing CLP resulted in a decrease in proinflammatory mediators, increased bacterial load in the liver, and earlier mortality. These data suggest that rather than the degree of proinflammatory cytokine production determining severity of infection in sepsis, it is the timing and magnitude of the anti-inflammatory predominance that determine bacterial load.
Whereas a prior study showed improved survival in a CLP model with continuous intravenous infusion of IL-1ra and sTNFR1, this study utilized a sublethal model of sepsis as well as antibiotic therapy to clear infection (30). Thus inflammation triggered by mediators like TNF-α and IL-1β was the likely cause of death in that study. Our data clearly show that in a lethal sepsis model, treatment with IL-1ra, to block the initial proinflammatory response and alter the timing of the anti-inflammatory response, is associated with increased severity of infection and mortality.
It has been suggested that serum cytokine levels may not be representative of local tissue inflammation and end-organ damage (4). This may be because cytokines that are produced locally have the greatest effect on tissues (12). Therefore, we performed our studies at the tissue level, using the liver as a model. The rationale for choosing the liver is that it is an important source of cytokines during sepsis, it participates in bacterial clearance via the reticular-endothelial system, and it is often injured as a result of sepsis (12, 24). Our previous work has shown that liver end-organ injury, as indicated by elevated serum alanine aminotransferase levels and histopathological evidence of injury, occurs at 24 h after CLP (11). This coincides with the onset of mortality and other organ injury. The use of this model allowed us to observe the transition from an initial proinflammatory response to an anti-inflammatory state at the tissue level. We were also able to correlate the degree of immunosuppression with the amount of accumulation of bacteria in the liver. Both the development of the anti-inflammatory state and the accumulation of bacteria occurred at time points consistent with increased mortality.
Many studies have shown an initial brisk proinflammatory response following onset of sepsis (38). It is clear that inflammatory cytokines are released in tissues and their local levels are related to the severity of the initial localized inflammatory response (13). Studies have shown serum TNF-α and IL-1β levels to be predictive of organ injury (7) and mortality (17, 25, 37). In addition, previous studies report a strong correlation between severity of sepsis and serum TNF-α levels (23). However, inflammatory cytokines are also necessary to control the initial infection. It is therefore not surprising that although therapies directed against TNF-α (2) and IL-1β (27) have improved survival in murine endotoxemia models, blocking TNF-α and IL-1β has failed to improve survival in patients with sepsis (8, 29). In addition, mice that are resistant to endotoxin have been shown to have increased mortality with true sepsis (16). These observations suggest that infection is an important factor in determining severity of illness. Therefore, we used a true infection model of sepsis rather than an endotoxemia model. This study showed that the initial brisk inflammatory response is associated with low bacterial load. As the inflammatory response is superceded by an anti-inflammatory response with its resultant immunosuppression, the bacterial load increases. Furthermore, blocking the initial proinflammatory response with IL-1ra resulted in increased severity of infection. Our results provide further evidence of the importance of inflammatory mediators such as TNF-α and IL-1β in the control of infection.
Whereas some studies have shown that proinflammatory mediators predict mortality (23), other studies have shown that anti-inflammatory mediators may be better predictors of mortality in sepsis (14, 21, 31, 33). It has become clear that the measurement of any single circulating mediator cannot reflect functional status and that it is the balance between proinflammatory and anti-inflammatory mediators that is related to outcome in severe sepsis (38). To evaluate this balance, we calculated the ratios of inflammatory cytokines and their respective, specific inhibitors. The use of ratios to reflect net pro- or anti-inflammatory activity is based on the fact that TNF-α and the type I IL-1 (IL-1RI) receptor interact with their respective soluble inhibitors on a 1:1 stoichiometric basis (28). The hypothesis that a balance between IL-1β and IL-1ra is important to maintain a state of homeostasis has recently been supported by the observation that IL-1ra knockout mice develop an inflammatory arthropathy similar to rheumatoid arthritis (19). In addition, in patients with inflammatory bowel disease, the ratio of IL-1β to IL-1ra is higher than in normal subjects and is related to severity of disease (9). Patients with acute respiratory distress syndrome have been shown to have increased IL-1ra:IL-1β ratio in bronchoalveolar lavage fluid (28). It appears that the sole essential function of IL-1ra is the competitive inhibition of the IL-1RI (20). Previous studies have shown that from 50- to 500-fold excess of IL-1ra to IL-1β is necessary to inhibit IL-1β's action on effector cells in vitro (15). Our results from liver samples indicate that significantly less excess of IL-1ra is necessary to have decreased IL-1β levels. To our knowledge, no prior studies have evaluated the ratio of IL-ra:IL-1β in the liver. The function of the sTNFR1 is more complex. At low levels, it does not appear to negate the effects of TNF-α, whereas at higher levels, it inhibits TNF-α activity (35).
Our findings show that the ratio of anti- to proinflammatory mediators is more predictive of the severity of infection than are the absolute measurements of any individual mediator. In addition, the transition to anti-inflammatory predominance is associated with a period of immunosuppression during which there is increased severity of infection, increased end-organ damage, and increased mortality.
This work was supported by a Veterans Administration Merit Review Grant and by National Institutes of Health Grants ES-09607 and HL-60316 (to G. W. Hunninghake).
We are grateful for the statistical advice and guidance from Dr. Bridgett Zimmerman.
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Copyright © 2005 the American Physiological Society