INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Septic shock has become one of the most frequent causes of morbidity and mortality in intensive care units (1, 2). Treatment of sepsis consists of support of blood pressure, organ blood flow, and ventilation in an intensive care environment, along with an emphasis on antibiotics and eradication of sources of infection. Despite significant advances in therapies available and understanding of pathogenesis, the mortality from septic shock has improved little over the last several decades (3). Studies of sepsis in humans are difficult because the seriousness of the disease mandates immediate intervention and because the heterogeneity of patient presentations imposes substantial limitations on clinical trials (4). Thus, animal models have been used extensively to explore the pathogenesis of sepsis and to generate preclinical data for therapeutic interventions. For both purposes, it is important to document that the animal model reproduces the relevant physiologic parameters present in patients with septic shock (5).

Shock in sepsis is caused by persistent vasodilation with a normal or high cardiac output, which results in hypotension, hypoperfusion, and a high mortality (1, 6). In the initial phases of sepsis, venodilation and capillary leak can cause filling pressures to be low (7). Adequate fluid resuscitation is necessary to produce the typical hyperdynamic state with increased cardiac output and decreased systemic vascular resistance (6, 8).

Some large animal models have been shown to reproduce the hyperdynamic hemodynamics present in septic humans (9). Natanson and colleagues have established a canine model entailing implantation of an infected peritoneal clot that reproduces not only the hyperdynamic state and hypotension seen in sepsis, but also the time course and extent of myocardial depression (9). These investigators also documented that resuscitation with fluids and treatment with antibiotics improve mortality in this model (12). Other models using continuous endotoxin administration in cynomolgus monkeys (10) and sheep (11) have also documented the presence of a hyperdynamic state.

Not all animal models, particularly those without adequate fluid repletion, reproduce the typical hyperdynamic hemodynamics seen in resuscitated patients (5). In addition, it is well recognized that supportive therapy can alter survival in an animal model considerably (5, 13). Therapeutic interventions that are effective in untreated models may not work as well when combined with antibiotics and other supportive measures (5, 13). In addition, acute administration of microbial toxins to animals may lead to an exaggerated release of host cytokines that may not replicate many of the important features of clinical sepsis in patients (14).

These difficulties can be particularly acute in rodent models of sepsis. Unlike humans, rodents are resistant to endotoxin, and use of the high doses of endotoxin necessary to produce hypotension and mortality in mice may lead to toxic effects not seen at the lower doses that lead to sustained inflammatory responses in endotoxin-sensitive species such as humans (15). In addition, interventions that protect rodents in models of endotoxin infusion may not be similarly protective in more clinically relevant models such as peritonitis (16).

Nonetheless, given the ready availability of transgenic technology, appropriate rodent models of sepsis, particularly those involving mice, are desirable for studies of pathogenetic mechanisms and therapy. We set out to develop a clinically relevant model of murine sepsis employing peritonitis with fluid resuscitation and antibiotic administration, and we hypothesized that such a model would produce the vasodilation and hyperdynamic response typical of human sepsis.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The study was performed in accordance with National Institutes of Health (NIH) guidelines for the use of experimental animals, and the protocol was approved by the institutional Animal Care and Use Committee. Animals were made septic by cecal ligation and puncture (CLP), treated with antibiotics, and resuscitated with fluids. Blood pressure was measured using manometric catheters placed in the carotid artery, and stroke volume and cardiac output were measured in the same animals using echocardiography.

CLP

Sepsis was induced surgically by CLP as previously described (17). C57/BL6 mice weighing 20 to 25 g (approximately 12 wk old), obtained from Charles River Laboratories (Wilmington, MA), were anesthetized, the cecum ligated, and a puncture wound made with an 18-gauge needle. For sham operations, laparotomy was performed in a similar manner with ligation and puncture omitted. In some experiments (n = 12), mice were killed 12 h after CLP, and blood was sent for cultures.

Mortality Experiments

Survival was tested in a cohort of animals separate from those undergoing hemodynamic studies. Mice were anesthetized and made septic by CLP as described previously; control mice underwent sham ligation. After CLP the mice were treated once for perioperative discomfort with buprenorphine (1 mg/kg intramuscularly), but after that no further anesthesia was given. Treated mice were resuscitated with fluids (50 ml/kg 0.9 N saline at the time of surgery and every 6 h thereafter subcutaneously) and treated with appropriate antibiotics (ceftriaxone 30 mg/kg intramuscularly and clindamycin 30 mg/kg intramuscularly) every 6 h. Mice were observed continuously and survival recorded each hour for 48 h.

Hemodynamic Experiments

For the hemodynamic studies, blood pressure measurements and echocardiographic measurements were made simultaneously for the entire time course of the experiment in anesthetized animals. The mice were anesthetized by intramuscular injection of ketamine 100 mg/kg and acepromazine 2.5 mg/kg, and a constant level of anesthesia was maintained throughout the experiment by supplemental doses of 20/0.5 mg as needed. The carotid artery was cannulated for measurement of systemic blood pressure with a 1.4-F high-fidelity micromanometric pressure transducer (Millar Instruments, Houston, TX), and pressure was recorded continuously using MacLab (ADInstruments, Mountain View, CA). The temperature of the mice was kept at 36 to 37° C throughout the experiments using a heating lamp. Echocardiography was performed at 6-h intervals using a 12-MHz transducer and a Hewlett-Packard 5500 echocardiography machine (Hewlett-Packard, Palo Alto, CA). Aortic outflow tract diameter was determined by parasternal long axis M-mode, and pulse Doppler was used to measure aortic outflow tract velocities from both the apical four-chamber and suprasternal views. Stroke volume was calculated by multiplying aortic area by the time-velocity integral of aortic outflow. Cardiac output was then calculated by multiplying stroke volume by heart rate (HR).

Materials

Ketamine, acepromazine, and buprenorphine were obtained from the animal facility at Rush Medical College and given undiluted. Ceftriaxone (Roche, Nutley, NJ) and clindamycin (Upjohn, Peapack, NJ) were obtained from the hospital pharmacy and reconstituted in normal saline.

Reproducibility of Echocardiographic Measurements

Intraobserver variability for measuring baseline stroke volume was assessed by comparing two separate measures in each of the 24 animals. The correlation coefficient was 0.981, with a delta of 2.5 ± 1.7%. Interobserver variability was assessed by comparing 30 paired measures by two independent observers. The correlation coefficient between the paired measurements of stroke volume was 0.952, with a delta of 3.2 ± 2.1%.

Data Analysis

Data are reported as mean ± SD, with n indicating the number of animals. Mortality experiments were analyzed using Kaplan-Meier survival analysis and log-rank testing for significance. Hemodynamic experiments were analyzed using repeated measures analysis of variance (ANOVA). Values of p < 0.05 were considered significant.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Survival

All of the septic mice tested were bacteremic with gram-negative enteric organisms (Escherichia coli and others); some mice grew anaerobic organisms as well (see Table 1). Survival studies were done in three separate experiments. There were no deaths in sham-ligated control animals (n = 30). Survival in untreated septic mice (n = 30) was 0%. Survival increased to 24% with fluid resuscitation (n = 28, p < 0.05 versus untreated) and to 30% with antibiotic treatment alone (n = 33, p < 0.05 versus untreated). In mice treated with fluids and antibiotics, however, survival was significantly improved (n = 28, 46%, p < 0.01 versus untreated and p < 0.05 versus fluids alone and antibiotics alone by Kaplan-Meier log-rank analysis; Figure 1).

View larger version (25K): [in this window] [in a new window]   Figure 1.   Survival in septic unresuscitated and resuscitated mice. Survival curves for mice made septic by CLP. Three separate experiments were performed with 10 mice per group in each experiment. Sham-operated control mice had no mortality. Untreated septic mice had 0% survival at 48 h. Septic mice resuscitated with fluids and with antibiotics alone (circles, solid line) had improved survival compared with untreated mice (p < 0.05 by Kaplan-Meier analysis). Septic mice treated with both fluids and antibiotics had improved survival compared with both untreated mice (p < 0.01) and mice treated with either fluids or antibiotics alone (p < 0.05). Abx = antibiotics; NoRx = no therapy.

Hemodynamics

Hemodynamics were measured in anesthetized mice resuscitated with fluids and treated with antibiotics. Mice were anesthetized with ketamine/acepromazine 100 mg/2.5 mg intramuscularly, and a constant level of anesthesia was maintained with additional doses as needed. Survival was not analyzed in this cohort, and animals surviving to 48 h were killed. Premorbid values (defined as values obtained within 2 h of animal death) were excluded. The number of animals evaluated at each time point (dropouts were due to mortality or to technical difficulties in blood pressure measurement) is listed in Table 1. A sample tracing showing simultaneously acquired Doppler envelopes and aortic blood pressure tracings is shown in Figure 2. A sample M-mode echocardiographic tracing is shown in Figure 3.

View larger version (30K): [in this window] [in a new window]   Figure 2.   Sample tracing showing simultaneously acquired aortic flow Doppler signals and aortic blood pressure tracings. A calibrated signal from micromanometric catheter was fed into the echocardiography machine and displayed. Calibration numbers on the right represent flow velocities in cm/s; calibration for the blood pressure is not shown. Blood pressure values were measured using the original signal from the data analysis package.

View larger version (27K): [in this window] [in a new window]   Figure 3.   Sample M-mode echocardiographic tracing showing systolic and diastolic left ventricular dimensions in a septic animal.

In sham-operated control mice (n = 12), mean HR was 420 ± 31 beats/min, mean arterial pressure () was 100 ± 8 mm Hg, mean stroke volume was 26 ± 3 µl, and mean cardiac output was 12.5 ± 6.6 ml/min at baseline. These values did not change significantly over 48 h (Figure 4). End-diastolic dimension was 3.37 ± 0.42 mm (Table 1).

View larger version (51K): [in this window] [in a new window]   Figure 4.   Hemodynamics in septic and control mice. Septic animals are shown with square symbols and a dotted line, control mice with round symbols and a solid line. (a) Heart rate (HR) in septic and control mice. (b) HR normalized to baseline values in septic and control mice. (c) Mean arterial pressure (MAP) in septic and control mice. (d ) Normalized MAP in septic and control mice. (e) Stroke volume (SV) in septic and control mice. (f  ) Normalized SV in septic and control mice. (g) Cardiac output (CO) in septic and control mice. (h) Normalized CO in septic and control mice. HR was unchanged in control mice and increased significantly in septic mice (p < 0.01 by repeated measures ANOVA). MAP was unchanged in control mice and was significantly decreased in septic mice (p < 0.01). SV did not change significantly in either group. CO increased significantly (p < 0.01), largely owing to the increase in HR.

In septic mice (n = 20), was significantly decreased (102 ± 14 to 65 ± 19 mm Hg, p < 0.02), starting at 12 h after CLP. HR increased significantly, from 407 ± 70 to 524 ± 76 beats/min (p < 0.01), peaking between 24 and 30 h. Stroke volume (24 ± 4 µl at baseline) did not change significantly over time. Cardiac output increased significantly, from 11.6 ± 2.0 to 17.1 ± 1.5 ml/min (p < 0.01), largely because of the increase in HR (see Figure 2). End-diastolic dimension was 3.26 ± 0.45 mm at baseline, and did not change significantly over the course of the study (Table 1).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

We have established a murine peritonitis model with fluid resuscitation and antibiotic administration that replicates the mortality seen in patients with septic shock. We have, for the first time, performed continuous hemodynamic monitoring using echocardiography and continuous micromanometric pressure monitoring to demonstrate that the model reproduces the hyperdynamic cardiovascular response seen in clinical sepsis.

The current model uses CLP to produce intra-abdominal sepsis. The time course of mortality in this model is similar to that seen in clinical septic shock. Some models that lead to significant mortality within the first 12 h may not be entirely relevant to human sepsis (5). In the current model, all animals tested were bacteremic 6 h after the procedure. The onset of hypotension was slightly later, usually 12 to 18 h after the insult. Although the precise onset of bacteremia in this model cannot be determined, it should be recognized that this is the case with patients admitted with septic shock.

A particularly important feature of this murine model was development of a hyperdynamic state. The increased cardiac output resulted largely from tachycardia whereas stroke volume was relatively constant, a pattern similar to human sepsis and to endotoxin challenge of volunteers (20). Aggressive fluid resuscitation was implemented, because both in the clinical setting and in animal models, the development of a hyperdynamic state is dependent on volume repletion. In addition, patients with sepsis are immediately treated with fluid infusions, and virtually all of the studies of hemodynamics in septic patients have been performed after substantial fluid loading. In our model, fluid repletion was done using fixed volumes, in part because filling pressures are difficult to evaluate in mice, and in part because compliance abnormalities complicate the assessment of preload in sepsis (9); the maintenance of stroke volume and increases in cardiac output observed, along with the maintained end-diastolic dimensions, suggests that adequate volume was given. Further studies, however, will be needed in order to characterize myocardial performance in this model more thoroughly.

The employment of fluid resuscitation and administration of appropriate antibiotics reproduces the supportive therapy carried out in the clinical arena. This supportive regimen improved survival compared with untreated animals, a finding similar to that in the canine peritonitis model (9). We have recently reported use of CLP to compare survival in septic mice deficient in inducible nitric oxide synthase (iNOS) with septic wild-type mice (21). Despite improved microvascular reactivity and systemic blood pressure, survival in iNOS-deficient septic mice was increased only when mice were resuscitated with fluids and given antibiotics. This suggests that in the absence of fluid resuscitation, profound hypotension and decreased cardiac output may cause irreversible organ system failure. When the septic animals are resuscitated with fluids and treated with antibiotics, however, improved hemodynamics may allow time for the antibiotics to take effect (12).

Close examination of the survival curves reveals a small degree of early mortality (compared with controls) in animals treated with antibiotics alone. It is well recognized that killing of bacteria by antibiotics can result in release of substantial amounts of endotoxin and inflammatory cytokines (22); this may exacerbate the inflammatory response and cause hemodynamic deterioration, especially in the absence of fluid resuscitation. Additional studies will be needed to clarify the mechanisms involved.

Limitations

As with all animal models, this study does have certain limitations. Although efforts were made to keep the level of anesthesia as light as possible compatible with animal comfort, awake hemodynamic measurements were not possible in mice cannulated for measurement of blood pressure. Although ketamine has been reported to have less of a cardiodepressor effect than other regimens (23), in combination with other agents, ketamine anesthesia is known to decrease HR, blood pressure, and myocardial contractility in murine models (24, 25). The use of anesthesia does represent a significant limitation, yet it should be noted that control animals also received anesthesia. Furthermore, septic animals would be expected to be more susceptible to the effects of anesthesia than control animals. Thus, use of anesthesia might be expected to reduce cardiac output in septic mice and bias toward the null hypothesis. In addition, this is a peritonitis model with polymicrobial bacteremia; although the hemodynamic profile of sepsis does not appear to be dependent on whether the organism is gram-negative or gram-positive (26), its applicability to other forms of sepsis, such as gram-positive sepsis, is unproven. Finally, it is difficult to control the magnitude of the septic challenge in the CLP model.

Clinical Implications

A small animal model of sepsis that reproduces the vasodilation, hypotension, increased cardiac output, and response to treatment seen in patients with septic shock would be useful for studies of pathophysiology and treatment, but no current models replicate all of these features. We have established a murine model with fluid resuscitation and antibiotic administration and have shown that this model replicates the mortality seen in patients with septic shock. We have also demonstrated that the model reproduces the hyperdynamic state seen in clinical sepsis. Although no one model can replicate all of the features of human sepsis, we believe that this model may be useful for initial evaluation of new therapies for septic shock. In particular, the hemodynamic effects of those therapies may be tested. Finally, use of transgenic mice in this model should allow for mechanistic studies to dissect the pathogenic mediators of sepsis.