A Novel Porcine Model of Ventilator-associated Pneumonia Caused by Oropharyngeal Challenge with Pseudomonas aeruginosa

VENTILATOR-ASSOCIATED pneumonia (VAP) is a frequent respiratory complication in the intensive care unit,¹  associated with increased morbidity and healthcare costs.^2–4  VAP is commonly caused by aerobic Gram-negative bacteria, and among these pathogens Pseudomonas aeruginosa is the most prevalent.³ 

Animal models of VAP can be employed to study the pathophysiology of the disease and to evaluate the efficacy and safety of preventive and therapeutic interventions. Throughout the years, several animal models have been developed, which differ considerably in the underlying pulmonary injury, causative pathogens, methods for bacterial inoculation, and time to onset of the disease.⁵ 

In the ’80s, Johanson et al.^6–8  described a fundamental model of VAP in intubated baboons (Papio papio) with acute lung injury, and on mechanical ventilation (MV) up to 8 days. VAP was mostly caused by P. aeruginosa and Staphylococcus aureus.⁶  Nevertheless, new VAP models in pigs and sheep gradually replaced the use of primates in this field of investigation. In particular, Marquette et al.^9,10  developed a model of tracheobronchial stenosis and VAP in pigs on MV for 4 days. In a later report by the same group, mechanically ventilated pigs without tracheobronchial injury developed VAP in 90% of cases.¹¹  Importantly, in these porcine models, Pasteurella multocida and Streptococcus suis were the predominant causative organisms. Similar colonization by endogenous oropharyngeal pathogens was demonstrated in models developed at the National Institutes of Health in healthy sheep^12,13  and pigs.¹⁴  In these latter models, the animals were positioned with a tracheal orientation above horizontal to enhance microaspiration of oropharyngeal pathogens. Finally, a few investigators focused on the human etiology of VAP and induced severe pneumonia in healthy pigs through massive bronchial inoculation of Escherichia coli,^15,16 P. aeruginosa,¹⁷  and methicillin-resistant S. aureus.¹⁸ 

Importantly, the aforementioned ovine and porcine models do not reproduce the most prevalent human etiology,^9,11–14  or the primary pathogenic mechanism of the disease,^15–18 via aspiration of colonized oropharyngeal contents. Therefore, the primary aim of this study was to characterize a novel animal model of VAP, which reflected the key pathogenic mechanisms of the disease, in pigs challenged with P. aeruginosa into the oropharynx. Additionally, we assessed in a limited number of animals, the endogenous colonization of the respiratory tract and development of VAP during antibiotic treatment and in the absence of bacterial challenge.

The Institutional Review Board and Animal Ethics Committee of the University of Barcelona, Barcelona, Spain, approved the protocol. Animals were managed according to the National Institutes of Health guidelines for the use and care of animals.¹⁹ 

Seven female Large White–Landrace pigs (weight, 30.5 ± 1.8 kg; range, 27 to 33 kg) were premedicated with an intramuscular dose of 2 mg/kg of azaperone and induced with an intravenous loading dose of 8 mg/kg sodium thiopental. Pigs were then orotracheally intubated with a 7.5-mm I.D. endotracheal tube (ETT) comprising a high-volume low-pressure cuff (Hi-Lo; Covidien, Boulder, CO). The ratio between the ETT cuff outer (OD) diameter and tracheal internal diameter was 1.9 ± 0.8 (see Supplemental Digital Content 1, https://links.lww.com/ALN/B41, which describes methods to compute this ratio). Animals were connected to a SERVO-i mechanical ventilator (Maquet, Wayne, NJ). A continuous infusion of midazolam (1 to 2.5 mg kg⁻¹ h⁻¹) and fentanyl (5 to 10 μg kg⁻¹ h⁻¹) was administered. Boluses of 50 mg of intravenous sodium thiopental were administered as needed to optimize sedation. A loading dose of vecuronium (0.4 mg/kg), followed by a maintenance dose of 0.3 to 0.4 mg kg⁻¹ h⁻¹ was administered to produce muscle paralysis. We surgically cannulated the femoral artery as previously reported.²⁰  A 7-French Swan–Ganz catheter (Edwards Lifesciences, Irvine, CA) was inserted into the jugular vein. A no. 8 Foley catheter was introduced into the bladder through surgical minipelvectomy. After surgical preparation, the pigs were placed in the prone position and the custom-made surgical bed was oriented approximately 30 degrees in the anti-Trendelenburg position, to model the semirecumbent position in humans (fig. 1). A fluoroscopic image was taken to assess tracheal orientation. Fluid balance was maintained through infusion of lactated Ringer’s and 0.9% NaCl solutions. In order to prevent pneumonia caused by endogenous oropharyngeal flora, 1 g of ceftriaxone was administered intravenously 30 min before intubation, and then 50 mg/kg every 12 h for the entire duration of the study. Every 24 h, arterial and mixed venous blood gases, pulmonary mechanics, hemodynamics, urine output, and ventilatory settings were assessed. Additionally, complete blood count was obtained and reviewed.

Before bacterial instillation and every 24 h thereafter, we took arterial blood and tracheal secretions for P. aeruginosa cultures. Each animal was challenged immediately after surgical preparation and 4 h thereafter. Five milliliter of approximately 10⁷ to 10⁸ colony-forming unit (cfu)/ml of a log-phase culture of P. aeruginosa was slowly instilled, over 5 min, into the oropharynx of the animals. We employed a respiratory isolate of ceftriaxone-resistant P. aeruginosa, derived from P. aeruginosa ATCC 27853 and selected in a multistep process, under increasing concentrations of ceftriaxone, until reaching a maximum of 256 mg/l. Before each challenge, and up to 10 min thereafter, 5 cm H₂O of positive end-expiratory pressure (PEEP) was applied and the internal cuff pressure was increased to 40 cm H₂O to avoid rapid aspiration of pathogens.

Pigs were mechanically ventilated in volume control, tidal volume (V_T) of 10 ml/kg, Fio₂ 0.4, no PEEP, and respiratory rate adjusted to maintain Paco₂ between 40 and 45 mmHg. Inspiratory gases were conditioned through a Conchatherm III heated humidifier (Hudson RCI, Temecula, CA). Internal ETT cuff pressure was maintained at 28 cm H₂O through an external mechanical device.²¹  Open endotracheal suctioning via a 12-French sterile suction catheter (Oppo-Cath Suction Catheter; Pennine Healthcare Ltd., Derby, United Kingdom) was performed every 6 h, or when considered clinically indicated. During the procedure, the catheter was gently inserted into the ETT and slowly advanced up to the proximal trachea. Then the catheter was withdrawn applying intermittent suction and continuous twisting motion. On each tracheal suctioning, a score from 0 to 4 (none, scant, minimal, moderate, large) was used to estimate the quantity of aspirated secretions. Additionally, quality of airway secretions (normal or purulent) was recorded.

Airway pressure and respiratory flow rates were measured as previously reported.²⁰  Flow and pressure signals were recorded on a personal computer for subsequent analysis with dedicated software (Colligo; Elekton, Milan, Italy). The static elastance of the respiratory system, total inspiratory resistance, inspiratory airflow resistance, and inspiratory tissue resistance were calculated through the rapid occlusion method by using standard formulae.²⁰ 

Arterial and venous pressures were measured with disposable pressure transducers (TrueWave Pressure Transducer; Edwards Lifescience). Through the Swan–Ganz catheter pulmonary artery pressure, central venous pressure, pulmonary artery wedged pressure, core blood temperature, and cardiac output were measured. The systemic and pulmonary vascular resistances and venous admixture were calculated using standard formulae.⁹ 

Seventy-two hours after tracheal intubation (68 h after bacterial challenge), the animal was euthanized. The ETT was clamped and the internal cuff pressure increased to 40 cm H₂O. The animal was positioned supine and, under strict asepsis, the lungs were exposed, excised, and placed on sterile drapes. We took two samples from the most affected region of each of the five lobes (fig. 2) for histological and microbiological assessments. Lung histology was evaluated according to previously published methods and a six-point injury score.²²  Quantitative cultures of Gram-negative pathogens in MacConkey agar were performed using standard methods.¹³  Ultimately, bacteria were identified by mass spectrometry through a Microflex LT (Bruker Daltonics GmbH, Leipzig, Germany) and bacterial identification was performed using the MALDI BioTyper 2.0 software (BrukerDaltonics). We assessed the clinical pulmonary infection score at baseline and before autopsy as previously reported²³  (see Supplemental Digital Content 2, https://links.lww.com/ALN/B42, which depicts in table 1 the variables to calculate the modified clinical pulmonary infection score). VAP was confirmed according to a histological injury score of 3 or greater, associated with a quantitative P. aeruginosa lobar culture of 3 log cfu/g or greater.²⁴ 

We assessed in three female Large White–Landrace pigs (weight, 31.6 ± 1.5 kg; range, 30 to 33 kg) the endogenous colonization of the respiratory tract and development of VAP during antibiotic treatment and in the absence of P. aeruginosa oropharyngeal challenge. The animals were prepared and managed as described in the section Animals with oropharyngeal P. aeruginosa challenge. Upon autopsy, lung samples were plated in MacConkey and blood agar using standard microbiological methods.¹³  Bacteria were identified as described in section Animals with oropharyngeal P. aeruginosa challenge.

Data are reported as the mean ± SD, unless otherwise specified. One-way repeated measures ANOVA and Friedman test were employed to analyze effects of time on normally or not-normally distributed continuous variables. Post hoc multiple comparisons were analyzed using paired Student t tests and Wilcoxon signed ranks test, with Bonferroni correction. To assess the differences in bacterial concentrations and histologic scores between lobes, a restricted maximum likelihood analysis was used, which contained both fixed (lobar colonization) and random effects (subject). If the overall F test was significant (P ≤ 0.05), multiple comparisons between lobes were tested, using Bonferroni correction to control for the experiment-wise error rate. Categorical variables were analyzed using Fisher exact test. A two-sided P value less than 0.05 was considered statistically significant. All statistical analyses were performed using SAS software (version 9.2; SAS Institute, Cary, NC).

Six of the seven studied pigs completed the 72-h study; one pig was euthanized after 57 h due to severe respiratory and hemodynamic instability. Pigs were placed on a surgical bed oriented 27.3 ± 1.4 degrees above horizontal, resulting in a tracheal orientation of 13.4 ± 1.5 degrees.

Data are reported in table 1. Pigs were generally normothermic, although slightly hypothermic after surgery—the highest mean body temperature was 38.6° ± 0.9°C, after 48 h of tracheal intubation. Leukocyte count did not significantly increase. We aspirated increased amount of tracheal secretions over time, which appeared progressively infected; thus, after 72 h, 80% of the aspirated secretions were purulent. Throughout the study period, the hemodynamic parameters remained stable. The mean respiratory rate was 17.1 ± 1.2 breaths/min (range, 14 to 20 breaths/min) and V_T 304.8 ± 18.5 ml (range, 270 to 330 ml). Pao₂/Fio₂ decreased up to 309.2 ± 107.1, after 48 h of tracheal intubation (P = 0.004). The static elastance of the respiratory system worsened during the study reaching 41.5 ± 5.8 cm H₂O/l after 72 h. The clinical pulmonary infection score at baseline and before the autopsy was 0.85 ± 0.37 and 7.0 ± 1.82, respectively (P = 0.016) (table 2).

At the beginning of the study, tracheal secretions were not colonized by P. aeruginosa (fig. 3). Conversely, after bacterial challenge the concentration of P. aeruginosa in the tracheal secretions progressively increased, eventually reaching 6.39 ± 0.34 log cfu/ml at the end of the study. Bordetella bronchiseptica and E. coli were seldom found in tracheal secretions (table 2). After bacterial challenge, blood cultures were positive in 4 of 40 samples.

In lobar cultures, P. aeruginosa was the only isolated pathogen. The right medium lobe (RML) presented the highest P. aeruginosa burden (4.14 ± 1.01 log cfu/g; fig. 4A). Likewise, histological assessment showed, on average, a pattern of confluent pneumonia in the aforementioned lobe (histologic injury score: 4.14 ± 0.69; figs. 4B and 5). Accordingly, VAP was confirmed in six of seven pigs. In one pig, abscessed histological pneumonia was found in the RML, nevertheless VAP was not ultimately confirmed due to a P. aeruginosa tissue concentration of 2.86 log cfu/g. VAP developed in the RML in 55.6% of the cases; in the right lower lobe in 22.2%; in the left upper or lower lobes in 11.1%; and never in the right upper lobe (P = 0.032). Multilobar infection was found in three of the six pigs that developed VAP. Among these pigs, the RML was always affected with an associated coinfection of the right lower lobe in two cases and left upper lobe in one case.

All three animals completed the 72-h study. As shown in figure 6, diverse airway bacterial communities colonized respiratory secretions, at different concentrations. The most common Gram-negative microbes in tracheal secretions were E. coli (25% of isolated pathogens) and B. bronchiseptica (8.3% of isolated pathogens); whereas, S.aureus (19.4% of isolated pathogens) and S. suis (19.4% of isolated pathogens) were the most common Gram-positive bacteria (fig. 7). On autopsy, E. coli and B. bronchiseptica were found in 13 and 20% of the lung tissue samples, respectively. Only one pig developed B. bronchiseptica VAP in the RML.

In this study, we developed an animal model of VAP, in which the main pathogenic mechanism is through pulmonary aspiration of oropharyngeal secretions colonized by P. aeruginosa. The main results of our study are the following: (1) P. aeruginosa, instilled into the oropharynx, rapidly translocate into the airways and colonize tracheal secretions; (2) the first signs of infection develop approximately after 48 h of MV; (3) the right medium and lower lobes are the most commonly affected pulmonary sites, which strongly suggests a gravity-dependent dissemination of the infection; (4) ultimately, VAP is not associated with severe hemodynamic impairment.

Johanson et al.^6–8,25,26  developed the first animal model of VAP in primates, which has been crucial in understanding the pathophysiology of the disease and developing preventive strategies. This model closely reflected the human etiology, and the main pathogenic mechanism via aspiration across the ETT cuff of colonized oropharyngeal secretions. Nevertheless, the main drawbacks of the model were the scarce animal availability and the high costs of the animals and dedicated settings. Additionally, strict legislations and policies related to the use of research primates make this model impractical nowadays.

In the last two decades, investigators developed alternative VAP models in pigs and sheep. Only one model of VAP associated with tracheobronchial stenosis has been described.^9,10  Interestingly, after injury of the airways, a high incidence of VAP was unintentionally found, mainly caused by endogenous bacteria. This model emphasized the roles of mucus clearance impairment and MV in the development of respiratory infections. The use of this model was marginal, but provided essential insights to develop models in healthy pigs and sheep. Indeed, shortly thereafter Marquette et al.¹¹  demonstrated that the majority of healthy pigs on invasive mechanically ventilation developed VAP within 48 to 72 h. VAP was mainly caused by endogenous oropharyngeal pathogens, such as P. multocida and S. suis. Zanella et al.¹⁴  found similar results in pigs positioned as in our current study and mechanically ventilated for 72 h. Although these models mimic aspiration of endogenous oropharyngeal pathogens, characteristic human bacteriology is not reproduced. Also, bacterial challenge is not controlled, thus methodological reproducibility is limited and significant discrepancies could occur between subjects or treatment groups. In our model, pulmonary colonization by the most prevalent endogenous pathogens was proficiently controlled with ceftriaxone. Indeed, in our preliminary studies Pasteurella and Streptococcus species were never found in the lung tissue and VAP hardly developed as previously reported.¹¹  Also, after bacterial challenge with P.aeruginosa, we seldom found endogenous Gram-negative pathogens in tracheal secretions, that is, E. coli and B. bronchiseptica. It is likely that the vast inoculum of P. aeruginosa could have altered the dynamics of the oropharyngeal and pulmonary microbiota and favored an overgrowth of this exogenous pathogen.

Several reports from the National Institutes of Health^12,13  depicted a VAP model in sheep positioned with a tracheal orientation above horizontal. Similar to the aforementioned models, VAP was caused by endogenous oropharyngeal pathogens, the time to VAP onset was faster than the pig model (approximately 24 h),¹²  and the VAP incidence approximated 100%. Nevertheless, sheep are ruminants and opposed to pigs or humans, the large production of oropharyngeal secretions could overestimate risks of pulmonary aspiration across the ETT cuff and lung infections.

Finally, Goldstein et al.,^15,16  Luna et al.,¹⁷  and Martinez-Olondris et al.¹⁸  developed VAP models via intrabronchial instillation of nosocomial pathogens. As a result, severe bilateral pneumonia consistently developed after 12 to 24 h, and it was likely intensified by the MV.²⁷  These models reproduced human etiology, and they have been used to study local and systemic inflammatory responses,²⁸  as well as effectiveness of antibiotic and novel treatments.^17,29  Yet, in humans, the oropharynx is the primary source of infection³⁰  and the dependent lung segments are mostly affected.²⁴  Therefore, these models do not mimic pulmonary aspiration of oropharyngeal pathogens, and the characteristic intrapulmonary distribution of the infection. In our model, the oropharynx was initially colonized, and in 78% of the cases, pneumonia developed in the right medium and lower lobes. A possible explanation of these findings is that pigs were prone, and colonized oropharyngeal fluids could have easily gravitated into the main right bronchus, which is wider and with a less acute angulation in comparison with the left one (fig. 2), and into the RML, which arises from the ventrolateral side of the right bronchus. Additionally, these findings are in agreement with previous postmortem human studies on VAP, which demonstrated a greater incidence of pneumonia in the dependent lung segments and a gravity-dependent dissemination of the infection.²⁴ 

As corroborated by the gradual colonization of respiratory secretions (fig. 3), a continuous microaspiration of colonized oropharyngeal secretions across the cuff likely occurred in our model, regardless of the tight control of the internal cuff pressure. We used an ETT comprising a polyvinyl chloride high-volume low-pressure cuff (Hi-Lo). This is the most commonly used cuff design in critical care, although several in vitro^31–34  and in vivo studies³⁵  have reported limited sealing efficacy. Importantly, we recently corroborated substantial microaspiration across the Hi-Lo cuff while it was inflated at an internal cuff pressure of 30 cm H₂O.³²  When the cuff OD is larger than the tracheal diameter,³¹  folds form along the cuff surface and fluids leak through these folds. In our model, the ratio between the ETT cuff OD and tracheal diameter was 1.9; accordingly, several folds may have formed. Additionally, in our model, we did not apply PEEP because laboratory^36,37  and clinical studies³⁸  have shown reduction of lung infections with its use. As a result, microaspiration of oropharyngeal contents was likely enhanced.^33,34,39 

Our study has several additional features that merit consideration. Previous reports¹¹  used a quantitative lobar culture of 4 log cfu/g or greater for the diagnosis of pneumonia. Conversely, we used a quantitative P. aeruginosa lobar culture of 3 log cfu/g or greater as a threshold for the diagnosis of VAP, because we administered ceftriaxone that may affect cultures of lower airway samplings and decrease diagnostic sensitivity.^40–42  Importantly, the development of pneumonia was associated with clinical signs of infection and the worsening of respiratory parameters, without hemodynamic compromise. These results are consistent with previous reports,^11,17  and underline a compartmentalized rather than systemic inflammatory response of VAP. Importantly, we used a tidal volume of 10 ml/kg and no PEEP. A few reports in pigs demonstrated that MV without PEEP^27,36  and tidal volumes as low as 6 to 8 ml/kg⁴³  could potentially result in ventilator-induced lung injury. Therefore, although in our model the plateau airway pressure was always below 13 cm H₂O (table 2), pulmonary injury caused by the mechanical ventilator could have played a confounding additional role on lung colonization and VAP.

Our VAP model has several advantages: reproducibility, ease of induction, it reflects human etiology, and has a consistent time course of infection. Indeed, the first clinical signs of pneumonia appeared approximately after 48 h of MV and pneumonia was consistently confirmed after 72 h. An additional strength of our study was the assessment of the endogenous respiratory colonization in a limited number of animals treated with ceftriaxone and without bacterial challenge. Our model could conveniently be employed to study pathophysiologic mechanisms of the disease and diagnostic strategies and to test the effectiveness and safety of novel strategies aimed at preventing translocation of oropharyngeal pathogens into the lungs. ETTs comprising novel leak-proof cuffs or aspiration of subglottic secretions, as well as new agents for oropharyngeal decontamination, could be tested in this model. Conversely, due to the time required for VAP development, this model is not feasible for pharmacological studies.

A few limitations of our experimental model need to be discussed. First, we studied a small population; nevertheless, our findings were consistent and VAP developed in 85% of the cases. Second, our results may not be applicable in subjects intubated with different ETTs. In particular, ETTs comprising narrower OD cuffs may reduce aspiration of colonized oropharyngeal secretions. Third, we did not assess biofilm formation within the ETT. Yet, in our model, translocation of pathogens from within the ETT could have played an additional role in the development of VAP.⁴⁴  Fourth, the animals were kept prone whereas human patients are normally kept supine. This prone position was necessary because pigs develop diffuse areas of atelectasis when they are supine for a prolonged period of time. Fifth, animals were deeply sedated and paralyzed. In critical settings, a deep sedation/paralysis is only applied to the most critical patients, and during the very first period of MV; interestingly, these patients also present the greatest risk for VAP development.⁴⁵  Sixth, pigs were healthy at the time of intubation; hence innate host’s respiratory defenses could have been more effective in comparison with that of severely ill patients.

In conclusion, we demonstrated that VAP consistently developed within 72 h after oropharyngeal P. aeruginosa challenge in tracheally intubated and mechanically ventilated pigs, positioned in anti-Trendelenburg. The main advantage of this experimental model is that it closely reflects primary pathogenic mechanisms of human VAP and bacteriology. Thus, it could be used to study the pathophysiology of the disease, including diagnostic and preventive strategies.

Support was provided by the Institut d’Investigacions Biomèdiques August Pi iSunyer, Barcelona, Spain; Ministerio de Ciencia e Innovación (PS09/01249), Madrid, Spain; 2009 Alain Harf Award on Applied Respiratory Physiol ogy, European Society of Intensive Care Medicine, Brussels, Belgium; Fundació Catalana de Pneumologia, Barcelona, Spain; Sociedad Española de Neumología y Cirugía Torácica, Madrid, Spain; and Centro de Investigación Biomedica En Red-Enfermedades Respiratorias, Mallorca, Spain.

The authors declare no competing interests.