Main

Streptococcus pneumoniae is the most common causative pathogen of community-acquired pneumonia (CAP), inducing highly severe infections and often resulting in unfavorable outcomes. Approximately one-third of hospitalized patients with pneumococcal pneumonia require admission to the intensive care unit (ICU)1.

In the last decade, ventilatory management and therapy of these patients have improved, yet pneumococcal CAP remains associated with high mortality1,2, particularly in patients with septic shock or those who require mechanical ventilation3. In more recent years, macrolide resistance in S. pneumoniae has increased substantially. In several regions worldwide, macrolide-resistant pneumococci have become even more common than penicillin-resistant S. pneumoniae4, particularly after the introduction of 7-valent or 13-valent pneumococcal conjugate vaccines and the rapid decline of penicillin-resistant strains5.

Among several S. pneumoniae virulence factors, the capsule is critical for the development of infection. Based on variations of the capsule molecular structure6, a total of 93 S. pneumoniae serotypes have been characterized; however, only roughly 20 serotypes can cause human infections. Per epidemiology data, serotypes 1, 4, 5, 7F, 8, 12F, 14, 18C and 19A are more likely to cause invasive pneumococcal disease, as defined by isolation of S. pneumoniae from a normally sterile body site (e.g., blood)6. These serotypes comprise the most recent pneumococcal conjugate vaccine7. Furthermore, the Spanish National Institute of Health has shown that the most common invasive serotypes in adults included 3, 19A and 18.

Given the pivotal role of S. pneumoniae in CAP9, several animal models have been developed to better understand the pathogenesis of pneumococcal disease and evaluate novel, preventive and therapeutic strategies. Specifically, mice, rats and rabbits10 have gained increasing interest for studying S. pneumoniae disease. A main limitation of these small animal models is the difficulty in reproducing the severity of human disease or prolonged mechanical ventilation. Severe infections kill small animals quickly and, as a result, prevent investigators from performing a comprehensive assessment of disease dynamics. To date, no reliable model of severe pneumococcal pneumonia reproducing intensive care settings is available. However, in the last decade, our group has developed several other models of respiratory infections caused by nosocomial pathogens, that is, Pseudomonas aeruginosa and methicillin-resistant Staphylococcus aureus11,12.

We now describe a pioneering model of pneumococcal pneumonia caused by penicillin- and macrolide-resistant S. pneumoniae serotype 19A in mechanically ventilated pigs. We fully characterize the infection, focusing on the dynamics of clinical variables, pulmonary mechanics, hemodynamics, inflammatory markers and microbiology studies. Furthermore, using this model and as a proof of concept, we aim to describe the pharmacokinetics of ceftriaxone and levofloxacin in pigs, two first-line antibiotics for pneumococcal pneumonia.

Results

All pigs not challenged with S. pneumoniae completed the study (n = 4/4 pigs, 100%). Conversely, 76-h survival in pigs infected with S. pneumoniae was 70% (n = 7/10 pigs), because two pigs were euthanized at 34 and 48 h, respectively, for severe respiratory/hemodynamic instability, and one pig was euthanized at 55 h due to severe hemodynamic instability.

Tracheal secretions, bronchoalveolar lavage and blood cultures

Tracheal secretions were collected at baseline, 24, 48 and 72 h in both the preliminary study (control pigs) and the main study (pigs infected with S. pneumonia) (Fig. 1). In control pigs, tracheal secretions at baseline were obtained in only one pig, and those were colonized by Bordetella bronchiseptica. At 24, 48 and 72 h, all tracheal secretions in control pigs were colonized by at least one pathogen (n = 12 samples) (Fig. 2). In infected pigs, five samples, which were colonized by endogenous flora, were obtained at baseline. By contrast, at 24 and 48 h, all tracheal aspirates in infected pigs were colonized by S. pneumoniae at a concentration >106 c.f.u./ml (n = 19 samples); whereas, at 72 h, six out of seven (86%) tracheal aspirates were colonized by S. pneumoniae at a concentration >106 c.f.u./ml (Fig. 2). The mean S. pneumoniae concentration across study measures (i.e., from 24 to 72 h) was 6.31 × 107 c.f.u./ml, vastly overtaking all other pathogen concentrations. Importantly, at 24 h, pneumococcal pneumonia was clinically confirmed in all infected pigs by ≥106 c.f.u./ml in tracheal secretions, plus at least one clinical feature. Pathogens in tracheal secretions varied between infected and control pigs, specifically at 24 h (Fig. 2). As expected, S. pneumoniae was consistently the most frequent pathogen in infected pigs at each assessment time point. Conversely, Streptococcus suis, S. aureus and B. bronchiseptica were the most frequently isolated pathogens in control pigs.

Fig. 1: Study design of preliminary, main and pharmacokinetic studies.
figure 1

The timeline of experimental events is displayed. The number of pigs used for each study is reported. The preliminary and main studies were designed to compare data from microbiological assessments and other measurements between control pigs and pigs infected with S. pneumoniae serotype 19 A. Of note, ceftriaxone pharmacokinetic studies were planned to achieve an epithelial lining fluid concentration above S. pneumoniae minimum inhibitory concentration for ≥60% of the time within 24 h in at least two consecutive pigs. For levofloxacin, we aimed for a ratio of area under the curve to minimum inhibitory concentration of bound levofloxacin >40.

Full size image
Fig. 2: Bacterial colonization of tracheal secretions.
figure 2

On the upper left side, tracheal secretion bacterial diversity during the main study is shown for S. pneumoniae–infected pigs. On the lower left side, tracheal secretion bacterial diversity during the preliminary study is shown for control pigs. In the right section, sequential assessment of quantitative bacterial concentration in tracheal secretions in infected (upper plot) and control (lower plot) pigs is shown (mean ± s.e.m. bars). In infected pigs, the concentration of S. pneumoniae (red bars) in tracheal secretion increased after bacterial challenge (P < 0.001), whereas the concentrations of other pathogens (gray bars) did not vary between study time points (P = 0.53). In controls, bacterial concentration did not vary during study times (P = 0.59). Data are reported as percentage of isolated bacteria among total tracheal aspirate samples as a pool of pigs combined. The number of included pigs, analyzed samples, positive cultures and isolated species are indicated for each group and time of assessment. The percentage of colonization between infected and control group variables was analyzed by using Fisher’s exact test and the Cochran–Mantel–Haenszel test. Restricted maximum likelihood (REML) analysis based on a repeated measures approach with time of assessment as factors was performed. P values—corrected through Bonferroni methods—for comparisons between bacterial burden per study time are reported. B. bronchiseptica, Bordetella bronchiseptica; E. cloacae, Enterobacter cloacae; S. aureus, Staphylococcus aureus; S. dysgalactiae, Streptococcus dysgalactiae; S. epidermidis, Staphylococcus epidermidis; S. hycus, Staphylococcus hycus; S. pneumoniae, Streptococcus pneumoniae; S. suis, Streptococcus suis.

Full size image

Bronchoalveolar lavage (BAL) was performed to quantify bacterial burden at baseline and at 24 and 72 h. In control pigs, colonization occurred in 50% (n = 2 out of 4) of BAL samples at baseline. However, at 24 and 72 h, colonization occurred in 100% of BAL samples (n = 4 samples per time point) (Fig. 3). In infected pigs, 6 out of 10 (60%) BAL samples were colonized at baseline. After bacterial challenge (at the 24-h time point), the S. pneumoniae concentration significantly increased up to 1.20 × 106 c.f.u./ml. All infected pigs presented with an S. pneumoniae concentration >103 c.f.u./ml in BAL fluids at 24 h (n = 10 samples); however, at the end of the study, only 86% of these pigs presented with the aforementioned S. pneumoniae concentration (n = 6 out of 7 samples). Pathogens varied between infected or control pigs at 72 h (P = 0.04) (Fig. 3). Concomitant pathogens progressively increased in infected pigs throughout the study time (P = 0.02), yet not notably in control pigs (P = 0.08).

Fig. 3: Bacterial colonization of BAL.
figure 3

The upper left section depicts BAL bacterial diversity during the main study in infected pigs. On the lower left side, BAL bacterial diversity during the preliminary study is shown for control pigs. In the right section, sequential assessment of quantitative bacterial concentration in BAL fluids in infected (upper plot) and control (lower plot) pigs is shown (mean ± s.e.m. bars). In infected pigs, the concentration of S. pneumoniae (red bars) in BAL fluids increased after bacterial challenge (P < 0.001), whereas for other pathogen concentrations (gray bars), the burden increased progressively between study times (P = 0.02), specifically between baseline and 24 h (P = 0.02). In controls, bacterial concentration did not vary during study times (P = 0.08). Data are reported as percentage of isolated bacteria among total BAL samples as a pool of pigs combined. The number of included pigs, analyzed samples, positive cultures and isolated species are indicated for each group and time of assessment. The percentage of colonization between infected and control group variables was analyzed by using Fisher’s exact test and the Cochran–Mantel–Haenszel test. REML analysis based on a repeated measures approach with time of assessment as factors was performed. P values—corrected through Bonferroni methods—for comparisons between bacterial burden per study time are reported. A. baumanii, Acinetobacter baumannii; R. nasimurium, Rothia nasimurium.

Full size image

Finally, of the 35 assayed blood samples from infected pigs (10 samples at baseline, 10 at 24 h, 8 at 48 h and 7 at 72 h), 8 (22.8%) tested positive for S. pneumoniae. S. pneumoniae bacteremia was found in five (50%) infected pigs during the experiment; specifically, one pig tested positive at 24 and 48 h, two pigs at 48 h and two pigs at 48 and 72 h. Conversely, controls never developed bacteremia (n = 0 out of 16 samples) (P = 0.045 versus infected pigs).

Pulmonary tissue burden

Macroscopic findings were observed in lungs upon necropsy (Supplementary Fig. 1). Histological features of infected pigs corroborated confluent pneumonia in all lung tissue samples (Fig. 4). S. pneumoniae was the most commonly isolated pulmonary pathogen (n = 43 out of 50 samples; 86%) in the infected group. The mean ± s.e.m. pulmonary S. pneumoniae concentration among the 10 infected pigs was 1.26 × 105 ± 2 × 102 c.f.u./g, and it did not differ among the 50 sampled lobes (i.e., five lobes/pig; P = 0.92) (Fig. 5). Fifteen of fifty (30%) analyzed lobes in infected pigs presented concomitant colonization by other pathogens. Nevertheless, the mean ± s.e.m. concentration of these other pathogens among 50 sampled lobes was 2-log lower than S. pneumoniae (1.26 × 103 ± 102 c.f.u./g), and it did not differ among the sampled lobes (P = 0.086). The associated median pulmonary histological score was 4 (interquartile range 4–4) (n = 50 samples). Thus, pneumococcal pneumonia was microbiologically/histologically confirmed in 8 of 10 infected pigs (80%). In the remaining two pigs, pathology studies confirmed confluent pneumonia (i.e., a score ≥3); however, the S. pneumoniae concentration was below the microbiological threshold (i.e., <103 c.f.u./g) to ultimately confirm pneumonia. Pneumococcal pneumonia was always multi-lobar. In control pigs, S. suis and S. aureus were the most commonly encountered pathogens. The mean ± s.e.m. pulmonary concentration of pathogens in control pigs was 7.94 ± 20 c.f.u./g and slightly differed between the sampled lobes (P = 0.05).

Fig. 4: Lung tissue microscopic findings from infected pigs (nos. 137, 145 and 179) and a control pig (no. 295).
figure 4

Upper figures show ×40 magnification, and lower figures show ×200 magnification. Confluent pneumonia was the most common histological pattern in infected pigs: polymorphonuclear infiltrate is identifiable within contiguous secondary lobes, separated by a septum (white arrows indicate the septum margins). Findings in control pigs included mild bronchiolitis, composed of histocytes with some polymorphs, and interlobular septal edema.

Full size image
Fig. 5: Pulmonary microbiological burden.
figure 5

The left section depicts pulmonary bacterial variety. In the right section, box plots display pulmonary concentration per lobes. Horizontal bars represent the median, boxes represent the interquartile range, whiskers represent the range and plus signs denote the means. In the upper graph, red box plots depict lobar S. pneumoniae concentration data, whereas gray ones depict other concomitant pathogens per lobes among infected pigs. Neither pulmonary S. pneumoniae concentration (P = 0.92) nor concomitant pathogen concentrations (P = 0.09) differed between pulmonary lobes. In the lower right section, the pulmonary concentration of other pathogens is shown for the control group. Significant differences were found between the right upper lobe (RUL) and the left upper lobe (LUL) and left lower lobe (LLL) (P = 0.05). Data are reported as percentage of isolated bacteria among lung tissue samples as a pool of pigs combined. The number of included pigs, analyzed pulmonary lobes, positive cultures and isolated species are indicated for each group and time of assessment. REML analysis based on a repeated measures approach with lung lobes as factors was performed. P values—corrected through Bonferroni methods—for comparisons between bacterial burden per lung lobes are reported. S. castoreus, Streptococcus castoreus. RML, right middle lobe; RLL, right lower lobe.

Full size image

Clinical findings, lung function and hemodynamics

Clinical and hemodynamic data are reported in Tables 1 and 2. Pneumonia development was not associated with specific changes in body temperature or white blood cells. Conversely, a significant drop in platelet count was observed in infected pigs. A significant increase in prothrombin time was another sign of systemic infection. In addition, creatinine was higher in control than in infected pigs at baseline, albeit without clinical relevance. Because of a significant decrease in systemic vascular resistance in infected pigs, mean ± s.e.m. arterial pressure decreased to 66.9 ± 6.8 mmHg at 24 h, whereas cardiac output significantly increased. In addition, mean pulmonary arterial pressure was significantly different between infected and control pigs. Norepinephrine was administered in 9 of 10 challenged pigs at some point during the experiment to sustain hemodynamic stability, whereas control pigs did not need vasopressor support during the entire experiment (P = 0.003). Venous saturation was kept within the normal range throughout the study in both infected and control groups. Finally, gas exchanges worsened in infected pigs, specifically at 24 h after bacterial challenge (Supplementary Fig. 2). A progressive and significant increase in ventilatory support was therefore necessary to sustain pulmonary function in those pigs (Supplementary Table 1).

Table 1 Clinical data
Full size table
Table 2 Hemodynamic parameters
Full size table

Inflammatory biomarkers

Systemic inflammatory biomarkers significantly increased at 24 h in infected pigs compared with baseline due to S. pneumoniae challenge but decreased thereafter (Supplementary Fig. 3). There was a steady increase in tumor necrosis factor α throughout the study in the serum of infected pigs. In BAL fluids of infected pigs, IL-1β, IL-6 and IL-8 concentrations significantly changed over time, reaching peak levels after 24 h of bacterial challenge and returning to baseline levels at 72 h (Supplementary Fig. 4).

Pharmacokinetic studies

A study of pharmacokinetics of ceftriaxone and levofloxacin was also performed (Fig. 6 and Supplementary Table 2). Six pigs were necessary to fulfill the expected goals of maintaining an unbound ceftriaxone epithelial lining fluid (ELF) concentration above S. pneumoniae minimum inhibitory concentration (MIC) for ≥60% of the time within 24 h. Increasing doses of ceftriaxone—50, 100 and 150 mg/kg—were therefore tested (i.e., two pigs per dose). Conversely, for levofloxacin, the first tested dose of 10 mg/kg (n = 2 pigs) was sufficient to achieve a ratio of area under the curve to MIC of bound levofloxacin >40.

Fig. 6: Epithelial lining fluid (ELF) pharmacokinetics of ceftriaxone and levofloxacin.
figure 6

In the upper portion of the figure, ceftriaxone (CTX) pharmacokinetics are shown. A dose of 150 mg/kg of ceftriaxone was necessary to achieve a concentration of unbound ceftriaxone concentration above S. pneumoniae minimum inhibitory concentration (MIC) for ≥60% of the time between daily dosing (y-axis dotted line, ‘60% T>MIC’). Levofloxacin (LVX) pharmacokinetics are shown in the lower portion of the figure. A dose of 10 mg/kg was sufficient to achieve the ratio between the area under the curve for 0–24 h and S. pneumoniae MIC of ≥40 (y-axis dotted line).

Full size image

Discussion

We developed a novel model of macrolide-resistant, multi-lobar pneumococcal pneumonia in mechanically ventilated pigs. Characterization of the model includes S. pneumoniae bacteremia in 50% of pigs and an important systemic and pulmonary release of inflammatory markers that causes septic shock and pulmonary/hemodynamic derangement. The survival rate at 76 h was 70% in infected pigs, allowing for a relatively prolonged follow-up after the development of infection.

To the best of our knowledge, this is the first model of penicillin- and macrolide-resistant pneumococcal pneumonia in animals ventilated for days, with severe hemodynamic impairment and requiring the use of vasoactive drugs. In the late 1970s, the first ground-breaking model of pneumococcal pneumonia was developed in anesthetized rats, via instillation of S. pneumoniae serotype 3 into the left main bronchus10. The animals were spontaneously breathing, and infection even spread into the right lung throughout the study, which lasted up to 132 h. A detailed description of pneumococcal pneumonia in mice, rats and rabbits goes beyond the scope of this article, given previous comprehensive appraisals of animal models of S. pneumoniae infections10. Nevertheless, to determine the limitations of our novel model and identify possible refinements, taking a closer look at previously developed models of multi-lobar pneumococcal pneumonia that cause significant hemodynamic compromise is essential. The mouse models presenting with the most severe pulmonary infections were developed through intra-tracheal instillation of S. pneumoniae. Azoulay-Dupuis and colleagues13 developed a model of severe pneumococcal pneumonia in Swiss mice via tracheal cannulation and instillation of S. pneumoniae type 3 with microliter syringes. In another model, Iwasaki and collaborators14 lightly sedated a mouse and introduced an otoscope into the mouse’s oropharynx for direct visualization of the trachea and bacterial challenge into the airways.

In comparison with our model, though, several dissimilarities are evident. First, neither mice nor rats underwent mechanical ventilation, and circulatory shock was not supported by administering vasoactive drugs. Second, severe pneumonia developed in just 8–12 h in our swine model, whereas mice and rats required days to develop lethal severity. Furthermore, our model used a clinically relevant, penicillin- and macrolide-resistant S. pneumoniae serotype (19 A). There are well-known difficulties in causing pneumonia in rodents with penicillin-resistant S. pneumoniae, and scientists have ultimately used immunocompromised or young rodents to develop infections with such a strain10.

Alternatively, rabbits can be used to study pneumonia by penicillin-resistant S. pneumoniae in immune-competent animals15,16. For example, Piroth and collaborators16 undertook an extremely high intrabronchial challenge approach of 1010 c.f.u./ml of penicillin-resistant S. pneumoniae serotype 9V. After 72 h, the authors reported a survival rate of 50% and progression from lobar to multi-lobar pneumonia. Yet, even then, rabbits did not receive mechanical ventilation or care that would be provided to patients in ICU.

With respect to models of pneumococcal pneumonia in larger animals, Kraft et al. developed a pneumococcal pneumonia model in adult baboons (Papio cynocephalus) that were infected with escalating doses of S. pneumoniae serotype 19A17. Although this model is characterized by a good association between dose inoculation and cytokine response, the baboons were not mechanically ventilated and were diagnosed by using traditional methods, and no microbiological data of lung burden are available. Similarly, Reyes and collaborators18 described an innovative model in adult baboons that underwent bacterial challenge with 109 c.f.u./ml of S. pneumoniae serotype 4 in the right middle lobe. The baboons developed severe pneumonia and bacteremia but not septic shock; no mechanical ventilation was necessary. Furthermore, it is uncertain whether these nonhuman primates developed concomitant colonization by endogenous pathogens. Lim et al.19 also developed an S. pneumoniae model in pigs by challenging each pulmonary lower lobe with 108 c.f.u./ml. Unlike the previous studies, these experiments lasted only 6 h, and no microbiological information was reported.

In our study, we aimed at creating a highly severe model of pneumococcal pneumonia, similar to that observed in patients requiring intensive care. Indeed, between 1.2% and10.0% of patients requiring hospital admission for CAP caused by S. pneumoniae will require ICU admission, and 28-d mortality in patients admitted to the ICU is 17%. This figure increases to 25% in patients who require invasive mechanical ventilation and 40% in patients who develop septic shock20. The ideal treatment for severe pneumococcal pneumonia is limited to antibiotic therapy and potential ventilatory support until pulmonary function has recovered. Furthermore, current therapeutic options for severe pneumococcal pneumonia still lack an evidence-based guideline for best practice. None of the previously described models underwent mechanical ventilation for >6 h. Therefore, our reliable model of pneumococcal pneumonia may serve as a means to obtain experimental results that would be likely unachievable in clinical settings, particularly because this animal model can be used to harvest various tissues post mortem21.

We used prophylactic antibiotic therapy with doxycycline and aztreonam to avoid colonization by species-specific endogenous pathogens. As clearly demonstrated by our results obtained from control pigs, these antibiotics adequately prevented infection, and all pigs survived the 76-h experiment. Nevertheless, in pigs with pneumococcal pneumonia, the concentration of endogenous flora increased, although always marginally when compared to S. pneumoniae. Prophylactic antibiotics were not effective enough to prevent infection of an already injured lung by S. pneumoniae. Endogenous flora probably occurs in mechanically ventilated patients, as suggested by post-mortem studies22. The interplay between endogenous flora and S. pneumoniae cannot be fully described with our current study design. Lung damage caused by S. pneumoniae may facilitate the colonization of the lower airway by endogenous flora. In this context, studies have shown that patients with greater lung damage (i.e., patients with acute respiratory distress syndrome) have a higher risk of infection by more than one pathogen than patients with lower pulmonary insult23. Significant differences between baseline and positive BAL cultures 24 h after bacterial challenge corroborate this hypothesis. Therefore, it is possible that the resulting, life-threatening pulmonary insult performed by two consecutive bacterial challenges significantly thwarted respiratory defenses and facilitated endogenous colonization in our model. Future studies should therefore consider the underlying pathophysiological mechanisms for concurrent bacterial growth. Furthermore, in view of the high S. aureus colonization in infected pigs, our model may set the groundwork to test breakthrough fifth generation antibiotics active against both pathogens (e.g., ceftobiprole24 or ceftaroline25) and potentially detect their effect on both S. pneumoniae and S. aureus burden.

An additional strength of our model is the use of an S. pneumoniae strain resistant to penicillin and macrolides. Our model contributes to providing insight into macrolide resistance in CAP. Previous investigations have reported macrolide-resistant S. pneumoniae rates to be between 15% and 35%26. These figures are alarming, given that macrolides are first-line antibiotic treatment in patients with CAP27. In addition, results related to the effects of macrolide resistance on patient outcomes have been conflicting. Our model, however, could help explore these controversial matters or even address potential immunomodulatory benefits associated with macrolides28 while even preventing the bactericidal effects of such drugs. Finally, as in previous models29, our model could allow investigators to easily study the deleterious cardiac effects and potential remodeling caused by severe pneumococcal pneumonia.

As for other noteworthy clinical features of our model, we observed hyperdynamic septic shock at 24 h, characterized by a fall in mean arterial pressure (MAP) and systemic vascular resistance and an increase in cardiac output. This is in line with the clinical status of patients admitted to the ICU with pneumococcal pneumonia, who often require the use of vasopressors. We also found sepsis-induced coagulopathy, as corroborated by prolonged prothrombin time and thrombocytopenia. However, possibly because of prompt hemodynamic support during the development of infection, no other clinical signs of organ failure were evident. In addition, we confirmed systemic inflammatory response on the basis of a considerable upsurge of inflammatory markers IL-1ß, IL-6 and IL-8. This model could therefore be useful in the testing of novel strategies aimed at modulating inflammation during the early disease stages. Moreover, novel antimicrobial agents with Gram-positive or broader-spectrum features (e.g., ceftaroline and ceftobiprole or omadacycline, respectively) could be tested in future studies.

Finally, another major strength of our study is the assessment of ELF concentrations of ceftriaxone and levofloxacin. These analyses serve as proof of concept for the potential use of this novel model. In particular, as detailed in Methods, we aimed for specific endpoints commonly applied in clinical settings to ensure bactericidal efficacy, as is normally done in the first phases of clinical trials. Interestingly, in comparison with dosages in humans, a fourfold increase in ceftriaxone dose was necessary to achieve a concentration of free antibiotic above S. pneumoniae MIC for 60% of the time. This was probably related to the substantial severity and resulting hyperdynamic septic shock of our model30, largely augmenting the distribution volume of ceftriaxone. These data also highlight the potential risks of under-dosing in critically ill patients and the need for new reliable clinical data to validate ELF concentrations. In addition, these findings are in line with previously reported lower pulmonary concentrations and efficacy attained with standard daily dosing of 2 g of ceftriaxone in the treatment of S. aureus pneumonia31,32. Conversely, required levofloxacin doses were similar to doses used in patients.

This study presents some limitations that deserve further discussion. First, unlike in the most probable clinical scenario, we used a high challenge dose in ventilated pigs, which resulted in an immediate injurious response. These methods were aimed at developing severe pneumococcal pneumonia quickly, and the model should be considered to have captured only certain features of pneumococcal pneumonia. Second, we found that S. aureus and S. suis concurrently colonized the respiratory system. An alternative methodological approach could consist of further increasing the doxycycline dose to avoid colonization by the most frequent Gram-positive pathogens. Third, because of severe respiratory/hemodynamic instability, we found a survival rate of 70%. This substantial mortality may be a limitation for future drug-testing studies, by reducing the number of animals included in the analysis, raising ethical concerns and creating potential bias in results. Fourth, only female pigs were used, and thus the impact of sex on the model is unknown. Finally, potential viral colonization or coinfection has not been studied.

In conclusion, we have developed a model of penicillin- and macrolide-resistant S. pneumoniae serotype 19 A in mechanically ventilated pigs. Pulmonary infection caused bacteremia in half of the pigs and was associated with severe hemodynamic compromise.

Methods

This study was conducted at the animal facilities of the University of Barcelona, Spain. All procedures were performed following European Directive 2010/63/UE and Spanish RD 53/2013 regulations related to the Guide for the Care and Use of Laboratory Animals. The study protocol was approved by the Animal Experimentation Ethics Committee of the University of Barcelona (approval reference number: DAAM 42/12). To ensure adequate animal welfare, the pigs were housed in groups before the experiments. The pigs were tested for common pathogens before the experiments, identifying normal and non-pathogenic flora (data not shown). Figure 1 depicts the study design and aims of this multi-phase project. Further methodological details are available in the Supplementary Information.

Preliminary study

We conducted preliminary analyses in four healthy Large White-Landrace female pigs (mean ± s.e.m.: 31.5 ± 3.0 kg; range: 28–34 kg; Specipig SL) that were surgically prepared and managed as reported below, but were not challenged intra-bronchially with S. pneumoniae. Throughout the study, 100 mg/kg of doxycycline was administered every 12 h to hinder colonization by endogenous Gram-positive bacteria. In addition, 50 mg/kg of aztreonam was administered every 8 h to prevent endogenous colonization by Gram-negative bacteria. We assessed clinical data, pulmonary mechanics and hemodynamic parameters at baseline and 24 h thereafter, as well as collected tracheal secretions for quantitative microbiology studies (detailed below). In addition, we performed a BAL, as reported below, to quantify bacterial burden at baseline and at 24 and 72 h. Finally, after 76 h, pigs were euthanized, and lung biopsy specimens were taken to quantify bacterial concentration.

Main study

The study was conducted on 10 healthy Large White-Landrace female pigs (mean ± s.e.m.: 33.7 ± 1.9 kg; range: 32–37 kg).

Animal preparation and mechanical ventilation

The pigs were sedated, orotracheally intubated and connected to a mechanical ventilator (SERVO-I; Maquet). Pigs were initially ventilated in volume-control, square-wave inspiratory flow, duty cycle 0.33, tidal volume 8 ml/kg, positive end-expiratory pressure of 3–5 cm H2O and respiratory rate adjusted to maintain arterial partial pressure of carbon dioxide within normal range. Inspiratory gases were conditioned via a heated humidifier (MR850; Fisher and Paykel) with a heated inspiratory circuit. The pigs were under continuous anesthesia with a steady infusion of midazolam at 0.2–0.8 mg kg−1 h−1, propofol at 4–9 mg kg−1 h−1 and fentanyl at 5–10 µg kg−1 h−1. Respecting strict sterile technique, we performed an ultrasound-guided femoral artery cannulation for systemic MAP monitoring and blood sample collection. We surgically inserted an 8-French mm catheter into the internal jugular vein (Introflex percutaneous sheath introducer; Edwards Lifesciences) and placed a 7-Fr Swan-Ganz catheter (Swan-Ganz PAC; Edwards Lifesciences) through the introducer. Furthermore, to monitor urinary output throughout the study, we inserted a no. 16 Foley catheter into the bladder via a surgical midline mini-pelvectomy. Throughout the study, 100 mg/kg of doxycycline every 12 h and 50 mg/kg of aztreonam every 8 h were administered to hinder colonization by endogenous bacteria. After surgical preparation, we placed the pigs in the prone position for bacterial challenge; thereafter, we cyclically moved the pigs into a right- or left-lateral position every 6 h. Throughout the study, we assessed gas exchanges, pulmonary mechanics, hemodynamics, urine output and ventilatory settings every 12 h; we also obtained a complete blood count, basic metabolic panel, liver function, renal panel and coagulation tests.

Bacterial challenge

We prepared a culture of S. pneumoniae (ST276) serotype 19 A, characterized by resistance to penicillin, macrolides and tetracyclines with MICs of ceftriaxone of 1 µg/ml, erythromycin >256 µg/ml, azithromycin >256 µg/ml, doxycycline 6 µg/ml and levofloxacin 0.75 µg/ml. We initially attempted to induce severe pneumococcal pneumonia through single bronchoscopic challenge with 15 mL of 108 c.f.u./ml of S. pneumoniae into each lobe of four pigs set in the prone position: two pigs on the bed oriented horizontally and the other two at an angle above horizontal. After this challenge, the pigs remained healthy and did not develop pneumonia (data not shown). We therefore modified our methods accordingly: shortly after surgical preparation and 4 h thereafter, two separate slow instillations of 15 ml of 108 c.f.u./ml of the aforementioned log-phase culture of S. pneumoniae were performed into each pulmonary lobe of the pigs as they laid on a bed oriented 15° above horizontal.

Microbiological assessments and clinical definition of pneumonia

Before bacterial challenge and every 24 h thereafter, we collected tracheal secretions for quantitative microbiology studies. In addition, before bacterial challenge, and after 24 and 72 h, we performed a BAL of the right middle lobe for quantitative microbiology studies and quantification of inflammatory parameters. Finally, before bacterial challenge and every 24 h thereafter, we assessed blood cultures. At 24 h, clinical diagnosis of pneumonia was confirmed by an association with S. pneumoniae colonization ≥106 c.f.u./ml in tracheal secretions, plus at least one of the following clinical features: body temperature >38.5 °C or <36 °C, a white blood cell count >14,000/mm3 or <4,000/mm3 and purulent secretions.

Respiratory measurements

We collected respiratory mechanical measurements daily as previously described33. Airway pressure was measured proximally to the endotracheal tube, while respiratory flow rates were measured with a heated pneumotachograph. Flow and pressure signals were recorded on a personal computer for subsequent analysis with dedicated software (Colligo; Elekton). The static elastance of the respiratory system was calculated by using standard formulae and applying the rapid occlusion method34.

Hemodynamic measurements

After assessment of pulmonary variables, we evaluated hemodynamics and gas exchange (arterial and mixed venous blood) as previously described33. Stroke volume, systemic vascular resistance, pulmonary vascular resistance and venous admixture were computed as previously reported33. In addition, we computed vasopressor dependency index35.

Inflammatory parameters

Before bacterial challenge and at 24, 48 and 72 h thereafter, blood was drawn to measure serum IL-1β, IL-6, IL-8, IL-10 and tumor necrosis factor α concentrations. Furthermore, before bacterial challenge and at 24 and 72 h, the aforementioned cytokines were also quantified in BAL. Inflammatory marker concentrations are reported as log pg/liter.

Necropsy, post-mortem microbiological and histological studies and the definition of pneumonia

76 h after tracheal intubation (72 hours after the initial bacterial challenge), pigs were euthanized. We prematurely discontinued the study in cases of severe refractory respiratory instability (ratio between the partial pressure of arterial oxygen and inspiratory fraction of oxygen <100, irrespective of maximal ventilatory support) or refractory hemodynamic instability (MAP <50 mmHg, irrespective of norepinephrine dosing >0.3 µg kg−1 min−1). We collected two samples from the most-affected region of each of the five lobes for histological and quantitative microbiological studies. Lung histology was evaluated according to previously published methods and a six-point injury score36. For each lobe, pneumonia was microbiologically confirmed by a quantitative S. pneumoniae lobar culture ≥103 c.f.u./g and histologically confirmed per an injury score ≥3.

Pharmacokinetic analyses

In eight additional pigs, severe pneumococcal pneumonia developed as previously described. After pneumonia was clinically confirmed, different doses of ceftriaxone and levofloxacin were administered to assess pharmacokinetics in the ELF. Initially, we administered 50 mg/kg of ceftriaxone in two pigs and 10 mg/kg of levofloxacin in two additional pigs. We sequentially collected plasma and performed BAL in the right middle lobe to measure antibiotic concentrations in the ELF37. Plasma and BAL urea concentrations were determined with a validated enzymatic assay38. Our aim was to achieve a concentration of unbound ceftriaxone in the ELF greater than S. pneumoniae MIC for ≥14.4 h (60% of the time between daily dosing)32. With respect to levofloxacin, our aim was to achieve a ratio between the area under the concentration-time curve from 0 to 24 h post-dose of bound levofloxacin and S. pneumoniae MIC >4039.

Statistical analysis

Continuous variables were described as means and standard deviations or standard errors and median (25th–75th quartiles). Categorical variables were described as frequencies and percentages. Continuous variables were analyzed by using an REML analysis, based on repeated measures approach (PROC MIXED) and including times of assessment and pulmonary lobes as factors. A compound of symmetry or univariate (co)variance structure was used to model the within-subjects’ errors. For each continuous variable, the overall F test was first assessed for significance (P ≤ 0.05). Each pair-wise comparison was also performed, corrected by using the Bonferroni test to control for the experiment-wise error rate, and considered significant if its P value was ≤0.05. We tested the assumption in PROC MIXED with respect to normality of the model residuals. In cases of non-normally distributed residuals, we used the Friedman test. Categorical variables were analyzed by using Fisher’s exact test and the Cochran–Mantel–Haenszel test. All statistical analyses were performed by using SAS software (version 9.4; SAS Institute).

Reporting Summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.