A long-lasting porcine model of ARDS caused by pneumonia and ventilator-induced lung injury

This group comprised six animals. Briefly, after animal preparation, we performed 3 h of VILI with a ventilator setting that ensured harmful pulmonary dynamic strain (> 1.5) [12]. Following this, we instilled 15 mL of 10⁷ colony-forming unit (CFU)/mL of MDR P. aeruginosa, susceptible to both meropenem and levofloxacin, into each lobe via bronchoscopy (see Additional file 1: Table S1). Injurious mechanical ventilation resumed with the same previous setting, albeit with a positive end-expiratory pressure (PEEP) of 5 cmH₂O. We diagnosed ARDS when arterial partial pressure of oxygen (PaO₂)/fraction of inspired oxygen (FiO₂) < 150 mmHg was achieved. Once criteria for ARDS were met, ventilator mode was changed to volume control (VC) and pigs were ventilated according to a lung protective strategy. Tidal volume was set to aim for normocapnia with a DP that did not exceed 15 cmH₂O [13].

The control arm included four animals which presented bilateral pneumonia and received treatment with protective ventilator parameters (no significant strain). Like the previous group, 3 h of ventilation was applied before the bacterial challenge, with a setting that assured a non-harmful pulmonary strain (< 1.5) [12]. After that, we instilled 15 mL of 10⁷ CFU/mL of MDR P. aeruginosa into each pulmonary lobe. We continued applying protective ventilation until the end of the experiment. Apart from a DP < 15 cmH₂O [13], we maintained the thresholds for the most important VILI parameters under a safe range per literature [5, 12, 13] (Table 1).

Pneumonia was clinically diagnosed in both groups based on a decline in PaO₂ and one of the following signs of infection: body temperature > 38.5 °C, a white blood count > 14·10⁹/L and purulent secretions [14]. At pneumonia diagnosis, we started antimicrobial therapy with 25 mg/kg of meropenem every 8 h and 10 mg/kg of levofloxacin every 24 h, as dual therapy recommended by international guidelines [15].

Table 1 and Fig. 2 display parameters related to mechanical ventilation, gas exchange and respiratory mechanics observed during the induction period. During this period, we performed a 15-mL P. aeruginosa challenge of 7.77 ± 0.11 log CFU/mL in each pulmonary lobe in both groups.

After the bacterial challenge, we continued injurious ventilation in the VILI group for 28.04 ± 3.03 h until the Berlin criteria were met, with the mean PaO₂/FiO₂ being 129.62 ± 6.1 mmHg at ARDS diagnosis. Bilateral opacities were identified by X-ray in all six pigs. In contrast, the pneumonia-without-VILI group never met the Berlin criteria for ARDS. PaO₂/FiO₂ did not decrease below 300 mmHg at any time point (mean PaO₂/FiO₂ was 389.90 ± 3.97 mmHg) (p < 0.0002) (Fig. 2).

No significant differences were found in the hemodynamic parameters between groups during the induction period, as shown in Fig. 3.

Pneumonia was clinically and microbiology established at 11.99 ± 0.67 and 7.25 ± 1.12 h post-inoculum in the pneumonia-with-and-without-VILI groups, respectively (p = 0.01). Additional file 1: Table S2 shows the achieved criteria of pneumonia diagnosis. No differences in P. aeruginosa burden were observed between study groups at pneumonia diagnosis, neither in tracheal secretions (p = 0.46) nor in BAL samples (p = 0.57) (Fig. 4).

Animals comprising the pneumonia-with-VILI group maintained a PaO₂/FiO₂ below 300 mmHg until the end of experiment (model development period). The mean duration of an ARDS diagnosis in animals that survived was 58.7 ± 1.3 h. For those who were killed due to clinical instability, it was 29.9 h and 16.3 h each. During this period (model development), the mean PaO₂/FiO₂ was 187.80 ± 28.46 mmHg, with the lowest measure recorded at 83.00 ± 5.45 mmHg. Figure 2 displays the evolution of respiratory system mechanics during model development. Respiratory system compliance was lower in animals with VILI compared with those without VILI (14.54 ± 1.30 and 21.77 ± 0.32 cmH₂O/L) (p = 0.011). Furthermore, DP and mechanical power showed significant differences between groups (p = 0.011 and p = 0.001, respectively). Mean values were 17.65 ± 0.50 cmH₂O and 26.87 ± 1.72 J/min in the group of pneumonia with VILI and 13.46 ± 0.34 cmH₂O and 10.09 ± 0.27 J/min in the group of pneumonia without VILI. After the ARDS diagnosis time point, PaCO₂ in animals with VILI was significantly higher than in those without VILI (69.28 ± 10.95 and 38.85 ± 1.076 mmHg, p = 0.027). This also was observed in exhaled minute volume (11.71 ± 0.32 and 7.02 ± 0.10 L/min, p < 0.001). Furthermore, mean EVLW was 840.2 ± 61.73 mL/kg in the pneumonia-with-VILI group and 385.6 ± 6.90 mL/kg in the pneumonia-without-VILI group (p = 0.013), as shown in Fig. 2. Ventilator setting adjustments are displayed in Additional file 1: Fig. S1.

Mean pulmonary pressure increased in pigs with ARDS: Mean values were 26.23 ± 0.73 mmHg compared with 22.79 ± 0.28 mmHg in in the pneumonia-without-VILI group (p = 0.056). Mean arterial pressure in pigs with ARDS remained at 72.48 ± 1.23 mmHg. In the other group, it was 84.79 ± 5.21 mmHg (p = 0.003). Pigs from the pneumonia-with-VILI group presented a higher vasopressor dependency index (VDI); however, it was not significant (0.63 ± 0.16 vs 0.03 ± 0.02, p = 0.158). After an ARDS diagnosis, systemic vascular resistances were lower in the pneumonia-with-VILI group than in the pneumonia-without-VILI group (907.3 ± 44.76 dynes/s/cm⁻⁵ 1432.71 ± 124.41 dynes/s/cm⁻⁵; p = 0.02). Figure 3 shows other hemodynamic parameters. Laboratory findings and fluid balance are reported in Additional file 1: Table S3 and Fig. S2.

We performed X-ray images of two different projections at baseline and the end of the experiment. While normal lung appearance was found at baseline, bilateral opacities were observed at the end of the experiment in both groups. X-ray images were also performed at ARDS diagnosis in animals with VILI, revealing bilateral opacities too. At necropsy, macroscopic findings were more severe in those with VILI than in animals presenting pneumonia without VILI, as shown in Additional file 1: Fig. S3.

Figure 5 displays the main lung findings. Macroscopically, all pigs presented with findings consistent with bilateral pneumonia. However, animals that developed ARDS also had widespread hepatization, severe edematous appearance, pleural effusion and adherence. Importantly, in the pneumonia-with-VILI group, we found higher vascular permeability, as revealed by both a higher lung/body weight ratio (2.71 ± 0.26 in the pneumonia-with-VILI group vs. 1.61 ± 0.13% in the pneumonia-without-VILI group, p = 0.009) and lung wet-to-dry ratio (7.56 ± 0.73 in the pneumonia-with-VILI group vs. 4.03 ± 1.03% in the pneumonia-without-VILI group, p = 0.029).

In the pneumonia-with-VILI group, all except one (83%) presented hyaline membrane formation consistent with diffuse alveolar damage (DAD). However, DAD and hyaline membranes were absent in all the animals in which pneumonia was induced without VILI. Furthermore, when compared with the pneumonia-without-VILI group, animals that developed ARDS also presented a rise in intra-alveolar (p = 0.033) and interstitial neutrophil infiltration (p = 0.005), as well as increased septal edema (p = 0.038). Alveolar edema, intra-alveolar fibrin and hemorrhage, proliferation of alveolar type II cells and epithelial denudation were reported in both groups without differences. In Additional file 1: Fig. S4, the same features are displayed by dependent and non-dependent areas.

As shown in Additional file 1: Fig. S5, systemic IL-6 and IL-8 had significant time-dependent changes (p < 0.0001 and p = 0.024). At pneumonia diagnosis, plasma IL-6 was higher in the pneumonia-with-VILI group (579.94 ± 138.79 vs. 356.79 ± 233.66 pg/mL, p = 0.038). However, after the ARDS diagnosis (model development period), these differences disappeared. IL-6 and IL-8 also presented an increase in BAL, albeit without significant differences between groups.

Figure 5 displays the main microbiological results. Throughout the model development period, animals from the pneumonia-with-VILI group presented a lower burden of P. aeruginosa in tracheal aspirates (4.24 ± 0.68 log CFU/mL in the pneumonia-with-VILI group and 6.01 ± 0.32 log CFU/mL in the pneumonia-without-VILI group, p = 0.027). However, it was not different for lung tissue or BAL. At the end of the experiment, tracheal aspirate cultures from both groups tested > 3 log CFU/mL. In lung tissue, mean values were below 3 log CFU/mL in both groups.

Given the clinically relevant insults used (pneumonia and VILI) as well as the physiological characteristics exhibited during the 96-h experiment period, we believe that the model presented has higher accuracy in reproducing human ARDS compared with others [7, 8, 16, 17].

In line with our study, the repeated lavage model presents remarkable gas exchange disorders and diminished compliance due to surfactant depletion. Specifically, our ARDS model shows a mean PaO₂/FiO₂ which, throughout the model development, ranges in moderate severity with periods of severe instability. Moreover, notably diminished and progressively deteriorating compliance is found throughout the model development period as well. Surfactant depletion is not the main mechanism of respiratory failure in adult ARDS; hence, this model fails to reproduce other clinical features of ARDS. For example, in contrast from our model, only modest epithelial injury signs are present; permeability changes, scarce; and clinical features described, reversible, lasting less than 24 h [9, 10, 18]. Our model, however, presents a histology that demonstrates that severe epithelial damage and disorders in oxygenation are maintained up to approximately 60 h. In the surfactant depletion model, some authors reported an increase in BAL cytokines, but the role of inflammation is unclear [9]. We demonstrated, nonetheless, an increase in both systemic and pulmonary inflammation and signs of sepsis-associated cardiovascular dysfunction.

Similar flaws are identified in the oleic acid (OA) model. This one mimic ARDS caused by fat embolism, an exceptional etiology of syndrome [10]. In the OA model, the pathophysiology of pulmonary injury is unknown and unlikely similar to the most common causes of ARDS such as pneumonia, the risk factor used in our model [6]. OA administration raises pulmonary arterial pressure and intrapulmonary shunt [19‐21] and produces a marked increase in EVLW [20]. We also showed a rise in the mean pulmonary arterial pressure and EVLW. We tracked changes in PaO₂/FiO₂ as surrogate of pulmonary shunt, identifying periods with a value below 100 mmHg (i.e., severe ARDS). In the OA model, an inflammatory response is shown with an increase in some inflammatory markers in lung such as IL-6 and IL-1 beta [22‐24]. At histological examination, however, only mild signs of lung injury are found with no hyaline membrane formation [20, 23, 25]. In our model, though, we demonstrated DAD in 83% of histological specimens of pigs with ARDS. In the OA model, the severity of lung injury is somewhat unpredictable because animals present different responses per OA doses and sudden, intense hemodynamic instability during infusion [23, 25]. Like the repeated lavage model, the oleic acid-induced lung injury is also reversible within several hours, contrasting with our 60-h ARDS period [21, 23].

Systemic administration of bacterial lipopolysaccharides (LPS) was one of the earliest approaches used to study sepsis-induced ARDS. LPS administration markedly decreases PaO₂ and systemic arterial pressure. It also provides important information about host inflammatory responses in BAL and serum [7, 26]. For that reason, it is a good choice for pathophysiological studies of infection and ARDS [7, 26]. Our model follows this approach; however, we instill viable bacteria—rather than bacterial toxins—into the lungs, thus fulfilling the complete clinical picture of severe pneumonia. During the model development, a high burden of P. aeruginosa was found in BAL and tracheal aspirates. While we did not find differences in BAL samples, we did identify a lower bacterial burden in tracheal aspirates from animals with ARDS in comparison with those who did not develop it. The interpretation of this finding is not clear. We speculate that the plasma leak into the lungs of pigs with ARDS prompted tracheal aspirations more frequently and promoted bacterial clearance. Histological alterations observed in the LPS model are mainly neutrophil parenchymal infiltration and mild edema, features that resemble those seen in pneumonia. However, in this model, hyaline membranes are lacking [7, 27], suggesting milder epithelial damage and lower pulmonary permeability compared with our model. Other limitations of this model are short duration, significant differences in endotoxin response between species and variability in biological activity of endotoxin preparations [8]. In contrast, the duration of our model is enough to perform multiple analyses, given its reproducible nature as well.

The smoke/burn is a reproducible model that resembles the clinical time course of ARDS during the first 24 h. It can be used in studies up to 96 h [28]. However, smoke inhalation as etiology of ARDS is not frequent, and the pathophysiology of lung injury is different from sepsis [6]. Although some studies achieve severe respiratory failure [28], this does not last longer 24 h [29, 30]. Inflammatory markers as IL-6 are found elevated in BAL and plasma [28, 31]. Histology observations demonstrate a presence of inflammatory infiltrate, mild hemorrhage and edema, albeit no hyaline membranes [28, 31, 32].

Injurious ventilation has been used as a unique or double-hit model to reproduce ARDS in animals [8, 32]. In rodent animal models, VILI causes an impairment in respiratory mechanics [33, 34], tissue injury [33, 35] and pulmonary permeability [34], but oxygenation disorders are mild and do not usually meet ARDS criteria [36‐39]. These findings in small animals are difficult to translate in terms of humans. Contrarily, in larger animal models, VILI provokes severe respiratory failure. Unlike our model, though, it does not last longer than 24 h [38‐42]. In these models, inflammation caused by VILI was also evaluated by several methods including cytokine analysis in BAL fluid, plasma and lung tissue, and either immunohistochemistry or immunofluorescence [8, 33-44]. While some studies confirmed that VILI induces a pulmonary inflammatory response [33‐39], others did not [44]. These differences might be explained by either the intensity of VILI observed between studies or the primary pulmonary insult performed before aggressive ventilation was started [45]. Interpretations of our findings should consider that we used a DP of up to 30 cmH₂O to ensure a clinically significant pulmonary strain and that such ventilation was performed in lungs affected by a primary inflammatory insult (pneumonia). Interestingly, as observed in our model, other authors reported the presence of hyaline membranes and diffuse alveolar damage in animals ventilated in an injurious fashion [40].

In our model, the group with low pulmonary strain—no VILI—presented oxygen disorders that were mild and did not meet the Berlin criteria for ARDS. On the other hand, pigs comprising the pneumonia-with-VILI group had sustained life-threatening gas exchange disorders, as well as other remarkable disturbances. In the current model, however, we did not find clinically relevant extrapulmonary organ dysfunctions other than hemodynamic failure (sepsis-induced vasoplegia). The latter was more prominent in the pneumonia-with-VILI group. Of note, while pigs with ARDS exhibited a high inflammatory response, this was not clearly influenced by the presence of VILI during model development. Importantly, high mortality (33.3%) found only in the group of pigs ventilated with the high pulmonary strain was driven by respiratory acidosis and refractory respiratory failure.

The findings described between groups merit clinical interpretation. Refractory respiratory failure—either manifested as severe hypoxemia or respiratory acidosis—represents a non-negligible cause of death and morbidity in ARDS [46, 47]. Given the extreme oxygenation and ventilation disorders found in the group of pigs with VILI, maintaining protective ventilation was challenging after the ARDS diagnosis (model development period). This is revealed by an increase in mechanical power throughout all the experiment which, consequently, might have dramatically diminished the chances of pulmonary recovery in the group with VILI. Our results, therefore, suggest that a high pulmonary strain might ease progression to ARDS in pneumonia and perhaps raise respiratory-attributable death.