Ventilator-induced lung injury is aggravated by antibiotic mediated microbiota depletion in mice

Mechanical ventilation (MV) is a live-saving intervention in acute respiratory failure. Patients in intensive care units (ICU) are often subjected to MV, which can induce local and systemic inflammatory responses, thereby leading to the development of ventilator-induced lung injury (VILI) [1, 2]. VILI contributes to mortality and morbidity in patients with respiratory failure [1, 3, 4].

Critically ill patients in the ICU frequently receive antibiotics [5], and exposure to broad-spectrum antibiotics alters the microbiota. The microbiota is an “ecological community of commensal, symbiotic and pathogenic microorganisms” [6], which can be found e.g. in the gut, on the skin, and in the respiratory tract [7]. Recent studies revealed a significant impact of the physiological gut microbiota on the homeostasis of the immune system. The intestinal bacterial flora was shown to be essential for the calibration of local inflammatory responses of the gut [8]. Further, intact gut microbiota calibrates the systemic homeostasis between inflammatory and anti-inflammatory components and is essential for an effective immune response e.g. by altering IL-10 activity [9]. Recent studies suggest that the gut microbiota can also affect the pulmonary inflammatory response and associated lung injury [10‐12]. For instance, lung inflammation, organ damage and mortality have been shown to be increased in antibiotic-treated mice infected with Streptococcus pneumoniae compared to controls, suggesting that the gut microbiota modulates local inflammatory responses in the lungs [13]. Translocation of commensal bacteria and their metabolites, including short-chain fatty acids, from the gut into the bloodstream was suggested as a potential underlying mechanism of the gut-lung interaction [14‐16]. Moreover, Clarke et al. revealed that components of the microbiota, after translocation from the gut into the bloodstream, also regulate the inflammatory activity of neutrophilic granulocytes [16]. This might be helpful for the host in the case of infection, but might be harmful within the context of autoimmunity or tissue trauma-induced inflammation caused by e.g. MV.

Currently, little is known about the effect of the microbiota on local stimulation of the immune system and pulmonary inflammatory phenotype in sterile lung inflammation. Although great effort is made to minimize antibiotic exposure in general, certain groups of patients are exposed to long and frequent antibiotic treatment. This includes patients that are rendered immunosuppressed by e.g. chemotherapy and that are often treated with antibiotics more frequently, for longer periods and under certain circumstances even for prophylaxis. Hence, by depleting the microbiota by antibiotic treatment prior to MV we examined the impact of the microbiota on the pulmonary inflammatory response to MV and thereby its influence on the development of VILI. Some of the results presented here were previously reported in the form of abstracts [17‐19].

In the current study we found that disruption of the microbiota by antibiotic therapy prior to mechanical ventilation (MV) increased the susceptibility to ventilator-induced lung injury (VILI). Mechanical ventilation is a life-saving intervention without an alternative for patients with respiratory failure, but it may aggravate or even induce VILI [29]. Further, mechanically ventilated critically ill patients are frequently exposed to antibiotic treatment (reviewed in [30]) or receive selective digestive decontamination (SDD) [31]. Antibiotics change the microbiota composition, which impacts the local and systemic inflammatory response towards various stimuli [32‐36]. Taking this into account it is of great interest to determine the effect of antibiotic therapy and associated microbiota alteration on the development of VILI.

To investigate this interaction, we treated mice for 6–8 weeks with a broad-spectrum antibiotic regimen to generate mice with a virtually depleted gut microbiota. Using the described cocktail, we showed a significant decline in relevant bacterial species in the gut, but as part of the antibiotic applied is absorbed in the gut, the general microbiome of the animals is likely to be changed, which has to be considered in the interpretation of our results. It has been reported that treatment with antibiotics renders patients susceptible to pulmonary infections including multidrug-resistant pathogens [37, 38]. Recently, we have shown that antibiotic treatment lowers pulmonary IgA production and enhances susceptibility of mice towards P. aeruginosa [12]. However, the impact of the microbiota on the development of a hyperinflammatory sterile lung injury like VILI remains unclear. Using the described antibiotic cocktail including systemically effective antibiotics, we efficiently depleted the microbiota of the gut, but probably also of other organs including the lungs.

Prolonged antibiotic treatment prior to ventilation significantly aggravated VILI, as indicated by increased pulmonary permeability, decreased pulmonary compliance reflecting lung edema, and macroscopic signs of aggravated lung injury. An extensive proinflammatory response towards MV is a major driver of VILI, and alteration of the microbiota was reported to have a significant impact on the regulation of pulmonary inflammation [39, 40]. We found elevated protein levels of KC and IL-6 in antibiotic-treated mice subjected to VILI, but did not detect any effect of microbiome depletion on the inflammatory response in non-ventilated mice. It remains unclear if the aggravated inflammatory response to VILI in antibiotic-treated mice may be the consequence of aggravated lung injury due to MV and associated inflammation or if the absence of an intact microbiome prior to VILI may have rendered the lung more reactive leading to an increased inflammatory response.

In this study, we implemented severely injurious ventilation in mice applying a tidal volume of 34 ml/kg, which was much higher than the standard of lung-protective ventilation tidal volume of 6 ml/kg in humans with acute respiratory distress syndrome (ARDS). At first view, this might outrange the stress and strain applied during MV in patients with ARDS. However, functional residual lung capacity in ARDS is severely reduced, which is referred to as the “baby lung” in patients with ARDS [41]. Even with a tidal volume of 6 ml/kg it is unclear how much lung strain is produced during MV but available data show that lung stress and strain are significantly increased [42]. Further, the high grade of tissue inhomogeneity in lungs in ARDS, in which open, atelectatic, and collapsed but recruitable lung areas coexist, locally results in lung stress exceeding the measured airway pressure by far [43]. In the applied model, we used previously healthy lungs, sufficiently recruited at the beginning of MV. Therefore, the lungs are lacking consolidations and relevant atelectatic lung regions that account for the reduced functional residual capacity resulting in elevated lung strain in patients with ARDS even under lung-protective ventilation. Further, in the recruited healthy lungs homogenously ventilated local lung stress is minimized. Therefore simulating “clinically relevant” lung stress and strain accruing in patients with ARDS requires an experimental approach that provokes significant lung stress/strain in healthy lungs with previously normal residual capacity and in the context of homogenously open and ventilated lung tissue. Moreover, the lung injury to be investigated has to be prevalent in a feasible timeframe. Taking all this into account, ventilation with 34 ml/kg in healthy mice is a reasonable experimental approach for our study meeting the American Thoracic Society (ATS) criteria for lung injury research in animal models [21].

The complex interaction between gut and lung is still not fully understood. One possible pathway encompasses altered communication between these two compartments over the “gut-lung axis” (reviewed in [44, 45]). For instance, the risk of developing airway diseases is increased after antibiotic-induced alterations in the gut microbiota in early childhood [46‐48]. The intestinal microbiota might influence pulmonary immune responses by distributing bacterial metabolites via the bloodstream, which may have direct immunomodulatory effects [14, 49]. One of these immunomodulating mediators might be short-chain fatty acids (SCFA) which result from digestion and converting of indigestible nutrients by the commensal bacteria and have been shown to have anti-inflammatory action (as reviewed by Marsland [14]). Zhang et al. showed that prophylactic administration of the SCFA sodium butyrate reduced lung injury after sepsis and attenuated lung lesions [50]. Reduced anti-inflammatory properties could be a possible explanation for the aggravated inflammatory response to VILI that we observed. However, further studies are warranted to dissect in the current context whether elevated cytokines mediate the aggravation of lung injury or result from ventilation-induced tissue injury leading e.g. to the release of proinflammatory DAMPs [51, 52].

The approach of using long-term broad-spectrum antibiotics is a well-established method to deplete the microbiome in mice with no reports of any remarkable toxic effect to the animals [20, 26‐28]. Accordingly, histopathological analysis of naive lungs in microbiota-depleted mice in our study did not reveal any signs of pulmonary toxicity of the applied antibiotic regimen. To minimize a possible influence of circulating bacterial components arising after antibiotic lysis of bacteria, we stopped the antibiotic therapy 3 days prior to ventilation to “wash out” all possible remaining antibiotic substances or circulating bacterial metabolites. Taking this into account, an underlying toxic effect of the antibiotics on the lung tissue is very unlikely.

Nevertheless, as yet unknown effects of such antibiotic intervention might have influenced the results of this study. Studies using germ-free mice and mice recolonized after antibiotic treatment may offer further insights and allow to dissect if the microbiome itself or determine whether other potential antibiotic-related effects have affected the susceptibility to VILI. Thus, despite all limitations of an experimental animal model, we judge the complex model used in the present study to be suitable to examine the influence of the microbiota on the development of lung injury, best modeling the situation of ICU patients.

In summary, microbiota depletion by antibacterial therapy prior to MV enhanced inflammatory responses towards MV and rendered mice susceptible to the development of VILI. The findings highlight a potentially relevant interaction of the microbiome with the inflammatory response in lung injury.