Background
Community-acquired pneumonia (CAP) is a significant cause of morbidity and mortality worldwide, with
Streptococcus pneumoniae being the most prevalent causative pathogen [
1,
2]. Since the 1950s, the in-hospital mortality rate of CAP has remained about 12–13% in most high-income countries [
3]. Severe forms of CAP necessitate admission to the intensive care unit (ICU) and result in mortality rates ranging from 18 to 38% [
4‐
6]. Antibiotic intervention within 4 h of hospital arrival is associated with reduced mortality compared to a delayed start of treatment in CAP [
7,
8]. To improve survival in severe pneumonia, CAP is nowadays described as a medical emergency and early and aggressive treatment is therefore proposed on an empiric basis [
9‐
14]. However, the pathophysiological differences in the course of pneumonia resulting from early as opposed to late treatment are unknown.
Differences in survival of CAP patients receiving antibiotic therapy result from a wide range of contributors, which can be pathogen, drug or host related. The host’s immune response may aggravate detrimental pulmonary barrier failure and lung edema development [
15‐
19]. Particularly, activation of lung resident cells (e.g., alveolar macrophages and epithelial cells) by pathogen-associated molecular patterns (PAMPs) results in local inflammation, which in turn promotes attraction of inflammatory cells like polymorphonuclear leukocytes (PMNs) into the lungs [
20‐
22]. As professional phagocytes, PMNs are crucial for antimicrobial defense; however, PMNs also cause host tissue injury, leading to increased permeability of the alveolar–capillary barrier [
23,
24]. As a further consequence of pulmonary barrier failure, CAP can progress to life-threatening sepsis and multiorgan dysfunction [
25]. However, specific reasons for host-related differences in survival that depend on timely versus delayed antibiotic treatment remain unclear to date. Analysis of processes contributing to host-related therapy failure and high mortality rates due to delayed treatment are needed to foster the development of new adjunctive therapies. We hypothesized that antibiotic treatment decelerates exaggerated immune responses, but does not relevantly reduce established lung barrier dysfunction and lung edema. In the current study, we performed in-depth examination of mice with severe pneumococcal pneumonia receiving early versus late antibiotic treatment.
Methods
A detailed methodology and materials are presented in Additional file
1: Materials and methods.
Study approval
Female C57BL/6 N mice (8–10 weeks old; Charles River, Germany) were housed under specific-pathogen-free conditions. All animal studies have been ethically reviewed and approved by the State Office for Health and Social Services, Berlin, Germany.
Murine bacterial pneumonia
Mice were transnasally inoculated as previously described in detail [
26] with 5 × 10
6 colony-forming units (CFUs) of
S. pneumoniae serotype 3 in 20 μl PBS. Control mice received sham infection (PBS). Starting at 24 h or 48 h p.i., ampicillin (0.4 mg/mouse) or 0.9% NaCl (solvent control) was injected intraperitoneally every 12 h (Additional file
1: Figure S1). Body weight, temperature and murine pneumonia symptoms (scaling from 0 (absent) to 1 (present) and 2 (severe)) were assessed twice daily. Detailed information about the scoring system for murine pneumonia symptoms is presented in Additional file
1: Materials and methods, and Table S3. The summary of scored symptoms provides individual clinical scores. Mice meeting the inclusion criteria at 24 h p.i. (body weight loss more than 10% and/or body temperature < 37.0 °C) were analyzed. Numbers of analyzed mice per group per time point are summarized in Additional file
1: Tables S1 and S2. At defined endpoints, mice were anesthetized and exsanguinated prior to analysis. EDTA-blood was used for hemogram (measured with Scil Vet abc; scil animal care company GmbH) and bacterial load determination. The remaining blood was collected in serum separator tubes (BD) and sera were stored at − 80 °C.
Bronchoalveolar lavage
Mice were sacrificed prior to lavage. Airways were washed twice with 800 μl PBS. Bronchoalveolar lavage (BAL) suspensions were used to determine the bacterial burden. Supernatant (BAL fluid (BALF)) was frozen at − 80 °C. BAL cells were used for leukocyte differentiation.
Bacterial loads
Serial dilutions of EDTA-blood and BAL suspensions were plated onto Columbia agar with 5% sheep blood (BD) and incubated for 24 h at 37 °C under 5% CO2 prior to CFU counting.
Histological analysis
Mice were sacrificed prior to histological analysis. Lungs were immersion-fixed in 4% buffered formalin, embedded in paraffin, cut into 2-μm sections and stained with hematoxylin and eosin (H&E) as described previously [
27,
28]. Lung inflammation was scored based on consideration of specific parameters graded on a scale of 0 (absent) to 4 (severe), including the degree of inflammation, infiltration of neutrophils, pleuritis and steatitis. A lung edema score was assessed as a sum of distribution and degree of interstitial (perivascular) and alveolar edema graded on a scale of 0 (absent) to 5 (massive).
Blood clinical analytes
AST and urea levels were measured using Cobas-8000-C701 (Roche) by LABOKLIN (Bad Kissingen, Germany).
Leukocyte differentiation
BAL cells were preincubated with blocking antibody and stained with anti-CD11c, anti-CD11b, anti-F4/80, anti-Ly6G, anti-Ly6C and anti-MHCII antibodies. Cells were measured and analyzed with FACS Canto II (BD) and FACSDiva software. Cell numbers were calculated using CountBright Absolute Counting Beads (ThermoFisher Scientific).
Pulmonary vascular leakage
Mouse serum albumin (MSA) levels in BALF and serum were measured by ELISA (BETHYL). The alveolar–capillary barrier permeability index was calculated as the ratio of BALF/serum MSA.
Multiplex assay and ELISA
Mouse cytokine/chemokine levels were measured in BALF and serum with the ProcartaPlex Custom Mix & Match according to the manufacturer’s instructions (AffymetrixBioscience). BALF levels of CXCL2 and CXCL5 were measured by ELISA (RD).
Cytokine-immune cell network
Details for construction of the network are described in Additional file
1: Materials and methods.
Data analyses
GraphPad Prism 6 software was used for statistical analyses. Survival curves were analyzed using the Kaplan–Meier method and log-rank (Mantel–Cox) test. The Kruskal–Wallis test/Dunn’s multiple comparisons test and one-way ANOVA/Dunnett’s multiple comparisons test were used for comparison to S. pneumoniae-infected mice at the start of therapy. Grouped analyses were performed using two-way ANOVA/Sidak’s multiple comparison test. p < 0.05 was considered significant.
Discussion
Upon CAP diagnosis, commencement of immediate treatment is considered critical since a delay increases the likelihood of lethality [
7,
9]. However, some patients present themselves in a phase of advanced disease or diagnosis is belated. In these cases, antibiotic treatment can become insufficient to avoid development of acute lung injury; thus, adjuvant therapeutic options are needed. Understanding the pathophysiology of acute lung injury in pneumonia with delayed treatment is a prerequisite to successful therapy development. Therefore, we established a mouse model of CAP comprising early and late antibiotic treatment of bacterial pneumonia mirroring different treatment commencements in CAP patients. Validity of the model was achieved as: severe pneumonia was established before early and late treatment start; treatment success rates in terms of survival outcome depended on timely treatment; and antibiotic treatment efficacy in terms of bacterial elimination was independent of timely treatment. Our model settings are thus in line with observations made in immunocompetent ICU patients with CAP who received an adequate antibiotic regimen and management of comorbidities, but nonetheless showed high mortality rates [
32,
33].
Aiming to identify host-related factors responsible for detrimental disease outcome, we suspected an inadequate immune response, as described and reviewed by Mizgerd [
34], as the underlying mechanism following late therapy. Indeed, histopathological analyses revealed that only late antibiotic treatment failed to reduce neutrophilic infiltration into the pleura and adjacent mediastinal adipose tissue. In contrast, recruitment of inflammatory cells into lung tissue and air spaces appeared unaffected by early and late antibiotic interventions but, nevertheless, alveolar levels of inflammatory mediators were found to be repressed.
In order to assist interpreting these confounding observations, we implemented our data in a cytokine–cell network model. The network analysis illustrated that antibiotic treatment suppressed proinflammatory cytokines more efficiently than inflammatory chemokines (such as CXCL1, CXCL5 and GM-CSF). This indicates that recruiting mechanisms like chemokine gradients were less affected by antibiotic treatment. However, pronounced antibiotic-dependent reduction in inflammatory mediators suggests that, contrary to cellular numbers, the activation state and effector functions of alveolar immune cells were dampened by antibiotic-governed bacterial eradication.
An intact blood–air barrier is crucial for gas exchange, and fatal barrier breakdown during streptococcal pneumonia, leading to acute respiratory distress syndrome (ARDS), is well described (reviewed in [
35‐
37]). Accordingly, we observed that an intact lung barrier was associated with favorable therapy outcome in antibiotic-treated mice exhibiting severe pneumococcal pneumonia. Furthermore, differences in survival correlated with extended exposure time to pneumococci, resulting in a prolonged local and systemic contact with infection-induced inflammatory mediators. In line, early as opposed to late treatment prevented systemic inflammation.
The evident local and systemic inflammation might as well have occurred due to release of proinflammatory cytosolic components upon cell death processes. Necroptosis, a form of programmed necrosis, can be initiated by
S. pneumoniae in pulmonary immune cells like macrophages and epithelial cells [
38,
39]. In our pneumonia model, alveolar macrophage numbers remained stable upon infection, whereas numbers of inflammatory macrophages significantly increased with infection time, without showing significant differences between treatment and control groups. Due to their slow replenishment rates [
40], stable numbers of alveolar macrophages rather speak against necroptosis. Monocyte-derived macrophages that were actively recruited to infected lung tissue could potentially have undergone necroptosis and been replaced by newly recruited cells. To date, it remains likely but unproven that higher rates of monocyte-derived macrophage and epithelial cell necroptosis, following a late antibiotic regimen, are one of the underlying mechanisms for elevated inflammation and vascular leakage. Correspondingly, the presence of PAMPs as well as necroptosis-independent DAMPS could have likely regulated the activation status of local and systemic immune cells, thereby altering levels of inflammatory mediators following antibiotic intervention.
In severe pneumonia, only early antibiotic therapy, administered prior to barrier breakdown, prevented systemic inflammation, development of pleuritis, steatitis and elevated AST levels, which was followed up by restoration of fitness and rescue of mice from fatal outcome. Gracia et al. [
41] have shown in
S. pneumoniae-infected rats that early and late antibiotic regimens similarly eliminated bacterial burdens whereas only early therapy helped to prevent lung damage. However, the authors started early therapy just 1 h p.i., an important difference to our protocol whereby infected mice already developed pneumonia prior to antibiotic treatment. We therefore conclude that our model represents the patients’ situation in a more compatible fashion.
Nonetheless, our mouse model remains a species-different approach to a highly complex human disease. It does not include therapeutic measures conducted in the ICU such as additional intake of fluids, (high flow) oxygen or macrolide therapy. Furthermore, laboratory mice which did not experience pathogen contact before generally show a modified immune response as humans do, for example, since they are missing mucosal memory cells [
42]. Nevertheless, mice exhibited many parallels to the patients’ course of infection; from a contained local infection, pneumonia developed into life-threatening systemic inflammation as often seen in CAP patients. Further studies from our laboratory likewise described processes in murine pneumonia which resembled the human course of pneumonia [
43‐
45].
As a consequence, we stress the importance of defining CAP as an emergency and to identify and treat CAP patients at risk prior to lung barrier failure and systemic inflammation [
13,
14,
46]. Besides many promising studies [
47‐
52], specific adjunctive therapies for CAP are still missing. Such therapies should be coadministered with antibiotics in order to reduce lung barrier failure, thereby alleviating the disease course and finally reducing mortality and the length of hospital stay.
Our mouse model of antibiotic-treated CAP will aid in investigating how barrier-stabilizing interventions can prevent fatal disease progression in mice. Conclusively, proof-of-concept analyses indicating to which extent this mouse model adequately mirrors a significant range of patients must follow.
Acknowledgements
S. pneumoniae PN36 was kindly provided by S. Hammerschmidt, University of Greifswald, Germany. Furthermore, the authors thank Angela Linke for excellent technical assistance with histological lung processing and Dres. Birgitt Gutbier and Katrin Reppe for managing the in-laboratory animal facilities.