Requirement for Serratia marcescens Cytolysin in a Murine Model of Hemorrhagic Pneumonia

Abstract

Serratia marcescens, a member of the carbapenem-resistant Enterobacteriaceae, is an important emerging pathogen that causes a wide variety of nosocomial infections, spreads rapidly within hospitals, and has a systemic mortality rate of ≤41%. Despite multiple clinical descriptions of S. marcescens nosocomial pneumonia, little is known regarding the mechanisms of bacterial pathogenesis and the host immune response. To address this gap, we developed an oropharyngeal aspiration model of lethal and sublethal S. marcescens pneumonia in BALB/c mice and extensively characterized the latter. Lethal challenge (>4.0 × 10⁶ CFU) was characterized by fulminate hemorrhagic pneumonia with rapid loss of lung function and death. Mice challenged with a sublethal dose (<2.0 × 10⁶ CFU) rapidly lost weight, had diminished lung compliance, experienced lung hemorrhage, and responded to the infection with extensive neutrophil infiltration and histopathological changes in tissue architecture. Neutrophil extracellular trap formation and the expression of inflammatory cytokines occurred early after infection. Mice depleted of neutrophils were exquisitely susceptible to an otherwise nonlethal inoculum, thereby demonstrating the requirement for neutrophils in host protection. Mutation of the genes encoding the cytolysin ShlA and its transporter ShlB resulted in attenuated S. marcescens strains that failed to cause profound weight loss, extended illness, hemorrhage, and prolonged lung pathology in mice. This study describes a model of S. marcescens pneumonia that mimics known clinical features of human illness, identifies neutrophils and the toxin ShlA as a key factors important for defense and infection, respectively, and provides a solid foundation for future studies of novel therapeutics for this important opportunistic pathogen.

INTRODUCTION

Serratia marcescens is a facultative anaerobic, rod-shaped, Gram-negative bacterium that is ubiquitous in water, in soil, and on plant surfaces. S. marcescens is also an antibiotic-resistant opportunistic pathogen and is among the top 10 causative agents of bloodstream bacterial infections in North America, with a mortality rate of 41% (1–3). In newborns and immunocompromised and intensive care patients, S. marcescens can cause severe infections such as pneumonia (4), bloodstream infections (5), and urinary tract infections, surgical site infections, and ocular infections (6). S. marcescens infections are most often associated with the hospital environment (5), but community-acquired infections are now increasingly diagnosed (7).

S. marcescens has acquired notoriety in the last 20 years because of its resistance to multiple antibiotics (7, 8). An 8-year surveillance study in Taiwan identified multiple S. marcescens strains that were resistant to the antibiotics ciprofloxacin and levofloxacin (9). Multiple studies have revealed an alarming increase in S. marcescens resistance to carbapenem and other β-lactam antibiotics (8, 10, 11). As such, S. marcescens is considered a member of the carbapenem-resistant Enterobacteriaceae (CRE) (12). A recent study found that S. marcescens has an intrinsic resistance to polymyxin, which is considered to be the final alternative for treating CRE pathogens (13). These studies underscore the importance of understanding how S. marcescens causes disease in order to identify prophylactic vaccine candidates or adjunct therapeutic approaches for prevention or treatment of nosocomial infections.

Since the 1950s, multiple case reports have described the pathological patterns of pneumonia caused by S. marcescens (14–18), yet little is known in regard to the molecular basis for the observed pathology and the host factors that are required to resolve infection in the lungs. To address this gap in knowledge, we developed and characterized a mouse model of S. marcescens pneumonia. Our model strongly recapitulated the clinical features of human disease, such as bronchopneumonia, loss of lung function, neutrophil infiltration, edema, and hemorrhage. What is more, using this model, we ascertained the importance of the neutrophil responses to limit severe infection and generated findings that suggest that the bacterial cytolysin ShlA was critical for the observed pathology. These studies enhance our knowledge and understanding of S. marcescens pneumonia and lay a foundation for future studies focused on this important emerging pathogen.

MATERIALS AND METHODS

Bacterial strains.

The tetracycline-resistant Serratia marcescens clinical isolate UT-383 was obtained from Jan Patterson (Division of Infectious Disease, Department of Medicine, The University of Texas Health Sciences Center at San Antonio [UTHSCSA]) and is defined here as “wild type” (WT). Isogenic, unmarked shlA and shlB deletion mutants were constructed by deleting the majority of the corresponding coding regions via allelic exchange, leaving behind the first 5 and the last 5 codons. Briefly, fragments carrying ∼750 nucleotides upstream and ∼750 nucleotides downstream of either the shlA or shlB open reading frame were PCR amplified and cloned into the kanamycin resistance-encoding suicide vector pSR47s. Clones were originally isolated in Escherichia coli DH5α λpir, and then, after sequence confirmation, a plasmid containing the deletion allele was moved into UT-383 via triparental conjugation with E. coli MT607/pRK600, as previously described (19). After isolation of Tet^r Kan^r bacteria, bacteria were grown in the absence of kanamycin selection, and sucrose-resistant isolates were selected and confirmed by sequencing for the deletion of the shlA or shlB coding sequence with no accompanying upstream or downstream mutations.

Culture conditions.

S. marcescens was grown on Luria-Bertani (LB) agar plates (LB-Lennox formulation, consisting of 10 g tryptone, 5 g yeast extract, and 5 g NaCl LB per liter) and incubated overnight at 30°C. A single colony was transferred to LB broth and incubated at 30°C with rolling overnight. The next day, the culture was back diluted at 1:50 in fresh LB broth for 3 h at 30°C. Growth at 30°C has been previously shown to be optimal for cytolysin expression (20). The infectious inoculums (75-μl volume) were prepared by diluting the bacteria with sterile phosphate-buffered saline (PBS) to the final desired bacterial concentration. For all experiments, the amount of CFU inoculated was determined by serial dilution of the inoculum and plating on LB agar plates.

Mouse infection.

All animal experiments were approved by the UTHSCSA Institutional Animal Care and Use Committee (protocol 12030X) and performed in agreement with the NIH Guide for the Care and Use of Laboratory Animals. Female 6- to 8-week-old BALB/c mice were obtained from Charles River Laboratories (Frederick, MD). Oropharyngeal aspiration was performed as previously described (21). Briefly, each mouse was anesthetized using 2% vaporized isoflurane and hung upright by its incisors with its back supported by a plank. The tongue was gently pulled outward using blunt forceps. An inoculum of 75 μl (∼1.0 × 10⁶ for a sublethal dose and ∼5.0 × 10⁶ for a lethal dose) was pipetted into the pharynx behind the tongue. Immobilization of the tongue (continued to be held gently with forceps), accompanied by coverage of the nares with a finger, achieved forced inhalation. Mock-infected animals received 75 μl of PBS. All mice recovered fully within 45 s of challenge.

Survival, weight change, and organ CFU assays.

Infected mice were monitored for up to 8 days, with survival and weight loss measured daily. For determination of CFU within organs, mice were sacrificed, and organs were aseptically removed, rinsed, and homogenized in PBS (TH homogenizer; Omni International, Marietta GA). Serial dilutions of tissue homogenates were applied to nutrient agar plates and incubated overnight at 30°C. Tissue CFU values were extrapolated from colony counts and normalized to the prehomogenization weight of the organ.

Lung function analysis.

Changes in mouse airway function after S. marcescens infection were determined using the Flexivent system (Scireq, Montreal, PQ, Canada). Mice were anesthetized with Avertin (Sigma-Aldrich, St. Louis, MO), and then pulmonary mechanic measurements were acquired after administration of aerosolized PBS. A positive end-expiratory pressure of 3.0 cm H₂O was maintained during experiments. A series of 12 repeated measurements to obtain dynamic compliance and dynamic resistance was repeated in duplicate. The 3 lowest or highest values of dynamic compliance and dynamic resistance measurements, respectively, were averaged per series. The two series were then averaged together.

Histology.

Following euthanasia, mice were perfused with PBS, their lungs were excised and inflated with 1 ml of 10% buffered formalin (Sigma-Aldrich, St. Louis, MO), and then the lungs were processed for paraffin embedding and sectioning. Four-micrometer sections were stained with hematoxylin and eosin (H&E) (UTHSCSA Histopathology Laboratory, Department of Pathology).

Cytokine analysis.

At the indicated times, mice were sacrificed, and the lungs were removed and placed in a solution of PBS containing protease inhibitor cocktail (Roche Diagnostics, Indianapolis, IN). After homogenization, samples were centrifuged to pellet debris, and then the supernatants aliquoted and stored at −80°C. Levels of cytokines and chemokines were assayed using the Mouse Cytokine 20-Plex panel (Life Technologies, Carlsbad CA) following the manufacturer's protocol, using a Bio-Plex 200 system with Bio-Plex Manager 6.0 software (Bio-Rad, Hercules CA).

Depletion of neutrophils and inflammatory monocytes.

BALB/c mice were treated via intraperitoneal injection with 100 μg of either control IgG2b specific to KLH (clone LTF-2) or neutrophil-depleting anti-Ly6G (clone Rb6-8c5). Both antibodies were obtained from BioXCell (West Lebanon, NH). Mice received these treatments or PBS at 24 h prior to bacterial inoculation.

Bronchoalveolar lavage fluid (BALF) and cytospins.

BALB/c mice were sacrificed and tracheotomies performed with an 18-gauge Angiocath Autoguard catheter (Becton Dickinson, UT). A 3-way stopcock with an attached 3-ml syringe was connected to the catheter, and lungs were lavaged 3 times with 0.5 ml of ice-cold PBS. Cells were pelleted by centrifugation at ∼1,000 relative centrifugal force (RCF) for 10 min and resuspended in 0.5 ml PBS. Cell concentration was determined using the Cellometer slide chamber (Nexcelom Biosciences, Lawrence MA). Differential cell counting was done by counting a minimum of 400 cells per sample and reported as the relative numbers of macrophages, lymphocytes, and polymorphonuclear cells (PMNs) (22). For cytospins, cells were diluted to a concentration of ∼10⁵ cells in 250 μl and fixed to Shandon Cytospin coated slides using a Cytospin 4 centrifuge (Thermo Scientific). Slides were stained for differentiation of blood cell types with the Protocol Hema-3 stain set (Fisher Scientific) and inspected using microscopy for total number of PMNs and monocytes.

Albumin analysis.

Vascular leakage in BALF fluid was assessed using a mouse albumin enzyme-linked immunosorbent assay (ELISA) quantitation set (Bethyl Laboratories Inc., Montgomery, TX) as described previously (23). Cleared BALF and albumin standards were applied to wells preabsorbed with an anti-mouse albumin antibody and allowed to incubate at room temperature for 1 h, and then wells were washed and a biotinylated detection antibody added. After another 1 h of incubation, wells were washed, and a solution containing streptavidin-conjugated horseradish peroxidase was added to wells and left for 1 h. Wells were washed, a detection solution containing 3,3′,5,5′-tetramethylbenzidine was added, and wells were incubated in the dark for 30 min. After the reaction was stopped, absorbance was measured at 450 nm.

Immunofluorescence and detection of apoptotic cells.

Prior to their excision, lungs from euthanized mice were inflated with 0.5 ml of optimal cutting medium (Tissue-Tek OCT; Sakura, Torrance CA) in 2 M sucrose (1:1, vol/vol) using a 19-gauge needle through the trachea. Inflated lungs were surgically removed, transferred to molds containing OCT, and snap-frozen. Tissue sections 8 μm in thickness were cut, placed on slides, and then fixed with acetone. Slides were wrapped in aluminum foil and stored at −80°C. After thawing, lung sections were fixed again in cold acetone (−20°C), in 70% ethanol (−20°C), and then in PBS to hydrate the sections. The sections were blocked at 25°C for 30 min in 3% goat serum–3% bovine serum albumin (BSA). The primary antibody used was rat anti-mouse Ly6G (clone 1A8; BD Biosciences) diluted at 1:50 in the blocking solutions and incubated over the tissue sections at 25°C for 40 min. After incubation, sections were washed with a solution of PBS–0.05% Tween 20. The secondary antibody used was goat anti-rat antibody conjugated to Alexa Fluor 488 (Jackson Immuno Research, West Grove, PA), diluted at 1:900 in blocking solution and incubated over sections at 25°C for 30 min. Slides were then washed and mounted with ProLong Gold Antifade reagent containing DAPI (4′,6′-diamidino-2-phenylindole) (Life Technologies, Carlsbad CA).

To label dead cells (cells positive by terminal deoxynucleotidyltransferase-mediated dUTP-biotin nick end labeling [TUNEL]), sections were stained using the ApopTag kit (EMD Millipore, Billerica MA) as follows. Equilibration buffer was incubated on the section for 1 min at 25°C, and then a terminal deoxynucleotidyl transferase solution (10% enzyme plus 10% PBS plus 80% reaction buffer) was applied for 20 to 60 min at 37°C. A termination buffer was applied for 10 min at 25°C and then sections were then washed 3 times with PBS. In a dark humidity chamber, an antidigoxigenin solution (53% blocking solution plus 47% antidigoxigenin conjugated to fluorescein isothiocyanate [FITC]) was applied and allowed to incubate for 30 min at 25°C. Slides were then washed 4 times for 2 min in PBS and then stained for Ly6G and mounted as described above.

Microscopy and image capture.

Stained tissue sections were visualized using either an EC Plan Neofluar 10× or Plan-Apochromat 20×/0.8 objective on a Zeiss AxioImager Z1 epifluorescence microscope (Carl Zeiss, Thornwood, NY). Images were captured using a Zeiss AxioXam MRm Rev3 and/or MRc camera and analyzed using Zeiss AxioVision release 4.5 software.

Western blots.

Whole lung homogenates were generated as described above but prepared in radioimmunoprecipitation assay (RIPA) buffer (150 mM NaCl, 1% NP-40, 0.5% deoxycholate [DOC], 0.1% SDS, and 50 mM Tris) (23). After homogenization, each biological sample was separated by SDS-PAGE and proteins transferred onto nitrocellulose using a semidry electrophoretic system (Bio-Rad, Hercules, CA). An antibody targeting myeloperoxidase (MPO) (polyclonal IgG; Abcam, Cambridge, MA) was applied to membranes at a 1:500 dilution. Membranes were subsequently probed using a secondary antibody against MPO (goat anti-rabbit antibody) conjugated to horseradish peroxidase (Jackson ImmunoResearch, West Grove, PA), at a 1:5,000 dilution, and developed using standard chemiluminescence methods. To confirm protein load, membranes were stripped and reprobed with antibody against actin (Bethyl Laboratories, Montgomery, TX).

Statistical analysis.

Prism 5 (GraphPad Software, La Jolla, CA) was used for graphing and statistical analysis. Survival curves were made using the Kaplan-Meier method and significance calculated using the log rank test. Mann-Whitney U tests were applied for two-group comparisons, and nonparametric analysis of variance (ANOVA) (Kruskal-Wallis) and Dunn's post hoc analysis were used for multiple-group comparisons.

RESULTS

S. marcescens induces acute illness and compromises lung function in mice.

In this study, we aimed to establish a murine model of S. marcescens pneumonia that mimics key aspects of human disease and to exploit this model to elucidate mechanisms of pathogenesis and the host response. First, we characterized illness caused by a wild-type (WT) clinical isolate following challenge via forced aspiration. Mice inoculated with 6.43 × 10⁶ CFU of the bacterium displayed 100% mortality within 24 h (Fig. 1A). In contrast, all mice challenged with the lower dose of 9.72 × 10⁵ CFU survived (Fig. 1A). Daily weight monitoring postinfection demonstrated that mice exposed to the lower sublethal inoculum lost >20% of their weight during the first 72 h, and this was followed by detectable recovery at 120 h (Fig. 1B).

FIG 1.

Impaired lung function and increased morbidity and mortality in mice with respiratory S. marcescens infection. (A and B) Female 6- to 8-week-old BALB/c mice were infected via forced aspiration with 75 μl of S. marcescens wild-type (WT) strain UT-383 (low dose [LD], 9.72 × 10⁵ CFU; high dose [HD], 6.43 × 10⁶ CFU) or UT-383 with deletion of either shlA (ΔshlA) (LD, 8.00 × 10⁵ CFU; HD, 5.98 × 10⁶ CFU) or shlB (ΔshlB) (LD, 9.60 × 10⁵ CFU; HD, 5.98 × 10⁶ CFU) at the indicated doses or given 75 μl of PBS, and then animal survival (A) and weight change (B) were monitored daily for up to 9 days postinoculation. Data shown are representative of 4 independent experiments with 5 or 6 animals per group. (C and D) For lung function experiments (n = 3), mice were administered either PBS or WT S. marcescens at the indicated doses and times, with lung compliance (C) and airway resistance (D) measured by pressure-volume analysis. Average values + standard errors of the means (SEM) are shown for weight change and lung function analyses. Student's t test was used to analyze statistical differences among the groups: *, P < 0.05; **, P < 0.01; n.s., not significant.

One suspected major virulence factor of S. marcescens is the ShlA cytolysin, which has been shown to lyse red blood cells (RBCs), epithelial cells, and fibroblasts in vitro (24). Secretion and activation of ShlA require the membrane-bound ShlB protein (21, 25), such that bacteria lacking shlB do not secrete toxin and are noncytolytic (26, 27). Mice infected with either a ΔshlA (4.91 × 10⁶ CFU) or ΔshlB (5.98 × 10⁶ CFU) mutant all survived (Fig. 1A) but lost weight at a rate equivalent to that observed for mice infected with the sublethal WT dose (∼1.0 × 10⁶ CFU) (Fig. 1B). Mice infected with the ΔshlA (8.00 × 10⁵ CFU) or ΔshlB (9.60 × 10⁵ CFU) mutant at this lower dose also all survived and experienced ∼10% weight loss through the first 48 h postinfection (Fig. 1B). These results indicate that S. marcescens causes morbidity and illness in mice in a dose-dependent manner and suggest this occurs through the activity of the ShlA cytolysin.

In humans, S. marcescens pneumonia leads to decreased pulmonary function (28). To determine how S. marcescens impacted the mouse airway, lung dynamics were measured in mechanically ventilated mice after inhalational exposure to S. marcescens at different doses. Mice inoculated with the sublethal and lethal WT inoculum dose showed significantly decreased compliance compared to that of the PBS-treated control mice at 24 and 6 h postinfection, respectively (Fig. 1C). Compliance was worse in mice that were challenged with the lethal dose at 6 h versus the sublethal dose at 24 h, although the values obtained did not reach statistical significance (P = 0.052). No trends or significant differences were observed for baseline airway resistance levels among the three cohorts (Fig. 1D). These results demonstrate that mouse S. marcescens respiratory tract infection is accompanied by gross changes in health and diminished pulmonary function, features typical of severe bacterial pneumonia.

S. marcescens infection induces inflammation and hemorrhage in mouse lungs.

Autopsies of individuals who have succumbed to S. marcescens pneumonia have shown that two distinct pathological patterns occur. In immunocompetent patients, the primary reaction was an acute, necrotizing bronchopneumonia with focal hemorrhage. In contrast, immunosuppressed neutropenic patients presented a diffuse pneumonia with extensive pulmonary hemorrhage (14, 16). To determine which outcome was mimicked by our mouse model and to assess the resulting pulmonary damage, mice were inoculated with WT S. marcescens or the ΔshlA or ΔshlB mutant at a dose that corresponded to sublethal challenge (Fig. 2). Inspection of H&E-stained lung sections showed that mice exposed to WT bacteria had substantial airway consolidation and immune cell infiltration compared to the case for mice that were inoculated with PBS. Mice infected with the ΔshlA or ΔshlB mutant had minimal damage to the lungs, with very little histopathology observed beyond 48 h. Of note, lung pathology and consolidation were dramatically elevated when mice were challenged with the WT at the lethal dose (Fig. 2). Thus, healthy mice infected with S. marcescens pneumonia displayed pathological hallmarks consistent with those observed in nonneutropenic patients, and the pathology was also bacterial dose and likely cytolysin dependent.

FIG 2.

S. marcescens-infected lungs show inflammatory cell infiltrates, consolidation, and hemorrhage. Mice were inoculated by oropharyngeal aspiration with 4.5 × 10⁶ CFU or 1.3 × 10⁶ CFU of WT S. marcescens or with 1.2 × 10⁶ CFU of ΔshlB or 1.1 × 10⁶ CFU of ΔshlA bacteria, and then animals were sacrificed at 24, 48, and 72 h for pathological evaluation of their lungs. Images demonstrate that WT-infected mice have hemorrhage, airway consolidation by immune cell infiltration, destruction of perivascular tissue, cell metaplasia, edema, and bronchiole infiltration, while mice infected with ΔshlB and ΔshlA bacteria have considerably less inflammation and tissue damage. Boxes show edema (LD, 48 h) and bronchiole infiltration (LD, 48 h). Selected bronchioles are labeled with “Br.”; arrowheads indicate blood vessels. Results are representative of 2 separate experiments with 4 or 5 mice per experiment. Bar, 60 μm.

Sublethal S. marcescens infection is rapidly cleared from the lungs but leads to robust proinflammatory cytokine production.

To further characterize the development and resolution of S. marcescens pneumonia, we measured the bacterial burden in the lungs kinetically following challenge with the sublethal dose of WT bacteria. This revealed that the S. marcescens tissue burden was greatest at 4 h postinoculation and diminished rapidly thereafter (Fig. 3A). A similar result was observed when we determined the amount of bacteria present in bronchoalveolar lavage fluid (BALF) (Fig. 3B). This confirmed that following sublethal challenge, airway-localized bacteria were most abundant at the earliest time point postinoculation and then decreased rapidly over time. In contrast, mice exposed to the lethal dose of S. marcescens had bacterial titers that increased logarithmically over time until mice succumbed to infection (Fig. 3C).

FIG 3.

Kinetics of S. marcescens infection. Mice were inoculated by forced aspiration with WT S. marcescens. Over a 72-h time course, mice were sacrificed and evaluated for lung tissue (A) and BALF (B) CFU titers following sublethal-dose challenge as well as lung CFU levels following lethal dose challenge (C). Each circle represents an individual mouse, and the bar represents the median value. Individual groups were compared using Dunn's multiple-comparison posttest: ***, P < 0.001; **, P < 0.01; *, P < 0.05. Data are representative of 2 or 3 separate experiments. †, mouse that succumbed to infection following lethal challenge with the WT.

We next evaluated the levels of 20 different cytokines and chemokines in lung homogenates from sublethally infected mice at 4, 12, 24, 48, and 72 h postinoculation (Table 1). We observed four patterns of production within infected lungs. The first pattern, which included KC, granulocyte-macrophage colony-stimulating factor (GM-CSF), interleukin-6 (IL-6), macrophage inflammatory protein 1 (MIP-1), and tumor necrosis factor alpha (TNF-α), showed levels peaking at 4 h and then rapid decline immediately thereafter. A second pattern, which included IL-1β, fibroblast growth factor (FGF), gamma interferon (IFN-γ), IL-5, IL-12, IP-10, and monocyte chemoattractant protein 1 (MCP-1), showed cytokine/chemokine levels that peaked at 12 h, also followed by a rapid decline. The third pattern, observed for IL-10, IL-1α, IL-2, and vascular endothelial growth factor (VEGF), had peak levels of production at 12 h and then a slow decrease for the remaining time points. Finally, none to very low and therefore possibly physiologically irrelevant levels of IL-4, IL-17, IL-13, and MIG were observed.

TABLE 1.

Cytokine and chemokine levels in mouse lung homogenates following S. marcescens challenge^a

^aAlternate shaded groupings indicate cytokines and chemokines with similar production profiles.

^bBold denotes values with a statistically significant difference versus the PBS control group as determined using Dunn's multiple-comparison post-Kruskal-Wallis statistic test.

S. marcescens infection induces dynamic changes in neutrophil and macrophage populations in mouse lungs.

Given the rapid upregulation of neutrophil-recruiting chemokines such as KC and MIP-1α in S. marcescens-infected lungs, we evaluated the extent of neutrophil infiltration during infection. Compared to the PBS control, visual inspection of BALF cells from low-dose WT S. marcescens-infected mice showed that infiltrates were composed largely of neutrophils, whose numbers significantly peaked at 24 h and then diminished through 72 h (Fig. 4A). Unexpectedly, monocyte numbers dropped precipitously early after infection, with almost no macrophages detected at 12 h. Monocyte numbers did not increase during the 72-h time course (Fig. 4B). Of note, macrophage depletion was not observed in mice infected with the ΔshlA or ΔshlB mutant (see Fig. S2 in the supplemental material). Thus, neutrophils are most likely the key cell involved in bacterial clearance, and ShlA was most likely responsible for macrophage depletion.

FIG 4.

Dynamic changes in lung neutrophil and monocyte populations in mice infected with S. marcescens. (A and B) Mice were inoculated with ∼1 × 10⁶ CFU of WT, ΔshlA, or ΔshlB S. marcescens or with PBS by forced aspiration. BALF was collected at the indicated time points for differential cell counting. The concentration of neutrophils (A) or monocytes (B) (the y axis is log scale) in Hema-3-stained cytospins of BALF is shown, with values for PBS shown as white circles at t = 0 h. Each circle represents an individual mouse, and the bar represents the median value. Individual groups were compared using Dunn's multiple-comparison posttest: ***, P < 0.001; **, P < 0.01. (C to F) Fluorescent images showing neutrophils (Ly6G, green) and all nucleated lung cells (DAPI, blue) present in lung tissue taken at 48 h postinoculation with PBS (C) or WT (D), ΔshlA (E), or ΔshlB (F) bacteria, all at a sublethal dose. BALF values and images shown are representative of two independent experiments.

To confirm the identify of the infiltrating cells, we performed in situ immunofluorescence analysis of infected lung tissue sections taken from animals at 48 h postinoculation using antibodies against Ly6G, a marker for neutrophils (29). Numerous Ly6G⁺ cells were detectable in the lung sections from infected animals, whereas almost none was observed in PBS-treated animals (Fig. 4C and D). Of note we observed the presence of DNA-containing (i.e., DAPI⁺) fibers extended from the aggregates of cells located within the bronchioles (Fig. 4D). We subsequently costained cells by TUNEL assay to determine whether Ly6G⁺ cells had degraded DNA typical of dying cells. The immunofluorescence examination showed a large degree of overlap between the Ly6G⁺ cells and TUNEL⁺ staining (see Fig. S1A to D in the supplemental material). The release of antimicrobial DNA and histones by dying neutrophils, or NET formation, is modulated in part by myeloperoxidase (MPO), which is itself an antimicrobial (30, 31). MPO immunoblots of lung homogenates from infected animals (1 × 10⁶ CFU) showed an increase in MPO levels as soon as 12 h after infection (see Fig. S1E in the supplemental material). These results demonstrate that neutrophils are rapidly recruited to S. marcescens-infected lungs and that recruitment is coincident with evidence of NET formation, followed by a rapid decline in bacterial titers. Mice infected with either the ΔshlA or ΔshlB mutant had neutrophils present, but we did not observe evidence of NET formation (Fig. 4E and F).

Neutropenic mice are exquisitely susceptible to S. marcescens and display extensive pulmonary hemorrhage.

Our data suggest that neutrophils are essential for eliminating S. marcescens. To determine how neutrophil deficiency affected susceptibility to S. marcescens infection and whether our experimental model might be used to examine disease that occurs in neutropenic/immunosuppressed patients (16), we depleted the mice of neutrophils using a monoclonal antibody against Ly6G and challenged them with the sublethal dose of S. marcescens. Ly6G-depleted mice developed overwhelming fulminate infection and succumbed within 24 h of challenge, with 75% mortality during the first 12 h (Fig. 5A). In contrast, PBS- and isotype antibody-treated control mice showed 0% and 10% mortality, respectively (Fig. 5A). Bacterial burden analysis of BALF showed that Ly6G-depleted mice had significantly more bacteria than PBS- or isotype control-treated mice (Fig. 5B), consistent with the notion of neutrophil-mediated protection.

FIG 5.

Neutrophils are essential for host protection against S. marcescens pneumonia. (A) Survival (n = 5 to 9 mice) of and (B) airway bacterial burden (n = 8 or 9 mice per group) in mice that were challenged by forced aspiration with 1.0 × 10⁶ CFU of WT S. marcescens 24 h after being injected intraperitoneally with a neutrophil-depleting antibody (clone Rb6-8c5, 100 mg), an isotype control antibody, or PBS are shown. For panel A, the log rank (Mantel-Cox) test gave an overall P value of <0.0001 versus PBS and the isotype control antibody. For panel B, each circle indicates one animal, bar indicates median value. Individual groups were compared using Dunn's multiple-comparison posttest: ***, P < 0.001; **, P < 0.01; *, P < 0.05.

Mutants deficient in ShlA and ShlB fail to cause lung hemorrhage.

Close examination of tissue samples from mice infected with the sublethal (Fig. 6A) and lethal (see Fig. S3A in the supplemental material) doses showed signs of extensive hemorrhage within the lungs at 48 h postinfection. Analysis of BALF from mice infected with the sublethal WT dose showed that hemorrhage, as determined by the presence of RBCs in the BALF, began as early as 12 h postinfection (Fig. 6B). In contrast, mice infected with equivalent doses of the ΔshlA and ΔshlB mutants had substantially reduced levels of visible RBCs (Fig. 6B; see Fig. S3 in the supplemental material). An important role for the cytolysin in loss of vascular integrity was independently suggested by the measurement of albumin in BALF from animals infected with the sublethal dose (Fig. 6C). These results indicate that mice exposed to inhalational S. marcescens have a major compromise of the alveolar epithelial barrier and suggest that development of hemorrhagic bronchopneumonia occurs in a cytolysin-dependent manner.

FIG 6.

S. marcescens respiratory infection causes profound vascular leakage. Mice were inoculated by forced aspiration with the indicated doses of WT S. marcescens or the ΔshlA or ΔshlB mutant and sacrificed at the indicated times for determination of the presence of red blood cells in tissue (A) or in airway lavage fluid (B) or for the presence of serum albumin in BALF (C). Albumin data show the average values from 6 to 12 mice + SEM. Kruskal-Wallis tests gave an overall P value of 0.0037 (C), while individual groups were compared using Dunn's multiple-comparison posttest: *, P < 0.05. Data are representative of two separate experiments. Bar, 60 μm.

DISCUSSION

In this study, we have described a novel mouse model of Serratia marcescens pneumonia that mimics human S. marcescens respiratory infection; we subsequently used this model to dissect aspects of bacterial pathogenesis and the host response. Mice intratracheally challenged with a clinical S. marcescens isolate via forced aspiration developed illness remarkably similar to that observed in humans with pneumonia. Features included bacterial persistence and histopathological evidence of bronchopneumonia, including extensive tissue consolidation and diminished lung function. Hallmarks of the innate immune system that were observed included high levels of proinflammatory cytokines and chemokine production, infiltration of neutrophils, and NET formation. Neutrophils proved to be critical for infection resolution. Infected mice displayed a profound pulmonary hemorrhagic response to infection, which worsened under conditions of neutropenia and at the higher infectious dose. In addition to showing concordance with human disease, the murine model proved useful for studying bacterial pathogenesis, and elucidated the likely importance of S. marcescens ShlA cytolysin and its transporter ShlB for bacterial virulence. Despite inherent limitations of the mouse model due to differences in anatomy and physiology, this model should prove extremely tractable for studying S. marcescens pneumonia and identifying critical host protective factors.

Prior studies found that mice cleared S. marcescens from their lungs, without apparent signs of illness (32, 33). Possible reasons for the difference between our findings and these prior reports include differences in the bacterial or mouse strains used, our use of a higher infectious dose, and our in-depth examination of pulmonary mechanics and tissue architecture after lower-dose inoculation. Another study did report murine hemorrhagic pneumonia following intranasal inoculation of S. marcescens, but as the study focused primarily on generation of protective immunity against a bacterial protease, details of the pneumonia model were scant (34). Therefore, our study represents the first in-depth characterization and successful use of a murine model of S. marcescens respiratory illness.

During S. marcescens pneumonia, human patients show a reduction in pulmonary function (28), which our model recapitulates. The significant decrease in lung compliance seen in S. marcescens-infected mice, similar to that observed in mice exposed to respiratory syncytial virus (RSV) and influenza virus (35, 36), indicates that lung elasticity is decreased, such that more effort is required for lung expansion. Our model explains why pulmonary mechanics are disturbed during S. marcescens infection: inflammation, tissue consolidation, and fluid accumulation all result following bacterial infection. What is more, ours is the first study to demonstrate that murine lung function is compromised following pneumonia with any Gram-negative bacterium, as prior studies have evaluated this symptom only in rat models (37, 38). Thus, this study validates the use of this apparatus to measure lung function during bacterial pneumonia in the more genetically pliable mouse model.

One striking feature of S. marcescens murine respiratory illness is the pulmonary hemorrhage, which is also observed in humans (16). It is generally assumed that alveolar hemorrhage results when infiltrating neutrophils disrupt the alveolar-capillary barrier, allowing extravasation of erythrocytes into the airways, and from neutrophil-released enzymes and reactive oxygen species (ROS) (39, 40). However, we observed that mice depleted of neutrophils showed a more pronounced pulmonary hemorrhage. This suggests that the hemorrhagic pathophysiology of S. marcescens pneumonia results from an S. marcescens process such as cytolysin production. In support of the latter hypothesis, mice infected with cytolysin-deficient mutants showed minimal to no hemorrhage. There is precedent for this notion in the literature, as ShlA has been shown to be cytotoxic for epithelial cells (24), and the streptococcal pore-forming toxin β hemolysin/cytolysin can also cause alveolar epithelial and endothelial cell injury (41, 42). Further examination of the S. marcescens ShlA cytolysin should determine how the organism uses this virulence factor to influence bacterial dissemination into the bloodstream. Importantly, our conclusions are limited by the fact that we did not include a complemented version of S. marcescens to rule out any inadvertent mutations introduced during the creation of the ΔshlA or ΔshlB mutant. While we made various attempts to create a complemented mutant to address this, we were unsuccessful.

Patients in a state of neutropenia, for example, neonates, elderly people, or those with chronic granulomatous disease, are at serious risk for complications such as meningitis, myocarditis, osteomyelitis, pneumonia, and death from S. marcescens lung infections (16, 43–45). In support of a protective role for neutrophils, we observed high numbers of infiltrating neutrophils early after infection, and neutropenic mice were highly susceptible to otherwise nonfatal infection. During bacterial pneumonia, neutrophils kill the invading bacteria by phagocytosis and release of degradative enzymes and ROS (46). NETs are another critical antimicrobial defense mechanism employed by neutrophils (47). Our immunofluorescence analysis of lung sections from infected mice showed evidence of NET formation after 48 h. We speculate that NET production may be induced by the S. marcescens ShlA cytolysin, as the shlB and shlA cytolysin mutants failed to induce NETs. Support for this idea also comes from research showing that Mannheimia haemolytica, a bovine respiratory pathogen, produces a leukotoxin capable of triggering NET formation by bovine neutrophils (48).

Alveolar macrophages (AMs) compose 95% of the immune cell population in the airways and mediate innate responses to respiratory pathogens (49, 50). The depletion of AMs early after S. marcescens infection would thus prevent production of AM cytokines and leukotrienes, effectively allowing the bacteria to establish a foothold at a lower dose of infection, such as would occur in a hospital setting. In our model, we did detect proinflammatory cytokines at early time points postinfection, suggesting that nonmacrophage lung cells are capable of compensating for the absence of AMs. We suggest that epithelial cells are responsible for the observed early cytokine production and for triggering neutrophil recruitment, as epithelial cells can produce IL-8 in response to influenza virus and IL-6, IL-1α, IL-1β, and TNF-α in response to Haemophilus influenzae (51, 52). Ongoing studies are focused on elucidating the effects of S. marcescens on epithelial cell cytokine production.

In conclusion, our results describe a murine model of S. marcescens nosocomial pneumonia that reproducibly mimics the human condition. Given the increasing frequency of antibiotic resistance in S. marcescens clinical isolates, especially against “last-resort” carbapenem antibiotics (11, 12, 53), there is an urgent need to identify novel prophylactic or therapeutic agents that are effective against nosocomial pneumonia. Our model offers a way to reproducibly model human S. marcescens pneumonia in a commonly used inbred mouse strain and is thus suitable for studies to search for lung-specific bacterial pathogenesis factors, new antibiotic/antimicrobials, and host responses during infection and recovery.