Introduction

Acinetobacter baumannii is a Gram-negative opportunistic pathogen that causes diverse infections including pneumonia, urinary tract infection (UTI), bone and soft tissue infection, and septicemia1,2,3,4,5. While becoming an increasingly more common cause of community-acquired infections, A. baumannii still primarily causes hospital-acquired infections in critically ill and immunocompromised patients, ~25% of which are polymicrobial6,7,8,9,10,11. These infections are associated with an alarming mortality rate, up to 80% in some populations, largely owing to extremely high rates of multi-drug resistance8,12,13. Notably, A. baumannii isolates exhibit the highest rates of multi-drug resistance of all Gram-negative pathogens, leading the World Health Organization to classify the bacterium at its highest priority for research and development of new treatments13,14. There is consequently an urgent need to better understand the virulence mechanisms employed by A. baumannii to guide the development of novel therapeutic approaches to combat infections.

While A. baumannii can cause a variety of infections, it is most commonly associated with pneumonia4,15. In fact, A. baumannii causes up to 10% of all hospital-acquired pneumonia (HAP) cases in the United States, highlighting its importance in clinical settings16,17. Despite this, little is known regarding the pathogenesis of this bacterium in the respiratory tract18. A major hindrance in the ability to investigate A. baumannii pneumonia is the lack of available clinically-relevant murine infection models. This is, in large part, due to the low virulence of most strains in immunocompetent mice. This is a shared feature among many pathogens that commonly cause HAP, including Pseudomonas aeruginosa and Staphylococcus aureus, for which animal models closely mimicking human infection are not available19,20. An acute infection model requiring a very high, and rather artificial, inoculum of 108–109 bacteria introduced intranasally or intratracheally is most often used to investigate these pathogens10,20,21. Wild-type (WT) mice will typically either succumb to infection or clear the organism by 72 h, thus requiring early readouts of infection such as bacterial pulmonary titers at 24–36 h. While this model may serve as a useful tool to study pathogenesis early during infection, the quick bacterial clearance does not allow for the study of bacterial virulence mechanisms at later timepoints. Importantly, A. baumannii respiratory infection in humans results in an average length of hospital stay of ~30 days, and this number is much higher in cases caused by multi-drug resistant strains, highlighting the need for a long-term infection model22,23. In this pursuit, some laboratories have used antibody or cyclophosphamide treatments to render mice neutropenic24,25,26,27,28,29. These treatments initially make mice more susceptible to A. baumannii infection, enabling the study of bacterial pathogenesis up to 7 d post-infection (dpi) using lower inoculums (~107 bacteria). However, these models do not achieve stable neutropenia in mice which leads to clearance of infection. To maintain neutropenia over longer periods, multiple injections would be necessary, which can lead to fluctuating neutrophil levels, thereby altering the overall course of disease30,31,32,33,34. In all, there is an urgent need for alternative infection models to study bacterial pathogenesis during long-term infection by relevant clinical isolates.

Previous reports have used genetically immunocompromised mice to study the role of the host immune response to A. baumannii infection. One example is mice carrying a mutation in toll-like receptor 4 (tlr4). TLR4 recognizes the lipid A moiety of bacterial lipopolysaccharide (LPS) and lipooligosaccharide (LOS), the main component of the outer membrane of most Gram-negative bacteria30,31,32. The recognition of lipid A by TLR4 triggers a signaling cascade through MyD88- or TRIF-dependent pathways, resulting in increased inflammatory cytokine and type 1 interferon production, respectively33. The role of TLR4 during A. baumannii infection has been examined in murine septicemia, acute pneumonia, UTI, and catheter-associated UTI (CAUTI) models34,35,36. In the acute pneumonia model, Knapp et al. showed that tlr4 mutant mice had increased A. baumannii CFU in the lungs with reduced inflammatory cytokines compared to WT mice35,37. Using a bloodstream infection model, Lin et al. demonstrated that WT C3H/FeJ and tlr4 mutant C3H/HeJ mice had similar bacterial burdens34,38. However, all WT mice succumbed to infection by day 4, whereas all tlr4 mutant mice survived. This could be attributed to WT mice experiencing septic shock associated with increased inflammatory cytokines. Finally, in a UTI model, our laboratory found that tlr4 mutant C3H/HeJ mice were more susceptible to infection than WT C3H/HeN mice36. Moreover, we found that C3H/HeJ mice in the UTI model formed small intracellular populations in urothelial cells referred to as Acinetobacter baumannii intracellular reservoirs (ABIRs), which could seed a recurrent infection upon catheterization at higher rates relative to WT mice. In addition to playing a significant role during murine infection, TLR4 is relevant in clinical settings as well. In fact, numerous studies have identified links between tlr4 polymorphisms and infection outcomes from A. baumannii pneumonia in humans39,40,41. In all, these studies demonstrate the key role of TLR4 in controlling A. baumannii infection and disease progression and highlight the clinical relevance of the associated signaling cascade.

In this work we describe a novel murine model of A. baumannii pneumonia that employs tlr4 mutant mice and low bacterial inoculums (105 bacteria). Using this model, we show that A. baumannii can establish chronic infection lasting several weeks, which is more clinically relevant than shorter-term models. We additionally demonstrate that our model enables the discovery of virulence factors not detectable in the acute infection model. Finally, we illustrate how our model can be employed to assess the efficacy of antibiotics over the course of infection and investigate polymicrobial infections.

Results

tlr4 mutant mice are susceptible to chronic infection at low inoculums

To assess if tlr4 mutant mice could serve as permissive hosts for long-term respiratory infection, we performed intranasal inoculations of WT (C3H/HeN) and tlr4 mutant (C3H/HeJ) mice with a range of inoculums of a modern A. baumannii respiratory isolate, G636, and sacrificed groups of mice every 3 days starting at 24 h post-infection (hpi). At the highest inoculum of 108 colony forming units (CFU), WT mice cleared infection by day 4, consistent with previously published results using the acute pulmonary infection model (Fig. 1A)21. tlr4 mutant mice infected with 108 CFU also cleared infection relatively early after inoculation, with most mice having no detectable bacteria in the lungs by ~7 dpi. Strikingly, while WT mice infected with a lower dose of 105 bacteria cleared infection after 1 day, tlr4 mutant mice maintained detectable bacteria in the lungs out to the latest timepoint tested in this experiment, 19 dpi (Fig. 1D). A subsequent extended experiment revealed that some tlr4 mutant mice infected with 105 G636 maintain detectable levels of CFU in the lungs even out to 3–4 weeks pi (Fig. S1). Despite this long infection course in tlr4 mutant mice at this dose, dissemination to distal organs was rarely detected. This is consistent with the clinical manifestations of non-ventilator A. baumannii pneumonia, as less than 20% of patients will develop subsequent bacteremia42. Notably, by employing confocal microscopy, we were able to visualize bacteria in tlr4 mutant mice infected with the 105 inoculum; at early timepoints (4 hpi and 2 dpi), bacteria were identified inside cells in the bronchoalveolar lavage fluid (BALF), as well as extracellularly, consistent with our previously published results in the acute infection model (Fig. S2)43. We also screened other inoculums of G636 in tlr4 mutant mice above and below 105 CFU (Fig. S3). As inoculums increased above 105 CFU however, we noted quicker time to bacterial clearance; mice infected with 106 and 107 CFU began clearing infection by 10 and 13 dpi, respectively. tlr4 mutant mice infected with inoculums lower than 105 CFU also gave less optimal results; the 104 inoculum resulted in slightly earlier clearance, and the 103 inoculum resulted in inconsistent infection. Consequently, we selected the 105 CFU inoculum as the optimal dose for studying long-term infection.

Fig. 1: Low inoculums of modern respiratory A. baumannii clinical isolates result in chronic lung infection in tlr4 mutant mice.
Fig. 1: Low inoculums of modern respiratory A. baumannii clinical isolates result in chronic lung infection in tlr4 mutant mice.
Full size image

Groups of female C3H/HeN (WT) or C3H/HeJ (tlr4 mutant) mice were intranasally inoculated with 108 G636 (A), 108 G654 (B), 108 Ab19606 (C), 105 G636 (D), 105 G654 (E), or 105 Ab19606 (F). Beginning at 24 hpi, groups of mice were sacrificed every 3 days, and bacteria in the lungs were quantified. Each data point indicates an individual mouse, and the connecting line intersects each timepoint at the mean. The limit of detection (10 CFU) is indicated by the dashed line. Source data are provided as a Source Data file.

We next evaluated whether a second modern A. baumannii respiratory isolate, G654, exhibits similar behavior to G636 (Figs. 1E and S1). As with G636, detectable levels of bacteria were present in the lungs out to 19 dpi with the 105 inoculum in tlr4 mutant mice, while WT mice cleared this inoculum within 1 day. At the higher inoculum of 108 CFU, G654 infection was inconsistent in tlr4 mutant mice and infection was cleared early after infection in WT mice (Fig. 1B). Finally, we screened a third modern respiratory isolate, G803, which is phylogenetically and geographically unrelated to G636 and G654, for its ability to establish long-term infection with a 105 inoculum in tlr4 mutant mice. G803 was also able to establish long-term infection, as CFU were recovered from the lungs up to 16 dpi (Table S1 and Fig. S4A).

Given possible genotypic and phenotypic differences between older and modern isolates, we also screened commonly employed older, lab-domesticated strains (Ab19606 and Ab17978) to see if these can similarly persist in tlr4 mutant mice at low inoculums. In contrast to infection kinetics observed with the modern isolates, Ab19606 infection was cleared after 1 day regardless of inoculum size or mouse background (Fig. 1C, F). Alternatively, Ab17978 was able to establish persistent infection in tlr4 mutant mice with the low inoculum out to the latest timepoint tested for this strain, 7 dpi, with similar kinetics to those of the modern respiratory isolates tested (Fig. S4B). This finding that Ab17978 and Ab19606 have differential infection phenotypes highlight strain-dependent differences among A. baumannii isolates and is in agreeance with other infection models classifying Ab19606 as a low virulence strain44,45.

In all, these results indicate that 3 modern A. baumannii respiratory isolates can cause infection out to ~3 weeks pi at lower and likely more clinically-relevant inoculums than previously used in the literature. Importantly, this infection duration noted here with low inoculums of modern respiratory isolates in tlr4 mutant mice is the longest reported for A. baumannii in any animal model to date. We therefore chose to further characterize these conditions as a model to study pulmonary pathogenesis, referred to hereafter as the “chronic respiratory infection model.”

Lower A. baumannii inoculums result in a decreased immune response in tlr4 mutant mice

Given the unexpected result that tlr4 mutant mice exhibit chronic infection at lower inoculums, while WT and tlr4 mutant mice clear infection at higher inoculums, we sought to characterize the host immune response in these different conditions. We intranasally infected groups of WT and tlr4 mutant mice with 105 or 108 bacteria or mock infected them with phosphate-buffered saline (PBS). Then, at early timepoints of 4 hpi and 2 dpi and a later timepoint of 7 dpi, BALF was collected for immune cell quantification (Figs. 2 and S5). Regarding alveolar macrophages (AMs), initial reductions were noted at early timepoints with the large inoculum. However, AM numbers rebounded by 2 dpi and minimal differences were noted after this timepoint (Fig. 2A–C). In WT mice, the number of polymorphonuclear leukocytes (PMNs) was increased with the higher inoculum relative to lower and mock inoculums at every timepoint (Fig. 2D–F). This result at 4 and 48 hpi aligns with previous reports comparing A. baumannii acute murine pulmonary infection to mock infection46,47,48,49,50,51,52,53. However, the result at 7 dpi that neutrophils were still significantly elevated in WT mice relative to mock-infected controls diverges from some previous reports that indicate that neutrophil levels of WT mice regress to those of mock-infected mice following 48 hpi47,50. However, other reports showed still increasing levels of neutrophils at 72 hpi, with one showing a decrease only after this timepoint52,53. These differences can likely be attributed to use of different mouse (C3H/HeN, C57BL/6, A/J) or A. baumannii (G636, ATCC17961, ATCC BAA-1605, A112-II-a) strains in these previous papers. However, neutrophil levels are still decreasing following 48 h of high inoculum infection in C3H/HeN mice in line with these previous studies (~2 log10 between 2 and 7 dpi). At the higher inoculum, WT mice also had increased PMNs relative to tlr4 mutant mice early during infection in line with previous results35. Interestingly, at the lower inoculum, while PMN counts trended higher at 4 hpi for WT mice relative to tlr4 mutant mice, no significant differences were noted between these groups at any timepoint. This result suggests that, despite neutrophil influx being a predominant mechanism of A. baumannii clearance, PMN numbers alone may not account for differences in clearance between WT and tlr4 mutant mice at the lower inoculum47,50,53,54,55,56,57.

Fig. 2: Lower intranasal A. baumannii inoculums result in reduced lung neutrophil influx.
Fig. 2: Lower intranasal A. baumannii inoculums result in reduced lung neutrophil influx.
Full size image

Groups of female C3H/HeN (WT) or C3H/HeJ (tlr4 mutant) mice were intranasally inoculated with 105 G636, 108 G636, or mock inoculated with PBS. At 4 h (A, D), 2 d (B, E), and 7 d (C, F) pi, alveolar macrophages (AMs) (A–C) and polymorphonuclear leukocytes (PMNs) (D–F) in the BALF were enumerated by flow cytometry. Shown are pooled results from at least two independent experiments, and each data point represents an individual mouse. The horizontal line represents the mean, and the standard error of the mean (SEM) is indicated by error bars. *p < 0.05; two-way analysis of variance (ANOVA), Tukey’s test for multiple comparisons. Shown are statistically different significances between different strains at the same inoculum and between different inoculums given to the same strain. Source data and statistical test details are provided in the Source Data file. p values: 4 hpi AMs WT, 105 vs. Mock = 0.0242; 4 hpi AMs tlr4 mutant 105 vs. 108 = 0.0003; 7 dpi AMs, WT Mock vs. 108 = 0.0272; 4 hpi PMNs, WT Mock vs. 108 = 0.0037; 4 hpi PMNs, WT 105 vs. 108 = 0.0020; 4 hpi PMNs, WT 108 vs. tlr4 mutant 108 = 0.0402; 2 dpi PMNs, WT Mock vs. 108 = < 0.0001; 2 dpi PMNs, WT 105 vs. 108 = < 0.0001; 2 dpi PMNs, tlr4 mutant Mock vs. 108 = 0.0190; 2 dpi PMNs, tlr4 mutant 105 vs. 108 = 0.0313; 7 dpi PMNs, WT Mock vs. 108 = < 0.0001; 7 dpi PMNs, WT 105 vs. 108 = < 0.0001; 7 dpi PMNs, WT 108 vs. tlr4 mutant 108 = < 0.0001.

Given the established importance of PMNs in A. baumannii infection, we decided to further assess the role of PMNs in the chronic infection model. We rendered tlr4 mutant mice neutropenic with cyclophosphamide prior to infection with 105 G636 via a previous established protocol by Manepalli et al.29 (Fig. S6). Neutropenic tlr4 mutant mice had an initial bacterial burden of 107 CFU at 1 dpi. This is 100-fold higher than numbers seen in the chronic infection model at the same timepoint (see Fig. 1). As expected though, as neutrophil numbers rebound following cyclophosphamide treatment (3–4 days post-treatment), bacterial CFU began to approach levels similar to those observed in the chronic infection model29,58,59. This result implies that while PMNs are not effective in efficiently clearing infection in the chronic infection model, they still maintain a role in controlling bacterial proliferation.

To further evaluate the host response, we quantified 13 common inflammatory cytokines in the BALF (Table S2). At the higher inoculum, WT and tlr4 mutant mice exhibited significantly increased levels of IL-1α, IFN-γ, TNF-α, MCP-1, IL-1β, IL-6, and IL-17A early during infection relative to the lower inoculum while levels dissipated by 7 dpi, consistent with bacterial clearance (see Fig. 1A). WT mice infected with the high inoculum had significantly increased levels of IFN-β relative to tlr4 mutant mice at 4 hpi and increased levels of IL-1α, IFN-γ, TNF-α, MCP-1, IL12-p70, IL-1β, IL-6, IL-27, and IL17A at 2 dpi, likely leading to the earlier clearance observed. At the lower inoculum, although WT mice clear infection within nearly 24 h and tlr4 mutant mice maintain infection out to at least 3 weeks, minimal significant differences in inflammatory cytokines were observed (see Fig. 1D). In fact, the only significant difference noted was the increased levels of GM-CSF at 4 hpi in WT mice relative to tlr4 mutant mice. Other inflammatory cytokines that trended higher at early timepoints in WT mice at the lower inoculum include the inflammasome-associated cytokines IL-1α and IL-1β, as well as TNF-α and IL-6. These elevated levels of inflammatory cytokines early during infection in WT mice could possibly account for the earlier clearance. Later during infection, however, tlr4 mutant mice had elevated, albeit not significantly higher, amounts of TNF-α, IL-1α, and IL-6 relative to WT mice, consistent with persistent infection.

The chronic respiratory infection model results in alveolitis and airway epithelium attenuation

We next assessed if the chronic respiratory infection model is associated with lung pathology. tlr4 mutant mice were infected with 105 G636 or mock infected with PBS, and mice were sacrificed at 4 hpi, 2 dpi, 7 dpi, 14 dpi, and 21 dpi. Lungs were then sectioned, stained with hematoxylin and eosin (H&E), and scored for pathological changes by a licensed veterinary pathologist (Fig. 3). The chronic model resulted in a significant increase in alveolitis at 2 dpi (Fig. 3A) and airway epithelium attenuation at 7 and 21 dpi (Fig. 3B) relative to mock-infected mice. Moreover, trends toward higher levels of these phenotypes were observed at most timepoints (Fig. 3C–E).

Fig. 3: The chronic respiratory infection model results in lung pathology.
Fig. 3: The chronic respiratory infection model results in lung pathology.
Full size image

Groups of female C3H/HeJ (tlr4 mutant) mice were inoculated with 105 G636 or mock-inoculated with PBS, and at 4 hpi, 2 dpi, 7 dpi, 14 dpi, and 21 dpi, lungs slices were prepared, H&E stained, and scored for alveolitis (A) or airway epithelium attenuation (B). Shown are three biological replicates with 2/3 mice/replicate (n = 6–7). The bar represents the mean, each mouse is indicated by a dot, and the SEM is indicated by error bars. *p < 0.05; two-way ANOVA, Bonferroni’s test for multiple comparisons. Representative tissue sections of mock-infected tissue (C) and alveolitis (D) and attenuated airway epithelium (E) in infected tissue are shown. Scale bars = 500 µm and inset scale bars = 200 µm.

We also examined other potential inflammatory pathologies in these experiments, revealing alveolar histiocytosis, perivascular infiltrate, peribronchiolitis, bronchoalveolar lymphoid tissue (BALT), and pleuritis in both infected and mock-infected animals (Fig. S7). These findings were not surprising, however, as tlr4 mutation results in changes in lung development and morphology, such as airspace enlargement and emphysema60. Moreover, naïve C3H/HeJ mice have been shown to have basal increased levels of inflammation relative to wild-type C3H background mice, including increased terminal airspace, inflammatory cytokines, and immune cells possibly due to colonization by environmental bacteria61. Although we did identify pathologies associated with the chronic infection model, these basal pathologies in C3H/HeJ mice may lead to an inability to identify other potential pulmonary damage associated with infection. Finally, we noted no incidence of airway smooth muscle hypertrophy, airway squamous epithelium metaplasia, airway goblet cell hyperplasia, and fibrosis in infected or mock-infected mice. In all, these results demonstrate that the chronic infection model results in significant alveolitis and airway attenuation, indicative of chronic infection.

InvL is a critical virulence factor for long-term infection

The acute pulmonary infection model has been widely used to characterize A. baumannii virulence factors21. While the acute model is valuable for identifying bacterial proteins required at early timepoints, these mice clear infection within 3-4 dpi, not allowing for the identification of factors required for prolonged infection. As a proof of principle, we sought to determine if the chronic respiratory infection model could identify proteins required for bacterial persistence in the lungs. We hypothesized that prolonged adherence to respiratory epithelium would be required for persistence, so we first tested individual mutants lacking previously identified A. baumannii adhesins (Bap, Ata, FhaBC, and InvL) for attenuation in the chronic infection model (Fig. S8)62,63,64,65,66,67,68,69,70,71. tlr4 mutant mice were infected with 105 G636 WT or mutant bacteria, and mice were sacrificed at 1 or 14 dpi for lung colony-forming units (CFU) quantification. This experiment indicated a possible role for InvL in long-term infection as mice began to clear bacteria in the lungs by 14 dpi. We also performed more extensive analyses with G636 WT, invL mutant (ΔinvL), and complemented invL mutant (invL+) strains in the chronic infection model, sacrificing mice at 1, 7, 14, and 21 dpi to quantify CFU in the lungs (Fig. 4A–D). At 1 dpi, no defect was identified between WT and mutant-infected mice. However, by 7 dpi, mutant-infected mice had significantly reduced CFU. This significant reduction in CFU was also observed at 14 and 21 dpi. At 7 and 14 dpi, genetic complementation partially rescued the phenotype, as no significant difference was detected between WT and complemented strains, although the complemented mutant did not have significantly higher CFU than the mutant strain. On day 21, full genetic complementation was observed. As high variability within groups was observed during long-term infection, we also examined each of the three pooled experiments from Fig. 4 using WT, ΔinvL, invL+ strains individually (Fig. S9) to assess if results were consistent among biological replicates. Indeed, early during infection in these 3 independent experiments, minimal differences were seen with the ΔinvL mutant strain relative to the WT strain. By 21 dpi, all mice in 2/3 groups infected with ΔinvL mutant had cleared infection. Alternatively, all groups infected with WT and invL+ strains had mice still infected at that timepoint. In all, these results indicate that InvL is specifically important for long-term infection in the chronic model.

Fig. 4: InvL is a critical virulence factor for long-term respiratory infection, but dispensable in the acute infection model.
Fig. 4: InvL is a critical virulence factor for long-term respiratory infection, but dispensable in the acute infection model.
Full size image

Female C3H/HeJ (tlr4 mutant) mice were infected with 105 G636, G636 ΔinvL, or G636 invL+. Groups of mice were then sacrificed at 1 dpi (A), 7 dpi (B), 14 dpi (C), and 21 dpi (D), and CFU in the lungs were quantified. Shown are the pooled results from three independent experiments. For the acute infection model, groups of female C57BL/6 mice were infected with 109 G636, G636 ΔinvL, or G636 invL+. 24 hpi, mice were sacrificed, and CFU in the lungs (E), spleen (F), and kidneys (G) were enumerated. Shown are the pooled results of two independent experiments. Each data point represents an individual mouse, the horizontal line represents the mean, and the SEM is indicated by error bars. The limit of detection (10 CFU) is indicated by the dashed line. *p < 0.05; Kruskal–Wallis H test with Dunn’s test for multiple comparisons; ns not significant. Source data and statistical test details are provided in the Source Data file.

We next compared results from the chronic respiratory infection model to the acute infection model. We infected C57BL/6 mice with 109 G636 WT, ΔinvL, or invL+ strains. 24 hpi, mice were sacrificed, and CFU in the lungs, spleens, and kidneys were quantified (Fig. 4E–G). As opposed to results seen in the chronic infection model, the ΔinvL mutant had no significant defect in bacterial load in the lungs. Additionally, no defect was noted in dissemination to the spleen and kidneys, indicating that InvL is dispensable in the acute infection model. In all, these results highlight the differences in required bacterial genes between these disparate pulmonary infection models and show the importance of continuing to explore models that can better approximate clinical disease. Additionally, these experiments establish InvL as the first known A. baumannii virulence factor required for long-term infection.

The chronic infection model can be used to study the outcome of antibiotic treatment

The acute pulmonary infection model has been employed extensively to assess effects of antibiotic treatment21. However, this model only allows us to estimate the efficacy of antibiotics by measuring the initial reduction in the bacterial burden at 24–36 hpi due to rapid bacterial clearance by the host. A clear limitation of this model is that it does not inform if bacterial infection is cleared, or if persistent bacteria remain in the lung. The chronic respiratory infection model therefore represents a novel platform that could be used to track the kinetics of A. baumannii clearance due to antibiotic treatment. As a proof of principle, we assessed the effect of tigecycline, colistin, and imipenem in the chronic model with strains G636 and G654 at antibiotic concentrations similar to those previously used to determine treatment efficacy in mice (Figs. 5A, B and S10A–D)72,73,74,75,76,77,78. We additionally assessed the effect of apramycin, a drug with demonstrated efficacy and safety in mice that is currently in Phase I clinical trials for use in humans (Fig. 5C, D)79,80,81. Minimum inhibitory concentrations (MICs) for G636 and G654 for these and other commonly used antibiotics are listed in Table S3. Colistin was ineffective for both strains in the acute infection model, while initial reductions in CFU were noted in the chronic infection model (Fig. S5A, B). However, bacterial numbers appeared to stabilize over time in the chronic infection model, consistent with the development of bacterial persisters (discussed below). Imipenem showed limited efficacy against both strains in both models (Fig. S10C, D), as expected given the strains’ resistance in vitro (Table S3). At 24 hpi, tigecycline and apramycin treatment resulted in initial reductions in CFU in both the chronic and the acute infection models relative to PBS-treated mice (Fig. 5A–D). However, the chronic model enabled us to differentiate the efficacy of both antibiotics at later times. Apramycin treatment ultimately led to clearance after 3–5 days, demonstrating the efficacy of this antibiotic. However, with tigecycline treatment, although there were initial reductions in CFU, bacterial numbers leveled out over time indicative of treatment failure. The behavior of bacteria in presence of tigecycline over time is consistent with the development of persisters. Notably, the efficacy of tigecycline and apramycin against A. baumannii cannot be distinguished at 24 hpi. These results indicate that the chronic model can be used to determine outcome of infection with therapeutic intervention, a significant advantage over the currently employed acute infection model.

Fig. 5: The chronic respiratory infection model can be used to study outcomes of antibiotic treatment.
Fig. 5: The chronic respiratory infection model can be used to study outcomes of antibiotic treatment.
Full size image

Groups of female C3H/HeJ (tlr4 mutant) mice were infected with 105 G636 (A, C) or 105 G654 (B, D) and sacrificed at 1, 3, and 5 dpi (long-term). Additionally, groups of female C57Bl/6 mice were infected with 109 G636 (A, C) or 109 G654 (B, D) and sacrificed at 24 hpi (acute). Mice in both infection models were treated intraperitoneally with PBS or 100 mg/kg tigecycline (tig) every 12 h (A, B) or PBS or 500 mg/kg apramycin (apr) every 12 h (C, D) with all treatments beginning 4 hpi. At each timepoint, CFU were quantified in the lungs. Shown are the pooled results from two independent experiments, each data point represents the mean, and the SEM is represented by error bars. The limit of detection (10 CFU) is indicated by the dashed line. *p < 0.05; two-tailed Mann–Whitney U test. Source data and statistical test details are provided in the Source Data file.

Use of the chronic infection model to study bacterial co-infections reveals that Staphylococcus aureus exacerbates ongoing A. baumannii infection while Klebsiella pneumoniae leads to earlier clearance

Approximately 25% of A. baumannii pulmonary infections are polymicrobial, and two of the most commonly co-infecting pathogens are Staphylococcus aureus and Klebsiella pneumoniae9. We thus sought to assess the impact of secondary infections with these two bacteria on the outcome of A. baumannii infection in the context of the chronic respiratory infection model. For these experiments, we first established a primary A. baumannii infection by inoculating tlr4 mutant mice with 105 CFU of strain G636. Following 14 days of A. baumannii infection, we inoculated mice with 5 × 107 CFU of S. aureus strain Newman or K. pneumoniae strain TOP52, mock-treated mice with PBS, or left mice untreated. One and 2 days post-secondary infection, mice were sacrificed, and bacterial CFU were quantified in the lungs, spleens, and kidneys (Figs. 6 and S11, S12). Secondary infection with S. aureus led to a resurgence of A. baumannii CFU in the lungs of many mice, though the overall mean CFU in these mice were not significantly different from mock-infected and untreated groups (Fig. 6A, D, E). A. baumannii were also identified in the spleens and kidneys of some mice that received the secondary S. aureus infection, even though A. baumannii bacteremia rarely occurs in the context of this chronic respiratory infection model (Fig. S11). Additionally, S. aureus trended toward increased numbers in the lungs, spleens, and kidneys in the context of polymicrobial infection with A. baumannii relative to monomicrobial infection (Fig. S12A–C). Contrarily, secondary infection with K. pneumoniae significantly decreased A. baumannii CFU in the lungs relative to mock-infected and untreated groups at 1 dpi (Fig. 6A, D, E). Additionally, polymicrobial infection with A. baumannii and K. pneumoniae resulted in significantly reduced K. pneumoniae CFU recovered in the lungs at 1 and 2 dpi, as well as in the spleens and kidneys at 2 dpi relative to K. pneumoniae monomicrobial infection (Fig. S7D–F). Although understanding the interactions between these bacteria is beyond the scope of this work, these experiments indicate that S. aureus exacerbates A. baumannii infection, while K. pneumoniae attenuates infection in the context of the chronic infection model. Additionally, these results demonstrate the ability of the model to be used to study longer-term aspects of polymicrobial interactions that were not previously able to be done with the acute infection model.

Fig. 6: Bacterial secondary infection alters the course of chronic A. baumannii pneumonia.
Fig. 6: Bacterial secondary infection alters the course of chronic A. baumannii pneumonia.
Full size image

Female C3H/HeJ (tlr4 mutant) mice were intranasally inoculated with 105 G636, and groups of mice were sacrificed at 1, 7, and 14 dpi. At 14 days post-A. baumannii infection, groups of mice were either not inoculated (untreated), inoculated with PBS (mock-infected), infected with S. aureus, or infected with K. pneumoniae. Subsequently, on days 15 and 16 post-A. baumannii infection (1 and 2 days post-secondary infection), mice were sacrificed, and A. baumannii CFU were quantified in the lungs (A, D, E), spleen (B), and kidneys (C). In A–C, each data point represents the mean, the SEM is represented by error bars, and the limit of detection (10 CFU) is indicated by the dashed line. In D and E, each data point represents an individual mouse, the horizontal line represents the mean, and the SEM is indicated by error bars. Shown are the pooled results from at least two independent experiments. *p < 0.05; Kruskal–Wallis H test with Dunn’s test for multiple comparisons. The limit of detection (10 CFU) is indicated by the dashed line. Source data and statistical test details are provided in the Source Data file.

Discussion

A. baumannii has emerged as a significant cause of nosocomial pneumonia and is of major clinical importance due to its extremely high rates of multidrug resistance8,12,13. Despite this, our understanding of A. baumannii respiratory pathogenesis is hindered by a shortage of clinically relevant infection models. Here, we aimed to address this significant gap in the field by developing a novel respiratory infection model. In this pursuit, we found that, at likely more clinically-relevant inoculums, tlr4 mutant mice maintain long-term respiratory infections by A. baumannii. We then demonstrate the versatility of this model which enabled (1) the identification of a bacterial virulence factor required for long-term respiratory infection, which is not required in acute models, (2) the study of kinetics of bacterial clearance upon treatment with clinically-relevant antibiotics, and (3) the exploration of the impact of secondary infections with two commonly co-isolated respiratory pathogens.

Importantly, use of a long-term immunocompromised mouse model in this manuscript maintains clinical relevance, as A. baumannii primarily affects immunocompromised and critically ill patients, causing extended infections particularly in patients in intensive care units (ICUs)7,8,10,11. Furthermore, studies have indicated a significant link between human tlr4 mutations and poor clinical outcomes from A. baumannii infection, further supporting the use of a tlr4 mutant murine model to study pathogenesis39,40,41. For example, Behairy et al. demonstrated that tlr4 polymorphisms were associated with significantly increased A. baumannii infections, but not infections with other Gram-negative pathogens, in Egyptian ICUs39. Similarly, Chatzi et al. showed that tlr4 polymorphisms were associated with increased risk of A. baumannii infections in Greek ICUs40. Finally, using the large study population of the Chinese National Hospital Infection Surveillance Network, He et al. demonstrated a significant link between tlr4 mutations and A. baumannii infection41.

While use of tlr4 mutant mice enables long term A. baumannii pulmonary infection and has clear relevance to human disease, it is important to note that the chronic infection model described here has limitations. First, it is not possible for one model to reflect the state of all populations that are susceptible to A. baumannii infection and all associated diseases/outcomes; A. baumannii is capable of infecting individuals with a range of immunocompromisation and infection duration is often the result of multi-drug resistance in combination with immunocompromised status82. Moreover, A. baumannii can cause a range of disease from mild to severe pneumonia depending on the individual83. Therefore, careful consideration should be taken when choosing a murine model, especially in the case of studying host responses which will drastically differ depending on if immunosuppression is used and the mechanism of immunosuppression. Related to this limitation, an important aspect of this model that likely enabled bacterial persistence was the decreased immune response associated with low inoculums in tlr4 mutant mice. In the case of studying exacerbated/lethal host responses that can occur in severe human infections and pathologies associated with specific immune mechanisms, future studies may need to employ mice lacking specific immune factors of interest rather than TLR4. For these circumstances, the previously described acute infection model using C57BL/6 mice and congenic mutants will likely still serve as a valuable tool21. Second, use of a tlr4 mutant murine model may impact potential bacterial virulence mechanisms that are able to be studied. For example, when studying bacterial mutations potentially altering bacterial LOS, a tlr4 mutant mouse model may have limited use. A third limitation of this model is that its use may be strain-dependent. While we established that 3 modern respiratory isolates and one lab-domesticated strain, Ab17978, establish long-term infection, another lab-domesticated strain, Ab19606, was cleared by 24 hpi. Therefore, strain-to-strain variability could affect the ability to use the model, and preliminary experiments may be required to adapt the model to strains of interest. While the chronic infection model does have some limitations, similar to all murine models, it is the first model enabling the study of A. baumannii persistence past 1 week pi. It therefore has the potential to change how we address questions associated with specifically long-term infection, which is extremely relevant in the case of difficult to treat multidrug resistant infections due to treatment failures. Future work, however, will be aimed at identifying potential other methods of extending infection in mice that can be used in conjunction with the chronic infection model reported here. Such alternatives include the use of aged mouse models (klotho mutation) or mixing bacteria with gastric mucin, both of which have been shown to extend infection kinetics in sepsis models84,85,86,87,88,89.

In this study, we found that InvL is required for chronic infection, and, more importantly, at the later stages of infection. However, InvL was dispensable in the context of the acute infection model. There are multiple possible reasons for this discrepancy. First, the massive bacterial dose required for the acute infection model may mask potential defects that can now be detected with a smaller, more clinically-relevant inoculum. This is unlikely, as WT and invL mutant bacteria behave similarly at early time points in our model. An alternative reason could be that adhesins required early during infection/interaction with the healthy airway differ from those required during persistent interaction with a more inflamed or damaged airway. It is well-established that the airway extracellular matrix (ECM) is altered by bacterial infection, lung damage, and/or inflammation90,91. Long-term lung damage and inflammation results in increased fibronectin, collagen, laminin, and fibrinogen in the ECM92,93,94,95,96,97. Moreover, specific pathogens elicit different inflammatory responses, resulting in distinct changes to the lung ECM. For example, in an acute mouse model of pneumonia, Pseudomonas aeruginosa induces versican deposition in the lungs, while Escherichia coli induces robust versican and hyaluronan deposition98,99. We previously showed that InvL can bind α5β1 integrin, collagen V, and fibrinogen71. However, whether A. baumannii infection or the associated inflammation induces production of these protein(s) during pulmonary infection is unknown. Future work will investigate this possibility, as well as assess which InvL-host protein interactions are essential for chronic infection.

Herein, we demonstrate the potential to use the chronic respiratory infection model to study the efficacy of antibiotic treatments over time. One intriguing finding from these experiments is that with antibiotics such as colistin and tigecycline, an initial decrease in CFU (~10–100 fold) recovered from the lungs at 1 dpi was observed. However, following this decrease, the number of bacteria in the lungs appeared to stabilize over time. It is tempting to speculate that this is the result of the formation of bacterial persisters, defined as bacterial cells that become tolerant to antibiotics despite undergoing no genetic changes100,101,102. Importantly, the commonly used acute infection model does not allow for the study of bacterial persisters due to the short time course of the model. Given that persister cells represent a major cause of treatment failure and chronic infection, the chronic infection model presented here represents a unique platform that is desperately needed to understand this aspect of A. baumannii pathogenesis. Furthermore, our model offers new possibilities to study efficacy of novel antibiotics in murine models before committing to expensive clinical trials.

While a significant portion of A. baumannii infections are polymicrobial, the acute infection model has limitations for use with polymicrobial infections. First the quick clearance of the bacteria usually only allows inoculation at a single timepoint, thus not enabling investigation of secondary infections. Second, the high required infectious dose often means that typical inoculums for bacteria used in these experiments must be adjusted, so mice do not succumb to infections at early timepoints. Here, we applied the chronic respiratory infection model to assess the result of secondary infection with two pathogens commonly co-isolated with A. baumannii, S. aureus and K. pneumoniae. We found opposite results with these different bacteria; S. aureus secondary infection trended toward exacerbation of A. baumannii infection, while K. pneumoniae secondary infection led to reduced A. baumannii numbers. The potential synergism of A. baumannii and S. aureus in the chronic infection model aligns with previous reports. For example, using a Tn-Seq-based approach, Li et al. demonstrated that the 49% of genes required by S. aureus for monomicrobial infection in a murine systemic infection model became non-essential upon A. baumannii co-infection103. Another recent report showed that S. aureus can support A. baumannii growth in vitro by providing acetoin as a carbon source104.

Although we found that K. pneumoniae secondary infection led to reduced A. baumannii numbers in the lungs in the chronic infection model, one study has shown that K. pneumoniae could cross-feed A. baumannii through products of sugar fermentation in vitro and demonstrated that co-infection led to reduced survival of Galleria mellonella relative to monomicrobial infection with either pathogen105. This, in part, shows that these two pathogens can have beneficial interactions for the bacteria. There are two potential reasons however for the reduction of CFU for both bacteria in the context of the chronic infection model reported here; (1) bacterial competition or (2) the host response to the secondary infection. Regarding bacterial competition, there have been several lines of evidence pointing to direct bacterial killing between diverse A. baumannii and K. pneumoniae strains mediated by the type VI secretion system106,107,108,109. In addition to direct killing, this bacterial competition could be indirect as well, as both A. baumannii and K. pneumoniae may be competing for similar nutrients in the lung microenvironment. With respect to the immune response, a difference between this work and the above study is that the microenvironment encountered in the mammalian lung is not perfectly modeled by the wax moth110. Our results may therefore be the result of TLR4-independent host response elicited by the combination of both bacteria that is not recapitulated by a G. mellonella model. While understanding the precise mechanism behind the in vivo interactions between A. baumannii and commonly co-isolated pathogens is outside the scope of the current study, these results highlight the practicality of applying the chronic respiratory infection model to better understand polymicrobial infections.

In this study, we have validated several different uses for the chronic respiratory infection model. However, there are also other potential uses for this model that were not previously investigable. For example, we can now perform experiments differentiating between virulence factors required for establishment of infection and factors required for maintenance of infection, assessing bacterial evolution during long-term infection, investigating changes in the pulmonary microbiome due to infection over time, and analyzing the long-term outcomes of novel therapies such as newly developed phage cocktails. Additionally, while this model was initially developed to study Acinetobacter respiratory infections, it has the potential to be applied to research with other respiratory pathogens in cases where suitable animal models are lacking. In all, this work describes the longest-term infection model available to investigate A. baumannii host-pathogen interactions to date, which will ultimately aid in the development of novel therapeutics to combat infection by this increasingly multidrug-resistant bacterium.

Methods

Ethical regulations

This research complies with all relevant ethical regulations and has been approved by the Washington University School of Medicine Institutional Animal Care and Use Committee (IACUC; protocol number: 23-0071) and Institutional Biosafety Committee (IBC; protocol number: 12098 Ver.25).

Bacterial plasmids, strains, and growth conditions

Plasmids and strains used in this study are detailed in Table S4. Bacterial cultures were grown at 37 °C in Lennox broth/agar supplemented with 10 μg/mL chloramphenicol, 50 μg/mL apramycin, 100 μg/mL ampicillin, 50 μg/mL kanamycin, 10 μg/mL tetracycline, or 10% sucrose when appropriate. All strains used in this manuscript will be made available upon request.

Murine pneumonia models

All animal experiments were approved by the Washington University Animal Care and Use Committee, and we have complied with all relevant ethical regulations. For all rodent housing rooms, the light cycle is kept at 12 h:12 h light:dark, with lights on at 6 a.m. and lights off at 6 p.m. Temperature range is kept between 20–24°C, and humidity range is 30–70%. The acute pneumonia model was performed similar to previously described experiments21,111. Briefly, overnight cultures were subcultured at a 1:200 dilution and grown shaking at 37 °C for 3 h to mid-exponential growth phase. Six- to eight-week-old female C57BL/6 mice (Charles River Laboratories, Wilmington, MA) anesthetized with 4% isoflurane were intranasally inoculated with approximately 109 CFU that were twice-washed in PBS. At 24 hpi, mice were sacrificed, and CFU in the lungs, spleen, and kidneys were quantified by serial dilution plating the homogenized organs. For experiments with female C3H/HeN (Envigo International Holdings, Indianapolis, IN) and C3H/HeJ (Jackson Laboratory, Bar Harbor, ME) mice, A. baumannii, S. aureus, and K. pneumoniae inoculums were prepared and mice were intranasally inoculated as described above, with the exception that inoculums of approximately 103, 104, 105, 106, 107, or 108 CFU were used for A. baumannii, and 5 × 107 CFU was used for S. aureus and K. pneumoniae. Following, at the indicated timepoints, mice were sacrificed and bacteria in the lungs, spleen, and kidneys were quantified as described above. For co-infections, A. baumannii was distinguished from S. aureus and K. pneumoniae by plating on LB agar supplemented with 10 μg/mL chloramphenicol. For antibiotic treatment experiments the indicated mice were treated intraperitoneally with PBS or 100 mg/kg tigecycline every 12 h, PBS or 5 mg/kg colistin every 8 h, PBS or 500 mg/kg apramycin every 12 h, or PBS or 100 mg/kg imipenem every 12 h with all treatments beginning 4 hpi. Antibiotics for intraperitoneal treatments were dissolved in PBS, and the injection volume was 100 μl. For cyclophosphamide treatments, one treatment of 300 mg/kg dissolved in PBS was intraperitoneally administered 3 days prior to infection.

Flow cytometry

Flow cytometry was performed similarly to previously described methods43. Briefly, BALF samples were collected in PBS supplemented with 1 mM EDTA, and cells were collected by centrifugation at 300 × g for 5 min. Cells were then resuspended in Pharm Lyse Buffer (BD Biosciences, Franklin Lakes, NJ) and incubated for 3 min at room temperature to lyse red blood cells. Cells were subsequently washed in fluorescence-activated cell sorting (FACS) buffer (PBS supplemented with 1% heat inactivated fetal bovine serum and 0.1% sodium azide) and blocked with TruStain FcX PLUS (BioLegend, San Diego, CA; Ref: 156603) for 15 min at 4 °C. Samples were then stained with a 1:100 dilution with anti-CD45-BV605 (BioLegend; Ref: 103139), anti-CD11b-Alexa700 (BioLegend; Ref: 101222), anti-CD11c-APC (BioLegend; Ref: 117309), anti-SiglecF-PerCP5.5 (BioLegend: Ref: 155525), and anti-Ly6G-BV421 (BioLegend; Ref: 127628) for 30 min at 4 °C, markers used to define total AMs and PMNs. In some experiments, CD45 antibody was not included in the staining panel. Following, cells were washed in FACS buffer and fixed in 2% paraformaldehyde (PFA). Samples were read on a LSR II Fortessa cytometer (BD Biosciences) or an Aurora cytometer (Cytek Biosciences, Fremont, CA). Total cell counts in the BALF were calculated using Precision Count Beads (BioLegend) according to the manufacturer’s instructions.

Histopathology of lung slices

Lungs were perfused with in PBS followed by either inflation with 10% neutral buffered formalin and paraffin embedding or inflation with optimal cutting temperature (OCT) compound (Fisher Scientific) diluted 1:1 in 4% PFA followed by snap freezing. Lungs were sectioned at 4–5 µm thickness, and staining with hematoxylin and eosin (H&E). Histopathological evaluation and scoring were performed by a board-certified veterinary pathologist, and lungs were assessed for changes in alveolitis, alveolar histiocytosis, perivascular infiltrate, peribronchiolitis, BALT, loss and attenuation of airway epithelium, pleuritis, airway smooth muscle hypertrophy, airway squamous epithelium metaplasia, airway goblet cell hyperplasia, and fibrosis.

Generation of constructs and strains used in this study

Primers used in this study are listed in Table S5. DNA fragments were assembled using either the In-Fusion HD EcoDry Cloning Kit (TaKaRa Bio, Mountain View, CA) or NEBuilder HiFi DNA Assembly Master Mix (New England Biolabs, Ipswich, MA). To generate the vector for generation of the invL mutational construct, pEX18Tc was amplified without the tetracycline resistance cassette (primers: 5’ pEX18 marker swap and 3’ pEX18 marker swap), the apramycin resistance cassette was amplified from pKD4-Apr (primers: 5’ Apr for pEX18Ap and 3’ Apr for pEX18Ap), and the amplicons were assembled, generating pEX18Ap112,113. The pEX18Ap mutational constructs were then made by amplifying the pEX18Ap vector (primers: 5’ pEX18Tc and 3’ pEX18Tc), a ~1000 bp region upstream of the genes of interest (invLKO primers: 5’ F1 G636 invLKO and 3’ F1 G636 invLKO; bapKO primers: 5’ F1 G636 bapKO and 3’ F1 G636 bapKO; ataKO primers: 5’ F1 G636 ataKO and 3’ F1 G636 ataKO; fhaBCKO primers: 5’ F1 G636 fhaBCKO and 3’ F1 G636 fhaBCKO), and a ~1000 bp region downstream of the genes of interest (invLKO primers: 5’ F2 G636 invLKO and 3’ F2 G636 invLKO; bapKO primers: 5’ F2 G636 bapKO and 3’ F2 G636 bapKO; ataKO primers: 5’ F2 G636 ataKO and 3’ F2 G636 ataKO; fhaBCKO primers: 5’ F2 G636 fhaBCKO and 3’ F2 G636 fhaBCKO), followed by assembly of these amplicons. Mutational constructs were then transformed into G636, and strains with the integrated plasmid were selected for by apramycin treatment. Counterselection for double crossover was performed by plating these strains on LB agar without NaCl supplemented with 10% sucrose. Mutants were then confirmed by PCR analyses and whole-genome sequencing.

The invL complementation construct was generated by amplifying the putative promoter region (~300 bp upstream) along with the invL open reading frame (primers: 5’ G636 fdeCKO Comp and 3’ G636 fdeCKO Comp-His6 v2) and the pUC18T-miniTn7T-Apr vector (primers: Tn7 linear Fwd-His6 and Tn7 liner Rev)114. These amplicons were then assembled, generating pUC18T-miniTn7T-Apr::G636 invLKO comp. To generate the gfp integration construct, the gfp cassette was amplified from PB-FLuc+GFPd2 (primers: 5’ d2EGFP for pUC18T-mTn7 and 3’ d2EGFP for pUC18T-mTn7) and pUC18T-mTn7-Apr was amplified (primers: 5’ pUC18T-mTn7 for d2EGFP and 3’ pUC18T-mTn7 for d2EGFP). These fragments were then assembled, generating pUC18T-miniTn7T-Apr::gfpd2. pUC18T-miniTn7T-Apr::G636 invLKO comp and pUC18T-miniTn7T-Apr::gfpd2 were introduced into G636 ΔinvL and G636, respectively, using a four-parental conjugation technique, as previously described114,115,116,117. Selection was achieved using LB supplemented with apramycin and chloramphenicol, and insertion of the respective fragments at the mTn7 site in the resulting G636 invL+ and G636-gfp strains was confirmed by PCR analyses.

Statistical methods

All statistical analyses were performed using GraphPad Prism version 9, and p values < 0.05 were considered statistically significant. The two-tailed Mann–Whitney U test was used when comparing two groups, and the Kruskal–Wallis H test with Dunn’s test for multiple comparisons was used when comparing more than two groups. Two-way ANOVA with Tukey’s or Bonferroni’s test for multiple comparisons was used when comparing two or more groups with two independent variables. Source data are provided as a source data file.

Reporting summary

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