Host Fate is Rapidly Determined by Innate Effector-Microbial Interactions During Acinetobacter baumannii Bacteremia

Abstract

Background. Acinetobacter baumannii is one of the most antibiotic-resistant pathogens. Defining mechanisms driving pathogenesis is critical to enable new therapeutic approaches.

Methods. We studied virulence differences across a diverse panel of A. baumannii clinical isolates during murine bacteremia to elucidate host-microbe interactions that drive outcome.

Results. We identified hypervirulent strains that were lethal at low intravenous inocula and achieved very high early, and persistent, blood bacterial densities. Virulent strains were nonlethal at low inocula but lethal at 2.5-fold higher inocula. Finally, relatively avirulent (hypovirulent) strains were nonlethal at 20-fold higher inocula and were efficiently cleared by early time points. In vivo virulence correlated with in vitro resistance to complement and macrophage uptake. Depletion of complement, macrophages, and neutrophils each independently increased bacterial density of the hypovirulent strain but insufficiently to change lethality. However, disruption of all 3 effector mechanisms enabled early bacterial densities similar to hypervirulent strains, rendering infection 100% fatal.

Conclusions. The lethality of A. baumannii strains depends on distinct stages. Strains resistant to early innate effectors are able to establish very high early bacterial blood density, and subsequent sustained bacteremia leads to Toll-like receptor 4–mediated hyperinflammation and lethality. These results have important implications for translational efforts to develop therapies that modulate host-microbe interactions.

In the last decade, Acinetobacter baumannii has emerged to become a predominant cause of nosocomial infections in the United States and across the globe [1–4] and is one of the few bacteria of which strains have developed resistance to all available antibiotics [5, 6]. In particular, bacteremia caused by carbapenem-resistant A. baumannii isolates results in >50% mortality rates owing to inadequate available therapies [7–11]. Furthermore, in contrast to other resistant bacteria, few antibiotics are in the pipeline to treat extremely drug-resistant A. baumannii [6, 12]. There is a critical need for new strategies to prevent and treat infections caused by highly resistant A. baumannii.

Rational design of new prevention and/or treatment strategies requires a much deeper understanding of relevant host-pathogen interactions. In vitro virulence factors, such as surface hydrophobicity, iron siderophores, outer membrane proteins, biofilm-inducing polysaccharides [13–15], and phospholipase D [16] are described as being dominant during lung, skin, or ascites/abscess infection by A. baumannii. Regrettably, studies defining host factors that determine outcome are more limited. During lung infection, depletion of neutrophils converted survivable inocula to lethal inocula [17, 18], and macrophage depletion resulted in increased tissue bacterial burden [19]. However, the mechanisms that enable A. baumannii to cause disease during bacteremia are not well described.

We sought to define microbe-host interactions that determined survival during bloodstream infection caused by a representative collection of clinical isolates of A. baumannii. Previously, we demonstrated a role for lipopolysaccharide (LPS) in A. baumannii virulence and showed that mice lacking functional Toll-like receptor 4 (TLR4) were resistant to sepsislike disease and death after bloodstream infection [20]. Here we extend those findings to demonstrate that early clearance kinetics of the bacteria by the host are critical in determining virulence and that distinct A. baumannii isolates display different susceptibilities to host innate effector mechanisms. These findings have important implications for vaccine and other immune-based therapies.

MATERIALS AND METHODS

Mouse and Bacterial Strains

Multiple clinical isolates of A. baumannii were used, of diverse strain types and origins (Table 1). Strains were typed by standard multilocus sequence typing methods, as described elsewhere [28–30]. Single colonies of freshly streaked A. baumannii isolates were cultured overnight in tryptic soy broth (TSB) at 37°C, subcultured 1:100 to log phase, washed 3 times in phosphate-buffered saline (PBS), and quantitated by optical density to determine appropriate inocula for injection or in vitro assay. Wild-type (C3H/FeJ) mice were used for all in vivo experiments (Jackson Laboratories). Mice were challenged with defined inocula of A. baumannii isolates in 250 µL of sterile PBS, injected intravenously. Mice were bled after infection at 1 hour by tail vein nicking and 5 hours by terminal bleed via cardiac puncture, or at 24 hours, either by terminal bleeding via cardiac puncture or by tail vein nicking for experiments in which survival was monitored. For survival experiments, mice were euthanized when they appeared moribund, which was defined as the inability to ambulate when stimulated tactilely. Colony-forming units (CFUs) were quantified by serial dilutions on tryptic soy agar (TSA). All animal experiments were approved by the Institutional Committee on the Use and Care of Animals at the Los Angeles Biomedical Research Institute and followed the National Institutes of Health guidelines for animal housing and care.

Table 1.

Acinetobacter baumannii Clinical Strains and Relative Virulence in C3H/FeJ Mice

Strain Name	Strain Type (MLST)	Source	Inoculum, CFUs	Survival, %
Hypervirulent
HUMC1 [21]	ST2	Clinical blood and lung isolate	1.5–2 × 107	0
LAC-4 [22, 23]	ST10	Clinical blood isolate	1.5–2 × 107	0
Virulent
HUMC4 [21]	ST2	Clinical lung isolate (tracheostomy)	2 × 107	100
			5 × 107	0
HUMC5 [21]	ST2	Clinical lung isolate (BAL)	2 × 107	100
			5 × 107	0
HUMC6 [21]	ST2	Clinical lung isolate (sputum)	2 × 107	100
			5 × 107	0
HUMC12 [21]	ST2	Clinical wound isolate	2 × 107	100
			5 × 107	0
C14 [24]	ST107	Clinical wound isolate; pmrB mutant	2 × 107	100
			5 × 107	0
AB0061 [25]	ST2	Clinical blood isolate	3 × 107	60
AB0068 [25]	ST10	Skin swab sample	3 × 107	20
UH7807	ST2	Clinical lung isolate (sputum)	2 × 107	100
			5 × 107	0
Hypovirulent
ATCC 17978 [26]	ST112	Clinical CSF isolate	8 × 108	100
R2 [24]	ST112	ATCC 17978 derivative; pmrB mutant	1 × 108	100
Not hypervirulent (tested only at 2–3 × 107)
ATCC 31 Clone B	NA	NA	2 × 107	100
ATCC 125	NA	NA	2 × 107	100
ATCC 152 Clone A	NA	NA	3 × 107	100
AB0057 [25]	ST1	Clinical blood isolate	3 × 107	100
AB0071 [25]	NA	Clinical blood isolate	2 × 107	100
AB0072 [25]	NA	Skin swab sample	2 × 107	100
AB0074 [25]	NA	Clinical blood isolate	2 × 107	100
AB0093	NA	NA	2 × 107	100
METRO 9	NA	Clinical isolate (catheter)	2 × 107	100
UH2207	ST2	Clinical lung isolate (sputum)	2 × 107	100
UH4907	Novel	Clinical throat isolate	2 × 107	100
UH5107	Novel	Clinical lung isolate (sputum)	2 × 107	100
UH5207	ST78	Clinical isolate (catheter)	2 × 107	100
UH6507	Novel	Clinical isolate (urine)	2 × 107	100
UH7007	ST2	Clinical isolate (urine)	2 × 107	100
UH7507	Novel	Clinical wound isolate	2 × 107	100
UH8107	ST2	Clinical isolate (BAL)	3 × 107	100
UH8307	Novel	Clinical wound isolate	2 × 107	100
UH8407	ST2	Clinical isolate (blood)	2 × 107	100
UH9007	ST2	Clinical isolate (blood)	3 × 107	100
UH9707	ST2	Clinical isolate (catheter)	3 × 107	100
AB5075 [27]	ST2	Clinical isolate (blood)	2 × 107	100

Abbreviations: BAL, bronchoalveolar lavage; CFUs, colony-forming units; CSF, cerebrospinal fluid; MLST, multilocus sequence typing; NA, not available.

^a All mice were monitored for 7 days in survival experiments; mice that were not moribund by 3–4 days after infection typically survive intravenous infection with A. baumannii.

In Vitro Bacterial Growth Curves

A. baumannii strains were cultured overnight in TSB at 37°C, passaged by placing 100 µL of overnight culture in 10 mL of TSB, and serially sampled to determine optical density and bacterial density by quantitative culturing. Optical density was measured at an absorbance of 600 nm (Implen OD600 DiluPhotometer).

Complement Susceptibility Assay

The susceptibility of A. baumannii strains was measured in vitro by suspending 1 × 10⁶ CFUs of log-phase A. baumannii isolates in Hank's balanced salt solution (HBSS) containing 50% BALB/cJ mouse serum with or without heat inactivation at 57°C for 30 minutes. After 2-hour incubation at 37°C, CFUs were quantified by serial dilutions on TSA, and counts were compared.

Macrophage Phagocytosis Assay

The susceptibility of bacteria to uptake by macrophages was measured in vitro by seeding 1 × 10⁶ RAW264.7 cells onto glass coverslips in 1 mL of Dulbecco's modified Eagle medium with 10% fetal bovine serum (heat-inactivated) supplemented with 100 U of interferon γ in 12-well tissue culture plates. The cells adhered to the cover glass overnight in a humidified incubator at 37°C with 5% carbon dioxide. Each well was subsequently washed 3 times with HBSS and then seeded at a 10:1 multiplicity of infection with 1 × 10⁷ CFUs of log-phase A. baumannii isolates in 1 mL of HBSS with 20% complement-active CD-1 mouse serum (Innovative Research). The plates were briefly centrifuged at 250g to cohere the RAW264.7 and bacterial cells. Wells with no RAW264.7 cells were used as a control. After a 1-hour incubation at 37°C, bacteria that did not associate with macrophages (remained in the supernatant) were quantitated by serially diluting supernatant samples on TSA plates. In parallel, macrophages adhered to coverslips were washed, fixed with 100% methanol, and stained using Hema-3 Manual Stains (Fisher Scientific), and the percentage of cells harboring bacteria and bacterial counts per cell were quantitated by microscopy.

Microscopy

To visualize macrophages, arbitrary locations on each coverslip were imaged on a Zeiss AxioImager microscope. Images that included ≥4 visible macrophages within the frame were analyzed using the ImagePro software under the ×100 objective. Macrophages harboring visible adherent or internalized bacteria were counted, and divided by total macrophages to determine the percentage of cells with associated bacteria. The number of bacteria associated with each macrophage was also quantitated and recorded. A minimum of 100 cells per coverslip were counted.

In Vivo Depletion of Innate Immune Effectors

Mice were depleted of innate immune effector cells 2 days before infection. Macrophages were depleted by intraperitoneal injection of 250 µL (approximately 1.25 mg) of room temperature liposomal clodronate or liposomal sterile PBS as a control (both purchased from ClodronateLiposomes.org). The complement pathway was systemically inactivated by intraperitoneal injection of cobra venom factor (15 µg per mouse; Complement Technology), which results in depletion of C3 and C5 components and prevents complement activity for 3–6 days [31, 32]. Mice were made neutropenic by intraperitoneal injection of cyclophosphamide (230 mg/kg [approximately 6 mg per mouse]; Baxter).

Statistics

Survival was compared by using the nonparametric log rank test. Categorical variables were compared using the Wilcoxon rank sum test for unpaired comparisons or the Wilcoxon signed rank test for paired comparisons. In the absence of evidence of normality, nonparametric approaches were used to compare data sets.

RESULTS

Different Virulence Levels of A. baumannii Strains in a Mouse Model of Bloodstream Infection

We used a representative collection of A. baumannii clinical isolates from different sources and of dissimilar strain backgrounds to determine the spectrum of virulence during bloodstream infection in mice (Table 1). One of 3 patterns was evident when mice were intravenously infected with different A. baumannii isolates. Among the 33 strains tested, we found 2 strains that we characterized as “hypervirulent” because they caused 100% lethal bloodstream infection within 3 days with a 2 × 10⁷ inoculum (Figure 1 and Table 1). We also collected 6 strains that we characterized as “virulent”; they did not cause death when given intravenously at a 2 × 10⁷ inoculum, but they caused 100% fatal infection at a 5 × 10⁷ inoculum (Figure 1 and Table 1). Two other strains resulted in partial lethality at an intermediate dose and were also classified as virulent. Finally, we used 2 strains that we characterized as “hypovirulent” because they did not cause death when given intravenously at inocula of ≥1 × 10⁸ (Figure 1 and Table 1). The remaining strains tested were all found to be nonlethal at an inoculum of 2–3 × 10⁷. Although they were not hypervirulent, we did not discriminate whether they were virulent (lethal at a 5 × 10⁷ inoculum) or hypovirulent. These results define distinct inoculum thresholds at which various A. baumannii isolates result in bacteremia-associated lethality.

Figure 1.

Survival of C3H/FeJ mice infected with different clinical isolates of Acinetobacter baumannii. Mice (5–8 per group) were intravenously infected via the tail vein with 2 × 10⁷ (left) or 5 × 10⁷ (right) colony-forming units (CFUs) of bacilli. Experiments were repeated up to 3 times for each strain shown. *P < .05 vs HUMC6, C14, and ATCC 17978; ^†P < .05 vs ATCC 17978.

Different Patterns of Clearance From the Bloodstream After Intravenous Inoculation of A. baumannii Strains

To conduct a more detailed investigation of virulence, we selected representative strains of each category of virulence (hypervirulent, HUMC1 and LAC-4; virulent, HUMC6 and C14; hypovirulent, ATCC 17978). We sought to determine whether there was a relationship between in vivo bacterial blood density and strain virulence over time. Mice were infected with 2 × 10⁷ of each strain, which is a lethal inoculum only for the hypervirulent strains, and blood was serially sampled thereafter. Surprisingly, we found evidence that 2 different stages of bacterial blood density are related to virulence. Within the first hour of infection, the 2 hypervirulent strains HUMC1 and LAC-4 maintained extraordinarily high blood bacterial density (>10⁷ CFUs/mL), which reflects essentially no net clearance of organism from the blood (2 × 10⁷ organisms administered into approximately 2 mL of blood in the mouse) (Figure 2). Furthermore, these bacterial densities were maintained during the subsequent 18–24 hours for both strains, reflecting lack of net clearance of the organism during that time. In contrast, the virulent strains were reduced to 100–1000-fold lower bacterial densities at 1 hour after infection, but, like the hypervirulent strains, they underwent minimal additional clearance in the subsequent 24 hours (Figure 2). Finally, the hypovirulent strain ATCC 17978 reached a bacterial density at 1 hour that was similar to the virulent strains, but it underwent nearly 100-fold additional clearance during the subsequent 24 hours (Figure 2).

Figure 2.

Bacterial counts of different Acinetobacter baumannii strains in the blood of intravenously infected mice. Different strains of log-phase A. baumannii were inoculated (approximately 2 × 10⁷ colony-forming units [CFUs]) intravenously into mice (5–10 per group), and blood samples were harvested, serially diluted, and plated to quantitate CFUs at the time points shown. Solid lines represent serial sampling of blood from the same mice; dashed lines, blood samples collected from a separate group of identically infected mice; error bars, interquartile ranges.

In Vitro Growth Rates of A. baumannii Strains and Susceptibility of Strains to Complement Killing in Mouse Serum

We sought to determine whether differences in bloodstream bacterial levels could be explained by different growth rates or serum susceptibilities of the various strains. Growth curves under nutrient-rich conditions in vitro demonstrated similar growth rates of the strains within 4 hours (Figure 3A). We then tested the serum susceptibilities of 4 A. baumannii strains by comparing bacterial CFUs after 2-hour incubations in 50% complement-inactive mouse serum (heat inactivated) versus 50% normal serum. Hypervirulent (HUMC1 and LAC-4) and virulent (HUMC6) strains displayed low susceptibility to complement-mediated killing (5%–10%), whereas the hypovirulent strain (ATCC 17978) displayed moderate susceptibility (40%) (Figure 3B). Although ATCC 17978 was more susceptible to direct killing by serum complement, this was unlikely to be the only explanation for differences in virulence because hypervirulent and virulent strains had similar, low susceptibility to mouse complement despite achieving very different bacterial densities at both 1 and 24 hours after infection (Figure 2).

Figure 3.

In vitro growth rates and complement susceptibilities of different Acinetobacter baumannii strains. A, Overnight cultures of each strain were subcultured 1:100 in triplicate, and consecutive aliquots were removed and measured at an optical density of 600 nm to determine bacterial densities at each time point shown. B, To assess complement susceptibility, strains were incubated in Hank's balanced salt solution containing either 50% mouse serum or 50% heat-inactivated mouse serum for 2 hours. Viable remaining bacteria were quantitatively cultured. The percentage of complement killing was calculated as follows: (CFU_HI − CFU_CA)/CFU_HI, where CFU_HI indicates heat-inactivated and CFU_CA, complement-active. Results are the average of 4 independent experiments, each done in triplicate. Error bars represent interquartile ranges. *P < .05 vs all other strains.

Distinct Susceptibilities of A. baumannii Strains to Macrophage Uptake In Vitro

We next asked whether different strains of A. baumannii interacted differently with macrophages. To this end, we coincubated A. baumannii strains for 1 hour with RAW264.7 macrophages that were previously adhered to coverslips, before washing out extracellular bacteria. Visualized colocalization of bacteria with macrophages after washing was interpreted as either attachment to the macrophage or internalization of the bacteria. Initial experiments demonstrated that this interaction required the presence of serum, presumably to opsonize bacteria. Control wells therefore had serum (to control for growth and any direct serum killing), but no macrophages.

Macrophage colocalization with A. baumannii correlated inversely with strain virulence. Very few macrophages associated with the hypervirulent strains (HUMC1 and LAC-4), a 10-fold higher percentage associated with the virulent strain (HUMC6), and virtually all macrophages associated with the hypovirulent strain (ATCC 17978) (Figure 4A and 4C). Similarly, individual macrophages that did successfully bind to or ingest bacteria were able to take up 3-fold more HUMC6 than HUMC1 or LAC-4 per cell, and nearly 10-fold more ATCC 17978 per cell (Figure 4B and 4C). The numbers of viable bacteria remaining in the supernatant after macrophage incubation was also measured. Consistent with the colocalization observations, the presence of macrophages resulted in a greater reduction in supernatant CFUs for the virulent strain compared with the hypervirulent strains, and a still greater reduction in CFUs for the hypovirulent strain compared with all strains (Figure 4D).

Figure 4.

Susceptibility of different Acinetobacter baumannii isolates to macrophage uptake. A. baumannii strains were coincubated with RAW264.7 macrophages for 1 hour in Hank's balanced salt solution supplemented with 20% CD-1 mouse serum. Two experiments were repeated in triplicate. A, Median percentage of macrophages associated with bacteria. B, Median number of bacteria per bacteria-positive macrophage. C, Representative microscopic images. D, Supernatants from wells containing or not containing macrophages were serially diluted and plated to determine the percentage of bacteria removed from the supernatant due to association with macrophages. *P < .05 vs HUMC6 and ATCC 17978; ^†P < .05 vs all other groups. Error bars represent interquartile ranges.

Effect of In Vivo Depletion of Innate Immune Effectors on Bacterial Density of the Hypovirulent Strain

Given the enhanced susceptibility of ATCC 17978 (hypovirulent strain) to complement-mediated killing and macrophage uptake in vitro, we next hypothesized that interfering with these innate immune mechanisms during infection would result in increased blood density and would change lethality of the bacteria. Mice were depleted of macrophages with liposomal clodronate [33, 34] and/or depleted of functional complement with cobra venom factor [31, 32]. To rule out the possibility that the liposomes themselves would modulate bacterial density, or interfere with the ability of cobra venom factor to deplete complement, negative control mice were treated with empty liposomes or empty liposomes plus cobra venom factor. Empty liposomes had no effect on bacterial blood density, which was very low at 24 hours after infection with A. baumannii ATCC 17978 (Figure 5A). Mice depleted of macrophages or complement alone had a 10-fold increase in blood bacterial density, which nevertheless remained below 10³ CFUs/mL (Figure 5A), far below the lethal threshold (approximately 10⁷CFUs/mL) achieved by hypervirulent A. baumanni HUMC1. Dual depletion of macrophages and complement synergistically increased blood bacterial density by 1000-fold, but the density still remained far below the lethal threshold (Figure 5A), and all the mice survived with minimal clinical evidence of illness (normal behavior, no ruffled fur). Of note, this synergy between liposomal clodronate and cobra venom factor was detected despite the fact that empty liposomes somewhat antagonized the effect of cobra venom factor (Figure 5A).

Figure 5.

Impact of innate immune effector disruption on bacterial densities of a hypovirulent Acinetobacter baumannii isolate. A. baumannii ATCC 17978 was inoculated into the tail veins of C3H/FeJ mice (5 mice per group) at an inoculum of 2 × 10⁷ CFUs, at which the organism is nonfatal in normal mice. A, Mice were given sterile phosphate-buffered saline (PBS) placebo, or empty liposome (L, negative control), or depleted of macrophages with liposomal clodronate (LC) and/or of complement with cobra venom factor (CVF), 2 days before ATCC 17978 challenge. Bacterial density in the blood at 24 hours increased with depletion of macrophages and complement, but all mice in this group survived. The limit of detection was 2 × 10¹ colony-forming units [CFUs]/mL (dashed line). *P < .05 vs PBS control; ^†P < .05 vs all other groups. B, Mice were depleted of neutrophils with cyclophosphamide (Cyclo) alone or in combination with macrophage depletion (LC) and/or complement depletion (CVF). All mice in the triple-depletion group died within 24 hours, and blood samples to determine bacterial counts were obtained at death. None of the other mice died. The limit of detection was 2 × 10¹ CFUs/mL (dashed line). *P < .05 vs PBS; ^†P < .05 vs cyclophosphamide and PBS; ^‡P < .05 vs all other groups.

We next determined the impact of depleting neutrophils with cyclophosphamide, either alone or in combination with liposomal clodronate and/or cobra venom factor. Cyclophosphamide depletion of neutrophils alone significantly increased bacterial blood density (Figure 5B). Dual depletion of neutrophils along with either macrophages or complement further increased bacterial blood density to between 10⁵ and 10⁶ cells/mL, still below the lethal threshold (Figure 5B). However, triple disruption of neutrophils, macrophages and complement resulted in a synergistic increase to lethal levels of bacterial density, 10⁷-fold above background (Figure 5B). None of the mice with these high bacterial densities survived past 24 hours.

Effects of In Vivo Depletion of Innate Immune Effectors on Bacterial Density of Virulent and Hypervirulent Strains

We next determined the effect of depleting innate immune effectors on the density of virulent and hypervirulent strains. Mice were similarly selectively depleted of innate effectors in different combinations and infected with a sublethal dose (1.3 × 10⁷ CFUs) of the virulent strain HUMC6. In this experiment, an earlier end point (5–6 hours) was necessary to ensure that all groups of mice were still alive. Depletion of either macrophages (liposomal clodronate) or neutrophils (cyclophosphamide) alone had no impact on the bacterial density of HUMC6 at 5–6 hours after infection (Figure 6A). In contrast, complement depletion (cobra venom factor) markedly increased bacterial density (Figure 6A), as did combinations of complement with macrophage (liposomal clodronate) or neutrophil (cyclophosphamide) depletion. Triple depletion of complement, macrophages, and neutrophils resulted in bacterial densities akin to those observed in triple depleted mice infected with A. baumannii ATCC 17978 (Figure 6A).

Figure 6.

Impact of innate immune effector disruption on bacterial densities of virulent and hypervirulent Acinetobacter baumannii isolates. C3H/FeJ mice (5 per group) were treated with phosphate-buffered saline (PBS), liposomal clodronate (LC) to deplete macrophages, cobra venom factor (CVF) to deplete complement, cyclophosphamide (Cyclo) to deplete neutrophils, or various combinations. A, A. baumannii HUMC6 was inoculated into the tail veins of the mice at an inoculum of 1.3 × 10⁷ CFUs. Blood was harvested at 5–6 hours, at which time none of the mice appeared moribund. The limit of detection was 2 × 10² colony-forming units [CFUs]/mL (dashed line). *P < .05 vs PBS. B, A. baumannii HUMC1 was inoculated into the tail veins of the mice at an inoculum of 1.6 × 10⁷ CFUs. Blood was harvested at 5–7 hours, at which time mice disrupted of macrophages all appeared moribund but none of the other mice appeared severely ill. The limit of detection was 2 × 10⁴ CFUs/mL (dashed line). *P < .05 vs PBS.

Different results were seen when effectors were depleted in the setting of infection with the hypervirulent strain, A. baumannii HUMC1. Groups of depleted mice were inoculated with a dose of 1.6 × 10⁷ CFUs of HUMC1 and euthanized 5–6 hours after infection. Bacterial density of the hypervirulent strain in the blood at this time point reflected its resistance to clearance by nondepleted, control mice. Complement depletion had no impact on bacterial density (Figure 6B). Neutrophil depletion resulted in a trend to increased blood bacterial density (P = .06), whereas macrophage depletion resulted in a >100-fold increase in bacterial density, to saturation levels of bacteria that could not be further increased by any combination of effector disruption (Figure 6B). These observations underscore the differences in susceptibility to innate immune clearance mechanisms of various A. baumannii strains that display distinct virulence levels.

DISCUSSION

We have found that different clinical isolates of A. baumannii have profound differences in their in vivo virulence in a murine bacteremia model. The 100% lethal dose of hypervirulent strains, for example, was >25-fold less than that of hypovirulent strains. To our surprise, the primary driver of virulence was the ability of the organism to evade innate immune effectors such that a high bacterial density was established very early after infection and then maintained in blood. Isolates that manifested different levels of virulence displayed distinct susceptibilities to specific host defense effectors; depending on the isolate, interruption of these effectors had distinct impacts on bacterial growth in vivo.

In vitro assays modeling innate immune effector mechanisms revealed different susceptibilities among isolates that correlated with bacterial levels during bacteremia. Hypovirulent strains displayed a greater sensitivity to direct complement killing in vitro than virulent and hypervirulent strains. Interestingly, inhibition of complement in vivo led to increased bacterial numbers of not only the hypovirulent strain (ATCC 17978) but also the virulent strain (HUMC6), suggesting a more complex interaction between complement and HUMC6 in vivo than direct killing alone. Likewise, in vitro macrophage uptake assays revealed a hierarchy of susceptibilities of hypervirulent to hypovirulent strains that was reflective of their increasing thresholds of lethal inocula.

We found two stages of effector clearance of A. baumannii. Notably, much of host fate depends on the first hour of infection. Hypervirulent strains (HUMC1 and LAC-4) were able to achieve high levels of bacterial density within 1 hour of inoculation, reflecting little to no net clearance of bacteria from the blood. That high level of bacteria then persisted during the subsequent 23 hours. At the same inocula, virulent strains (HUMC6 and C14) underwent a dramatic decrease in CFUs in the blood after 1 hour but were able to persist during the next 23 hours without net clearance. In contrast, the hypovirulent strain (ATCC 17978) was effectively cleared by three distinct innate effectors, such that disruption of any single defense mechanism resulted in only a modest increase in bacterial density, but multiple disruptions synergistically increased bacterial density. Recognizing that the hypovirulent strain could be converted into a hypervirulent strain by triple disruption of complement, macrophages, and neutrophils underscores that the strains themselves do not have intrinsic differences in their ability to trigger sepsis other than the ability to achieve and maintain a high bacterial density in blood. Disruption of innate effectors allowed hypovirulent and virulent strains to achieve much higher blood densities, which resulted in lethality caused by these isolates at normally nonlethal inocula. Disruption of innate effectors during infection with the hypervirulent strain had less relative impact on bacterial density, because the hypervirulent strain was already highly resistant to all three innate effectors, and only disruption of macrophages substantially increased its ability to proliferate in the host.

Unexpectedly, we found that disruption of different effectors had distinct impacts on hypovirulent, virulent, and hypervirulent strains. The hypovirulent strain had very low bacterial burden and seemed to be equally susceptible to complement-, macrophage-, and neutrophil-mediated clearance. In contrast, the virulent strain had 1000-fold higher bacterial density in the absence of effector depletion, indicating some intrinsic level of resistance to innate effectors. Further clearance of the virulent strain was disproportionately dependent on complement depletion, with a synergistic increase in CFUs when both macrophages and neutrophils were also depleted. For the hypervirulent strain, bacterial density was 1000-fold higher than the virulent strain even with no effector depletion, indicating profound resistance to innate effectors even with no depletion. Macrophage depletion further exacerbated the bacterial density, with minimal additional impact of complement or neutrophil depletion. Thus, a complex interplay between strains and innate effectors governs bacterial density during A. baumannii infection. Functional redundancy is evident in effector clearance of less virulent strains. In contrast, virulent strains are resistant to one or more innate mechanisms, enabling higher baseline bacterial density and greater sensitivity to depletion of individual residual effector mechanisms.

These results demonstrate a remarkable range of abilities to manifest virulence across a varied panel of A. baumannii clinical isolates and elucidate how host-microbe interactions lead to clinical outcome of these infections. We have shown elsewhere that the lethality of bloodstream A. baumannii infections in mice is driven by the interaction between bacterial LPS and host TLR4-mediated sepsis pathways [20]. The current results elucidate bacterial strain differences that lead to different abilities to drive sepsis. Our results demonstrate that the primary bacterial virulence mechanism during bacteremia is escape from host effectors that enable persistence in the bloodstream. Persistence in the bloodstream leads to the production of sufficient quantities of LPS to trigger TLR4-mediated sepsis. Thus, promising approaches are to develop interventions that enhance the ability of innate effectors to clear A. baumannii from blood, such as antibody-based therapies and/or vaccines [21, 35]. Alternatively, therapies aimed at preventing LPS-TLR4 interactions from triggering sepsis are likely to be broadly effective across A. baumannii isolates.