Correspondence: Michael McConnell, Hospital Universitario Virgen del Rocío, Instituto de Biomedicina de Sevilla, Laboratorio 208, Avenida Manuel Siurot s/n, 41013 Sevilla, Spain. Tel.: +34 955923100; fax: +34 955013292; e-mail: firstname.lastname@example.org
Acinetobacter baumannii has emerged as a medically important pathogen because of the increasing number of infections produced by this organism over the preceding three decades and the global spread of strains with resistance to multiple antibiotic classes. In spite of its clinical relevance, until recently, there have been few studies addressing the factors that contribute to the pathogenesis of this organism. The availability of complete genome sequences, molecular tools for manipulating the bacterial genome, and animal models of infection have begun to facilitate the identification of factors that play a role in A. baumannii persistence and infection. This review summarizes the characteristics of A. baumannii that contribute to its pathogenesis, with a focus on motility, adherence, biofilm formation, and iron acquisition. In addition, the virulence factors that have been identified to date, which include the outer membrane protein OmpA, phospholipases, membrane polysaccharide components, penicillin-binding proteins, and outer membrane vesicles, are discussed. Animal models systems that have been developed during the last 15 years for the study of A. baumannii infection are overviewed, and the recent use of these models to identify factors involved in virulence and pathogenesis is highlighted.
Acinetobacter baumannii has become an increasingly important human pathogen because of the increase in the number of infections caused by this organism and the emergence of multidrug-resistant (MDR) strains. The majority of infections caused by A. baumannii are hospital-acquired, most commonly in the intensive care setting in severely ill patients. In addition, A. baumannii has emerged as a cause of infections acquired in long-term care facilities, in the community, and in wounded military personnel (Anstey et al., 1992, 2002; Leung et al., 2006; Scott et al., 2007; Schafer & Mangino, 2008; Sebeny et al., 2008; Sengstock et al., 2010). The types of infections produced by this pathogen include, but are not limited to, pneumonia (both hospital and community-acquired), bacteremia, endocarditis, skin and soft tissue infections, urinary tract infections, and meningitis. In most cases, it is thought that infections are acquired after exposure to A. baumannii that persists on contaminated hospital equipment or by contact with healthcare personnel that have been exposed to the organism through contact with a colonized patient (Maragakis et al., 2004; Crnich et al., 2005; Dijkshoorn et al., 2007; Asensio et al., 2008; Rodríguez-Baño et al., 2009). However, despite the increasing clinical importance of A. baumannii infections, relatively little is known about the factors that contribute to its pathogenesis. Of the studies addressing A. baumannii that have been carried out over the preceding decades, the majority either describe the epidemiology, risk factors, and outcomes of infections caused by this bacteria or aimed to optimize antibiotic regimens for the treatment of infections produced by MDR strains. While these studies provide important information regarding the epidemiology and clinical management of A. baumannii infections, they do not address the underlying biological basis for the increasing success of this organism as a human pathogen. Fortunately, a number of studies have begun to address characteristics of A. baumannii that may have contributed to its clinical emergence and to explain how this bacterium produces human disease on a molecular level. These studies have begun to provide important insight into this important human pathogen and may reveal targets for developing novel treatment and prevention strategies.
Human infections caused by A. baumannii
Hospital-acquired pneumonia represents the most common clinical manifestation of A. baumannii infection. These infections occur most typically in patients receiving mechanical ventilation in the intensive care setting. It is thought that ventilator-associated pneumonia caused by A. baumannii results from colonization of the airway via environmental exposure, which is followed by the development of pneumonia (Dijkshoorn et al., 2007). The crude mortality rate of ventilator-associated pneumonia caused by A. baumannii has been reported to be between 40% and 70% (Fagon et al., 1996; Garnacho et al., 2003), although the mortality directly attributable to A. baumannii infection has been the subject of controversy. Recently, however, a handful of studies and a systematic review have concluded that nosocomial infection with A. baumannii is associated with increased attributable mortality (Falagas et al., 2006b; Abbo et al., 2007; Falagas & Rafailidis, 2007; Lee et al., 2007) . Community-acquired pneumonia caused by A. baumannii, although much less frequent than nosocomial infection, has also been described (Anstey et al., 1992, 2002; Chen et al., 2001; Leung et al., 2006). Community-acquired pneumonia is associated with mortality rates between 40% and 60% and is often associated with underlying host factors such as alcohol abuse or chronic obstructive pulmonary disease.
Acinetobacter baumannii is also a common cause of bloodstream infections in the intensive care setting (Wisplinghoff et al., 2004). The most common sources of A. baumannii bloodstream infections are lower respiratory tract infections and intravascular devices (Seifert et al., 1995; Cisneros et al., 1996; Jang et al., 2009; Jung et al., 2010), although wound infections and urinary tract infections have also been reported as foci of infection (Seifert et al., 1995). Risk factors associated with acquiring A. baumannii bloodstream infections include immunosuppression, ventilator use associated with respiratory failure, previous antibiotic therapy, colonization with A. baumannii, and invasive procedures (García-Garmendia et al., 1999; Jang et al., 2009; Jung et al., 2010). Inappropriate empirical antibiotic therapy, comorbidities, neutropenia, and the presence of disseminated intravascular coagulation have all been associated with poorer clinical outcomes after the acquisition of A. baumannii bloodstream infections (Cisneros et al., 1996; Falagas et al., 2006a; Erbay et al., 2009). Crude mortality rates for A. baumannii bloodstream infections have been reported to be between 28% and 43% (Seifert et al., 1995; Wisplinghoff et al., 2004). As with other types of infections caused by this pathogen, the emergence of drug resistance has posed an increasing challenge to clinicians. This point is illustrated by a retrospective study performed in the United Kingdom between 1998 and 2006 in which carbapenem resistance rates rose from 0% in 1998 to 55% in 2006 in A. baumannii isolates causing bacteremia (Wareham et al., 2008). In addition to clinical complications, the emergence of drug resistance has also resulted in an additional economic burden on health systems. Lee et al. (2007) reported that bacteremia caused by MDR strains required $3758 in additional medical costs and 13.4 additional days of hospitalization per patient compared with bacteremia with non MDR strains in a tertiary care hospital in Taiwan.
Acinetobacter baumannii is an important cause of burn infections, although it can be difficult to differentiate between infection and colonization of burn sites. Because of the high rates of multidrug resistance and the poor penetration of some antibiotics into burn sites, these infections can be extremely challenging for clinicians. Recent studies reporting high incidences of A. baumannii infection in burn units have underscored the importance of A. baumannii in this patient population (Albrecht et al., 2006; Chim et al., 2007; Keen et al., 2010a, b), although the prevalence of A. baumannii burn site infection likely varies considerably depending on institution and geographic location. Acinetobacter baumannii has also emerged as an important cause of burn infection in military personnel as evidenced by a recent report characterizing bacterial infections in a military burn unit which identified A. baumannii as the most common cause of burn site infection (22%), with 53% of isolates demonstrating multidrug resistance (Keen et al., 2010b). Burn infection can be especially problematic as it can delay wound healing and lead to failure of skin grafts, and wound site colonization can progress to infection of the underlying tissue and subsequent systemic spread of the bacteria (Lyytikainen et al., 1995; Roberts et al., 2001; Trottier et al., 2007). Despite the potentially serious complications that can result from A. baumannii burn infection, data regarding clinical outcomes of infected burn patients do not provide a clear picture of the mortality attributable to the pathogen in this patient population. A case–control study of burn patients that acquired A. baumannii bloodstream infections demonstrated that infected patients had an overall mortality of 31%, whereas uninfected controls had a mortality of 14% (Wisplinghoff et al., 1999). In contrast, a retrospective cohort study found that although A. baumannii was a common cause of burn site infection, it was not an independent risk factor for mortality (Albrecht et al., 2006).
Soft tissue infections caused by A. baumannii have emerged as a significant problem in military personnel sustaining war-related trauma in Iraq and Afghanistan (Murray et al., 2006; Johnson et al., 2007; Scott et al., 2007; Sebeny et al., 2008). Like other infections caused by this organism, the treatment of these infections has been complicated by multiresistant strains. Skin and soft tissue infections related to war injury can produce cellulitis and necrotizing fasciitis, which require surgical debridement in addition to antibiotic therapy (Sebeny et al., 2008). To identify the source of infection in military treatment facilities, Scott et al. (2007) screened patient skin samples, soil samples, and treatment areas within the facilities for the presence of A. baumannii. Their findings demonstrating that A. baumannii was present on the skin of only one of 160 patients (0.6%), in only one of 49 soil samples (2%), but in all of the treatment areas suggest that the source of infection is within the treatment facilities. Separate studies assessing skin colonization have reported higher rates (Griffith et al., 2006; Doi et al., 2010), although differences in bacterial identification methodology between studies should be taken into account as they may affect reported colonization rates. In addition to military personnel, skin and soft tissue infections caused by A. baumannii were also identified in wounded survivors of the tsunami that occurred in Southeastern Asia in December of 2004 (Garzoni et al., 2005; Maegele et al., 2005). In the nonmilitary setting, A. baumannii has been reported to be a cause of surgical site infections in some institutions (Cisneros et al., 1996; Rodríguez-Baño et al., 2004) and an infrequent cause of skin and soft tissue infections in the ICU setting (Sader et al., 2002; Gaynes & Edwards, 2005).
Acinetobacter baumannii is an increasingly important cause of meningitis, with the majority of cases occurring in patients recovering from neurosurgical procedures (Siegman-Igra et al., 1993; Katragkou et al., 2006; Ng et al., 2006; Ho et al., 2007; Huttova et al., 2007; Metan et al., 2007; Paramythiotou et al., 2007; Sacar et al., 2007; Rodríguez Guardado et al., 2008; Krol et al., 2009; Cascio et al., 2010), although rare cases of community-acquired A. baumannii meningitis have been reported (Chang et al., 2000; Taziarova et al., 2007; Lowman et al., 2008; Ozaki et al., 2009). Clinical features of A. baumannii meningitis are consistent with those of bacterial meningitis caused by other organisms and include fever, altered consciousness, headache, and seizure (Rodríguez Guardado et al., 2008). Mortality rates associated with A. baumannii meningitis are difficult to estimate because of a limited number of studies with adequately sized study populations. A retrospective study identified 51 cases of postsurgical A. baumannii meningitis in two tertiary care hospitals between 1990 and 2004 (Rodríguez Guardado et al., 2008). These cases represented 10.9% of all meningitis cases at these institutions and had a crude mortality of 33%. A similar study evaluating postsurgical A. baumannii meningitis in 28 patients reported a crude mortality of 71% (Metan et al., 2007).
Osteomyelitis caused by A. baumannii occurs predominantly in military personnel sustaining war-related trauma and has become as a significant problem in U.S. military operations in Iraq and Afghanistan (Davis et al., 2005; Schafer & Mangino, 2008). A study describing 18 cases of A. baumannii osteomyelitis in wounded soldiers in a military tertiary care center reported that all patients required surgical debridement of necrotic bone and that three cases were associated with bacteremia (Davis et al., 2005). Mortality in this cohort was 0%, although it should be noted that the young age of the patients (median age; 26 years) may have contributed to the low mortality rate.
In addition to the above-mentioned infections, A. baumannii is an infrequent cause of endocarditis. Individual case reports have described A. baumannii endocarditis associated with prosthetic valves (Olut & Erkek, 2005; Menon et al., 2006; Kumar et al., 2008) and intravascular catheters (Bhagan-Bruno et al., 2010).
Antibiotic resistance: contribution of genome plasticity and effect on pathogenesis
The ability of A. baumannii to acquire antibiotic resistance mechanisms has allowed this organism to persist in hospital environments and has facilitated the global emergence of MDR strains. Especially alarming are reports describing infections caused by pandrug-resistant strains with resistance to all clinically used antibiotics (Taccone et al., 2006; Valencia et al., 2009). These strains represent a challenge for clinicians treating these infections and necessitate the development of novel strategies for preventing and treating infections caused by this organism. There are a number of reviews that provide comprehensive information on antibiotic resistance mechanisms and clinical aspects of A. baumannii infection (Chopra et al., 2008; Peleg et al., 2008b; Vila & Pachón, 2008; Fishbain & Peleg, 2010; Gordon & Wareham, 2010) . The major resistance mechanisms that have been identified in A. baumannii for different antibiotic classes are summarized in Table 1.
Table 1. Major resistance mechanisms found in Acinetobacter baumannii
Found in AbaR1 from the A. baumannii strain AYE (Fournier et al., 2006).
Found in AbaR2 from the A. baumannii strain ACICU (Iacono et al., 2008).
Mutations in the PmrA/B two-component system (LPS modification), mutations in LPS biosynthesis genes
Recent technical and computational advances have facilitated the global genomic comparative analyses of clinical isolates and have shown the remarkable capacity of A. baumannii to acquire and rearrange genetic determinants that play a critical role in its pathobiology (see Table 2 for important sequenced strains of A. baumannii). The first report describing this type of analysis showed that the MDR phenotype of A. baumannii AYE is because of the acquisition of the 86-kb AbaR1 resistance island (Fournier et al., 2006). The acquisition of this island, which includes 45 resistance genes as well as genetic traits coding for DNA mobilization functions (transposases) and is absent in sensitive strains, could be explained by horizontal gene transfer from unrelated sources. Other A. baumannii strains such as the European Clone (EC) II strain ACICU harbor the AbaR2 resistance island (Iacono et al., 2008). More recently, the comparative genome-wide analysis of ACICU and three strains belonging to the A, B, and C types determined by pulse-field gel electrophoresis isolated during an outbreak at the National Institutes of Health Clinical Center (Snitkin et al., 2011) confirmed the presence of discrete genomic regions dedicated to antimicrobial resistance. This report also showed that A. baumannii has the capacity to adapt to hospital environments not only by horizontally acquiring genetic traits responsible for the evolution of non-MDR ancestors into MDR outbreak strains, but also by rearranging preexisting genes. Acinetobacter baumannii strains can shuffle, add, and/or delete genes coding for important virulence factors, particularly those associated with cell-surface products, such as surface proteins and O-antigens and adhesins, and the expression of the functions needed to acquire essential nutrients such as iron (Snitkin et al., 2011). Taken together, these observations indicate that non-MDR strains may serve as a source of antigenic variants that could play a critical role in the diversification and emergence of MDR A. baumannii clinical isolates. Such a possibility is supported by a recent report (Imperi et al., 2011) showing that A. baumannii has a relatively small-sized core genome and a rather large accessory genome that hosts numerous antibiotic resistance and virulence determinants and is likely acquired by horizontal gene transfer processes.
Table 2. Important sequenced strains of Acinetobacter baumannii and associated iron uptake systems
A small number of studies have characterized the effect of acquisition of antibiotic resistance on the fitness and virulence of A. baumannii. A colistin-resistant strain, isolated after growth of a colistin-sensitive strain in sub-inhibitory concentrations of colistin, showed decreased in vitro and in vivo growth compared with the parental strain (López-Rojas et al., 2011). Additionally, the LD50 of the colistin-resistant strain was 10-fold higher than the parental strain in a mouse model of intraperitoneal sepsis. Targeted gene sequencing showed that the colistin-resistant strain had acquired a mutation in the pmrB gene, a mechanism that has previously been described to confer resistance to colistin (Adams et al., 2009). A separate study in which the fitness and virulence of an A. baumannii strain in which ciprofloxacin resistance was induced by growth in the presence of ciprofloxacin was compared with the parental strain showed similar results (Smani et al., 2012). The ciprofloxacin-resistant derivative induced less cell death, reduced in vitro and in vivo growth, and reduced mortality in a mouse model of peritoneal sepsis. Taken together, these studies indicate that, at least in some cases, the acquisition of antibiotic resistance in A. baumannii comes at a biological cost.
Natural habitats of A. baumannii
According to the current taxonomy, the genus Acinetobacter includes 27 valid species [J.P. Euzéby taxonomy site (http://www.bacterio.cict.fr/a/acinetobacter.html)], the most recent addition of Acinetobacter indicus sp. nov. (Malhotra et al., 2012) and nine provisional species based on DNA–DNA hybridization, all of which encompass strains found in a wide range of ecological niches. However, the most medically relevant species belong to the A. baumannii complex, which includes A. baumannii and the genomic species 3 and 13TU, which were recently renamed Acinetobacter pitti sp. nov. and Acinetobacter nosocomialis sp. nov., respectively (Nemec et al., 2011). Because of the widespread presence of Acinetobacter in different ecological niches, one of the main misconceptions regarding the natural habitat of A. baumannii is its ubiquitous presence in nature and consequent isolation from water, animal, and soil samples. Reports describing these types of isolations should be carefully considered, particularly if they refer to strains that were not identified to the species level according to the current taxonomy using validated methods, which are more accurate than those used some time ago, especially before 1986 when the taxonomy of the genus underwent major revisions (Bouvet & Grimont, 1986). Equally misleading is the concept that A. baumannii is a normal component of the human flora. On the basis of the ecology, epidemiology, and antibiotic phenotype of different isolates, Towner proposed the existence of three major Acinetobacter populations (Towner, 2009). One of them, which consists mainly of A. baumannii and closely related members of the A. baumannii complex, is represented by strains isolated from medical environments and equipment, medical personnel, and hospitalized patients. In general, these isolates tend to be resistant to multiple antibiotics, although strains such as the clinical isolates ATCC 19606T and ATCC 17978, which are sensitive to most antibiotics, clearly belong to this group. The second population is represented by strains that can be found in human and animal skin flora as well as in spoiled food samples. Members of this group include Acinetobacter johnsonii, Acinetobacter lwoffii, and Acinetobacter radioresistens. The last group includes antibiotic-sensitive isolates obtained from environmental sources such as soil and wastewater samples and mainly comprises Acinetobacter calcoaceticus and A. johnsonii. Although most members of the these last two groups are sensitive to antibiotics, some A. radioresistens, A. johnsonii, and A. calcoaceticus isolates have been found to contain carbapenemase resistance genes (Figueiredo et al., 2011). While this grouping makes sense considering the current understanding of the taxonomy, epidemiology, and ecology of different members of the Acinetobacter genus, the fact that strains such as ATCC 19606T, which is the A. baumannii-type strain (Bouvet & Grimont, 1986), and ATCC 17978, which was the first to be fully sequenced (Smith et al., 2007), are sensitive to most if not all antibiotics used in human medicine may, at first glance, contradict the existence of these groups.
The ability of A. baumannii to resist desiccation and persist on hospital materials and medical devices (Villegas & Hartstein, 2003) has played a critical role in the emergence of this bacterium as a relevant human pathogen. However, many host, environmental, and bacterial factors affecting the virulence phenotype of A. baumannii remain to be identified and characterized. For example, it was observed that exposure of A. baumannii to ethanol enhances not only its growth in media containing ethanol, but also serves as an environmental signal that controls responses to salt tolerance and increased pathogenicity when tested in Caenorhabditis elegans (Smith et al., 2004). Further genomic and mutagenesis analysis of the strain ATCC 17978 showed that the enhanced ethanol-mediated virulence response in C. elegans worms and Dictyostelium discoideum amebae relates to genes located in pathogenicity islands, some of which code for novel gene products (Smith et al., 2007). Interestingly, some of the mutants harbor mutations impairing the expression of ABC transporters, an uncharacterized urease activity, and transcriptional regulators. The latter finding suggests that ethanol could play a global regulatory function, a hypothesis that is supported by the data obtained using global RNA-sequencing (Camarena et al., 2010). This study, which resulted in the identification of 49 ethanol-induced genes coding for metabolic functions, stress responses and virulence functions, suggests that ethanol affects the pathobiology of A. baumannii. Such findings could be significant because the presence of ethanol in clinical settings may have an impact as previously reported (Edwards et al., 2007). More recently, it was reported that A. baumannii also senses and responds to light, an unexpected observation considering that this is a nonphotosynthetic microorganism (Mussi et al., 2010). This observation led to the hypothesis that the outcome of certain infections, such as surface-exposed wound infections, could depend on the exposure of bacteria to light and temperatures lower than 37 °C.
Adherence and biofilm formation appear to contribute to pathogenicity
Acinetobacter baumannii has a remarkable capacity to survive and prosper in hospital environments, most likely due to its ability to interact with different types of surfaces, including abiotic substrata normally found in medical settings, such as furniture, linen, and medical equipment (Neely et al., 1999; Neely, 2000; Villegas & Hartstein, 2003; Borer et al., 2005). Such behavior is in accordance with the described capacity of A. baumannii clinical isolates to survive long stretches of time under highly desiccated conditions on abiotic surfaces, a property that is uncommon among other Gram-negative pathogens (Wendt et al., 1997; Jawad et al., 1998). Acinetobacter baumannii also adheres to and colonizes indwelling devices such as catheters and respiratory equipment (Villegas & Hartstein, 2003), as well as biotic surfaces such as those of human epithelial cells (Fig. 1a), which may be a target during respiratory infections, or Candida albicans filaments (Fig. 1b; Lee et al., 2006). The latter type of interaction could represent the capacity of A. baumannii to interact with other components of the human microbial flora under particular environmental conditions and serve as a reservoir, as was shown with Helicobacter pylori (Salmanian et al., 2008).
The adherence of A. baumannii is variable among clinical isolates as it has been shown that strains belonging to the EC II strain are more adherent than the EC I strain to human bronchial epithelial cells, although no significant differences were observed between outbreak and nonoutbreak strains (Lee et al., 2006).
Generally, the adherence of A. baumannii to biotic and abiotic surfaces results in the development of biofilms, which are complex multicellular three-dimensional structures with cells in intimate contact with each other and encased in an extra-cellular matrix that can be comprised of carbohydrates, nucleic acids, proteins, and other macromolecules (Costerton et al., 1999). It is hypothesized that A. baumannii persists in medical environments, resists antimicrobials, and causes disease because of its capacity to form biofilms on solid surfaces (Donlan & Costerton, 2002; Gaddy & Actis, 2009). Some A. baumannii clinical isolates form complex biofilm structures on the surface of liquid media, which are known as pellicles (Fig. 2; Martí et al., 2011; McQueary & Actis, 2011). Pellicle formation and biofilm formation on abiotic surfaces are quite variable among A. baumannii clinical isolates with no apparent correlation between the nature of different types of substrata and bacterial surface properties (McQueary & Actis, 2011). Furthermore, there are significant variations not only in the amount of biofilm formed on abiotic surfaces but also in the type of cell arrangements formed on these surfaces. Some cell arrangements are simple monolayers of bacteria attached in an organized or random manner while others are complex multilayered structures encased within a biofilm matrix (McQueary & Actis, 2011).
A number of A. baumannii gene products have been shown to play a role in biofilm formation and adherence to abiotic surfaces. Initial studies showed that pilus production mediated by the CsuA/BABCDE usher-chaperone assembly system is required for attachment and biofilm formation on abiotic surfaces by the A. baumannii ATCC 19606T strain (Tomaras et al., 2003). This operon seems to be widespread among clinical isolates, an indication that the pili assembled by this system could be a common factor among different clinical isolates. However, the ATCC 19606T strain has the capacity to produce alternative pili that may participate in the interaction of this pathogen with bronchial epithelial cells (de Breij et al., 2009). Preliminary observations (C.N. McQueary & L.A. Actis, unpublished results) also indicate that A. baumannii ATCC 17978 has pili that are different from those described in the ATCC 19606T strain. The ATCC 17978 pili are long and thin and they tend to bundle; this observation is congruent with the fact that ATCC 17978 cells do not produce CsuA/B, which is considered the pilin subunit of the CsuA/BABCDE-mediated pili, and with the fact that this strain forms much weaker biofilms on plastic when compared with the ATCC 19606T strain.
In strain 307-0294, mutational loss of a large outer membrane protein, which has high similarity to the staphylococcal biofilm-associated protein (Bap), resulted in a diminishment of the volume and thickness of biofilms formed by this strain on glass (Loehfelm et al., 2008). On the basis of its cellular location and participation in biofilm formation and development, the Bap protein, which is conserved among different clinical isolates, appears to be needed for cell-to-cell interactions that support biofilm development and maturation (Loehfelm et al., 2008). The two latter processes also depend on the capacity of A. baumannii clinical isolates to produce and secrete poly-β-1-6-N-acetylglucosamine (PNAG), an exopolysaccharide produced by almost all tested strains that is critical for the formation of fully developed biofilms on glass by cells cultured statically (Choi et al., 2009).
In the ATCC 19606T strain, a two-component regulatory system comprised of a sensor kinase encoded by bfmS, and a response regulator encoded by bfmR is involved in bacteria–surface interactions (Tomaras et al., 2008). Insertional inactivation of bfmR resulted in a loss of expression of the csuA/BABCDE operon and the ensuing lack of pili production and biofilm formation on plastic when cells were cultured in rich medium (Tomaras et al., 2008). Inactivation of the bfmS sensor kinase gene resulted in a diminishment, but not abolishment of biofilm formation. When the BfmRS system was not expressed, the composition of the culture medium still influenced the interaction of cells with abiotic surfaces (Tomaras et al., 2008). This indicates that BfmR could crosstalk with other sensing components and suggests that multiple and different environmental stimuli could control biofilm formation via the BfmRS regulatory pathway.
In contrast to the ability to adhere to abiotic surfaces, much less is known regarding the A. baumannii factors that play a role in adherence to and biofilm formation on biotic surfaces. As mentioned above, this pathogen attaches to human epithelial cells and C. albicans filaments, in a process that involves at least the outer membrane protein OmpA. While OmpA could also play a role in biofilm development on plastics, this outer membrane protein is critical for the interaction of the pathogen with human and Candida cells when the latter are in a filamentous form (Gaddy et al., 2009). The ATCC 19606T-C. albicans filament interactions are independent of the pili assembled by the csu usher-chaperone system and lead to apoptotic death of the fungal filaments (Fig. 3). These results suggest that there is no direct correlation between biofilm formation on abiotic and biotic surfaces, and that there is wide variation in the cell-surface and cell–cell interactions that result in adherence and biofilm formation by different A. baumannii clinical isolates. In spite of this information, the role of pili in bacterial virulence and the pathogenesis of the infections A. baumannii causes in humans remains to be confirmed using appropriate derivatives and experimental infection models.
Adherence and biofilm formation are well-orchestrated processes that respond to a wide range of cellular and environmental cues (Stanley & Lazazzera, 2004). For instance, the ability of A. baumannii to form biofilms could depend on the presence and expression of antibiotic resistance traits, such as the blaPER-1 gene. A positive correlation was found between the presence and level of expression of this gene and the amount of biofilms formed on plastic and the adhesiveness of bacteria to human epithelial cells (Lee et al., 2008). However, an independent study found that only two of 11 isolates carrying the blaPER-1 gene formed stronger biofilms when compared with isolates lacking this genetic determinant (Rao et al., 2008), which brings to question the relevance of the presence and expression of this gene in biofilm formation by A. baumannii isolates. Environmental cues such as temperature and the concentration of extracellular free iron, which are relevant for the interaction of A. baumannii with the host, also affect the amount of biofilm formed by this pathogen on abiotic surfaces. Acinetobacter baumannii ATCC 19606T formed more biofilm when cultured in LB broth at 30 °C or in M9 minimal medium at 37 °C in the presence of the synthetic iron chelators 2,2′-dipyridyl (DIP) or ethylenediamine-di-(o-hydroxyphenyl) acetic acid (EDDHA; Tomaras et al., 2003). However, a global transcriptomic analysis later showed a significant down-regulation of some of the csu genes in cells cultured under iron-chelated conditions (Eijkelkamp et al., 2011). This apparent discrepancy might be explained by the fact that these two studies used different strains (ATCC 19606T vs. ATCC 17978). In ATCC 17978 cells, the csu operon does not appear to be active and iron-regulated cell products other than pili may play a role in adherence and biofilm formation (McQeary Zimbler and Actis, unpublished observations). The chelating agent EDTA also affects the interaction between clinical A. baumannii isolates with biotic and abiotic surfaces, as it has been shown that it significantly reduces bacterial attachment and biofilm formation on human respiratory epithelial cells and plastic surfaces (Lee et al., 2008). The molecular mechanisms by which these chelators produce this effect remain to be elucidated.
Cell population density is another mechanism by which bacteria control adherence and biofilm formation. Accordingly, environmental and clinical isolates produce quorum sensing signaling molecules (González et al., 2001, 2009). Interestingly, these studies showed that a large proportion of the tested isolates produce one or more quorum sensors that seem to belong to three types of molecules. Although none of these sensors could be assigned to a particular species, the Rf1-type sensor is more frequently found in isolates belonging to the A. calcoaceticus-baumannii complex. More detailed studies showed that the A. baumannii M2 clinical isolate produces an N-acyl-homoserine lactone [i.e. N-3-hydroxy-dodecanoyl-homoserine lactone (3-OH-C12-HSL)], the product of the abaI autoinducer synthase gene, which is important for the formation of fully developed biofilms on abiotic surfaces (Niu et al., 2008). This autoinducer also plays a role in the ability of this strain to move on semisolid media, as described in the next section. Finally, as mentioned above, the observation that light affects biofilm formation on abiotic surfaces was unexpected considering that A. baumannii is a chemotroph not known to conduct photosynthesis (Mussi et al., 2010). This response is mediated by the BlsA photoreceptor protein, which contains a BLUF domain and uses FAD to sense light. The mechanisms by which BlsA transduces the light signal and controls gene expression are not known (Mussi et al., 2010). The A. baumannii response to light seems to have a global effect on the physiology of A. baumannii, affecting not only biofilm formation but also motility and virulence. Furthermore, the differential response to illumination is modulated by temperature changes, which result in differential transcription of blsA at 28 and 37 °C and hence differentially affect light-controlled phenotypes (Mussi et al., 2010).
In conclusion, biofilm formation and adherence in A. baumannii clinical isolates involves a range of bacterial factors and multiple signals or cues. However, the medical relevance of data obtained using in vitro models is not clear, considering the lack of correlation between the biofilm phenotype of different clinical isolates and their outbreak, epidemic and antibiotic resistance nature (de Breij et al., 2010). By comparison with other bacterial pathogens, such as Pseudomonas aeruginosa, very little is known about the A. baumannii products involved in pathogenicity mechanisms and the cellular and environmental signals that control them within the vertebrate host.
Motility on semi-solid surfaces: does it impact on pathogenicity?
Motility in A. baumannii is counterintuitive considering that the name of this genus (acinetobacter, nonmotile bacterium) implies the inability of members of this genus to move. However, this phenotype was reported more than 30 years ago when Henrichsen described the influence of the environment on the ability of A. calcoaceticus to move on agar plates (Henrichsen, 1975; Henrichsen & Blom, 1975). This issue recently resurfaced because of the observation that A. baumannii displays differential motility in response to illumination (Mussi et al., 2010), quorum sensing (Clemmer et al., 2011), and iron chelation (Eijkelkamp et al., 2011). Although the type of motility displayed by Acinetobacter strains has not been elucidated unequivocally, current phenotypic and genetic evidence suggests that this pathogen moves on semi-solid surfaces by expressing twitching motility rather than gliding, sliding, swimming, or swarming motility (Barker & Maxted, 1975; Henrichsen, 1984; Eijkelkamp et al., 2011). This possibility is supported by the observation that iron affects motility and the expression of pil-com ATCC 17978 genes (Eijkelkamp et al., 2011) that participate in the assembly and function of type IV pili, which are known to be involved in twitching motility (Mattick, 2002). Furthermore, insertional inactivation of pilT, which codes for an ATPase activity involved in pilus retraction in other bacteria (Merz et al., 2000; Mattick, 2002), resulted in a significant reduction in the motility of the A. baumannii M2 strain (Clemmer et al., 2011). However, a pilT::Km derivative of strain M2 was still motile on the surface of 0.35% Eiken agar, suggesting that the strain has type IV pili-independent motility functions. Such a possibility is supported by the reduced motility of M2 transposon insertion derivatives affected in the expression of genes that do not code for type IV pilus assembly and function (Clemmer et al., 2011). Sequence analysis of some of these derivatives showed the potential involvement of a lipopeptide, degradation of peptidoglycan, synthesis of O-antigen, OmpA, and the histidine kinase sensor BfmS, which regulates the expression of the csuA/BABCDE operon via the BfmR response regulator (Tomaras et al., 2008).
Most if not all movement displayed by A. baumannii occurs on the surface of semi-solid media and diminishes as the concentration of agar or agarose in the medium increases (Clemmer et al., 2011; McQueary & Actis, 2011). Furthermore, different types of agar affect the outcome of the motility response with complex patterns that manifest as either well-defined brunches, with some of them resembling ditching motility already described in Acinetobacter anitratus (Barker & Maxted, 1975), or circular cell expansion from the inoculation point with an even pattern of cells on the surface of the agar (Clemmer et al., 2011; McQueary & Actis, 2011). Different strains display different patterns, and not all tested strains move on semi-solid surfaces (Fig. 4; Clemmer et al., 2011; McQueary & Actis, 2011). Currently, we do not know whether motility depends on different motility systems or variations in the capacity of different strains to sense the appropriate environmental cues controlling this complex multicellular process. For instance, the availability of free iron, which is well known for its impact in bacterial human infections (Weinberg, 2009), plays a role in the pil-com-dependent motility of A. baumannii, as mentioned before (Eijkelkamp et al., 2011). Cell population density is another factor that controls motility as inactivation of the abaI gene (specifying the autoinducer 3-OH-C12-HSL) resulted in a drastically reduced motility response, which was corrected by the supplementation of purified autoinducer (Clemmer et al., 2011). Light, particularly blue light, is a third environmental signal that controls A. baumannii motility on semi-solid media by mechanisms that remain to be elucidated (Mussi et al., 2010).
In conclusion, A. baumannii is capable of moving on the surface of semi-solid media by processes that are mediated, at least in part, by type IV pili-dependent mechanisms, which are affected by environmental and cell signals that also affect bacterial virulence. However, it is unclear whether motility plays a significant role in the virulence of A. baumannii and the pathogenesis of the serious infections that it causes in the human host. This is because not all clinical isolates display motility when tested under laboratory conditions (which undoubtedly do not reflect those encountered by this pathogen in the human host) and because appropriate nonmotile isogenic derivatives have not been tested in relevant animal models.
Iron acquisition from heme and via siderophores
Although iron is abundant in environmental and biological systems, ferric iron is relatively unavailable to cells because of its poor solubility under aerobic conditions and its chelation by low-molecular-weight compounds such as heme and by high-affinity iron-binding proteins such as lactoferrin and transferrin. In response to iron limitation, most aerobic bacteria express high-affinity iron acquisition systems that mainly include the production, export, and uptake of Fe3+ chelators known as siderophores. In addition, some bacteria utilize heme or hemoglobin as an iron source, and some are able to remove iron from transferrin or lactoferrin (Crosa et al., 2004; Wandersman & Delepelaire, 2004). Acinetobacter baumannii does not bind transferrin (Echenique et al., 1992) and does not carry genetic determinants coding for the proteins involved in the acquisition of iron from transferrin and lactoferrin (Smith et al., 2007). However, the ATCC 19606T strain uses heme as an iron source (Zimbler et al., 2009). The chromosomal cluster annotated as A1S_1608-A1S_1614 in strain ATCC 17978 seems to be a polycistronic operon that could be involved in the transport of heme from the periplasm into the cytoplasm (Smith et al., 2007). A more recent genomic analysis (Antunes et al., 2011a) showed that different strains can use this compound as an iron source expressing potential heme uptake and utilization systems (Table 2). Taken together, these observations indicate that the A. baumannii genome contains genes coding for products devoted to the capture and utilization of heme, a host product that could be available to bacteria at sites where extensive cell and tissue damage are produced by infections such as necrotizing fasciitis (Brachelente et al., 2007; Charnot-Katsikas et al., 2009; Corradino et al., 2010) or in severely injured patients (Peleg et al., 2008b). Acinetobacter baumannii may also acquire ferrous iron (Table 2), which would be available under low-oxygen tension conditions, because fully sequenced and annotated genomes show the presence of genes coding for a Feo transport system (Antunes et al., 2011a), the function of which remains to be tested experimentally.
Acinetobacter baumannii is also capable of acquiring ferric ions under iron-limited conditions via siderophores. The best-characterized system is that mediated by the siderophore acinetobactin, which was initially described in the ATCC 19606T strain and has a molecular structure highly related to anguibactin (Fig. 5), a high-affinity iron chelator produced by the fish pathogen Vibrio anguillarum 775 (Yamamoto et al., 1994, 1999; Dorsey et al., 2004; Mihara et al., 2004). The main difference between these two siderophores is that acinetobactin contains an oxazoline ring derived from threonine, while anguibactin has a thiazoline group derived from cysteine. In spite of this difference, these siderophores, which are produced by two unrelated bacterial pathogens found in different environments, are also functionally related because both of them crossfeed derivatives affected in their production (Dorsey et al., 2004). Genetic and functional analyses (Dorsey et al., 2004; Mihara et al., 2004) have shown that a 26.5-kb chromosomal region harbors genes coding for all functions needed for the biosynthesis, transport, and secretion of acinetobactin with the exception of entA, which codes for the biosynthetic enzyme 2,3-dihydro-2,3-dihydroxybenzoate dehydrogenase. This enzyme is needed for the production of the dihydroxybenzoic acid (DHBA) moiety present in the acinetobactin molecule (Fig. 5). This observation indicates that A. baumannii could contain more than one locus involved in siderophore biosynthesis, a possibility that is supported by the recent information made available by genomic analysis of fully sequenced and annotated A. baumannii isolates (Table 2; Antunes et al., 2011a; Eijkelkamp et al., 2011). The report by Eijkelkamp et al. (2011) describes five gene clusters predicted to be involved in siderophore production and utilization for iron acquisition. The 26.5-kb cluster mentioned above related to acinetobactin, which was named as the acinetobactin cluster, and cluster 1 are present in the genomes of all strains analyzed with the exception of 8399 and the environmental strain Acinetobacter baylyi ADP1 (Table 2). Cluster 2 is present only in A. baumannii ATCC 17978 and A. baylyi ADP1 while cluster 4 seems to be unique to isolate 8399. Cluster 5 is found in all genomes analyzed except that of ATCC 17978 and ADP1 . It is important to note that the SDF strain does not contain any gene cluster related to siderophore-mediated iron acquisition functions.
Taken together, the available experimental and in silico observations indicate that A. baumannii can acquire iron either by using heme as an iron source or by capturing this metal with acinetobactin and/or one or more additional siderophore-mediated systems. A recent report by Gaddy et al. (2012) showed that the acinetobactin-mediated system plays a critical role in the ability of A. baumannii ATCC 19606T to persist and cause cell damage and animal death when tested using human epithelial cells, Galleria mellonella caterpillars, and mice infection models, indicating that iron acquisition functions play a critical role in virulence. Whether the other iron acquisition systems listed in Table 2 also play a virulence role or provide an ecological advantage for their persistence in particular ecological niches are interesting possibilities that remain to be examined.
Compared to other Gram-negative pathogens, relatively few virulence factors have been identified for A. baumannii. The recent sequencing of numerous A. baumannii complete genomes (Table 2) and the application of methods for manipulating the bacterial genome to generate gene-deficient mutants, together with the use of animal models, have been crucial in the identification of bacterial factors that contribute to pathogenesis (Table 3). These studies have begun to shed light on how this pathogen persists in the environment, interacts with host cells, and causes host cell damage, although there are undoubtedly numerous additional factors that have yet to be identified.
Probably the best-characterized virulence factor of A. baumannii identified to date is OmpA. Evidence that A. baumannii OmpA contributes to virulence was obtained in a random transposon mutagenesis screen that detected A. baumannii mutants that were deficient in inducing apoptosis in a human laryngeal epithelial cell line (Choi et al., 2005). Purified OmpA localized to the mitochondria and induced apoptosis through the release of the proapoptotic molecules cytochrome c and apoptosis-inducing factor, suggesting that this may be one pathway by which A. baumannii induces damage to human airway cells during infection. The OmpA protein also plays a role in adherence and invasion of epithelial cells may contribute to the dissemination of A. baumannii during infection, as the bacterial loads in blood of mice with experimentally induced A. baumannii pneumonia were significantly higher in mice infected with a wild-type strain than in mice infected with an equivalent amount of an isogenic ompA mutant (Choi et al., 2008). The OmpA protein also contributes to the ability of A. baumannii to persist and grow in human serum as it has been shown that OmpA interacts with soluble inhibitors of the alternative complement pathway and allows the bacteria to avoid complement-mediated killing (Kim et al., 2009). However, OmpA is unlikely to be the only factor that contributes to serum resistance as different A. baumannii strains, all of which contain putative ompA genes, have significantly different capacities for growth and survival in human serum (Antunes et al., 2011b). As mentioned before, in the environment, OmpA may also facilitate the persistence and survival of A. baumannii by assisting biofilm formation and surface motility (Gaddy et al., 2009; Clemmer et al., 2011). Because of the multiple functions of OmpA in the pathogenesis of A. baumannii, it may be an attractive target for the development of novel treatment and prevention strategies.
Acinetobacter baumannii LPS contains a lipid A moiety, the carbohydrate core, and the repetitive O-antigen. The role of LPS in A. baumannii pathogenesis was recently investigated using a mutant lacking the LpsB glycotransferase that results in a highly truncated LPS glycoform containing only two carbohydrate residues bound to lipid A (Luke et al., 2010). This mutant showed decreased resistance to human serum and decreased survival in a rat model of soft tissue infection compared with the isogenic parent strain, indicating a role for the surface carbohydrate residues of LPS in pathogenesis. It has also been demonstrated that CD14 and Toll-like receptor 4 play a role in clearing A. baumannii from the lung through detection of LPS (Knapp et al., 2006), suggesting that A. baumannii LPS activates the innate immune response. Studies further addressing the inflammatory response and the mortality elicited by LPS mutants in animal models would be of interest for further characterizing the contribution of A. baumannii LPS to virulence.
The structures of the capsular polysaccharides from two clinical isolates of A. baumannii were recently reported revealing a linear aminopolysaccharide consisting of three carbohydrate residues in one strain and a branched pentasaccharide in the other (Fregolino et al., 2011). In addition to LPS, the capsular polysaccharide has also been identified as a pathogenicity factor in A. baumannii. Mutants that were deficient for growth in human ascites fluid because of transposon insertion in the ptk or epsA gene failed to produce a capsule-positive phenotype. These mutants showed decreased growth in both human serum and ascites compared with the wild-type counterpart. Additionally, capsule-deficient strains were completely cleared by 24 h postinfection in a rat model of soft tissue infection, whereas the isogenic parental strain persisted with > 107 bacteria mL−1 of exudative fluid (Russo et al., 2011). Thus, the capsular polysaccharide appears to play an important role in protecting bacteria from the host innate immune response.
Bacterial phospholipases are lipolytic enzymes that catalyze the cleavage of phospholipids. These enzymes are thought to contribute to the pathogenesis of Gram-negative bacteria by aiding in the lysis of host cells, via cleavage of phospholipids present in the host cell membrane, and by degrading phospholipids present at mucosal barriers to facilitate bacterial invasion. Disruption of one of the two phospholipase D genes present in A. baumannii genome resulted in reduced survival in serum and a reduced capacity for invading epithelial cells (Jacobs et al., 2010). In addition, in a murine model of pneumonia, mice infected with the phospholipase D-deficient strain had lower bacterial burdens in the heart and liver at 48 h postinfection than mice infected with the parental strain, suggesting that phospholipase D contributes to the ability of A. baumannii to disseminate from the lungs. The A. baumannii genomes sequenced to date also demonstrate the presence of two putative phospholipase C genes, and culture supernatants from different A. baumannii strains contain phospholipase C activity, indicating that these genes are expressed and the resulting proteins are secreted from the bacterial cell (Antunes et al., 2011b). In a separate study, inactivation of one of the phospholipase C genes resulted in a modest decrease in the ability of the mutant strain to induce cell death in an epithelial cell line compared to the parental strain (Camarena et al., 2010).
Penicillin-binding proteins (PBPs) are most commonly associated with binding to and inactivating β-lactam antibiotics. However, PBPs also participate in the final steps of the biosynthesis of the peptidoglycan layer and thus contribute to bacterial cell stability (Sauvage et al., 2008). An A. baumannii mutant with a transposon insertion in the pbpG gene, which encodes the putative low-molecular-weight penicillin-binding protein PBP7/8, demonstrated reduced growth on ascites plates (Russo et al., 2009). While this mutant grew similarly to its wild-type counterpart in laboratory media, it showed reduced growth in human serum and decreased survival, both in a rat model of soft tissue infection and a mouse model of pneumonia. Thus, the mutant had reduced fitness in conditions that mimic the environment encountered by the bacterium during the infectious process. Electron microscopy of the mutant and wild-type strains showed a difference in bacterial morphology, supporting the idea that a lack of PBP7/8 may affect cell stability, possibly via effects on the peptidoglycan layer (Russo et al., 2011).
Outer membrane vesicles (OMVs) are vesicles secreted from the outer membrane of various Gram-negative bacteria and consist of outer membrane and periplasmic proteins, phospholipids, and LPS. OMVs have been reported to participate in bacterial virulence by delivering virulence factors to the interior of host cells, to facilitate horizontal gene transfer, and to protect the bacteria from the host immune response (Ellis & Kuehn, 2010). Multiple strains of A. baumannii secrete OMVs during growth in vitro (Kwon et al., 2009; Jin et al., 2011; McConnell et al., 2011b; Rumbo et al., 2011) . A proteomic analysis of purified A. baumannii OMVs identified the OmpA protein as well as putative proteases and a putative hemolysin as potential virulence factors. Moreover, OMVs could deliver OmpA to the interior of eukaryotic cells and induce cell death, exposure of cells to OMVs isolated from an ompA-deficient strain did not produce cell death (Jin et al., 2011). Interestingly, OMVs were also able to facilitate the horizontal transfer of the OXA-24 carbapenemase gene, indicating that OMVs may play a role in the spread of antimicrobial resistance in A. baumannii (Rumbo et al., 2011).
Animal models of A. baumannii infection
As with most bacterial diseases, experimentation in animals is necessary for characterizing both the invading pathogen and the host response. Animal models provide a means for evaluating novel treatments, characterizing the host immune response, and identifying bacterial virulence factors. In the first report describing an animal model for A. baumannii infection, a mouse model of acute pneumonia was described (Joly-Guillou et al., 1997). Since this report, the number of animal models used for characterizing A. baumannii infection has multiplied, allowing the study of various types of infectious pathologies in a wide range of animal species including rodent, nonrodent, and even nonmammalian models of infection. These studies have begun to yield information about A. baumannii biology and its interaction with the host.
A major challenge in the use of animal models to characterize infectious pathologies is to accurately mimic human disease. Although it is unrealistic to expect that infection of laboratory animals, which in many cases are inbred strains, can completely reproduce the human disease process, efforts should be made to simulate these conditions as closely as possible. An important aspect to consider when employing an animal model of A. baumannii infection is the selection of the animal strain. In the case of mouse models of A. baumannii infection, a number of inbred strains of mice have been employed including C57BL/6, Balb/c, and A/J strains. These strains, although they are likely to give a more homogeneous response to infection, may not accurately reflect the variable response that likely occurs in humans. It is worth noting that a number of outbred strains of mice, such as CD-1 and Swiss Webster mice, have been successfully used in models of A. baumannii infection (Joly-Guillou et al., 1997; Braunstein et al., 2004; Crandon et al., 2009; Koomanachai et al., 2009; Song et al., 2009). The animal strain employed can also dramatically affect the outcome of infection. For instance, individual mouse strains have different susceptibilities to infection with A. baumannii, which can result in significant differences in survival, tissue bacterial load, clinical signs, inflammatory cell infiltration, and cytokine response (Joly-Guillou et al., 1997; Qiu et al., 2009a). For these reasons, the effect of the animal strain employed should be carefully considered.
A second important aspect that should be considered is the A. baumannii strain that will be employed in an animal model. Although well-defined, sequenced strains of A. baumannii are available from the American Type Culture Collection (e.g. ATCC 19606T and ATCC 17978; see Table 2), their clinical background and epidemiology are unknown. For animal studies, it may be more appropriate to use more recently isolated strains that have known clinical relevance, for example strains belonging to the European clones I–III. The majority of studies performed using animal models have employed uncharacterized clinical isolates for which there are often few epidemiological and microbiological data, making the generalization of results difficult. This point is illustrated by the fact that in mouse models of A. baumannii infection, the virulence of individual strains can vary widely. In these models, the use of different strains resulted in significant differences in mortality, tissue bacterial load, histological score, and inflammatory cytokine levels (Eveillard et al., 2010; McConnell et al., 2011a). These findings underscore the importance of strain selection, which would ideally include strains with proven potential to cause infections and spread epidemically. In addition, once a tentative virulence trait has been identified in a reference strain, ideally studies using large collections of well-defined strains should be performed to assess the prevalence of this trait in clinically relevant isolates.
Pneumonia models of infection
Studies employing animal models of A. baumannii pneumonia have overwhelmingly employed mice, presumably because of their low cost and ease of handling. The mouse pneumonia model has been used to measure a number of variables related to infection such as mortality, tissue bacterial load, cytokine levels in both bronchoalveolar lavage fluid and serum, inflammatory cell infiltration into the lung, and histological score. In the majority of models developed to date, pneumonia is produced either through direct instillation of a bacterial suspension (typically containing between 107 and 109 CFU) into the trachea or by intranasal instillation of the inoculum. For intratracheal instillation, a technique developed for inducing pneumonia with Haemophilus influenzae has been employed (Esposito & Pennington, 1984). Animals are suspended vertically, and the bacterial suspension is instilled with a microliter syringe using a blunt-tipped needle (Joly-Guillou et al., 1997; Rodríguez-Hernández et al., 2000). Correct positioning of the needle in the trachea is confirmed by localizing the tracheal cartilage with the tip of the needle. After instillation of the bacterial suspension, animals are maintained in a vertical position to facilitate inoculation of the lower airways. For intranasal inoculation, a bacterial suspension is applied to the nares of anesthetized mice and inhaled (Knapp et al., 2006; van Faassen et al., 2007; Crandon et al., 2009; Qiu et al., 2009a, b; Jacobs et al., 2010). While most studies employ intratracheal or intranasal infection, an oral inoculation technique for A. baumannii has recently been described in which a bacterial suspension was instilled in the oral cavity and the nares were completely blocked to facilitate inhalation of the inoculum (Koomanachai et al., 2009).
There is no evidence supporting the idea that one inoculation technique (intratracheal vs. intranasal) more accurately reproduces human disease than the other. It could be argued that the intranasal technique may more closely mimic the clinical situation in which a patient's upper airway is colonized before producing pneumonia, and however, there is no literature support for this presumption. In animal models of pneumococcal pneumonia, the intranasal route produces a bronchopneumonia, whereas the intratracheal route results in a lobar pneumonia (Azoulay-Dupuis et al., 1991; Canvin et al., 1995). However, these results have not been confirmed in the case of A. baumannii models. In mice, both routes of infection produce an acute pneumonia that is accompanied by an increase in inflammatory cytokine levels in the lungs and histological changes consistent with pneumonia (Joly-Guillou et al., 1997; Knapp et al., 2006; Qiu et al., 2009a; Eveillard et al., 2010). Both intratracheal and intranasal infection have been reported to produce mortality in mice. The bacterial and mouse strains used, as well as whether or not the animals were rendered neutropenic prior to infection, can influence the level of mortality (Joly-Guillou et al., 1997; Braunstein et al., 2004; van Faassen et al., 2007; Crandon et al., 2009; Qiu et al., 2009a; Eveillard et al., 2010).
Because of the low virulence of most A. baumannii strains in mice, and for this reason the high infection doses that must be used to establish infection, a number of studies have employed neutropenic mouse models of pneumonia to facilitate infection (Joly-Guillou et al., 1997, 2000; Wolff et al., 1999; Braunstein et al., 2004; Crandon et al., 2009; Koomanachai et al., 2009; Song et al., 2009; Eveillard et al., 2010; Yuan et al., 2010). To induce transient neutropenia, mice are treated with cyclophosphamide, usually in two doses starting 4 days before infection. The first description of the neutropenic model in C3H/H3N mice reported weight loss in the infected animals, bacterial growth in the lungs, and 85% mortality between days 2 and 3 (Joly-Guillou et al., 1997). Studies employing the neutropenic mouse model have almost exclusively had the objective of characterizing antibiotic efficacy. However, a recent study used this model to compare the virulence of different clinical isolates of A. baumannii (Eveillard et al., 2010). The authors report that mortality and tissue bacteria load varied depending on the infecting strain. A caveat is that the neutropenic mouse model may not accurately reproduce some features of infection in non-neutropenic patients, as neutrophils play an important role in the early stages of A. baumannii infection in mice (van Faassen et al., 2007).
A commonly used alternative to rendering mice transiently neutropenic is the use of porcine mucin in the inoculum. In these models, the cultured bacteria are mixed with porcine mucin, typically to a final concentration of 5% mucin, before instillation of the inoculum into the lungs (Rodríguez-Hernández et al., 2000, 2001; Montero et al., 2002, 2004; Pachón-Ibañez et al., 2006, 2010; Beceiro et al., 2009; Chiang et al., 2009; Pichardo et al., 2010). Mucin has long been known to increase the virulence of numerous bacterial species in mouse models (Olitzki, 1948). Its use thus allows for the number of bacteria in the inoculum to be reduced dramatically, in some cases by as much as 1000-fold for some bacterial species (Batson et al., 1950). Although the presence of mucin during the establishment of infection has no role in human infection, the use of lower amounts bacteria in the inoculum followed by bacterial growth and the development of pneumonia may more accurately mimic human disease than the rapid instillation of large quantities of bacteria into the lung. Porcine mucin has been used exclusively with the intratracheal route of infection, likely due to the difficulty of administering a viscous solution via the intranasal route.
In a mouse pneumonia model that did not require the induction of neutropenia or the use of porcine mucin, A/J mice were more susceptible to intranasal infection with the ATCC 17961 strain of A. baumannii than were C57BL/6 mice (Qiu et al., 2009a, b). An inoculum of 1.5 × 108 bacteria in saline produced 100% mortality by day 3 postinfection. Infection of the A/J strain resulted in reduced early neutrophil infiltration into the lung, lower levels of pro-inflammatory cytokines (IL-1β, MIP-2, and TNF-α), increased bacterial replication, and increased extrapulmonary dissemination compared to C57BL/6 mice (Qiu et al., 2009a).
The mouse pneumonia model has recently been employed for characterizing host factors involved in respiratory infections. The availability of transgenic mice with well-defined genetic changes has facilitated the characterization of these factors to the molecular level. Mice deficient in CD14 and Toll-like receptors 2 and 4 have been used to characterize the role of these molecules in respiratory infections (Knapp et al., 2006). This study concluded that CD14 and Toll-like receptor 4 play a role in clearing A. baumannii from the lung through detection of LPS, while Toll-like receptor 2 appears to down-regulate the immune response to A. baumannii pulmonary infection. Another study characterizing the role of NADPH phagocyte oxidase in pulmonary infection showed that mice lacking the gene encoding this enzyme had higher postinfection bacterial loads and increased mortality after infection, demonstrating the importance of the neutrophil response in controlling A. baumannii infection in mice (Qiu et al., 2009b).
Although the overwhelming majority of animal models of A. baumannii pneumonia have employed mice, models using other species have been developed to study the biology of this pathogen during pulmonary infection. A rat pneumonia model previously used to characterize Escherichia coli respiratory infection has been adapted for the study of A. baumannii (Russo et al., 2008). In this model, Long Evans rats are anesthetized before surgical exposure of the trachea and instillation of the inoculum into the trachea via a needle. Infected rats demonstrated many characteristics that mimic human infection including bacterial growth in the lung, decreased arterial oxygenation (as assessed by PaO2/FiO2), increased cytokine levels (TNF-α, IL-1β, and CINC-1) in bronchoalveolar lavage fluid, neutrophil infiltration and mortality. One advantage of the rat model is that, because of their larger size, rats can be infected via direct instillation of the inoculum into the trachea, which may increase the reproducibility of infection by ensuring that the entire inoculum is instilled into the airways. It should also be noted that the rat model described by Russo et al. (2008) did not require the use of porcine mucin or the use of animals that had been rendered transiently neutropenic prior to infection. A guinea-pig model of pneumonia was developed to characterize the efficacy of antibiotic combinations in the treatment of MDR A. baumannii (Bernabeu-Wittel et al., 2005). This model uses an inoculation technique similar to that described above for inducing pneumonia in rats in that the inoculum is instilled directly into the surgically exposed trachea via a needle. However, in contrast to the rat model, the inoculum contained 5% porcine mucin to increase the virulence of the infecting strain. Histopathological studies of infected lungs demonstrated that infection produced acute inflammation and resulted in neutrophil infiltration into the peribronchial and perivascular spaces. Lung bacterial loads 24 h postinfection in untreated animals receiving an inoculum of approximately 3.2 × 109 were between 7.9 × 106 and 2.0 × 107 CFU g−1 lung tissue, and mortality at this time point was between 83% and 92%, depending on the A. baumannii strain used (Bernabeu-Wittel et al., 2005).
Skin and soft tissue infection models
Both mouse (Shankar et al., 2007; Dai et al., 2009; DeLeon et al., 2009; Thomas-Virnig et al., 2009) and rat (Uygur et al., 2009) models of A. baumannii burn infection have been developed. In these models, a nonlethal, full-thickness burn is produced on the dorsal surface of the anesthetized animal after shaved skin is either submerged in a heated water bath or exposed to preheated metal blocks. Infection is subsequently initiated either through topical application or subcutaneous injection of a solution containing A. baumannii at the burn site. The objective of the studies employing the burn model has exclusively been to evaluate the efficacy of different experimental treatments. For this reason, the most common parameter measured in these studies is the bacterial concentration present at the burn site. In addition to measuring bacterial concentrations at the burn site, some studies have also measured bacterial loads at different anatomical sites (underlying muscle, lung, and blood) and survival (Shankar et al., 2007; Dai et al., 2009; Uygur et al., 2009). Because of the fact that these studies aimed to evaluate treatment efficacy, there are few data available regarding the host response to burn infection by A. baumannii.
Animal models for studying soft tissue infection by A. baumannii have predominantly employed two distinct techniques for initiating infection. The first technique involves injection of a solution containing A. baumannii directly into the thigh muscle. This approach has been used in mice to measure tissue bacteria loads in the muscle after treatment (Dijkshoorn et al., 2004), and in a neutropenic rat model in which survival and bacterial loads in distant organs were measured (Pantopoulou et al., 2007). In the neutropenic rat model, bacteria disseminated to distant organs (lung, liver, and spleen) and infection produced 100% mortality at 2 days in untreated control mice, suggesting that this model results in sepsis secondary to infection of the thigh muscle. A variation on this model is the implantation of filter paper impregnated with a suspension of A. baumannii in muscle via a surgical incision (Fetiye et al., 2004). These approaches have been used to evaluate the efficacy of treatments on A. baumannii infection, and no data regarding the host response or the bacterial factors involved in the pathogenesis of the infection were reported.
The second approach used for producing soft tissue infection utilizes the formation of a subcutaneous pouch to mimic the formation of an abscess (Russo et al., 2008). In this model, a subcutaneous space is produced on the dorsal surface of an anesthetized rat by injection of air. The space is then filled with 1 mL of an oil mixture, and the rats are maintained for 7 days to allow for the accumulation of exudative fluid which is then inoculated with the desired quantity of bacteria (5 × 106 CFU). It has been reported previously that this model results in neutrophil infiltration, indicating that the response to infection is similar to that seen during natural infection (Dalhoff et al., 1982, 1983). An advantage of this model is that the pouch fluid can be repeatedly sampled at different time points, allowing for continuous measuring of bacterial concentrations and the host response. While both inoculation techniques appear to share some characteristics with human infection, it is not clear how well these models mimic soft tissue infection produced by traumatic injury. Further study characterizing the host response to infection in these models will help to provide insight into this issue.
Animal models of sepsis
Compared to pneumonia and burn and soft tissue infection models, relatively few models of A. baumannii sepsis have been developed. In the first model of sepsis, a bacterial suspension is administered intraperitoneally in Balb/c mice (Ko et al., 2004). This approach produced mortality between 87.5% and 100% with an inoculum of approximately 2 × 107 CFU of a clinical isolate. More recently, a similar model has been employed to characterize the efficacy of vaccines for preventing infection by A. baumannii (McConnell & Pachón, 2010; McConnell et al., 2011a) and to evaluate the effect of antibiotic resistance on virulence and fitness (López-Rojas et al., 2011) and the role of the acinetobactin-mediated iron uptake system in virulence (Gaddy et al., 2012). In this model, the bacterial inoculum is first combined with porcine mucin to a final concentration of 5% before intraperitoneal instillation to increase the virulence of the infecting strain, allowing for the use of inocula containing fewer bacteria. In this model, different A. baumannii strains had widely varying LD50 values (up to an 80-fold difference in the strains tested), suggesting that this model can be used to characterize the virulence of different isolates. After intraperitoneal instillation of the inoculum, bacteria rapidly disseminated to distant organs (spleen, kidney, and lung), with bacterial loads in these tissues reaching approximately 1 × 105 CFU g−1 of tissue at 1 h postinfection. Serum levels of pro-inflammatory cytokines (TNF-α, IL-1β, and IL-6) were elevated after infection, consistent with the cytokine storm produced during the development of septic shock. Bacteria continued to reproduce in distant organs until death occurred between 18 and 48 h, depending on the strain used for infection and the number of bacteria in the inoculum. Importantly, 100% mortality was obtained by 24 h after infection with 5 × 104 CFU of the ATCC 19606T strain (McConnell et al., 2011a). Because of the high mortality that can be reproducibly achieved with relatively small inocula (compared to mouse pneumonia models), this model is also convenient for the use of vaccine efficacy studies. In a rat model of A. baumannii sepsis, the efficacy of various antibiotic regimens has been determined (Cirioni et al., 2009). Similar to the mouse model of sepsis, the inoculum is injected intraperitoneally, however without the addition of porcine mucin. In this study, bacterial loads in the blood, peritoneum, spleen, liver, and mesenteric lymph nodes were measured, in addition to plasma endotoxin and cytokine levels. Infection resulted in 100% mortality 48 h after infection, although it should be noted that a high inoculum was used (2 × 1010 CFU), which may not be physiologic.
The advantages of the animal models of sepsis developed for A. baumannii are a facile inoculation procedure, high mortality, and the possibility of measuring tissue bacterial loads in multiple organs. In addition, these models share some important aspects with human sepsis including high serum levels of pro-inflammatory cytokines, bacterial dissemination, and high mortality. However, an obvious difference between these sepsis models and the human disease process is the source of infection. Intraperitoneal instillation of the inoculum produces sepsis that is secondary to intraperitoneal infection, which does not represent a natural route of infection in A. baumannii disease. This approach rapidly produces fulminant sepsis in mice and rats, whereas in human disease, the development of A. baumannii sepsis is typically more progressive. This aspect may complicate studies aiming to characterize the host response during the early phases of A. baumannii sepsis. It is worth noting that a number of the models described above for the study of pneumonia, burn, and soft tissue infections have been shown to result in disseminated infection (Rodríguez-Hernández et al., 2001; Montero et al., 2004; Pantopoulou et al., 2007; Shankar et al., 2007; Dai et al., 2009; Uygur et al., 2009). These models may therefore better mimic clinical scenarios in which sepsis results from dissemination of the infection from a discrete infectious focus.
Other mammalian models of A. baumannii infection
Pachón-Ibañez et al. (2010) have recently reported the development of a rabbit model of A. baumannii meningitis. In this model, anesthetized rabbits are immobilized in a stereotactic frame, and the inoculum is instilled into the intracisternal space. The authors report that features of meningitis in the cerebrospinal fluid developed by 12 h postinoculation. Bacterial loads, lactate dehydrogenase concentrations, and white blood cell counts were quantified in cerebrospinal fluid after obtaining the sample by cisternal puncture. In addition, brain edema was quantified by determining brain weight after desiccation. A major advantage of this model is that the large size of rabbits allows for repeated sampling of the cerebrospinal fluid for monitoring the evolution of meningitis in a single animal.
A rabbit endocarditis model has been described for A. baumannii that was developed based on a model initially used for characterizing staphylococcal endocarditis (Perlman & Freedman, 1971; Rodríguez-Hernández et al., 2001, 2004). In this model, an intracardiac catheter is inserted into the left ventricle of New Zealand rabbits 4 days prior to inoculation. For the establishment of infection, mice are inoculated via injection of a suspension containing A. baumannii into an ear vein. The authors report that 100% of control rabbits were bacteremic 24 h postinoculation and that A. baumannii was isolated from the valves of all infected animals. Valve bacterial loads and valve sterility rates, in addition to quantitative blood cultures, were used to evaluate the efficacy of antibiotic regimens.
There have been two reports describing the development of animal models of A. baumannii osteomyelitis. In a mouse model, osteomyelitis can be produced by surgically exposing the tibia and inserting a 0.25-mm stainless steel pin that has been coated with a suspension of A. baumannii into the bone metaphysis (Crane et al., 2009). This approach produces a chronic, but nonlethal osteomyelitis. The authors used histology to show that A. baumannii resulted in osteoblastic bone formation at the site of infection, in contrast to Staphylococcus aureus that produces an osteolytic response. An attempt to produce osteomyelitis in rats has been made by inoculating induced segmental bone defects with A. baumannii (Collinet-Adler et al., 2010). The authors report that in this model, the bacteria were cleared without producing a clinically relevant osteomyelitis, although bony lysis and new bone formation were detected in some animals.
Nonmammalian models of A. baumannii infection
Mammalian models of A. baumannii infection allow characterization of a host response to infection similar to that found in human infections. In addition, because drug pharmacokinetics and pharmacodynamics can be determined in mammalian species using techniques similar to those used in humans, these models also permit the study of antibiotic regimens for the treatment of A. baumannii infection. For these reasons, models based on mammalian species clearly have advantages for study designs that aim to mimic the human disease process. However, recently developed nonmammalian models of infection offer the advantage of lower cost for both purchasing and maintaining the animals, and avoidance of ethical issues that are associated with the use of mammalian species (Table 4). In addition, nonmammalian models, because of the ease of infecting large groups, may be more suitable for high-throughput screening of bacterial mutants.
Table 4. Overview of Acinetobacter baumannii infection models
Mouse, rat, guinea pig
Drug pharmacokinetics and pharmacodynamics, tissue bacterial loads, survival, cytokine levels, histology
Optimization of antibiotic therapy, characterization of the host response, identification of bacterial virulence factors
Burn and soft tissue infection
Bacterial load, survival, abscess formation
Optimization of antibiotic therapy, identification of virulence factors
Identification of virulence factors, interactions with eukaryotic cells
Galleria mellonella is the larval stage of the greater wax moth, and it has recently been developed for use as model for infection by A. baumannii. (Peleg et al., 2009). In this model, caterpillars are inoculated with a suspension containing A. baumannii via injection into the caterpillar hemocoel. Postinfection mortality was shown to depend upon the number of bacteria present in the inoculum as well as the active expression of the acinetobactin-mediated iron acquisition system (Gaddy et al., 2012). In addition, infection of the caterpillars with A. baylyi and Acinetobacter lwoffii, which are thought to be less pathogenic in humans, produced less mortality than infection with an inoculum containing an equal amount of A. baumannii. Together, these data may suggest that this model can be used to characterize the virulence of different strains. Treatment of infected caterpillars with antibiotics that had activity against the A. baumannii strain in vitro increased survival, indicating that this model may also be useful for evaluating antibiotic efficacy.
Caenorhabditis elegans is an unsegmented nematode that feeds on bacteria. Two reports have used C. elegans to characterize A. baumannii. Smith et al. (2007) used an assay in which worms are placed onto lawns of A. baumannii, and the proliferation and brood sizes of the worms are measured. With this approach, these authors identified 114 A. baumannii mutants that supported growth of the worms on ethanol-containing media, which has been reported to increase bacterial virulence (Smith et al., 2007). A subsequent study demonstrated that A. baumannii reduced the ability of C. albicans to kill C. elegans during co-infection, demonstrating that this model can be used to study eukaryote–prokaryote interactions (Peleg et al., 2008a).
Dictyostelium discoideum is an ameba that is found in soil environments and has recently been used to measure the virulence of A. baumannii mutants (Smith et al., 2007). In this model, D. discoideum is co-cultured with A. baumannii and, because of the slower growth rate of the amebae, A. baumannii produces a lawn in which plaques are formed because of the consumption of the bacteria by D. discoideum. Bacteria with increased virulence, or bacteria grown under conditions that increase the expression of virulence traits, results in a reduction in the number of plaques formed by D. discoideum. Conversely if the virulence of a strain is reduced, for example by the introduction of a mutation, an increase in the number of plaques formed by the ameba is seen. This model was successfully used to identify A. baumannii transposon mutants with reduced virulence (Smith et al., 2007).
With the increasing clinical importance of A. baumannii because of a rise in the number of colonizations/infections caused by this organism and the emergence of MDR strains, continued study of this pathogen is required. The use of animal models has already yielded important data regarding the optimization of antibiotic regimens for the treatment of resistant strains and has begun to identify virulence factors that have contributed to the success of A. baumannii as a pathogen. With the sequencing of multiple A. baumannii genomes and the increasing availability of transgenic animals, it is hoped that these models can be used to continue to identify bacterial and host factors involved in infection at the molecular and cellular levels. These studies will provide insight into the basic biology of this important human pathogen and aid in the identification of targets for the development of novel treatment and prevention strategies.
The authors thank Pilar Pérez-Romero for critical reading of the manuscript and Miguel Gallardo for technical assistance. This work was funded by the Ministerio de Economía y Competitividad, Instituto de Salud Carlos III – co-financed by European's Development Regional Fund ‘A way to achieve Europe’ ERDF, Spanish Network for the Research in Infectious Diseases (REIPI RD06/0008/0000). M.J.M. is supported by the Subprograma Miguel Servet from the Ministerio de Economía y Competitividad of Spain. U.S. Public Health AI070174 and NSF 0420479 grants (awarded to L.A.A.) and Miami University funds supported the work reported in this review.