A. baumannii is an emerging MDR pathogen frequently isolated from clinical settings. The healthcare and economic impact of A. baumannii infections, particularly in intensive care units, highlight the urgent need to develop new approaches to treat and prevent such infections. Here, we have defined a novel group of A. baumannii candidate vaccine antigens based on their predicted subcellular location, prevalence, sequence conservation and predicted solubility.
The resistance of A. baumannii
to multiple antibiotics occurs through intrinsic mechanisms, and through the gain of laterally acquired resistance genes [16
]. Given its propensity to rapidly and efficiently develop resistance, vaccination represents a viable alternative strategy to prevent infections caused by A. baumannii
. In fact, several conventional vaccinology approaches have been used for the identification of potential vaccine targets. Whole-cell [17
], OMV [18
] and outer membrane preparations [19
] have been shown to confer active and passive protection in a murine model of disseminated sepsis against different strains. These approaches have the advantage of providing responses to several surface-exposed epitopes simultaneously, however endotoxin contamination remains a limiting factor for their development and translation to human use. In contrast, subunit preparations offer a viable alternative for vaccine development. Such vaccines stimulate the production of opsonizing and/or functional antibodies, which target specific components present on the bacterial cell surface or secreted virulence factors that interact with host epithelial cells and cause disease. When soluble, subunit antigens are relatively easy to obtain on a large scale and production processes are highly reproducible, assisting approval from regulatory agencies.
Reverse vaccinology takes advantage of the many genome sequence datasets available in the public domain and represents a targeted approach for the discovery of novel surface antigens. The use of a sequenced based approach ensures the evaluation of all putative proteins encoded within the genome of a given strain, however is limited by the availability of effective search tools to predict protein subcellular location. The accuracy of genome annotation and the quality of training datasets therefore impact the analysis outcome. Conversely, proteomic approaches allow the identification of bacterial surface antigens, but may be limited by sensitivity (i.e. for detection of poorly expressed proteins) or lack of expression of certain antigens under some growth conditions. Thus, methods for bacterial growth, cell fractionation and protein preparation may lead to variable or even inconsistent results.
In this study, we adopted an approach that combines these methods and provides an extra filtering step in our definition of candidate vaccine antigens for MDR A. baumannii
, and we predict this will be more informative than data obtained from each method employed as a standalone technique for vaccine antigen selection. Ten complete and thirty-one draft A. baumannii
genomes were used for reverse vaccinology analysis, leading to the identification of 234 putative outer membrane or secreted proteins based on our strict criteria to define protein subcellular localization. The strains employed covered a wide spectrum of geographic locations and disease states, and contained representative strains from the dominant A. baumannii
clones currently circulating the globe. The proteomic approach involved the analysis of three A. baumannii
clinical isolates, and resulted in the identification of 122 secreted or OMV-associated proteins. Overall, sixty-two proteins were identified from both approaches. This list was further refined by the removal of twenty proteins predicted to contain a beta-barrel structure (Table S4
). Such proteins are generally buried within the outer membrane, and interact with the immune system via external loops presented as conformational epitopes. Many of these eliminated proteins were also predicted to be involved in phenotypes associated with a high level of redundancy, such as iron acquisition and transport functions. For example, BauA (ABAYE1093), is a siderophore receptor involved in the transport of acinetobactin [20
] and intracellular survival within epithelial cells [21
Our final refined list of A. baumannii
proteins contained 42 putative antigens. This included OmpA (ABAYE0640) and Bap (ABAYE0792), both of which have been examined previously as vaccine candidates against A. baumannii
]. OmpA is an outer membrane and secreted lipoprotein that contributes to biofilm formation, interaction with epithelial cells [22
] and induction of apoptosis by epithelial cells [23
]. Immunization of diabetic mice with OmpA in combination with aluminum hydroxide adjuvant has been shown to induce high anti-OmpA antibody titers and resulted in reduced tissue bacterial burden and improved survival of mice following intravenous infection with A. baumannii
]. Despite this promising level of protection, others have shown OmpA is soluble and active when recovered from the supernatant of bacterial cultures, but insoluble when expressed as a recombinant protein [24
]. Vaccination of mice with Bap, a surface-exposed adhesin involved in biofilm formation, has also been performed. Immunization with Bap conferred protection in a murine model of sepsis and led to a significant increase in survival, as well as a reduction in bacterial counts in the liver and spleen of infected mice [8
Based on sequence and structural analysis, the 42 candidate antigens could be divided into five groups. The major group comprised outer membrane lipoproteins (n=18). Outer membrane lipoproteins have previously been identified as major vaccine targets against other bacterial pathogens [11
]. For example, factor H-binding lipoprotein elicits antibodies against N. meningitidis
serogroup B [25
], and constitutes an important component of Bexsero, the first vaccine developed using reverse vaccinology [26
]. Ten proteins predicted to be associated with adhesion, including five putative fimbrial proteins and Bap, were also identified. Fimbrial proteins have previously been shown to constitute effective vaccines for some pathogens (e.g. Bordetella pertussis
] and uropathogenic E. coli
in animal infection models [28
]). Furthermore, our detection of these proteins using proteomics is consistent with their role in biofilm formation [30
]. Nine enzymes/toxins were identified in our final list, and given that most MDR A. baumannii
strains are hemolytic [32
], these proteins may also represent important targets for vaccine development. Enzymes/toxins play an important role in the scavenging of nutrients by bacterial pathogens, and many are toxic to the human host. Functional antibodies that block the activity of toxins generally reduce the severity of infection, and several highly effective, licensed vaccines use inactivated toxins (toxoids) to raise protective antibodies (e.g. anthrax, diphtheria, pertussis and tetanus toxins). Two solenoid repeat proteins were identified in our analysis. These proteins often interact with other proteins and include Tetratrico Peptide Repeat proteins, Pentatrico Peptide Repeat proteins and Sel1-like repeat proteins. Characterized examples from this family include the Helicobacter
cysteine-rich protein (Hcp), and the newly described c5321 protective antigen against extraintestinal pathogenic E. coli
]. Finally, three hypothetical proteins with no sequence or structural homology to any characterized proteins in the NCBI non-redundant protein database were identified.
The prevalence and variability of each of the 42 antigens was also assessed using the ten complete genome sequenced strains as well as an additional 33 strains for which a draft genome sequence was available. The antigens could be classified into three groups based on prevalence and amino acid sequence conservation, with antigens in Group I (n=41) representing the most likely candidates for vaccine antigens. However, antigens in Group II and III include many of the potential adhesins, and thus we cannot rule out their use in a potential multi-component subunit vaccine. We note also that some proteins that offer potential as vaccine candidates may have been missed in our analysis. Such an example is Ata, an autotransporter protein that plays an important role in biofilm formation and binding to extracellular matrix and basal components [33
]. Vaccination with Ata has been shown to attenuate infection in a pneumonia murine passive model using immunocompetent and immunocompromised mice [7
]. Ata (A1S_1032) is present in the ST92 strains and was identified by reverse vaccinology (Table S2
), but we were unable to detect its expression in vitro
by HPLC-MS/MS analysis. Another example is PKF (ABAYE0936), a secreted serine protease that confers resistance to complement mediated killing and biofilm formation [35
]. Predicted to be a periplasmic protein by PSORTb, PKF was initially discarded by the genomic approaches but identified by the proteomic analysis (Table S5
). Even though our combined approach could miss some potential vaccine antigens, the standalone approaches are complementary and thus provide a backup list of potential targets for future consideration. We note also that a limitation of our approach was the inability to infer the correct orientation of hypothetical outer membrane proteins with respect to the extracellular or periplasmic space by sequence and structural analysis alone. Thus, further analysis of these proteins is required to confirm their suitability as vaccine candidate antigens.
Murine sepsis models using subcutaneous or intramuscular immunization and either intravenous [6
] or intraperitoneal [8
] challenge have previously been used to test the efficacy of potential vaccine candidates against A. baumannii
. The 42 vaccine antigens identified in this study could feasibly be tested in a murine sepsis model to examine their ability to protect against bloodstream infections, one of the most important clinical manifestations of MDR A. baumannii
]. A recent international surveillance study reported the association of A. baumannii
with 8.8% of ICU infections (ranging from 3.7% to 19.2% according to geographical region) [36
]. A. baumannii
has also emerged as an important cause of infections resulting from injuries sustained by military personnel during recent operations in Iraq and Afghanistan [37
]. Therefore, a broadly protective vaccine against A. baumannii
may have a major impact on some high risk groups, including ICU patients, injured military personal, patients undergoing elective surgery, diabetics and hemodialysis patients.
In conclusion, this study provides the first comprehensive analysis of A. baumannii strains for vaccine purposes, and has identified potential antigens that could form a framework for the design of a novel and broadly protective vaccine targeted against MDR A. baumannii. Given the rapid emergence and dissemination of MDR A. baumannii strains in healthcare settings across the globe, such a vaccine would address an urgent and currently unmet need. Our combined use of reverse vaccinology and proteomics provides an excellent example of the high throughput power of these complementary strategies for the identification of potential vaccine targets when an appropriate collection of genome sequences from epidemiologically relevant strains is available. Future work will now be targeted towards the characterization of the proteins identified in our analyses, as well as their evaluation in animal infection models as vaccine antigens.