Pasteurella multocida is a Gram-negative bacterial pathogen which is the causative agent of a range of diseases in animals, including fowl cholera in avian species, hemorrhagic septicemia in ungulates, shipping fever and pneumonia in cattle, atrophic rhinitis in swine, and snuffles in rabbits 123. The bacterium also causes infection in humans, primarily through dog and cat bites. Fowl cholera, which is generally caused by serotypes A:1, A:3 or A:4 4, is a severe systemic disease which occurs in domestic poultry and wild birds and results in significant economic losses to poultry industries worldwide. Current vaccines against fowl cholera include bacterins 5, which provide only limited protection against homologous serotypes and live attenuated strains, which have been observed to revert to virulence 6. Therefore, there is a need for more effective vaccines to control diseases caused by P. multocida.

The surface of Gram-negative bacteria is critical for interaction of the bacterium with the host cell environment as it mediates nutrient uptake, secretion of toxins and other products and is involved in avoidance of the host immune system 7. Furthermore, it is the bacterial surface molecules that are the targets for host immunity. Indeed, bacterial surface proteins have been shown to be important for conferring protective immunity in a range of infection models 89. Recently, the P. multocida PlpB protein was identified as a cross-protective antigen 1011 and this protein is located in the P. multocida outer membrane 12. Outer membrane proteins also promote adherence to host cell surfaces and are therefore likely to be involved in P. multocida virulence 13.

In an effort to identify novel immunogens to which chickens respond during natural infection, and as a first step towards developing protective recombinant vaccines against fowl cholera, a detailed antigen profiling analysis was undertaken. The effectiveness of this approach, which has been termed 'reverse vaccinology' or 'reverse immunology', has been reported by several authors with Plasmodium falciparum 14, Streptococcus pneumoniae 15, Treponema pallidum 16, Neisseria meningitidis 17, and Chlamydia pneumoniae 18.

We utilised a range of bioinformatics analyses of the annotated P. multocida Pm70 genome sequence and previously published experimental data 12 to select genes likely to have vaccine potential. The central premise of this work was that protective antigens are likely to be surface exposed or secreted by the bacteria and therefore accessible to the host immune response. We used PSORTB 19 and ProteomeAnalyst 20 to predict all outer membrane and secreted proteins and LipoP 21 to predict all lipoproteins. We included all lipoproteins as a large number of lipoproteins was observed in the proteomics analysis of the P. multocida outer membrane 12 and we believe that precise sub-cellular prediction of lipoprotein location (inner or outer membrane) is presently unreliable. These bioinformatics analyses identified a combined list of 129 proteins (Figure 1, see Additional file 1). The prediction programs PSORTB and ProteomeAnalyst can be used to predict protein localisation in Gram-negative organisms, while LipoP specifically predicts Gram-negative lipoproteins. PSORTB and ProteomeAnalyst use different methods for prediction of localisations. PSORTB uses primary sequence analysis algorithms and gives a high precision, but low sensitivity, prediction. Proteome analyst on the other hand utilises analysis of text annotations for the closest homologue of any given query sequence and is medium-high precision and sensitivity; it thus makes a higher number of predictions. These analyses were complemented by the previously published proteomic analysis of the P. multocida outer membrane 12 which identified three proteins (PM0612, PM1357 and PM1746) not predicted by bioinformatics.

The large number of candidate genes necessitated the adoption of a high-throughput cloning strategy, which utilised by the Gateway™trogen Inc., Carlsbad, CA) cloning and expression system to clone PCR-amplified P. multocida ORFs. An attempt was made to clone all 129 ORFs; however, only 105 were successfully cloned and expressed in E. coli. Each of the recombinant expression clones was assessed for levels of protein expression and solubility. Because of the large inventory of proteins that were required to be expressed and purified in a time and cost effective way, protein expression was analysed using a novel inclusion body assay developed on a TECAN liquid handling robot.

To evaluate the range of P. multocida proteins recognised by the chicken antibody response, serum samples from chickens that had been repeatedly infected with P. multocida and then treated with antibiotics were tested for reactivity against 105 recombinant proteins by Western blot. Pooled serum samples from P. multocida infected chickens (8 from X-73, 3 from VP161 – a highly virulent A:1 strain 22) and four serum samples from control chickens were reacted in immunoblot assays with comparable amounts (3 μg/lane) of recombinant proteins. The immune chicken sera generated during P. multocida strain X-73 or VP161 infection recognised a total of 12 (Table 1) recombinant proteins, ten of which were not recognised by naive sera. To our knowledge, six of these antigens have not been previously identified as capable of eliciting immune responses from chickens, demonstrating the utility of this approach for the identification of novel immunoreactive bacterial antigens (Figure 2). Although immunological reactivity to unique antigens alone does not necessarily indicate a role in protective immunity against pasteurellosis, these novel antigens constitute prime candidates as potential vaccine targets for pasteurellosis, and warrant inclusion in studies to elucidate their potential role in pathogensis. The results have important implications in understanding avian immune responses to P. multocida and in elucidating putative elements of a protective immune response.

One drawback of this immunogen discovery assay is that it is limited to protein coding regions that can be expressed in E. coli. In addition, antibodies that recognise only conformational epitopes that are denatured in SDS-PAGE will not be detected. Furthermore, a lack of antibody recognition to recombinant proteins may be the result of a poor immune response to these antigens during infection or alternatively, a reflection of differential expression of antigens by the VP161 and X-73 strains within the host. Four of the six novel proteins (PM0442, PM0659, PM0979, PM1614) were identified as putative lipoproteins characterised by a cysteine residue after a signal peptide. PM1993 was predicted to contain a signal peptide processed by signal peptidase 1. These results do not preclude the existence of additional immunogens present in other serotypes.

In summary, this work represents a systematic and comprehensive approach for the characterisation of the immune response associated with pasteurellosis by targeted computational screening of genomic sequences with predictive antigenic properties, followed by an automated high throughput expression and purification strategy using a robotic platform. It has resulted in the identification of 12 immunogenic determinants which are clearly expressed in vivo during infection, six of which were unique to this study, and thus constitute candidates for the development of a protective vaccine against fowl cholera. We are currently evaluating the protective effects of these P. multocida proteins against bacterial challenge in chickens.