Pasteurella multocida is a facultative anaerobic Gram-negative bacterium responsible for economically significant infections in a broad range of animal species [1]. P. multocida commonly presents as a commensal bacterium in the respiratory tract but can become a primary or secondary pathogen depending on health status, animal species, climatic conditions, and poor husbandry practices [1, 2]. The organism causes a variety of respiratory diseases, including pneumonic pasteurellosis of ruminants and pigs, porcine progressive atrophic rhinitis (PAR), avian cholera, bovine haemorrhagic septicaemia (HS), snuffles in lagomorphs, and human infections via bites or scratches from carnivores [2]. Pneumonic and septicaemic pasteurellosis remain important problems in pigs, cattle, and water buffaloes worldwide. The disease severity varies from acute to subacute to chronic in different animal hosts [3] and depending on the bacterial serotype [4]. Generally, this organism can be classified into five capsular serotypes (A, B, D, E, and F) and 16 lipopolysaccharide (LPS) serotypes (1–16) [3]. Associations of certain serotypes of P. multocida with specific disease symptoms or geographical areas have been systematically shown by Davies et al. [5–9]. How these serotypic preferences and host predilections contribute to the virulence of P. multocida is not fully understood.

Recently, several comparative genomic studies have revealed different genes and genetic elements between virulent strains of P. multocida and avirulent strains [10–12]. The pathogenicity and virulence of P. multocida involve multiple genes related to the bacterial capsule, endotoxins (LPS), dermonecrotic exotoxin or P. multocida toxin, phages, plasmids, type IV pili, adhesins, extracellular enzymes, and outer membrane proteins (OMPs) [2, 13, 14]. Several genes encoding these features of P. multocida, including exbB, exbD, fimA, fur, hgbA, hgbB, hsf1, hsf2, nanB, nanH, oma87, ompA, ompH, pfhA, plpB, plpE, pmHAS, psl, ptfA, sodA, sodC, tadD, tbpA, tonB, and toxA, have been used in molecular typing studies to define virulence-associated gene profiles. The products of the exbB, exbD, fur, and tonB genes are needed for energy transfer from the inner membrane to the outer membrane iron transporters [15]. The hgbAB and tbpA genes encode outer membrane haemoglobin and transferrin transporters that are tightly regulated by the product of the fur gene [16]. The fimA, hsf1, hsf2, pfhA, ptfA, and tadD genes encode adhesion and colonisation factors that allow P. multocida to colonise host mucosal membranes [2, 17–19]. The products of the nanB, nanH, ompH, plpB, and psl genes are OMPs involved in nutrient acquisition and transport. The sialidase proteins NanB and NanH cleave sialic acids from the host membrane for use as nutrients and in modifying the cell envelope to evade the host immune system [2, 20]. OmpH, a porin, is a general transporter that facilitates the diffusion of various molecules [21]. The superoxide dismutases sodA and sodC are important in the oxidative stress response and in protection against oxidative stress [22]. The ompA and oma87 genes are required for the biosynthesis and integrity of the outer membrane, the outermost surface of the bacterium that interfaces with the external environment [23]. The highly abundant OmpA protein possesses a C-terminal globular domain that interacts with the peptidoglycan layer of the cell wall, providing stability and integrity to the outer membrane. The OmpA protein is also involved in adherence to host cells through heparin and fibronectin binding [24]. The outer membrane lipoproteins PlpB, PlpE and Psl (P6-like) may stimulate immune protection against the bacterium in animal hosts, but their precise functions remain unclear [25–27]. The toxA gene encodes a dermonecrotic toxin (DMT) that blocks chemotaxis-induced migration of dendritic cells to lymph nodes and restricts the progress of the adaptive immune response [2]. Some of these virulence-associated genes appear only in virulent strains of the bacterium and are highly correlated with disease severity. Others may also be present in non-pathogenic strains but are highly expressed only in virulent strains in vivo, as shown by Li et al. [28].

The existing evidence suggests that the virulence-associated genes described above may be required for pathogenic strains of P. multocida to survive under host conditions. Several studies have screened these genes in a range of P. multocida isolates in attempts to identify their involvement in virulence and host specificity [29–33]. These studies revealed the prevalence of the toxA gene in P. multocida strains (mainly capsular type D) associated with porcine PAR and frequent observation of the tbpA gene in the strains that affect cattle. Many of these genes, particularly the OMP-encoding genes, have been widely studied, and some have been used as the basis for the development of vaccines that can heterologously protect against infections by multiple strains of P. multocida [13, 14, 34–39]. This evidence has led to the hypothesis that molecular variations within the same virulence-associated genes may affect host specificity and virulence. Verma et al. found seven classes of ompA alleles in 46 isolates of P. multocida [40]. Associations of these ompA alleles with capsular type and disease status have been proposed. Another recent study identified associations between ompH alleles and the capsular types of 83 isolates of P. multocida and found variable ompH alleles within capsular type A isolates [41]. This information is important for detailed typing of virulence isolates and will be helpful in the monitoring of P. multocida infections. However, the prevalence of specific virulence-associated genotypes and OmpA types in pigs, cattle, and water buffaloes in Thailand has not been determined. Therefore, this study aimed to characterise the virulence-associated gene profiles of 191 bovine and porcine P. multocida isolates collected in Thailand over a 24-year period and to develop a protein-sequence-based OmpA typing scheme using molecular genetics and bioinformatics techniques. The results of this study provide an alternative molecular typing of P. multocida isolates and serve as a basis for the development of OmpA type-specific vaccines.

Clustering of 25 virulence gene profiles of 191 bovine and porcine isolates of P. multocida revealed two major clusters: bovine and porcine (Fig. 1). Ninety-nine percent of the isolates had the profiles of 17 virulence genes in common (the presence of exbB, exbD, fimA, fur, hgbA, hsf2, nanB, oma87, ompA, ompH, plpB, psl, ptfA, sodA, sodC, and tonB and the nearly complete absence of toxA). The dendrogram in Fig. 1 shows that the profiles of eight virulence genes (hgbB, hsf1, tadD, nanH, pfhA, plpE, pmHAS, and tbpA) could be used to separate the isolates into two major clusters. The bovine cluster (shown in dark blue in Fig. 1) included five genes (hsf1, hgbB, nanH, pfhA, and toxA) that were shared across the seven subgroups. The profiles of the eight virulence genes varied across the four subgroups of the porcine cluster (shown in green in Fig. 1). The profiles of the Thai bovine isolates of P. multocida were less diverse than those of the porcine isolates.

In this study, the bovine isolates (C cluster) were divided into six subgroups — C1, C2, C3, C4, C5, and C6 — based on the profiles of the genes hgbB, hsf1, tadD, and pmHAS. The third subgroup, C3, accounted for 76% of the bovine isolates. There was no clear separation between the isolates obtained from cattle and those obtained from water buffaloes or between samples collected in different years. The subgroups C1, C2, C3, C6, and C0 displayed similar virulence gene profiles lacking pmHAS, plpE, and tadD (Fig. 1). The subgroups C4 and C5 additionally displayed the presence of the pmHAS and tadD genes, respectively. Unlike the porcine cluster, almost all of the bovine isolates possessed the hsf1, hgbB, pfhA, and tbpA genes; thus, these genes could be key to the success of these strains in causing acute HS.

The second cluster of porcine isolates was divided into four subgroups: P1, P2, P3, and P4 (Fig. 1). Subgroup P4 (40%), followed by subgroup P1 (37%), included the highest number of porcine isolates. Most isolates of subgroup P1 had the nanH, pfhA, plpE, pmHAS, and tadD genes in common and showed complete loss of tbpA. The pfhA gene was only present in subgroup P1, whereas hsf1 and hgbB were present in some isolates of this subgroup. All members of the P1 subgroup belonged to capsular type A, including two bovine type A isolates: 1C and 12C. Similarly, nearly all members of subgroup P4, with the exception of the 104P strain, also belonged to capsular type A. These strains had the nanH, pmHAS, hgbB, and tadD genes in common. Two other subgroups of the porcine cluster, subgroups P2 and P3, were isolates of capsular type D. All isolates of subgroup P2 showed the presence of hsf1, nanH, hgbB, and tadD genes together with the lack of pfhA, pmHAS, and tbpA genes. Some isolates in this subgroup had the plpE gene. The third subgroup, P3, only shared the tadD, hgbB, and nanH genes; the other five genes were absent. These results demonstrate the presence of variable virulence gene-associated profiles in porcine isolates of P. multocida in Thailand.

The ompA gene was successfully amplified and sequenced from 186 isolates of P. multocida in this study. The obtained nucleotide sequences were trimmed to a length of 940 nucleotides and were confirmed to correspond to the ompA gene. The obtained sequences covered nearly the full length of the transmembrane domain of this gene. Forty-one available ompA nucleotide sequences were also downloaded from the NCBI nucleotide database and compared with the sequences identified in this study. After multiple sequence alignment and phylogenetic construction of 227 DNA sequences, the phylogenetic tree of the ompA gene of P. multocida (Fig. 2a and Additional file 1: Table S1) clearly showed two major clusters of the bovine (cluster C) and porcine (cluster P) isolates, consistent with the clustering of the virulence-associated genes shown in Fig. 1. The bovine cluster consisted of a single large group of 97 bovine isolates (79 Thai isolates and 18 database strains) of capsular type B. The second porcine cluster consisted of five subgroups: P1, P2, P3, P4, and P5. Sixty-five of the porcine isolates were almost equally clustered in subgroups P1 and P2. Subgroup P3 members (13 strains) included only the samples in the NCBI database. Thirty isolates of subgroup P4 were further divided into subgroups P4I (9 isolates) and P4II (21 isolates). With the exception of a single capsular type D isolate, 87P (subgroup P2), the isolates in subgroups P1, P2, P4I, and P5 were capsular type A. All isolates in the P4II subgroup were capsular type D. These results demonstrate the presence of molecular differences in the ompA genes of the two capsular types of porcine P. multocida isolates. Interestingly, two bovine isolates of capsular type A (1C and 12C) also grouped within the porcine cluster (subgroups P5 and P4I).

Analysis of nucleotide substitutions within the ompA gene was conducted on 95 unique ompA sequences, returning a dN of 1.3667, a dS of 1.0341, and a dN/dS ratio of 0.8751. The dN and dS of each codon position in 227 nucleotide sequences of the ompA genes analysed in this study were compared (Fig. 3). The results showed that selection pressure had operated unequally on different regions of the ompA genes of P. multocida. Codon positions 61–220 accumulated more nonsynonymous than synonymous changes, as did positions 409–457. Codon positions 247–325 displayed more synonymous changes, as did positions 460–508 and 781–820, whereas positions 544–780 showed a counterbalance between synonymous and nonsynonymous substitutions.

The OmpA proteins encoded by the ompA gene sequences in this study contained 314 amino acids. Compared to the amino acid residues in the full-length OmpA proteins in the Uniprot database, 39 amino acid residues were not included in this study. Twenty-two amino acids were missing from the N-terminus, whereas the C-terminus was shortened by 16 amino acids. The sequences of the transmembrane domain were obtained, and the equally trimmed 186 sequences of this domain were aligned and summarised in a logo plot (Fig. 4a), which showed three variable regions at amino acid positions 24–42, 80–93, and 132–147. These regions correlated well with the dN/dS results for regions 60–159, 248–296, and 415–455, respectively (Fig. 3).

Approximately 252 amino acids within the transmembrane domain were compared (Fig. 4a); the resulting protein phylogenetic tree is shown in (Fig. 2b and Additional file 1: Table S1). The tree shows two major clusters of OmpA proteins from the Thai bovine and porcine isolates. The first cluster consists of two subgroups (P1 and P2) of the porcine isolates associated with capsular type A. The second cluster contains a very large subgroup C (C1 and C2) of the bovine isolates of capsular type B and a small subgroup of the porcine (P3 and P4) isolates. The members of subgroup P3 are all porcine isolates associated with capsular type D, whereas the members of subgroup P4 are capsular type A isolates, including the bovine isolate 12C. Although this protein tree differs slightly from the nucleotide tree derived for the ompA gene, the distinctions between animal hosts and capsular types are clearly shown.

The multiple aligned OmpA sequences were subjected to structural prediction. The consensus OmpA structure of P. multocida strains in Thailand was predicted based on the template structures of the OmpA-like domain from Acinetobacter baumannii (4g4x_A) and the OmpA transmembrane domain from Escherichia coli (1qjp) (Fig. 4b). The predicted OmpA structure of the Thai isolates of P. multocida contained two domains: a globular C-terminal peptidoglycan-binding domain and a transmembrane ß-barrel N-terminal domain. The transmembrane domain consisted of eight antiparallel ß-sheets interspersed with four short periplasmic turns and four long extracellular loops, similar to the OmpA proteins of other bacteria. Six OmpA structures (21P, 32P, 42P, 91P, 34C, and 84C) were predicted as representative of six major OmpA protein clusters (P1, P2, P3, P4, C1, and C2) from the tree shown in Fig. 2b. Superposition of these structures revealed eight conserved transmembrane ß-sheets and four hypervariable loop regions (Fig. 5a and b). The fourth loop was more structurally conserved than the other three loops (Fig. 5c). Individual models of these six OmpA proteins are presented in Fig. 6. These representative structures were classified into two clusters, similar to the protein clustering shown in Fig. 2b. By computing the molecular surface, the extracellular loops of these six proteins were determined to be rich in charged amino acid residues (middle and bottom rows in Fig. 6). Taken together, the data show that the observed variations in the nucleotide sequences of the ompA genes contribute to the structures of the hypervariable extracellular loop regions of the OmpA proteins.

The amino acid sequences corresponding to the four hypervariable extracellular loops (L1, L2, L3, and L4) were extracted from the alignment and trimmed to the same length. The first loop (L1) was closest to the N-terminus, whereas the last loop (L4) was close to the C-terminal domain. A difference in at least one amino acid position within the same loop was considered to represent a distinct loop sequence type (LST). This study developed the OmpA protein sequence-based typing scheme using the observed LSTs. Each LST consisted of four loop types: loop 1 sequence type (L1ST), loop 2 sequence type (L2ST), loop 3 sequence type (L3ST), and loop 4 sequence type (L4ST), which was formularised as LST = L1ST: L2ST: L3ST: L4ST. The use of LST molecular typing on these isolates revealed 30 LSTs comprising 21 L1STs, 11 L2STs, 14 L3STs, and 10 L4STs. Fourteen LSTs were found in 186 Thai bovine and porcine isolates; the other 16 LSTs were present only in the database strains. Five LSTs (LST8, LST12, LST15, LST18, and LST19) included 80% of the Thai isolates. The bovine OmpA proteins in subgroup C1 (capsular type B) harboured LST19 (11:5:5:3), whereas subgroup C2 (capsular type B) belonged to the LST8 group (6:5:5:3). The porcine subgroups P1, P2, and P3 had LST12 (8:3:3:3), LST15 (9:7:3:3), and LST18 (10:2:2:3), respectively. Clearly, the porcine strains of capsular types A (subgroups P1 and P2) and D (subgroup P3) could be differentiated based on their LSTs.

To understand the importance of these variations, the four hypervariable loop sequences of the six subgroups (C1, C2, P1, P2, P3, and P4) were subjected to antigenicity prediction. The prediction algorithm was based on the physicochemical properties of experimentally determined antigens. No loops of the porcine isolates of subgroups P1, P2, and P3 displayed antigenic epitopes. Subgroup P4, which contained the bovine isolate 12C and other porcine isolates, displayed a possibly antigenic first loop. Intriguingly, the bovine subgroups C1 and C2 each had three possibly antigenic loops (L1, L2, and L3). This antigenic variation could have resulted from high levels of nonsynonymous substitution within the loop regions.

Analysis of 25 virulence-associated gene profiles identified 16 genes that may have important functions in all 191 isolates of P. multocida. The absence of these genes is expected to greatly affect the survival and likely the virulence of P. multocida. Some of these genes may also appear in non-pathogenic strains but may be highly expressed in virulent strains under in vivo conditions, as shown by Li et al. [28]. However, the near absence of the toxA gene in this study (it appeared in only one porcine isolate) could reflect successful eradication of toxA-positive porcine isolates in Thailand due to the heavy use of antibiotics and vaccines in past years. None of the bovine isolates possessed the toxA gene. This finding is consistent with the results of a similar study of Indian and South Asian cattle strains associated with HS, in which the infectious bacteria did not carry the toxA gene but harboured the tbpA gene [42]. Clustering of the 191 isolates analysed, which was based mainly on the profiles of eight genes, showed that the profiles of the Thai bovine isolates of P. multocida were less diverse than those of the porcine isolates. The results were supported by pulsed-field gel electrophoresis (PFGE) and multilocus sequence typing (MLST) of eight field isolates and one vaccine strain of P. multocida associated with HS cases from Thailand [43]. All isolates shared the same MLST sequence type 122 and the same PFGE pattern, whereas the method used in the current study was able to distinguish additional minor groups of the bovine isolates.

The prevalence rates of tbpA, pfhA, and hgbB genes in the bovine isolates are also consistent with the results of a study of Japanese cattle strains that exhibited associations of tbpA, pfhA, and hgbB genes with strains collected from diseased animals rather than healthy ones [32]. The virulence-gene-associated profiles of the porcine isolates were highly diverse and showed clear separation between isolates of capsular types A and D. These isolates were associated with chronic and subchronic pneumonic pasteurellosis that persisted and multiplied for long periods (several generations) within the host’s respiratory tract, which could explain why more variations accumulated in these porcine isolates.

PCR amplification of virulence-associated genes P. multocida typically uses primers that specifically bind the conserved regions of the virulence-associated genes identified in different isolates of P. multocida. This strategy provides a rapid broad scanning of these genes across a large number of isolates, but the results of this method do not provide details on how the genes differ. This study selected the ompA gene, which encodes a highly abundant cell-surface-adhesive OMP, for further molecular evolutionary studies. The phylogenetic analyses of the ompA gene and its encoded protein conducted in this study confirmed that there is a clear separation between a few clusters of the bovine isolates and more diverse porcine clusters and a distinction between capsular types that is consistent with that previously shown by Verma et al. [40]. The results support the idea that low genetic diversity exists in the clonal population of the bovine isolates of capsular type B in Thailand. This conclusion is consistent with previous studies that compared 16S rRNA genes and found a monophyletic relationship and low genetic diversity of bovine strains of P. multocida in the UK [9] and India [44]. Interestingly, two bovine isolates of capsular type A (1C and 12C) collected from water buffaloes and cattle in 1989 and 1995 were grouped within the porcine cluster collected between 2005 and 2009, similar to the gene profiles shown in Fig. 1, suggesting that these bovine isolates were transferred from porcine hosts.

The ompA genes displayed a dN/dS ratio of less than one, indicating that these genes are under purifying or stabilising selection that does not allow the genes to change drastically over time, consistent with the calculation by Johnson et al. [11]. The selection pressures on different regions of the ompA genes of P. multocida were also observed to be unequal based on the fact that the multiple aligned OmpA sequences showed the presence of limited variable regions, a characteristic that could have caused the observed clustering in the protein phylogenetic tree. Mapping the aligned amino acid sequences of the transmembrane domain of the OmpA proteins in this study to the predicted OmpA structures clearly confirmed that the four hypervariable extracellular loops have experienced different selection pressures. Three of these loops were predicted to be more antigenic in the bovine cluster than in the porcine cluster, and these three loops contained more charged residues (positive or negative charges) than other loops, which could determine the specificity of adherence of the bacterium to host cells or to extracellular matrix molecules. These findings suggest that these three extracellular loops are potentially important in host adaptation and that they may play roles in the development of acute HS in cattle and water buffaloes. However, these strong antigenic epitopes might not be necessary for the development of porcine chronic pneumonic pasteurellosis; the adherence function of the loop regions of the OmpA proteins would likely be more important for this prolonged disease. Therefore, the amino acid sequences of these four loops were used to develop the LST scheme through which P. multocida strains were directly classified based on the specific sequences of the potential virulence-associated regions of their OmpA proteins.

Evidence for an adherence function of the OmpA protein in P. multocida was obtained in this study. OmpA of the bovine strain of P. multocida serotype A:3 adheres to heparin, heparin sulphate, type IX collagen, and fibronectin in the extracellular matrix layer of Madin-Darby bovine kidney (MDBK) cells [24, 45]. The OmpA protein preferentially bound to the N-terminal heparin-binding domain of fibronectin [45]. In our study, a heparin-binding site that includes a K/R-X-K/R motif was identified in the second extracellular loop (L2) of the protein (Fig. 4a). This second loop was the most enriched in positively charged amino acids (lysine, arginine, and histidine), accounting for an average of 25% of the loop length. Variations in the amino acid sequence of this loop could indicate certain levels of binding optimisation for specific host cell adherence. Recently, Katoch et al. [46] amplified and sequenced the ompA genes of four bovine strains of P. multocida (two of capsular type A and two of capsular type B). The authors found two different sequence types of the ompA genes that were specific to the strains of capsular types A (allele II) and B (allele I). These differences occurred within the four extracellular loop regions. The authors compared the in vitro and in vivo adhesion and invasion capacities of the two groups of strains and found that the capsular type B group with allele I of the ompA gene was more invasive than the other group. Unlike other studies, the current study could also differentiate porcine isolates of capsular types A and D, and bovine isolates of capsular type B based on the protein sequence-based LSTs (not the ompA alleles), and the differences in the LSTs could be directly linked to the molecular function of the OmpA protein. These assigned LSTs could potentially be used to determine the virulence of P. multocida isolates and further classify other P. multocida isolates e.g. bovine and porcine isolates of the same capsular type A. However, in vitro and in vivo assays may be required to confirm these findings.

Molecular variation of the OmpA proteins could be used to classify P. multocida isolates into different capsular types, hosts, and, possibly, pathogenicity levels. Taking into account this protein sequence-based typing together with virulence-associated gene profiles will provide a better understanding of P. multocida virulence. Effective monitoring and potential strain-specific subunit vaccines could be developed based on these loop oligopeptides.