Comparative pangenomics of Pasteurella multocida uncovers adhesins specific to capsular serogroup B strains associated with bovine host tropism Authors: Venugopal Nandish, Pillenahalli Sadashivappa Pooja, Anand Shirisha, Suresh Bindu, Mandrira Ramakrishna Namrutha, V. G. Shanmuga Priya, Mohammed Mudassar Chanda, Sathish Bhadravati Shivachandra Published: 2026 ABSTRACT Haemorrhagic septicaemia (HS), an acute and highly fatal disease of bovines in tropical countries of Africa and Asia, including India, is primarily caused by Pasteurella multocida serogroup B:2 strains. Although P. multocida strains are ubiquitous and known to infect a wide range of host species with varied clinical manifestations hitherto, the molecular basis of host tropism remains unknown. In this study, we present an integrated genomic, structural, and molecular dynamics framework to investigate how P. multocida serogroup B strains selectively adhere to bovine tissues. Comparative pangenomic analysis identified 115 genes significantly enriched in serogroup B strains, among which five lineage-specific putative adhesins were predicted through a multi-layer computational pipeline. In silico adhesin-fibronectin docking and interface energetics analyses identified preferential binding of these five adhesins to bovine fibronectin FnIII 9–10 domains, with markedly weaker interactions observed for non-bovine (caprine, pig, chicken and rabbit) homologues. Molecular dynamics simulations further revealed exceptional structural stability and favourable binding energetics for three adhesins (PACFBAAN_02021, PACFBAAN_00422 and PACFBAAN_00429), whereas two adhesins (PACFBAAN_01225 and PACFBAAN_01232) exhibited substantially lower adhesive performance. Together, these adhesins represent a potential molecular signature of bovine host tropism and warrant experimental validation for diagnostics, anti-adhesion therapeutics, and subunit vaccine development.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12866-026-05020-9. Introduction Pasteurella multocida is a Gram-negative bacterial pathogen of major veterinary and public health significance, causing distinct disease manifestations across diverse hosts such as haemorrhagic septicaemia (HS) in bovines, pneumonia in caprine, fowl cholera in poultry, snuffles in rabbits and atrophic rhinitis in swine [1, 2]. The pathogenic mechanisms of P. multocida vary considerably among its capsular serogroups and host species. For instance, serogroup A strains (predominantly A:1, A:3 and A:4) are widely distributed and infect multiple hosts (caprines, poultry and rabbits), whereas serogroup D:1 strains are mainly associated with porcine infections and serogroup B:2 strains are strongly linked to haemorrhagic septicaemia in bovines [2, 3]. In addition to animal infections, P. multocida can also cause zoonotic disease in humans following animal (cat and dog) bites or scratches, leading to septic arthritis, meningitis, endocarditis, pneumonia, and systemic sepsis [4–6]. The differences in host range and disease manifestations are attributed to substantial variation in virulence determinants involved in host colonization [7, 8]. Several virulence factors have been implicated in the pathogenicity of P. multocida, including Pasteurella multocida toxin (PMT), iron acquisition systems, and surface-associated adhesins [3]. Differences in these surface proteins arise directly from nucleotide sequence variation in the genes encoding them [9], suggesting the existence of capsular or somatic serogroup-specific adhesins that modulate host recognition and drive species-restricted colonization. However, it is likely that many key virulence factors are yet to be identified, including those required for the initial attachment and invasion of host cells [3]. Fibronectin (Fn) is a key extracellular matrix glycoprotein frequently targeted by bacterial adhesins during the earliest stages of host colonization [10]. Fn type III domains (FnIII 9–10) constitute the principal cell-binding module, encompassing the RGD motif and an adjacent synergy site that together mediate integrin binding, cell adhesion, and downstream signalling [11]. P. multocida has been reported to interact with fibronectin through these FnIII 9–10 domains, and species-specific sequence variation within these regions can influence adhesin binding affinity and specificity, thereby shaping host tropism [12]. A highly versatile capability of P. multocida strains to cause pathogenesis in a wide range of host species strongly suggests the presence of an armamentarium of adhesins that play a vital role in tissue colonization. However, to date, only a small number of potential adhesins, such as hemagglutinins [13], type IV fimbriae [14] and the outer membrane protein OmpA [15], widely conserved across serogroups, have been identified. In recent years, the advent of whole-genome sequencing (WGS) has substantially expanded our understanding of P. multocida genetic diversity and population structure. Comparative genomic studies have revealed an open pangenome characterized by a highly conserved core genome and remarkably variable accessory genome. This accessory repertoire is strongly shaped by horizontal gene transfer, mobile genetic elements, and lineage-specific adaptations that dictate host tropism and pathogenicity [16]. Pangenome analyses have demonstrated that while essential metabolic functions remain conserved across the species, the distribution of virulence factors, antimicrobial resistance genes, and outer membrane proteins varies substantially between host-adapted lineages [17]. Despite these advances, the accessory genomic determinants and adhesin repertoires responsible for the strict bovine host tropism of the classical serogroup B lineage remain poorly defined. In this context, the present study aimed to identify the genetic and structural determinants underlying the bovine-restricted colonization exhibited by P. multocida serogroup B strains. We employed a comparative pangenomic approach in which publicly available genomes of bovine-restricted serogroup B strains were systematically compared with genomes from the multi-host serogroup A lineage to identify lineage-specific genetic features potentially responsible for host specificity, with a particular emphasis on adhesin-encoding factors. To functionally interrogate this hypothesis, we integrated structural modelling with molecular docking analyses of serogroup B-specific adhesins against FnIII 9–10 domains derived from multiple host species. By comparing interaction energetics and interfacial residue contributions across hosts, this strategy was designed to directly test whether preferential binding to bovine fibronectin provides a molecular explanation for the bovine-restricted adhesion and colonization of serogroup B strains. Materials and methods Genome selection and data retrieval A curated dataset of 20 P. multocida strains was selected from 2,440 publicly available whole-genome sequences in the NCBI Assembly database (https://www.ncbi.nlm.nih.gov/) (accessed on 8 September 2025). Strains were selected based on (i) clearly documented capsular serogroup classification, (ii) reliable host and clinical metadata, and (iii) acceptable assembly quality, as assessed by CheckM v1.2.3 completeness and contamination estimates reported in NCBI records. The final dataset comprised serogroup A strains (n = 10) isolated from diverse hosts (sheep, cattle, avian, pig and human) and serogroup B strains (n = 10) exclusively from bovine hosts. While the majority of the selected bovine serogroup B strains were isolated from classical haemorrhagic septicaemia cases, three strains associated with bovine respiratory disease were also included in the dataset, as nasopharyngeal colonization and respiratory infection represent critical early steps in HS pathogenesis prior to systemic dissemination [18], inclusion of these respiratory strains provides important insight into host-restricted colonization. To minimize sampling bias and evaluate genetic relatedness among the selected strains, a whole-genome phylogenetic analysis was performed using the Type (Strain) Genome Server (TYGS) (https://tygs.dsmz.de/). The analysis employed the genome BLAST distance phylogeny (GBDP) method, which calculates intergenomic distances based on whole-genome BLAST comparisons. All genome assemblies were subjected to a standardized annotation to ensure uniformity and reproducibility across the selected P. multocida strains. Genome annotation was performed using Prokka v1.14.6 under default parameters. Prokka integrates multiple curated databases and employs the Prodigal algorithm for gene prediction, enabling the identification and annotation of coding sequences (CDSs) and other key genomic features [19]. The tool generated a comprehensive set of annotation files for each genome. Among these, the general feature format (.gff) files were selected, as they provide the standardized annotation framework required for pangenome construction and downstream analyses. Pangenome construction and comparative genomic analyses The pangenome of the selected P. multocida strains was assembled using Panaroo v1.3.3, an advanced graph-based pangenome analysis framework designed to produce high-quality gene clustering by addressing common assembly and annotation inconsistencies. The general feature format (.gff) files generated from Prokka annotation were used as input for Panaroo, ensuring standardized gene feature representation across all genomes. The analysis was performed in strict mode using a 95% protein sequence identity threshold for gene clustering. Genes present in ≥ 95% of strains were defined as core genes, whereas genes present in < 95% of strains were categorized as accessory genes. Panaroo incorporates an integrated error-correction algorithm that identifies and resolves fragmented, redundant, or misannotated genes, thereby enhancing the accuracy of orthologous group delineation [20]. Paralogous genes were handled through graph-based ortholog clustering, where paralog copies within the same genome were separated into distinct clusters based on sequence similarity and gene neighbourhood context. The default length difference cutoff (0.98) was retained to prevent clustering truncated genes with full-length homologues. This refinement facilitates the generation of gene presence-absence matrix, representing the full spectrum of core, accessory, and unique genes across the analyzed P. multocida genomes. A flower plot illustrating the distribution of core and accessory genes was generated using Python (Matplotlib v3.10.0). Genome-wide association analysis and functional annotation of capsular serogroup-specific genes To identify genetic determinants specifically associated with P. multocida serogroup B, a genome-wide association study (GWAS) was performed using Pyseer v1.3.10 [21]. The Panaroo-derived pangenome gene presence-absence matrix was used as the genetic input, while the serogroup classification (A or B) was assigned as the binary phenotype. Pyseer performed a gene-wise association analysis to test for statistical associations between gene presence and serogroup type. Given the curated dataset size, this analysis was intended to identify candidate lineage-associated genes within the analyzed dataset rather than to infer population-wide associations. Genes identified as significantly associated with serogroup B strains were translated into amino acid sequences using the Bio.Seq module of the Biopython package [22]. The resulting protein sequences were then functionally annotated using DIAMOND BLASTp v2.1.9 against the UniProt/Swiss-Prot database [23]. In silico prediction of adhesin proteins To identify putative adhesins, the translated protein sequences were submitted to the SPAAN (Software program for the Prediction of Adhesins and Adhesin-like proteins using Neural networks) server. SPAAN analyses sequence-derived features, including amino acid composition and physicochemical properties, to generate an adhesin probability score (Pad value) for each protein [24]. Proteins exhibiting a predicted adhesin probability (Pad value) greater than 0.5 were classified as putative adhesins and retained for downstream analyses. Subcellular localization and topology-based screening of predicted adhesins The adhesins predicted by SPAAN were further subjected to a comprehensive screening based on subcellular localization, membrane topology, and domain architecture to ensure their biological relevance. SignalP v6.0 was employed to predict the presence of N-terminal signal peptides indicative of secretion pathways [25]. DeepTMHMM v1.0.57 was utilized to identify transmembrane topology and classify residues as intracellular, membrane-spanning, or extracellular [26]. PSORTb v3.0.3 was used to determine subcellular localization. In addition, HMMER was used to identify conserved adhesin-associated domains through searches against the Pfam database [27]. Candidates that were corroborated by any of these tools were considered biologically relevant adhesins and retained for further analyses. Structural modelling and validation of putative adhesins and host fibronectin receptors Structural modelling The fibronectin sequences of bovine (P07589), caprine (A0A8C2PDJ5), pig (F1SS24), chicken (P11722), and rabbit (Q28749) were retrieved from the UniProt database (https://www.uniprot.org/). BLAST analysis revealed that the fibronectin sequences of cattle and buffalo, as well as those of sheep and goat, exhibited high sequence similarity and were therefore grouped and collectively referred to as bovine and caprine, respectively. As the interaction of P. multocida with fibronectin occurs through fibronectin type III (FnIII) 9–10 domains [12], only these domains were extracted using HMMER for structural modelling to ensure biological relevance and computational efficiency. The putative adhesins and the respective host fibronectin domains were modelled using AlphaFold3 (https://alphafoldserver.com) server to obtain high-confidence models which were evaluated based on AlphaFold predicted template modelling (pTM) scores. All predicted structures were further subjected to atomistic refinement using the GalaxyRefine server to improve side-chain conformations and overall structural quality prior to validation and docking analyses [28]. Validation of putative adhesins The predicted structural models were validated using various computational tools to assess stereochemical quality, structural integrity and reliability. Ramachandran plot analysis was performed using PROCHECK to examine backbone dihedral angles, with the percentage of residues in favoured regions used as a quality metric. Global and local model quality was assessed using the QMEANDisCo (Qualitative Model Energy Analysis Distance Constraints) scoring functions provided by the SWISS-MODEL assessment pipeline. QMEANDisCo is a knowledge-based scoring function that evaluates the agreement between predicted structural features and experimentally determined protein structures. Non-bonded atomic interactions were assessed using ERRAT through the SAVES v6.1 server (https://saves.mbi.ucla.edu), while MolProbity v4.4 (http://molprobity.biochem.duke.edu) was employed to evaluate steric clashes, rotamer conformations, and Ramachandran outliers [29]. Structural visualization and analysis were conducted using PyMOL v3.1.3 (Schrödinger, LLC) and UCSF ChimeraX v1.9. Molecular docking and interface energetics calculations Molecular docking analysis was performed to investigate the binding interactions between the modelled putative adhesins and the fibronectin type III (FnIII) 9–10 domains of different host species, including bovine, caprine, pig, chicken, and rabbit. Docking was conducted using the HADDOCK2.4 (High Ambiguity Driven biomolecular DOCKing) web server (https://wenmr.science.uu.nl/haddock2.4/) which employs data-driven restraints and energy-based optimization to predict biologically relevant protein-protein complexes. Putative binding residues were identified using the COACH-D server (https://aideepmed.com/COACH/) based on the top-ranked predicted binding pocket. Specifically, residues within the highest-confidence consensus binding cluster were defined as active residues. These residues were incorporated into the semi-flexible docking protocol of HADDOCK. Surface-exposed neighbouring residues were automatically treated as passive residues according to HADDOCK default settings, enabling interface-guided complex assembly [30]. The resulting complexes were evaluated based on HADDOCK scores, interface root-mean-square deviation (i-RMSD), and binding energy, with the top-ranked models selected for further analyses [31]. To evaluate the specificity of adhesin-fibronectin interactions, a negative control docking analysis was performed using a non-adhesin protein (PACFBAAN_01220) selected from the 115 serogroup B-associated genes. The protein structure was modelled using AlphaFold3 (pTM = 0.65) and docked against the bovine fibronectin FnIII 9–10 domain using identical HADDOCK parameters and active residue definitions. All docking protocols were kept consistent to enable direct comparison with the candidate adhesins. HADDOCK-generated complexes were further refined using the implicit solvent optimization protocols available in the HADDOCK2.4 refinement interface (https://rascar.science.uu.nl/haddock2.4/refinement/1). Binding affinities (ΔG) for top-ranked complexes were predicted using PRODIGY (PROtein binDIng enerGY prediction) (https://prodigy.embl.de/) based on interface residue contacts and structural parameters [32]. For comprehensive interface characterization, PDBePISA (Protein Data Bank in Europe-Proteins, Interfaces, Structures and Assemblies) (https://www.ebi.ac.uk/pdbe/pisa/) was employed to characterize the protein-protein interfaces by calculating the interface area (Ų) and solvation free energy gain (ΔiG) upon complex formation [33]. Molecular dynamics simulations and free energy landscape (FEL) mapping Molecular dynamics (MD) simulations were performed using GROMACS v2024.5 with the OPLS-AA all-atom force field to evaluate the dynamic stability of adhesins docked to bovine fibronectin domain complexes [34]. Each complex was enclosed in a cubic simulation box and solvated with the SPC/E water model. Na⁺ and Cl⁻ ions were added to neutralize the system and approximate physiological ionic strength. Energy minimization was carried out using the steepest descent algorithm to remove steric clashes. Equilibration was carried out in two consecutive phases: a 1 ns constant-volume (NVT) ensemble at 300 K maintained using a V-rescale thermostat, followed by a 1 ns constant-pressure (NPT) ensemble at 1 bar with the Parrinello–Rahman barostat under isotropic pressure coupling [35]. A 100 ns production MD simulation was performed under identical thermodynamic conditions with trajectory snapshots recorded every 20 ps for downstream analyses. Structural stability and conformational dynamics were quantified by computing the backbone root mean square deviation (RMSD) and per-residue root mean square fluctuation (RMSF). The radius of gyration (Rg) was used to monitor global compactness, while solvent-accessible surface area (SASA) measurements probed variations in solvent exposure and surface hydrophobicity. Inter and intramolecular hydrogen bond occupancies between protein-protein interfaces were tracked to evaluate the persistence of stabilizing interactions within the complexes. For statistical comparison of conformational stability among the five adhesin-fibronectin complexes, RMSD, Rg, and SASA values were extracted from the equilibrated portion of each trajectory corresponding to the final 20 ns of simulation. One-way ANOVA (analysis of variance) was performed to assess differences among complexes, followed by Tukey’s honestly significant difference (HSD) post-hoc test with a family-wise error rate (FWER) of 0.05 to identify statistically significant pairwise differences. Principal component analysis (PCA) was performed on the covariance matrix of Cα-atom fluctuations, and the dominant motions were extracted from the first two principal components. The free energy landscape (FEL) was reconstructed as a function of these components (PC1 and PC2), revealing distinct low-energy clusters corresponding to the most stable conformations. The FEL plots were generated using the ggplot2 package in R [36]. To further quantify conformational basin characteristics, PC1-PC2 energy surfaces were extracted from the corresponding fel.dat files containing 1024 sampled conformations. Conformations with free energy within 4 kJ/mol of the global minimum were assigned to the dominant basin, and their number was used to estimate the relative population of stable conformational states for each complex. Binding free energies were calculated using the MM/GBSA (Molecular Mechanics/Generalized Born Surface Area) method, a molecular mechanics-based approach used to estimate protein-protein binding free energies, implemented in the HawkDock server (https://cadd.zju.edu.cn/hawkdock/) [37] in accordance with the equation:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\Delta\mathrm{G}=\Delta\mathrm{E}_\mathrm{vdw}+\Delta\mathrm{E}_\mathrm{ele}+\Delta\mathrm{G}_\mathrm{GB}+\Delta\mathrm{G}_\mathrm{SA}$$\end{document} Where ΔEvdw and ΔEele denote the van der Waals and electrostatic interaction energies respectively, ΔGGB denotes the polar solvation energy calculated using the generalized born (GB) model and ΔGSA​ corresponds to the nonpolar solvation energy estimated from the solvent-accessible surface area (SASA). Protein-protein interaction analysis and visualization The PDBsum server (https://www.ebi.ac.uk/thornton-srv/databases/pdbsum/) was used to identify and assess specific molecular interactions within the docked complexes [38]. The analysis provided detailed information on interfacial residues, hydrogen bonds, salt bridges, and hydrophobic contacts contributing to complex stabilization. Structural visualization of the interacting residues was performed using PyMOL v3.1.3 (Schrödinger, LLC), UCSF ChimeraX v1.9 and LigPlot [39]. Results Annotated P. multocida genomic datasets A total of 20 P. multocida genomes representing serogroups A and B were compiled for comparative genomic analyses. Genome completeness and contamination estimates previously assessed using CheckM v1.2.3 confirmed high assembly quality, with completeness values ranging from approximately 94% to 99% and low contamination levels (≤ 1.2%). The dataset included both complete genomes and high-quality draft assemblies, with contig numbers varying across strains (Table 1). Table 1Clinical metadata and assembly quality metrics of the P. multocida strains selected for comparative analysisSl no.SerogroupStrainAssembly AccessionHostReported diseaseYear of IsolationIsolation SourceAssembly levelNo. of ContigsCompleteness (%)Contamination (%)Country1ANIVEDIPm10GCA_027951325.1sheepPneumonia2019Nasal DischargeContig level11895.601.19India2ANIVEDIPm22GCA_027951355.1sheepPneumonia2019Nasal DischargeContig level10095.621.16India3ANIVEDIPm31GCA_027951305.1sheepPneumonia2019Nasal DischargeContig level13494.590.97India4ACQ7GCA_009739575.1CattlePneumonia2013Pneumonic LungsComplete genome198.850.16China5A190,067GCA_048568305.1CattleNA2019NAComplete genome298.840.72Germany6AP2237GCA_024055505.2AvianNA2022NAComplete genome198.190.05USA7AP2192GCA_023573045.1AvianNA2022NAComplete genome298.120.20USA8AHN05GCA_013012805.1PigPorcine Pneumonia2017LungsContig level5198.230.53China9APM1463GCA_037021625.1PigPorcine Pneumonia2011NAComplete genome298.570.76Australia10AHuN001GCA_018139065.1HumanNA2021NAComplete genome198.580.93China11B36,950GCA_000234745.1BovineBovine Respiratory Disease2014NAComplete genome197.510.70Germany12BP52GCA_009663815.1BovineHaemorrhagic Septicaemia2019NAContig level6199.100.21India13B12,601GCA_003261355.1BovineHaemorrhagic Septicaemia2006Nasal SwabsComplete genome198.160.72Denmark14BNCTC10323GCA_900638665.1BovineHaemorrhagic Septicaemia2018NAComplete genome198.540.16Myanmar15BPm1GCA_022179625.1BovineBovine Respiratory Disease2017Oronasal SecretionComplete genome197.490.93Japan16B17BRD-035GCA_019916005.1BovineBovine Respiratory Disease2017NAComplete genome198.850.31Australia17BPM3GCA_004114645.1BovineHaemorrhagic Septicaemia2015BloodScaffold level2699.110.19Pakistan18BBAUTB2GCA_003268295.1BovineHaemorrhagic Septicaemia2016Throat SamplesContig level15899.090.68Bangladesh19BVTCCBAA264GCA_000296345.2BovineHaemorrhagic Septicaemia2011Small IntestineContig level7894.140.21India20BNIVEDIPm34GCA_022869065.1BovineHaemorrhagic Septicaemia2019TissueContig level3699.160.16IndiaNA indicates information was not available in the public database Table 1Clinical metadata and assembly quality metrics of the P. multocida strains selected for comparative analysisSl no.SerogroupStrainAssembly AccessionHostReported diseaseYear of IsolationIsolation SourceAssembly levelNo. of ContigsCompleteness (%)Contamination (%)Country1ANIVEDIPm10GCA_027951325.1sheepPneumonia2019Nasal DischargeContig level11895.601.19India2ANIVEDIPm22GCA_027951355.1sheepPneumonia2019Nasal DischargeContig level10095.621.16India3ANIVEDIPm31GCA_027951305.1sheepPneumonia2019Nasal DischargeContig level13494.590.97India4ACQ7GCA_009739575.1CattlePneumonia2013Pneumonic LungsComplete genome198.850.16China5A190,067GCA_048568305.1CattleNA2019NAComplete genome298.840.72Germany6AP2237GCA_024055505.2AvianNA2022NAComplete genome198.190.05USA7AP2192GCA_023573045.1AvianNA2022NAComplete genome298.120.20USA8AHN05GCA_013012805.1PigPorcine Pneumonia2017LungsContig level5198.230.53China9APM1463GCA_037021625.1PigPorcine Pneumonia2011NAComplete genome298.570.76Australia10AHuN001GCA_018139065.1HumanNA2021NAComplete genome198.580.93China11B36,950GCA_000234745.1BovineBovine Respiratory Disease2014NAComplete genome197.510.70Germany12BP52GCA_009663815.1BovineHaemorrhagic Septicaemia2019NAContig level6199.100.21India13B12,601GCA_003261355.1BovineHaemorrhagic Septicaemia2006Nasal SwabsComplete genome198.160.72Denmark14BNCTC10323GCA_900638665.1BovineHaemorrhagic Septicaemia2018NAComplete genome198.540.16Myanmar15BPm1GCA_022179625.1BovineBovine Respiratory Disease2017Oronasal SecretionComplete genome197.490.93Japan16B17BRD-035GCA_019916005.1BovineBovine Respiratory Disease2017NAComplete genome198.850.31Australia17BPM3GCA_004114645.1BovineHaemorrhagic Septicaemia2015BloodScaffold level2699.110.19Pakistan18BBAUTB2GCA_003268295.1BovineHaemorrhagic Septicaemia2016Throat SamplesContig level15899.090.68Bangladesh19BVTCCBAA264GCA_000296345.2BovineHaemorrhagic Septicaemia2011Small IntestineContig level7894.140.21India20BNIVEDIPm34GCA_022869065.1BovineHaemorrhagic Septicaemia2019TissueContig level3699.160.16IndiaNA indicates information was not available in the public database Clinical metadata and assembly quality metrics of the P. multocida strains selected for comparative analysis NA indicates information was not available in the public database Whole-genome phylogenetic analysis revealed that the selected strains span distinct lineages and are not clonally redundant (Fig. 1). Serogroup A strains were distributed across multiple branches of the phylogeny as they originated from diverse host species. In contrast, serogroup B strains formed a more closely related cluster; however, variation in branch lengths indicates substantial underlying genomic divergence among the strains. Fig. 1Whole-genome phylogenetic analysis of selected P. multocida strains representing capsular serogroup A (blue) and serogroup B (red) Whole-genome phylogenetic analysis of selected P. multocida strains representing capsular serogroup A (blue) and serogroup B (red) Following genome retrieval, all P. multocida strains were uniformly annotated, ensuring consistent and comparable gene predictions across datasets. Prokka generated a complete set of annotation files for each genome, including .gff, .gbk, .faa, .fsa, .tsv and .tbl files. Among these, .gff files were selected for downstream analysis, as they provide a compact, structured representation of all genomic features. Each .gff file begins with contig-level sequence-region descriptors, followed by line-wise annotations of CDSs and other features including coordinates, strand orientation, locus tags, and predicted products information essential for accurate gene comparison and pan-genome construction. Comparative pangenome analysis Pangenome analysis generated several key output files, including gene_presence_absence.Rtab which provides a binary matrix indicating the presence or absence of each gene across all genomes, gene_presence_absence.csv which summarizes gene clusters along with their functional annotations, summary_statistics.txt which reports core and accessory genome metrics and pan_genome_reference.fa which contains representative sequences of each gene cluster. The summary statistics indicated a total of 1,816 core genes shared across all strains and 2,087 accessory genes (soft core + shell + cloud). These accessory genes constitute the variable genome fraction and represent strong candidates for uncovering genetic determinants linked to host restriction and virulence (Fig. 2). Fig. 2Flower plot representation of the P. multocida pangenome. The central circle represents core genes (n = 1,816). Each petal shows the number of accessory genes present in an individual genome (Serogroup A in blue; Serogroup B in green). The plot was generated using Python (Matplotlib v3.10.0) Flower plot representation of the P. multocida pangenome. The central circle represents core genes (n = 1,816). Each petal shows the number of accessory genes present in an individual genome (Serogroup A in blue; Serogroup B in green). The plot was generated using Python (Matplotlib v3.10.0) P. multocida serogroup B associated genes and functional annotation Pyseer identified 115 statistically significant genes from a total of 2,087 accessory genes that were enriched in serogroup B strains and absent in serogroup A strains within the analysed dataset, highlighting their potential association with the serogroup B lineage. The corresponding nucleotide sequences for these significant gene clusters were retrieved from the Panaroo generated .fa reference file (Supplementary Material 1) to enable further functional characterization and downstream analyses. These nucleotide sequences were subsequently translated into amino acid sequences for functional annotation. Functional annotation of the 115 serogroup B-associated genes provided putative functional descriptions for their predicted protein products, and the resulting annotations are summarized in supplementary material 2. The annotated genes include proteins with diverse predicted functions as well as several hypothetical proteins lacking characterized homologs. Predicted adhesin proteins Considering the association of adhesins in host colonization, the translated protein sequences derived from the serogroup B-associated genes were screened for adhesin-like properties. The SPAAN server predicted a total of 18 proteins with Pad values > 0.5, which were classified as putative adhesins (Table 2) and retained for further functional characterization. Table 2Putative adhesins predicted by SPAAN with scores > 0.5Sl NoPad-valueCluster IDProtein names (Functional Annotation)10.631782> group_2022Hypothetical protein20.647892> group_537Putative quercetin 2,3-dioxygenase30.775149> group_2023Probable tail fiber protein40.705226> group_2026Major outer membrane protein50.607925> group_651Hypothetical protein60.606143> group_1334Hypothetical protein70.696645> group_1295Hypothetical protein80.656193> group_523Hypothetical protein90.715192> group_1016Hypothetical protein100.614485> group_1347Hypothetical protein110.629211> group_1326Hypothetical protein120.735838> group_1327Uncharacterized protein130.867247> group_274Mu-like prophage FluMu protein gp42140.599438> group_1066Hypothetical protein150.683164> group_438Mu-like prophage FluMu protein gp47160.511022> group_347tRNA nuclease170.519274> group_2027Prepilin peptidase-dependent protein D180.526701> group_1861Hypothetical protein Table 2Putative adhesins predicted by SPAAN with scores > 0.5Sl NoPad-valueCluster IDProtein names (Functional Annotation)10.631782> group_2022Hypothetical protein20.647892> group_537Putative quercetin 2,3-dioxygenase30.775149> group_2023Probable tail fiber protein40.705226> group_2026Major outer membrane protein50.607925> group_651Hypothetical protein60.606143> group_1334Hypothetical protein70.696645> group_1295Hypothetical protein80.656193> group_523Hypothetical protein90.715192> group_1016Hypothetical protein100.614485> group_1347Hypothetical protein110.629211> group_1326Hypothetical protein120.735838> group_1327Uncharacterized protein130.867247> group_274Mu-like prophage FluMu protein gp42140.599438> group_1066Hypothetical protein150.683164> group_438Mu-like prophage FluMu protein gp47160.511022> group_347tRNA nuclease170.519274> group_2027Prepilin peptidase-dependent protein D180.526701> group_1861Hypothetical protein Putative adhesins predicted by SPAAN with scores > 0.5 Topology and cellular localization analyses of the 18 putative adhesins identified five high-confidence adhesin candidates. Subcellular localization predictions using DeepTMHMM and PSORTb revealed four candidates as surface-associated or outer membrane proteins. The remaining candidate lacked a definitive localization prediction but contained adhesion-associated Ig-like/Cna_B domains detected through HMMER searches. These five candidates were retained for subsequent structural and functional analyses (Supplementary Material 3). For standardized identification throughout the analyses, the predicted adhesin candidates were mapped to the P. multocida strain P52 reference genome and assigned corresponding NCBI RefSeq locus tags where available. One candidate (group_1016) lacked an official RefSeq locus tag, likely due to differences between RefSeq and Prokka annotation pipelines, as certain predicted open reading frames may not be annotated in the RefSeq assembly. This gene was therefore identified using a locally generated Prokka annotation of the strain P52 genome. To maintain consistency across downstream analyses, locus tags derived from the Prokka re-annotation were used for adhesin identification. The complete Prokka annotation of strain P52 genome is provided in supplementary material 4, and the summarized localization and topology features of the adhesin candidates, along with their locus tags, are presented in Table 3. Table 3Features of putative adhesin candidates from P. multocida serogroup B strainsSl NoCluster IDNCBI RefSeq locus tagsProkka locus tagNumber of Amino acidsSignalPDeepTMHMMPSORTb/HMMER1> group_2026FQ481_RS05610PACFBAAN_02021356SP prob = 0.97OutsideOuter membrane (score 10.0, SP detected)2> group_523FQ481_RS07615PACFBAAN_00422303No clear SPOutsideSignal peptide detected, but localization uncertain3> group_1016-PACFBAAN_00429134No SP detected-Unknown / [Ig-like / Cna_B domains (HMMER)]4> group_274FQ481_RS10485PACFBAAN_01225633No SP detectedOutsideOuter membrane (score 9.64)5> group_438FQ481_RS10515PACFBAAN_01232355No SP detectedOutsideUnknown / [Baseplate_J domain (HMMER)]SP Signal peptide Table 3Features of putative adhesin candidates from P. multocida serogroup B strainsSl NoCluster IDNCBI RefSeq locus tagsProkka locus tagNumber of Amino acidsSignalPDeepTMHMMPSORTb/HMMER1> group_2026FQ481_RS05610PACFBAAN_02021356SP prob = 0.97OutsideOuter membrane (score 10.0, SP detected)2> group_523FQ481_RS07615PACFBAAN_00422303No clear SPOutsideSignal peptide detected, but localization uncertain3> group_1016-PACFBAAN_00429134No SP detected-Unknown / [Ig-like / Cna_B domains (HMMER)]4> group_274FQ481_RS10485PACFBAAN_01225633No SP detectedOutsideOuter membrane (score 9.64)5> group_438FQ481_RS10515PACFBAAN_01232355No SP detectedOutsideUnknown / [Baseplate_J domain (HMMER)]SP Signal peptide Features of putative adhesin candidates from P. multocida serogroup B strains SP Signal peptide Structural modelling and validation of adhesins and host fibronectin receptors The AlphaFold3 predicted structures of the putative adhesins and the corresponding host fibronectin (FnIII 9–10) domains exhibited well-defined folds consistent with fibronectin-binding protein architectures (Fig. 3). The adhesin models showed stable secondary structure arrangements dominated by β-sheets and connecting loops, features commonly observed in fibronectin-interacting proteins. Model confidence assessment indicated that all adhesin structures, except PACFBAAN_00429, exhibited pTM scores ranging from 0.6 to 0.88, whereas the fibronectin receptor domain models showed pTM values between 0.73 and 0.81 (Table 4), indicating variable levels of confidence across the predicted protein structures. Fig. 3Three-dimensional structural models of putative adhesins and different hosts fibronectin (FnIII 9–10) domains. Panel A: PACFBAAN_02021, Panel B: PACFBAAN_00422, Panel C: PACFBAAN_00429, Panel D: PACFBAAN_01225, Panel E: PACFBAAN_01232, Panel F: Bovine, Panel G: Caprine, Panel H: Pig, Panel I: Chicken, Panel J: Rabbit Three-dimensional structural models of putative adhesins and different hosts fibronectin (FnIII 9–10) domains. Panel A: PACFBAAN_02021, Panel B: PACFBAAN_00422, Panel C: PACFBAAN_00429, Panel D: PACFBAAN_01225, Panel E: PACFBAAN_01232, Panel F: Bovine, Panel G: Caprine, Panel H: Pig, Panel I: Chicken, Panel J: Rabbit Table 4Structural validation of putative adhesins and host fibronectin receptorsValidation indexNumber of Amino acidspTM scoreQMEANDisCo GlobalProCheckERRATMolProbityRamachandran FavouredBad BondsC-Beta DeviationsClash ScoreRamachandran OutliersRotamer OutliersMolProbity ScorePutative Adhesins PACFBAAN_020213560.880.53 ± 0.0594.92%0190.060.180.28%0.71%0.92 PACFBAAN_004223030.600.72 ± 0.0598.01%00100.000.210.00%1.99%0.81 PACFBAAN_004291340.390.38 ± 0.07100.00%0194.252.340.00%0.00%1.01 PACFBAAN_012256330.810.50 ± 0.0593.03%1293.523.763.17%1.89%1.82 PACFBAAN_012323550.810.72 ± 0.0598.02%0098.531.300.28%0.00%0.85Fibronectin (FnIII 9–10 domain) Receptors Bovine1920.810.84 ± 0.0696.84%0188.330.341.05%0.00%0.82 Caprine2000.750.82 ± 0.0693.43%0587.030.001.52%1.68%1.10 Pig1940.750.80 ± 0.0697.38%1277.842.731.05%3.01%1.55 Chicken1950.760.82 ± 0.0697.93%1182.951.360.52%1.78%1.07 Rabbit2010.730.79 ± 0.0698.49%0280.950.330.00%0.00%0.62 Table 4Structural validation of putative adhesins and host fibronectin receptorsValidation indexNumber of Amino acidspTM scoreQMEANDisCo GlobalProCheckERRATMolProbityRamachandran FavouredBad BondsC-Beta DeviationsClash ScoreRamachandran OutliersRotamer OutliersMolProbity ScorePutative Adhesins PACFBAAN_020213560.880.53 ± 0.0594.92%0190.060.180.28%0.71%0.92 PACFBAAN_004223030.600.72 ± 0.0598.01%00100.000.210.00%1.99%0.81 PACFBAAN_004291340.390.38 ± 0.07100.00%0194.252.340.00%0.00%1.01 PACFBAAN_012256330.810.50 ± 0.0593.03%1293.523.763.17%1.89%1.82 PACFBAAN_012323550.810.72 ± 0.0598.02%0098.531.300.28%0.00%0.85Fibronectin (FnIII 9–10 domain) Receptors Bovine1920.810.84 ± 0.0696.84%0188.330.341.05%0.00%0.82 Caprine2000.750.82 ± 0.0693.43%0587.030.001.52%1.68%1.10 Pig1940.750.80 ± 0.0697.38%1277.842.731.05%3.01%1.55 Chicken1950.760.82 ± 0.0697.93%1182.951.360.52%1.78%1.07 Rabbit2010.730.79 ± 0.0698.49%0280.950.330.00%0.00%0.62 Structural validation of putative adhesins and host fibronectin receptors Global model quality evaluation using QMEANDisCo revealed scores ranging from 0.50 ± 0.07 to 0.84 ± 0.06, demonstrating good agreement between the predicted models and experimentally determined protein structures. Stereochemical evaluation using PROCHECK showed the presence of 93–100% of residues within the favoured regions of the Ramachandran plot, with only 0–1 bad bond across all structures. ERRAT scores exceeded 90 for most structures, and MolProbity scores ranged from 0.6 to 1.8, consistent with values observed for high-resolution experimental structures. Among the adhesin candidates, PACFBAAN_00429 exhibited a comparatively lower AlphaFold confidence score (pTM = 0.39). However, post-refinement stereochemical validation demonstrated excellent geometric quality, with 100% residues in Ramachandran favoured regions, no bad bonds, an ERRAT score of 94.25, and favourable MolProbity metrics, supporting its suitability for exploratory interaction analyses. Collectively, these validation metrics suggest that the predicted adhesin and fibronectin (FnIII 9–10) domain models possess acceptable stereochemical quality for downstream docking and interaction analyses. Molecular docking analysis Molecular docking analysis revealed preferable binding of putative adhesins with bovine fibronectin, with favourable HADDOCK scores (-106.8 +/- 1.6 to -133.3 +/- 1.0) and substantially larger buried surface areas (1707.9 +/- 64.2 to 2035.5 +/- 55.9 Ų). In contrast, docking against fibronectin domains from caprine, pig, chicken, and rabbit sources yielded consistently less favourable scores (-52.9 +/- 0.4 to -95.9 +/- 1.5) and significantly reduced interface areas, generally < 1600 Ų, indicating weaker adhesive contacts (Table 5). Notably, negative control docking using a non-adhesin protein against bovine fibronectin exhibited markedly weaker interactions, with a less favourable HADDOCK score (-51.5 +/- 9.5) and reduced buried surface area (~ 1354.3 +/- 114.6 Ų). Table 5HADDOCK docking scores and interface energetics of putative adhesins with fibronectin (FnIII 9–10) domains from different host species Table 5HADDOCK docking scores and interface energetics of putative adhesins with fibronectin (FnIII 9–10) domains from different host species HADDOCK docking scores and interface energetics of putative adhesins with fibronectin (FnIII 9–10) domains from different host species Binding affinity estimates derived from PRODIGY corroborated these findings, with adhesin-bovine fibronectin complexes exhibiting the most favourable predicted binding energies (ΔG = -10.1 to -14.6 kcal mol⁻¹), whereas interactions with fibronectin from other hosts showed comparatively reduced affinities (ΔG: -4.9 to -9.3 kcal mol⁻¹). Consistently, the negative control exhibited lower binding affinity (ΔG = -7.6 kcal mol⁻¹) compared to adhesin- bovine fibronectin complexes, confirming that the observed preferential interactions are specific to the identified adhesins rather than a non-specific property of serogroup B-associated proteins. PDBePISA interface analysis further supported these observations (Table 5). Adhesin complexes with bovine fibronectin produced the most favourable solvation free energy gains (ΔiG = -8.4 to -10.9 kcal/mol) and interface areas consistently exceeding 1000 Ų. Although PACFBAAN_01225 yielded a slightly weaker ΔiG (-3.7 kcal/mol), it still maintained a substantial interface (1061.5 Ų). Conversely, adhesin interactions with fibronectin from other hosts were characterized by lower solvation free energy gains (ΔiG = 2.1 to -9.8 kcal/mol) and reduced interface areas, suggesting weakened interfacial stability (Table 5). Molecular dynamics simulations and free energy landscape (FEL) mapping Molecular dynamics (MD) simulations of the adhesin-bovine fibronectin complexes revealed clear differences in structural stability among the five candidate adhesins. RMSD trajectories showed that PACFBAAN_02021, PACFBAAN_00422 and PACFBAAN_00429 rapidly equilibrated within the first 5–10 ns and remained stable throughout the simulation with backbone deviations of ~ 0.3–0.6 nm. In contrast, PACFBAAN_01232 had moderate structural drift, and PACFBAAN_01225 showed large fluctuations (> 1.5 nm), suggesting substantial conformational rearrangements (Fig. 4A). Statistical analysis confirmed significant differences among complexes (ANOVA: F = 3536.79, p = 3.45 × 10⁻²⁸). Tukey post-hoc comparisons revealed no significant difference between PACFBAAN_02021 and PACFBAAN_00422, suggesting comparable stability. PACFBAAN_00429 showed intermediate RMSD behaviour, while PACFBAAN_01232 exhibited moderately elevated deviations consistent with increased flexibility. In contrast, PACFBAAN_01225 differed significantly from all other complexes, reflecting pronounced structural instability. Fig. 4Molecular dynamics (MD) simulation analysis of putative adhesins and bovine FnIII 9–10 complex. Panel A: Root mean square deviation (RMSD) showing the structural stability of each adhesin-fibronectin complex during the simulation. Panel B: Radius of gyration (Rg) indicating the compactness of the protein complexes over time. Panel C: Solvent accessible surface area (SASA) reflecting changes in surface exposure that may influence protein-protein interactions. Panel D: Hydrogen bond profiles illustrating the stability of intermolecular interactions between adhesins and fibronectin. Panels E, F, G, H, I: Root mean square fluctuation (RMSF) representing residue-level flexibility of the adhesin proteins during the simulation Molecular dynamics (MD) simulation analysis of putative adhesins and bovine FnIII 9–10 complex. Panel A: Root mean square deviation (RMSD) showing the structural stability of each adhesin-fibronectin complex during the simulation. Panel B: Radius of gyration (Rg) indicating the compactness of the protein complexes over time. Panel C: Solvent accessible surface area (SASA) reflecting changes in surface exposure that may influence protein-protein interactions. Panel D: Hydrogen bond profiles illustrating the stability of intermolecular interactions between adhesins and fibronectin. Panels E, F, G, H, I: Root mean square fluctuation (RMSF) representing residue-level flexibility of the adhesin proteins during the simulation Radius of gyration profiles indicated compact conformations for PACFBAAN_02021, PACFBAAN_00422 and PACFBAAN_00429 (~ 2.5–2.9 nm), whereas PACFBAAN_01232 showed moderate expansion and PACFBAAN_01225 exhibited the least compact structure (> 3.0 nm) (Fig. 4B). Differences among complexes were significant (ANOVA: F = 1873.46, p = 1.96 × 10⁻²⁵), with Tukey post-hoc analysis showing significantly lower Rg values for PACFBAAN_00429 compared with PACFBAAN_01232 and PACFBAAN_01225. Similarly, solvent accessible surface area (SASA) analysis showed the lowest solvent exposure for PACFBAAN_00429 (160–240 nm²), intermediate values for PACFBAAN_02021, PACFBAAN_00422 and PACFBAAN_01232 (~ 270–340 nm²), and the highest exposure for PACFBAAN_01225 (~ 600 –400 nm²) (Fig. 4C). These differences were also significant (ANOVA: F = 6389.62, p = 9.38 × 10⁻³¹), with Tukey post-hoc comparisons confirming significantly lower SASA values for PACFBAAN_00429 and the highest exposure for PACFBAAN_01225. Intramolecular hydrogen-bond analysis further supported these structural trends. PACFBAAN_00429 maintained a stable H-bond network (~ 200–230), whereas PACFBAAN_02021, PACFBAAN_00422 and PACFBAAN_01232 exhibited higher but stable counts (~ 340–450). In contrast, PACFBAAN_01225 showed the highest H-bond numbers (> 450) accompanied by pronounced temporal fluctuations, consistent with its larger and more flexible architecture (Fig. 4D). Residue-level flexibility assessed by RMSF analysis revealed minimal fluctuations for PACFBAAN_00429, followed by PACFBAAN_02021 and PACFBAAN_00422 (< 0.3 nm across most residues) (Fig. 4E-G). PACFBAAN_01232 showed localized loop flexibility, whereas PACFBAAN_01225 exhibited extensive residue fluctuations across multiple domains (Fig. 4H-I), consistent with its elevated RMSD and SASA values. Free energy landscape (FEL) analysis further characterized the conformational behaviour of the adhesin–fibronectin complexes. PACFBAAN_02021 and PACFBAAN_00422 occupied deep and well-defined minima with dominant basin populations of 253 conformations (~ 24.7%), consistent with stable conformational states (Fig. 5A-B). PACFBAAN_01232 formed the largest basin population (354 conformations; ~34.6%) but across a broader energy surface, suggesting transitions among metastable conformations (Fig. 5E). In contrast, PACFBAAN_00429 occupied a smaller basin (40 conformations; ~3.9%), reflecting increased conformational variability (Fig. 5C), whereas PACFBAAN_01225 exhibited the smallest basin population (25 conformations; ~2.4%), consistent with pronounced conformational heterogeneity and reduced structural stability (Fig. 5D). Fig. 5Conformational free energy landscapes of putative adhesins and bovine FnIII 9-10 complex. Panel A: PACFBAAN_02021, Panel B: PACFBAAN_00422, Panel C: PACFBAAN_00429, Panel D: PACFBAAN_01225, Panel E: PACFBAAN_01232 Conformational free energy landscapes of putative adhesins and bovine FnIII 9-10 complex. Panel A: PACFBAAN_02021, Panel B: PACFBAAN_00422, Panel C: PACFBAAN_00429, Panel D: PACFBAAN_01225, Panel E: PACFBAAN_01232 MM/GBSA-based binding free-energy prediction provided quantitative insight into the thermodynamic forces governing the adhesin-bovine fibronectin interactions. PACFBAAN_02021, PACFBAAN_00422 and PACFBAAN_00429 had the most favourable binding free energies (-68.12, -68.85, and − 67.88 kcal/mol, respectively), supported by strong van der Waals interactions (-111.16 to -78.35 kcal/mol), highly attractive electrostatic components (-517.50 to -127.73 kcal/mol), and stabilizing nonpolar solvation energies. These energetic profiles are fully consistent with their structurally stable behaviour observed in MD analyses. PACFBAAN_01232 exhibited a moderately favourable binding affinity (-53.00 kcal/mol), characterized by favourable van der Waals stabilization (-111.80 kcal/mol) partially offset by an unfavourable electrostatic term (+ 152.91 kcal/mol), aligning with its intermediate structural stability. In contrast, PACFBAAN_01225 showing a comparatively favourable binding free energy (-67.34 kcal/mol), driven by strong van der Waals interactions (-121.97 kcal/mol) and electrostatic contributions (-579.74 kcal/mol), exhibited extensive fluctuations, conformational heterogeneity, and reduced stability throughout the MD trajectory, suggesting that the complex fails to maintain a stable conformation under dynamic conditions. Overall, the combined MD and MM/GBSA analyses ranked PACFBAAN_02021, PACFBAAN_00422 and PACFBAAN_00429 as the stable adhesin-fibronectin complexes, whereas PACFBAAN_01232 showed intermediate stability and PACFBAAN_01225 exhibited pronounced structural instability. Protein-protein interaction analysis and visualization A comprehensive interface analysis of the five adhesin-bovine fibronectin complexes revealed distinct interaction architectures that reflect their relative structural stability. PACFBAAN_02021 formed an interface spanning ~ 1031 − 978 Ų involving 18 − 13 residues (Chains A/B: receptor/adhesin). The interface was stabilized by four hydrogen bonds (GLU136-ARG334, GLU136-LYS344, GLU164-ARG334 and GLN184-ASN346), and four salt bridges, (GLU136-ARG334, GLU136-LYS344, GLU164-ARG334 and GLU164-LYS344) (Fig. 6A). PACFBAAN_00422 showed the most extensive interaction network, with an interface of ~ 1112 − 1078 Ų involving 21 − 19 residues, supported by nine hydrogen bonds (ARG25-PRO54, ARG25-ASP57, ARG25-PRO54, GLY104-ASN122, GLU164-ARG128, GLU164-ARG128, GLU164-GLN131, GLN184-GLN131 and GLN185-GLN131) and three salt bridges (ARG25-ASP57, ASP106-ARG130 and GLU164-ARG128) (Fig. 6B). Fig. 6Molecular interaction of putative adhesins and bovine FnIII 9–10 complex (cyan). Panel A: PACFBAAN_02021 (orange). Panel B: PACFBAAN_00422 (green). Panel C: PACFBAAN_00429 (pink). Panel D: PACFBAAN_01225 (yellow). Panel E: PACFBAAN_01232 (dark blue). Note: Right panels present two-dimensional interaction diagrams generated using LigPlot+, indicating hydrogen bonds and salt bridges Molecular interaction of putative adhesins and bovine FnIII 9–10 complex (cyan). Panel A: PACFBAAN_02021 (orange). Panel B: PACFBAAN_00422 (green). Panel C: PACFBAAN_00429 (pink). Panel D: PACFBAAN_01225 (yellow). Panel E: PACFBAAN_01232 (dark blue). Note: Right panels present two-dimensional interaction diagrams generated using LigPlot+, indicating hydrogen bonds and salt bridges PACFBAAN_00429, exhibited a compact interface spanning ~ 808–871 Ų with 18 − 14 residues, forming eleven hydrogen bonds (ASN18-ASN103, VAL19-ASN103, VAL19-HIS104, GLN95-GLN107, ARG130-ASP54, ARG146-ASP54, LYS175-ILE53, LYS175-ASP54, LYS175-ASP74, GLU176-ARG77 and GLN184-THR96) and five salt bridges (ARG130-ASP54, ARG146-ASP54, LYS175-ASP74, GLU176-ARG77 and GLU176-HIS104), consistent with its strong structural stability observed in MD analyses (Fig. 6C). PACFBAAN_01232 formed an interface spanning ~ 1081–1140 Ų, involving 24 − 21 residues, stabilized by eight hydrogen bonds (ASP51–ARG215, ASP61–ASN265, ALA63-ASN265, ASN68-ARG215, ASN68-ARG215, GLY104-GLN186, ILE105-GLN186 and ALA112-MET1) and a single salt bridge (ASP51-ARG215), reflecting an intermediate level of interfacial stability with MD results (Fig. 6E). In contrast, PACFBAAN_01225 presented an interface of ~ 1111–1208 Ų involving 27 − 19 residues, forming ten hydrogen bonds (LYS47-GLU93, ASN48-GLU93, GLU74-PHE272, ARG94-ASP273, LYS96-HIS269, THR125-LYS115, THR125-GLU268, THR127-HIS269, THR127-HIS269, GLU177-ARG497) and three salt bridges (LYS47-GLU93, ARG94-ASP273 and GLU177-ARG497) (Fig. 6D). Despite its sizable interface, PACFBAAN_01225 exhibited pronounced conformational fluctuations and poor retention of interfacial contacts throughout the MD trajectory, indicating limited dynamic stability. Discussion Haemorrhagic septicaemia (HS), caused predominantly by P. multocida serogroup B:2, remains a major bacterial disease of bovines in tropical regions, leading to substantial economic losses due to high mortality [8, 40, 41]. Although P. multocida is a genetically diverse pathogen capable of infecting multiple host species, HS-causing serogroup B:2 strains exhibit a striking host restriction to bovines [42, 43]. Adhesion to host tissues represents a critical early determinant of host specificity and pathogenesis; however, adhesion mechanisms in P. multocida and other members of the family Pasteurellaceae remain inadequately characterized [12]. Earlier studies have demonstrated adherence of P. multocida to porcine respiratory epithelial cells and mucus, primarily mediated by outer membrane components [44]. Subsequent ligand blotting analyses have also suggested the presence of several putative fibronectin-binding outer membrane proteins in P. multocida [10]. Nevertheless, the specific adhesins mediating, lineage specific interaction and their contribution to the bovine-restricted nature have not been defined. In this context, the present study was undertaken to investigate the molecular basis of host specificity in serogroup B strains by integrating comparative pangenomics with structural and molecular interaction analyses to identify lineage-specific adhesins that preferentially bind bovine fibronectin. Although a substantial number of P. multocida genome assemblies were available in public databases, only a subset of these genomes met the stringent quality and metadata criteria required for robust comparative genomic analysis. Many publicly available assemblies lacked reliable serogroup classification or essential metadata such as host origin, disease association, and geographic source. Inclusion of such genomes in presence-absence-based pangenome analyses can introduce substantial noise due to annotation inconsistencies. As several genomes included in this study were draft assemblies, potential biases associated with gene fragmentation and misannotation were minimized by using the Panaroo pangenome pipeline. Panaroo employs a graph-based pangenome construction approach that corrects common assembly and annotation artefacts by integrating gene neighbourhood context and sequence similarity to refine ortholog clustering. This framework can remove spurious gene calls and split incorrectly clustered paralogs, thereby reducing false gene presence-absence signals that may arise from fragmented assemblies. In addition, Panaroo applies stringent filtering to remove low-confidence gene clusters and assembly-induced artefacts, improving the accuracy of accessory genome identification [20]. Genome-based studies of HS associated P. multocida serogroup B strains have shown the presence of distinct accessory gene repertoires in HS-causing lineages. Comparative analyses of ten HS isolates and two vaccine strains, evaluated against non-HS reference genomes (Pm70, 3480, 36950, and HN06), identified a conserved core genome of approximately 1,824 genes together with discrete HS-associated gene sets, including 96 genes conserved across all HS strains and an additional 59 genes restricted to Asian serogroup B:2 lineages [45]. Broader pan-genomic investigations across P. multocida strains from diverse hosts further revealed 1,602 core genes and extensive genomic plasticity, with large dispensable and strain-specific gene fractions enriched in virulence-associated loci, including surface polysaccharide biosynthesis and adhesion-related genes [16]. Similarly, open pan-genome architectures and disease-associated phylogenomic clustering based on gene presence-absence patterns have been reported across strains causing distinct clinical manifestations like pneumonia, fowl cholera, HS and snuffles [17]. Extending this genomic framework, our comparative pangenomic analysis of P. multocida genomes representing serogroups A and B delineated a conserved core genome of 1,816 genes and a highly diversified accessory genome comprising 2,087 genes. Notably, the exclusive enrichment of 115 genes in serogroup B strains, and their complete absence in serogroup A strains, provides strong genomic evidence for lineage-specific adaptation. Given that serogroup B strains are predominantly bovine-associated, the inclusion of bovine-derived serogroup A strains mitigates host-serogroup confounding. Accordingly, the identified genes are interpreted as serogroup B-associated candidates requiring validation in larger and more diverse genome collections. Adhesion-related proteins are widely recognized as key determinants of host specificity and early colonization across different P. multocida serogroups [8]. Adhesin-profiling and comparative GWAS studies in host-restricted bacterial pathogens have demonstrated associations of distinct host niches with differential adhesion strategies. In Shiga toxin-producing Escherichia coli (STEC), cattle-associated lineages are characterized by selective enrichment of a limited subset of surface-exposed adhesins (ehaA, stgABC, yadLMN, and iha) that support persistent colonization, whereas human-associated strains exhibit broader adhesin repertoires (eae, cah, ypjA, and paa), reflecting selective pressures linked to invasion and immune evasion [46]. Consistent with this paradigm, functional prioritization of serogroup B-specific genes in P. multocida revealed an adhesin repertoire comprising five candidate adhesins, four (PACFBAAN_02021, PACFBAAN_00422, PACFBAAN_01225 and PACFBAAN_01232) of which were predicted to be surface-associated based on PSORTb and DeepTMHMM analyses, while one candidate (PACFBAAN_00429) showed no definitive localization prediction but was retained due to the presence of Ig-like/Cna_B adhesion-related domains. The convergence of these adhesins supports a model in which bovine tropism in P. multocida is mediated by constrained, lineage-specific adhesion strategies rather than broad adhesin redundancy. Selective binding of host fibronectin by bacterial adhesins is a well-established mechanism driving tissue tropism, systemic dissemination and disease severity in pathogens that exhibit strong host restriction [47]. While the severe systemic manifestations of haemorrhagic septicaemia are primarily mediated by capsule-associated virulence and endotoxin activity, successful colonization of host epithelial surfaces represents the critical initial step in disease establishment. Previous studies have demonstrated that outer membrane proteins, fimbriae, and fibronectin-binding adhesins function as key virulence determinants mediating initial host-pathogen interaction and influence host tropism in P. multocida infections [8, 48]. Early biochemical and proteomic studies, followed by phage display based functional genomics, identified PM1665 as a novel fibronectin-binding adhesin that targets the integrin-binding FnIII 9–10 domains, establishing fibronectin as a validated host receptor in P. multocida adhesion [10, 12]. Importantly, the five candidate adhesins identified in the present study show no significant sequence homology to PM1665, indicating that they represent structurally distinct adhesin candidates rather than variants of the previously described protein. Guided by this precedent, we prioritized the five serogroup B specific adhesins as the most plausible mediators of host-ECM interaction and evaluated their ability to bind fibronectin. Docking and interface analyses demonstrated a pronounced species-specific interaction pattern, with all five adhesins binding to bovine fibronectin FnIII 9–10 domains than to fibronectin from non-bovine hosts, supported by favourable docking scores, expanded interfacial contacts and consistently stronger binding energetics (-10.1 to -14.6 kcal/mol). This observation suggests the possibility of convergent targeting of a functionally important fibronectin domain by structurally distinct adhesins. Molecular dynamics simulations further validated these observations by resolving the dynamic stability of adhesin-fibronectin complexes, an approach widely used to distinguish transient docking poses from functionally stable interactions [49]. Among the five candidates, PACFBAAN_02021, PACFBAAN_00422 and PACFBAAN_00429 formed stable complexes with bovine fibronectin, whereas PACFBAAN_01232 exhibited intermediate stability and PACFBAAN_01225 showed pronounced conformational instability, defining a clear hierarchy of adhesin performance. The predicted selective binding of bovine fibronectin by this subset of serogroup B adhesins likely facilitates efficient attachment to bovine respiratory and endothelial tissues, promoting rapid systemic dissemination and contributing to the severe, fast-progressing nature of HS. This convergence on functionally dominant adhesins parallels observations in Mycobacterium tuberculosis, where integrative computational screening similarly refined broad genomic candidates to a small set of extracellular matrix-binding adhesins with validated host interactions [50]. Such hierarchical deployment of adhesins represents a common evolutionary strategy among host-restricted pathogens, enabling robust colonization under variable host environments and immune pressures [46]. Collectively, the identified adhesins define a coherent molecular signature that likely underpins the bovine-restricted ecology of P. multocida serogroup B strains. Such constrained, lineage-specific adhesion strategies are increasingly recognized as a hallmark of host-restricted pathogens and may explain the paradoxical combination of high virulence and narrow host range observed in HS, in contrast to the broader host adaptability of other P. multocida lineages [12, 47]. The serogroup-specific distribution, predicted surface accessibility, and computationally inferred receptor-binding properties of these adhesins suggest that they may represent a molecular signature associated with the bovine host tropism of P. multocida serogroup B strains. These findings highlight candidate adhesins that require future experimental validation for potential applications in diagnostics, anti-adhesion therapeutics, and recombinant subunit vaccine development aimed at preventing early host colonization [51, 52]. Despite the strength and internal consistency of the integrative computational framework, the absence of experimental validation represents a limitation of this study. The present study focused specifically on fibronectin (FnIII 9–10 domains) as a representative extracellular matrix receptor. While fibronectin is a well-established adhesion target for many bacterial pathogens, it is likely that additional extracellular matrix components may also contribute to pathogenesis. Although molecular dynamics simulations were performed using an explicit solvent model, the system represents a simplified approximation of the in vivo environment, lacking factors such as pH fluctuations, membrane interactions, and competing host proteins. Therefore, the MD simulations reflect interaction stability within a controlled computational framework rather than the full physiological environment. Accordingly, in vitro and in vivo approaches including fibronectin-binding assays, site-directed mutagenesis, surface plasmon resonance (SPR), and bovine infection models will be essential to confirm the functional relevance of these adhesin-fibronectin interactions. Nonetheless, this study provides computational insights into the molecular basis of host-restricted adhesion in P. multocida serogroup B strains and establishes a framework for future experimental studies aimed at developing targeted diagnostics and intervention strategies against HS. Conclusion This study presents an integrated comparative genomic and computational framework to identify candidate adhesins associated with bovine-adapted P. multocida serogroup B strains. By combining comparative pangenomics with multi-layer adhesin prediction, adhesin-fibronectin docking, interface energetics, and molecular dynamics simulations, we identified five lineage-specific adhesins uniquely present in serogroup B genomes that show preferential binding to bovine fibronectin. Among these, three putative adhesins (PACFBAAN_02021, PACFBAAN_00422 and PACFBAAN_00429) demonstrated relatively greater structural stability, well-defined energy landscapes, and favourable binding free energies, suggesting their potential to mediate early host-pathogen interactions in bovines. Collectively, these findings provide computational insights into the molecular basis of host specificity in serogroup B strains. However, this study is entirely based on in silico analyses, the identified adhesins should be considered as candidate molecules that serve as starting points for downstream experimental validation. Future studies employing biochemical binding assays and host-cell adhesion experiments will be necessary to confirm the predicted adhesin-fibronectin interactions and to elucidate their role in host-specific colonization. Supplementary Information Supplementary Material 1. Supplementary Material 1. Supplementary Material 2. Supplementary Material 2. Supplementary Material 3. Supplementary Material 3. Supplementary Material 4. Supplementary Material 4.