https://orcid.org/0000-0002-8389-6345
Figures
Abstract
Mannheimia haemolytica is an opportunistic bacterial pathogen associated with the economically costly bovine respiratory disease. Two genotypes have been described, of which genotype 2 is more strongly associated with disease. Several previous studies have investigated the genomic differences between the genotypes and/or the major serotypes (1, 2 and 6) of M. haemolytica, however we still lack a clear basis for the greater disease association of genotype 2 (serotypes 1 and 6) and demonstrations of phenotypic differences are scarce. This work builds upon previous investigations to identify genes that differentiate the two genotypes with a particular focus on genes that may play a role in virulence and fitness in the respiratory tract microbiome. We identified 422 genotype differentiating genes in a collection of 206 unique M. haemolytica genomes (61 genotype 1, 145 genotype 2). Genotype differentiating genes included genotype-associated variants of a TonB-dependent siderophore receptor homolog, transferrin binding protein B, leukotoxin A, and IgA1 proteases. We also identified a genotype 1 associated lytic transglycosylase, and a genotype 2 specific highly immunogenic outer membrane lipoprotein. Genotype 2 genomes were significantly larger in size and contained more predicted protein coding genes than genotype 1 genomes. These results expand our knowledge of what differentiates the genotypes 1 and 2 of M. haemolytica and provides information that can be used as the basis for laboratory investigations of corresponding phenotypic differences.
Citation: Deschner D, Hill JE (2025) Identification of genes that differentiate Mannheimia haemolytica genotypes 1 and 2 using a pangenome approach. PLoS One 20(10): e0325338. https://doi.org/10.1371/journal.pone.0325338
Editor: Brenda A. Wilson, University of Illinois Urbana-Champaign, UNITED STATES OF AMERICA
Received: May 10, 2025; Accepted: September 29, 2025; Published: October 16, 2025
Copyright: © 2025 Deschner, Hill. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All relevant data are within the paper and its Supporting Information files, and in S1 Dataset (FigShare repository DOI:10.6084/m9.figshare.28872536).
Funding: The author(s) received no specific funding for this work.
Competing interests: The authors have declared that no competing interests exist.
Introduction
Mannheimia haemolytica is a Gram-negative opportunist bacterial pathogen associated with bovine respiratory disease (BRD) [1,2]. The beef industry is estimated to lose more than $3 billion per year globally due to BRD [3,4]. Two genotypes of M. haemolytica have been described based on whole genome sequencing and single-nucleotide polymorphism (SNP) analysis and multiple serotypes have also been defined [5,6]. The three most common serotypes (1, 2 &, 6) have been found to correspond to genotypes with serotypes 1 & 6 corresponding to genotype 2 (G2) and serotype 2 corresponding to genotype 1 (G1) [1,5,7,8]. Genotype 2 has been more frequently associated with clinical cases of BRD, however the genetic and phenotypic basis for this apparent association is unclear [1,9–14]. In addition to virulence factors such as leukotoxin, lipopolysaccharide, adhesins and proteases [11,15–17], adaptations that confer increased fitness, such as greater ability to acquire iron and other nutrients, could contribute to disease association by allowing a pathogenic strain to outcompete non-pathogenic strain(s) present in the same environment [18].
Reports of specific phenotypic differences between M. haemolytica G1 and G2 are scarce. G2 strains corresponding to serotype 1 have been demonstrated to invade differentiated bovine bronchial epithelial cells in vitro while G1 strains (serotype 2) could not [19]. The only other phenotypic differences described to date are distinct matrix-assisted laser desorption/ionization – time of flight mass spectrometry (MALDI-TOF) profiles and colony morphology [20,21].
Genotypic differences between G1 and G2 have been investigated previously using whole genome sequences. A pangenome analysis of 69 M. haemolytica genomes identified 112 G1 specific, and 179 G2 specific genes. This study was focused on outer membrane protein encoding genes and identified seven G2 specific predicted outer membrane protein encoding genes that were hypothesized to play a role in the greater disease association of G2 [22]. Another study examined 11 M. haemolytica genomes for potential differences in virulence factors between the serotypes 1, 2 and 6 [23]. The authors reported more prophages in serotypes 1 and 6 than in serotype 2, which could influence virulence due to the presence of virulence factors in these prophages. Self-targeting CRISPR spacer sequences were detected in serotypes 1 and 6, which may enhance immune evasion or adhesion during infection by modulating sialic acid expression on the cells surface. ARG-carrying ICEs were more prevalent in serotypes 1 and 6 than serotype 2, which could enhance survival of antibiotic exposure [23].
The objective of our current study was to take advantage of the greater number of M. haemolytica whole genome sequences now available to identify genetic differences between the two genotypes. In our examination of >200 M. haemolytica genomes, we gave particular attention to genes encoding proteins that may contribute to the reported differences in disease association such as those involved in virulence and/or fitness in the bovine respiratory tract. We designed an approach that allowed the detection of both genes with genotype associated presence/absence patterns and sequence variants of genes conserved in both genotypes. This contrasts with previous studies that have largely focused on the presence/absence of genes within the genotypes. Our results showed that G2 genomes are significantly larger in size and contained more predicted protein coding genes than genotype 1 genomes. We identified a total of 422 genotype differentiating genes, including several whose predicted functions may partially explain the differences in observed disease association and that have not to our knowledge been reported previously. This list of genes will inform future research to demonstrate phenotypic differences between G1 and G2.
Materials & methods
Genome sequences
M. haemolytica whole genome sequences defined as complete by NCBI were downloaded from the NCBI Genome database on March 25, 2024, using the NCBI Datasets command line tools [24]. NCBI defines “complete” as all chromosomes are gapless and contain runs of nine or less ambiguous bases, no unplaced or unlocalized scaffolds, and all expected chromosomes present. To avoid duplication all.gca files that had a matching.gcf file were removed. To identify G1 and G2 genomes, the sequences were genotyped using an in silico PCR assay in Unipro UGENe v45.1 and primers from a genotype specific LAMP assay with default parameters (Mismatches = 3, Min Perfect Match = 15, Max Product Size = 5000, Use ambiguous bases = True, Extract annotations = Inner, and Temperature settings = Primer 3) [25,26]. Primers from the LAMP assay included a pair targeting G1 specific adhesin pseudogene B1, G2 adhesin G (CP017538.1 locus tag: BG586_06285; Forward: 5’-GCA CAT TAA AAT ATA GCA GCT TTG-3’; Reverse: 5’-CTA AGC CAG AGT GAT CCG-3’), and leukotoxin D (lktD) (CP005972.1 locus tag: F382_07400; Forward: 5’-ACT GAA AA TTC ACA CTA TAG GTG-3’; Reverse: 5’-GCT AAT ATG TTT AAT TCG ACC AGT T-3’). The lktD primers were used to confirm identity of genomes as M. haemolytica. In initial testing G1 specific primers failed to produce any predicted products in our G1 reference strain and so were discarded. New G1 specific primers targeting the G1 specific adhesin pseudogene B1 were developed using Primer 3 v 4.1.0 (CP017495.1 locus tag BG548_01640; Forward: 5’-ACT AAT CTG AAG AGC GGC GT-3’; Reverse: 5’-TTG GTG GTT GCT GCT TGT AC-3’) [27–29]. G1 and G2 NCBI genomes were combined with a set of G1 and G2 genome sequences previously generated by our lab (BioProject: PRJNA1088094)(30) and reference genomes (G1 = CP017518; G2 = CP017519) for both genotypes.
A phylogenetic tree was constructed using MASH with the ani command in PanTools to verify genotype designations [30].
Pangenome construction, annotation, and classification
A pangenome was constructed using PanTools v 4.3.2 with FASTA files, gff3 annotation files, and genotype designation as inputs [30–32]. Genomes were assigned a new unique identifier based on their input order during pangenome construction (Table A in S1 File). Genome metrics such as genome size, GC content, number of genes, etc. were determined using the metrics command.
PanTools groups protein sequences from the input genomes into homology groups based on their similarity. Homology groups also contain the nucleotide sequences encoding the constituent protein sequence(s). Homology groups were generated using the group command with the relaxation parameter set to 2. The relaxation parameter is a combination of four sub-parameters (intersection rate, similarity threshold, mcl inflation, and contrast) where a lower number represents a stricter set of thresholds. We elected to utilize a relaxation parameter of 2 to allow for detection of minor variations that might be expected when examining sub-populations of the same species and avoid grouping of more divergent genes as might occur with a looser parameter.
Homology groups were classified by PanTools as core (present in ≥ 90% of all genomes), unique (present in only one single genome), or accessory (present in >1 genome and <90% of all genomes). Based on their prevalence within each genotype, homology groups were classified as either shared (present in ≥ 90% genomes of one genotype, regardless of their prevalence in the other genotype), exclusive (solely present in genomes of one genotype), or specific (present in ≥ 90% genomes of one genotype, and no genomes of the other genotype) [30]. To capture homology groups strongly associated with one genotype that did not meet the strict PanTools definition of specific we used the classification “genotype associated” for homology groups that were present in ≥ 95% of one genotype while also being present in > 0% and ≤ 5% of the other genotype.
Genomes were functionally annotated using InterProScan v 5.64–96.0 with SignalP v 4.1 database integrated [33–35]. Biosynthetic gene cluster annotations were generated using antiSMASH 7.1.0 with default parameters [36]. InterPro and antiSMASH annotations were added to the pangenome, with InterProScan annotations including InterPro, Pfam, GO, and SignalP annotations [30].
Multiple sequence alignments of constituent nucleotide sequences were generated for all homology groups in the pangenome with MAFFT and consensus sequences from the alignments were generated to represent the genotype specific and associated homology groups [30,37]. Consensus sequences were used as queries for BLASTx searches using the bacterial/archaeal translation code against the NCBI refseq_select database [38–42]. All other parameters were kept as default. Searches were performed March 20–25, 2025. The top hit for each search was manually recorded to provide further evidence of the predicted function, facilitate human interpretation, and identify functions that may not have been annotated by any of the other approaches. The entire pangenome database is available in [37].
The nucleotide sequences of several previously characterized genes upregulated in response to iron limitation, and previously identified proteases were used as BLAST queries against the pangenome using the blast command in PanTools to determine if these previously characterized genes were present in our dataset [15,30,43]. Both alignment threshold and minimum identity were set to 90. Query nucleotide sequences are available in Table B in S1 File.
Pangenome structure analysis
The structure of the pangenome was predicted using the pangenome_structure command in PanTools. Heap’s law α value was generated using the heaps_law.R script with α < 1 indicating an open pangenome (i.e. adding additional genomes will consistently add new homology groups to the pangenome) and α > 1 indicating a closed pangenome (i.e. adding additional genomes will not add new homology groups to the pangenome) [44].
Tet(H) gene abundance
Tet(H) genes were used as a proxy for the presence of integrative and conjugative elements (ICE) due to the strong correlation between presence of tet(H) and presence of ICE [12,45]. The reference tet(H) nucleotide sequence (Accession Number: Y15510.1 (nucleotides 2099–3301)) was downloaded from the comprehensive antibiotic resistance database (CARD) on July 26, 2024, and used as nucleotide BLAST query against the pangenome database using the inbuilt PanTools blast command. Minimum identity and alignment threshold were both set to 90%.
Leukotoxin operons
Reference nucleotide sequences for each of the component genes of the M. haemolytica leukotoxin operon (lktCABD) were downloaded from GenBank (AF314508.2) and used as queries against the pangenome BLAST database with percent identity and alignment threshold both set to 90%. Consensus nucleotide sequences of G1 and G2 versions of all leukotoxin operon genes were extracted and translated in Geneious 2024.0.5. Predicted protein sequences for the G1 and G2 versions of all leukotoxin operon genes were aligned against one another in Geneious 2024.0.5.
Annotation of carbohydrate active enzymes
Carbohydrate active enzymes (CAZymes) were annotated using a local installation of dbCAN v 4.1.1 software with integrated SignalP v 4.1 database and DIAMOND, HMMER, and HMMER dbcan_sub prediction algorithms [46]. Annotation was performed on each genome sequence individually using a custom BASH script [37]. dbCAN annotations that had supporting annotations from at least three tools and contained a predicted signal peptide were selected for further analysis.
Protein structure prediction and visualization
Protein sequences were submitted to the AlphaFold Server for structural prediction [47]. Models with the highest ranking score were downloaded and visualized in PyMol version 2.5.2 (Schrödinger, LLC).
Statistics
Welch’s t-test for unpaired data was used to compare G1 and G2 genome characteristics with P ≤ 0.05 considered significant and 95% confidence intervals for the difference in means were reported. A one-way ANOVA (Kruskal-Wallis) with Dunn’s test for multiple comparisons was used to determine if there was a difference in genome size between G1 and G2 with or without ICE. P ≤ 0.05 was considered significant. These analyses were performed in GraphPad PRISM v 10.4.1.
A Fisher’s exact test and odds ratio comparison was performed to determine if the prevalence of tet(H) was different between genotypes. Both tests were performed in Microsoft Excel for Microsoft 365 using Real Statistics Resource Pack. P ≤ 0.05 was considered significant.
Results & discussion
Description of genome collection
M. haemolytica genomes designated as complete were retrieved from NCBI and after removal of duplicates, 121 genomes were available for further analysis. In silico PCR with modified primers from a previously published genotyping assay resulted in the expected amplicons for 108/121 genomes. No in silico genotyping PCR products were obtained for 13 genomes and so they were removed from further analyses leaving 108 genomes.
Three of the 13 isolates that could not be genotyped also failed to produce any predicted product for the lktD in silico PCR. Two of these genomes have since been removed by the RefSeq curation team for either containing too many frameshift mutations (GCF_900474405) or failing to pass a completeness check (GCF_023516295). All three genome sequences that failed to produce an lktD product were from sheep isolates (GCF_900638155/NCTC 10643, GCF_900474405/NCTC 10609) or from culture collections from the 1960s (GCF_023516295/ATCC 29701). Of the other ten genomes that were not genotyped by in silico PCR, four were from a German study examining macrolide resistance in M. haemolytica (GCF_030370195, GCF_030370215, GCF_030370355, and GCF_030370375) [45], and three were from sheep isolates collected in 1951 or 1968 and sequenced as part of the United States Department of Agriculture (USDA) M. haemolytica sequencing project (GCF_023516315, GCF_023516375, GCF_023516395 in BioProject: PRJNA172854). The last three isolates were all generated from non-nasopharyngeal cattle samples as part of either the USDA sequencing project (GCF_000819525, GCF_023516255) or a German study (GCF_020971705) and had a CheckM completeness <92.5%.
The in silico PCR assay we used was selected for identification of G1 and G2 genomes as defined by Clawson et al. [5]. The original LAMP assay from which we selected primers for this process was developed using M. haemolytica isolates from North American cattle [25]. Thus, the most likely explanation for genomes to be un-typable by this approach (besides sequence quality issues) is that they represent genotypes other than G1 or G2.
The NCBI genomes were combined with 99 genomes from our previous work (BioProject Accession PRJNA1088094) and reference genomes for G1 and G2 as described in the Methods. One additional genome was removed following discovery that it was likely a chimera, leaving a final dataset 208 M. haemolytica whole genomes with 62 G1 and 146 G2 representatives (Table A in S1 File). The reference genomes for G1 and G2 were also represented in the collection of genomes downloaded from NCBI resulting in the collection representing 206 unique M. haemolytica genomes. Most genomes were from isolates originating from feedlot cattle in either Canada or the United States between 2003 and 2021. Consistent with the in silico PCR identifications, a pangenome derived phylogenetic tree showed two distinct clades corresponding to isolates classified as G1 or G2 (S1 Fig).
Pangenome structure
The structure of the pangenome was analyzed and the predicted Heap’s law α was 0.9020 (pangenome_structure seed = 1755274650463). An α value ≤ 1 indicates that the pangenome of M. haemolytica is open, i.e. the addition of new genomes to the pangenome will result in the addition of new homology groups. Open pangenomes are generally expected for bacteria due to high frequency of horizontal gene transfer and other genetic recombination events [48,49].
The publication of 42 new complete and annotated M. haemolytica whole genome sequences between our original data collection on March 26, 2024, and initial peer review of this study (September 2, 2025) provided us with an opportunity to examine the impacts of adding new genomes to the pangenome. All 42 of the new genome sequences were genotyped using the same in silico PCR method resulting in the addition of 21 G2 and 3 G1 genomes (18/42 were untypable). The addition of these 24 genomes resulted in reduction of G2 specific and associated homology groups (from 266 to 238) and a reduction of G1 specific and associated homology groups (from 156 to 137). The small reduction in the number of specific and associated homology groups was expected since the inclusion of new genomes increases the number of genomes a homology group must be present in, or absent from, to be classified as genotype specific or associated. We also determined the Heap’s law α value for the expanded pangenome to be 0.9011 (pangenome_structure seed = 1757648129495, [37]). The small reduction in genotype specific and associated homology groups and similarity of the α values of the original and expanded pangenomes supports the appropriateness of our initial pangenome to represent the genetic diversity of the established M. haemolytica genotypes and be useful for identification of genotype differentiating genes.
Genome properties & Tet(H) abundance
The two genotypes were compared to one another on a broad set of genome metrics, such as genome size, GC content, etc., to determine if there were any high-level differences in the genomic structure and content of the two genotypes. G2 genomes were larger, had more predicted protein-coding sequences, and had slightly lower GC content than G1 genomes. No differences were identified in the number of tRNA or rRNA genes or the proportion of the genome covered by genes (Table 1).
We hypothesized that the larger size of G2 genomes was attributable to a greater prevalence of integrative and conjugative elements (ICEs), large (~20 kb to >500 kb) mobile genetic elements that are known to be associated with antimicrobial resistance and are more frequently identified in G2 genomes [5,50]. To test this hypothesis, we used the proxy measure of the presence of tetracycline resistance gene tet(H) to determine the prevalence of ICEs due to the strong association of tet(H) with ICEs [12,45].
Tet(H) was identified significantly more frequently in G2 (67/146, 45.9%) than G1 (1/62, 1.6%) (Fisher’s p-value = 4.39 x 10-12 Odds Ratio: 0.02 ± 1.02), however, G2 genomes were larger than G1 regardless of the presence of ICE (Fig 1). The larger size and higher gene count of G2 genomes regardless of presence of ICE (Table 1) could indicate that G2 genomes carry a larger repertoire of genes that may allow them to be more adaptable to a wider variety of environments than the smaller G1 genomes. Considering average bacterial gene sizes, the difference in the median genome size of G2 without ICE and G1 (~100 kb) could represent dozens of G2-specific genes. Additionally, the greater prevalence of ICEs in G2 suggests that this more disease associated genotype is also more likely to possess antimicrobial resistance genes.
One way ANOVA (Kruskal-Wallis) followed by Dunn’s multiple comparisons test (corrected P values are shown). Red bars indicate mean.
Pangenome analysis
The pangenome calculation resulted in the identification of “homology groups”: clusters of sequences from input genomes that exceed a given similarity threshold. We then examined the distribution of homology groups across all M. haemolytica genomes to identify those that are core, unique, accessory, shared, genotype specific or genotype associated as defined in the methods. A total of 4,038 homology groups were generated of which 2,007 (49.7%) were core, 1,473 (36.5%) were accessory, and 558 (13.8%) were unique (Pangenome_Classification in Table C in S1 File).
For the current study, our interest was homology groups that differentiate the genotypes: those that are found in the large majority of representatives of one genotype and in none, or a small minority, of the other genotype. PanTools identified 135 G1 specific and 233 G2 specific homology groups. The PanTools specific category definition is strict, however, requiring a homology group to be present in >90% of one genotype and zero of the other. To identify homology groups that do not meet this strict criterion but are nevertheless strongly associated with one genotype, we examined accessory groups not initially classified as specific for those that were present in ≥ 95% of the one genotype and > 0% and ≤ 5% of the other genotype. These were classified as G1 associated (n = 21) or G2 associated (n = 34) (Tables C and D in S1 File). Identification of genotype specific and associated genes within this relatively large genome collection reinforces previous findings that the two genotypes carry distinct sets of genes [22]. The pangenome results also provide evidence of the internal coherence of the genotypes, since when considered separately, G1 and G2 genomes included 2,169 and 2,342 “genotype-core” homology groups (present in >90% of genomes of the genotype), respectively.
To provide an overview of the potential phenotypic differences between the two genotypes, all specific and associated homology groups were assigned to one of nine general functional categories based on functional annotations and the results of manual BLASTx searches against the NCBI refseq_select database (Fig 2, Tables D and E in S1 File). For both genotypes, the largest number of specific and associated homology groups were classified as MGE proteins (transposases, toxin-antitoxin plasmid maintenance systems, prophage associated proteins, etc.) or hypothetical. The larger number of G2 specific and associated MGE proteins and antimicrobial resistance genes is consistent with greater ICE prevalence in G2 [9,12,22,23,45]. The large number of genotype specific and associated hypothetical proteins in both genotypes is also unsurprising as ~22% of all domain families in Pfam v32.0 are defined as “domains of unknown function” and a recent examination of >30,000 bacterial genomes and metagenomes showed that 40–60% of genes are of unknown function [51,52]. Homology groups predicted to encode unknown/hypothetical proteins, and MGE proteins were excluded from further analysis in this study but certainly represent an opportunity for future investigation.
G1 = Genotype 1, G2 = Genotype 2. Specific groups are those found in ≥90% of the genomes of the given genotype and no genomes of the other genotype. Associated groups are present in ≥95% of the genomes of one genotype > 0% and ≤5% of the genomes of the other genotype. MGE proteins include transposases, toxin-antitoxin plasmid maintenance systems, prophage associated proteins, etc. Hypothetical proteins include those annotated as hypothetical or domain of unknown function.
Based on inspection of the complete inventory of pangenome homology groups (Tables C and D in S1 File), we identified a subset of specific/associated homology groups with functional annotations that may be associated with reported differences between disease association of G1 and G2 (Table 2). These genome features were related to iron acquisition, host immune evasion, immunogenicity and carbohydrate modification.
Iron acquisition
Iron is essential for bacterial growth and bacterial pathogens must compete with each other and with the host for the scant iron typically available in their environment [53,54]. Animal pathogens like M. haemolytica are faced with concentrations of available iron well below their needs, which has necessitated the evolution of mechanisms for acquiring iron including siderophores, transferrin binding proteins, and haem binding proteins [55]. Given the essential nature of iron, and the generally iron limited conditions of the respiratory tract, the ability to acquire iron more readily than other bacteria provides a considerable advantage. The identification of several genotype associated, and specific homology groups potentially involved in iron acquisition was of interest since these differences may be associated with fitness in the respiratory tract and thus with disease association (Table 2).
Homology groups 2042480 (G2 specific) and 2042481 (G1 associated) were both annotated as encoding a TonB-dependent siderophore receptor. TonB dependent receptors (TBDTs) are transporters often involved in iron acquisition, specifically in the uptake of small, secreted iron scavenging molecules known as siderophores [56,57]. Homology groups 2042480 and 2042481 also produced BLAST hits to a previously characterized siderophore (ferric enterobactin) receptor (frpB) homolog that was upregulated in M. haemolytica in response to iron limitation [43]. The consensus amino acid sequences of these homology groups were 84.4% identical, and both contained consensus TonB-dependent transporter (TBDT) motifs. Predicted 3D structures demonstrated the canonical 22 β-strand barrel structure with an N-terminal plug domain characteristic of TBDTs (AlphaFold ranking_score = 0.95 for both) (Fig 3). The amino acid sequence differences between the G1 and G2 sequences were concentrated in the extracellular loops responsible for substrate binding, suggesting that these two proteins may differ in their ability to bind their substrate [56,57] or they may bind different substrates (Fig 3). Identification of a putative siderophore receptor in M. haemolytica is unexpected as M. haemolytica is not known to produce or utilize siderophores and has been demonstrated to be unable to utilize ferric enterobactin so it is likely that the identified TBDTs have some other function besides siderophore uptake [11,43].
Predicted 3D structures of Genotype 1 (G1, Purple) & Genotype 2 (G2, Orange) FrpB homolog proteins conform to canonical TonB-dependent receptor structures. Below the 3D structure alignment is a 2D representation of tertiary structure of G1 and G2 frpB homologs. Each grey rectangle represents one of the 22 β-strand sheets, and the size of the extracellular/periplasmic loops are shown. Differences in the amino acid sequence for select extracellular loops are shown in further detail. Difference in extracellular loops may contribute to differences in substrate binding efficiency and/or result in interactions with completely different substrates. G1 AlphaFold ranking_score = 0.95; G2 AlphaFold ranking_score = 0.95.
G2 specific homology group 2041941 was also predicted to encode a TBDT. No similar G1 specific/associated protein was identified and the consensus sequence of 2041941 did not correspond to previously identified genes upregulated in response to iron limitation [43]. The identification of another G2 specific TBDT suggests that G2 may have a greater ability to uptake iron and/or other substrates. TBDT have been implicated in transport of a variety of substrates in gram-negative bacteria including siderophores as well as vitamin B12 and carbohydrates [57]. Further examination is required to determine the function of this TBDT and whether it provides a competitive advantage to G2 over G1 strains.
Other iron acquisition proteins include transferrin binding proteins A and B (tpbAB), both of which are found in M. haemolytica [11]. Transferrin is an iron-carrying glycoprotein found in mammalian circulatory systems that transports iron throughout the body and serves as a sink for any excess free iron released into the circulatory system [58]. TpbB is a surface lipoprotein that extends outwards from the outer membrane to the extracellular space where it binds to transferrin and brings the transferrin closer to TbpA, a TBDT in the outer membrane that transports the transferrin into the periplasm [58]. To investigate any genotype associations of transferrin binding proteins, previously characterized tbpA and tbpB nucleotide sequences from M. haemolytica were used as BLAST queries against the pangenome database. Core homology group 2044546 contained hits for tbpA while tbpB hits were solely identified in G2 specific homology group 2044545 with no hits in any G1 genomes. These results suggest that tbpA is highly conserved within M. haemolytica and likely plays a key role in iron metabolism. In contrast, while no BLAST hits were produced to the query tbpB nucleotide sequence in G1 genomes, G1 specific homology group 2040773 was predicted to encode for a transferrin binding protein. The amino acid sequences of G1 and G2 variants of the putative TbpB proteins (homology groups 2040773 and 2044545) were compared and found to be 56.5% identical and 81.2% similar. Despite the relatively low similarity of these two proteins, the predicted 3D structures were very similar to one another as well as the crystal structure for the TbpB from Actinobacillus pleuropneumoniae (G1 ranking_score = 0.88, G2 ranking_score = 0.93, Fig 4). A difference in the TbpB structure between the genotypes was observed in the N lobe with G2 appearing to have a longer alpha-helix structure in the N-lobe handle region than G1 or A. pleuropneumoniae. This region of the protein contains residues important for binding of iron-loaded transferrin [59]. Genotype variation of a transferrin binding protein further suggests they may differ in their ability to acquire iron, which may partially explain observed differences in disease association between the two genotypes. The genotype variants of TbpB identified here may also reflect host preferences of the genotypes as TbpB is selective for host transferrin and differences between TbpB proteins has been partially attributed to this host selectivity [58].
(A) The crystal structure of TbpB from Actinobacillus pleuropneumoniae (PDB accession 3HOL [59]) was used as a reference. Major domains include the N-lobe and C-lobe, each containing a beta-barrel and a “handle”. The reference structure is rainbow coloured from N terminus (blue) to C terminus (red). (B) Alignment of G1 and G2 variants. (C) G2 aligned to reference. (D) G1 aligned to reference. The G1 variant has an elongated alpha-helix in the N-lobe handle domain relative to G2 and the reference. (G1 ranking_score = 0.88, G2 ranking_score = 0.93).
IgA protease
IgA proteases are metalloproteases that cleave immunoglobulin A (IgA), a critical component of the mucosal immune system. Two IgA proteases (IgA1 and IgA2) have been previously identified and characterized in a G2 (serotype 1) M. haemolytica isolate [15]. We identified a G2 specific IgA protease (homology group 2042189) and a G1 associated IgA protease (2040817) (Table 2). To determine the relationship of these sequences to the previously reported IgA proteases, a BLAST search was performed using previously reported nucleotide sequences for IgA1 and IgA2 as queries against the pangenome database (IgA1:NZ_DS264709 Locus Tag: MHA_RS00165IgA2: NZ_DS264612.1 Locus Tag: MHA_RS09765). At least one BLAST hit was returned for IgA2 in all but five of the included genomes with hits grouped into five homology groups (2041035, 2041036, 2041399, 2041400, and 2043723). Homology groups 2041305 and 2041306 were accessory, while 2041399 and 2041400 were unique suggesting some minor variations are likely present within the pangenome for this IgA2 protease. Homology group 2043723, however, was a core group suggesting that this IgA2 protease is a core gene of M. haemolytica and is not genotype differentiating. In contrast, a BLAST hit was returned for IgA1 in 146/209 of the genomes with hits grouped into one of three homology groups (2040817, 2041298, and 2042189). Homology group 2041298 was found in only one genome so was not examined any further. Group 2040817 is a G1 associated group while 2042189 is a G2 specific group and thus these groups appear to represent genotype variants of an IgA protease.
Comparison of the consensus sequences of the two IgA1-like homology groups revealed that the nucleotide sequence of 2042189 (G2 specific) is 4,512 bp consistent with previous descriptions of IgA1 in G2 M. haemolytica [15], while the sequence of 2040817 (G1 associated) was only 2,281 bp. An alignment of the two variants showed that they were 94% identical over 2,281 shared positions and that the G1 associated sequence is a truncated version of the other. The M. haemolytica IgA1 protease contains three functional domains: a large, N-terminal peptidase domain, pertactin-like passenger domain and a C-terminal autotransporter domain [15]. The product of the truncated G1 associated gene would include the peptidase domain but lack part of the pertactin-like passenger domain and the autotransporter, and thus may not encode a complete or functional IgA1 protease.
The observation of a G2 specific IgA1 sequence and possibility of genotype variation remains of interest and warrants further investigation since differences in IgA protease activity could relate to difference(s) in their ability to evade the host immune system and thus may contribute to differences in disease association.
Highly immunogenic outer member lipoprotein
Pasteurella lipoproteins (Plp) are a group of several immunogenic outer membrane lipoproteins identified in M. haemolytica [60] and PlpE has been investigated as a potential vaccine target for M. haemolytica because of its immunogenicity [61–63]. Aside from the immunogenic role of PlpE relatively little is known regarding its function although it is important as a vaccine target and thus clinically and diagnostically significant.
G2 specific homology group 2044609 was predicted to encode PlpE, however no corresponding G1 homology group was identified. Previous investigations of PlpE sequence diversity across serotypes showed that the serotype 1 and 6 (G2) sequences were highly conserved whereas serotype 2 (G1) PlpE sequences were extremely heterogenous, however, antibodies raised against serotype 1 PlpE cross reacted with serotype 2 bacteria [64]. To identify potential G1 variants of the G2 plpE, a reference plpE gene (NZ_CP023044.1, c780551-779481, locus tag: CKG22_RS03975) was used in a BLASTx search with relaxed parameters. When percent identity threshold was reduced to 70% and alignment threshold to 50% a hit to G1 specific homology group 2042820 was observed.
The identification of a sequence encoding a G2 specific PlpE and possible G1 variant suggests that the genotypes may differ in immunogenicity, however further research is needed to investigate the prevalence and immunogenicity of Plp variants across M. haemolytica genotypes.
Leukotoxin operon
Leukotoxin (Lkt) is a critical virulence factor in M. haemolytica and is produced by both genotypes (all serotypes) [65]. Lkt is encoded in an operon consisting of four genes (lktCABD). lktA encodes the protoxin, lktC encodes the acylating enzyme essential for post-translational modification and activation of the protoxin, and the products of lktB and lktD are involved in secretion of the toxin [66,67]. lktCABD genes were identified and the predicted protein sequences for the Lkt operon genes from G1 and G2 were aligned (Fig 5). The predicted proteins LktC, LktB and LktD were 95.2%, 99.6% and 100% identical, respectively, suggesting these proteins are under conservation pressure, likely due to their roles in regulating the secretion and function of Lkt [67,68]. In contrast, the predicted LktA protein sequences were only 89% identical. Differences in the predicted LktA proteins of G1 and G2 were concentrated in the N-terminal and C-terminal regions of the protein (Fig 5). In leukotoxins the N-terminus is involved in pore-forming activities while the C-terminus is involved in export and host immune system recognition [67]. As such any differences in the N-terminal and/or C-terminal regions may potentially influence the activities of the Lkt and thus may result in virulence differences between the two genotypes. Genome 9 was excluded from the LktA analysis as it represents an LktA gene knockout construct of genome 8.
Percent sequence identities were calculated with a sliding window size of 10. A larger peak indicates a greater number of matching amino acids within the window for that position. Locations of pore forming domain and export signal domain are indicated above the identity plot. Schematic below the identity plot shows identical (black) and non-identical (light grey) residues in the alignment.
Sequence diversity in LktA within M. haemolytica has recently been reported to be greater than within other species in the genus [69]. Davies et al. examined the sequence diversity of leukotoxin in 31 M. haemolytica, 6 Mannheimia glucosida, and 4 Pasteurella [Bibersteinia] trehalosi isolates identified eight M. haemolytica allelic variants of lktA [66]. Five of the alleles (A6-A10) were identified solely in ovine isolates, allele A3 was only identified in bovine isolates, while A1 and A2 alleles were identified in both ovine and bovine isolates. The A1 allele was subdivided into 5 groups (A1.1-A1.5) with A1.1 solely occurring in bovine isolates from serotype 1 (n = 3) and 6 (n = 1) isolates (corresponding to G2). In bovine isolates from serotype 2 (corresponding to G1), the authors reported detection of alleles A2, A2.1, A2.2, and A3. LktA sequences from genomes in our current study were compared with representative amino acid sequences for each allele and subgroup downloaded from NCBI Genbank. Within G2, 141/146 LktA sequences were 100% identical to the A1.1 allele and another 4/146 were >99.5% identical to A1.1, consistent with the observations of Davies et al. (Tables F-H in S1 File). The only exception was the LktA from Genome 2, that shared 100% amino acid identity with A1.3 and A1.5 (Table I in S1 File). Interestingly, all but one of the G1 LktA sequences were 99.5–100% identical to the A8.1 allele (Genome 157 was identical to A2) (Tables F, I and J in S1 File). The A8.1 allele was found in serotype 2 (G1) isolates by Davies et al., however, all the isolates in that study were ovine (n = 3), and A8.1 was also detected in ovine serotype 7 isolates (n = 3). The identification of A8.1 across many bovine origin genomes and nearly complete absence of A2 alleles suggests that G1 from Canada and the U.S.A has diverged from the G1 found in the United Kingdom where most of the isolates used by Davies et al. were collected and further demonstrates that A8.1 is not restricted to ovine isolates [66]. Whether there is a relationship between specific LktA alleles and respiratory disease severity in cattle warrants further investigation.
CAZymes
Carbohydrate active enzymes (CAZymes) are enzymes whose function has been biochemically verified to be involved in the metabolism of carbohydrates. We chose to investigate any differences in CAZymes between the two genotypes since these may indicate differences in ability to metabolize certain carbohydrates, and thus in the nutritional niches of the genotypes [70]. We were particularly interested in enzymes predicted to be secreted into either the periplasmic or extracellular space (i.e., contain a signal peptide) and are thus most likely to be involved in interactions with the external environment. A total of six CAZymes containing signal peptides and that were annotated by a minimum of three of the annotation tools were identified in the pangenome (Table K in S1 File) (50). At least one CAZyme was predicted in all genomes except for Genome 86/GCF_022404935, which was removed from this analysis. CAZymes identified were members of families GH23 (lytic transglycosylase), GH33 (sialidase), GH13_19 (amylose glucosidase), GT9 (glycosyltransferase), GH102 (lytic transglycosylase) and GH103 (lytic transglycosylase). Only GH103 (homology group 2042729) had a difference in prevalence between the two genotypes, being present in 60/61 (98.4%) G1 genomes and 2/146 (1.4%) G2 genomes (Table 2). The GH103 family consists of a group of lytic transglycosylases that cleave the β-1,4 linkage between N-acetylmuramoyl and N-acetylglucosaminyl residues in peptidoglycan and belong to family 3 of the transglycosylases with structures similar to lysozyme [71,72]. Numerous biological functions have been attributed lytic transglycosylases, however the only clear role is that they are involved in cell wall remodelling [71–74]. Whether this genotype difference in lytic transglycosylases is associated with phenotypic differences in colony morphology of the two genotypes [21] or other differences in cell wall maintenance will require further investigation.
Conclusions
We identified a total of 422 Mannheimia haemolytica genotype differentiating genes (135 G1 specific, 21 G1 associated, 233 G2 specific, and 33 G2 associated), including 125 with unknown functions and 132 that are associated with mobile genetic elements. Genotype associated genes and variants that may contribute to differences in disease association include TonB-dependent transporters, IgA proteases, a highly immunogenic outer membrane protein (PlpE), a lytic transglycosylase, and genotype variants of the leukotoxin gene. The list of genotype differentiating genes provides a shortlist of targets that warrant further investigation to provide a mechanistic understanding of phenotypic differences between the M. haemolytica genotypes and inform rational strategies for M. haemolytica and bovine respiratory disease diagnostics.
Supporting information
S1 Fig. Phylogenetic tree of study genomes created using MASH with the ani command in PanTools.
G1 and G2 clusters are indicated.
https://doi.org/10.1371/journal.pone.0325338.s002
(PDF)
Acknowledgments
The authors are grateful to Dr. David Palmer (Department of Chemistry, University of Saskatchewan) for assistance with the protein structure alignments, and to Dr. Maarten Voordouw (Department of Veterinary Microbiology, University of Saskatchewan) for assistance with statistical analysis and approaches.
References
- 1. Rice JA, Carrasco-Medina L, Hodgins DC, Shewen PE. Mannheimia haemolytica and bovine respiratory disease. Anim Health Res Rev. 2007;8(2):117–28. pmid:18218156
- 2. Highlander SK. Molecular genetic analysis of virulence in Mannheimia (pasteurella) haemolytica. Front Biosci. 2001;6:D1128-50. pmid:11532607
- 3. DeDonder KD, Apley MD. A literature review of antimicrobial resistance in pathogens associated with bovine respiratory disease. Anim Health Res Rev. 2015;16(2):125–34.
- 4. Blakebrough-Hall C, McMeniman JP, González LA. An evaluation of the economic effects of bovine respiratory disease on animal performance, carcass traits, and economic outcomes in feedlot cattle defined using four BRD diagnosis methods. Journal of Animal Science. 2020;98(2).
- 5. Clawson ML, Murray RW, Sweeney MT, Apley MD, DeDonder KD, Capik SF, et al. Genomic signatures of Mannheimia haemolytica that associate with the lungs of cattle with respiratory disease, an integrative conjugative element, and antibiotic resistance genes. BMC Genomics. 2016;17(1).
- 6. Biberstein EL, Gills M, Knight H. Serological types of Pasteurella hemolytica. Cornell Vet. 1960;50:283–300. pmid:13800458
- 7. Al-Ghamdi GM, Ames TR, Baker JC, Walker R, Chase CCL, Frank GH, et al. Serotyping of Mannheimia (Pasteurella) haemolytica Isolates from the Upper Midwest United States. J VET Diagn Invest. 2000;12(6):576–8.
- 8. Mason C, Errington J, Foster G, Thacker J, Grace O, Baxter-Smith K. Mannheimia haemolytica serovars associated with respiratory disease in cattle in Great Britain. BMC Vet Res. 2022;18(1):5. pmid:34980139
- 9. Klima CL, Alexander TW, Hendrick S, McAllister TA. Characterization of Mannheimia haemolytica isolated from feedlot cattle that were healthy or treated for bovine respiratory disease. Can J Vet Res. 2014;78(1):38–45. pmid:24396179
- 10. Arnal JL, Fernández A, Vela AI, Sanz C, Fernández-Garyzábal JF, Cid D. Capsular type diversity of Mannheimia haemolytica determined by multiplex real-time PCR and indirect hemagglutination in clinical isolates from cattle, sheep, and goats in Spain. Vet Microbiol. 2021;258:109121. pmid:34020174
- 11.
Confer AW, Martino P. Mannheimia haemolytica. CABI Compend. 2017;CABI Compendium: 70912. https://doi.org/10.1079/cabicompendium.70912
- 12. Klima CL, Zaheer R, Cook SR, Booker CW, Hendrick S, Alexander TW, et al. Pathogens of bovine respiratory disease in North American feedlots conferring multidrug resistance via integrative conjugative elements. J Clin Microbiol. 2014;52(2):438–48. pmid:24478472
- 13. Holman DB, McAllister TA, Topp E, Wright A-DG, Alexander TW. The nasopharyngeal microbiota of feedlot cattle that develop bovine respiratory disease. Vet Microbiol. 2015;180(1–2):90–5. pmid:26249828
- 14. Noyes NR, Benedict KM, Gow SP, Booker CW, Hannon SJ, McAllister TA, et al. Mannheimia haemolytica in Feedlot Cattle: Prevalence of Recovery and Associations with Antimicrobial Use, Resistance, and Health Outcomes. Veterinary Internal Medicne. 2015;29(2):705–13.
- 15. Ayalew S, Murdock BK, Snider TA, Confer AW. Mannheimia haemolytica IgA-specific proteases. Vet Microbiol. 2019;239:108487. pmid:31767097
- 16. Singh K, Ritchey JW, Confer AW. Mannheimia haemolytica: bacterial-host interactions in bovine pneumonia. Vet Pathol. 2011;48(2):338–48. pmid:20685916
- 17. Kisiela DI, Czuprynski CJ. Identification of Mannheimia haemolytica adhesins involved in binding to bovine bronchial epithelial cells. Infect Immun. 2009;77(1):446–55. pmid:18981250
- 18. Hacker J, Carniel E. Ecological fitness, genomic islands and bacterial pathogenicity. EMBO Reports. 2001;2(5):376–81.
- 19. Cozens D, Sutherland E, Lauder M, Taylor G, Berry CC, Davies RL. Pathogenic Mannheimia haemolytica Invades Differentiated Bovine Airway Epithelial Cells. Infect Immun. 2019;87(6):e00078-19. pmid:30962401
- 20. Loy JD, Clawson ML. Rapid typing of Mannheimia haemolytica major genotypes 1 and 2 using MALDI-TOF mass spectrometry. J Microbiol Methods. 2017;136:30–3. pmid:28267571
- 21. Wynn EL, Schuller G, Loy JD, Workman AM, McDaneld TG, Clawson ML. Differentiation of Mannheimia haemolytica genotype 1 and 2 strains by visible phenotypic characteristics on solid media. J Microbiol Methods. 2020;171:105877. pmid:32088258
- 22. Clawson ML, Schuller G, Dickey AM, Bono JL, Murray RW, Sweeney MT, et al. Differences between predicted outer membrane proteins of genotype 1 and 2 Mannheimia haemolytica. BMC Microbiol. 2020;20(1):250. pmid:32787780
- 23. Klima CL, Cook SR, Zaheer R, Laing C, Gannon VP, Xu Y, et al. Comparative Genomic Analysis of Mannheimia haemolytica from Bovine Sources. PLoS One. 2016;11(2):e0149520. pmid:26926339
- 24. O’Leary NA, Cox E, Holmes JB, Anderson WR, Falk R, Hem V, et al. Exploring and retrieving sequence and metadata for species across the tree of life with NCBI Datasets. Sci Data. 2024;11(1):732. pmid:38969627
- 25. Dassanayake RP, Clawson ML, Tatum FM, Briggs RE, Kaplan BS, Casas E. Differential identification of Mannheimia haemolytica genotypes 1 and 2 using colorimetric loop-mediated isothermal amplification. BMC Res Notes. 2023;16(1).
- 26. Okonechnikov K, Golosova O, Fursov M, Ugene team. Unipro UGENE: a unified bioinformatics toolkit. Bioinformatics. 2012;28(8):1166–7. pmid:22368248
- 27. Untergasser A, Cutcutache I, Koressaar T, Ye J, Faircloth BC, Remm M, et al. Primer3--new capabilities and interfaces. Nucleic Acids Res. 2012;40(15):e115. pmid:22730293
- 28. Kõressaar T, Lepamets M, Kaplinski L, Raime K, Andreson R, Remm M. Primer3_masker: integrating masking of template sequence with primer design software. Bioinformatics. 2018;34(11):1937–8. pmid:29360956
- 29. Koressaar T, Remm M. Enhancements and modifications of primer design program Primer3. Bioinformatics. 2007;23(10):1289–91. pmid:17379693
- 30. Jonkheer EM, van Workum D-JM, Sheikhizadeh Anari S, Brankovics B, de Haan JR, Berke L, et al. PanTools v3: functional annotation, classification and phylogenomics. Bioinformatics. 2022;38(18):4403–5. pmid:35861394
- 31. Sheikhizadeh S, Schranz ME, Akdel M, de Ridder D, Smit S. PanTools: representation, storage and exploration of pan-genomic data. Bioinformatics. 2016;32(17):i487–93. pmid:27587666
- 32.
Anari SS, de Ridder D, Schranz ME, Smit S. Pangenomic read mapping. Cold Spring Harbor Laboratory. 2019. doi: https://doi.org/10.1101/813634
- 33. Jones P, Binns D, Chang H-Y, Fraser M, Li W, McAnulla C, et al. InterProScan 5: genome-scale protein function classification. Bioinformatics. 2014;30(9):1236–40. pmid:24451626
- 34. Blum M, Chang H-Y, Chuguransky S, Grego T, Kandasaamy S, Mitchell A, et al. The InterPro protein families and domains database: 20 years on. Nucleic Acids Res. 2021;49(D1):D344–54. pmid:33156333
- 35. Nielsen H. Predicting Secretory Proteins with SignalP. Methods Mol Biol. 2017;1611:59–73. pmid:28451972
- 36. Blin K, Shaw S, Augustijn HE, Reitz ZL, Biermann F, Alanjary M, et al. antiSMASH 7.0: new and improved predictions for detection, regulation, chemical structures and visualisation. Nucleic Acids Res. 2023;51:W46–W50.
- 37.
Deschner D, Hill JE. Genomic differentiation of Mannheimia haemolytica genotypes. FigShare. 2025. doi: https://doi.org/10.6084/m9.figshare.28872536
- 38. Camacho C, Coulouris G, Avagyan V, Ma N, Papadopoulos J, Bealer K, et al. BLAST+: architecture and applications. BMC Bioinformatics. 2009;10:421. pmid:20003500
- 39. Altschul SF, Madden TL, Schäffer AA, Zhang J, Zhang Z, Miller W, et al. Gapped BLAST and PSI-BLAST: a new generation of protein database search programs. Nucleic Acids Res. 1997;25(17):3389–402. pmid:9254694
- 40. Altschul SF, Gish W, Miller W, Myers EW, Lipman DJ. Basic local alignment search tool. J Mol Biol. 1990;215(3):403–10. pmid:2231712
- 41. Zhang Z, Schwartz S, Wagner L, Miller W. A greedy algorithm for aligning DNA sequences. J Comput Biol. 2000;7(1–2):203–14. pmid:10890397
- 42. Morgulis A, Coulouris G, Raytselis Y, Madden TL, Agarwala R, Schäffer AA. Database indexing for production MegaBLAST searches. Bioinformatics. 2008;24(16):1757–64. pmid:18567917
- 43. Roehrig SC, Tran HQ, Spehr V, Gunkel N, Selzer PM, Ullrich HJ. The response of Mannheimia haemolytica to iron limitation: implications for the acquisition of iron in the bovine lung. Vet Microbiol. 2007;121(3–4):316–29. pmid:17240088
- 44. Tettelin H, Riley D, Cattuto C, Medini D. Comparative genomics: the bacterial pan-genome. Curr Opin Microbiol. 2008;11(5):472–7. pmid:19086349
- 45. Kostova V, Hanke D, Kaspar H, Fiedler S, Schwarz S, Krüger-Haker H. Macrolide resistance in Mannheimia haemolytica isolates associated with bovine respiratory disease from the German national resistance monitoring program GERM-Vet 2009 to 2020. Front Microbiol. 2024;15:1356208. pmid:38495516
- 46. Zheng J, Ge Q, Yan Y, Zhang X, Huang L, Yin Y. dbCAN3: automated carbohydrate-active enzyme and substrate annotation. Nucleic Acids Res. 2023;51(W1):W115–21. pmid:37125649
- 47. Abramson J, Adler J, Dunger J, Evans R, Green T, Pritzel A, et al. Accurate structure prediction of biomolecular interactions with AlphaFold 3. Nature. 2024;630(8016):493–500. pmid:38718835
- 48. Matthews CA, Watson-Haigh NS, Burton RA, Sheppard AE. A gentle introduction to pangenomics. Brief Bioinform. 2024;25(6):bbae588. pmid:39552065
- 49. Tettelin H, Masignani V, Cieslewicz MJ, Donati C, Medini D, Ward NL, et al. Genome analysis of multiple pathogenic isolates of Streptococcus agalactiae : Implications for the microbial “pan-genome”. Proc Natl Acad Sci USA. 2005;102(39):13950–5.
- 50. Deschner D, Voordouw MJ, Fernando C, Campbell J, Waldner CL, Hill JE. Identification of genetic markers of resistance to macrolide class antibiotics in Mannheimia haemolytica isolates from a Saskatchewan feedlot. Appl Environ Microbiol. 2024;90(7):e0050224. pmid:38864630
- 51. Vanni C, Schechter MS, Acinas SG, Barberán A, Buttigieg PL, Casamayor EO, et al. Unifying the known and unknown microbial coding sequence space. Elife. 2022;11:e67667. pmid:35356891
- 52. Lobb B, Tremblay BJ-M, Moreno-Hagelsieb G, Doxey AC. PathFams: statistical detection of pathogen-associated protein domains. BMC Genomics. 2021;22(1):663. pmid:34521345
- 53. Messenger AJ, Barclay R. Bacteria, iron and pathogenicity. Biochemical Education. 1983;11(2):54–63.
- 54. Frawley ER, Fang FC. The ins and outs of bacterial iron metabolism. Mol Microbiol. 2014;93(4):609–16. pmid:25040830
- 55. Sheldon JR, Laakso HA, Heinrichs DE. Iron Acquisition Strategies of Bacterial Pathogens. Microbiol Spectr. 2016;4(2). pmid:27227297
- 56. Ferguson AD, Deisenhofer J. TonB-dependent receptors-structural perspectives. Biochim Biophys Acta. 2002;1565(2):318–32. pmid:12409204
- 57. Silale A, van den Berg B. TonB-Dependent Transport Across the Bacterial Outer Membrane. Annu Rev Microbiol. 2023;77:67–88. pmid:36944260
- 58. Pogoutse AK, Moraes TF. Iron acquisition through the bacterial transferrin receptor. Crit Rev Biochem Mol Biol. 2017;52(3):314–26. pmid:28276700
- 59. Moraes TF, Yu R, Strynadka NCJ, Schryvers AB. Insights into the bacterial transferrin receptor: the structure of transferrin-binding protein B from Actinobacillus pleuropneumoniae. Mol Cell. 2009;35(4):523–33. pmid:19716795
- 60. Ayalew S, Shrestha B, Montelongo M, Wilson AE, Confer AW. Identification and immunogenicity of Mannheimia haemolytica S1 outer membrane lipoprotein PlpF. Vaccine. 2011;29(47):8712–8. pmid:21875637
- 61. Ayalew S, Confer AW, Blackwood ER. Characterization of immunodominant and potentially protective epitopes of Mannheimia haemolytica serotype 1 outer membrane lipoprotein PlpE. Infect Immun. 2004;72(12):7265–74. pmid:15557652
- 62. Confer AW, Ayalew S, Panciera RJ, Montelongo M, Wray JH. Recombinant Mannheimia haemolytica serotype 1 outer membrane protein PlpE enhances commercial M. haemolytica vaccine-induced resistance against serotype 6 challenge. Vaccine. 2006;24(13):2248–55. pmid:16430995
- 63. Pandher K, Confer AW, Murphy GL. Genetic and immunologic analyses of PlpE, a lipoprotein important in complement-mediated killing of Pasteurella haemolytica serotype 1. Infect Immun. 1998;66(12):5613–9. pmid:9826333
- 64. Ayalew S, Blackwood ER, Confer AW. Sequence diversity of the immunogenic outer membrane lipoprotein PlpE from Mannheimia haemolytica serotypes 1, 2, and 6. Vet Microbiol. 2006;114(3–4):260–8. pmid:16386856
- 65. Oppermann T, Busse N, Czermak P. Mannheimia haemolytica growth and leukotoxin production for vaccine manufacturing — A bioprocess review. Electronic Journal of Biotechnology. 2017;28:95–100.
- 66. Davies RL, Whittam TS, Selander RK. Sequence diversity and molecular evolution of the leukotoxin (lktA) gene in bovine and ovine strains of Mannheimia (Pasteurella) haemolytica. J Bacteriol. 2001;183(4):1394–404. pmid:11157953
- 67. Narayanan SK, Nagaraja TG, Chengappa MM, Stewart GC. Leukotoxins of gram-negative bacteria. Vet Microbiol. 2002;84(4):337–56. pmid:11750142
- 68. Larsen J, Pedersen AG, Davies RL, Kuhnert P, Frey J, Christensen H, et al. Evolution of the leukotoxin promoter in genus Mannheimia. BMC Evol Biol. 2009;9:121. pmid:19480674
- 69. Christensen H, Kuhnert P, Foster G, Bisgaard M. Mannheimia indoligenes sp. nov., proposed for clade V organisms of Mannheimia. Int J Syst Evol Microbiol. 2024;74(5). pmid:38739681
- 70. Drula E, Garron M-L, Dogan S, Lombard V, Henrissat B, Terrapon N. The carbohydrate-active enzyme database: functions and literature. Nucleic Acids Res. 2022;50(D1):D571–7. pmid:34850161
- 71. Dik DA, Marous DR, Fisher JF, Mobashery S. Lytic transglycosylases: concinnity in concision of the bacterial cell wall. Crit Rev Biochem Mol Biol. 2017;52(5):503–42. pmid:28644060
- 72. Martinez-Bond EA, Soriano BM, Williams AH. The mechanistic landscape of Lytic transglycosylase as targets for antibacterial therapy. Curr Opin Struct Biol. 2022;77:102480. pmid:36323133
- 73. Suvorov M, Lee M, Hesek D, Boggess B, Mobashery S. Lytic transglycosylase MltB of Escherichia coli and its role in recycling of peptidoglycan strands of bacterial cell wall. J Am Chem Soc. 2008;130(36):11878–9. pmid:18700763
- 74. Scheurwater E, Reid CW, Clarke AJ. Lytic transglycosylases: bacterial space-making autolysins. Int J Biochem Cell Biol. 2008;40(4):586–91. pmid:17468031