Pathogenomic analysis and characterization of Pasteurella multocida strains recovered from human infections

TL;DR

This study identifies specific strains of Pasteurella multocida that cause human infections and reveals genetic traits linked to their ability to infect humans, including a new database for analyzing these bacteria.

Contribution

The study identifies a narrow range of P. multocida strains causing human infections and discovers a novel L-fucose pathway linked to human predilection in subsp. septica.

Findings

Human P. multocida isolates are predominantly capsule type A and LPS type L1/L3, with some lacking capsule biosynthesis genes.

A novel LPS outer-core locus and mobile genetic elements were identified in human isolates.

A new L-fucose uptake pathway was found to be over-represented in subsp. septica strains, suggesting a novel host predilection mechanism.

Abstract

Pasteurella multocida is an upper respiratory tract commensal in several mammal and bird species but can also cause severe disease in humans and in production animals such as poultry, cattle, and pigs. In this study, we performed whole-genome sequencing of P. multocida isolates recovered from a range of human infections, from the mouths of cats, and from wounds on dogs. Together with publicly available P. multocida genome sequences, we performed phylogenetic and comparative genomic analyses. While isolates from cats and dogs were spread across the phylogenetic tree, human infections were caused almost exclusively by subsp. septica strains. Most of the human isolates were capsule type A and LPS type L1 and L3; however, some strains lacked a capsule biosynthesis locus, and some strains contained a novel LPS outer-core locus, distinct from the eight LPS loci that can currently be identified using an LPS multiplex PCR. In addition, the P. multocida strains isolated from human infections contained novel mobile genetic elements. We compiled a curated database of known P. multocida virulence factor and antibiotic resistance genes (PastyVRDB) allowing for detailed characterization of isolates. The majority of human P. multocida isolates encoded a reduced range of iron receptors and contained only one filamentous hemagglutinin gene. Finally, gene-trait analysis identified a putative L-fucose uptake and utilization pathway that was over-represented in subsp. septica strains and may represent a novel host predilection mechanism in this subspecies. Together, these analyses have identified pathogenic mechanisms likely important for P. multocida zoonotic infections.

Pasteurella multocida can cause serious infections in humans, including skin and wound infections, pneumonia, peritonitis, meningitis, and bacteraemia. Cats and dogs are known vectors of human pasteurellosis, transmitting P. multocida via bite wounds or contact with animal saliva. The mechanisms that underpin P. multocida human predilection and pathogenesis are poorly understood. With increasing identification of antibiotic-resistant P. multocida strains, understanding these mechanisms is vital for developing novel treatments and control strategies to combat P. multocida human infection. Here, we show that a narrow range of P. multocida strains cause disease in humans, while cats and dogs, common vectors for zoonotic infections, can harbor a wide range of P. multocida strains. We also present a curated P. multocida-specific database, allowing quick and detailed characterization of newly sequenced P. multocida isolates.

Figures

Maximum-likelihood core-genome phylogenetic tree of the 402 P. multocida strains used in this study. A core-genome alignment was generated using Roary using 281,664 sites from a core-genome containing 1,681 genes, and a maximum-likelihood phylogenetic tree was generated using IQ-TREE using a general time reversible nucleotide substitution model, with rate heterogeneity modeled with empirical base frequencies and a FreeRate model distribution (GTR + F + R9). Branch supports were estimated using 1,000 bootstrap replicates (see Supplemental Material 2 for bootstrap values). All strains underwent in silico capsule and LPS genotyping using Assembly2Feature, with genotypes annotated for each node; inner ring shows host, middle ring shows capsule genotype, and outer ring shows LPS genotype; L9 indicates a novel LPS locus described in detail in the text. No color represents non-typable or no available information. Colored circles on node tips show country of origin. Scale bar represents number of nucleotide substitutions per site. Full metadata are given in Supplemental material 3.

Heat map showing the presence of capsule and LPS biosynthesis genes in P. multocida strains recovered from humans, cats, or dogs. P. multocida genome sequences were searched against the P. multocida virulence factor and antibiotic resistance database using Assembly2Feature. Colored squares represent relative amino acid identity between proteins encoded by query and reference sequences. A maximum-likelihood core-genome phylogenetic tree of P. multocida strains isolated from humans, cats, and dogs is shown on the left. The core-genome alignment was generated using Roary, and the maximum-likelihood tree was generated using IQ-TREE modeled with empirical base frequencies and a FreeRate model distribution allowing for a proportion of invariable sites (GTR + F + I + G4), with 1,000 bootstrap replicates. Scale bar represents number of nucleotide substitutions per site. Both closed and incomplete genomes were included in the analysis, with closed genomes shown by a superscript C. Contig breaks in the capsule and LPS loci are represented by a red line.

Heat map showing the presence of various proteins of interest in the genomes of P. multocida strains isolated from humans, cats, and dogs. P. multocida genome sequences were searched against the P. multocida virulence factor and antibiotic resistance database using Assembly2Feature. Colored squares represent relative amino acid identity between proteins encoded by query and reference sequences. For reference, a maximum-likelihood core-genome phylogeny of P. multocida strains isolated from humans, cats, and dogs is shown. The core-genome alignment was generated using Roary, and the maximum-likelihood tree was generated using IQ-TREE modeled with empirical base frequencies and a FreeRate model distribution allowing for a proportion of invariable sites (GTR + F + I + G4), with 1,000 bootstrap replicates. Scale bar represents number of nucleotide substitutions per site. Both closed and incomplete genomes were included in the analysis, with closed genomes indicated by a superscript C. Some genes/regions in the draft genomes may not be detected due to contig breaks.

Tables

TABLE 1. P. multocida strains sequenced in this studyc

Strain	Host	Subspecies	Capsule genotype	LPS genotypea
Past1	Human	septica	A	L1b
Past3	Human	multocida	F	L3
Past4	Human	septica	A	L1
Past5	Human	septica	A	L3
Past6	Human	septica	No capsule locus	L3
Past7	Human	septica	A	L3
Past9	Human	septica	A	L1
Past10	Human	septica	A	L1b
Past11	Human	septica	A	L1b
Past13	Human	septica	A	L3
Past15	Human	gallicida	A	L1
Past18	Human	septica	A	L1
Past19	Human	septica	A	L1
Past22	Human	septica	F	L1b
Past23	Human	septica	A	L1
Past26	Human	septica	A	NT
Past28	Human	septica	A	L3
Past29	Human	multocida	A	Novel LPS locusL9
Past30	Human	septica	A	L1b
Past31	Human	septica	A	NT
Past33	Human	multocida	No capsule locus	L2
Past34	Human	septica	A	L1b
Pm1476	Cat	septica	A	L1b
Pm1612	Cat	multocida/gallicida	A	L3
Pm1613	Cat	multocida/gallicida	A	L3
Pm1616	Cat	multocida/gallicida	No capsule locus	L3
Pm1617	Cat	multocida/gallicida	No capsule locus	L3
Pm1618	Cat	septica	A	L1
Pm1620	Cat	multocida/gallicida	A	L3
Pm1621	Cat	multocida	No capsule locus	L3
Pm1622	Cat	septica	A	L1
Pmc-A	Cat	septica	A	L1
Pmc-B	Cat	septica	A	L1
Pmc-C	Cat	septica	A	L1
PI31	Dog	septica	A	L1
PI32	Dog	multocida/gallicida	A	L3

TABLE 2. Genes over-represented in P. multocida subsp. septica genomes compared with subsp. multocida/gallicida genomes

Roary gene groupa	Gene name	Percentage presence in 14 subsp. septica strains	Percentage presence in 130 subsp. multocida and gallicida strains	Bonferroni adjusted P value	Predicted function
group_2831	fucU	92.85	0	6.61E−14	L-Fucose mutarotase
fucK	fucK	92.85	14.61	1.99E−05	L-Fuculokinase
ydjF	fucR	92.85	15.38	3.25E−05	L-Fucose operon regulator
fucA	fucA	92.85	15.38	3.25E−05	L-Fuculose-phosphate aldolase
fucI	fucI	92.85	15.38	3.25E−05	L-Fucose isomerase
aldA	aldA	92.85	15.38	3.25E−05	Aldehyde dehydrogenase A
mglA_2		92.85	15.38	3.25E−05	Putative monosaccharide ABC-transport system subunit, contains a MglA domain
rbsC_4		92.85	15.38	3.25E−05	Putative monosaccharide ABC-transport system subunit, contains anAraH domain
rbsB_3		92.85	15.38	3.25E−05	Putative monosaccharide ABC-transport system subunit, contains an ABC sugar binding-like domain
group_1251		92.85	14.61	1.99E−05	Hypothetical protein, contains DUF302
group_2091		92.85	1.53	6.84E−12	Putative phage transcriptional regulator, contains dinD and Bro-N domains
group_2624		85.71	0.76	5.59E−11	Hypothetical protein, contains a IS200-like transposase domain

TABLE 3. Genes under-represented in P. multocida subsp. septica strains compared with subsp. multocida/gallicida genomes

Roary gene groupa	Gene name	Percentage presence in 14 subsp. septica strains	Percentage presence in 130 subsp. multocida and gallicida strains	Bonferroni adjusted P value	Predicted function
group_34		0	99.23	7.57E−15	Hypothetical protein
group_2101		0	93.08	4.12E−10	Hypothetical protein
btuB_4	PM1282	0	90	1.01E−08	Hemin-binding receptor
dcuD_2	dcuC	0	90	1.01E−08	Anaerobic C4-dicarboxylate transporter
group_736		0	87.69	7.34E−08	Hypothetical protein, contains a sulfatase-like domain
yidK		0	87.69	7.34E−08	Hypothetical domain, contains a YidK domain
yeaD_1	yeaD	0	86.92	1.34E−07	D-Hexose-6-phosphate mutarotase
chuR_1		0	86.92	1.34E−07	Putative sulfatase maturation enzyme, contains an AslB domain
group_1535		7.14	93.85	1.27E−08	Hypothetical protein
mtnN	mtn/pfs	14.28	99.23	5.59E−11	5′-Methylthioadenosine/S-adenosylhomocysteine nucleosidase
group_743		14.28	98.46	3.86E−10	Hypothetical protein, contains a STE14 domain
torD		14.28	96.92	7.50E−09	C-Type cytochrome
torC		14.28	96.92	7.50E−09	Trimethylamine-N-oxide reductase-specific chaperone
torA		14.28	96.92	7.50E−09	Trimethylamine-N-oxide reductase
group_1521		14.28	96.15	2.51E−08	Hypothetical protein, contains a DUF1706 domain
group_2890		14.28	95.38	7.43E−08	Hypothetical protein
group_1044		14.28	93.85	4.89E−07	Hypothetical protein

Full text

INTRODUCTION

Pasteurella multocida is a Gram-negative coccobacillus that is part of the normal flora of the upper respiratory tract of many mammal and bird species, including cats and dogs (1–4). P. multocida causes many distinct animal diseases, including fowl cholera in birds, hemorrhagic septicemia in ungulates, atrophic rhinitis in pigs, snuffles in rabbits, and bovine respiratory disease/pneumonia in most livestock animals (5). P. multocida also causes human disease, including skin and wound infections, following animal bites or wound contact with animal saliva, as well as more invasive diseases such as pneumonia, peritonitis, meningitis, and bacteraemia (6, 7). Companion animals such as cats and dogs are often identified as the disease vector, but there are reports of human pasteurellosis occurring without direct contact with animals (6–9).

Very little is known about the strains capable of causing human pasteurellosis, and there have been no detailed comparative genomic investigations of P. multocida strains isolated from humans. However, P. multocida disease in animals has been studied extensively. P. multocida strains can be differentiated by multiplex PCR into three subsp. (multocida, gallicida, and septica) (10), five capsule genotypes (A, B, D, E, and F) (11), and eight lipopolysaccharide (LPS) genotypes (L1 to L8) (12). P. multocida diseases and host predilection are often correlated with capsule and LPS structure (5). Genomic comparison of P. multocida capsule type B and LPS type L2 (B:L2) strains isolated from cattle suffering from hemorrhagic septicemia with strains isolated from poultry and pigs identified 92 genes unique to P. multocida B:L2 strains, some of which are predicted to be disease-specific virulence factors (13). Some prophages and integrative conjugative elements (ICE) were also specific to the hemorrhagic septicemia strains (13). A larger investigation of 656 P. multocida genomes found associations between capsule type and some P. multocida outer membrane proteins and virulence factors (14), identifying possible host predilection determinants. Additionally, several ICEs have been identified in multidrug-resistant strains by whole-genome sequencing or by PCR (15, 16).

Here, we have determined and characterized the genomes of P. multocida strains isolated from human infections, as well as strains isolated from cats and dogs. Comparative genomics of these strains, together with publicly available P. multocida genomes, revealed human P. multocida isolates were predominately subsp. septica, capsule type A and LPS types L1 and L3, although we identified some human strains that had no capsule biosynthesis locus or contained a novel LPS locus. The subsp. septica strains were phylogenetically distinct from other P. multocida strains. Furthermore, subsp. septica strains had a reduced set of genes encoding outer-membrane proteins, and gene-trait analysis identified several genes over-represented in subsp. septica strains representing putative human predilection mechanisms.

RESULTS

Whole-genome sequencing and assembly of P. multocida isolates

We whole-genome sequenced 22 P. multocida isolates from human infections (recovered by Monash Medical Centre in Victoria, Australia), 12 from the oral cavity of cats, and 2 from wounds on dogs (recovered by University of Queensland in Queensland, Australia) (Table S1). Sequencing was performed using Illumina (28 isolates) or a combination of Illumina and Nanopore sequencing (eight isolates), with all genomes assembled de novo. Hybrid sequencing resulted in seven closed genomes (Table S2). Initially, we performed in silico typing of the 36 newly sequenced isolates using average nucleotide identity determined using fastANI or by searching for current multiplex PCR (mPCR) target genes using Assembly2Feature. All strains were confirmed to be P. multocida (Table S3). Most of the 22 human isolates were subsp. septica (18/22), capsule genotype A (16/22), and LPS genotypes L1 (6/22) or L3 (6/22) (Table 1; Table S3). Similarly, the 14 cat and dog isolates were mostly capsule type A (11/14) and LPS types L1 (5/14) or L3 (7/14); however, the cat and dog isolates were a mix of subsp. septica (7/14) and subsp. multocida/gallicida (7/14) (Table 1; Table S3).

P. multocida subsp. septica strains are divergent from other P. multocida strains

A maximum-likelihood core-genome phylogenetic tree was generated using the genomes of the 36 P. multocida isolates sequenced in this study, alongside 366 publicly available genomes (Fig. 1). The core-genome phylogeny was generated using 281,664 nucleotide sites from a core-genome containing 1,681 genes. The P. multocida species split into two clades; all 43 subsp. septica strains included in the phylogeny clustered into the smaller clade, and the larger clade contained only subsp. multocida/gallicida strains. The strain GS2020-X2, although clustering close to subsp. septica clade, had an ANI value < 98 to all P. multocida subspecies type strains and could not be identified to subspecies. The subsp. septica clade contained predominantly capsule type A and LPS type L1 and L3 strains, with a few capsule type F and LPS type L7 strains (Fig. 1). Human infections were significantly over-represented by subsp. septica strains (27 of 35 isolates, P < 0.0001, Fisher’s exact test). Furthermore, capsule type A strains (27 of 35 isolates, P < 0.0013, Fisher’s exact test) were also significantly over-represented in P. multocida isolates recovered from human infection, indicating subsp. septica and capsule type A strains are most likely to cause human disease. Identification of several LPS genotypes (L1, L2, L3, L6, L7, and L9) in the human isolates indicates that strains of many different LPS types can be associated with human disease. Importantly, P. multocida strains isolated from the upper respiratory tract of cats and dogs were found throughout the phylogeny, showing cats and dogs can harbor a wide range of P. multocida strains as normal flora.

Maximum-likelihood core-genome phylogenetic tree of the 402 P. multocida strains used in this study. A core-genome alignment was generated using Roary using 281,664 sites from a core-genome containing 1,681 genes, and a maximum-likelihood phylogenetic tree was generated using IQ-TREE using a general time reversible nucleotide substitution model, with rate heterogeneity modeled with empirical base frequencies and a FreeRate model distribution (GTR + F + R9). Branch supports were estimated using 1,000 bootstrap replicates (see Supplemental Material 2 for bootstrap values). All strains underwent in silico capsule and LPS genotyping using Assembly2Feature, with genotypes annotated for each node; inner ring shows host, middle ring shows capsule genotype, and outer ring shows LPS genotype; L9 indicates a novel LPS locus described in detail in the text. No color represents non-typable or no available information. Colored circles on node tips show country of origin. Scale bar represents number of nucleotide substitutions per site. Full metadata are given in Supplemental material 3.

Human, cat, and dog isolates contained novel P. multocida plasmids, prophages, and an integrative conjugative element

The genomes of the P. multocida isolates sequenced in this study were investigated for mobile genetic elements. Five plasmids (named pBAC1 and pAL1941–pAL1944; Table S4), 14 prophages (named Pmp1 to Pmp14; Table S5), and an ICE (named ICEPmu3) were identified. One of the plasmids identified in five of the human isolates had >96% nucleotide identity with the widely used P. multocida-E. coli shuttle vectors pPBA1100 and pAL99 (17, 18), matching to a region containing the native P. multocida plasmid pBAC1 (17, 19); as such, these plasmids were identified as pBAC1 plasmid (Table S4). Plasmids pAL1942 and pAL1944 were nearly identical to previously known P. multocida plasmids (Table S4). The pAL1944 assembly was split into three contigs, but each contig matched to a predicted conjugative plasmid (GenBank accession CP049757) suggesting they are from the same plasmid (Fig. S1). Plasmid pAL1943 shared significant nucleotide identity with several known P. multocida plasmids but contained novel regions (Fig. S1; Table S4). pAL1941 had no matches when searched against the Plasmid Database and no matches with >20% nucleotide identity in GenBank, strongly suggesting pAL1941 is a novel P. multocida plasmid (Fig. S1). Twelve novel P. multocida prophages were identified, with several of the prophages sharing regions of significant nucleotide identity (Fig. S2). Two prophages, Pmp6 and Pmp12, had regions with identity to the previously identified P. multocida temperate bacteriophages F108 and P86, respectively (20, 21). ICEPmu3 shared regions of nucleotide identity and colinear genes with several ICEs identified in P. multocida genomes, including ICEPmu1, despite being shorter than previously identified P. multocida ICEs (Fig. S3). All plasmids, 12 of the 14 prophages, and ICEPmu3 were identified in the genomes of P. multocida strains isolated from human infection, with only 2 prophages identified in the genomes of strains isolated from animals. All mobile genetic elements contained no identifiable antibiotic resistance or virulence genes.

P. multocida human, cat, and dog isolates had altered carbohydrate surface structures and lacked genes encoding several outer membrane proteins

The human P. multocida isolates were analyzed for virulence and/or host predilection factors associated with human diseases. The genomes of cat and dog isolates were included in the analysis as companion animals are common reservoirs for P. multocida zoonotic infection. We compiled a curated database of P. multocida virulence factor and antibiotic resistance genes (named the PastyVRDB; see Supplemental material 1 additional methods and Table S6). Each genome was searched for the presence/absence of homologs of features in the PastyVRDB using Assembly2Feature. Incomplete genomes were included in this search, and as such, an absence of a feature may be due to missing sequence. Several incomplete genomes had sequence breaks in important loci, namely, the capsule biosynthesis and/or the LPS outer-core locus, and were excluded from further analysis.

In eight genomes (five isolated from human infections—Past6, Past33, P1591, NCTC 11620, and NCTC 11619—and three strains isolated from cats—Pm1616, Pm1617, and Pm1621), no capsule locus was identified (Fig. 2). While the genomes of Pm1616, Pm1617, P1591, and NCTC 11620 were incomplete, in all of these isolate genomes, the two genes typically flanking the capsule locus in P. multocida, grxD and DUF441, were immediately adjacent to each other on a single contig. To confirm these strains did not produce a capsule, capsule quantification assays were performed alongside P. multocida strain VP161 and a known acapsular strain VP161 hyaD mutant (22). No measurable capsule was produced by the strains Past6, Past33, Pm1616, Pm1617, and Pm1621 (Fig. S4).

Heat map showing the presence of capsule and LPS biosynthesis genes in P. multocida strains recovered from humans, cats, or dogs. P. multocida genome sequences were searched against the P. multocida virulence factor and antibiotic resistance database using Assembly2Feature. Colored squares represent relative amino acid identity between proteins encoded by query and reference sequences. A maximum-likelihood core-genome phylogenetic tree of P. multocida strains isolated from humans, cats, and dogs is shown on the left. The core-genome alignment was generated using Roary, and the maximum-likelihood tree was generated using IQ-TREE modeled with empirical base frequencies and a FreeRate model distribution allowing for a proportion of invariable sites (GTR + F + I + G4), with 1,000 bootstrap replicates. Scale bar represents number of nucleotide substitutions per site. Both closed and incomplete genomes were included in the analysis, with closed genomes shown by a superscript C. Contig breaks in the capsule and LPS loci are represented by a red line.

Almost all strains contained all genes required for the assembly of the LPS inner core polysaccharide (Fig. 2). Most human isolates contained an L1 or L3 outer core locus, with some strains containing an L6 or L7 locus (Fig. 2). Three strains, Past29, 161215033201-1, and NCTC 11619, could not be assigned an LPS genotype. Further examination of the LPS outer-core biosynthesis locus, located between priA and fpg, revealed that Past29 and NCTC 11619 contained six genes predicted to be LPS associated, including two with 98% identity to the gatB/gctD and hptE genes in the L6 LPS locus and four genes with no significant identity to any genes previously identified in a P. multocida LPS locus. The four novel LPS genes included a homolog of wecB in E. coli that encodes a UDP-N-acetylglucosamine 2-epimerase, a gene encoding a hypothetical protein containing an OCH1 domain found in mannosyltransferases, a gene encoding a hypothetical protein with a glycosyltransferase family A domain, and a gene encoding a hypothetical protein with a family 25 glycosyltransferase domain. There are currently eight P. multocida LPS outer-core genotypes (12); the outer-core loci of Past29 and NCTC 11619 represent a new LPS outer-core genotype we have now designated L9. Further analysis of the strain 161215033201-1 revealed only one glycosyltransferase gene between priA and fpg, encoding a heptosyltransferase with 98% identity to HptE present in many P. multocida LPS types. Previous LPS structural data have shown that HptE adds the first heptose to the LPS outer core (23).

A search of other known virulence genes revealed that all P. multocida subsp. septica strains isolated from humans (27 isolates), cats (7 isolates), and dogs (2 isolates) encoded only a single filamentous hemagglutinin secretion partner pair, PfhaB1 and PfhaC1 (Fig. 3). In comparison, 46% (164 of 358) of subsp. multocida/gallicida strains encoded two filamentous hemagglutinin secretion partner pairs. Most subsp. septica strains, including those isolated from humans, lacked genes encoding the iron receptors HemR, HasR, and Pm1282 (Fig. 3). Genes encoding transcriptional regulators of virulence factors and the AI-2 quorum sensing system were well conserved in the human, cat and dog P. multocida isolates (Fig. 3). A homolog of Pmorf0222, which encodes a virulence associated protein, was only found in three genomes of isolates recovered from humans, cats, and dogs (Fig. 3). Two phylogenetically independent Tad-loci were identified, with both loci identified in the genomes of both subsp. septica and subsp. multocida/gallicida strains. Comparison of representative Tad-loci from strains HB03 and Past33 showed most of the encoded structural proteins shared <50% amino acid identity, but the Flp1 pilin, TadA, and TadC all shared >85% amino acid identity. No antibiotic resistance genes included in the PastyVRDB were identified in human, cat, and dog P. multocida isolates.

Heat map showing the presence of various proteins of interest in the genomes of P. multocida strains isolated from humans, cats, and dogs. P. multocida genome sequences were searched against the P. multocida virulence factor and antibiotic resistance database using Assembly2Feature. Colored squares represent relative amino acid identity between proteins encoded by query and reference sequences. For reference, a maximum-likelihood core-genome phylogeny of P. multocida strains isolated from humans, cats, and dogs is shown. The core-genome alignment was generated using Roary, and the maximum-likelihood tree was generated using IQ-TREE modeled with empirical base frequencies and a FreeRate model distribution allowing for a proportion of invariable sites (GTR + F + I + G4), with 1,000 bootstrap replicates. Scale bar represents number of nucleotide substitutions per site. Both closed and incomplete genomes were included in the analysis, with closed genomes indicated by a superscript C. Some genes/regions in the draft genomes may not be detected due to contig breaks.

A putative L-fucose uptake and utilization system was over-represented in P. multocida subsp. septica strains

To identify genes that may be associated with P. multocida disease in humans, gene-trait analysis was performed using Scoary to identify genes that were over- or under-represented in P. multocida strains isolated from human infections, as well as in subsp. septica and capsule type A strains as these traits were associated with human infection. No genes were identified as over- or under-represented in P. multocida strains isolated from human infections, and only the capsule synthase gene hyaD was over-represented in capsule type A strains. Overall, 12 genes were identified as over-represented and 17 genes were under-represented in subsp. septica strains ([Tables 2 and 3](#T2 T3)). Over-represented gene groups in subsp. septica strains included an L-fucose utilization system (fucRIKUA and aldA), a monosaccharide ABC-transport system (Groups mglA2, rbsC4, and rbsB3), and a putative prophage transcriptional regulator (Group_2091) (Table 2). The genes for L-fucose metabolism and the monosaccharide transport system were co-located on the genome. Under-represented gene groups in subsp. septica strains included torACD that encodes a trimethylamine N-oxide (TMAO) dehydrogenase system, an anaerobic C4-dicaboxylate transporter, a putative sulfatase and sulfatase maturation protein, a 5′-methylthioadenosine/S-adenosylhomocysteine nucleosidase, and a hemin-binding receptor (Table 3). As human P. multocida isolates are typically subsp. septica, these gene groups may represent novel predilection and pathogenesis mechanisms involved in P. multocida zoonotic infection.

DISCUSSION

In this study, we performed whole-genome sequencing of P. multocida strains isolated from human infections and from the upper respiratory tract of cats and wounds on dogs, as companion animals are a common reservoir for P. multocida zoonotic infections (6). Comparative genomics was performed to characterize these isolates and identify novel host predilection or pathogenic mechanisms. Most P. multocida strains isolated from humans were subsp. septica, capsule genotype A, and LPS genotype L1 or L3 (Table 1), supporting a non-WGS study of P. multocida strains isolated from humans in Hungary where most strains were subsp. septica, capsule type A, and LPS types L1 and L3 (24).

Phylogenetic analysis with strains sequenced in this study and publicly available P. multocida genomes showed the P. multocida species separated into two clades based on subsp. type, with all subsp. septica strains clustering away from subsp. multocida/gallicida strains (Fig. 1). Furthermore, our phylogenetic or ANI data did not support the differentiation of subsp. multocida and gallicida strains as different subspecies. The distinction between subsp. septica and subsp. multocida/gallicida strains has been previously shown with a small number of strains in phylogenies generated using 16S rRNA identity and MLST sequences (24, 25). Interestingly, previous whole-genome nucleotide variant/polymorphism phylogenies have not shown this split between subspecies (14, 26).

Our whole-genome analysis showed P. multocida isolates recovered from human infections were predominantly subsp. septica. P. multocida strains isolated from cats and dogs, considered the most common vectors for zoonotic disease, harbored strains found throughout the phylogeny, not just subsp. septica strains (Fig. 1; Table 1), indicating that only the subsp. septica subset of cat and dog strains commonly cause human disease. Furthermore, we identified several plasmids, prophages, and a novel ICE, primarily in human P. multocida isolates. Interestingly, none of these elements contained identifiable resistance genes. Given the plasmids identified in this study do not have identifiable antibiotic resistance genes, investigation of any fitness advantage provided by these plasmids is warranted.

Investigation of known P. multocida virulence genes across the available human, cat, and dog isolates identified several strains with mutations, deletions, and/or absent genes in the capsule or LPS biosynthesis loci. No capsule locus was found in the genomes of five strains isolated from human infections and three strains isolated from the mouths of cats. Five of the strains that lacked a capsule locus had no measurable capsule in capsule quantification assays, confirming these strains do not produce a capsule layer (Fig. S4). Two of the strains that lacked a capsule locus, namely, P1591 and NCTC 11620, were not available to us to test in the capsule assays. These strains are currently represented by incomplete genomes, which would need to be closed to confirm that these strains lack a capsule locus. Two of the human P. multocida isolates sequenced that did not produce a capsule, Past6 and Past33, were recovered from a peritoneal dialysis portal site and a skin wound, respectively. While extracellular capsule is considered essential for prevention of phagocytosis and for resisting complement mediated killing by P. multocida (27, 28), these data indicate that capsule may not be essential for P. multocida to cause some forms of disease in humans, particularly if the host is immunocompromised.

Analysis of the LPS genes in each genome revealed that three strains had an incomplete LPS outer-core locus (Fig. 2). Strain 16125033201-1, isolated from a human lung infection, only encoded the heptosyltransferase HptE in the outer-core locus. HptE transfers heptose to the LPS inner core glucose (29–31). The lack of any other transferase genes in the locus suggests that the LPS structure would terminate at a distal heptose, a sugar that is usually strongly recognized by the mammalian innate immune response (32). The LPS outer-core structure has been determined for several genotype L3 fowl cholera isolates, but none had heptose as the terminal sugar (12, 30). The LPS outer-core locus present in the genome of strains Past29 and NCTC 11619 is novel and not included in the current diagnostic mPCR that is used to identify the LPS genotype (12). A similar locus is present in the chicken isolate strain 31971 (GenBank accession CP097790). This novel locus contained two genes encoding homologs of glycosyltransferases encoded in other LPS loci, HptE and the L6 glycosyltransferase GatB, and four more genes that have not previously been associated with LPS biosynthesis, encoding a WecB homolog, a mannosyltransferase-like protein, a glycosyltransferase family A protein, and a family 25 glycosyltransferase protein. WecB catalyses the reversible conversion between UDP-GlcNAc and UDP-ManNAc (33). The presence of a mannosyltransferase-like protein and a WecB homolog suggests ManNAc addition to the LPS outer-core of strains Past29 and NCTC 11619. The new P. multocida LPS outer-core genotype, denoted L9, suggests Past29 and NCTC 11619 produce a new outer-core structure. As such, this new L9 locus should be included in the mPCR for identification of the P. multocida LPS genotype, and structural analysis of the LPS outer core produced by these strains is warranted.

P. multocida strains recovered from humans, cats, and dogs had a reduction in the number of outer membrane/surface proteins. The filamentous hemagglutinins PfhaB1 and PfhaB2 were both identified as important for virulence in mice by signature-tagged mutagenesis (STM) (34); however, our analysis suggests only one filamentous hemagglutinin is required for P. multocida colonization of humans, cats, and dogs. Furthermore, strains isolated from humans, cats, and dogs lacked 1–4 genes encoding iron receptors compared with strain Pm70, with no subsp. septica strains encoding the hemin-receptor Pm1282. P. multocida strains typically have several iron receptors that provide functional redundancy (35, 36). As such, loss of a subset of these iron receptors is not expected to impact the ability of these cells to import iron and may improve the fitness of these strains by reducing the number of immune targets. The iron receptor HgbA has been shown to be important for virulence in mice by STM (34), and most of the human, cat, and dog isolates encoded a HgbA homolog. Finally, most strains would likely encode functional Flp-pili, which is expected as Flp-pili subunits have been identified previously as important for P. multocida virulence in mice by STM (37). However, we identified two phylogenetically distinct Flp-pili, with the two types distributed randomly across the phylogeny (Fig. S5), suggesting significant horizontal gene transfer of this locus.

Our study found no genes unique to strains isolated from human infection, suggesting that P. multocida strains able to cause disease in humans do not require an exclusive set of genes to do so. Nevertheless, genes encoding a putative L-fucose utilization pathway (fucRIKUA and aldA) were over-represented in subsp. septica strains. This pathway allows the bacterium to metabolize L-fucose into dihydroxyacetone phosphate and L-lactate, with L-lactate metabolized into pyruvate, with both molecules entering central metabolism pathways (38–40). Together with the ABC monosaccharide transport system, these pathways would allow the bacterium to utilize L-fucose as a carbon source, a sugar commonly found on mucins and human cell surface glycoproteins (41, 42). In Campylobacter jejuni strain NCTC 11168, disrupting L-fucose uptake resulted in reduced fitness compared with the wild-type parent strain, showing L-fucose is important for some bacterial strains to cause disease in humans (43). Previous comparative genomics of three P. multocida strains had identified this L-fucose utilization system as associated with virulent P. multocida strains (44).

Overall, we have characterized P. multocida isolates recovered from human infections and from cats and dogs. This investigation showed P. multocida subsp. septica strains were genetically distinct from subsp. multocida/gallicida strains. Most isolates recovered from human infections were subsp. septica and encoded a type A capsule. Although novel mobile genetic elements were identified, none contained identifiable resistance or virulence genes. Most human, cat, and dog P. multocida isolates encoded a reduced repertoire of outer membrane proteins and iron receptors, and some lacked a capsule biosynthesis locus. Two strains isolated from human infections encoded a novel LPS outer-core biosynthesis locus. Finally, a putative L-fucose uptake and utilization pathway was identified in subsp. septica strains that may be a human host predilection mechanism.

MATERIALS AND METHODS

Bacterial strains and growth conditions

The strains used in this study are listed in Table S1. P. multocida isolates were cultured on heart infusion (HI) agar plates (Oxoid) or horse blood agar plates (Blood agar base no.2 from Oxoid, defibrinated horse blood from Australian Ethical Biologicals Pty Ltd.) or in HI broth for liquid cultures. P. multocida on solid media were grown at 37°C for 24 h to 48 h. Broth cultures were grown at 37°C for up to 24 h with shaking at 200 rpm.

DNA extraction and polymerase chain reaction

Genomic DNA was extracted using the HiYield Genomic DNA Mini Kit (Real Biotech Corporation) as per the manufacturer’s instructions or by cetyltrimethyl ammonium bromide (CTAB) method as previously described (45). DNA yield was assessed using agarose gel electrophoresis and Qubit Fluorometry (Life Technologies).

Whole-genome sequencing and de novo genome assembly

Genomic DNA was submitted to either Micromon Genomics (Monash University, Australia) and sequenced using a 150-bp paired-end protocol on an Illumina MiSeq v2, or to Ramaciotti Centre for Genomics (University of New South Wales, Australia) and sequenced using a 150-bp paired-end protocol on an Illumina MiSeq v3, or prepared with a Native Barcoding Expansion Kit (SQK-NBD112.24) and sequenced on a GridION flow cell (FLO-MIN112) using a GridION X5 Mk1 (Oxford Nanopore Technologies).

Quality control of Illumina sequencing was determined using FastQC (v0.11.9). Reads were trimmed to remove low-quality reads and Illumina adapter sequences using Trimmomatic v0.39 (46). Genomes sequenced by Illumina only were de novo assembled using Unicycler v0.4.8 (47). The Nanopore FAST5 data were processed using Guppy v6.2.7 (Oxford Nanopore Technologies) running the High Accuracy Calling mode. Fastq reads were processed using Filtlong v0.2.1 (https://github.com/rrwick/Filtlong) to remove reads <500 bp and the worst 5% of reads by quality. Genomes sequenced by Nanopore were assembled de novo using Flye v2.9.1 with the –nano-raw argument (48). Each Nanopore assembly was first polished with the Nanopore reads using Medaka v1.7.2 (https://github.com/nanoporetech/medaka) and then polished with the appropriate genomic DNA Illumina reads using Polypolish v0.5.0 (49) and then POLCA v3.4.2 (https://github.com/alekseyzimin/masurca). The genomes were annotated using Prokka v1.14.6 (50). Genome assembly quality control data were generated using QUAST v5.1.0 with the de novo-assembled genomes and trimmed reads as input (51).

Subspecies identification, capsule and LPS genotyping, and plasmid, prophage, and ICE identification

Species and subsp. septica identification was performed using fastANI v1.32 with a kmer size of 21 (52), using the P. multocida subspecies type strains, NCTC 10322 (subsp. multocida), NCTC 10204 (subsp. gallicida), and NCTC 11995 (subsp. septica) as references (https://lpsn.dsmz.de/genus/pasteurella). In silico typing of subsp. gallicida strains or capsule and LPS genotypes was performed using Assembly2Feature (https://github.com/LPerlaza/fromAssembly2Feature); searching for genes used in the current subspecies, capsule, and LPS multiplex PCRs (10–12). See Supplemental material 1 for further in silico typing methodology.

Plasmids were identified in genomes sequenced by Nanopore by generating a hybrid genome assembly with both Nanopore and Illumina data using Unicycler v0.4.8, with small contigs taken to be plasmids (53). Any small contig that had >1× sequencing depth compared with the largest contig and did not have a significant match (BLASTn) to the rest of the hybrid assembly was designated as a plasmid. Plasmids were identified in all genomes using PlasClass with a probability >0.2 used as a cut-off (54). Putative plasmids identified in draft genomes were searched against P. multocida strains with complete genomes using BLASTn, and contigs that matched to chromosomal DNA were excluded from further analysis. The sequencing depth ratio of predicted plasmid sequences compared with the chromosome, as generated in the hybrid Nanopore and Illumina assemblies using Unicycler, was used to estimate the plasmid copy number. Plasmid matches were identified in GenBank using BLASTn and the Plasmid Database using BLASTn with default settings (55). Putative ICEs were identified using ICEfinder (56). Putative prophages were identified using PhiSpy v4.2.21 (57) and annotated using Pharokka v1.3.2 (58). Putative virulence factor and antibiotic resistance genes were searched using ABRicate v1.0.1 (https://github.com/tseemann/abricate) against the ARG-ANNOT and virulence factor databases (59, 60). Positive matches in ABRicate required 80% nucleotide identity over 50% of the query sequence. Homologous regions between different plasmids, prophages, and ICEs were identified using Mauve v2.4.0 (61).

Generating core-genome maximum-likelihood phylogenies

Publicly available P. multocida whole-genome sequences (366 strains) were obtained from the BV-BRD database (Feb 2023) (62) and annotated using Prokka as described above. Core- and pan-genome analysis was performed using Roary v3.11.2 (63), with the arguments -s -ap to not split paralogs and to allow paralogs in the core-genome alignment. Core- and pan-genome analysis was performed using the 36 isolates sequenced in this study and 366 publicly available P. multocida genome sequences. IQ-TREE v1.6.12 (64) was used to generate a maximum-likelihood phylogenetic tree from the P. multocida core-genome alignments produced by Roary. IQ-TREE was run with 1,000 bootstrap replicates; bootstrap values are available in Supplemental material 2. A maximum-likelihood core-genome phylogenetic tree was generated for all P. multocida strains isolated from humans, cats, and dogs in the same way. The capsule and LPS genotypes of publicly available genomes were identified by in silico typing as described above and annotated on the tree with host and country of isolation. Trees were annotated in R using packages ggplot2 and ggtree (65). Strain names were used as node labels, with strain names over 12 characters shortened (ABT_RAWAL_2015_HSR replaced with ABT_HSR and USDA-ARS-USMARC with US). Full metadata for all strains included in the phylogeny are provided in Supplemental material 3.

Identification of virulence, resistance, and trait-specific genes

A targeted P. multocida virulence factor and antibiotic resistance gene search was performed with all strains recovered from humans, cats, and dogs using Assembly2Feature. A curated P. multocida-specific database of virulence factor and antibiotic resistance genes was compiled to use as the reference; the development of this database is described in more detail in Supplemental material 1 Additional Methods, and the full list of genes is included as Supplemental material 4. Assembly2Feature matches were found using the default >80% nucleotide identity cutoff with no minimum length. The nucleotide and protein identities reported by Assembly2Feature are normalized by gene length. For the heat maps, only matches that had a normalized nucleotide identity >50% were shown. Heat maps for Assembly2Feature outputs were generated using the ggplot2 package in R. Heat maps were shown next to the phylogeny of P. multocida strains isolated from humans, cats, and dogs, generated as described above. Trait-specific genes were identified using Scoary v1.6.16 (66). Scoary was run with an adjusted P value cutoff of 1E^−4^ and the argument –no_pairwise. For identification of protein function in this study, protein sequences were searched against the UniProt database using BLASTp to identify homologs and conserved domains identified using NCBI conserved domain search.

Assessing capsule production

Capsule production was assessed in different P. multocida strains using a capsule quantification assay as previously described (22), with minor modifications. Capsule extractions were performed by resuspending mid-exponential culture in 500 µL of PBS and 100 µL of capsule extraction buffer (500 mM citric acid pH 2.0, 1% wt/vol Zwittergent 3-10), before being heated at 50°C for 20 min. The extracted capsule (300 µL) was washed by mixing with 1.5 mL of absolute ethanol, incubating at 4°C for 30 min, and then pelleting the capsule by centrifugation for 13,000 × g. The supernatant was removed, and the capsule pellet was dried before resuspension in 300 µL of dH_2_O.

Supplementary Material

Reviewer comments