Gene Content and Diversity of the Loci Encoding Biosynthesis of Capsular Polysaccharides of the 15 Serovar Reference Strains of Haemophilus parasuis

Abstract

Haemophilus parasuis is the causative agent of Glässer's disease, a systemic disease of pigs, and is also associated with pneumonia. H. parasuis can be classified into 15 different serovars. Here we report, from the 15 serotyping reference strains, the DNA sequences of the loci containing genes for the biosynthesis of the group 1 capsular polysaccharides, which are potential virulence factors of this bacterium. We contend that these loci contain genes for polysaccharide capsule structures, and not a lipopolysaccharide O antigen, supported by the fact that they contain genes such as wza, wzb, and wzc, which are associated with the export of polysaccharide capsules in the current capsule classification system. A conserved region at the 3′ end of the locus, containing the wza, ptp, wzs, and iscR genes, is consistent with the characteristic export region 1 of the model group 1 capsule locus. A potential serovar-specific region (region 2) has been found by comparing the predicted coding sequences (CDSs) in all 15 loci for synteny and homology. The region is unique to each reference strain with the exception of those in serovars 5 and 12, which are identical in terms of gene content. The identification and characterization of this locus among the 15 serovars is the first step in understanding the genetic, molecular, and structural bases of serovar specificity in this poorly studied but important pathogen and opens up the possibility of developing an improved molecular serotyping system, which would greatly assist diagnosis and control of Glässer's disease.

INTRODUCTION

Haemophilus parasuis is a Gram-negative bacterium and a member of the family Pasteurellaceae. It colonizes the upper respiratory tracts of pigs as a commensal but is also often found associated with pneumonia and is the etiological agent of Glässer's disease (1, 2). Glässer's disease depends on invasion of the bacteria into the systemic compartment and is characterized by fibrinous polyserositis, polyarthritis, and meningitis (3–5). It imposes a significant economic and welfare burden on the global pig industry, resulting in a high demand for the use of antimicrobials (1, 6–8). The vaccines available against this bacterium are whole-cell bacterins, which are protective only against strains of the same serovar (9–11). Current vaccines are protective only against serovars 1, 4, 5, and 6 (12–14); therefore, knowing which serovar is causing infection in a herd is central to the management of Glässer's disease (15–17).

The current serotyping scheme is based on reactions between antisera and surface antigens that classify the bacteria into 15 serovars, with a considerable number of nontypeable isolates also being observed (18, 19). The original Kielstein-Rapp-Gabrielson serotyping scheme was designed in 1992 and used a gel immunodiffusion (GID) assay, but this has been superseded by an indirect hemagglutination assay (IHA) that has increased the proportion of typeable strains from 60% to 80% (12, 19–22). The two methods use different antigen preparation techniques, with GID using autoclaved antigens that are presumed to consist of thermostable polysaccharide components (19), while IHA uses boiled or saline extracts, which are thought to be composed primarily of lipopolysaccharide (LPS) components (20–22). The extracts used in the IHA protocol may vary from the original GID assay, as the production method differs between testing laboratories, and so additional bacterial components may remain in the extracts that are tested. This adds a source of variation to the serotyping process (21, 22). While there are differences between the two serotyping methods, there is strong evidence that the important antigens in both protocols are polysaccharides. The study of these polysaccharide components in relation to serovar has only just started with the publication of structures of the capsular polysaccharide and LPS for two reference strains, and their expression has been monitored in only a selection of the reference strains (23–26). There are some problems with the serotyping assay, including the difficulty of consistently producing specific antisera against several reference strains, variation in growth conditions, cross-reactions between serovars, and the very small number of laboratories that currently perform this test (8, 18, 20, 26). Molecular serotyping systems have been developed for other bacteria based on the genes involved in biosynthesis of extracellular polysaccharide structures such as LPS O antigens or capsular polysaccharides (27–30).

The study of capsule genetics in other bacteria (predominantly Escherichia coli) has led to a classification system that separates capsule loci into four groups (31) based on the genetics and biochemical properties of the polysaccharides (32). Capsular polysaccharides in groups 1 and 4 can be present on the cell surface as a capsule or as short oligosaccharides linked to lipid A core in LPS (33). Capsule loci for groups 1 and 4 contain genes encoding products involved in sugar biosynthesis, polymerization of the sugars into larger structures and translocation of these structures to the cell surface. These genes include, among others, glycosyltransferase genes, wzy (encoding a polymerase), wza (encoding an outer membrane lipoprotein), wzb (encoding a protein-tyrosine phosphatase), and wzc (encoding a tyrosine-protein kinase) (32). These genes are usually grouped by function within the locus, representing export regions (that are common across a number of serovars) and serovar-specific regions (32, 34, 35). wzx and wzy are typically required for both the LPS translocation pathway and capsular polysaccharide export, while wza, wzb, and wzc are specific to capsule biosynthesis (27, 33).

Analysis of the first complete H. parasuis genome sequence (strain SH0165) (36) identified a 14-kb polysaccharide biosynthesis region that was thought to encode an O antigen, with 12 predicted coding sequences (CDSs) in the same transcriptional direction. This locus was also found in the genome sequence of H. parasuis strain 29755 (37). The assignation of putative function to the predicted products of these genes was based on their similarity to other glycosyltransferases and polysaccharide processing/export proteins. In fact, the presence of the wza gene and homologues of wzb and wzc (ptp and wzs, respectively) strongly indicates that the locus is required for the biosynthesis of a polysaccharide capsule, rather than an O antigen (33, 38, 39). Furthermore, experimental evidence does not support the production of an O antigen in H. parasuis (23–25, 40). We therefore propose that this is a capsular polysaccharide biosynthesis locus.

Here we describe the complete DNA sequence and a detailed analysis of this locus from the 15 serovar reference strains. The locus contains considerable serovar-specific variation, which may help to form the basis for a molecular serotyping method and could facilitate the development of glycoconjugate vaccines.

MATERIALS AND METHODS

Bacteria.

Details of the 15 reference strains that are currently used for the production of antisera for the H. parasuis serotyping scheme are in Table 1 (18).

Table 1.

Background information on the serovar reference strains of H. parasuis that have been sequenced (18, 22, 63, 83)^a

Serovar	Reference strain	Country of origin	Isolation site	Diagnosis
1	No. 4	Japan	Nose	Healthy
2	SW140	Japan	Nose	Healthy
3	SW114	Japan	Nose	Healthy
4	SW124	Japan	Nose	Healthy
5	Nagasaki	Japan	Meninges	Septicemia
6	131	Switzerland	Nose	Healthy
7	174	Switzerland	Nose	Healthy
8	C5	Sweden	Unknown	Unknown
9	D74	Sweden	Unknown	Unknown
10	H555	Germany	Nose	Healthy
11	H465	Germany	Trachea	Pneumonia
12	H425	Germany	Lung	Polyserositis
13	IA-84-17975	United States	Lung	Unknown
14	IA-84-22113	United States	Joint	Unknown
15	SD-84-15995	United States	Lung	Pneumonia

^aThese strains were supplied by National Veterinary Institute, Technical University of Denmark, and Centre de Recerca en Sanitat Animal (CReSA), Universitat Autònoma de Barcelona.

Genome sequencing and draft genome sequence assembly.

A single colony of each strain was picked and passaged on chocolate agar plates (Colombia agar base and 7% defibrinated horse blood supplemented with 25 μg/ml NAD). After a minimum of four passages on chocolate agar, the strains were scraped from the plates for genomic DNA preparation. Genomic DNA was prepared using a blood and tissue DNeasy kit (Qiagen) as per the manufacturer's instructions. For library preparation 500 ng of genomic DNA was used, and modified Illumina protocols were followed (41, 42). Paired-end sequencing was performed at the Wellcome Trust Sanger Institute, Hinxton, Cambridge, United Kingdom. All strains were sequenced on an Illumina HiSeq 2000 analyzer for 75 cycles, with repeat sequencing being performed on the Illumina MiSeq for 150 cycles when required.

Draft genome sequences were assembled using a custom-made bioinformatics pipeline, by removing undetermined bases, using Cutadapt (43), Sickle (44), and custom Perl scripts to match paired-end reads. Draft genomes were finally assembled using Velvet (version 1.2.08) and Velvet Optimizer 2.2.0 (45, 46).

Comparison of the capsular polysaccharide biosynthesis loci from the reference strains.

The first and last genes from the published polysaccharide biosynthesis locus (36) were used to test for its presence in the 15 reference strains. A BLAST database was created using the genome sequencing data from the reference strains and queried using the HAPS0039 (starting at 49,282 bp on the SH0165 chromosome) and HAPS0052 (starting at 64,696 bp on the SH0165 chromosome) genes in a BLASTn search.

The potential capsule loci thus identified were visualized using the Artemis comparison tool (ACT) (47) to look for any variation in the locus between the reference strains. These comparisons were performed using a custom Perl script and NUCmer to order, concatenate, and align the draft contigs of the reference strains using exact matches (100% identity) of blocks of sequence (65 bp at a time) to the published genome of SH0165 (36, 48, 49). Default NUCmer settings were used (48, 49). For each reference strain, all open reading frames (ORFs) of more than 100 bp within the locus were examined for predicted CDSs and translated into amino acid sequences using the SIB ExPASy Bioinformatics Resources Portal (50). These sequences were then queried against the NCBI database (using BLASTp) to search for the possible functions of these proteins. The predicted functions were recorded based on the highest amino acid identity match, with an E value threshold of 1 × 10⁻⁶. If the predicted CDS was one previously identified in SH0165, then the highest non-H. parasuis match was also recorded. Annotation of this locus was performed using this information. Predicted gene functions between the reference strains of the 15 serovars were compared for similarity in composition to see if the same gene was found in multiple serovars. Each capsule locus was analyzed for the presence of promoters using BPROM (http://linux1.softberry.com/berry.phtml?topic=bprom&group=programs&subgroup=gfindb), where the top scores were found and mapped to within 150 bp preceding the start codon of each CDS of the 15 capsule loci.

The nucleotide sequences of all CDSs from the 15 capsule loci were aligned using MUSCLE and then viewed by eye using SeaView (17, 21) to group homologous genes present in the different serovars. A predicted function and gene name were given to these groups based on the BLASTp annotations (see Table S2 in the supplemental material). In a few cases where there were different gene function assignations for individual genes within a homology group, the most commonly assigned function and gene name were used.

To cluster the reference strains based on shared genes, the presence and absence of all predicted CDSs across the capsule loci of the 15 reference strains were recorded. This did not take into account gene order in the locus. The presence or absence of genes was recorded as 0 or 1, respectively. To be able to use the pairwise distance matrix function from the R package APE (51), this binary code was converted into a pseudo-nucleotide-based code (A or T, respectively), which could be treated as a nucleotide alignment. Using R, the pairwise distance matrix was generated based on the coded presence/absence alignment, and a neighbor-joining (NJ) tree was then generated based on these data to show the relationship between the gene compositions of the different loci using the APE, Phangorn, and SeqinR software packages (51–53).

Further analysis of the similarities between the individual genes present in the capsule biosynthesis loci of serovars 5 and 12 was performed by aligning the amino acid sequences derived from the CDSs therein using MUSCLE and viewing them using SeaView (54, 55). Amino acid differences that might influence capsule structure, which determines the difference between these two serovars, were recorded.

Comparison to polysaccharide loci of other bacteria.

In order to validate our assignation of the capsule as being from group 1, a comparative analysis between H. parasuis and the characterized capsule groups from different bacteria was carried out. A BLAST database was created using the 15 H. parasuis reference strains and queried using amino acid sequences (via tBLASTn) of known lipooligosaccharide (LOS), LPS, and capsule genes from better-studied bacteria (Haemophilus influenzae, E. coli, and Klebsiella pneumoniae). Proteins involved in LOS and capsule biosynthesis were selected from H. influenzae, as this bacterium is a somewhat closer relative of H. parasuis than the other two and has known polysaccharide structures and protein sequences (56–58). Using the capsule proteins from H. influenzae allowed us to look for any homology to a group 2 capsule within the H. parasuis genomes. E. coli was selected as a model organism for both LPS and the different capsule groups (groups 1 to 4), while K. pneumoniae is another example of a bacterium with a group 1 capsule locus (59–62). The full list of genes used in this series of BLAST queries can be found in Table S3 in the supplemental material.

Comparison of the capsule loci to multilocus sequence type (MLST) profiles.

The relationship between the capsule loci was examined based on the presence or absence of genes within the loci. We sought to compare this to the relationship with the rest of the genome by looking at the MLST profiles of the reference strains. The MLST housekeeping genes from the SH0165 strain were used to query the BLAST database of the reference strains (BLASTn). These genes were then selected from the genomes using the BLAST hit coordinates and a custom Perl script. The MLST housekeeping genes were concatenated in the order established by the MLST scheme (63) and aligned using MUSCLE, and the phylogenetic tree was built using RAxML (54, 64).

Nucleotide sequence accession numbers.

GenBank accession numbers for the individual CDSs from the capsule loci of the 15 reference strains are KC795279-KC795293, KC795295-KC795314, KC795316-KC795331, KC795333-KC795335, KC795351-KC795360, KC795362-KC795364 KC795336-KC795346, KC795348-KC795350, KC795365, KC795367-KC795379, KC795381-KC795384, KC795386-KC795400, KC795402-KC795417, KC853023, KC795418, KC795420-KC795423, KC795425-KC795436, KC795438-KC795441, KC795443-KC795455, KC795457-KC795476, KC795478-KC795491, KC795493-KC795508, KC795510-KC795528, KC795530-KC795533, KC795535-KC795546, and KC795548-KC795550 in order of serovar and gene position within the capsule locus.

RESULTS AND DISCUSSION

Identification of capsular polysaccharide biosynthesis loci.

We report the DNA sequences of the capsular polysaccharide biosynthesis loci in the reference strains of all 15 serovars of H. parasuis. The ACT comparisons of the 15 reference strains have been combined in Fig. 1. Strains are ordered by similarity, as determined by the phylogeny based on gene presence and absence for the locus. This is just one representation of the similarities (and differences) between the capsule loci, ordered here to maximize similarity of the loci based on gene composition.

Fig 1.

Comparison of the reference strains for all 15 serovars of H. parasuis for the capsule loci. Strains are ordered to show the greatest homology between the loci, based on the neighbor-joining tree of gene presence and absence among the capsule loci, as seen on the left of the ACT comparisons. Red areas represent 100% sequence identity, while white areas show variation (areas of nonidentical sequence).

The genes marking the start (HAPS0039) and end (HAPS0052) of the locus were found in all strains, with the genes being found on a single contig for all strains except the reference strain for serovar 2. We were able to align the capsule locus for the serovar 2 reference strain to another serovar 2 genome, which we also sequenced (isolate 9904791), in which the locus assembled onto one closed contig. All the predicted CDSs that were found were the same between the two different serovar 2 isolates.

The location of the capsule loci was the same for all 15 reference strains, based on the available contigs. The genes found at the 5′ end of the locus include genes for four potential antibiotic resistance proteins, a helix-turn-helix (HTH) transcriptional regulator, an outer membrane protein, and several hypothetical proteins. The capsule loci were all followed at the 3′ end by genes encoding an iron-sulfur cluster and several more hypothetical proteins.

Identification of conserved and variable regions in the capsular biosynthesis loci.

The capsule loci have been classified into two regions based on gene content (Fig. 2). Homologues of the wza-wzc genes are found at the 3′ end of the locus, and they are conserved in all of the references strains. These genes are involved in capsular polysaccharide surface expression and export, which matches the definition of region 1 from a model group 1 or 4 capsule locus (32). This is the main line of genetic evidence for the presence of a group 1 in H. parasuis (32). This region is followed by HAPS0052, which is predicted to be a HTH transcriptional regulator that may influence transcription of the capsule loci and is similar to iscR (65). This is in addition to the HTH transcriptional regulator found preceding the locus. In comparison, the 5′ end of the locus is variable, and so this has not been classified as a separate region. Only HAPS0039 is present at the 5′ end of the locus in all strains. HAPS0039 was the locus tag given for the SH0165 gene found at the beginning of the locus. We have renamed this gene funA as the first gene in the locus with function unknown. Region 2 has been designated between funA and wza, with variation in gene composition seen between most of the reference strains (Fig. 2), with the notable exception of the loci from serovars 5 and 12, which are identical in gene composition. This divergent region may correspond to potential serovar-specific regions usually found in capsule loci, including a variety of genes involved in synthesis of oligosaccharide repeat units. These may be distinctive for particular serovars and result in the production of different serovar-specific polysaccharide structures (32, 33, 66).

Fig 2.

Separation of the serovar 5 capsule locus into potential capsule regions with details of the predicted functions. This is based on the published strain SH0165 (36).

The presence of the capsule-specific genes wza, wzb (ptp) and wzc (wzs), indicates that this locus does not in fact encode an O antigen. Rather, we propose that the locus is required for the biosynthesis of a group 1 polysaccharide capsule (33). With the reassignment of the function of this locus, we contend that H. parasuis expresses a LPS molecule lacking O antigen, sometimes called LOS. This fits with newly reported LOS structures for serovars 5 and 15, which were found to be identical, both consisting of lipid A, a single phosphorylated 3-deoxy-D-manno-octulosonic acid (Kdo), and a globotetraose terminal sequence, with no O-antigen-like polysaccharide chains identified (23).

Gene compositions and homology groups of the 15 reference strains' capsule loci.

In total 268 CDSs were found within the capsule loci of the 15 reference strains (an average of 17.9 CDSs per locus), with 86 individually definable genes being annotated. Full details of the predicted gene functions for this locus ordered by reference strain can be found in Table S1 in the supplemental material. A schematic diagram of the polysaccharide biosynthesis loci of the 15 reference strains, which compares the compositions of the variable regions and the diverse range of gene functions, is shown in Fig. 3, with predicted genes colored by function and with the addition of possible promoter locations. One gene was found in the opposite transcriptional direction and appears to encode a transposase (KC795330) in the serovar 3 reference strain. Six genes were present in all 15 reference strains, while many CDSs were unique to a single reference strain based on the amino acid sequences. These genes are candidates for the development of a molecular serotyping assay and represent the huge diversity in capsule structures that might be produced by this bacterium. Ten of the CDSs identified had no significant BLASTp matches and so have been labeled as function unknown (FUN) genes (67). CDSs predicted to encode hypothetical proteins have also been labeled as FUN genes. Of the top hits of these 268 CDSs, 44% have top BLAST hits from the family Pasteurellaceae. After the Pasteurellaceae, the top hits were from the Enterobacteriaceae and the Moraxellaceae, with approximately 10% of the genes being from each family. These three bacterial families are all gammaproteobacteria, but the diversity in capsule genes found within these groups suggests that some of these genes were acquired through horizontal gene transfer.

Fig 3.

Schematic of each capsule locus for all 15 reference strains of H. parasuis. Serovars are ordered based on highest identity of gene composition. Direction of CDS is represented by the direction of the arrow; the gene name assigned is inside the arrow. Coloring represents the predicted function of the gene. Genes are grouped to show similarity between serovars where possible. Arrows above the CDSs represent predicted promoters from BPROM output.

An example of the diversity of genes found within a single capsule locus is found in the serovar 14 reference strain, which contains genes encoding a predicted acetyltransferase, a dehydrogenase, an oxidoreductase, two aminotransferases, two polysaccharide biosynthesis proteins, two epimerases, two glycosyltransferases, and three hypothetical proteins within region 2.

The transcriptional organization of the loci can be predicted from analysis of promoter-like sequences. A −35 box promoter consensus sequence precedes the funA gene in all the capsule loci apart from that from serovar 13. At least three additional promoters have been predicted for the majority of the loci. Eleven of the capsule loci have predicted promoter sequences preceding the wza, wzb, and wzc genes, suggesting that they are transcribed separately from the 5′ end of the locus (Fig. 3). Additional promoter sequences were predicted within the capsule loci, which suggests that transcription is organized differently in the different loci (Fig. 3).

While the first attempt at annotation of the capsule loci used BLASTp to assess protein similarity, we also used nucleotide alignment tools to determine similar/homologous genes with high sequence identity among the reference strains. The main function/homology groups that we identified were genes for polysaccharide biosynthesis, epimerases, and transferases. The remaining genes were considered “not classified,” as they were similar to different genes in the public databases covering a wide range of possible functions. A summary of these functions is presented in Table S2 in the supplemental material, which gives the number of groups found and a tally of unique or orphan genes. Most of the genes were grouped with genes of similar function (214 CDSs), with 48 unique CDSs that did not align well with any other genes. Only 10 of these 48 genes shared sequence identity with genes from the Pasteurellaceae.

Serovar-specific genes.

With the exception of serovars 5 and 12, serovar-specific genes were identified in 12 other serovars. The CDSs in serovar 1 have all been identified in at least one other capsule locus, and so it contains no serovar-specific CDSs. Across the remaining serovars, the majority of serovar-specific CDSs either have no known function or encode glycosyltransferases, based on the unique CDSs from the homology group analysis. This wide range of CDSs, particularly the glycosyltransferases, suggests that the structures of the capsules between the serovars could vary greatly. This is in addition to the 26 glycosyltransferase CDSs that are found among all the serovars, which could create a wide range of polysaccharide structures. Verification of the actual serovar specificity of these CDSs is required to be certain that they could be candidates for a molecular serotyping assay by analyzing further capsule loci from other serotyped field isolates.

Similarities between the capsule loci.

Region 2 of the capsule locus is identical in gene composition in serovars 5 and 12, the only two regions that are identical among the reference strains (Fig. 3). This was not expected, as serovars 5 and 12 can be distinguished from one another in the serotyping scheme. However, vaccine-induced cross-protection between serovars 5 and 12 has been reported on several occasions (11, 22, 68), and it is also a common cross-reaction in the serotyping results of field strains (20, 21). Alignments of the CDSs revealed a total of 48 amino acid substitutions among the 12 of the 14 CDSs that are found in both serovars. Of these differences, 40 amino acid changes are likely to influence the structure of the proteins produced. These changes could be sufficient to alter the polysaccharide structures produced, leading to definition as separate serovars. These amino acid changes have been found in all CDSs except iscR and wzb, with the highest number being in wbgY (encoding a glycosyltransferase) and wzx (encoding flippase). In addition to these amino acid substitutions, promoter analysis has predicted an additional promoter in the serovar 5 strain. At the whole-genome level, the reference strains for serovars 5 and 12 are not identical, with the serovar 12 strain sharing 81.7% nucleotide identity with the complete SH0165 strain (serovar 5) (our unpublished data). Further differences between these serovars may be due to a modification of the capsule encoded by a gene outside the main capsule locus or differential expression of capsule genes (29, 69, 70).

There are other serovars that are very similar to each other at this locus, with serovars 1, 2, 7, and 11 forming a group and serovars 8 and 10 forming another, whose capsule loci differ by only three CDSs. For these serovar groupings, we would expect the capsular polysaccharide structures to be quite similar based on the high identity in gene composition. This similarity in gene composition may also indicate that they have diversified from a single locus within the serovar group. Other serovars share some CDSs and intergenic regions but to a lesser extent than the examples given above, and so we are less certain of their relationships. These similarities in gene composition (as seen in Fig. 1) are summarized in Fig. 4, which, irrespective of gene order, shows the presence and absence of all the genes in the 15 reference strains.

Fig 4.

Summary of genes present in the capsule loci of the 15 reference strains. Strains are ordered by identity, based on the neighbor-joining tree of gene presence and absence among the capsule loci, as seen on the left of the diagram. The genes found in the capsule locus of the 15 reference strains are separated into six categories, colored by function. Unique genes have been excluded. Genes are ordered by decreasing frequency of presence across the 15 reference strains from left to right.

Genes that can be found in capsule and LPS biosynthesis loci from other bacteria.

The wzx and wzy genes are involved in the assembly of both O antigens and group 1 capsular polysaccharides, and so there are similarities between the biosynthetic pathways of these polysaccharide structures (32). We have found that these genes are not present in all capsule loci for H. parasuis based on the BLASTp annotation (see Table S1 in the supplemental material). The wzx allele is not present in three serovars (1, 3, and 11), while there is experimental evidence that at least one of these strains produces a capsule (26), and so the wzx gene may have been replaced by another gene that plays a similar role within the locus. Alternatively, the presence of the wza export protein may be sufficient for the production of the capsule in these serovars. Conversely, the wzy gene is absent from the majority of capsule loci, but again there are many other hypothetical genes within the loci that might substitute for this gene.

So far, we have discussed the content of the capsule loci with respect to homologues specific to a group 1 capsule locus. Group 2 and 3 capsules are assembled in a processive way, using glycosyltransferases instead of the wzy product and an ABC transporter in place of the wzx product for export of the polysaccharide (32, 39). Within the capsule loci we identified one putative ABC transporter gene with low sequence identity in the reference strain for serovar 8 (funM8 KC795416) based on initial BLASTp annotation, but it aligned well with two hypothetical genes in serovars 6 and 10 (funM6 and funM10, respectively), and so these genes have been assigned the name funM, with no known function at this time (Fig. 3). One O-antigen ligase was predicted within the capsule locus of serovar 13 (waaL13 KC795502) (Fig. 3), but with only 27% sequence identity, this may be an inaccurate assignation (31, 39). Further study of the individual genes within the loci and their functional roles in the expression and structure of the capsule is required to understand the mechanisms of capsular polysaccharide biosynthesis in H. parasuis and the potential impact on serotyping and virulence (26).

To reinforce the assignation as a capsule locus, we looked for homologues to genes involved in capsule assembly for all groups, in LOS biosynthesis, and in O-antigen biosynthesis using tBLASTn analyses of the reference strains. The tBLASTn results can be found in Table S3 in the supplemental material. We have found homologues of LOS core proteins and group 1 capsule loci with high sequence identity throughout the reference strains. In comparison, proteins involved in the production of the O antigen were not found to have homologues with high sequence identity in this bacterium, with the exception of those that may also have a role in capsule production (e.g., Wzx, glycosyltransferases, and WbaP). Similarly, the only group 2 and 3 capsule loci homologues found are ABC transporters with low sequence identity (∼25%); this is in fact a large protein family that plays many roles within the cell aside from capsule biosynthesis and so may not be involved in capsule production (71). While group 4 capsule loci are produced in a similar fashion to a group 1 capsule loci, only two homologues of group 4 capsule genes were identified in the H. parasuis reference strains out of 12 known genes (see Table S3) (32). The lack of further proteins involved in O antigen and group 2, 3, or 4 capsule biosynthesis in the genomes of the H. parasuis reference strains supports our assignation of the capsule in this bacterium to group 1.

Relationship between capsule structure and genetics.

The capsule structures of serovars 5 and 15 were recently published, both of which contained the same main chain with a disaccharide repeating unit of β-glucose-6P and 2,4-diacetamido-2,4,6-trideoxy-d-galactopyranose substituted with α-Neu5R-3-α-GalNAc-1-P (serovar 15 strain) or α-Neu5R-3-α-Gal-1-P (serovar 5 strain) side chains, where R is an N-acetyl or N-glycolyl group (72). The relationship between the capsule genes identified in this study and the capsule structures show some similarities for both serovars. This includes the presence of the neuA gene in both loci, which is likely to be involved in the generation of the Neu5R groups in both capsule structures. For the serovar 5 structure, wcwK could be involved in phosphorylation of the UDP-Gal group, and wbgX could be involved in the synthesis of the reducing end sugar. For the serovar 15 structure, the ugeA product is likely to convert glucose to galactose, with the links between the Neu5 group and the main chain being uncertain at this time. The role of capD in the loci is also uncertain; with its product having five predicted transmembrane domains, it may encode an UndPP transferase (36). The wide gene variation within the capsule loci of the 15 reference strains shows that there is the possibility for a wide variety of capsule structures to be produced by this bacterial species. While the genes in the serovar 5 and 15 loci differ significantly, the capsule structures contain the same main chain with various side chains. This could suggest that while the genes differ at the nucleotide level, they could play similar roles, particularly for the glycosyltransferases. These genes differ between the loci, but the majority of the sugars are found in both structures. Alternatively, the additional genes found within the loci could be expressed under different conditions, such as in different growth phases or in response to the host environment, something that is highly likely given the presence of multiple promoters and homologues of the icsR transcriptional regulator.

The capsular polysaccharide acts as the barrier between the bacterium and the environment, including resistance to desiccation, adherence, antiphagocytic activity, and interaction with complement, which all contribute to its common association with virulence and reputation as a virulence factor (73–75). Several polysaccharide modifications can be found in the structures that aid in this role, including sialyl groups and phosphorylcholine (25, 76–78). The presence of sialyl groups in the newly published structures may be an important factor for virulence of this bacterium, as has been suggested for H. parasuis and shown in H. influenzae (25, 76, 77). Homologues of lsgB and neuA have been predicted for the capsule loci of serovars 5 and 15, which would allow the modification and transfer of Neu5 in the capsule structures. We can see that these genes are also present in other capsule loci (Fig. 3), with an additional gene, astA, also being predicted as a sialyltransferase. From this, we could predict that sialylation will be present in 11 of the capsule structures, with the exceptions of serovars 1, 2, 3, and 11. The link between genes involved in sialylation and clinical disease has previously been suggested for nine of the serovars, while the link between serovar and virulence is less clear (25, 79, 80).

Capsule origin and diversity.

This is the first group 1 capsule locus to be found within the genus Haemophilus, and no other members of the Pasteurellaceae have yet been recorded as containing a group 1 capsule (based on literature searches of 26 other Pasteurellaceae species). Group 2 or group 3 capsules have been recorded for other Pasteurellaceae, including H. influenzae and A. pleuropneumoniae as well as four other Pasteurellaceae species, but these are quite distantly related species, based on the molecular phylogenies that have been produced to date (58, 66, 81, 82). While many of the predicted CDSs show homology to closely related Pasteurellaceae species, a greater number match a diverse range of bacterial species.

We also investigated the relationship between the capsule loci in the context of the relationships between the strains derived from their MLST profiles (see Fig. S1 in the supplemental material) and showed that these two data sets indicate different interstrain relationships. For example, the most distantly related serovars are 8 and 10 according to the MLST tree, but they differ by only three CDSs in the capsule locus. The main similarity that can be seen is that serovars 5 and 12 are quite closely related, although not identical, in the MLST tree. This information combined with the wide diversity of BLAST hits used for the capsule annotation suggests that these capsule loci may have arisen through horizontal gene transfer and since diversified.

Concluding remarks.

Capsule gene sequences have been used to develop successful molecular serotyping assays for classifying E. coli (28), H. influenzae (30, 70), and Actinobacillus pleuropneumoniae (69). We suggest that the capsular polysaccharides are the dominant components in the determination of serovar in H. parasuis, and therefore that the genetic loci encoding proteins required for biosynthesis and assembly of capsule structures might be useful for differentiating serovars at the DNA level even though additional bacterial components may also be involved.

While the locus has not been experimentally validated at this time, we have attempted to show via extensive computational analyses that the genes in this locus show the greatest homology to those known to be involved in the group 1 capsular polysaccharide biosynthesis in better-studied organisms (32, 60). This fits with the current experimental data that show the presence of both LOS and capsule structures in this bacterium (23). Understanding this locus will greatly enhance our understanding of the basis for serotype in H. parasuis and will contribute to the field of H. parasuis microbiology.

In conclusion, we report the gene compositions and sequences of the putative capsule loci for all 15 reference strains of H. parasuis. These sequences are a great resource for understanding capsule biosynthesis in this bacterium and, with the large number of “orphan” genes found, may be useful for the development of a molecular serotyping assay for H. parasuis.