Haemophilus parasuis causes Glässer’s disease in pigs, with symptoms of fibrinous polyserositis, pericarditis, polyarthritis, and meningitis [1]. H. parasuis also causes septicemia and pneumonia without polyserositis and can be isolated from nasal passages of healthy swine. Introduction of conventionally raised pigs into segregated early weaning herds may result in infection and high economic losses because the latter lack immunity to H. parasuis[2, 3]. H. parasuis also remains a problem in many high health status herds. Economic losses in 2006 in the United States were estimated at $145 million dollars (Rodney B. Baker, Veterinary Diagnostic and Production Animal Medicine, Iowa State University, personal communication); [4].

Haemophilus parasuis strains are classified into 15 serovars based on immunodiffusion of heat-stable polysaccharide antigens [5, 6]. However, reagents for serotyping field isolates are not readily available, and a large number of isolates cannot be identified by serotyping and are designated as nontypeable (NT) [7]. Other serotyping methods, such as the indirect hemagglutination test [7–9] have been employed to identify NT isolates. Nonetheless, there are still NT isolates that do not have serovar-specific reagents and cannot be characterized. The virulence of each serovar was determined in specific pathogen free pigs [5].

Molecular typing techniques are increasingly used to identify field isolates including NT isolates. These methods include polymerase chain reaction-restriction fragment length polymorphism (PCR-RFLP) [10, 11], enterobacterial repetitive intergenic concensus-polymerase chain reaction (ERIC-PCR) [12, 13], restriction endonuclease analysis [14, 15], multilocus enzyme electrophoresis (MEE) [16], and multilocus sequence typing (MSLT) analysis [17]. The molecular typing methods have shown that considerable genetic diversity exists among strains of isolates of a particular serotype and that the genotyping techniques were more discriminating compared to conventional serotyping, especially for use in epidemiological studies. Each of these molecular typing techniques offers advantages and disadvantages.

For example, restriction endonuclease experiments [14, 15] found distinct patterns of isolates from animals with systemic disease compared to respiratory isolates from healthy animals but restriction enzymes are expensive. The PCR-RFLP method uses restriction enzymes and sometimes does not generate multiple bands [11]. Multilocus sequence typing (MSLT) is a technique that studies housekeeping genes [17]. However, the latter procedure requires isolation of genomic DNA, performing PCR, and sequencing of PCR products. Both ERIC-PCR [12, 13, 18–20] and MSLT analysis [17] could detect strain variation but not all strains were classified as virulent or avirulent.

Although ERIC-PCR has recently been extensively used to study the epidemiology of H. parasuis isolates [19–21], the random amplified polymorphic DNA (RAPD) technique has not been utilized for this purpose. However, RAPD has been used to distinguish other gamma-proteobacteria, including Salmonella spp. [22], E. coli O157 [23], and Klebsiella pneumonia[24]. Both ERIC-PCR and random amplified polymorphic DNA (RAPD) are global techniques since known primers can be easily synthesized, reagents are affordable and readily obtained, and the techniques have high levels of reproducibility.

In the PCR-based RAPD method, DNA does not have to be double-stranded, highly purified, or of high molecular weight [25]. Both ERIC-PCR and RAPD can utilize DNA from crude lysates [13, 26, 27] which shortens the time needed for completing the assays. However, most laboratories prefer to purify genomic DNA before it is used in genome-based techniques, especially if it is to also be used in assays other than the ERIC-PCR technique [16, 18–21, 28].

The ERIC-PCR technique uses higher annealing temperatures (approximately 50–58°C) and longer primers (20 nucleotides) than the RAPD method. These primers are specific for areas of the genome that are highly conserved and include an inverted repeat. The RAPD assay uses low stringency conditions of approximately 30–36°C annealing temperatures and short (10 nucleotide) primers. One or more of these arbitrarily chosen RAPD primers can anneal at multiple locations throughout the genome and amplify many products of the template DNA.

In addition to genomic-based methods, protein-based methods offer a different and complementary approach. Whole cell protein (WCP; [29–32] profiles or outer membrane protein profiles [33] of H. parasuis, which use a sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) technique have been described. These studies suggested that isolates from systemic sites had unique protein profiles. Isolates from respiratory sites had different protein profiles than the systemic isolates had. The 36–38.5 kDa proteins were described as virulence markers based on the isolation site of the strain [32].

This work analyzed the DNA and protein profiles of 46 H. parasuis reference and field isolates. Random amplified polymorphic DNA is a molecular typing technique that is often used to differentiate closely related strains. It is especially sensitive to strain variation when three optimized primers are employed [34–36]. Random amplified polymorphic DNA may detect single base changes in genomic DNA and genetic maps consisting of RAPD markers can be generated more efficiently than by using RFLP targeted PCR-based methods [28]. Intra-specific variation in the RAPD patterns can be observed for each primer and the sequence complexity of small plasmids is unlikely to contribute to the patterns [26]. However, bacteriophage and larger plasmids with transposons could possibly mediate horizontal gene transfer between strains and increase RAPD heterogeneity [18]. By using the relatively simple and economical RAPD technique, known primer sequences can be utilized by different laboratories, making it a standardized technique and amenable to epidemiological studies. However, interpretation of gel electrophoresis results could introduce some variability between laboratories. The objectives of this study were to compare the relatedness of the reference strains and field isolates based on the RAPD and WCP lysate profiles and to determine if clustering that occurred was related to the site of isolation or to the pathogenicity of the strain.

This study was undertaken to utilize the RAPD technique and SDS-PAGE protein profiles in order to compare 15 reference strains and 31 field isolates of H. parasuis to establish if a relationship existed between a particular clustering profile or if there was a relationship to the site of isolation or to the pathogenicity of the strain. The clinical origin and pathogenesis of a strain is an indication of its virulence, but conclusions as to its virulence cannot be made in our study because pathogenesis studies were not conducted in specific pathogen free pigs [5]. However, the virulence potential of H. parasuis strains, based on their serotype classification, isolation sites and the presence or absence of major proteins with molecular weights between 36 and 38.5 kDa, has been investigated [30, 33, 37]. Some of the expressed proteins in our recent field isolates may be called virulence markers but no direct association of the 40 kDa proteins could be made. Few laboratories have the ability to serotype H. parasuis isolates because of the lack of reagents. Therefore, a genome-based method and a phenotypic analysis of the reference strains and field isolates were emphasized in our study.

Neighbor joining analysis based on Dice coefficients of similarity was used to compare RAPD and protein (WCP lysate) profiles of the reference strains and field isolates. The samples were statistically discriminated better by using the composite RAPD technique than by using the WCP lysate technique but combining the results of both techniques gave a high discrimination index. Outgroups were included to compare the presence or absence of bands in these isolates to the bands in the more closely related H. parasuis isolates. The only monophyletic ingroup with the four “outgroups” was the SDS-PAGE dendrogram as determined by the neighbor joining analysis (Figure 5, Clade A3). The results suggest that the four outgroup species selected may have been too closely related to H. parasuis to act as a true outgroup. Dijkman et al. [20] were also unable to discriminate A. minor and A. porcinus strains from H. parasuis strains in an ERIC-PCR technique. Additionally, Olvera et al. [18] could not demonstrate that A. indolicus and A. minor strains were outgroups to H. parasuis strains when they used the variation of the partial hsp60 sequence of H. parasuis as a classification tool.

Others have shown that the geographic distribution or age of the isolate may cause the “outgroup” to act as an ingroup [38] and that if the isolates in the study were too closely related, then the outgroups could be rerooted to locations within phylogenetic trees [39]. A fourth possibility for the lack of outgroup observance in the dendrograms could be that horizontal gene transfer has occurred between the outgroup species and H. parasuis[40], which would cause unexpected similarities and unusual phyletic patterns [18]. This theory is supported by the presence of bacteriophages in H. parasuis[41–43], E. coli[44], P. multocida[45], M. haemolytica[46], and P. trehalosi[47], plasmids in H. parasuis[48] and A. pleuropneumoniae[49], and a DNA uptake sequence in H. parasuis[50].

Although isolates from known systemic sites [51] (lung in an animal with polyserositis, joint, brain, heart, or septicemia) were able to be separated into groups by the RAPD technique described here, the composite diagram of the three individual primers ultimately showed a limited degree of relatedness based on pathogenicity among the reference strains and the 31 field strains. The strains showed high heterogeneity with the RAPD method which indicated possible horizontal transfer of genes or chromosomal recombination between unrelated and potentially transient strains.

Wang et al. [25] compared RAPD and MEE and found that RAPD data that combined five primers was more discriminatory than MEE tests that used 34 enzymes. The ERIC-PCR technique is a comparable method to RAPD. Zulkifli et al. [52] found RAPD to be more discriminatory than ERIC-PCR. Some H. parasuis isolates were not able to be assayed by using the ERIC-PCR [20] because they gave no or very poor results. Recent studies have found a high diversity of H. parasuis strains isolated in various geographic areas but have not been able to assign a clear correlation between virulence or serovar and ERIC-PCR clusters [19–21]. This conclusion agrees with other H. parasuis ERIC-PCR studies [12, 18]. Macedo et al. [21] reported that the highest diversity was in the NT isolates which discriminated 23 genotypes. Our RAPD dendrogram also indicated high diversity of the H. parasuis strains, with only field isolates 1 and 13 being identical. Although there was no definite correlation between serovar and pathogenicity, most of the isolates that were serotypeable and from diseased animals clustered in Clade C.

Other genomic methods such as MEE and MLST [16, 17], also did not completely discriminate field isolates of H. parasuis. Blackall et al. [16] found 34 different electrophoretic types from 40 field isolates and 8 reference serovars, which clustered into 2 major subdivisions, which were not associated with virulence. Olvera et al. [17] concluded that subgroups of 120 field isolates and 11 reference serovars clustered into branches containing avirulent, nasal isolates and virulent, systemic isolates. However, 36 additional clinical isolates did not cluster within the virulent branch.

Two different studies [53, 54] combined serotyping and IHA methods and concluded that isolates of serovars 4, 5, 13, and NT isolates were the most prevalent in 2004 and 2005, with serovar 4 the most frequently isolated from the respiratory tract while NT isolates were usually systemic isolates. This study’s field isolates were known to be systemic except for isolates 25 and 26, and included serovars 2, 4, 5, 12, and 13, identified by available serotyping reagents. The serovars used in this study were the six most prevalent in the United States and Canada [51, 55]. The range of NT (15-31%) to the frequency of identification of serovars 2, 4, 5, 12, 13, and 14 (76-41%), respectively, by immunodiffusion [32] compares to the frequencies of our “Unk” (51.6%) and six identified serovars (48.3%). Some of our field isolates may have lost the expression of their polysaccharide capsule in vitro and may not be able to be serotyped presently [12, 51] as can be inferred from field isolate 30, which was serotype 4 in 1999 but “Unk” in our study. Field isolate 30 may have lost an enzyme involved in the polysaccharide capsule synthesis. All of our field isolates of known serotype were associated with animals with systemic disease. The majority of field isolates of known serotype were in clade C of the RAPD experiment except for isolates 7, 9, and 23 and in clades B and C of the WCL experiment. Rapp-Gabrielson and Gabrielson [51] and Olvera et al. [17] noted that the distribution of H. parasuis serovars isolated from healthy animals may differ from that found in diseased animals and that more than one serovar could be isolated from the same animal or same isolation site. Our study also identified isolates with different serovars within the same farm site (field isolates 9–11) and in from the same isolation sites in the same animal (field isolates 19–22). “Unk” isolates with septicemia were the most prevalent, indicating that NT isolates harboring bacteriophages, plasmids, or repetitive elements may influence genotyping methods [40]. There may not have been a correlation between serotype and RAPD because only a small number of genes is involved in serotyping while the entire genome is analyzed with the RAPD technique [22].

Our SDS-PAGE results agree with those of Oliviera and Pijoan [30] who reported that isolates from systemic sites were usually virulent and clustered together as shown by using a computer-based analysis of protein profiles from serovars 1, 2, 4, 5, 7, 12, 13, 14 and nontypeable (NT) isolates. Their results are similar to protein profiles described in our study for field isolates and their isolation sites and pathogenesis as shown in the WCP lysate dendrogram of Figure 5 and Table 2. The field strains clustered in Subclade A1 and Clades B and C were primarily systemic. Ruiz et al. [33] found different OMP profiles between isolates from healthy pigs and those from diseased pigs. However, they concluded that respiratory isolates were more heterogeneous than systemic isolates.

Four studies have stated that a protein of approximately 36–38.5 kDa may be associated with Glässer’s disease [29, 30, 33, 56]. In this work, a protein band was observed at approximately 40 kDa in all of the field isolates and thirteen of fifteen of the reference strains (Figure 4). The results shown for the WCP lysate dendrogram (Figure 5) imply that protein expression may be related to age or number of passages of the isolatein vitro, because reference strains clustered together, as did the “old” field strains (26–29) isolated in 1999 (Figure 5, Subclades A2 (C-G, J-O), A3 (A-B, H-I), and A1 (26–29), respectively). The phenotypic change of an isolate after serial passage was also reported by Rapp-Gabrielson and Gabrielson and Oliviera et al. [12, 57]. Although we had only seven samples from North Carolina, three isolates (27–29) from 1999 grouped together in Subclade A1 of the SDS-PAGE neighbor joining dendrogram (Figure 5). Our WCP lysate patterns easily discriminated between A. pleuropneumoniae serotype 1 and H. parasuis as well as the other three outgroup strains (Figure 2B). Identical H. parasuis field isolates (H. parasuis IA84-29755 and 31) (Figure 5), bands did not match sufficiently to obtain identity in the protein profile computer analysis. This may have been because the bands were not fully “matched” in the Gel Compar II program. They were, however, in the same clonal branch of Subclade A3.

Oliviera and Pijoan [30], Kielstein and Rapp-Gabrielson [5], Rosner et al. [58] and Blackall et al. [59] did not find any correlation between virulence and serotype of the isolate. However, the results reported in this study seem to indicate an association of virulence with isolates of Clade C in the WCP lysate analysis. There also seemed to be more serotypeable isolates among the recent field isolates of Clade C. This work also found that the genome-based RAPD composite analysis method (Figure 3) showed more heterogeneity than the protein-based WCP lysate analysis, which grouped isolates according to length of passage in vitro and to possible geographic origin (Figure 5). The underlying genome did not change as much as the protein expression did over time [10].

The recent field isolates from this study were obtained from swine diagnosed mostly with septicemia caused by serovars 2, 4, 5, 12, and 13. All of the isolates from diseased animals grouped into clades in the RAPD neighbor joining dendrogram containing systemic isolates (Figure 3, Clades A and C) or subclade or clades (Subclade A1 and Clades B and C) in the WCL neighbor joining dendrogram containing systemic isolates (Figure 5). Bootstrap values were low for both dendrograms. We did not raise bootstrap cut-off values because others have reported that gains and losses of genes may not be reflected when higher cut-off values are used in the analysis [60].

In order to estimate the discriminatory ability of the primers in the RAPD typing system and of the protein profiles, we used Simpson’s index of diversity. The Simpson’s index of diversity calculation assumes that samples are randomly selected from the population and that all groups are equally represented in the population. Samples in this study were from a few respiratory sites and mostly from diseased animals. Additionally, certain strains may be overrepresented because of their increased pathogenicity in diseased animals. However, if Simpson’s assumptions were not met, a decrease in discrimination would be expected. This was not the case in our study because differences between strains and isolates were seen in both the composite RAPD or WCP lysate results as shown in Table 3.

The results of this study suggested that reference strains, “old” strains isolated in 1999, and recent field strains isolated in 2004 clustered by age of isolate when using WCL methods but not by using RAPD methods. Both the RAPD and the SDS-PAGE methods clustered strains from systemic sites. There was no strong correlation between site of isolation and genotype or between the RAPD and WCL techniques in this study. The RAPD technique showed the high heterogeneity of the H. parasuis isolates, whereas the protein profiles indicated that the number of passages in vitro of an isolate may affect its protein expression. The protein profiles of H. parasuis and A. pleuropneumoniae were unique and this WCP lysate technique may be useful as a tool to differentiate the two NAD-dependent swine respiratory organisms. The protein studies suggested that expressed genes of the organism may help to elucidate the virulence factors involved in the infection. Moreover, the relatively low cost, including supplies and equipment and relatively short amount of time required to perform the RAPD and WCP lysate methods are more advantageous when compared to other genomic or protein methods.