Molecular serotyping of Haemophilus parasuis isolated from diseased pigs and the relationship between serovars and pathological patterns in Taiwan

Background Haemophilus parasuis is the etiological agent of Glässer’s disease, and causes severe economic losses in the swine industry. Serovar classification is intended as an indicator of virulence and pathotype and is also crucial for vaccination programs and vaccine development. According to a polysaccharide biosynthesis locus analysis, H. parasuis isolates could be classified by a molecular serotyping assay except serovars 5 and 12 detected by the same primer pair. The aim of this study was to identify H. parasuis isolates from diseased pigs in Taiwan by using a molecular serotyping assay and to analyze the relationship between serovars and pathological patterns. Methods From August 2013 to February 2017, a total of 133 isolates from 277 lesions on 155 diseased animals from 124 infected herds serotyped by multiplex PCR and analyzed with pathological data. Results The dominant serovars of H. parasuis in Taiwan were serovars 5/12 (37.6%), 4 (27.8%) and 13 (15%) followed by molecular serotyping non-typable (MSNT) isolates (13.5%). Nevertheless, the serovar-specific amplicons were not precisely the same sizes as previously indicated in the original publication, and MSNT isolates appeared with unexpected amplicons or lacked serovar-specific amplicons. Most H. parasuis isolates were isolated from nursery pigs infected with porcine reproductive and respiratory syndrome virus. The percentage of lung lesions (30.4%) showing H. parasuis infection was significantly higher than that of serosal lesions. Discussion Collectively, the distribution of serovars in Taiwan is similar to that found in other countries, but MSNT isolates remain due to genetic variations. Furthermore, pulmonary lesions may be optimum sites for H. parasuis isolation, the diagnosis of Glässer’s disease, and may also serve as points of origin for systemic H. parasuis infections in hosts.


INTRODUCTION
Haemophilus parasuis, a part of normal upper respiratory microbiota, is the etiological agent of Glässer's disease which induces sudden death, polyserositis, polyarthritis, meningitis and poor production performance, resulting in severe economic losses in the swine industry (Amano et al., 1994;Moller & Kilian, 1990;Vahle, Haynes & Andrews, 1997;Zhang et al., 2014). Vaccination is an effective strategy for preventing increased mortality and economic losses caused by virulent H. parasuis (Miniats, Smart & Ewert, 1991a;Smart & Miniats, 1989). However, only partial protection is observed with heterologous H. parasuis challenges due to poor cross-protection (Miniats, Smart & Rosendal, 1991b;Nielsen, 1993;Takahashi et al., 2001). Thus, serotyping of H. parasuis is very important, not only for epidemiological research but also for choosing efficacious inactivated whole-cell bacterial vaccines.
Previous studies established the serovar and pathotype of H. parasuis are based on differences at the genome level (Brockmeier et al., 2014;Howell et al., 2013Howell et al., , 2017. A multiplex PCR (mPCR) based on a polysaccharide biosynthesis locus analysis was employed to molecularly serotype H. parasuis serovars (Howell et al., 2015). As a result, 14 of 15 serovars of H. parasuis (serovars 5 and 12 could not be differentiated) were identified using this assay (Howell et al., 2015). Using the molecular typing assay many of the CSNT and CSCR isolates were successfully typed in a recognized serovar.
Although Glässer's disease is common in Taiwan, serotyping of pathogenic H. parasuis isolates from Taiwanese pigs is not clear. The principal aim of this study was to molecularly serotype H. parasuis isolated from Taiwanese diseased pigs, and correlate serovars with pathological patterns.

Bacterial isolate collection and identification
Haemophilus parasuis field isolates were collected from diseased pig herds between August 2013 and February 2017 in Taiwan (Table S2). Lesions suspected of being caused by H. parasuis in diseased pigs were located in the meninges, pleura, pericardia, peritonea, synovial cavities of joints and lungs. Lesions were swabbed and plated on chocolate agar (at 37 C, 5% CO 2 , 18-72 h for growth rate variation for various isolates), blood agar (at 37 C, 16-24 h) and MacConkey agar (at 37 C, 16-24 h). The bacterial isolates were identified by colony morphology, Gram stain (Gram negative bacillus), nicotinamide adenine dinucleotide dependence (only growing on chocolate agar) and virulence-associated trimeric autotransporter group 3 colony PCR (Pina et al., 2009).

Molecular serotyping mPCR
The molecular serotyping assay for H. parasuis isolates was modified from a previously published method (Howell et al., 2015). The species-specific (sp-sp) amplicon was used as an internal control. A loopful of bacteria from a passaged plate of pure culture was resuspended in 30 mL ultrapure H 2 O, which was heated to 100 C for 30 min then centrifuged at 4,000Âg for 1 min. The supernatant was used in the mPCR reaction. Isolates from various lesions or pigs from the same herd were serotyped. If they belonged to same serovar, they were considered one isolate.
Each PCR reaction was performed in a total volume of 30 mL containing ultrapure H 2 O, 1 Â DreamTaq buffer, 250 mM dNTP, 0.2 mM concentrations of forward and reverse serovar-specific primers, 0.04 mM concentrations of forward and reverse sp-sp primers, 1.25 U of DreamTaq DNA polymerase (Thermo Fisher Scientific, Waltham, MA, USA) and one mL of supernatant. The thermocycling conditions consisted of 5 min at 94 C, 30 cycles of 30 s at 94 C, 30 s at 58 C and 1 min at 72 C, and then a final extension at 72 C for 5 min. The molecular serotyping mPCR amplicons were stained with ethidium bromide and analyzed using a 20 cm-long 2% agarose gel. A 50 bp DNA ladder RTU (GeneDireX, Las Vegas, NV, USA) and Bio-1D software (Vilber Lourmat, Collégien, France) were used to estimate molecular size. The electrophoresis conditions were an electric field six V/cm (300 V, 50 cm full-length electric field) and 3 h. The results were confirmed by twice repeating tests.

Sequencing and analysis of unexpected PCR-amplified products
Unexpected amplicons of the molecular serotyping mPCR products were cloned using a TA cloning kit (Yeastern Biotech Co., Ltd. Taipei, Taiwan) and sequenced using an ABI 3730XL DNA analyzer (Applied Biosystems, Foster City, CA, USA). Sequence data were analyzed using MEGA7 (Molecular Evolutionary Genetics Analysis Version 7.0) software and Basic local alignment search tool (BLAST) database.

Pathological examination
Cases of sick animals or fresh, complete carcasses were subjected to necropsy for gross morphological examinations and H&E staining. Histopathological examination focused primarily on meningeal, pleural, pericardial, peritoneal and synovial cavities of joints, and lungs. Typical meningeal gross lesions were characterized by yellow to white exudate accumulation in the subarachnoid space, on pia mater and in the sulci (Fig. S1). Meningeal histopathological lesions were principal neutrophils and few mononuclear inflammatory cells infiltration on pia mater with fibrin and cellular debris deposit (Fig. S2). Serosal lesions were characterized by yellow to white exudate accumulation in pleural, pericardial, abdominal and joint synovial cavities and yellow to white fibrin covering the visceral and parietal serosa (Figs. S3-S5). The histopathological lesions of serositis were principal neutrophils and few mononuclear inflammatory cells infiltration with fibrin deposit (Fig. S6). In typical cases, H. parasuis resulted in bronchopneumonia with numerous neutrophils, mononuclear inflammatory cells, erythrocytes, cellular debris and fibrinous exudate accumulation in alveoli (Figs. S7 and S8). Due to disease duration, lesions varied in field. Other lesions infected H. parasuis were also involved to determine pathological patterns, including chronic fibrous serositis with angiogenesis and mononuclear inflammatory cells infiltration (Fig. S9), and meningitis with principal macrophages infiltration (Fig. S10).

Detection of porcine reproductive and respiratory syndrome virus
Nucleic acid extraction of pulmonary tissue was performed on a MagNA Pure LC 2.0 by using the MagNA Pure LC total nucleic acid isolation kit (Roche Applied Science, Indianapolis, IN, USA). Following cDNA synthesis was using PrimeScript TM RT reagent kits (Takara, Kyoto, Japan). Porcine reproductive and respiratory syndrome virus (PRRSV) reverse transcription real-time PCR was performed as previously described (Lin et al., 2013).

Statistical analysis
Fisher's exact test was used to compare the frequency of H. parasuis infected lesions and the percentage of various lesion patterns using GraphPad Prism software (GraphPad Software, La Jolla, CA, USA). Variables were considered significant at a 0.05 level (two-sided).

H. parasuis isolates, origins and pathological lesion patterns
A total of 133 isolates of H. parasuis were isolated from August 2013 to February 2017. The isolates were taken from 277 lesions on 155 diseased animals from 124 infected herds. Isolates from a herd serotyped as a single serovar were calculated as one isolate. Of 155 H. parasuis cases, 12 cases (7.7%) belonged to suckling pigs ( 3-week-old), 133 cases (85.2%) belonged to nursery pigs (4-to 12-week-old), seven cases (4.5%) belonged to growing pigs (13-to 26-week-old) and one case belonged to a breeding boar. Age information for two cases was unknown. A total of 86 cases (55.5%) had H. parasuis isolated from lung lesions with or without serosal lesions.
A total of 108 animals were necropsied with complete pathological examination and further correlated to pathological pattern and isolation proportion (Table S3). Of the H. parasuis infected animals, 54.6 % had serositis and pulmonary tissue lesions, 41.7% had serosal lesions only and 3.7 % displayed only pulmonary lesions (Fig. 1).

Identification of serovar-specific amplicons
The amplicons generated from molecular serotyping mPCR analyzed using Bio-1D software were not precisely the same sizes as previously indicated in the original description of this assay (Howell et al., 2015). The product size of a specific amplicon found in serovar 4 was mentioned at 320 bp in the original publication but the PCR generated an amplicon of nearly 350 bp which might be confused with serovar 6. In serovar 5, the PCR results generated an amplicon larger than 450 bp mentioned in the original publication which might be confused with serovar 7. The product size of serovar 9 serovar-specific primers, mentioned at 710 bp in the original publication, was smaller than the 700 bp ladder marker and might be confused with serovar 8. In light of these conflicting results, the isolates were serotyped again to confirm the sizes, and the amplicons were subsequently sequenced. Comparisons of the molecular serotyping original publication described, Bio-1D software analyzed, and BLAST product sizes are shown in Table 1 (Howell et al., 2015). The product sizes analyzed by Bio-1D software, BLAST and sequence are more consistent. The unexpected PCR products of the MSNT isolates were cloned for sequencing ( Table 2). The MSNT group 2 amplicon was 297 bp; this product was generated with the serovar 13 specific forward and the serovar 14 specific reverse primers targeting gltP gene as a marker of serovar 13 in the polysaccharide biosynthesis locus. These primers were paired because the target sequences in the respective serovars shared homologous segments. The other PCR generated an amplicon product of the MSNT group 2 isolate determined to be 836 bp, and was identified as a serovar 13 specific product. The Bio-1D software measured a 500 bp product of the MSNT group 3 isolate as 499 bp; this amplicon was identified as a serovar 7 specific product. The 300 bp product found in the MSNT group 4 isolate, (sequencing results indicated it was 297 bp) was generated by pairing the serovar 13 specific forward primer with the serovar 14 specific reverse primers. This result was the same as that generated using the same primer pair of DNA isolated from MSNT group 2. Serovar distribution based on molecular serotyping assay and sequencing The molecular serotyping assay combined with sequencing results reduced the percentage of isolates classified as MSNT from 30.1% to 13.5% (Fig. 3). The dominant serovars were serovar 5/12 (37.6%), serovar 4 (27.8%) and serovar 13 (15%) followed by serovar 14 (2.3%), serovar 7 (1.5%), serovar 1 (0.8%), serovar 2 (0.8%) and serovar 9 (0.8%). Combining the sequencing results showed serovar 13 is a common serovar.  Serovar 13 300 297 S13F, S14R Notes: † Serovar could not be defined without any serovar-specific product sequence result. ‡ There was no serovar-specific product. § Cloning of serovar-specific product was failed.

Relationship between pathological lesion patterns and serovars
The distribution of H. parasuis serovars in lesions from necropsied animals were serovar 5 (42.6%), serovar 4 (21.3%), serovar 13 (20.4%) and MSNT group 1 (13%). These categories were further subdivided into animals displaying both serosal and pulmonary lesions, those with only pulmonary lesions, and those with lesions found only in serosa. The respective percentages of lesions vs. serovars, and the pattern of lesions in infected animals were showed in Fig. 1 and Table S4. Necropsied animals with both serosal and pulmonary lesions were the most frequent; animals with pulmonary lesions alone were the least frequent (p < 0.0001). Serovars 4 and 5/12 showed similar results, the MSNT group 1 both serosal and pulmonary lesions were more frequent than serosal lesions alone. Serovar 13 had more serosal lesions than the combination of serosal and pulmonary lesions. A nine herds (7.3%) had populations infected with two H. parasuis serovars. One herd contained a population with lesions infected with serovars 1 and 4. Three herds were infected with serovars 4 and 5. Serovar 5, 13 and 5, 14 co-infections were seen in single herds. The infected lesions were located in animals displaying a variety of tissue lesion patterns. Other four herds contained individuals co-infected with two H. parasuis serovars. One clinical case showed pleural and pulmonary lesions coinfected with H. parasuis serovars 4 and 7, respectively. A separate herd contained one case of pulmonary lesions with serovar 5, as well as pleural, pericardial and peritoneal lesions infected with H. parasuis serovar 13. One case was co-infected with serovars 5 (pulmonary) and 13 (pleura and pericardium). A fourth case showed coinfection with serovar 4 and an MSNT group 1 isolate taken from separate pulmonary lesions.
Indirect hemagglutination assay was applied to H. parasuis serovar differentiation to decrease the proportion of H. parasuis isolates classified as CSCR (Cai et al., 2005;Del Rio, Gutierrez & Rodriguez Ferri, 2003). De-encapsulation due to multiple passages results in non-reaction with antisera and cross reactivity of isolate antigens to diagnostic (immune-based) test reagents are the primary factors behind CSNT and CSCR H. parasuis isolates, respectively (Cai et al., 2005;Kielstein & Rapp-Gabrielson, 1992;Oliveira, Blackall & Pijoan, 2003;Turni & Blackall, 2005). The presence of CSNT and CSCR isolates confounds epidemiological surveys used to assess H. parasuis isolate population profiles, and impairs efforts to generate effective vaccines against this pathogen. The correlation between the capsule and serovar of H. parasuis is well established; a multiplex serotyping PCR was developed with this in mind (Howell et al., 2015(Howell et al., , 2013. This protocol can be employed to type isolates previously classified as CSNT and CSCR via traditional (immunological) methods. The mPCR serotyping reduced the incidence of CSNT and CSCR H. parasuis but does not completely eliminate the issue of CSNT and CSCR isolates. One reason may be the sequence similarity of different serovar-specific primers and serovar-specific products. Another factor may be deletions and/or unknown sequences within certain antigenic markers (Ma et al., 2016). This underscores the importance of MSNT isolate whole-genome sequencing for in silico serotyping and improving the molecular serotyping assay. Emergence of MSNT isolates by the molecular serotyping assay may be due to insufficient or incomplete sequence data for H. parasuis from Asia. When this assay was developed, there were only nine Asian H. parasuis isolates in a 212-isolate database (seven from Japan, two from China). Investigating the sequences and gene composition of Asian H. parasuis isolate capsule loci may be key for assaying and serotyping MSNT isolates. Besides, absence of serovar-specific markers in polysaccharide biosynthesis loci in MSNT isolates may create antigenic variation impairing vaccine strategies. Therefore, it is also important to study the antigenic variation due to gene mutation and/or absence in polysaccharide biosynthesis loci in the future.
Thus far, molecular serotyping has been challenging as there are 15 serovars, making it difficult to design serovar-specific primers yielding differential results. Some primer pairs produce amplicons from different H. parasuis serovars that vary by less than 20 bp-a difference that is hardly detected especially when the amplicon size is larger than 600 bp. In our study, electrophoresis using longer agarose gels was performed to enhance the ability of the procedure to discriminate DNA fragment sizes. Bio-1D software was applied to more accurately measure product size based on the intensity of the bands and decrease human operation error. In the case of molecular serotyping tests resulting in ambiguities, serovar-specific primer pairs may be used (in simplex PCR format) to confirm or classify hard-to-identify serovars. According to sequence analysis, the product sizes described in the original publication were not accurate (Howell et al., 2015). The corrected product sizes are important to avoid mis-serotyping.
According to a previous study, pigs were infected with H. parasuis serovars 1, 5, 10, 12, 13 and 14 showed high mortality. Pigs challenged with serovars 2, 4, 8 and 15 showed polyserositis. Pigs inoculated with serovars 3, 6, 7, 9 and 11 resulted in no clinical symptoms or lesions (Kielstein & Rapp-Gabrielson, 1992). Serovars 5/12, 4, 13 and 7 are the most common serovars in most countries worldwide (Table S1). A previous study showed that serovar 4 and 13 have a higher prevalence in systemic infection than in only respiratory disease (Luppi et al., 2013). Our data also showed similar results of serovar 4, 5/12, 13 and MSNT isolates. There may be some correlation between serovar and virulence because serovars are defined by capsule which can directly interact with host cells and has been proven to be a key virulence factor relating to phagocytosis resistance ). Besides, it should be considered if the impact of some isolates resulting in only pulmonary lesions are underestimated in field due to absence of serositis and typical pulmonary lesions. The role and economic impact of H. parasuis in pulmonary infection animals related to porcine respiratory disease complex in field is also worthy of investigation in the future. In our study, serovars 7 and 9 caused serositis with or without respiratory lesions. A serovar 5 isolate was isolated from an animal with only bronchopneumonia lesions and another with lesions in both the serosa and lung tissues in the same herd. Therefore, the results show clinical manifestations of Glässer's disease are influenced by multiple factors, including host, stress, environment, co-infection with different serovars or other pathogens and gene differences between infecting H. parasuis isolates (Boerlin et al., 2013;Howell et al., 2014;Li et al., 2009). In general, most Glässer's disease cases in nursery pigs were co-infected with PRRSV in our data. This may be because PRRSV can cause immunosuppression by reducing non-specific bactericidal activity of pig alveolar macrophages and stimulating interleukin-10 production, which down-regulates inflammatory cytokines (Drew, 2000;Flores-Mendoza et al., 2008;Suradhat & Thanawongnuwech, 2003). The previous studies showed PRRSV does not result in increased Glässer's disease by experimental challenge (Segales et al., 1999;Solano et al., 1997). However, significant association between H. parasuis and PRRSV in field was reported (Palzer et al., 2015). Recently studies also showed PRRSV can induce bronchopneumonia with Bordetella bronchiseptica which is a part of normal upper respiratory microbiota and predispose to colonization with H. parasuis (Brockmeier, 2004;Brockmeier et al., 2001). Co-infection of pig alveolar macrophages with PRRSV and H. parasuis leads to pro-inflammatory mediated immunopathology by synergistic effect (Kavanova et al., 2015;Li et al., 2017). In the future, the synergistic effect between PRRSV and H. parasuis resulting in economic losses in field is worthy of further investigation. Other factors also interact with H. parasuis including the stress of weaning and maternal antibody reduction. However, highly virulent H. parasuis isolates might be considered primary pathogens (Aragon, Segalés & Oliveira, 2012). In our study, some H. parasuis isolates caused serositis and sudden death without co-infection, even in growing pigs and breeding boars.
A previous study showed H. parasuis can access the blood stream through invasion of the mucosal surface in the nasal cavity (Vahle, Haynes & Andrews, 1997). In our study, pulmonary lesions showed higher pathogenic H. parasuis infection rates than serosal lesions. These results are in accordance with a previous study from the Netherlands (Dijkman et al., 2012). H. parasuis invasion and survival in lung tissue is likely a key feature for the onset of disease Vahle, Haynes & Andrews, 1995). Our results show H. parasuis infected animals with lesions found in dual anatomical locations (pulmonary and serosal) occur at a higher rate than infected animals with lesions located in only one tissue type. A previous study also mentioned lung is one of the most successful sites for acute (serovar 12) and subacute (serovar 4) isolate recovery (Turni & Blackall, 2007). Therefore, lung is an important origin for H. parasuis isolation and a target organ for Glässer's disease diagnosis. Pulmonary infections may be an important step for H. parasuis systemic infections.
Others have reported isolation of multiple isolates from single pig farms (Cerda-Cuellar et al., 2010;Oliveira, Blackall & Pijoan, 2003;Olvera, Calsamiglia & Aragon, 2006a;Olvera, Cerda-Cuellar & Aragon, 2006b). Our results also show different serovars cause disease in a single herd, or even in a single animal, although the latter scenario is fairly uncommon. In most situations, Glässer's disease is caused by one isolate (Rafiee et al., 2000), but several isolates may be present at a given farm (Turni & Blackall, 2010). Therefore, it would be useful to develop universal or multivalent capsular vaccines against multiple serovars. The possibility of cross talk between different pathogenic H. parasuis isolates at a given site may be worthy of investigation.

CONCLUSIONS
The dominant serovars of H. parasuis in Taiwan are serovars 5/12, 4 and 13, followed by MSNT isolates. Proportions of isolates in those serovars resulting in both serosal and pulmonary lesions are significantly higher than pulmonary lesion. Pulmonary lesions may be most important for H. parasuis isolation and may serve as points of origin for systemic H. parasuis infections in hosts.