The bacterial strains and plasmids used in this study are described in Table 1. Escherichia coli plasmids were propagated in E. coli DH5α grown in Luria-Bertani medium (Oxoid) at 37°C. H. parasuis clinical strain SC096 was cultured on Trypticase Soy Agar (TSA) or Trypticase Soy Broth (TSB) (Oxiod) supplemented with 0.002% (w/v) nicotinamide adenine dinucleotide (NAD; Sigma) and 5% (v/v) inactivated bovine serum at 37°C in a 5% CO₂-enriched atmosphere. For selection of the plasmid-containing strains, the medium were supplemented with 30 μg/mL of kanamycin or gentamycin.

The oligonucleotides used for PCR are listed in Table 2. A DNA fragment encompassing the upstream region of the lgtB gene was amplified using the primer pair P1 and P2. The region downstream of the lgtB gene was amplified using the primer pair P3 and P4. A kanamycin resistance (Kan^R) cassette was amplified from pK18mobsacB using primers P9 and P10. These three fragments were connected by overlap PCR with primers P1 and P4 and then ligated into pMD-19T (simple) to obtain the plasmid pZQ001. Natural transformation was used to introduce pZQ001 into SC096 to obtain the lgtB mutant following a previously described method (Zhang et al., 2012a). The lex-1 mutant was constructed in the same manner with different primers. Primers P5 and P6 were used to amplify the upstream region of lex-1, primers P7 and P8 were used to amplify the downstream region, and primers P9 and P10 were used to amplify the Kan^R cassette. The three fragments were amplified using primers P5 and P8 by overlap PCR and then ligated into pMD-19T (simple) to obtain the plasmid pZQ002. Finally, the plasmid was introduced into SC096 using natural transformation then generated the lex-1 mutant. The mutants were confirmed by PCR and sequencing.

The pSF116 vector was constructed as follows. A 554-bp DNA fragment including the gentamicin resistance (Gm^R) gene was amplified from p34S-Gm using the primer pair P16 and P17. Then the Gm^R gene and pSF115 were digested with BamHI and SalI. The two digested products were ligated to obtain the complement vector pSF116. To construct the complementing plasmids pZQ003 and pZQ004, the lgtB and lex-1 genes were amplified from SC096, then cloned into KpnI and BamHI-digested pSF116 using the In-fusion HD cloning kit (Clontech Laboratories, TaKaRa Bio Inc., Shiga, Japan) respectively.

To construct an in-frame non-polar lgtB mutant, A DNA fragment about 500-bp encompassing the upstream region of the lgtB was amplified using the primer pair P24 and P25, whereas the downstream DNA region of the lgtB was amplified using the primer pair P26 and P27. An intact gentamicin resistant (Gm^R) cassette (from ATG to TAA) was amplified by PCR from p34S-Gm using primers P28 and P29. These three fragments were connected by overlap PCR with the primers P24 and P27, then purified and ligated into pMD19-T to give plasmid pZQ005. Then the plasmid was introduced into SC096 using natural transformation to construct a mutant containing a non-polar, in frame mutation in lgtB, ΔlgtB-np. This mutant was confirmed by PCR and sequenced using primer P22 and P23.

To construct a strain in which the wild-type lgtB gene is restored in the lgtB::kan mutant, the fragment containing upstream region of the lgtB and the intact lgtB was amplified using the primer pair P24 and P30. The downstream region fragment was amplified using the primer pair P26 and P27. The gentamicin resistant (Gm^R) cassette was amplified by PCR from p34S-Gm using primers P31 and P29. The three fragments were connected using primers P24 and P27 by overlap PCR and then ligated into pMD-19T (simple) to obtain the plasmid pZQ006. The plasmid was transformed into lgtB insertion mutant (ΔlgtB::Kan^R) to get the lgtB original locus complement strain. This complement strain was confirmed by PCR and sequenced using primer P22 and P23.

To obtain growth curves for the wild-type SC096 strain, lgtB or lex-1 mutant, cultures of each strain were grown overnight in TSB supplemented with NAD and 5% serum. Then the cultures were inoculated into fresh TSB medium supplemented with NAD and 5% serum at a ratio of 1: 100 and incubated at 37°C. The optical density at 600 (OD600) was measured at 1 h intervals.

Extraction of LOS with high purity was performed using a modified phenol-water extraction protocol accompanied by proteinase K digestion of the bacterial proteins and nuclease elimination of the nucleic acids (Hitchcock and Brown, 1983). The LOS preparations were treated with sodium dodecyl sulfate (SDS) loading buffer (100 mM Tris–HCl, pH 8.0, 2% β-mercaptoethanol, 4% SDS, 0.2% bromophenol blue, 0.2% xylene cyanole, and 20% glycerol), separated by SDS-polyacrylamide gel electrophoresis (PAGE; 4% stacking gel and 15% separating gel) at 100 V for 1 h and visualized by silver staining.

Porcine serum was collected from four healthy piglets (3–4 weeks old) from a farm free of Glässer's disease and was stored at −80°C. Some serum aliquots were treated at 56°C for 30 min to inactivate complement. The serum bactericidal assay was performed as described by (Zhang et al., 2012a). Briefly, bacterial suspensions [10⁷–10⁸ colony-forming units (CFU)/mL] were cultured with either fresh porcine serum or heat-inactivated serum at 1:1 ratios or different serum concentrations for 1 h. After incubation, 10-fold serial dilutions of the samples were generated, spotted onto plates and incubated at 37°C for 48 h. Then, the bacterial numbers were counted, and the survival ratios were calculated. The results were expressed as the means of triplicates from three independent experiments.

PK-15 cells were used for the adhesion and invasion assays following the previously described method (Xu et al., 2013). The cells (5 × 10⁵ CFU/mL) were seeded into 24-well tissue culture plates in Dulbecco's modified Eagle's medium (DMEM; Invitrogen) containing 10% heat-inactivated fetal bovine serum. The cells were cultured at 37°C in a humidified incubator with 5% CO₂ for 24 h, washed twice with PBS and then infected with approximately 1 × 10⁷ CFU of H. parasuis. The culture plates were incubated for up to 2 h at 37°C to allow bacterial adhesion. The cells were rigorously washed five times with PBS to eliminate non-specific bacterial attachment and were then incubated for 10 min at 37°C with 100 mL of 0.25% trypsin/EDTA. After incubation, 900 μL of ice-cold TSB was added, the cells were removed from the culture plates by scraping the bottoms of the wells. Bacterial enumeration was performed using serial 10-fold dilutions and plating on TSA plates. For the invasion assay, cell culture, bacterial infection, and bacterial counting were performed as described above for the bacterial adherence assay except that the extracellular bacteria were killed by incubation of the monolayer with DMEM containing chloromycetin (25 μg/mL) for another 2 h following the incubation with the bacteria and three washes with PBS. All of the above assays were performed in triplicate and repeated three times.

RNAs were isolated from the SC096 and ΔlgtB-c strain. The lgtB gene transcripts were analyzed by quantitative reverse transcription PCR (qRT-PCR). RNA was extracted using the bacterial RNA kit (Omega, USA) according to the manufacturer's instructions. The reactions were performed with the one-step SYBR® PrimeScript™ PLUS RT-PCR Kit (Clontech, USA). The 2(−ΔΔC(T)) method was used to relatively quantitate lgtB gene expression compared to the stably expressed rplM reference gene. The 7500 Real-Time PCR System (Applied Biosystems, Carlsbad, CA, USA) was used for the assay.

All experiments were repeated three times, and the results were expressed as means ± the standard deviations (SD). To determine significance of obtained results, comparison between groups was made using Student t-test. P < 0.01 were considered statistically significant.

Two putative lipooligosaccharide biosynthesis genes (lgtB and lex-1) from H. parasuis SC096 were sequenced and identified using the BLAST program. The lgtB gene had 789 bp and showed 100% identity with a glycosyltransferase from H. parasuis ZJ0906 (NCBI accession no. AGO15727), whereas lex-1 had 837 bp and showed 96% identity with another glycosyltransferase from H. parasuis ZJ0906 (NCBI accession no. AGO15728).

A global sequence comparison between the known lgtB and lex-1 proteins indicated that they belonged to glycosyltransferase family 25. lgtB showed 41, 42, and 41% identities with the β-1,4 galactosyltransferase from Pasteurella multocida (no. WP_014390682.1), the lipooligosaccharide biosynthesis protein lex-1 from Aggregatibacter actinomycetemcomitans (no. WP_053330004.1) and the lipooligosaccharide biosynthesis protein lic2B from H. influenzae (no. WP_048952714.1), respectively. lex-1 showed 45, 46, and 46% identities with the same sequences, respectively. Multiple alignments were created using CLUSTAL W (Figure 1).

To obtain the ΔlgtB mutant, pZQ001, which contained the ΔlgtB::Kan^R insertion, was introduced into the SC096 strain by natural transformation. To obtain the Δlex-1 mutant, pZQ002, which contained the Δlex-1::Kan^R insertion, was introduced in the same manner. To determine the phenotypes of the lgtB or lex-1 strains due to the inactivation of the lgtB and lex-1 genes, pZQ003 was introduced into the ΔlgtB mutant by transformation to obtain a ΔlgtB-Comp strain, and pZQ004 was introduced to obtain a Δlex-1-Comp strain. In the complemented strains, the intact lgtB or lex-1 with a gentamycin resistance cassette was inserted immediately downstream of ompP5. Analysis of the transformants indicated that the lgtB and lex-1 genes plus a kanamycin resistance cassette were integrated into the homologous chromosomes of the mutant strains (Figure 2A). Colony PCR was used to confirm the transformants (Figure 2B). The ability of these mutants to grow under standard conditions was tested, but only negligible changes in the growth rates were detected (Figure S1).

To evaluate the variations in the LOS glycoforms, the LOSs of the wild-type SC096 strain, mutants and complemented strains were assessed by SDS-PAGE. Electrophoretic analysis of the LOS from the ΔlgtB or Δlex-1 mutant indicated that the profiles had changed compared to the profile of the parental strain SC096 (Figure 3). The SC096 strain had a LOS band at 20 kDa (Lane 1), whereas the ΔlgtB mutant had a LOS with a distinctly smaller molecular mass (Lane 2). Specifically, the LOS of the ΔlgtB or Δlex-1 mutant (Lanes 2 and 4) migrated faster than the LOS from the wild-type SC096 strain, and the LOS from the ΔlgtB mutant migrated even faster than the LOS from the Δlex-1 mutant. The LOSs of the complemented strains (Lanes 3 and 5) were similar or identical to the LOS of the parent strain, which suggested that LOS synthesis was restored in the complemented strains. These results suggested that lgtB and lex-1 might affect the LOS structures.

To investigate whether the lgtB and lex-1 genes were involved in serum resistance, the survival rates of ΔlgtB, Δlex-1, and their complemented strains were assessed in 50% porcine serum (Figure 4A). Compared to the wild-type SC096 strain, the ΔlgtB and Δlex-1 mutants were both significantly more sensitive to pig serum (p < 0.01), resulting in survival rates of 15.0 and 54.46% in 50% pig serum, respectively. The ΔlgtB mutant was more susceptible to pig serum than the Δlex-1 strain (p < 0.01). Furthermore, lex-1 mutant showed significantly increased susceptibility to serum of different concentrations compared with the wild type strain SC096 (p < 0.01; Figure S2). Interestingly, complementation of the lex-1 mutant restored the serum resistance phenotype, whereas the survival rate following complementation of the lgtB mutant was 8.51%, representing an approximately 10-fold reduction in 50% porcine serum susceptibility. A qRT-PCR analysis was performed to confirm the transcription level of the lgtB gene in the complemented strain (Figure 4B). The ΔlgtB-c strain exhibited lgtB transcription level that was increased 27.05-fold relative to the SC096 strain, indicating that lgtB was overexpressed in the complemented strain. In order to confirm the phenotype of lgtB mutant, a new in-frame deletion non-polar mutant (ΔlgtB::Gm^R) and an original locus complemented strain were constructed (Figures 5A,B). The new mutant was as sensitive to pig serum as the previous, whereas the original locus complemented strain restored the serum resistance phenotype (Figure 5C). The results suggested that lgtB and lex-1 may be associated with serum resistance.

To assess the effects of the H. parasuis lgtB and lex-1 genes on host cell interactions, PK-15 cells were incubated with the wild-type, mutant, and complemented strains to compare the adherence and invasion abilities. As illustrated in Figure 6A, there was significantly less adhesion by the ΔlgtB and Δlex-1 mutants than the wild-type SC096 strain (p < 0.01). Similarly, the ΔlgtB and Δlex-1 mutants showed significantly less invasion efficiency (Figure 6B; p < 0.01). The adhesion and invasion levels were fully recovered in the complemented lex-1 strain. The complemented lgtB strain exhibited a 1.87-fold increase in adherence and a 1.64-fold increase in invasion. The results indicated that lgtB and lex-1 have an effect on the ability of the bacteria to interact with the host cells.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.