The bacterial strains and plasmids used in this study are listed in Table 1. E. coli was grown in LB medium (10 g/l tryptone, 5 g/l yeast extract, and 10 g/l NaCl; pH 7.0) at 37°C following a routine protocol. H. parasuis was grown in Trypticase Soy Agar (TSA) (Oxoid, Hampshire, UK) supplemented with 0.002% nicotinamide adenine dinucleotide (NAD) (Sigma Aldrich, USA) and 5% inactivated bovine serum at 37°C in 5% CO₂. The chemically defined medium was prepared according to a previously described protocol (Murphy and Brauer, 2011) with some modifications and had the following composition: 0.1 M NaCl, 2 mM K₂HPO₄, 2 mM KH₂PO₄, 4 mM Na₂EDTA, 4 mM NH₄Cl, 0.125 mM NaHCO₃, 0.006 mM thiamine hydrochloride, 0.001 mM thiamine pyrophosphate, 0.008 mM calcium pantothenate, 0.02% nicotinamide adenine dinucleotide, 0.375 mM hypoxanthine, 0.45 mM uracil, 0.15 mM glutathione, 0.012 mM biotin, 5% glucose, 0.1 mM Fe(NO₃), 2.5 mM MgCl₂, 0.6 mM CaCl₂, 6.25 mM Na acetate trihydrate, 1.125 mM L-alanine, 0.875 mM L-arginine hydrochloride, 0.2 mM L-asparagine, 3.75 mM L-aspartic acid, 0.35 mM L-cysteine hydrochloride hydrate, 0.15 mM L-cysteine, 7.5 mM L-glutamic acid, 0.35 mM L-glutamine, 0.225 mM glycine, 0.10 mM L-histidine, 0.225 mM L-isoleucine, 0.7 mM L-leucine, 0.35 mM L-lysine hydrochloride, 0.1 mM L-methionine, 0.15 mM L-phenylalanine, 0.45 mM L-proline, 0.475 mM L-serine, 0.425 mM L-threonine, 0.4 mM L-tryptophan, 0.4 mM L-tyrosine, and 0.525 mM L-valine, pH 7.2. When required, the medium was supplemented with kanamycin (30 μg/ml) and gentamicin (30 μg/ml) to grow E. coli and H. parasuis or erythromycin (10 μg/ml) to grow H. parasuis alone. Bacterial growth was determined by measuring the optical density at 600 nm.

A fadD mutant strain of E. coli (JW1794) was used for complementation studies (Baba et al., 2006). The H. parasuis fadD genes were amplified from the genomic DNA of the SC096 strain using the primers P1–P4, which contain NdeI and HindIII sites (as shown in Table 2). The PCR fragments were purified and cloned into the pMD19-T vector (Takara). The fadD sequences were confirmed by sequencing. The pSF226 (carrying fadD1) and pSF227 (carrying fadD2) plasmids were constructed by ligating NdeI-HindIII-digested fragments obtained from pMD19-T with the expression vector pBAD24m, which was digested with the same enzymes. The two recombinant vectors were introduced into the E. coli strain JW1794 for complementation. An empty vector was also transformed into JW1794 and used as a negative control in the complementation experiments. These strains were inoculated into M9 minimal medium supplemented with 0.1% fatty acid with Brij58 (0.02% arabinose was added if required), and the complementation results were determined after the cells were incubated for 72 h at 37°C.

To produce the expression plasmids pSF228 (pET28b carrying fadD1) and pSF229 (pET28b carrying fadD2), cloning vectors (pMD19-T) carrying the above-constructed fadD genes were digested using NdeI and HindIII. The fragments were gel-purified and ligated into pET28b, which was digested using the same enzymes, to generate the expression vectors. The plasmids were then transformed into the E. coli BL21 strain. The respective FadD1 and FadD2 proteins were expressed at high levels and purified by growing the FadD expression strains at 37°C in LB medium. When the OD₆₀₀ reached 0.8, the cultures were induced by the addition of 0.1 mM isopropyl-β-D-thio-D-galactoside (IPTG) and grown at 18°C for an additional 16 h prior to harvest. The cells were centrifuged and collected and then resuspended in lysis buffer containing 50 mM sodium phosphate, 300 mM NaCl, and 10 mM imidazole (pH 8.0). The cells were then lysed using ultrasonic disruption and centrifuged to remove unbroken cells and large debris. The supernatants were loaded onto a nickel-ion HisTrap HP affinity column in an ÄKTA Explorer FPLC system (GE). The column was washed with wash buffer (50 mM sodium phosphate, 300 mM NaCl, and 40 mM imidazole; pH 8.0), and the target proteins were eluted using elution buffer containing 300 mM imidazole. These protein solutions were dialyzed against lysis buffer without imidazole and analyzed using SDS-PAGE and MALDI-TOF.

Fatty acyl-CoA synthetase activity was monitored using Ellman's reagent as previously described (Kang et al., 2010) to detect the amount of free thiol. The reaction buffer mixture contained 150 mM Tris-HCl (pH 7.2), 10 mM MgCl₂, 2 mM EDTA, 0.1% Triton X-100, 5 mM ATP, 0.5 mM reduced CoA, and fatty acid substrate (30–300 μM). The total reaction volume was 450 μl and included 10 μg of purified protein. Briefly, to perform the reaction, each mixture containing all of the components listed above (excluding CoA) was assembled, and 405 μl of the mixture was pre-incubated at 37°C for 3 min. The reaction was initiated with the addition of 45 μl of 5 mM reduced CoA (diluted to a final concentration of 0.5 mM), which was pre-incubated at 37°C for 3 min, quickly mixed, and incubated at 37°C throughout the reaction. Immediately after mixing, a recording was taken at the zero time point by removing 75 μl from the 450 μl reaction mixture and adding it to 600 μl of 0.4 mM 5,5′-dithiobis (2-nitrobenzoic acid) (DTNB, dissolved in 0.1 M potassium phosphate at pH 8.0), and the absorbance value at 412 nm was then measured. Subsequently, 75-μl aliquots of the reaction mixture were taken at 1-min intervals and mixed with DTNB to obtain additional measurements. The extinction coefficient of CoA was assumed to be 1.36 × 10⁴ M⁻¹ cm⁻¹. All reactions involving FadDs were repeated to obtain triplicate data for each fatty acid at each different concentration. The results were analyzed using Hanes-Woolf plots for various amounts of [S] against [S]/V for each concentration of fatty acid.

A suicide plasmid was constructed as follows to disrupt H. parasuis fadD1: DNA fragments located approximately 500 bp upstream and 500 bp downstream of hpfadD1 (D1Up and D1Dn, respectively) were amplified from H. parasuis SC096 genomic DNA using the primers P5–P8 (Table 2). The marker gentamicin was amplified from p34S-Gm using the primers P9 and P10. All of the fragments were connected using overlap PCR with primers P5 and P8 and cloned into pMD19-T to obtain the plasmid pSF230, which carried the fadD1 disruption cassette. The fadD2 upstream and downstream fragments were amplified using the same method, except that the primers P11–P14 were used. The Kan^r gene was amplified from pK18mobsacB using the primers p15 and P16. These primers were ligated using overlap PCR with the primers P11 and P14 to obtain the plasmid pSF231, which carried a fadD2 disruption cassette. These plasmids were transformed into SC096 to generate the fadD1 or fadD2 mutants SF345 and SF346, respectively, via natural transformation, as previously described (Zhang et al., 2012). The fadD1 complementation strain was constructed as follows: an intact fadD1 fragment with its native promoter was amplified with primers P17 and P18. The PCR product of the fadD1 fragment was purified and digested with KpnI and BamHI. The fragment was then cloned into the complementation vector pSF115 (Zou et al., 2013) that was digested using the same enzymes to obtain the plasmid pSF232. The pSF232 plasmid was then transformed into SF345 (fadD1 mutant) to generate the fadD1 complementation strain SF347. A fadD2 complementation plasmid was constructed using a similar method with the primers P19 and P20 and the complementation vector pSF116 (Zhou et al., 2016). The fadD2 plasmid was transformed into SF346 (fadD2 mutant) to obtain the fadD2 complementation strain SF348.

To obtain a fadD1 site-directed mutation strain, a mutation plasmid was constructed as follows: a fragment containing D1Up and intact fadD1 was amplified using SC096 genomic DNA and the primers P5 and P21. Another fragment containing the intact fadD1 gene and D1Dn was amplified using the primers P22 and P8. The erythromycin resistance gene Em^r was amplified using the primers P23 and P24. The fragments were connected using overlap PCR with the primers P5 and P8. The resulting PCR product was cloned into pMD19-T to obtain the template plasmid pSF233, which carried an intact fadD1 gene containing upstream and downstream Em^r inserts at the end of fadD1. Five point mutations were generated using the primers P25–P34. For example, to place the T214A mutation in FadD1, an upstream fragment was amplified using the P5 primer and T214A point mutation primer P25 and using pSF233 as a template. The downstream fragment was amplified using the same template with primers P26 and P8. The two fragments were ligated using overlap PCR and cloned into pMD19T to obtain the plasmid pSF234, which carried the fadD1 recombination cassette containing the T214A mutation in FadD1. The mutation vector was then transformed into SF345 (fadD1 mutant) to generate the fadD1^T214A mutation strain SF350. The other four site-directed mutation strains (G216A, T217A, G219A, and E362A) were constructed using a similar method. To further confirm the essentiality of the fadD genes, fadD2 disruption strains were constructed by natural transformation using fadD1 site-directed mutation strains as recipients. The inability to obtain a mutation indicated that the selected mutation sites in FadD1 were essential for FadD1 activity and for supporting bacteria survival in a fadD2 mutant.

A cellular lipid assay was performed as previously described (Zhu et al., 2013) with some modifications. Briefly, the H. parasuis strains were grown to mid-logarithmic phase in TSB medium containing bovine serum and NAD. The cells were harvested and washed three times with water. The cells were resuspended in 0.8 ml of water, and 1 ml chloroform and 2 ml methanol were then added. An additional 1 ml of water and 1 ml of chloroform were added after the solution was shaken for 1 h at room temperature. The suspensions were vortexed for 10 s and then centrifuged. The lower organic phase was extracted and washed twice with 2 M KCl and once with water and then dried in nitrogen. The phospholipids were dissolved in 1.2 ml of dry methanol, and 0.2 ml of 25% (vol/vol) sodium methoxide (Sigma) was added. A 1.2-ml volume of 2 M HCl was added after the solution was incubated for 30 min at room temperature. The fatty acid methyl esters were extracted using petroleum ether. The extraction agent was removed using a stream of nitrogen, and the samples were analyzed using gas chromatography (GC)-mass spectrometry in an Agilent 7890-5975C. The GC was equipped with a (5%-phenyl)-methyl polysiloxane HP-5MS column (60 × 250 × 0.25 μm). A chromatogram was produced by initially holding the oven temperature at 110°C and then increasing the temperature to 140°C at a rate of 5°C/min. The temperature was then held for 5 min, increased at a rate of 5°C/min to 180°C, and held for 5 min at 180°C. The temperature was increased to 250°C at a rate of 5°C/min and held for 5 min. Finally, the temperature was increased to 280°C at a rate of 5°C/min and held at 280°C for 5 min. The following conditions were used for mass detector: capillary direct interface temperature, 260°C; ionization energy, 70 eV; and mass range, 35–550 amu. Helium was used as the carrier gas (flow rate, 1.5 ml/min). The components in essential oils were identified using the NIST11 MS data library.

The H. parasuis strains were cultured to mid-logarithmic phase in TSB medium containing heat-inactivated newborn bovine serum and NAD. Total RNA was extracted using a Bacterial RNA Isolation Kit (Omega). The RNA was adjusted to a concentration of 200–500 ng/μl (measured using a NanoDrop 8000, Thermo Scientific), and all samples were reverse-transcribed using a PrimeScript™ RT Reagent Kit (Takara) with the specific primers P35, P36, and P37. Quantitative real-time PCR analyses were performed using an Applied Biosystems PRISM model 7500 Sequence Detection system with SYBR® Premix Ex Taq™ (Takara). The results were relatively quantified using the comparative cycle threshold method. The endogenous internal control rplM was used for sample normalization, as previously described (Zhou et al., 2016). The amplification program was as follows: 30 s at 95°C followed by 40 cycles at 95°C for 5 s and 60°C for 34 s. Three independent experiments were performed for each strain, and three technical replicate RT-PCRs were performed for each sample.

The serum bactericidal assays were performed using porcine serum as previously described (Zou et al., 2013). Porcine serum stored at −80°C from four healthy piglets (3–4 weeks old) from a farm free of Glässer's disease collected in previous work (Zhou et al., 2016) was used for this study. The piglets were free of Glässer's disease. The serum was filter-sterilized (0.22 μM), and aliquots were stored at −80°C. The controls were performed using serum samples that were heated at 56°C for 30 min. Wild-type and recombinant H. parasuis strains were grown to mid-logarithmic phase. A 100 or 400-μl volume of fresh porcine serum or heat-treated serum was mixed with 100 μl of a diluted bacterial suspension (10⁸ CFU) to achieve a final concentration of 50 or 80% serum. The mixture was then incubated at 37°C for 1 h. The bacteria were serially diluted and plated on TSA plates containing heat-inactivated newborn bovine serum and NAD. To quantify the results, the plates were incubated for 24 h at 37°C in an atmosphere containing 5% CO₂. The percent survival was calculated as the ratio of the number of colonies grown from fresh serum to the number grown from heat-treated serum. Three independent experiments were performed for each H. parasuis strain.

To control the inoculum volume, the antimicrobial susceptibility of all of the H. parasuis strains was determined using the agar dilution method with an antibacterial determiner (SAKUMA, Japan). The results were interpreted according to the Clinical and Laboratory Standards Institute (CLSI) guidelines. The MIC value was defined as the lowest concentration that resulted in no visible colonies. The reference strains Actinobacillus pleuropneumoniae ATCC 27090 and Escherichia coli ATCC 25922 were used as quality controls for the MIC determination.

Student's t-test was used to determine the significance of differences. A p-value < 0.05 was considered statistically significant.

In this study, two H. parasuis genes, fadD1 (HAPS_0217) and fadD2 (HAPS_1695), were annotated and found to encode homolog of E. coli FadD, which is a FACS. FACSs belong to the ANL superfamily of adenylating enzymes. Members of this family catalyze two partial reactions: the first is the adenylation of a carboxylate to form an acyl-AMP intermediate, which is followed by a second reaction that usually leads to the formation of a thioester (Gulick, 2009). The reaction product of FadD is acyl-CoA (Figure 1A). A sequence comparison of members of the FACS protein family revealed the presence of an ATP-AMP signature motif and a FACSs conserved motif (Black and DiRusso, 2003).

HpfadD1 is located in a chromosomal cluster with resB and resC (a putative cytochrome c biogenesis-associated protein), whereas HpfadD2 is located at another locus and is likely to be monocistronic (Figure 1B). The gene cluster structures of both genes shared no similarity with the E. coli fadD locus. An alignment of the FadD homolog revealed that HpFadD1 was 65% identical to E. coli FadD and that the ATP/AMP and FACS motifs were conserved between them (Figure 1C). In contrast, HpFadD2 shared no similarity with EcFadD when analyzed using BLAST but was 22% identical to E. coli FadK, a short-chain acyl-CoA synthetase (Morgan-Kiss and Cronan, 2004). Interestingly, HpFadD2 was 25% identical to HpFadD1 and contained an ATP/AMP motif. This analysis indicates that FadD1 and FadD2 may have acyl-CoA synthetase activity.

We used an E. coli complementation system to determine whether FadD1 or FadD2 functions as an acyl-CoA synthetase. The E. coli JW1794 strain is a fadD mutant that cannot survive in minimal medium when exogenous fatty acid is the sole source of carbon (Baba et al., 2006). This deficiency in fadD can be restored by a plasmid carrying the gene encoding acyl-CoA synthetase. Hence, fadD1 and fadD2 were cloned into the arabinose-inducible vector pBAD24m to produce the expression constructs pSF226 and pSF227, respectively. These plasmids were then transformed into the E. coli strain JW1794. The resulting transformants were tested on M9 medium plates supplemented with different fatty acids as the sole carbon source.

An examination of the growth of the complementation strains incubated on M9 minimal medium plates revealed that the JW1794 strain carrying fadD1 grew on various fatty acids, whereas JW1794 carrying fadD2 grew only on C16:0 (Figure 2A). These strains grew only in the absence of the inducer arabinose, indicating that high levels of expression of either FadD may be toxic, consistent with previous findings (Bi et al., 2014). To confirm that this phenotype was consistent when these strains were grown in M9 broth, we determined the growth curves of these strains. We found that the HpfadD2 complementation strain survived when grown on minimal medium in which oleic acid was the sole carbon source, although it grew quite slowly compared with the HpfadD1 complementation strain (Figure 2B, the doubling times of the HpfadD1 and HpfadD2 complementation strains were 13.38 ± 2.16 h and 37.36 ± 12.51 h, respectively, P < 0.05). Altogether, these results indicate that FadD1 and FadD2 have acyl-CoA synthetase activity in vivo.

We produced His-tagged proteins to confirm the substrates of the two FACSs. FadD1 and FadD2 were expressed at high levels in E. coli grown in low-temperature induction conditions. The expression levels of the two enzymes were equivalent in whole cell protein extracts, but after ultrasonication and centrifugation, the lysed supernatant of the FadD2-expressing strain contained less of the target protein. Non-transparent inclusion bodies were not observed, indicating that FadD2 may be localized mainly in particulate fractions, as described in previous investigations that showed that EcFadD was located in both the soluble and particulate fractions of bacterial cells (Kameda and Nunn, 1981). The proteins were purified using nickel chelate chromatography in an AKTA explorer FPLC system. Target proteins were detected using SDS-PAGE (Figure 3A) and confirmed by mass analysis (Figure S1).

The fatty acid substrate specificity of each FadD was determined in vitro. FadD1 exhibited a high level of activity on C14:0, C16:0, and C18:1, with high V_max (78.42, 92.11, and 51.64 nanomoles of acyl-CoA formed/min/mg of protein, respectively) and low K_m values (41.48, 39.90, and 15.64 μM, respectively) when the substrate ranged from 30 to 300 μM (Figure 3B and Table 3). FadD2 showed clear activity only on C16:0 (17.78 nanomoles of acyl-CoA formed/min/mg of protein for V_max, 128.30 μM for K_m), and its level of activity was very low on C14:0 (1.45 nanomoles of acyl-CoA formed/min/mg of protein for V_max, 173.40 μM for K_m) (Figure 3C and Table 3). Moreover, oleic acid in different concentrations had distinct substrate inhibition effects on FadD2. A higher concentration of C18:1 resulted in lower FadD2 activity (Figure 3D). This may explain why the JW1794 strain carrying FadD2 could not grow on C18:1 (Figure 2A). The fact that FadD1 had higher k_cat/K_m values on all selected FAs indicated that FadD1 had a higher level of activity than FadD2 (Table 3). These results are in accordance with the E. coli complementation testing results (Figure 2). FadD1 and FadD2 are long-chain FACS, as they both have higher LCFA (long-chain fatty acid) and MCFA (medium-chain fatty acid) activity than SCFA (short-chain fatty acid, such as C8:0) activity (Figure 3E). Therefore, FadD2 is not a short-chain FACS as FadK is in E. coli, although they share some homology.

To determine the physiological functions of the two FACS in H. parasuis, fadD1, or fadD2 mutants were constructed using natural transformation, as described in the methods section (Figure 4). However, double mutants were not produced when the fadD1 disruption vector was transformed into the fadD2 mutant or when the fadD2 disruption vector was transformed into the fadD1 mutant. These transformation experiments were performed more than five times each; however, no positive colonies were obtained. Notably, the complementation strains of ΔfadD1 or ΔfadD2 were obtained as described in the methods section (Figure 4 and Figure S2). These results indicated that the natural transformation ability of the ΔfadD1 and ΔfadD2 strains was not significantly attenuated.

In a previous study, residue substitutions were introduced in the ATP/AMP signature motif of E. coli FadD to identify the specific residues that are critical for its activity (Weimar et al., 2002). To further determine whether fadD1 or fadD2 is essential for H. parasuis survival, we performed genomic site-directed mutagenesis of five conserved amino acids in the ATP/AMP motif of FadD1. Schematics for the constructed strains are shown in Figure 4A. All of the strains were confirmed by PCR and sequencing (Figure 4B). First, the fadD1^T214A, fadD1^G216A, fadD1^T217A, fadD1^G219A, and fadD1^E362A strains were constructed. Subsequently, all site-directed mutants were naturally transformed with pSF231 (a fadD2 disruption vector) to generate double-mutant strains. However, only fadD1^G216A–ΔfadD2 and fadD1^G219A–ΔfadD2 were obtained (Table 4). Moreover, the growth of E. coli JW1794 cells carrying plasmids encoding the mutant proteins (genes with point mutations contained in the expression vector pBAD24m) was also tested on oleate (Figure 4C). This assay demonstrated that FadD1^G216A or FadD1^G219A exhibited activity that was sufficient to support a similar level growth as that of JW1794 carrying wild-type FadD1 on minimal medium when oleic acid was the sole carbon source. Growth was summarized for of all of the complementation strains derived from the E. coli JW1794 fadD mutant in Table 5. In conclusion, these data indicated that a fadD2 mutant can be generated only when the strain has sufficient FadD1 activity. Hence, either fadD1 or fadD2 is required for the survival of this bacterium.

Next, the fadD1 and fadD2 mutants were tested to determine their physiological and virulence phenotypes. We tested the growth of the wild-type strain and the mutants in TSB medium containing bovine serum and NAD. Only negligible changes in growth rates were detected in cells grown in this rich medium (Figure 5A). An RT-PCR analysis of total H. parasuis RNA extracted from log-phase cells showed that the expression level of fadD2 was not significantly altered in the fadD1 mutant (Figure 5B, the relative expression levels of fadD1 in WT and ΔfadD2 were 0.71 ± 0.14 and 0.66 ± 0.10, respectively, P > 0.05). Similarly, the level of fadD1 transcription in the ΔfadD2 strain was equivalent to that in the wild-type strain (Figure 5C, the relative expression levels of fadD2 in WT and ΔfadD1 were 5.81 ± 0.58 and 5.31 ± 0.61, respectively, P > 0.05). When growth testing was performed using a chemically defined medium (CDM) containing oleic acid, the generation time for the fadD1 mutant was appreciably longer than that for the wild-type strain SC096 and its complementation strain (Figure 5D, the doubling times of SC096 and the ΔfadD1 and ΔfadD1 complementation strains were 3.31 ± 0.17, 4.98 ± 0.07, and 3.93 ± 0.25 h, respectively, P < 0.05). Smaller colonies of the fadD1 mutant grew on CDM agar, and these exhibited a phenotype consistent with the broth medium testing results. Additionally, no visible growth was detected when the wild-type strain was grown on CDM agar without fatty acids, indicating that acyl-CoA may be indispensable for the survival of H. parasuis (Figure 5E). Furthermore, the roles of the fadD genes in H. parasuis serum resistance were investigated. However, there was no significant difference in sensitivity between the wild-type strain and its mutants (Figure 5F, the survival rates of SC096, ΔfadD1 and ΔfadD2 in 50% porcine serum were 71.6 ± 6.6, 69.6 ± 4.9, and 71.9 ± 3.4%, respectively. The survival rates of SC096, ΔfadD1 and ΔfadD2 in 80% porcine serum were 23.6 ± 5.7, 24.5 ± 3.7, and 19.9 ± 5.1%, respectively, P > 0.05).

A summary of the MIC values of several antibiotics that were evaluated in this study is shown in Table 6. Interestingly, the fadD1 mutant was more susceptible to all of the selected quinolone antibiotics, including levofloxacin, enrofloxacin, nalidixic acid, and ciprofloxacin, whereas the fadD1 complementation strain exhibited a phenotype similar to the wild-type strain. Note that quinolone resistance can result from changes in bacterial membrane permeation (Neu, 1992). The free acid of nalidixic acid, for instance, is easily dissolved in chloroform but only slightly soluble in water, indicating that both the polarity of quinolone in the medium and the composition of phospholipids (i.e., the proportions of different fatty acids) may influence the diffusion of quinolone through the membrane. To analyze changes in phospholipids, we assayed the fatty acid composition of the cells using GC-MS. The phospholipid fatty acids in H. parasuis are composed mainly of palmitic acid, stearic acid, palmitoleic acid, myristic acid and oleic acid (Figure 6). However, no significant differences were observed in fatty acid composition between the wild-type strain and the mutants (Table 7). Hence, we propose that deleting fadD1 might affect the amount of all lipids rather than the overall composition.

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.