Experimental Identification of Actinobacillus pleuropneumoniae Strains L20 and JL03 Heptosyltransferases, Evidence for a New Heptosyltransferase Signature Sequence

We experimentally identified the activities of six predicted heptosyltransferases in Actinobacillus pleuropneumoniae genome serotype 5b strain L20 and serotype 3 strain JL03. The initial identification was based on a bioinformatic analysis of the amino acid similarity between these putative heptosyltrasferases with others of known function from enteric bacteria and Aeromonas. The putative functions of all the Actinobacillus pleuropneumoniae heptosyltrasferases were determined by using surrogate LPS acceptor molecules from well-defined A. hydrophyla AH-3 and A. salmonicida A450 mutants. Our results show that heptosyltransferases APL_0981 and APJL_1001 are responsible for the transfer of the terminal outer core D-glycero-D-manno-heptose (D,D-Hep) residue although they are not currently included in the CAZY glycosyltransferase 9 family. The WahF heptosyltransferase group signature sequence [S(T/S)(GA)XXH] differs from the heptosyltransferases consensus signature sequence [D(TS)(GA)XXH], because of the substitution of D261 for S261, being unique.


Introduction
Actinobacillus pleuropneumoniae is a non-motile Gram-negative bacterium causing porcine pleuropneumonia, a highly contagious respiratory disease transmitted through aerosols or close contact with infected animals including asymptomatic carriers [1]. This disease is often fatal and characterized by hemorrhagic, fibrinous and necrotic lung lesions; the clinical features ranging from acute to chronic [2], and it is an important cause economic losses worldwide in the porcine industry [2].
To date, two biovars have been described on the basis of NAD growth dependence; and fifteen serotypes based on capsular antigens [3,4]; the disease can be caused by any serotype, although differences in virulence have been described [5,6].
Studies aimed at the identification of A. pleuropneumoniae virulence factors showed that, as in other pathogenic bacteria, there is an array of these factors. However, the most relevant one appears to be the Apx toxins [5]. Other possible virulence factors include the capsule [7,8,9], outer membrane proteins involved in iron uptake [1,10,11], and lipopolysaccharides (LPS) [12,13]. Biofilm formation [14], autotransporter adhesion [15], and autotransporter protease synthesis [16] have been also described to contribute to the pathogenicity of this bacterium.
The LPS appears to play a role in virulence in different stages of the A. pleuropneumoniae infection, including adhesions to lower respiratory tract, induction of lesions, and persistence in the upper respiratory tract reviewed in [17]. In addition the LPS molecules appear to interact with ApxI and ApxII toxins [18].
Until now only one A. pleuropneumoniae heptosyltransferase was identified for a strain belonging to the serotype 1 [23]. Since the common heptasaccharide present in the core OS of the A. pleuropneumoniae studied strains is highly similar to the core OS of A. hydrophila [24] (Figure 1) we decided to use well defined core mutants of A-hydrophila AH-3 [25] producing surrogate acceptor LPS molecules to identify the heptosyltransferases involved in the core OS biosynthesis of genome strains belonging to serotypes 3 and 5b.

Bioinformatic and phylogenetic analysis
The core LPS structures of A. pleuropneumoniae studied strains contain between four and five heptose residues ( Figure 1) requiring an equal number of heptosyltransferases. These heptosyltransferases were predicted on the basis of the nature of the substrate heptose, either L,D-Hep or D,D-Hep, and the linkage to the corresponding substrate core residue. In order to attribute putative functions we performed a bioinformatic analysis based on the alignment (Clustal W) of heptosyltransferases whose function is experimentally proved with those of A. pleuropneumoniae strains L20 (serotype 5b) [26], JL03 (serotype 3) [27], and AP76 (serotype 7) [28]. After this alignment, we performed a phylogenetic analysis using the same heptosyltransferases whose function is experimentally proved and the ones obtained with A. pleuropneumoniae strains. The cladogram obtained is shown in Figure 2 as an indicative of the similarity degree among these proteins.
OpsX and WaaC transfer the first L,D-Hep residue to phosphorylated and unphosphorylated Kdo, respectively [25,29,30]. Indeed enzymes (APL_0904 and APJL_0916) for the phosphorylation at the 4-OH position of Kdo are also present in strain L20 or JL03, respectively. As expected these two heptosyltransferases, on the basis of amino acid similarity, cluster in two different branches when analyzed using Clustal W ( Figure 2). Aeromonas hydrophila strains ATCC 7966 and AH-3 and A. salmonicida A450 contain only the OpsX version (AHA_0042, OpsX-AH3 and ASA-0037) and as expected their core LPS oligosaccharides only contain phosphorylated Kdo [24]. The genome strain Pm70 of Pasteurella multocida contains both versions of heptosyltransferase I and analysis of the core LPS of strains Pm70 and VP161has shown that their core LPS contains two major oligosaccharides only differing in their Kdo phosphorylation [31,32]. No phosphorylated Kdo was reported in the core LPS of the four A. pleuropneumoniae strains studied [32] probably because in this study the core fraction was obtained by acid hydrolysis that could remove the phosphate from Kdo.  [24], A. salmonicida [33], and A. pleuropneumoniae serotypes 1, 5a, and 5b [22]. The genes involved in core LPS biosynthesis are shown for A. hydrophila and A. salmonicida [25,33]. doi:10.1371/journal.pone.0055546.g001 The same Clustal W analysis shows that known heptosyltransferases from Aeromonas hydrophyla and salmonicida transferring L,D-Hep II (WaaF) and L,D-Hep III (WahF) cluster with putative heptosyltransferases from A. pleuropneumoniae L20 (APL_1403 and APL_0421), and Pasteurella multocida Pm70 ( Figure 2) in agreement with the presence of a common trisaccharide of L,D-Hep in the inner core of these bacteria (Figure 1).
A fourth cluster encompasses A. hydrophila WahD (AHA_4226) and A. salmonicida WasD (ASA_0098) with A. pleuropneumoniae  Two proximal clusters includes WahC from A. hydrophila and APL_0981, APLJ_1001, and APP7_1037 from A. pleuropneumoniae genomic strains L20 (serotype 5b), J03 (serotype 3) and AP76 (serotype 7). WahC has been shown to be the heptosyltransferase responsible for the transfer of the terminal D,D-Hep residue of the A. hydrophila AH-3 core LPS [25] (Figure 1) and although their counterparts from A. pleuropneumoniae are actually included in a non-classified group of glycosyltransferases (GTnc, http://www. cazy.org) the present analysis could suggest a common function for all these proteins.
As a control for this analysis we used the WahE proteins from A. hydrophila and salmonicida involved in the transfer of a L,D-Hep residue by an a 1,6 linkage to L,D-Hep I [25,33] (Figure 1). As shown in Figure 2 no similar proteins from A. pleuropneumoniae or P. multocida fall in this cluster.

Functional identification of the two versions of A. pleuropneumoniae heptosyltransferase I
In order to unambiguously identify the functions of the above putative heptosyltransferases we decided to use LPS from several A. hydrophila AH-3 strain mutants as a source of surrogate LPS acceptor molecules. These mutants were previously constructed by either Km R insertion or internal in-frame deletion in such a way to avoid polar effects on downstream genes [25].
First, we established the function of APL_0186 by its expression under the control of the P ARA promoter (pBAD33-Gm-APL_0186). The core LPS from A. hydrophila AH-3DopsX mutant only contains a phosphorylated Kdo residue and is devoid of Oantigen [25]. The pBAD33-Gm-APL_0186 was introduced into A. hydrophila AH-3DopsX mutant and LPS from this strain was analyzed by SDS-PAGE after growth in inducing conditions. The APL_0186 expression plasmid fully complemented the opsX mutation leading to the production of full length LPS including O-antigen ( Figure 3A, lane 3). The same result was obtained when using APJL_0187 expression plasmid from A. pleuropneumoniae strain JL03 (serotype 3) ( Figure 3A, lane 4). These results strongly suggest that APL_0186 and APJL_0187 correspond to the OpsX version of heptosyltransferase I. By contrast, neither APL_1402 nor APJL_1427 were able to modify the LPS migration pattern of A. hydrophila AH-3DopsX mutant ( Figure 3A, lanes 5 and 6), suggesting that they could correspond to the WaaC version of heptosyltransferase I. In agreement with this hypothesis, expression of APL_1402 and APJL_1427 in K. pneumoniae 52145DwaaC [34] were able to restore the full length K. pneumoniae 52145 core LPS and O-antigen production ( Figure 3B, lanes 3 and 4). These results establish that A. pleuropneumoniae contains the two versions of heptosyltransferase I.

Heptosyltransferases II and III
The LPS from A. hydrophila AH-3DwaaF is devoid of O-antigen and contains a trisaccharide core [a-L,D-Hep-(1R6)-a-L,D-Hep-(1R5)-Kdo-P] [25]. This core structure is similar to the one expected from a A. pleuropneumoniae waaF mutant [a-L,D-Hep-(1R5)-Kdo-P] and it was expected that the putative WaaF enzymes APL_1403 and APJL_1428 would be able to recognized the A. hydrophila AH-3DwaaF core. As expected, expression of APL_1403 (pBAD33-Gm-APL_1403) or APJL_1428 (pBAD33-Gm-APJL_1428) into A. hydrophila AH-3DwaaF resulted in the production of full length A. hydrophila AH-3 core including Oantigen ( Figure 4, lanes 3 and 4). By contrast, expression of APL_0421 or APJL_0446 in this genetic background did not modify the mutant A. hydrophila AH-3DwaaF core LPS (Figure 4, lanes 5 and 6).

WahD
As mentioned above, the WahD from A. hydrophila AH-3 is involved in the transfer of the D,D-Hep residue linked to Glc by an a 1,6 linkage while WasD from A. salmonicida transfer the L,D-Hep to Glc using the same linkage ( Figure 1) [25,33]. As seen in Figure 2 these proteins cluster together and APL_0979 and APJL_0999 are more similar to Pasteurella multocida PM1144 than to Aeromonas WahD or WasD. It is difficult to infer the function solely from the protein clustering because the only expected difference between WahD and WasD homologues should be in the To further confirm the function of APL_0979 and APJL_0999, LPS from A. hydrophila AH-3DwahD, A. hydrophila AH-3DwahD + pBAD33-Gm-APL_0979, and A. hydrophila AH-3DwahD + pBAD33-Gm-APJL_0999, were isolated. These LPS samples were hydrolyzed by mild-acid treatment and the core OS fraction isolated by chromatography. MS analysis of the OS fraction from A. hydrophila AH-3DwahD showed a major signal at m/z 1311.44 corresponding to a core OS (Hex 1 HexN 1 Hep 4 Kdo 1 ) consistent with the loss of the outer-core trisaccharide fragment a-D,D-Hep-(1R6)-[b-D-Gal-(1R4)]-a-D,D-Hep from the full core LPS (Table 1). By contrast, MS analysis of the OS fraction from A. hydrophila AH-3DwahD + pBAD33-Gm-APJL_0999 showed a major signal at m/z 1857.617 (Table 1) in agreement with a full length AH-3 core (Hex 2 HexN 1 Hep 6 Kdo 1 ) ( Table 1) [25]. The second major signal at m/z 1503.498 was assigned to a core OS devoid of the terminal D,D-Hep and D-Gal from the wild type AH-3 core OS (Table 1). Similar results were obtained when analyzing the OS fraction from A. hydrophila AH-3DwahD + pBAD33-Gm-APL_0979 (data not shown). Taken together, these data allow to attribute a heptosyltransferase function involved in the transfer of the first D,D-Hep residue of the outer-core in A. pleuropneumoniae to APL-0979, and APJL_0999.

WahC
The A. hydrophila AH-3 WahC is responsible for the transfer of the terminal D,D-Hep residue of the AH-3 core LPS [25], and shows significant similarity to A. pleuropneumoniae APL_0981 and APJL_1001. No homologous protein from A. salmonicida was detected ( Figure 2) in agreement with the presence of terminal D,D-Hep in A. hydrophila but not in A. salmonicida (Figure 1). A major difference between these two groups of proteins is that while the AH-3 and ATCC 7966 WahC are assigned to the GT9 family in the CAZY classification of glycosyltransferases, the two A. pleuropneumoniae proteins are actually grouped as non-classified glycosyltransferases on the basis of weak similarity to established glycosyltransferase families. However, in the Kegg database, both of these proteins are shown as having a glycosyltransferase family 9 (heptosyltransferase) motif. On the basis of the presence of a terminal D,D-Hep in the A. pleuropneumoniae core OS of serotype 2, 5a, and 5b (genome strain L20) [22], we hypothesized that APL_0981 would be a WahC homologue. Furthermore, APL-0981 is identical to APP7_1037 (serotype 7), and shows only one amino acid residue different when compared with APJL_1001 (serotype 3) for which the core LPS structure remains unknown.
To test the above hypothesis APL_0981 and APJL_1001 were introduced and expressed into A. hydrophila AH-3DwahC. As previously shown, the LPS of the AH-3 mutant still contains Oantigen ( Figure 5, lane 6) and its core OS was devoid of the terminal D,D-Hep residue as demonstrated by MS analysis of its core OS fraction (Table 1) and methylation analysis [25]. A. hydrophila AH-3DwahC harboring pBAD33-Gm-APL_0981 or APJL_1001 recover full length core LPS since no differences in mobility of the lipidA-core band could be detected among LPS isolated from these strains and wild type AH-3 ( Figure 5, lanes 6,  7, and 8). Furthermore, MS analysis of the core LPS fraction isolated from A. hydrophila AH-3DwahC harboring pBAD33-Gm-APJL_1001 gave a major signal a m/z 1857.63 corresponding to full length core OS (Hex 2 HexN 1 Hep 6 Kdo 1 ), other signals were attributed to Hex 1 HexN 1 Hep 6 Kdo 1 (m/z 1695.573) and Hex 1 -HexN 1 Hep 5 Kdo 1 (m/z 1503.495) ( Table 1). Similar results were obtained with the core OS fraction from A. hydrophila AH-3DwahC harboring pBAD33-Gm-APL_0981 (data not shown). These results, strongly suggests that A. pleuropneumoniae APL_0981 and APJL_1001 are functional homologues of A. hydrophila WahD and are responsible for the transfer of the terminal outer-core D,D-Hep residue (Figure 1).

Discussion
The bioinformatic and phylogenetic analysis reported here shows that four L,D-heptosyltransferases (WaaC, OpsX, WaaF, and WaaQ) can be grouped on the basis of their amino acid similarity ( Figure 2 and Table 2), while the heptosyltransferase WahE would constitute a more distant group. The two other L,Dheptosyltransferases WahF and WasD (ASA 4222) appear to be related to D,D-heptosyltransferases (WahD and WahC) (see cladogram in Figure 2). It is not clear that differences between residues found to be important in WaaC with those in other heptosyltransferase are specifically significant for the catalytic function and interaction with ADP-L,D-Hep of E. coli heptosyltranferase I (WaaC) [35]. The key residues for the catalytic function and interaction with ADP-L,D-Hep were D 13 , K 192 , E 222 , D 261 , and H 266 , the last two (D 261 and H 266 ) present in the proposed heptosylytansferases signature sequence D 261 (T/S)(A/ G)XXH 266 . These conserved residues were shown to interact with ADP-2-deoxy-2-fluoro-heptose [35]. The only residues that were found conserved in all the heptosyltransferases analyzed in this study ( Table 2) were D 13 and H 266 . The K 192 was found to interact  with the O5 and O6 atoms of the heptose, and it was present in L,D-heptosyltransferases WaaC, OpsX, WaaF, and WaaQ. It has been suggested that in D,D-heptosyltransferases the K 192 could be substituted by other residues as by R in Helicobacter pylori [35,36]. This is also the case for D,D-heptosyltransferases WahD and WahC where K 192 is substituted by R or N or R, respectively. We also found K 192 in all the analyzed L,D-heptosyltransferases except WahF with N or T substituting K 192 ( Table 2). The E. coli WaaC E 222 residue, which interacts with the O2 and O3 residues of the heptose [35], was found in heptosyltransferases WahE, WahF, and WahD, being substituted by D in WaaQ and by other residues in OpsX, WaaF, WasD, and WahC (Table 2). E. coli WaaC D 261 and H 266 residues were previously shown to interact with O3 of the heptose and in addition H 266 contacts with the 2'' fluorine of the ADP-2-deoxy-2-fluoro-heptose [35]. They are found in all the heptosyltransferases analyzed in this work with the exception of WahF where D 261 is substituted by S ( Table 2 Table 2. This analysis stresses the importance of experimental identification of the function of hypothetical heptosyltransferases. Two approaches are possible; one is based on the construction of individual mutations and analysis of the core LPS defects [25]; and the second one, uses surrogate LPS molecules to test the function of individual heptosyltransferases. The first approach can be difficult in some bacterial species and has the problem that the  Table 1. Negative ion charge deconvoluted electrospray ionization mass spectral data of the core oligosaccharides released by mild acid hydrolysis from LPS of A. hydrohila AH-3DwahD, A. hydrohila AH-3DwahD (PBAD33-Gm-APJL_0999), A. hydrohila AH-3DwahC, and A. hydrohila AH-3DwahC (pBAD33-Gm-APJL_1001). conclusions are based on a negative feature represented by the shortage of the wild type core, in addition sometimes the interpretation of data is complicated by the presence of different core OS in the mutant strains. The second approach requires having surrogate LPS molecules highly similar to the predicted heptosyltransferase LPS acceptor molecule. The LPS surrogate acceptors would be provided by previously obtained mutants from other bacteria, where the acceptor (mutant strains) should be constructed in such a way that there are no downstream effects. The use of surrogate LPS molecules allows for a positive trait: the addition of a new residue to surrogate core LPS, and if adequate the analysis is facilitated because the modified surrogate core LPS will acquire also the corresponding O-antigen domain (Figures 3,  4, and 5) and further confirmed by composition and MS analysis ( Table 1). In this work we show the feasibility of this approach by identifying the A. pleuropneumoniae heptosyltransferases from strains L20 and JL03. The same functions are expected from the homologous proteins from strain AP76 ( Figure 2). Furthermore, our results based in the use of surrogate LPS acceptor from an A. hydrophila DwahC mutant ( Figure 5, lanes 6, 7, and 8) and MS analysis (Table 1) allow to establish unambiguously an heptosyltransferase function for APL_0981 and APJL_1001 although these two proteins are not included in the glycosyltransferase GT9 family in the CAZY classification. These two proteins are involved in the addition of the terminal D,D-Hep residue of the A. pleuropneumoniae core LPS (Figure 1), and then besides the sequence similarity should be included as GT9 family members.
There is an increasing interest in the development of antivirulence agents and one of the potentially interesting targets for them is the WaaC heptosyltransferase I [37]. Our and previous analyses of the heptosyltransferases involved in LPS biosynthesis clearly show that the most conserved region corresponds to the residues involved in catalytic function and interaction with the ADP-L,D-Hep. Thus, the search for antivirulent inhibitors of the WaaC and other heptosyltransferases interacting with these residues appears be the best way to proceed.

Bacterial strains, plasmids, and growth conditions
Bacterial strains, and plasmids used in this study are listed in Table 3. Aeromonas strains were routinely grown on tryptic soy broth (TSB) or tryptic soy agar (TSA) at 30uC unless stated otherwise. A. pleuropneumoniae strains were grown on Brain Herat Infusion (BHI) at 37uC and E. coli strains were grown on Luria-Bertani (LB) Miller broth and LB Miller agar at 37uC [38]. Kanamycin (50 mg/ml), rifampicin (100 mg/ml), or gentamycin (20 mg/ml) were added to the different media.

General DNA methods and computer analysis of sequence data
General DNA manipulations were done essentially as described [39]. DNA restriction endonucleases, T4 DNA ligase, E. coli DNA polymerase (Klenow fragment), and alkaline phosphatase were used as recommended by the suppliers. Deduced amino acid sequences used from different Enterobacteriaceae and Aeromonadaceae were obtained from the GenBank, EMBL, and SwissProt databases. ClustalW [40] was used for multiple-sequence alignments with a gap open penalty of 10, gap extension penalty of 0.05, no weight transition, with hydrophilic gaps and GPSNDQERK hydrophilic residues for proteins, and BLOSUM as selected weight matrix.

Plasmid constructions and mutant complementation studies
For complementation studies, the A. pleuropneumoniae genes (APL_0186, APL_0421, APL_0979, APL_0981, APL_1402, APL_1403, APJL_0187, APJL_0446, APJL_0999, APJL_1001, APJL_1427, and APJL_1428) were PCR-amplified by using specific primer pairs (Table 4) and ligated to the plasmid pBAD33-Gm. The plasmid constructions were transformed into E. coli LMG194 by electroporation, plated on gentamycin LB agar plates and incubated at 37uC. Plasmids with the amplified genes were independently transferred into the corresponding mutants by triparental mating using the mobilizing strain HB101/pRK2073. Transconjugants were selected on plates containing gentamycin and rifampicin and confirmed by PCR. Each gene was expressed from the arabinose-inducible and glucose-repressible pBAD33 Table 2. The presence of E. coli WaaC enzyme key residues for the catalytic function and interaction with ADP-L,D-Hep [35] in well characterized heptosyltransferases in different Gram-negative bacteria (Enterobacteriaceae and Aeromonadaceae). Comparison of the identified E. coli WaaC enzyme key residues based in amino acid alignment. promoter. Repression from the araC promoter was achieved by growth in medium containing 0.2% (w/v) D-glucose, and induction was obtained by adding L-arabinose to a final concentration of 0.2% (w/v). The cultures were grown for 18 h at 30uC in TSB medium supplemented with gentamycin and 0.2% glucose, diluted 1:100 in fresh medium (without glucose) and grown until they reached A 600 nm of about 0.2. Then L-arabinose was added, and the cultures were grown for another 2 h. Repressed controls were maintained in glucose-containing medium.

Preparation of oligosaccharide fraction
A portion of the LPS (,50 mg) from each strain was heated with aqueous 2% HOAc (6 mL) at 100 uC for 45 min. The precipitate was removed by centrifugation (13,000 g, 20 min) and the supernatant fractionated on a column (5662.6 cm) of Sephadex G-50 (S) in 0.05 M pyridinium acetate buffer pH 4.5 with monitoring using a differential refractometer (Knauer, Germany).

Mass spectrometry
Electrospray ionization mass spectra were run in the negative ion mode using a micrOTOF II instrument (Bruker Daltonics, USA). Mass spectra were acquired using standard experimental sequences as provided by the manufacturer. Samples (,50 ng mL -1 ) were dissolved in a 1:1 (v/v) water-acetonitrile mixture and sprayed at a flow rate of 3 mL min -1 . Capillary entrance voltage was set to 4.5 kV and exit voltage to 2150 V; drying gas temperature was 180uC. The spectra showing several charge states for each component were charge deconvoluted, and mass numbers given refer to monoisotopic molecular masses.

Phylogenetic analysis
The analysis was performed by running the phylogenetic inference software Protpars from the PHYLIP (Phylogeny Inference Package) package version 3.5c (Department of Genetics University of Washington, Seattle) [47], using heptosyltransferases whose function is experimentally proved from enteric bacteria, Aeromonas or Pasteurella multocida with those of A. pleuropneumoniae strains L20 (serotype 5b) [26], JL03 (serotype 3) [27], and AP76 (serotype 7) [28]. The cladogram is indicative by the separation and the large of the lanes from the values obtained running the program (the scale bar indicates an evolutionary distance of 0.1 aminoacid substitutions per position).