Identification and characterization of serovar-independent immunogens in Actinobacillus pleuropneumoniae

Despite numerous actions to prevent disease, Actinobacillus pleuropneumoniae (A. pleuropneumoniae) remains a major cause of porcine pleuropneumonia, resulting in economic losses to the swine industry worldwide. In this paper, we describe the utilization of a reverse vaccinology approach for the selection and in vitro testing of serovar-independent A. pleuropneumoniae immunogens. Potential immunogens were identified in the complete genomes of three A. pleuropneumoniae strains belonging to different serovars using the following parameters: predicted outer-membrane subcellular localization; ≤ 1 trans-membrane helices; presence of a signal peptide in the protein sequence; presence in all known A. pleuropneumoniae genomes; homology with other well characterized factors with relevant data regarding immunogenicity/protective potential. Using this approach, we selected the proteins ApfA and VacJ to be expressed and further characterized, both in silico and in vitro. Additionally, we analysed outer membrane vesicles (OMVs) of A. pleuropneumoniae MIDG2331 as potential immunogens, and compared deletions in degS and nlpI for increasing yields of OMVs compared to the parental strain. Our results indicated that ApfA and VacJ are highly conserved proteins, naturally expressed during infection by all A. pleuropneumoniae serovars tested. Furthermore, OMVs, ApfA and VacJ were shown to possess a high immunogenic potential in vitro. These findings favour the immunogen selection protocol used, and suggest that OMVs, along with ApfA and VacJ, could represent effective immunogens for the prevention of A. pleuropneumoniae infections in a serovar-independent manner. This hypothesis is nonetheless predictive in nature, and in vivo testing in a relevant animal model will be necessary to verify its validity. Electronic supplementary material The online version of this article (10.1186/s13567-017-0479-5) contains supplementary material, which is available to authorized users.


Introduction
Actinobacillus pleuropneumoniae (A. pleuropneumoniae) is a Gram-negative bacterium causing respiratory disease in pigs worldwide. A. pleuropneumoniae outbreaks are a constant threat to the pig farming industry, and are estimated to reduce the expected revenues in pig production by an average of 6.4€ per pig in affected herds [1]. To date, A. pleuropneumoniae has been divided into 16 serovars and 2 biovars, according to capsular antigens and nicotinamide dinucleotide (NAD) requirements [2][3][4], respectively. The spatiotemporal prevalence of individual serovars has been shown to vary, both regionally and over time [5].
The first generation of vaccines developed against A. pleuropneumoniae was based on inactivated whole cells (bacterins). Although in some cases effective in preventing both colonization and morbidity, these vaccines offered protection against little more than the specific serovar used for immunization [6,7]. A step forward was made with the characterization of the Apx toxins, a set of four pore-forming cytolysins (Apx I-IV) centrally involved in A. pleuropneumoniae pathogenesis, and secreted by all A. pleuropneumoniae serovars in various combinations [3,8]. Vaccines based on Apx toxins elicit a strong humoral response and are able to effectively reduce morbidity in vaccinated animals [9-12], but fail in preventing colonization and thus create the risk of the Open Access *Correspondence: miki@sund.ku.dk 1 Department of Veterinary and Animal Sciences, University of Copenhagen, Stigbøjlen 4, 1870 Frb. C., 1-20, Building: 301, Copenhagen, Denmark Full list of author information is available at the end of the article spreading of disease by asymptomatic carriers [6,11,13]. Nonetheless, vaccines based on Apx toxins in combination with various outer membrane (OM) proteins (see [6] for a complete review of A. pleuropneumoniae OM immunogens) are the most widely used for the prevention of A. pleuropneumoniae infection. Thus, despite the availability of vaccines for A. pleuropneumoniae, their efficacy is challenged by limited prevention of host colonization [10,14] or cross-serovar protection [11,13,15]. Hence a major challenge impeding development of an effective vaccine has been to identify immunogens capable of preventing colonization of the host by all known A. pleuropneumoniae serovars as well as the severity of the disease.
The selection of immunogens for vaccine development has for decades been a rather laborious process, usually requiring the isolation and testing of hundreds of potential immunogens from cultures with subsequent identification of the genes involved in their expression, and finally production of the chosen immunogens for in vitro and in vivo testing [16]. The advent of bioinformatics and the sequencing of bacterial genomes have allowed researchers to invert this workflow and identify instead potential immunogens at a genetic level. Once identified, candidate immunogens can be easily expressed using suitable vectors for subsequent testing. This alternative workflow has quite fittingly been called reverse vaccinology (RV) [17,18], and has offered immunologists the opportunity to significantly reduce the workload traditionally necessary for vaccine development [19].
Similar to RV, bacterial outer membrane vesicles (OMVs) represent a revolution in the way non-live vaccines are now conceived and developed. The OMVs are vesicles secreted by the majority of Gram-negative bacteria that serve numerous biological functions, from cargo delivery to immune modulation [20,21]. The OMVs are secreted during all phases of bacterial growth, and vary in size between 20 and 300 nm for Gram-negative bacteria [22,23]. Secretion of OMVs has been proven to be negatively regulated by various genes, inactivation of which often leads to an enhanced secretion of OMVs [24]. Being released from the bacterial OM, OMVs generally represent antigenic properties very similar to intact bacteria. In addition, OMVs have a significant adjuvant potential due to the presence of lipopolysaccharides (LPS) and immunomodulatory proteins on their surface [25,26]. The relatively high molecular mass of OMVs also allows for more cost-effective purification methods, as compared to recombinant antigens [27], leading to lower vaccine production costs [28,29]. Additionally, the structure and biogenesis of OMVs offer the possibility to engineer them in order to add desired antigens, either to their surface or as cargo [30,31]. However, immunization with OMVs also presents some risks. The presence of LPS on OMVs, for example, could pose a threat to the vaccinated individual, potentially causing Systemic Inflammatory Response Syndrome (SIRS) after immunization [32].
Both OMVs and proteins identified using RV are considered promising vaccines commodities, and have so far been utilized commercially in the widely used vaccine against Neisseria meningitidis [33]. Here we employ a similar methodology for the identification of serovar-independent immunogens as a first step in the development of a broadly protective vaccine against A. pleuropneumoniae.

Bacterial strains, genomes, and sequence analysis
The A. pleuropneumoniae strains and accession numbers for genomes used in this investigation are listed in Table 1. Unless otherwise stated, all sequence analysis programmes were used with default parameters. The pan-proteomes extracted from genomes of A. pleuropneumoniae JL03, L20 and AP76 strains (serovars 3, 5b, and 7, respectively) were analysed by PSORTb v3.0 [34]. Proteins whose subcellular localization was predicted as "OuterMembrane" with a score > 9.00/10.00 were included in a preliminary candidate immunogen list (Additional file 1). The conservation of candidate immunogen genes across all published A. pleuropneumoniae genomes (those available in Genbank as of 14/02/17) was verified using BLASTn. Minimum and maximum conservation rates were noted in the preliminary candidate immunogen list (Additional file 1), as was the predicted presence or lack of a signal peptide, as detected by Sig-nalP 4.1 [35], and the number of transmembrane helices (TMHs), as predicted by the TMHMM Server v. 2.0 [36].
Candidates with > 1 predicted TMH were discarded from further analysis. Preliminary candidate immunogens were screened for available data regarding their biological function and/or immunogenicity, narrowing down the initial list to 9 candidate immunogens ( Table 2). Finally, ApfA and VacJ were chosen for expression and characterisation. The ApfA and VacJ three-dimensional protein structures were predicted using the I-TASSER server [37]. Highest confidence models were chosen as predicted structures (ApfA model C-score: −0.40; VacJ model C-score: −3.74).

Construction of ΔdegS & ΔnlpI mutants
The A. pleuropneumoniae serotype 8 MIDG2331 strain was engineered for the construction of individual ΔnlpI and ΔdegS mutants. The nlpI or degS genes of the wild type (wt) strain were each replaced with a synthesised (Eurofins Genomics) trimethoprim resistance cassette consisting of the drfA14 gene from plasmid pM3389T [38] under control of the A. pleuropneumoniae sodC promoter [39], and followed by the 9 bp sequence required for efficient uptake of DNA during natural transformation of A. pleuropneumoniae [40]. Gene replacement constructs were designed by In-Fusion cloning, with the dfrA14 cassette flanked by approx. 500 bases of sequence upstream and downstream of the gene to be replaced, as previously described [41]. Sequences were amplified using AccuPrime Taq DNA high-fidelity Polymerase (ThermoFisher Scientific). Briefly, primers (degS_left_for + degS_right_rev; nlpI_left_for + nlpI_ right_rev; Table 3) were designed to amplify the left and right sequences flanking the gene to be deleted, with 15 bp extensions to allow directional cloning into the linear vector pJET1.2/blunt (ThermoFisher Scientific). The resulting plasmids, pJET-degS and pJET-nlpI were opened by inverse polymerase chain reaction (PCR), using degS_left_rev + degS_right_for and nlpI_left_ rev + nlpI_right_for primer pairs ( Table 3). The dfrA14 cassette was amplified and cloned into the inverse-amplified pJET-degS and pJET-nlpI plasmids, using primers with 15 bp extensions matching the ends of the linearized constructs (dfrA14_degS_for + dfrA14_degS_rev; dfrA14_nlpI_for + dfrA14_nlpI_rev; Table 3). The resulting plasmids, pJETΔdegS::dfrA14 and pJETΔnlpI::dfrA14 were linearized by digestion with FastDigest XbaI restriction enzyme (ThermoFisher Scientific) prior to natural transformation into A. pleuropneumoniae serotype 8 strain MIDG2331 [42], as previously described [41]. Transformants were selected on BHI (brain-heart infusion)-NAD containing 10 μg/mL trimethoprim, and the gene replacements were confirmed by PCR using primers designed to verify the location of the dfrA14 cassette insertion (degS_test + dfrA14_test_for; nlpI_ test + dfrA14_test_rev; see Table 3).

OMV isolation and analysis
For isolation of OMVs from A. pleuropneumoniae MIDG2331 wt, ΔdegS and ΔnlpI clones, 600 mL of BHI-NAD broth was inoculated for each with a 1/10 volume of overnight (ON) culture (to an initial OD 600 = 0.1 (OD: optical density) and incubated at 37 °C/200 rpm for Samples were analysed by transmission electron microscopy (TEM). Carbon-coated laminae were suspended into the OMV solutions for a few seconds and soaked with a droplet of 2% (w/v) phosphotungstic acid. Grids were soaked into a Sellotape/Chlorophorm solution to increase adherence and then placed on the laminae. After 1-2 min, the Grids were removed and left to dry on filter paper. Once dried, stained grids were examined using a Philips CM100 Transmission Electron Microscope (Philips). Additionally, aliquots of the OMV samples were quantified by tunable resistive pulse sensing (TRPS) using a qNano device (Izon Sciences Ltd) following the protocol recommended by the operator [44]. Data were analyzed using the proprietary data capture and analysis software, Izon Control Suite V.3.3.2.2001.

Expression and purification of recombinant ApfA and VacJ
The apfA and vacJ genes were amplified from genomic DNA using Phusion Hot Start II DNA high-fidelity Polymerase (ThermoFisher Scientific) and cloned into the pET-44 Ek/LIC Vector (Merck Millipore) by In-Fusion directional cloning (Clontech). Briefly, primers were designed to open the vector by inverse PCR (pET44_apfA_rev + pET44_apfA_for), and to amplify the genes (apfA_pET44_for + apfA_pET44_rev; vacJ_ pET44_for + vacJ_pET44_rev; see Table 3). The gene specific primers did not include the start or stop codons, as these were supplied by the vector. For all primers, 15 bp extensions were added to the 5′ ends, providing reciprocal regions of complementarity (30 bp total) for In-Fusion cloning. The In-Fusion reactions were initially transformed into Stellar competent cells (Clontech) and verified before being transformed into E. coli BL21(DE3) pLysS competent cells, with selection on BHI agar supplemented with 100 µg/mL Ampicillin. Selected clones were screened by PCR and sequencing, and plasmids with the correct inserts were designated pApfA expr n and pVacJ expr n (n = 1-3).
For large-scale purification of expressed ApfA and VacJ proteins, the appropriate clones were inoculated into 800 mL of pre-warmed 2YT-Amp broth, using a 1/10 dilution of an ON starter culture, and incubated at 37 °C, 150 rpm, until OD 600 reached 0.5-08. At this point, the incubation temperature was shifted to 20 °C, and 20 min later protein expression was induced by addition of 400 μL of 1 M IPTG The recombinant proteins were isolated from the supernatants by immobilized ion metal affinity chromatography (IMAC) using the ÄKTAxpress chromatography system (GE Healthcare). Supernatants were loaded and run on HisTrap SP-HP 5 mL columns (GE Healthcare) following the manufacturer's recommendations. The extinction coefficients of the recombinant proteins were calculated using the ExPASy ProtParam tool [45] and uploaded to the ÄKTAxpress managing program, in order to identify (by OD 280 absorbance) the fractions containing proteins after purification. Selected protein fractions were pooled on ice, transferred to a 6-8 kD dialysis membrane (Spectra/por), and dialysed ON in 4 l PBS. Protein profiles of the purified fractions were compared to that of the initial filtered supernatant by electrophoresis under denaturing conditions using NuPAGE ™ (PAGE: polyAcrylamide gel electrophoresis) 4-12% Bis-Tris Protein sodium dodecyl sulphate (SDS) Gel (Ther-moFisher Scientific), run for 1 h at 180 V, and stained with Coomasie Blue.
Tags were removed from the purified proteins by cleavage with human plasma thrombin (Merck Millipore)(~2 U thrombin/μg of protein, 37 °C, ON), and separation by size exclusion chromatography (SEC). Proteins were loaded and run on Superdex 75 PG columns (GE Healthcare) using the ÄKTAxpress chromatography system. Fractions containing ApfA and VacJ proteins were identified, collected and dialyzed as described above.

Analysis of serum samples by ELISA
Control pig serum samples were obtained from Danish pig herds participating in the specific pathogen free (SPF) system. All herd are tested for A. pleuropneumoniae serovars 1, 2, 5, 6, 7, 10 and 12, respectively, using ELISA (enzyme-linked immunosorbent assay) methods previously described [46]. Convalescent sera from pigs experimentally infected with A. pleuropneumoniae serotype 2 HK361 were also used. All animals were otherwise classified as pathogen-free. Details regarding all sera used in this study are provided in Table 4.
Antibody (IgG) responses to purified ApfA and VacJ were determined by ELISA. Briefly, 96-well flat bottom plates (TPP) were coated with 100 ng of protein/well in PBS, and incubated at 4 °C ON. Pig sera (negative controls and samples known to be positive for serovars 2, 6 and 12, respectively) were diluted threefold between 1:500 and 1:1 093 500 in diluent buffer, and incubated for 1 h at room temperature (RT). Bound antibodies were detected with peroxidase labelled rabbit anti-pig IgG (ThermoFisher Scientific) at a 1:15 000 dilution, followed by colorimetric detection (OD 450 ) with TMB2 PLUS2 substrate (Kem-En-Tec Diagnostics), according to the manufacturer's protocol. A list of the non-commercial buffers used is provided in Additional file 2. The four data sets (Control, A. pleuropneumoniae serotype 12, 2, 6) from ApfA and VacJ ELISA assays were transformed to log10 values and analysed by D' Agostino and Pearson ominubus normality test to verify the assumption of Gaussian distribution of the data sets (alpha = 0.05). Data sets were subsequently converted to area under curve (AUC) values and analysed by Ordinary one-way ANOVA (P < 0.05) to verify their statistical significance. When overall statistical significance was achieved by ANOVA, data were analysed by post hoc Holm-Sidak test to assess statistical differences between individual groups and controls. P < 0.05 were again considered statistically significant. All statistical analyses were performed using GraphPad Prism 6.0c (GraphPad Software, La Jolla California USA). A summary of the statistical results obtained is provided in Additional file 3.

2D SDS-PAGE and Western blot
Proteins from OMV isolates of A. pleuropneumoniae 2331 wt and ΔnlpI mutant, containing equal number of vesicles, were purified by Wessel-Flüge extraction [47]. Prior to IsoElectric Focusing (IEF), immobilized pH gradient (IPG) strips (pH 4-7, 7 cm, GE Life Sciences) were rehydrated in DeStreak rehydration solution (GE Life Sciences)(1.5% IPG buffer, pH 4-7, ON). The samples were applied using the cup-loading method and IEF was performed at a total of 7 kVh per strip on an Ettan IPGphor isoelectric focusing unit (GE Life Sciences). IPG strips were subsequently reduced, alkylated and proteins were separated in duplicates by SDS-PAGE as a second dimension using NuPAGE 4-12% Bis-Tris ZOOM gels (Invitrogen) in running conditions as described above

In silico selection of conserved immunogens
A list of all potential immunogens initially identified is shown in Additional file 1. Only proteins predicted to be localised in the OM, and having ≤ 1 TMH, were included; likely secreted candidates were discarded. Details regarding predicted function and immunogenicity (based on published data) were used to narrow down the list of potential immunogens to 9 candidates (Table 2). From the 9 candidates, ApfA and VacJ proteins were selected for expression and in vitro characterization.

Preparation of recombinant proteins and OMVs
Both apfA and vacJ were successfully cloned into the pET-44 Ek/LIC vector, resulting in plasmids pET-44_ apfA and pET-44_vacJ. Expression of protein from E. coli strains harbouring these plasmids yielded a total of 46 mg and 18 mg protein for ApfA and VacJ, respectively, prior to tag removal. Following tag removal, two proteincontaining fractions were isolated (Figure 1 The OMVs isolated by HF from wt, ΔdegS and ΔnlpI mutants were analysed by TEM ( Figure 2) and precisely quantified and measured by TRPS analysis ( Figure 3). Both ΔdegS and ΔnlpI mutants showed increased OMV secretion as compared to the wt. The size distribution of the OMVs was also affected, with a wider variation in diameter range recorded for both mutants (Figure 3).

Assessment of in vitro immunogenicity
Convalescent sera from animals naturally infected with A. pleuropneumoniae serovars 2, 6 or 12, respectively, showed a significantly higher median IgG response against both ApfA and VacJ than the control sera from uninfected pigs (Figure 4). Statistical analysis of the data sets showed the significance of the differences observed between groups in IgG levels for both ApfA and VacJ (P < 0.05; Additional file 3). Similarly, the Western blot (WB) analysis with sera from animals experimentally infected with A. pleuropneumoniae HK361 (serovar 2) showed a strong IgG response to both wt and ΔnlpI OMVs, and the presence of several particularly immunogenic proteins ( Figure 5, only ΔnlpI OMVs shown).

Discussion
The development of new vaccines is predicted to become increasingly important in the fight against bacterial pathogens, due to the emergence of multidrug-resistant strains, and the dwindling number of new antimicrobials released on the market during the recent decades [48]. Accordingly, several new strategies for the identification of potential immunogens have been proposed and implemented [16]. Most of these strategies make use of the wide range of genome sequences and bioinformatic tools, and are sometimes collectively characterized as "immunoinformatics" [49]. Despite their obvious hypothetical nature, these computational predictions and in silico analyses have proven effective in the development of at least one commercial vaccine [33]. Subcellular localization is intuitively important when selecting potential immunogens. Intracellular or periplasmic antigens are not highly accessible to the host immune system, and usually do not play a significant role in the development of adaptive immunity against extracellular bacteria [50]. On the other hand, secreted proteins are usually highly immunogenic and quite often centrally involved in the bacterial pathogenesis (e.g. toxins), but being detached from the cell they fail in many instances in providing protection against bacterial colonization when administered as vaccines [13,15,51]. These and other previously described considerations led us to focus our analysis on OM proteins. Some of the identified candidates are A. pleuropneumoniae proteins involved in iron acquisition from the host (see Table 2), and limiting access to free iron has been proven to severely affect bacterial growth during infection [52,53]. A. pleuropneumoniae represents no exception, and for instance the inactivation of the hgbA gene has been linked to reduced virulence in vivo [54]. PepN is another interesting candidate immunogen in our list, a protease able to degrade porcine IgA and IgG [55]. It is intuitive to understand how targeting antibody degradation could be beneficial in reducing A. pleuropneumoniae resistance to the host immune system. However, the most promising vaccine candidate antigens that we chose to investigate in this study were ApfA and VacJ.
Type IV fimbrial subunit protein ApfA is a well characterised virulence factor in A. pleuropneumoniae, whose potential as an immunogen has been preliminarily investigated in swine and mouse models of infection [11, 56,57]. ApfA is an OM pilin involved in the early stages of host colonization, and its operon has been proven to be overexpressed during cell adhesion in vitro [58]. Cell adhesion represents one of the first steps involved in the colonisation of host surfaces by a pathogenic bacterium [59], providing an ideal target for immunisation of the host with a view to blocking infection. To date, ApfA's potential as a protective immunogen remains controversial, with some data suggesting it may be able to confer partial protection [11,56], while others suggest ApfA may interfere with the development of an adaptive response when included in a multicomponent vaccine formulation [57]. Our data confirmed ApfA conservation among all known A. pleuropneumoniae serovars. Furthermore, we demonstrated that ApfA is not only expressed by A. pleuropneumoniae in vitro or during a challenge in an induced animal model, but also during a natural infection in pigs, eliciting a rather marked humoral immune response in animals infected by at least 3 different A. pleuropneumoniae serovars not previously investigated. Finally, protein domain profiling of ApfA allowed us to identify a highly conserved helical domain, common to many pilins across species, that could be potentially used as an OM "anchor" for the construction of chimeras for antigen enrichment of the bacterial OM or OMVs.
VacJ (virulence-associated chromosome locus J) is an OM lipoprotein partially conserved among several Gram-negative pathogenic bacteria, including A. pleuropneumoniae. VacJ is part of a proposed phospholipid transporter (VacJ/Yrb ABC transporter), the inactivation or deletion of which has recently been shown to increase OMV secretion in some species [60]. In A. pleuropneumoniae, VacJ has been shown to be involved in OM integrity, serum resistance and biofilm formation [61]. As its name suggests, VacJ has been proposed as potential immunogen, and proven necessary for intracellular motility of Shigella flexneri [62]. Furthermore, the protective potential of a VacJ immunization has been described in at least one case, for Pasteurella multocida [63], whose VacJ protein shares 59% homology with the A. pleuropneumoniae homolog. As for ApfA, our data have shown that VacJ is highly conserved, expressed by A. pleuropneumoniae during a natural infection, and able to induce a specific immune response in pigs.
The potential use of OMVs as immunogens has been extensively investigated in several bacterial species, including other Pasteurellaceae [64]. Bacterial OMVs have been shown to contain most of the antigens embedded in the OM, including LPS, toxins and other surface antigens [20]. Most notably, A. pleuropneumoniae OMVs were previously shown to carry Apx toxins and proteases [65]. Our data provides further insight into the immunogenicity of A. pleuropneumoniae OMVs, with WB profiles indicating the presence of several highly immunogenic antigens, recognised by sera from animals infected with A. pleuropneumoniae serovars different than that used for isolation of the OMVs. The identity of these antigenic determinants is yet to be clarified, but it is likely that they include conserved proteins and virulence factors in addition to the Apx toxins and proteases previously described [63]. Furthermore, the choice of HF as purification technique allowed the concentration of large volumes of bacterial cell-free culture supernatants, increasing OMV yield when compared to gel filtration and gradient centrifugation methods, traditionally employed in OMV isolation [27]. The ability to purify large batches of OMVs may prove very useful in the process of standardizing immunization protocols, where a large number of animals need to be vaccinated with the same sample.
In conclusion, our data support the potential of in silico identification of immunogens, and confirm the predicted high immunogenicity of A. pleuropneumoniae OMVs. Both ApfA and VacJ were shown to be expressed during infection by all the A. pleuropneumoniae serovars tested and thus may offer the long sought cross-serovar immunological potential. Accordingly, we hypothesize that OMVs, as well as ApfA and VacJ proteins, could be administered individually or as a combined vaccine for prevention of infection in pigs against all known A. pleuropneumoniae isolates. Nonetheless, in silico and in vitro analysis can only provide predictive information regarding the immunogenic potential of ApfA, VacJ and OMVs. In some instances in fact OM proteins have been described to produce a deleterious effect when employed as immunogens to vaccinate pigs [66], and accordingly in vivo testing in a relevant animal model (swine) will be required to verify the effective potential of ApfA, VacJ and OMVs as immunogens.