Treatment of A. pleuropneumoniae grown under iron restriction with the mild detergent sodium deoxycholate has been introduced as a suitable method for enrichment of antigenic outer membrane-associated lipoproteins . A preparation of proteins obtained by this method derived from apxIIA deletion mutants of A. pleuropneumoniae serotypes 1, 2 and 5 was used as a DIVA subunit vaccine and protected pigs from clinical signs upon homologous challenge with A. pleuropneumoniae serotype 2 and heterologous challenge with serotype 9 . However, the protein composition of this subunit vaccine has not been determined so far and, therefore, was addressed in this study using state-of-the-art proteomic technologies.
A combination of different bioinformatic tools predicted 93 outer membrane proteins for A. pleuropneumoniae serotype 5b  and searching the serotype 5b genome with the SLEP pipeline (http://www.caspur.it/slep
), which optimally combines several prediction tools, revealed even more outer membrane or secreted proteins. Our analyses led to the identification of 75 different proteins of which 15 proteins are membrane proteins and four are secreted proteins as predicted by the PSORTb algorithm. Thus several predicted outer membrane and secreted proteins were not identified by our approach, which is likely due to their low concentration or even absence in the vaccine preparation.
As a semi-quantitative measure we postulate a correlation between identified peptide matches and protein abundance. Of 64 proteins identified in total by UPLC Q-TOF MS/MS, 14 were identified by 10 or more peptide matches. These proteins were therefore designated as the "abundant protein fraction". Among these three Apx toxins and eight outer membrane proteins were found. This approach showed that the detergent extraction not only enriched outer membrane-associated lipoproteins like OmlA and TbpB , but also integral outer membrane proteins like FrpB, TbpA, OmpA1, OmpA2, HgbA and OmpP2. Only three of the highly enriched proteins are cytosolic (TufB and FusA) or periplasmic (YfeA). Their percentage is higher in the fraction of low abundant vaccine proteins and might be caused by lysed cells as well as by the release of membrane vesicles containing these proteins .
A combined interpretation of the "whole vaccine-proteome approach" and the 2-D DIGE analysis revealed different amounts for several proteins in the "detergent-wash" preparations of the three serotypes, indicating serotype dependent differences in gene expression. Additionally, the 2-DIGE analysis showed the presence of various isoforms for several proteins among and between the different strains that vary slightly in mass or isoelectric point. For proteins represented by different isoforms, a high ratio-value often means that only the respective isoform is abundant in one or the other strain. However, this does not necessarily refer to the total amount of the respective protein, which holds true e.g. for the outer membrane protein OmpP2. This protein was identified by comparable peptide numbers between the different serotypes on the one hand, but also appeared in several isoforms with high ratio-values in each strain on the other hand.
With an exception of ApxIIA, all secreted Apx toxins (including ApxIVA) produced by the A. pleuropneumoniae strains used for vaccine generation were detected, and they represent the most abundant protein fraction of the analyzed subunit vaccine. ApxIIA was not identified as apxIIA gene deletion mutants of A. pleuropneumoniae serotypes 1, 2 and 5 had been used for vaccine generation in order to generate a DIVA vaccine .
The Apx toxins are important for virulence [24, 25] and subunit vaccines based on Apx toxins have been developed, eg. Porcillis APP™ (Intervet, ). Therefore, the high content of ApxIA, ApxIIIA and ApxIVA in the analyzed subunit vaccine is presumably one important factor accounting for the protective efficacy of the "detergent-wash" vaccine.
ApxIA and ApxIIIA were identified on 2-DE gels as series of up to 10 spots with different isoelectric points, which is likely caused by charged posttranslational modifications (PTMs). Different isoforms of other proteins were also present within and between strains. As it cannot be excluded that these isoforms vary with respect to their protective efficacy for vaccination purposes, A. pleuropneumoniae-derived proteins might be superior in comparison to recombinant proteins.
ApxIVA has been considered as a DIVA antigen as it is immunogenic and encoded by most A. pleuropneumoniae isolates [27–29]. It was shown for serotype 7 strain AP76 that the apxIV A gene is interrupted by an insertion element preventing in vivo expression of apxIVA
. This is in accordance with our data showing no ApxIVA protein in the "detergent-wash" of AP76. Expression of the apxIVA gene has been considered for long as strictly induced in vivo
 and was shown recently to be upregulated by bronchoalveolar fluid [19, 31]. We show for the first time the translation of the apxIVA gene and identified abundant amounts of the respective protein in the "detergent-washes" of A. pleuropneumoniae serotypes 1, 2, and 5. This presence of ApxIVA in the analyzed subunit vaccine is likely to contribute to immune protection of vaccinated pigs .
As we have grown A. pleuropneumoniae for vaccine generation under iron restrictive conditions, one might argue that these conditions induce apxIVA expression. However, in contrast to many other proteins that were highly enriched in the subunit vaccine, iron restrictive induction of apxIVA has not been observed while analyzing the transcriptional profile of A. pleuropneumoniae serotype 5b during iron restriction grown in BHI-medium . We therefore speculate that the ApxIVA expression might be due to using customized growth conditions (iron-restriction in combination with supplemented PPLO-medium, see in Methods). Other reports describing the lack of apxIVA expression in culture either used BHI-medium  or PPLO-medium supplemented with horse serum  for growth of A. pleuropneumoniae.
The "detergent-wash" vaccine contains several proteins encoded by genes only expressed or upregulated in vivo as identified by IVET , SCOTS [34, 35] or transcriptional profiling . Proteins relevant for in vivo survival of A. pleuropneumoniae that were identified by STM [37, 38] are also present in the vaccine. Six of the abundant proteins were shown previously to be induced in vivo during acute infection (ApxIVA, TbpA, TufB, OmpA1, OmpA2, and HgbA, ). Further, TufB, AasP, DnaK, GapA and RplC had been identified as expressed in chronically infected lung by A. pleuropneumoniae
. Furthermore, the proteins OmpP2, ZnuA and APP7_2020 are considered as being virulence-associated, since the respective STM mutants were attenuated . A recent study on transcriptional profiling of A. pleuropneumoniae during the acute phase of infection detected four of the proteins identified in the "detergent-wash", namely TolB, ApxIVA, APP7_0979, and Irp . APP7_0979 is a conserved lipoprotein [12, 39] and was among the highest in vivo induced proteins .
The analyzed subunit vaccine was highly enriched in immunogenic proteins. A recent immunoproteomic analysis of secreted and outer membrane proteins from A. pleuropneumoniae serotype 3 revealed 30 immunogenic proteins  of which 18 were also identified in the analyzed subunit vaccine. Six of these proteins (ApxIIIA, FrpB, TufB, OmpA1, OmpA2 and YfeA) were among the 14 most abundant proteins; an additional five proteins among the 14 most abundant proteins were identified as immunogenic elsewhere including the ApxIA  and ApxIVA  toxin, as well as the iron repressible transferring binding proteins TbpA and TbpB  and the outer membrane lipoprotein OmlA .
As the immunoproteome of A. pleuropneumoniae has been extensively studied  we set out to identify cross-reactive antigens. Immunization of pigs with the presented subunit vaccine comprised of serotypes 1, 2, and 5 led to formation of antibodies against seven different serotype 7 proteins of which six (TufB, ApxIIA, D15, FrpB, OmpA1, OpmA2) were already known  except for UshA, which is a periplasmic protein catalyzing the degradation of external UDP-glycose to uridine, glycose-1-phosphate and inorganic phosphate .
Convalescent sera of pigs obtained upon experimental infection with serotype 7 detected 13 immunogenic proteins of the subunit vaccine of which seven were already described as immunogenic (TufB, D15, OmpA1, ApxIIIA, FkpA, ZnuA , and ApxIA ). The proteins RpsA, Tal, APP7_0756, FusA, YfeA, and UshA are cross-reactive immunogenic antigens that have not been described for A. pleuropneumoniae before. YfeA is the periplasmic binding protein of an ABC transport system for iron and manganese acquisition and deletion of this ABC system renders Yersinia pestis avirulent . In the porcine pathogen Haemophilus parasuis, which is related to A. pleuropneumoniae, YfeA has been recently identified as an cross-reactive immunogenic protein . Based on BLAST analysis, APP7_0756 is also the periplasmic component of an ABC transport system needed for iron acquisition.
In total, infection with a live A. pleuropneumoniae led to formation of more cross-reactive antibodies than immunization using the described subunit vaccine. Thus several proteins were only detected by convalescent sera (RpsA, FusA, ApxIA, ApxIIIA, Tal, FkpA, ZnuA, YfeA and APP7_0756), although they are present in the subunit vaccine. Live bacteria seem to induce an even stronger immune response than the subunit vaccine, which is in accordance with the findings that pigs surviving natural or experimental infection are at least partially protected from clinical symptoms upon infection with another serotype [46, 47].
The analyzed subunit vaccine was comprised of detergent extracted proteins from three different serotypes of A. pleuropneumoniae in order to increase the number of serotype-specific antigens. Additionally, the likelihood that a serotype specific antigen from one or the other serotype is cross-reactive against another serotype was increased. BLAST analysis showed that the identified immunogenic proteins were highly conserved between the different A. pleuropneumoniae serotypes explaining the observed cross-reactivity. Several of the identified proteins are ideally suited as vaccine components against A. pleuropneumoniae based on their conservation [14, 26].
The different Apx toxins of A. pleuropneumoniae are cross-reactive although sharing low homology. Immune sera of pigs vaccinated with the subunit vaccine, detected the ApxIIA toxin from serotype 7, although ApxIIA was not present in the subunit vaccine. On the other hand convalescent sera of pigs infected with serotype 7, which exclusively expresses the apxIIA gene, detected the ApxIA and ApxIIIA proteins of the vaccine. The Apx toxins share as a common structural feature a tandem array of a nine amino acid repeat  which might cause their cross-reactivity.