Transcriptional profiling of Actinobacillus pleuropneumoniae under iron-restricted conditions

Background To better understand effects of iron restriction on Actinobacillus pleuropneumoniae and to identify new potential vaccine targets, we conducted transcript profiling studies using a DNA microarray containing all 2025 ORFs of the genome of A. pleuropneumoniae serotype 5b strain L20. This is the first study involving the use of microarray technology to monitor the transcriptome of A. pleuropneumoniae grown under iron restriction. Results Upon comparing growth of this pathogen in iron-sufficient versus iron-depleted medium, 210 genes were identified as being differentially expressed. Some genes (92) were identified as being up-regulated; many have confirmed or putative roles in iron acquisition, such as the genes coding for two TonB energy-transducing proteins and the hemoglobin receptor HgbA. Transcript profiling also led to identification of some new iron acquisition systems of A. pleuropneumoniae. Genes coding for a possible Yfe system (yfeABCD), implicated in the acquisition of chelated iron, were detected, as well as genes coding for a putative enterobactin-type siderophore receptor system. ORFs for homologs of the HmbR system of Neisseria meningitidis involved in iron acquisition from hemoglobin were significantly up-regulated. Down-regulated genes included many that encode proteins containing Fe-S clusters or that use heme as a cofactor. Supplementation of the culture medium with exogenous iron re-established the expression level of these genes. Conclusion We have used transcriptional profiling to generate a list of genes showing differential expression during iron restriction. This strategy enabled us to gain a better understanding of the metabolic changes occurring in response to this stress. Many new potential iron acquisition systems were identified, and further studies will have to be conducted to establish their role during iron restriction.


Background
Actinobacillus pleuropneumoniae, etiological agent of porcine pleuropneumonia, causes great commercial losses to the swine industry worldwide [1]. Transmission of this highly contagious disease that affects pigs of all ages occurs mostly by aerosol and close contact with infected animals [2]. During 24 to 48 hours of the acute phase of the disease, formation of extensive and fibrinohemorrhagic lung lesions is often fatal. Animals that survive the disease may become asymptomatic carriers of the bacteria, developing localized and necrotizing lesions associated with pleuritis [3]. Based on differences of capsular polysaccharides, fifteen serotypes have been identified: serotypes 1 to 12 and 15 belong to biotype 1, which is NAD-dependent; serotypes 13 and 14 are classified in biotype 2, which is NAD-independent [4]. In North America, serotypes 1, 5 and 7 are prevalent, while serotypes 2 and 9 are most commonly found in Europe.
Despite many years of research, the total complement of bacterial components that are involved in infection by A. pleuropneumoniae has yet to be identified. Several virulence factors have been proposed: capsular polysaccharides, lipopolysaccharides (LPS), Apx toxins and various iron acquisitions systems [2]. However, the overall contribution of each component to the infection process remains unclear. Although less virulent, an acapsular mutant was still serum-resistant, showed higher adhesion to piglet tracheal frozen sections and could still be re-isolated from lungs of infected animals [5]. LPS apparently plays a role in adhesion in vivo, as these molecules show in vitro adhesion to many biological components [4]. The Apx toxins contribute to development of lesions typically associated with the disease [6] and mutants missing Apx toxins are avirulent in pigs and mice [7]. However, different A. pleuropneumoniae serotypes secrete different sets of Apx toxins, and the relative contribution of the four different Apx toxins (ApxI to IV) is still not clear.
Low availability of iron in the host represents a major stress for bacterial pathogens and is considered a signal that leads to significant changes in cell processes. Iron atoms are often linked with sulphur in Fe-S clusters in the catalytic core of enzymes involved in diverse functions such as respiration, ATP generation, and DNA replication and repair, which might account for this phenomenon. Iron is an essential element for almost every living organism. However in the host, molecules such as transferrin, lactoferrin, haptoglobin and hemoglobin in extra-cellular fluids bind free iron and iron-containing molecules very tightly [8]. While bacteria generally need free iron concentrations of about 10 -7 M, its concentration may be 10 -24 M in the mammalian host [9]. To counteract the effect of these iron-withholding mechanisms of the host, bacteria have evolved different iron acquisition systems, often relying on the secretion of siderophores, small (<1000 Da) molecules with high affinity for iron, or on surface receptors specific for iron-containing host proteins [10]. Studies in our laboratory have led to the identification, expression and characterization of the A. pleuropneumoniae hydroxamate siderophore receptor FhuA [11,12] and a hemoglobin binding receptor HgbA [13]. A. pleuropneumoniae also possesses a transferrin receptor complex composed of two outer membrane (OM) proteins: a 100 kDa TbpA may form a transmembrane channel enabling transport of iron across the OM; a 60 kDa lipoprotein TbpB acts as an auxillary molecule [2,14,15]. Energization of these OM transporters relies on the transduction of the proton motive force from the cytoplasmic membrane (CM) by the TonB-ExbB-ExbD complex [16] that is anchored in the CM and spans the periplasm. In A. pleuropneumoniae, two different TonB systems have been identified: genes tonB1-exbB1-exbD1 are transcriptionally linked to the tbpA-tbpB genes [17]; and a second system with genes tonB-exbB2-exbD2 was also identified [18]. Transport of iron across the CM is apparently accomplished by the AfuABC ABC transporter [19]. It has also been shown that A. pleuropneumoniae can use exogenous siderophores and may be able to secrete an iron chelator in response to iron stress [20].
The ferric uptake regulator Fur protein has been identified in many pathogenic bacteria, including A. pleuropneumoniae [21]. Using Fe 2+ as a cofactor, the Fur protein can interact with a specific sequence termed the Fur box in the promoter region of genes implicated in iron acquisition processes, thereby repressing transcription. When iron becomes scarce, the Fur protein loses its cofactor and becomes inactive. The fact that transcription of some genes seems to be under positive control of active Fur protein [22,23] was recently explained by the discovery of RyhB, a small non-coding RNA which belongs to the Fur regulon [24]. When transcribed, the RyhB RNA down-regulates the mRNA level of those genes that seemed to be positively regulated by Fur.
To better understand the mechanisms used by A. pleuropneumoniae that address iron restriction and to gain insights into strategies used by this pathogen under conditions mimicking the in vivo environment, we evaluated gene expression profiles of A. pleuropneumoniae grown under iron restriction. Our study identified 210 differentially expressed genes, of which 92 are up-regulated. Within the latter set, components of previously unrecognized iron acquisition systems were identified: a putative enterochelin-like siderophore receptor, a potential Yfe system for the acquisition of chelated iron, a putative hemoglobin acquisition system homologous to the N. meningitidis HmbR system, and a putative Fe 2+ -specific porin system.

Microarray analysis of mRNA levels during growth of A. pleuropneumoniae under iron-restricted conditions
To assess the response of A. pleuropneumoniae to iron restriction, the reference strain S4074 was grown in BHI broth containing 50 μg/ml EDDHA, a concentration sufficient to cause iron restriction [11]. This strain was chosen because it is the strain that has been the most studied over time, but also because major problems were encountered with RNA extraction from strain L20. Preliminary CGH studies conducted in our lab showed that 95% of the genes of the A. pleuropneumoniae 5b L20 genome are conserved between both strains. Growth curves established the optimum growth phase for RNA extraction (data not shown). At 50 μg/ml of EDDHA, bacterial growth is almost completely inhibited within an hour of addition. By adding the iron chelator at an optical density of 0.1, iron-restricted cultures and iron-rich cultures were harvested concurrently at an optical density of 0.3. Under these growth conditions, we identified 210 differentially expressed genes, with an estimated false discovery rate (FDR) of 3.22%: 118 were down-regulated (Table 1) and 92 were up-regulated (Table 2). In order to confirm that these variations were not caused by the chelator, control experiments where iron was supplemented to the restricted medium were conducted. Exogenous iron, in the form of FeCl 3 , was added to a final concentration of 50 μg/ml to the iron-depleted medium. Growth curves indicated that this concentration of FeCl 3 was sufficient to promote growth at a similar level as in the BHI broth. Under these conditions, the expression pattern was highly similar to that seen in BHI broth: we identified only 30 differentially expressed genes, out of 2025, with an estimated FDR of 2.5%, 26 of which were up-regulated, while only 4 were down-regulated (data not-shown). Only 12 genes significantly differentially expressed in the iron-supplemented medium were identified as such in the irondepleted versus BHI broth experiment, but with reversed levels of variation. Gene lldD (ap2032), which was up-regulated in the iron-depleted medium, was down-regulated in the iron-supplemented medium. Conversely, 11 genes that were down-regulated in the iron-depleted medium were up-regulated in the iron-supplemented medium. This indicates that the results obtained in the irondepleted versus BHI broth experiment can be attributed to iron restriction, and not to another effect of the chelator.

Validation of microarray results by qRT-PCR
Seventeen genes, representing a wide range of log 2 ratio values, were selected for transcript level analysis using qRT-PCR. Seven genes were overexpressed during iron restriction (tonB1, hgbA, omp64, fetB2, apxIC, PM0741, NMB1668); eight genes were repressed (nrfA, nrfC, nfrE, ompW, dcuB2, dmsA, torA, ccmC); two genes were not affected (pedD, ap1465). We also investigated the tran-script level of the exbB1, exbD1 and tbpA genes, all known to be transcriptionally linked to tonB1 [17] and previously used as positive controls to assess iron restriction [12]. However, they were not present on the AppChip1 as this region of the genome was in one of the few unsequenced areas when the microarrrays were designed. In all cases, genes that had been identified as up-or down-regulated with the microarrays were confirmed by the qRT-PCR experiments. The exbB1, exbD1 and tbpA genes were also up-regulated. Genes not affected showed low level of variation during qRT-PCR analysis, and show good correlation with other results (Fig. 1). Overall, there was good correlation between the log 2 ratios measured by microarray and log 2 ratios from qRT-PCR data (R 2 = 0.87). The log 2 ratios observed with qRT-PCR were usually superior to those observed with the microarray. This outcome has been observed before [25,26] and probably reflects the detection limit of microarrays as well as the complex normalization methods that are used prior to the analysis.

Genes expressed differentially under iron restriction
To evaluate the effect of iron restriction on the porcine pathogen A. pleuropneumoniae, we performed microarray hybridization experiments. Given that iron plays a vital role in metabolic pathways through its presence in the structure of numerous enzymes [27] and its implication in the regulation of genes associated with virulence [28], we recorded important changes in the transcriptome of the bacteria under iron-restricted conditions. A total of 210 genes showed differential expression and the functional classification of these genes provides a significant overview of changes occurring in the bacteria. Numerous microarray studies have investigated effects of iron restriction in many different pathogens, including E. coli [29], H. pylori [30], H. parasuis [31], N. gonorrhoeae [25], N. meningitidis [32], as well as Pasteurella multocida [33], a well known animal pathogen closely related to A. pleuropneumoniae. Many genes that were identified as being ironregulated in the P. multocida study were homologs of some genes that were also identified in our study (Table 3), thus emphasizing the importance of their regulation during iron restriction. A common feature in all these studies is the high induction of genes related to iron acquisition as the products of these genes are essential for survival of the bacteria.
(i) Down-regulated genes Down-regulated genes (Fig. 2) mostly belong to the functional class termed "Energy Metabolism"; 42 of the 118 repressed genes (35%) belong to this group, and they are amongst the most highly repressed. Almost all these genes encode proteins with Fe-S clusters, that use heme molecules as cofactors, or that are activated by Fe 2+ or other divalent cations. These include genes coding for the different subunits of formate dehydrogenase (bisC, hybA, fdhE),      fumarate reductase (frdABC), nitrate reductase (nfrABC), nitrate/trimethylamine oxidoreductase (TMAO) I and III (torAC and torYZ), dimethyl sulfoxide reductase (dmsA) and glycerol-3-phosphate dehydrogenase (glpAC). These enzymes as well as numerous others that encode either cytochrome components or functional partners (cydAB, ccp), cytochrome maturation proteins (ccmCF, nfrE) or iron-sulfur electron transport proteins (napF, hyaA, ykgF), are all implicated in the electron transport respiratory chain, either as electron donor or acceptor during aerobic and anaerobic respiration. Other genes in this category are involved in pathways of sugar metabolism such as fermentation (pflB, adh2), glycolysis or gluconeogenesis (pgk, pfkA, gapA) and the TCA cycle (fumC, maeA).
Many genes that are assigned to this category have been demonstrated or proposed to be members of the FNR regulon. The E. coli FNR transcriptional regulator is an oxygen-responsive activator implicated in the switch from aerobic to anaerobic metabolism in facultative anaerobes [34]. The oxygen-sensing domain of the FNR protein con-tains a Fe-S cluster, which is likely oxidized under aerobic conditions, thereby inactivating the FNR protein. Genes coding for fumarate and nitrate reductase are known to be influenced by FNR [35], as well as genes coding for anaerobic enzymes involved in the utilization of alternative terminal electron acceptors such as TMAO [36]. Sequence analysis in H. influenzae has identified conserved FNR binding motifs upstream of the cydAB genes [37]. These genes are usually considered to be up-regulated by the presence of the FNR protein, but FNR has also been implicated in the down-regulation of genes involved in aerobic respiration, such as genes coding for aerobic enzymes like NADH dehydrogenase and cytochrome oxidase [36]. In our study, although the A. pleuropneumoniae FNR homolog HlyX was observed to be up-regulated, all other putative members of the FNR regulon were shown to be down-regulated. In recent studies, genes aspA, coding for aspartate ammonia lyase, and dmsA, encoding a dimethyl sulfoxide reductase, were shown to be important for the virulence of A. pleuropneumoniae [38,39]. Both these genes, which are apparently under HlyX regulation [40],

8.
also showed down-regulation in our experiments. These results might indicate that another factor could interfere with HlyX regulation, or counter-balance the HlyX-inducing effect during iron restriction. The fact that most of the affected genes code for enzymes containing an Fe-S cluster in their structures or use iron as an activator [41] could explain this effect. Since studies have shown that the A. pleuropneumoniae FNR homolog may be involved in the activation of genes coding for virulence factors [42] and is essential for full virulence [40], the observed up-regulation of HlyX is not unexpected. Precise characterisation of the HlyX regulon in A. pleuropneumoniae will provide a better view of its role in pleuropneumonia. In E. coli, a second regulatory system, the ArcA/ArcB two component system, has also been shown to sense oxygen levels [43]. In our system, the baeS gene product, which has 51% identity with the P. multocida Pm70 ArcB protein, was also down-regulated, indicating that this system might be affected during iron restriction.
The overall picture of down-regulated genes shows that A. pleuropneumoniae has adopted strategies of economy for iron and energy. The principal components of the aerobic respiratory chain were all repressed, as well as key alternative final electron acceptors, probably since these processes implied extensive use of iron-containing enzymes. Genes involved in the synthesis of heme cofactors (bioD1, hemA, hemC and hemN) or quinones and menaquinones (menA, menB and ispH), which are important elements of the respiratory chain, showed down-regulation because lack of iron compromises these processes. Many components of the sugar phosphotransferase systems (PTSs) (manX, hisS, ptsBHI), which enable simultaneous transport and phosphorylation of sugars from phosphoe- nolpyruvate, as well as other genes involved in sugar transport (dcuB1,dcuB2,rbsB,glpF,glpT) were down-regulated under our experimental conditions. This outcome could hamper sugar uptake by the bacteria. Repression of the various PTSs might be caused by the repression of the pfkA gene, which codes for phosphofructokinase, a key enzyme in the pathway responsible for the conversion of glucose to phosphoenolpyruvate, and which serves as the primary source of phosphate for activation of PTSs [44]. The product of the mlc gene, which shows 70% homology with a probable Haemophilus ducreyi sugar metabolism repressor and which was also down-regulated, might be implicated in this down-regulation of PTSs. This repressor has been shown to repress the transcription of many PTSs, and is subject to a negative auto-regulation [44].
Considering these metabolic deficiencies, it is significant that some enzymes with ATPase activity as well as others involved in processes that are not of primary importance in adapting to iron restriction were down-regulated. As an example, four genes purA, purT, pyrD, pyrG for enzymes with ATPase activity that belong to the "Purine, Pyrimidines, Nucleosides and Nucleotides" functional class were down-regulated. Since bacteria are growing in a stressful environment and their metabolism seems highly compro-mised, expression of genes involved in the biosynthesis of molecules useful for replication is not essential.

(ii) Up-Regulated Genes
Many genes involved in cell metabolism were observed to be down-regulated by iron restriction, but cell metabolism was not highly represented in our set of up-regulated genes. Two genes showing high up-regulation during iron restriction were assigned to this category. The lldD gene showed a five-fold induction, and codes for L-lactate dehydrogenase, an enzyme required for conversion of lactic acid produced by fermentation to pyruvate. To compensate for defects of the respiratory chain, A. pleuropneumoniae might have started to rely on fermentation during iron restriction. The gene encoding the XylB xylose kinase involved in the degradation of xylose was also up-regulated. Considering that many PTSs were down-regulated, the use of this alternative sugar, for which PTS systems have seldom been implicated [45] may be reconciled. Several genes of the "Protein Fate" functional class also showed up-regulation. The two subunits of the Clp protease showed higher expression during iron restriction; this cytoplasmic protease is often involved in stress responses and protein quality control [46]. The genes prlC and def, encoding respectively an oligopepti-Functional classification of the differentially expressed genes according to TIGRFAMs dase and a peptide deformylase responsible for the hydrolysis of the N-formyl group of nascent polypeptide chains [47], were also up-regulated. This might indicate a higher turnover rate for native proteins requiring iron molecules in their structure, which might be unable to fold correctly in the absence of iron. The last gene of the "Protein Fate" functional class to be up-regulated was mopB, which codes for co-chaperonin GroES. This cochaperonin, essential for full function of GroEL, facilitates non-native protein folding [48]. Again, the absence of iron might cause the accumulation of incorrectly folded native oligopeptide chains, thereby leading to higher expression of the GroES co-chaperonin.
The major response of A. pleuropneumoniae to iron restriction was the induction of genes involved in iron transport, probably to counter-balance effects of EDDHA. Most genes with known functions, identified as up-regulated during iron restriction, were shown to be involved in iron acquisition and transport. The tonB1 gene showed the highest level of up-regulation, and genes exbB1, exbD1 and tbpA which are transcriptionally linked to tonB1 were shown by qRT-PCR analysis to be also up-regulated. The hgbA gene was over-expressed, as well as the hugZ heme utilization protein which is located immediately upstream of hgbA [13]. Among other known A. pleuropneumoniae iron acquisition-related genes, tonB2 also showed up-regulation, while genes of the fhu operon did not show any significant change in expression, in agreement with previous work done in our laboratory; expression of fhuA is not regulated by iron [12].
Previously unreported iron acquisition systems were also revealed by our experiments. We identified a gene cluster, composed of ORFs PM0741 and NMB1668, showing 43% identity with the HmbR hemoglobin receptor from N. meningitidis. The HmbR receptor was shown in N. meningitidis to be important for survival in an infant rat infection model [49]. HmbR binds hemoglobin with high affinity, is able to strip heme from hemoglobin and then transport it to the periplasm. In N. meningitidis, HmbR is subject to phase variation via frameshift mutations [50], and about half of all clinical isolates express HmbR [51]. In A. pleuropneumoniae, the HgbA receptor has been shown to be responsible for iron acquisition from hemoglobin, and a mutant strain with an internal hgbA deletion could not grow in an iron-restricted medium supplemented with hemoglobin, albeit from different species [13]. Apparently HgbA is the sole hemoglobin receptor in A. pleuropneumoniae serotype 1, but it is not the sole hemoglobin binding protein that was identified in A. pleuropneumoniae. In the same study that lead to the identification of HgbA, a 75 kDa protein that could bind hemoglobin and hemin was also isolated [52]. The putative A. pleuropneumoniae HmbR has an estimated molecu-lar weight of 76.7 kDa, and it is therefore tempting to speculate that those two proteins might share identity. In N. meningitidis, the hmbR gene is located downstream of hemO gene that codes for a heme oxygenase and that is considered essential for heme utilization by pathogenic Neisseriae [53]. No HemO homolog was found in the A. pleuropneumoniae genome, which might explain the apparent lack of iron acquisition from hemoglobin from other putative OM receptors than HgbA in A. pleuropneumoniae. Two other genes, located immediately downstream of the last NMB1668 ORF, and transcribed in the opposite direction, also showed up-regulation: ap2146 and ap2147; see Fig. 3. ORF ap2146 is predicted to code for the α subunit of a N-methylhydantoinase B/acetone carboxylase, while ORF ap2147 shares some region of homology with the periplasmic energy transducing protein TonB. Implication of the products of these ORFs in a potential iron-acquisition process involving the HmbR homolog remains to be assessed.
Our identification of a putative Yfe system was also of seminal interest. The Yfe system was first identified in Y. pestis and shown to allow chelated iron utilization in an E. coli mutant lacking enterobactin [54]. Two different operons encode the Yfe system, carrying genes yfeABCD and yfeE respectively; both operons were Fur-responsive. Later studies showed that the yfeABCD genes code for a periplasmic binding protein-dependent transport system belonging to the superfamily of ABC transporters [55], implicated in iron and manganese acquisition, and independent of TonB [56]. In A. pleuropneumoniae, homologs of components YfeABCD, showing respectively 63%, 76%, 75% and 66% homology with their counterparts in P. multocida, showed up-regulation during iron restriction, but were not present on the same operon. Gene yfeB can be found immediately downstream of the yfeA gene, in the same area as two other ORFs that were up-regulated during iron restriction and that could be implicated in iron acquisition. These ORFs, which are annotated as omp64, show good homology (32%) to the Moraxella catarrhalis CopB OM protein. Meanwhile, the yfeCD genes are located 160 kb downstream of the last omp64 ORF, and also show high homology with the corresponding Y. pestis Yfe proteins. The CopB protein has been implicated in iron acquisition from lactoferrin and transferrin; a mutant strain showed reduced ability to uptake iron from these proteins, with the more marked effect on transferrinbound iron acquisition [57]. In A. pleuropneumoniae, proteins responsible for the utilization of transferrin-bound iron were first identified by affinity methods [58]. Later studies showed that the tbpAB genes from A. pleuropneumoniae are transcriptionally linked to genes tonB1, exbB1 and exbD1, and these exb genes are essential for iron acquisition from transferrin [17]. It was also shown that both tbp genes are essential for iron acquisition from Genetic organization of some gene clusters identified during this study Figure 3 Genetic organization of some gene clusters identified during this study. a) Genomic region of the A. pleuropneumoniae 5b strain L20 genome surrounding ORFs encoding for genes PM0741 and NMB1668. ORFs ap2146 and ap2147 are located 260 pb downstream of the NMB1668 ORF, and are transcribed in the opposite direction. b) yfeAB and omp64 genes are separated by three ORFs that did not show differential expression. c) Genetic organization of a possible operon coding for a putative enterobactin-type ABC transporter system. The two ORFs separating fetB2 and NMB1993 are the putative cytoplasmic components of this hypothetical system. d) The Fe 2+ /Pb 2+ high affinity permease locus.  [59], and another recent study showed that a tonB1 mutant cannot use porcine transferrin, but is not attenuated in vivo [18]. As with the A. pleuropneumoniae HmbR homolog, it would be interesting to examine the presence of the CopB homolog in vitro and in vivo, and possible effects of mutations in this gene. Although the best amino acid homology was with the CopB protein, the overall alignment of the A. pleuropneumoniae omp64 gene product with the M. catharralis CopB protein is not strong, implying that Omp64 might have a role to play in iron acquisition, but its target might not be transferrin or lactoferrin. Another ORF (ap1453) also annotated omp64 showed up-regulation in our experiments. This second Omp64 also shows homology with the CopB OM protein (57%), but the overall alignment with CopB seems superior to that of the other Omp64 (ap0300 -ap0301). The ap1453 Omp64 also shows homology with the H. influenzae heme-hemopexin utilization protein C (41%). We hypothesize that the YfeABCD-Omp64 (ap0300-ap0301) proteins are components of a new iron acquisition system in A. pleuropneumoniae that are located in the OM (Omp64) and in the CM (YfeBCD), as determined by the PSORT algorithm [60]. The exact location of YfeA could not be determined precisely, but it is predicted not to be cytoplasmic.
Another cluster of genes was particularly interesting with regards to iron acquisition. Two ORFs, annotated fetB2, seem to encode a unique protein presenting similarities with members of the TroA superfamily of periplasmic metal binding proteins. Sequence analysis reveals homologies with other known or putative periplasmic binding proteins, some of which are involved in iron transport. Downstream, three putative genes appear to code for the components of an ABC transport system. One gene of this putative ABC transport system was also up-regulated: the NMB1993 gene coding for a putative ATPase component. These components show homology with the Ceu system (Campylobacter Enterochelin Utilization), and prediction of localization with PSORT indicates that fetB2 localizes to the periplasm, NMB1993 to the CM and/or cytoplasm; the two other components, which are not over-expressed in our system, were predicted to be in the CM. We demonstrated [20,61] that A. pleuropneumoniae uses different exogenous siderophores, including a catechol-type siderophore like enterochelin. Up to now, in A. pleuropneumoniae, the only identified siderophore OM receptor is FhuA, specific for ferric hydroxamates [11]. It is premature to conclude that the fetB2 and NMB1993 genes are part of this unidentified catechol-type siderophore acquisition system.
Three other up-regulated ORFs were identified as having a putative role in iron acquisition. ORFs ccrA1, ap0740 and ap0741 were classified as proteins of unknown function, but share homologies respectively with a family of periplasmic lipoproteins involved in iron transport, a family of iron-dependent peroxidase and a family of high affinity Fe 2+ /Pb 2+ permease. Since no clear homology with any known or characterized protein was established, hypotheses concerning their function and roles in iron acquisition have to be formulated with great care. Recently, Cowart showed [62] that bacterial reductases, by changing the state of free iron from Fe 3+ to Fe 2+ , could play a major role in iron acquisition. The presence of a possible Fe 2+ permease could indicate the existence of such a mechanism in A. pleuropneumoniae.
Considering that iron restriction conditions are encountered in vivo, we further examined the expression of known or putative virulence factors of A. pleuropneumoniae under such conditions. Aside from different iron acquisition systems, the Apx toxins are often regarded as essential for virulence of A. pleuropneumoniae. The Apx toxins are members of the RTX (Repeat in Toxin) family, and the genetic organization of the genes that are essential for the synthesis and secretion of the toxin generally follows the same order: apxC, apxA, apxB and apxD, which code respectively for the pretoxin activator, the pretoxin structure and the secretion apparatus [63]. These genes can be transcribed from two different transcripts: a major 3.5 kb transcript containing genes apxICA, and a minor 7.5 kb transcript with genes apxICABD [64]. During iron restriction, the first gene of the apxI operon, apxIC, showed slight up-regulation, but the three other genes were not over-expressed. The A. pleuropneumoniae strain used in this study possesses genes coding for the ApxI, II and IV toxins. Very little is known about the transcriptional regulation of the apxI operon, but it has been shown to be at least regulated by the combined activity of the Fur protein and calcium [21]. Under high calcium concentration, Fur seemed to act as an activator of the apxI operon, while it seemed to act as a repressor under low calcium concentration. Under our experimental conditions, it seems that Fur acts as a repressor since the apxIC gene was identified as being slightly up-regulated during iron-restriction, i.e. in the absence of Fur. The fact that it was the only gene of the apxI operon to show significant up-regulation might reflect the stringency of our analysis, but might also point towards the existence of fine post-transcriptional tuning of the apxI operon. The existence of such mechanisms of regulation could also explain why apxIA does not seem to be up-regulated, even though it is located on the same transcript as apxIC.
The A. pleuropneumoniae ureC has been implicated as a possible virulence factor, with a putative role in persistence of bacteria in vivo [65]. In our study, the ureC gene was identified as being down-regulated during iron restriction. Since it was shown that this gene might have more effect in the late stage of the disease, it seems clear that other in vivo factors, as with hlyX, may influence the regulation of the ureC gene.
Three ORFs which had approximately two-fold induction during iron deficiency also warrant attention. The Ssa1 protein (Serotype 1-Specific Antigen) was first identified in Mannheimia haemolytica and was associated with the serotype switch occurring in the upper respiratory tract of bovines following stressful events, potentially leading to development of disease [66]. The protein was shown to localize to the OM [67]. Sequence homology research on the A. pleuropneumoniae Ssa1 protein led to identification of an autotransporter domain at the C-terminal extremity of the protein; the A. pleuropneumoniae Ssa1 was also classified in the family of subtilisin-like serine proteases, although no experimental evidence of this activity could be found. Recently, autotransporter proteins such as Ig proteases have been implicated in virulence [68].
Autotransporter proteins belonging to the serine protease family have been identified in various Gram-negative bacteria. H. influenzae, a close relative of A. pleuropneumoniae, possesses at least two: an IgA1 protease and the Hap protein which has been shown to be involved in adhesion.
Little is known about the Ssa1 protease but its implication in virulence in M. haemolytica suggests that it could play a similar role in A. pleuropneumoniae.
Genes ftsK and ftsA, essential in the first steps of cell division, also showed higher expression during iron restriction. Considering that some genes involved in the "Purine, Pyrimidines, Nucleosides and Nucleotides" were down-regulated, this result was unexpected. However, it has been shown that the transcription of the ftsZ gene is subjected to regulation by antisense transcription of a 490-bp segment spanning the junction between the ftsA and ftsZ genes [69], which could probably explain the apparent overproduction of the ftsA mRNA. As for ftsK, although the protein is implicated in cell division, other functions have been suggested for this protein [70], and the observed up-regulation might not be linked with cell division.

Conclusion
In summary, the evaluation of differential gene expression in A. pleuropneumoniae during growth in an iron-restricted medium enabled us to gain a better understanding of the metabolic changes occurring in response to this stress. Transcript profiling using DNA microarrays is a powerful tool to determine the exact composition of the bacterial transcriptome in defined conditions, therefore leading to the putative identification of components that are essential during these conditions. It can also help identify components which are likely to be expressed during the infection process in the host, and that might be interesting targets for vaccines.
In the course of our study, many new potential iron acquisition systems were highlighted. Clearly, iron acquisition in A. pleuropneumoniae might rely on more systems that what was previously thought, and further studies will be necessary to evaluate the impact of these systems during the course of infection by A. pleuropneumonia.

Bacterial strains and growth conditions
Actinobacillus pleuropneumoniae serotype 1 strain S4074 was routinely grown on BHI medium supplemented with either 15 μg/ml (agar) or 5 μg/ml (broth) of NAD. For the microarray experiments, two flasks of BHI broth were inoculated with 500 μl of an overnight culture of A. pleuropneumoniae serotype 1 strain S4074 and grown at 37°C in an orbital shaker until an optical density of 0.1 was reached. To initiate iron restriction in one of the two cultures, EDDHA was added to a final concentration of 50 μg/ml. In the iron supplementation experiments, FeCl 3 was added to the iron depleted culture 5 min. after the addition of EDDHA to a final concentration of 50 μg/ml. The cultures were then re-incubated until they reached a final optical density of 0.3.

RNA extraction
RNA was harvested from cells at an optical density of 0.3. Ice-cold RNA degradation stop solution (95% ethanol, 5% buffer-saturated phenol), shown to effectively prevent RNA degradation and therefore preserve the integrity of the transcriptome [71], was added to the bacterial culture at a ratio of 1:10 (vol/vol). The sample was mixed by inversion, incubated on ice for 5 min, and then spun at 5000 g for 10 min to pellet the cells. Bacterial RNA isolation was then carried out using the QIAGEN RNeasy Mini-Kit. During the extraction, samples were subjected to an on-column DNase treatment, as suggested by the manufacturer. The RNA concentration, quality and integrity were assessed spectrophotometrically and by gel analyses.

Construction of the A. pleuropneumoniae 5b strain L20 microarray (AppChip1)
The draft genome sequence of A. pleuropneumoniae serotype 5b strain L20 [GenBank: CP000569] was used as a source of the genes used in this study. ORFs were identified using the Glimmer software package [72], and used to search for homologs among the bacterial gene subset of Genbank [73] using the BLASTP program [74]. PCR primers were designed for each of the 2025 ORFs of the genome of A. pleuropneumoniae using the Primer3 program [75] controlled by an automated script as described previously [76]. Primer-selection parameters were standardized and included a similar predicted melting temper-ature (60 ± 3°C), uniform length (25 nt), and a minimum amplicon size of 160 bp. Generation of PCR amplicons and fabrication of DNA microarrays were as described [76]. Details on the construction of this microarray (AppChip1) are available on the Institute for Biological Sciences website [77].

Microarray hybridizations
cDNA synthesis and microarray hybridizations were performed as described [78]. Briefly, equal amounts (15 μg) of test RNA and control RNA were used to set up a standard reverse transcription reaction using random octamers (BioCorp), SuperScript II (Invitrogen) and aminoallyl-dUTP (Sigma), and the resulting cDNA was indirectly labelled using a monofunctional NHS-ester Cy3 or Cy5 dye (Amersham). The labelling efficiency was assessed spectrophotometrically. Labelled samples were then combined and added to the A. pleuropneumoniae 5b strain L20 microarray. Nine hybridizations were performed for the iron-restriction experiments, including three pairs of microarrays for which Cy3 and Cy5 dyes were swapped, while 4 hybridizations were conducted for the iron supplementation experiments. Data were submitted to the Gene Expression Omnibus [79] [GEO:GSE4674 and GSE6366]. All slides were scanned using a Perkin-Elmer ScanArray Express scanner.

Microarray analysis and bioinformatics
The TM4 suite of software from The Institute of Genomic Research was used for the whole microarray analysis [80]. First, raw data were generated using SpotFinder v.3.0.0 beta. The integrated intensities of each spot, equivalent to the sum of intensities of all unsaturated pixels in a spot, were quantified and the integrated intensity of the local background was subtracted for each spot. The same operation was performed with the median spot intensities. The spot detection threshold was set so that spots for which the integrated intensity was less than one standard deviation above the background median intensity were set to zero. Raw spot data were converted from integrated intensities to median spot intensities using TIGR's Express Converter software, the latter being less influenced by outlier values than integrated intensities.
Data were normalized with TIGR's MIDAS software tool using locally weighted linear regression (lowess) [81][82][83]. Spots with median intensities lower than 1000 were removed from the normalized data set. Intensities for duplicate spots were merged to generate the final normalized data set, subsequently analyzed using TIGR's TMEV microarray analysis tool. The Significance Analysis of Microarray (SAM) algorithm [84], which is implemented in TMEV, was used to generate a list of differentially expressed genes. During SAM analysis, a false discovery rate of 3.22% was estimated for the iron-depleted versus BHI broth experiment, while a FDR of 2.51% was estimated for the iron-supplemented versus BHI broth experiment; this value estimates the proportion of genes likely to have been identified by chance. Functional classification of these genes was conducted using TIGR's Comprehensive Microbial Resource (CMR) [85]. Proteins were assigned to their corresponding pathways using the Meta-Cyc Metabolic Pathway Database [41]. Homologies were assessed using Blast tools [86] hosted on the NCBI and TIGR servers. Additional subcellular localization was determined with PSORTb [60]. Protein sequence alignments were performed using the ClustalW multiple sequence alignment algorithm [87].

Real-Time quantitative RT-PCR
Microarray results were verified by real-time quantitative RT-PCR (qRT-PCR), using the QuantiTect ® SYBR ® Green RT-PCR Kit (Qiagen). Reactions were performed with a 16-place Cepheid Smart Cycler ® System in a total volume of 25 μl. Oligonucleotide primers (Table 4) were designed using Primer3 software [75]. To ensure that amplification with these primers resulted in single amplicon of the anticipated size, they were PCR tested before proceeding to qRT-PCR analysis. Primer pairs which amplified fragments of 195 to 205 bp with a melting temperature of 60°C were selected. Seventeen genes (7 up-regulated, 8 down-regulated, 2 non-significant) were selected for analysis. Relative expression of each gene as determined by qRT-PCR was normalized to that of the ackA gene which showed a stable level of expression throughout the different microarray experiments (data not shown). Prior to the qRT-PCR, the RNA samples were subjected to a DNase treatment with TURBO DNase (Ambion, Austin, TX) to avoid DNA contamination in the samples. Quantitative measures were obtained using the method [88].

Authors' contributions
VD designed the transcript profiling experiments, carried out downstream data analysis, and drafted the manuscript. JHEN designed the AppChip1 and helped with the downstream data analysis. JWC and JH participated in the study design and revised the manuscript. MJ participated in the conception and supervised the design of the study and revised the manuscript. All authors read and approved the final manuscript.