Contribution of the Twin Arginine Translocation system to the exoproteome of Pseudomonas aeruginosa

The opportunistic pathogen Pseudomonas aeruginosa uses secretion systems to deliver exoproteins into the environment. These exoproteins contribute to bacterial survival, adaptation, and virulence. The Twin arginine translocation (Tat) export system enables the export of folded proteins into the periplasm, some of which can then be further secreted outside the cell. However, the full range of proteins that are conveyed by Tat is unknown, despite the importance of Tat for the adaptability and full virulence of P. aeruginosa. In this work, we explored the P. aeruginosa Tat-dependent exoproteome under phosphate starvation by two-dimensional gel analysis. We identified the major secreted proteins and new Tat-dependent exoproteins. These exoproteins were further analyzed by a combination of in silico analysis, regulation studies, and protein localization. Altogether we reveal that the absence of the Tat system significantly affects the composition of the exoproteome by impairing protein export and affecting gene expression. Notably we discovered three new Tat exoproteins and one novel type II secretion substrate. Our data also allowed the identification of two new start codons highlighting the importance of protein annotation for subcellular predictions. The new exoproteins that we identify may play a significant role in P. aeruginosa pathogenesis, host interaction and niche adaptation.

Scientific RepoRts | 6:27675 | DOI: 10.1038/srep27675 ref. 14). P. aeruginosa is a ubiquitous Gram-negative bacterium and an important opportunistic human pathogen. P. aeruginosa is responsible for chronic lung infections and mortality in cystic fibrosis patients, for life-threatening infections in immune-compromised humans, and causes 10% of all nosocomial hospital-acquired infections. So far only three extracellular proteins (or exoproteins) have been experimentally demonstrated to be Tat-dependent in P. aeruginosa: PlcH and PlcN, two phospholipases C and GlpQ, a phosphodiesterase involved in the hydrolysis of deacylated phospholipids 15,16 . All three proteins are produced under phosphate starvation conditions and are secreted by the T2SS/Xcp machinery [16][17][18] .
Here, we set out to identify the full repertoire of P. aeruginosa Tat-dependent proteins secreted into the growth medium (or exoproteome as defined in ref. 19) under nutrient rich and phosphate starvation conditions by two-dimensional (2D) gel analysis. Although no Tat-dependent exoproteins were identified under nutrient rich conditions we found that Tat substrates make a major contribution to the exoproteome during phosphate starvation, a condition that induces the synthesis of virulence factors 20 . We discovered and rigorously verified a number of new Tat substrates. We propose that these Tat-dependent exoproteins as well as other new T2SS exoproteins that we identified are likely to play significant roles in Pseudomonas virulence and colonization of the environment.

Results
Comparison of P. aeruginosa wild type and tat mutant exoproteomes. In numerous bacteria, inactivation of the Tat system leads to a defect in cell envelope integrity [21][22][23] . In E. coli this is linked to the export defect of two Tat-dependent cell wall amidases, AmiA and AmiC which are involved in peptidoglycan remodeling 24,25 . Consequently, E. coli tat mutants show a cell envelope defect that is characterized by leakage of periplasmic proteins, hypersensitivity to drugs and detergents, and a chain-forming phenotype 21,25 . Notably, in Salmonella Typhimurium, this Tat-dependent envelope defect has been linked to the virulence defect of the tat mutant 26 . P. aeruginosa PAO1 contains two cell wall amidase homologues, PA5538 (orthologue to E. coli AmiC) predicted to contain a Tat signal peptide and PA4947 (orthologue to E. coli AmiB) containing a typical Sec signal peptide. In order to exclude the possibility that the reduced virulence of the P. aeruginosa tat mutant is a result of similar Tat-dependent envelope defects and to ensure that the study of protein secretion was feasible in the P. aeruginosa tat mutant, we investigated whether the P. aeruginosa tat mutants of the PAO1 and PA14 wild type (WT) strains also displayed a cell envelope defect. We show that the P. aeruginosa tat mutant does not form chains of cells and is not sensitive to SDS (Supplementary Figure S1). These results indicate that the envelope of the P. aeruginosa tat mutant does not show the same Tat-dependent defects that have been observed in other bacteria and are associated with attenuated pathogenicity and the leakage of periplasmic proteins. Our results are supported by data recently obtained by Yakhnina et al showing that deletion of the cell wall amidase homologue PA5538 in the PAO1 or PA14 strains had no effect on cell growth or morphology 27 .
Next, to identify P. aeruginosa Tat-dependent exoproteins, PAO1 WT and tat mutants were grown in rich medium or under phosphate starvation and proteins were separated using 2D-gel electrophoresis. We chose phosphate starvation because it is a condition known to influence both the production of virulence factors and known Tat exoproteins 16,20 . The extracellular 2D-gel profiles of WT and tat strains grown in rich medium did not show any visible differences and were not analyzed further. In contrast, under phosphate starvation conditions the extracellular profiles of WT and tat mutant were noticeably different (Fig. 1). In all, a total of 36 different proteins were identified and their levels quantified in both the WT and tat mutant (Tables 1, 2 and Supplementary Table S1). The subcellular localizations and the mechanisms of secretion for each of the 36 proteins are listed in Table 1. The exoproteome analysis revealed that secreted proteins represent the major and most abundant part of the exoproteome with only minor contamination by proteins from other bacterial fractions. These include proteins such as flagella-related proteins, that are likely to originate from flagella that are shed from cells in shaking cultures 28 as well as proteins that are known to be secreted in outer membrane vesicles (OMV) 29,30 .
In agreement with our visual examination our quantitative mass spectrometry analysis revealed significant differences in the levels of at least 8 proteins between the WT and tat strains ( Table 2). Five proteins were present at 3-8-fold lower levels in the tat mutant strain and so represented potential Tat-dependent exoproteins (red spots). These proteins were identified as PlcH, a known Tat substrate, the putative alkaline phosphatase PA3910 and the hypothetical proteins PA4140, PA2377 and PA2699 ( Fig. 1 and Table 1). Notably, PA3910 is homologous to PhoD, a well-known Tat substrate in Bacillus subtilis and Streptomyces coelicolor 31,32 . In addition we found that the tat mutant exoproteome contained 2-fold lower levels of exotoxin A (Eta/ToxA) and the CbpD-fragment (CbpD-F), and 3-fold higher levels of the low molecular weight alkaline phosphatase LapA. We also noted that GlpQ, which we previously identified as a Tat substrate in P. aeruginosa PAK strain 16 , could be detected in the tat mutant exoproteome although at 1.7-fold lower levels than the WT. Altogether these results reveal for the first time the phosphate-starvation exoproteome of P. aeruginosa, and identify the fraction that is dependent on the Tat system for secretion.
Deletion of the Tat system modifies gene expression. The absence of the Tat system unexpectedly leads to a decreased amount of ToxA in the extracellular medium (Fig. 1). ToxA is a well-known T2SS substrate that is believed to be exported through the Sec pathway. Interestingly, in 2002, Ochsner and collaborators showed that the transcription of the toxA gene was decreased in a tat mutant strain to 25% of the WT 13 . Therefore, in order to differentiate Tat-dependent exoproteins from exoproteins whose genes are misregulated in the tat mutant, we performed quantitative RT-PCR (qRT-PCR) (Fig. 2). We detected significantly reduced transcript levels for the PA2377 and toxA genes in the tat mutant compared to the WT strain. Therefore, the reduced levels of these proteins in the tat exoproteome are likely to be caused by reduced transcription and not by impaired Tat-dependent protein export. Notably, PA2699, PA3910 and PA4140 expression levels were either higher or similar in the tat mutant suggesting that these genes may encode true Tat substrates. We also analyzed transcript levels for lapA whose gene product is present at 3-fold higher levels in the exoproteome of the tat mutant than in the WT (Table 2). We found that lapA expression was induced 5 fold in the tat mutant, and is therefore likely to explain the higher abundance of this protein in the tat mutant exoproteome (Fig. 2).
GlpQ is not a Tat-dependent exoprotein. The exoproteome analysis revealed that GlpQ was still present in the PAO1 tat mutant exoproteome, although at a 1.7-fold decreased level ( Fig. 1 and Table 2). This rather weak response was surprising because we previously identified GlpQ as a Tat-dependent exoprotein in P. aeruginosa strain PAK 16 . This is probably not due to strain differences because GlpQ from the P. aeruginosa PAO1 and PAK strains share 99.47% sequence identity and have the same signal peptide. Interestingly, while the N-terminal signal peptide of GlpQ contains two consecutives arginines, GlpQ is not predicted to be a Tat substrate by Tat prediction programs (Fig. 3a).
In order to re-investigate the Tat-dependence of GlpQ secretion we compared extracellular fractions of WT and tat mutant strains (Fig. 3b). Inspection of the gel at the size expected for mature GlpQ (39.2 kDa) showed two closely migrating bands in the WT (annotated a and b) which were identified as GlpQ and LapA by mass spectrometry. In the ∆ tat mutant however, overproduction of LapA, confirmed by immunoblot analysis, obscured the visibility of GlpQ. Next, analysis of GlpQ secretion in the ∆ lapA∆ tat double mutant confirmed that the overproduced protein corresponds to LapA and revealed that GlpQ is still present in the tat mutant (annotated c on Fig. 3b). We conclude from this experiment that GlpQ is still secreted in the tat mutant and therefore is not a Tat-dependent exoprotein.
PA2377 is a new Xcp-dependent exoprotein. Our qRT-PCR analysis showed that PA2377 transcription was decreased in the tat mutant giving a possible explanation for its absence from the tat exoproteome (Figs 1 and 2). However the N-terminal region of PA2377 shows some of the characteristics of a Tat signal peptide with a long signal peptide and a possible twin arginine motif RRxF albeit not localized next to the H-region ( Fig. 4a and ref. 6). Moreover in silico analysis of the PA2377 genomic region suggested that there is a second start codon AUG 2 (encoding M 24 ) associated with a ribosomal binding site (RBS) sequence and located 66 bp downstream of the annotated start codon AUG 1 (encoding M 1 ) (Fig. 4a). An algorithm developed by Kolaskar and Reddy for the identification of start codons predicts that AUG 2 is a good start codon while AUG 1 is not 33 . If AUG 2 was used  Table 1 and Supplementary Table S1. One of four replicate gels is shown; the other replicates are shown in Supplementary Figure S2.
PA2377 would possess a typical Sec-signal peptide. Finally, while PA2377 is annotated as a putative ABC transporter periplasmic iron-binding protein, we clearly identified it in the extracellular medium of the WT strain.
To investigate whether PA2377 synthesis is initiated from M 1 and/or M 24 we constructed the plasmid pJN2377H and two variants where each of the putative initiation codons was mutated to AUA causing a methionine-to-isoleucine replacement (pJN2377H-M 1 I and pJN2377H-M 24 I). Since PA2377 transcription was down-regulated in the tat mutant ( Fig. 2) we put PA2377 and its variants under the control of an arabinose inducible promoter (P BAD ) and with the addition of a C-terminal hexa-histidine (His 6 ) epitope-tag. Cells carrying pJN2377H and pJN2377H-M 1 I but not pJN2377H-M 24 I produced an exoprotein of ~50 kDa corresponding to the expected size of PA2377H (Fig. 4b). This result indicates that PA2377 synthesis is initiated from M 24 and not from M 1 , and demonstrates that PA2377 is a new exoprotein. The initiation of PA2377 translation from M 24 would suggest that PA2377 export occurs Tat-independently (Fig. 4a). To determine how PA2377 is secreted out of the cell, we compared the secretion profile of a WT strain, a ∆ tat mutant, a ∆ xcpT mutant and the respective complemented strains. As expected, the known Tat-and T2SS/Xcp-dependent exoprotein PlcH was secreted as a mature protein (mPlcH) in the WT strain, but accumulated as an uncleaved precursor (pPlcH) in the ∆ tat mutant and probably as a periplasmic mature protein in the ∆ xcpT mutant (Fig. 4c, lanes 7, 8, 3, and 5; and ref. 16). PlcH secretion could be restored in ∆ tat and ∆ xcpT mutant strains by cis-and trans-complementation respectively (Fig. 4c, lanes 10 and 12). Interestingly, PA2377H was secreted in the ∆ tat strain but accumulated largely intracellularly in the ∆ xcpT mutant (Fig. 4c lanes 9 and 5). These results indicate that PA2377H export is not Tat-dependent and that it is secreted via the T2SS machinery. Interestingly, the PA2377 homologue in E. carotovora, ECA2134, is predicted to contain a Sec signal peptide and was also shown to be present in the extracellular medium and secreted by the T2SS/Out machinery adding confidence to our data here 34 Table 2) and their transcript levels are similar in the WT and the tat mutant (Fig. 2). PA3910 and PA4140 are also predicted to contain a Tat signal peptide by the TatFind and TatP prediction programs (Fig. 5a).  To confirm Tat dependence, cellular localization and find out how PA3910 and PA4140 are secreted in the extracellular medium, both genes were replaced by genes encoding His 6 epitope-tagged proteins (PA4140H and PA3910H) on the chromosome of PA14 WT, ∆ tat and ∆ xcpT mutants and their respective complemented strains (Fig. 5b). These constructs allowed us to show that PA3910H and PA4140H are both secreted in a Tat-dependent manner (Fig. 5b lanes 2 and 7) via the Xcp/T2SS machinery (Fig. 5b lanes 4 and 9). However, we note that PA3910H secretion is not very efficient since the protein can still be detected in whole cells of the wild type and the tat complemented strain. The presence of the mature size protein in whole cells indicates that the protein has been exported through the IM and successfully processed but that subsequent secretion through the OM is limiting. We hypothesize that the presence of the histidine tag on PA3910H may impair periplasmic T2SS recognition. Taken together, these data confirm and support our proteomic data and show that PA3910 and PA4140 are authentic new Tat substrates in P. aeruginosa and that both are exoproteins that are secreted into the extracellular medium by the T2SS/Xcp machinery.
PA2699 is a new Tat-substrate translated from a corrected translation start site. PA2699 was identified as another candidate Tat-substrate since it is absent from the tat mutant exoproteome and its transcription is not downregulated in the tat mutant (Figs 1 and 2). However, the annotated PA2699 does not contain a Tat signal peptide. We inspected the genomic sequence around the annotated start codon AUG 1 (encoding M 1 ) and noticed that another putative start codon AUG 2 (encoding M −47 ) and a canonical RBS were located 141 bp upstream of AUG 1 in the same open reading frame (Fig. 6a). However, neither codon quite meets the threshold value for consideration as an initiation codon using the Kolaskar and Reddy prediction algorithm (both score 24 and a score of ≥ 26 is considered an initiation codon) 33 . If M −47 was used as translation start site PA2699 would then possess a typical Tat signal peptide with a RRHFL motif and a signal peptidase I cleavage site ANA (Fig. 6a). We first attempted to test whether PA2699 is a new Tat-dependent exoprotein by replacing the PA2699 chromosomal gene with a His 6 epitope-tagged protein. However, we were unable to detect the protein in phosphate depleted or rich media. Therefore, we next placed PA2699 under control of a P BAD promoter in the plasmid pJN2699H. Remarkably upon induction we could detect two distinct forms consistent with translation initiation from AUG 1 and AUG 2 . Independent inactivation of these AUG start codons in pJN2699H-M 1 I and pJN2699H-M −47 I showed that either AUG could be used as a translation initiation codon (Fig. 6b). Surprisingly, however neither form was localized in supernatant. To determine where both forms of PA2699 were localized in whole cells we further fractionated WT cells carrying pJN2699H. We found that both forms of PA2699H fractionated mainly with the heavy fraction consisting of protein aggregates and could be solubilized with 8 M urea (Fig. 6c). It is therefore likely that the presence of the His 6 epitope-tag may impair PA2699 conformation and thus correct folding and export.
To further address the Tat-dependence of PA2699, we next used an E. coli Tat reporter assay based on the amidase AmiA which allows the identification of Tat signal peptides 35 . When fused to a Tat signal peptide, AmiA is exported into the periplasm and allows the rescue of the SDS sensitive phenotype in an E. coli MCDSSAC reporter strain that lacks periplasmic AmiA and AmiC activities. Critically, AmiA does not rescue SDS sensitivity when fused to a Sec signal peptide 25 . When fused to AmiA, the putative Tat signal peptide of PA2699 (from M −47 to A −13 ; Fig. 6a) allowed the complementation of the chain-forming and SDS sensitive phenotypes of the E. coli reporter strain (Fig. 6d). The fusion protein failed to confer SDS resistance when produced in a tat mutant or when the conserved twin arginine was replaced by twin lysine, confirming that ssPA2699-AmiAH is exported in a Tat-dependent manner (Fig. 6d). Altogether these data show that PA2699 contains a cryptic upstream start codon that allows the production of a second isoform of the protein possessing a functional and in frame Tat signal peptide. This is in agreement with the data obtained in our proteomic analysis which indicated that PA2699 is a Tat-dependent exoprotein (Fig. 1).

Discussion
Our characterization of the P. aeruginosa exoproteome has provided the first global map of P. aeruginosa proteins that are secreted under phosphate starvation, a condition known to influence virulence factor production 20 . We identified 36 major proteins present in the supernatant of the WT PAO1 strain ( Fig. 1 and Table 1). Out of these  (Table 1). In addition, we identified new P. aeruginosa exoproteins whose localization was previously unknown. These include a probable protein binding component of an ABC sugar transporter PA3190, the putative alkaline phosphatase PA3910 and the hypothetical proteins PA2377, PA2451, PA2452, PA2699, PA3250, PA3734 and PA4140 (Table 1). We further used cell fractionation to verify that three of these (PA2377, PA3910 and PA4140) are indeed new exoproteins secreted by the T2SS (Figs 4-6).
The P. aeruginosa genome has been screened for the presence of Tat signal peptides, and depending on the stringency of the method used, between 18 and 57 candidate substrates have been identified 13,36,37 . To date, only five of these candidates have been experimentally shown to be Tat substrates: two secreted phospholipases 16 and three periplasmic proteins involved in pyoverdine biosynthesis and agmatine metabolism [38][39][40] . However, in silico approaches can only provide an estimate of the true number of Tat substrates and cannot predict whether a protein exported into the periplasm will subsequently be secreted through the OM. Therefore, to gain experimental evidence for new secreted Tat substrates, we compared the P. aeruginosa WT and tat mutant exoproteomes in nutrient-rich and phosphate starvation conditions. Although we did not detect Tat-dependent proteins under nutrient-rich conditions we identified and experimentally verified three new Tat substrates, PA3910, PA4140 and PA2699, produced and secreted in phosphate depleted medium. While PA3910 and PA4140 possess predicted Tat signal peptides, PA2699 does not (Figs 5 and 6). Interestingly, we show that translation of PA2699 can be initiated at two translation start sites M 1 and M −47 (Fig. 6a,b). Whilst PA2699 initiated at M 1 does not possess a signal peptide, PA2699 initiated at M −47 contains a fully functional Tat signal peptide (Fig. 6d). Phylogenetic analysis indicates that PA2699 is conserved in all Pseudomonas groups but that the upstream Tat signal peptide has been completely lost in the plant-associated Pseudomonads, P. fluorescens, P. syringae and P. chlororaphis ( Supplementary Figures S3 and S4). Our identification of PA2699 in the exoproteome of PAO1 and the fact that the PA2699 Tat signal peptide is conserved in some Pseudomonas species strongly supports the physiological relevance of the alternative initiation codon AUG 2 in P. aeruginosa. We additionally showed that PA2699 translation is also initiated from AUG 1 to produce an isoform lacking a Tat signal peptide, albeit when expressed from a plasmid. To our knowledge this could be the first example of a bacterial protein whose two distinct cellular localizations depend on two distinct start codons.
As discussed above we show that the genome annotation for PA2699 missed an authentic Tat peptide. We also show that the initiation codon is miss-assigned for PA2377. The correct initiation codon gives a typical Sec signal peptide rather than the long and unusual signal peptide encoded by the annotated gene (Fig. 4). Together, these results emphasize the limitation of genome annotation for the prediction of subcellular localization and export pathway targeting. We therefore propose that the number of proteins exported through the IM (thus possessing a signal peptide) is probably under estimated in P. aeruginosa because of initiation codon missannotation.
We found that the other two new Tat substrates, PA3910 and PA4140, are secreted via the general T2SS/ Xcp machinery (Fig. 5). Only two other P. aeruginosa exoproteins were previously known to be secreted via the Tat-T2SS/Xcp pathway, PlcH and PlcN (Figs 4 and 5 and ref. 16). These data strengthen the mosaic model for T2SS secretion where substrates can be exported into the periplasm by the Sec or Tat machineries before being secreted out of the cell by the T2SS/Xcp secreton. This would also suggest that Tat substrates are not an exception but rather a general feature of the T2SS. Although the Tat system exports only a few substrates in most organisms, the deletion of tat genes is associated with a large array of defects including virulence in plants and human pathogens 13,14 . However the molecular mechanisms underlying the loss of virulence in tat mutants are generally unknown. One exception comes from a study in Salmonella Typhimurium where the virulence defect of the tat mutant is shown to be likely indirect and primarily due to envelope defects 26 . Notably, we demonstrate in this work, that the P. aeruginosa tat mutants have an intact envelope (Supplementary Figure S1). Surprisingly, an orthologue of the E. coli and S. Typhimurium AmiC is encoded on the P. aeruginosa genome and is most probably a Tat substrate since it possesses a typical Tat signal peptide. This implies that the cell division mechanism is probably different between E. coli, S. Typhimurium and P. aeruginosa and strongly suggests that envelope defects cannot explain the attenuated virulence of the P. aeruginosa tat mutant.
In some organisms, the contribution of the Tat system to virulence is at least partially direct since individual substrates were shown to be involved in the Tat-dependent virulence 22,23 . In P. aeruginosa, the Tat system allows the export of PlcH, a known virulence determinant in a variety of mammal infection models 41,42 . PvdN and PvdP, both involved in the biosynthesis of Pyoverdine, an iron binding siderophore essential for P. aeruginosa full virulence, are also Tat substrates 13,38,43 . Although it has not been directly demonstrated, loss of PlcH secretion and/or Pyoverdine biosynthesis is probably one of the reasons for the attenuation of virulence of the P. aeruginosa tat mutant. Here we show that, the deletion of the tat genes affects, directly or indirectly, the secretion of multiple exoproteins, all of which are or could be involved in virulence. One of these proteins, ToxA, is reduced (a 2.46-fold decrease) in the tat mutant ( Fig. 1 and Table 2) due to transcriptional downregulation ( Fig. 2 and  ref. 13). Since ToxA has a major role in P. aeruginosa virulence 44 , its reduced level in the extracellular medium is likely to significantly contribute to the reduced virulence of the tat mutant. We also demonstrate that the absence of the Tat system induces a 7.7-fold decrease in the extracellular level of a new exoprotein, PA2377, also because of transcriptional downregulation (Figs 1 and 2 and Table 2). PA2377 is annotated as a hypothetical protein with similarity to the periplasmic iron-binding component of an ABC transporter. However, the absence of neighboring genes encoding an ABC transporter and the fact that we find PA2377 secreted in the external medium casts doubt on the predicted function. As mentioned by Coulthurst and collaborators, the similarity of PA2377/ ECA2134 to iron-binding proteins raises the possibility that it may be a novel iron-binding secreted protein involved in host interactions 45 . In agreement with this hypothesis it is interesting to note that PA2377 expression is upregulated when PAO1 cells are exposed to human airway epithelia 46 or are grown in acute and chronic murine models of infection 47 . Undoubtedly the molecular genetic mechanisms that lead to the differential gene expression of toxA, lapA, and PA2377 in the tat mutant will be interesting to study in greater detail.
We also show that Tat is directly required for the secretion of three new exoproteins (PA2699, PA3910 and PA4140) whose functions have not yet been demonstrated. Interestingly, transcriptome analyses of P. aeruginosa in interaction with human airway epithelial cells or grown in acute and chronic murine models of infection indicate that all three of these genes are upregulated during infection [46][47][48] . Moreover we found that these three proteins were only present in the exoproteome during phosphate starvation, a condition known to stimulate virulence factor production. Taken together, these findings suggest that these proteins might be important for P. aeruginosa infection of host cells and survival. PA2699 is predicted to encode a protein with homology to members of the YtcJ-like family (Pfam PF07969, amidohydrolase_3 family) which is part of the metal-dependent amidohydrolase superfamily. The second new Tat-dependent exoprotein, PA3910, is predicted to encode a phosphodiesterase/alkaline phosphatase D. PA3910 has also been named EddA for Extracelullar DNA (eDNA) degradation protein because it is part of a two gene operon with EddB/PA3909, an extracellular DNase involved in eDNA degradation 49 . Interestingly P. aeruginosa is able to grow on eDNA as sole source of carbon, nitrogen or phosphate 49 . Use of DNA as a nutrient may provide P. aeruginosa with a competitive advantage in environments where eDNA accumulates such as in biofilms or during interactions with other organisms 50  predicted to encode a phosphodiesterase/alkaline phosphatase D and because we show that it is extracellular, we hypothesize that PA3910 acts synergistically with PA3909 to degrade eDNA. Finally, the third new Tat-dependent exoprotein, PA4140, is predicted to encode a hypothetical protein with a substrate binding domain found in cholesterol oxidases (CHOs; Pfam 09129). CHOs are bacterial flavoproteins that catalyse the first step in the degradation of cholesterol 51 . Phylogenetic analysis showed that PA4140 is part of the family of class II CHOs where the FAD redox cofactor is covalently bound to the enzyme through a conserved histidine. This analysis also revealed that all class II CHOs possess a highly conserved twin arginine motif, which suggests that the use of the Tat pathway is a general feature for this family of enzymes (data not shown). In pathogenic bacteria CHOs have been shown to be involved in degradation of eukaryotic membranes and thus might help during the infection process 52 .
In conclusion, we present a global picture of the P. aeruginosa exoproteome under phosphate starvation, a condition known to stimulate virulence factor production. We have also identified novel exoproteins whose localization is affected directly or indirectly by the absence of a functional Tat system. These new "Tat-dependent" exoproteins are good virulence factors candidates, and their reduced levels could explain the attenuated virulence of the P. aeruginosa tat mutant. Future work will focus on the elucidation of the molecular mechanisms that may link these individual proteins to the attenuation of virulence in the P. aeruginosa tat mutant.

Experimental Procedure
Bacterial strains, plasmids and growth conditions. Bacterial strains, plasmids and growth conditions used and constructed in this study are described in Supplementary methods and listed in Supplementary Table S2. For clarity, PAO1 PA numbers were systematically used instead of PA14 PA numbers.

Analysis of RNA by quantitative Real Time PCR.
For qRT-PCR, total bacterial RNA was isolated with the SV Total RNA Isolation System (Promega) as described previously 53 . Next, RNAs were precipitated, treated with DnaseI (RTS Dnase kit -Ozyme), and cDNAs were synthesized by reverse transcription using Superscript III reverse transcriptase and random hexamers as described by the manufacturer (Invitrogen). To check for residual contaminating genomic DNA, control reactions without reverse transcriptase were analyzed using the 16S-RNAup/down primers. The cDNAs obtained were stored at − 20 °C until use. The expression levels of the different genes were assessed using SoFast EvaGreen Supermix (Bio-Rad) and the CFX96 Real Time System (Bio-Rad). Two microliters of a 1:25 dilution of the cDNA was used in a total volume of 15 μ L. The protocol was 98 °C, 2 min followed by 45 cycles of 98 °C, 5 s; 60 °C, 10 s; and 72 °C, 1 s. Data were acquired after each cycle at 72 °C. A melt curve was run at the end of the 45 cycles to test for the presence of a unique PCR reaction product. The primers used for the PCR amplification of cDNA, were designed using the primer3Plus program available at the website (http://www.bioinformatics.nl/cgi-bin/primer3plus/primer3plus.cgi/) and are shown in Supplementary Table S3. The gene transcription levels were normalized in each strain to the 16S ribosomal RNA gene (16SrRNA) and expressed as ratios to the values of the WT strain (set to 1). Samples were assayed in triplicate for each condition. Cellular fractionation, SDS-PAGE and immunoblotting. Whole cells and extracellular medium were prepared based on published procedure 54 . Briefly, cells were harvested and extracellular medium collected by centrifugation at 2000 × g for 10 min at room temperature. The pellet corresponding to whole cells was washed with Tris 40 mM pH 7.6, resuspended in SDS-PAGE loading buffer containing β -mercaptoethanol and heated at 95 °C for 10 min. Two third of the supernatant corresponding to the extracellular medium was carefully pipetted into a new tube and centrifuged once again for 5 min at 13,000 × g at room temperature. Next, extracellular proteins corresponding to two third of the supernatant were concentrated by adding TCA to a final concentration of 10% (vol/vol) and incubating on ice for at least 1h. The proteins were collected by centrifugation at 13,000 × g for 30 min at 4 °C, washed twice with ice-cold acetone, and collected again by centrifugation. The pellet corresponding to extracellular proteins was resuspended in SDS-PAGE loading buffer containing β -mercaptoethanol and heated at 95 °C for 10 min.
Fractionation of cells into spheroplast (cytoplasm and membranes) and periplasmic fractions were done as described 54 . Briefly cells corresponding to 5 uDO were pelleted and washed in 1 ml of 50 mM Tris-HCl (pH 7.6) before to be resuspended in 0,5 ml of 50 mM Tris-HCl (pH 7.6) − 200 mM MgCl 2 buffer. The mixture was incubated for 30 min at 30 °C with gentle shaking, left on ice for 5 min and then allowed to stand 15 min at room temperature. These steps were repeated a second time. The spheroplasts were then pelleted at 8,000 × g for 10 min at 4 °C and the supernatant was saved as periplasmic fraction. The pellet corresponding to the spheroplasts was resuspended in 1 ml of 50 mM Tris-HCl (pH 7.6), recentrifuged and the pellet, recovered in 1 ml of 50 mM Tris-HCl (pH 7.6), was disrupted by sonication 4 times for 15 s. Unbroken spheroplasts and protein aggregates (heavy fraction, HF) were collected by a low speed centrifugation at 2,000 × g for 10 min at 4 °C and the supernatant was saved as cytoplasm and membranes fraction (C + M). Protein aggregates present in the heavy fraction were solubilized in 1 ml of 50 mM Tris-HCl (pH7.6) 100 mM Glycine 8 M urea. The samples were incubated for 60 min on a roller drum at room temperature, and then ultracentrifuged at 120,000 × g for 60 min at 4 °C. The supernatant fraction containing the solubilized proteins and the pellet fraction containing insoluble material were retained. Each pellet was resuspended in SDS-PAGE loading buffer containing β -mercaptoethanol and heated at 95 °C for 10 min. Proteins corresponding to the periplasm, to the cytoplasm and periplasm fractions, or to insoluble material were precipitated as described above before to be resuspended in loading buffer containing β -mercaptoethanol and heated at 95 °C for 10 min.
Proteins were analysed by SDS-PAGE and immunoblotting as described in details in Supplementary methods. Extracellular proteome analysis and MALDI-TOF-TOF Mass Spectrometry. Extracellular proteome preparation and analysis were performed as described previously 55 . For preparation of the extracellular proteins, the equivalent of 200 OD units of cell culture supernatant was TCA precipitated overnight at 4 °C. The extracellular proteins were collected by centrifugation at 10,000 rpm for 90 min at 4 °C and the resulting protein pellet were washed 5-6 times with 96% ethanol (v/v) before to be dried. Separation of 200 μ g extracellular proteins was performed using the immobilized pH gradients in the pH range 3-10. In the second dimension, the equilibrated IEF strips were separated by 12,5% SDS-polyacrylamide gel electrophoresis and visualized with Coomassie brilliant blue. Quantitative image analysis was performed from the Coomassie-stained 2D gels using the DECODON Delta 2D software (http://www.decodon.com). The 2D gel images from wild type and the Δ tat mutant exoproteomes were aligned using a warp transformation. Before spot detection and quantification was performed, a fused 2D gel of both images was created using the 'union fusion' algorithm of Delta2D. Spot detection was performed in the fusion gel containing all spots present in both images according to the automatically suggested parameters for background subtraction, average spot size, and spot sensitivity. The resulting spot shapes were reviewed and manually edited in the fusion gel if necessary. This reviewed spot mask served as a spot detection consensus for all gel images, which was applied to both images to guide the spot detection and quantification. This enables spot quantification in all gels at the same locations resulting in 100% matching and in a reliable analysis of complete expression profiles. Normalization was performed by calculating the quantity of each single spot in percentage related to the total spot quantity per gel. Experiments were performed and quantified for two independent biological replicates and repeated technically twice. For identification of the proteins from 2D gels, spot cutting, tryptic digestion of the proteins and spotting of the resulting peptides onto the MALDI-targets (Voyager DE-STR, PerSeptive Biosystems) were performed using the Ettan Spot Handling Workstation (Amersham-Biosciences, Uppsala, Sweden) as described previously 56 . The MALDI-TOF-TOF measurement of spotted peptide solutions was carried out on a Proteome-Analyzer 4800 (Applied Biosystems, Foster City, CA, USA) as described previously 56 . The Mascot search was performed against the available P. aeruginosa database (http://www.pseudomonas.com).