Secretome Analysis Defines the Major Role of SecDF in Staphylococcus aureus Virulence

The Sec pathway plays a prominent role in protein export and membrane insertion, including the secretion of major bacterial virulence determinants. The accessory Sec constituent SecDF has been proposed to contribute to protein export. Deletion of Staphylococcus aureus secDF has previously been shown to reduce resistance, to alter cell separation, and to change the expression of certain virulence factors. To analyse the impact of the secDF deletion in S. aureus on protein secretion, a quantitative secretome analysis was performed. Numerous Sec signal containing proteins involved in virulence were found to be decreased in the supernatant of the secDF mutant. However, two Sec-dependent hydrolases were increased in comparison to the wild type, suggesting additional indirect, regulatory effects to occur upon deletion of secDF. Adhesion, invasion, and cytotoxicity of the secDF mutant were reduced in human umbilical vein endothelial cells. Virulence was significantly reduced using a Galleria mellonella insect model. Altogether, SecDF is a promising therapeutic target for controlling S. aureus infections.


Introduction
The Gram-positive pathogen Staphylococcus aureus is one of the leading causes of nosocomial infections [1]. Due to its acquisition of various resistance genes treatment of S. aureus infections have become increasingly difficult. Furthermore, the prevalence of methicillin-resistant S. aureus (MRSA) has been increasing in recent years. This has resulted in an alarming rise of communityassociated (CA-) MRSA infections in immunocompetent individuals [2,3,4]. In addition to its adaptive response to antibiotics [5], the success of S. aureus is based upon its huge array of virulence factors [6] helping S. aureus to avoid host immunity. These virulence factors have to be exported across the cytoplasmic membrane to reach their destined location: the membrane, the cell wall or the extracellular space. The main transport system is the Sec translocase, which is conserved in all three kingdoms of life [7,8]. Currently, the Sec pathway is best described in the Gramnegative bacterium Escherichia coli (as reviewed in [9,10]). The translocase consists of i) the heterotrimeric complex SecYEG, which forms a hydrophilic channel through the cytoplasmic membrane; ii) the motor protein SecA, an ATPase; and iii) the heterotrimeric complex SecDF-YajC. Proteins containing an Nterminal Sec signal peptide (SP) or a hydrophobic transmembrane segment are targeted to the translocase and transported through the channel in an unfolded state. For secreted proteins or membrane proteins with large hydrophilic loops, the driving energy is provided by the cycling of SecA, whereas ribosome-bound nascent chains are targeted mainly by inner membrane proteins [11] and are co-translationally exported powered by the translating ribosome. Small membrane proteins can also be inserted by YidC in a Sec-independent manner [12]. The auxiliary complex SecDF-YajC was shown to associate with SecYEG [13] as well as with YidC and is therefore believed to act as the linking molecule between SecYEG and YidC during Sec-dependent membrane protein insertion [12]. The integral membrane protein YajC was found to co-crystallize with the well-known E. coli multidrug exporter AcrB [14], which belongs to the resistancenodulation-cell division (RND) superfamily. Deletion of YajC only showed a weak phenotype and its exact function is still unknown [15,16].
SecDF also belongs to the RND superfamily and possesses the typical twelve transmembrane (TM) domains with two extracytoplasmic loops between TM1-2 and TM7-8, respectively [17]. Recently, Tsukazaki et al. resolved the crystal structure of the membrane protein SecDF of Thermus thermophilus [18]. Two conformations for the head subdomain P1 are observed, that seem to occur upon rotation by 120u [18]. Furthermore the P1 head was shown to interact with an unfolded preprotein, thereby preventing the substrate from backsliding and enhancing the translocation. This ATP-independent step in the later stage of protein translocation is driven by the proton motive force (PMF). Two conserved charged residues Asp519 and Arg247 are crucial for SecDF activity in E. coli and point mutations of the corresponding amino acids in T. thermophilus abolish ion channel activity [18].
In S. aureus, SecA and SecY have been shown to be essential [19,20,21]. Deletion of the non-essential secG leads to changes in the exoproteome, which are enhanced in a secG-secY2 double mutant [7]. SecY2 together with SecA2 belong to the accessory Sec pathway, which at present is known to export only one substrate, the serine-rich adhesin for platelets protein (SraP) [22]. However, virulence of the secG and secY2 single mutants and the secG-secY2 double mutant in mice is comparable to the parental strain [7].
We previously reported a S. aureus secDF mutant to have a pleiotropic phenotype influencing not only protein secretion, but also transcription and regulatory processes [23]. Resistance towards b-lactam and glycopeptide antibiotics was reduced. Furthermore, cell division was impaired and autolysis increased [23]. To determine the role of SecDF in the secretion of virulence factors and to assess its importance for pathogenesis, we performed a secretome analysis using isobaric tags for relative and absolute quantitation (iTRAQ) with subsequent LC-MS/MS. Major virulence determinants involved in adhesion to host proteins and cells, as well as in evasion of the host immune system were found to be decreased in the exoproteome of the secDF mutant. Important steps for establishing an infection were shown to be deficient in vitro in the secDF mutant in both methicillin sensitive and resistant S. aureus strains. Furthermore, secDF virulence was significantly reduced in a Galleria mellonella infection model.

Bacterial strains and growth conditions
Bacterial strains and plasmids used in this study are listed in Table 1. If not mentioned otherwise bacterial cultures were grown in Luria Bertani (LB) broth (Becton Dickinson, Difco Laboratories, Franklin Lakes (NJ), USA) at 37 uC. Bacterial cultures were grown under constant shaking and with a liquid-to-air ratio of 1:5 to assure good aeration. Media were supplemented with 50 mg/ml kanamycin (Sigma-Aldrich, St. Louis (MO), USA) or 10 mg/ml erythromycin (Sigma-Aldrich, St. Louis (MO), USA) when appropriate.

Sample preparation for secretome analysis
Strains were cultured until late exponential phase, which was shown to correspond to OD 600 1 in an earlier study [23] and centrifuged at 4 uC for 5 min. The supernatant (SN) was filtered (0.22 mM PES filter, Techno Plastic Products AG, Trasadingen, Switzerland) and mini EDTA-free complete protease inhibitors tablets (Roche, Rotkreuz, Switzerland) were added. For normalization purposes 500 pM enhanced GFP (Ams Biotechnology Ltd, Abingdon, UK) was added. The SN was concentrated by trichloroacetic acid (TCA) precipitation and washed twice with ice cold (220 uC) acetone. The pellet was resuspended in 0.5 M triethylammonium bicarbonate pH 8.5 (Sigma-Aldrich, St. Louis, USA) 0.5 mM tris(2-carboxyethyl)phosphine hydrochloride (TCEP, Sigma-Aldrich, St. Louis (MO), USA). Protein concentration was measured with the Quant-iT TM protein assay kit (Life Technologies, Invitrogen, Carlsbad (CA), USA).

Mass spectrometry
Dissolved samples were injected into an Eksigent-nano-HPLC system (Eksigent Technologies, Dublin (CA), USA) by an auto sampler and separated on a self-made reverse-phase tip column (200 mm6150 mm) packed with C 18 material (3 mm, 200 Å , AG, Bischoff GmbH, Leonberg, Germany). The column was equilibrated with 99% solvent A (1% ACN, 0.2% FA in water) and 1% solvent B (0.2% FA in ACN). Peptides were eluted using the following gradient: 0-3 min; 1-5% B, 3-57 min; 5-35% B, 57-63 min; 35-50% B and 63-70 min; 50-99% B, at a flow rate of 0.7 ml/min. High accuracy mass spectra were acquired with an AB Sciex 5600 (AB Sciex, Concord, Canada) in the mass range of 400-1250 m/z. Up to 40 data dependent MS/MS were recorded in high sensitivity mode of the most intense ions with charge state 2+, 3+ and 4+ using collision induced dissociation. Target ions already selected for MS/MS were dynamically excluded for 90 s after tree accuracies. After data collection, the peak lists were generated and analyzed using ProteinPilot TM 4.0 (AB Sciex, Concord, Canada). Data was searched against a SwissProt database (released January 2011). The following search parameters were used: Trypsin digestion, modifications of MMTS labelled cysteine, 4-plex-iTRAQ modifications of free amines at the Ntermini and of lysine, and 4-plex-iTRAQ modifications of tyrosine. Biological modifications and single amino acid exchanges were also included in the search. No normalization for iTRAQ ratios, such as BIAS correction, was applied, as normalization was performed by the addition of extrinsic GFP. The peptides without any 4-plex-iTRAQ label at the N-terminus or at a lysine were excluded from the analysis. The ProteinPilot cut-off score used was 1.3, which corresponds to a confidence limit of 95%. In total four biological replicates were analyzed in two independent iTRAQ experiments. The statistical analysis was assessed by the Student's t test.

Fibrinogen-and fibronectin-binding assay
The binding assay was adapted from O'Neill et al. [26]. Shortly, 100 ml two-fold dilutions of human fibrinogen (Merck, Calbiochem, Darmstadt, Germany) in 20 mM sodium citrate-HCl, pH 7.4 or human fibronectin (Merck, Calbiochem, Darmstadt, Germany) in PBS pH 7.4 were dispensed in flat-bottom 96-well Nunclon plates and incubated overnight at 4 uC. Following three PBS wash steps, the plates were blocked in 2 mg/ml bovine serum albumin (BSA, Thermo Scientific, Acros Organics, Geel, Belgium) in PBS for 2 h at 37 uC. Wells were washed three times in PBS. In the meantime bacterial cultures were grown to OD 600 1, washed twice in PBS and adjusted to OD 600 1 in PBS corresponding to ,10 8 cells/ml. Hundred ml of bacterial suspension was added per well (10 7 CFU/well), including a negative control (PBS) and incubated for 2 h at 37 uC. After repeated PBS washing steps, the cells were fixed with 25% formaldehyde (Applichem, Darmstadt, Germany), washed once with PBS and stained for 5 min in 0.5% crystal violet. Residual dye was removed with distilled water (dH 2 O) and plates were air-dried. The crystal violet stained cells were dissolved in 5% acetic acid and the absorbance measured at 570 nm. The experiment was performed with three technical and biological replicates, except for the positive control strain Cowan I in the Newman fibronectin binding assay, which was only performed with three technical replicates.

Tissue culture
Human umbilical vein endothelial cells (HUVECs) were purchased from Clonetics (Lonza, Basel, Switzerland) and maintained as previously described in M199 medium supplemented with antibiotics (penicillin 100 U/ml, streptomycin 100 mg/ml), glutamine at 2 mM, and 20% foetal calf serum, at 37 uC in a humidified 5% CO 2 atmosphere until they reached confluency [27]. One day before assays were performed, HUVECs were seeded at a density of 5610 4 cells/well into 48-well plates that were coated with 10 mg/ml fibronectin. Cells were used up to passage three.

Bacterial adherence and invasion to HUVECs
The capacity of the various strains to invade HUVECs was determined by the lysostaphin protection assay using conditions used for the MTT assays (see below) as previously described [28]. Adherence and invasion were expressed as % adherent and invading cells per well at the time point of measurement. Adherence and invasion of CHE482 were set to 100% for each run. Each experiment was performed in duplicate in three independent assays.

MTT assay
The ability of S. aureus to induce damage in HUVECs was assessed using an MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) assay. MTT is reduced by living cells to insoluble purple MTT formazan crystals using succinate, and the pyridine nucleotide cofactors, NADH and NADPH as substrates [29]. MTT production is therefore inversely related to cell death. MTT reduction results in a yellow to blue color change that can be quantified by measuring the absorbance at OD 570 [30]. MTT assays were performed as previously described with some modifications [31]. Briefly, wells containing HUVECs were rinsed twice with warm HBSS prior inoculation to remove medium containing antibiotics. The bacterial strains were grown overnight on sheep blood agar, resuspended in invasion medium (1% albumin and 25 mM HEPES, pH 7.3 in M199 without serum or Plasmids pCN34 S. aureus-E. coli shuttle vector, pT181-cop-wt repC aphA-3 ColE1, Km r , called pEmpty in this study [84] pCQ27 pCN34 derivative carrying secDF with its endogenous promoter, Km r , called pSecDF in this study [23] Abbreviations are as follows: antibiotics) to McFarland 0.5 (,5610 7 cells/ml), sonicated and diluted to the indicated multiplicity of infection (MOIs) [32,33,34]. Initial inocula were confirmed by serial fold dilution and plating on sheep blood agar. After 1 h or 3 h incubation at 37 uC in 5% CO 2 , respectively, the number of CFUs per well was determined to monitor growth. Then, the medium was aspirated, wells were washed with HBSS and 400 ml of fresh complete M199 medium containing 10 mg/ml lysostaphin was added [28,33,35]. After a total incubation time of 22 h at 37 uC in 5% CO 2 , 100 ml of 5 mg/ml MTT in HBSS were added to each well. 2 h later, the medium was removed and 150 ml of 0.04 M HCl in absolute isopropanol were added to solubilize the dye. Uninfected control wells which underwent the same washes were processed in parallel and served as negative control. Wells that only contained medium were used as background correction. Absorbance of 100 ml of the solution was measured at 570 nm. Specific cytotoxicity was calculated using the following formula: 1-(OD 570 experimental well/OD 570 control well). Each experiment was performed in triplicate in at least three independent assays.

Galleria mellonella virulence assay
To study the in vivo pathogenicity an invertebrate infection model was used as previously described by Peleg et al. [36]. Last instar larval stage G. mellonella were purchased at HRH Fishing Hebeisen, Zurich, Switzerland and stored at 4 uC until further use. Cells were grown until exponential phase (OD 600 0.5), washed twice in PBS and resuspended therein. Ten ml of bacterial suspension, corresponding to 10 6 CFUs, were injected into the last left proleg with a repetitive dispensing Tridak Stepper (Intertronic, Oxfordshire, UK) containing a 1 ml syringe with a 26-gauge needle. Thirty larvae were infected per strain. A control group was inoculated with 10 ml of PBS to assure the larvae were healthy and that death did not occur due to the needle prick or stress. Larvae were incubated at 37 uC and examined every 24 h for survival; they were considered dead when no movement occurred in response to touch. Additional larvae were inoculated separately to measure the bacterial burden which was monitored after 24, 48 and 72 hours. After incubating the larvae at 220uC, they were disinfected in 70% ethanol and rinsed with dH 2 O. The larvae were homogenized with the TissueLyser (Qiagen, Hombrechtikon, Switzerland) in 2-ml screw cap tubes containing a 5 mm stainless steel bead (Qiagen, Hombrechtikon, Switzerland) and 1 ml PBS, shaking for 10 min at 30 s 21 . Appropriate dilutions were plated on LB agar containing 50 mg/ml kanamycin to minimize growth of the normal flora of the larvae. Three independent experiments were pooled and analysed by log rank test. The survival-curves were plotted using the Kaplan-Meier method.

Deletion of secDF leads to an altered exoproteome
To determine the extent of the extracellular proteome affected by deletion of secDF, a quantitative secretome analysis was performed for the S. aureus wild type strain Newman and its mutant strain Newman DsecDF using iTRAQ. The SNs were collected during late exponential phase, where we found strongest SecDF expression (data not shown). Samples of four biological replicates were prepared for LC-MS/MS as described in materials and methods.
A total of 230 S. aureus proteins were quantified and their putative localization was determined with different bioinformatics tools ( Table 2, Table S1). Thirty-eight proteins had a predicted SP [37], 34 thereof were predicted Sec-type SPs [38]. These included seven cell wall proteins containing the LPXTG cell wall retention motif (SpA, SdrE, ClfA, IsdA, FnBPA, ClfB) [39,40] or LysM domains (Sle1/Aaa) [41]. In Newman DsecDF, in comparison to the wild type Newman, the extracellular levels of 27 proteins were altered significantly by at least two fold increase (two proteins) or two fold decrease (25 proteins); 21 of these proteins contained a Sec-type SP (Figure 1). Of the six remaining proteins, three were membrane proteins; OatA and the LytR-CpsA-Psr proteins NWMN_0925 (SA0908) and MsrR [42,43]. Furthermore, the SP containing, but Sec-independent enterotoxin A (SEA) and the delta hemolysin (Hld) were identified.
Peptides of the N-terminal region of the glycerol ester hydrolase Geh were identified and found to be reduced in Newman DsecDF. Because one of the four prophages in strain Newman is located in the Geh gene geh, leading to a truncated and inactive enzyme [50], this finding is not phenotypically relevant in the strain Newman background. Nonetheless, in other S. aureus strain backgrounds, deletion of secDF can be expected to reduce lipase activity.

Complementation of the altered virulence factor expression in the mutant strain Newman DsecDF and verification of its phenotype in a second strain background
To validate and further highlight the role of SecDF in S. aureus, a different, methicillin resistant strain background was included in the following confirmatory experiments. The low level methicillin resistant strain CHE482 represents a CA-MRSA [51,52,53]; its corresponding secDF mutant CHE482 DsecDF carries a transposon at position 1029 bp leading to a non-functional SecDF protein [54]. To complement the deletion of secDF, the plasmid pSecDF (pCQ27) containing secDF with its endogenous promoter from strain Newman was introduced into the mutant strains Newman DsecDF and CHE482 DsecDF, yielding Newman DsecDF pSecDF and CHE482 DsecDF pSecDF, respectively. The empty vector pEmpty (pCN34) was introduced into the wild type strains Newman and CHE482 and the secDF mutants Newman DsecDF and CHE482 DsecDF to ensure no additional effects were caused by the plasmid (Table 1).
To confirm the secretomics results, Western blot analysis of the SN from these strains were performed for selected proteins found to be altered in the secretome of Newman DsecDF. Specific antibodies were used for the Sec-dependent proteins CHIPS, Eap, FnBPA, LytM and SceD, as well as the Sec-independent SEA.
As expected, the SN of Newman DsecDF showed reduced protein amounts of CHIPS, FnBPA, Eap and SEA and increased amounts of LytM and SceD in comparison to the wild type and the complemented mutant (Figure 2A). This phenotype was validated and confirmed for CHIPS, FnBPA and SceD in strain background CHE482 ( Figure 2B).

Decreased adherence of the secDF mutant to attached human fibrinogen, fibronectin and endothelial cells
To establish an infection, the ability of S. aureus to adhere to host proteins and cells is essential and permits the bacteria to invade into the cells as shown previously for S. aureus [55]. Our secretomics screen revealed several factors belonging to the microbial surface components recognizing adhesive matrix molecules (MSCRAMM) or the secretable expanded repertoire adhesive molecules (SERAM) to be reduced in Newman DsecDF (Table 3). Therefore, binding of the secDF mutants to immobilized human fibrinogen and fibronectin, respectively, was studied in vitro. Ninety-six-well-plates were coated with various concentrations of human fibrinogen and fibronectin, respectively. Bacteria were allowed to adhere to the coated wells, washed, fixed and coloured with crystal violet as described in materials and methods.
Binding of Newman DsecDF pEmpty to fibrinogen was reduced by up to 50% in comparison to the wild type strain Newman pEmpty and the complemented mutant Newman DsecDF pSecDF   Figure 3A). In CHE482, binding of CHE482 DsecDF pEmpty was reduced at low concentrations of fibrinogen, but not at concentrations above 2 mg/ml ( Figure 3B). This phenotype was restored in both strains by the complementing plasmid pSecDF. Fibronectin binding proteins are not anchored to the cell wall in strain Newman due to a point mutation leading to an early stop codon and truncated proteins without sortase motif [56]. Therefore, strain Cowan I was used as a functional control in the binding assay of Newman to human fibronectin [56]. As expected, the binding capacity of strain Newman pEmpty to fibronectin was up to ten times lower as compared to strain Cowan I ( Figure 3A). However, deletion of secDF further reduced binding  [56,80,81,82] and are indicated by a ring. Additional protein bands due to unspecific binding to protein A or Sbi are indicated by an asterisk. doi:10.1371/journal.pone.0063513.g002 Table 3. Adhesins, which were reduced in the SN of Newman DsecDF.

Proteins
Binding to human fibrinogen Binding to human fibronectin Ref. to fibronectin in the Newman background and was partially restored by the complementing plasmid. In CHE482, the secDF mutant showed a similar concentration dependent phenotype for fibronectin as for fibrinogen, with the CHE482 DsecDF pEmpty displaying a reduced binding at lower fibronectin concentrations ( Figure 3B). These findings suggest the presence of additional fibrinogen and fibronectin binding proteins in the CHE482 background, possibly on mobile genetic elements, which reduce the effect of a secDF deletion on adhesion to fibrinogen and fibronectin.

MSCRAMM
In a next step, we investigated whether the secDF deletion would also affect adherence to host endothelial cells (human umbilical vein endothelial cells, HUVECs). Because of the point mutation in the FnBPs of strain Newman mentioned above, this strain has been previously shown to be weakly adherent to HUVECs [56]. Therefore, we used the secDF mutant of strain CHE482 to test adherence and found significantly reduced adherence as compared to the wild type strain ( Figure 4A). This effect was restored in the complemented mutant CHE482 DsecDF pSecDF. In accordance with the previous findings [56] adherence by strain Newman was only 12.766.6% of strain CH482.
secDF inactivation leads to reduced cytotoxicity Many of the virulence factors produced by S. aureus destroy host tissue to allow dissemination. Thus, we analyzed the ability of the secDF mutant to damage endothelial cells using a previously Figure 3. Impact of SecDF on fibrinogen and fibronectin binding. Binding properties of Newman and CHE482 strain sets to immobilized human fibrinogen and fibronectin was assessed as described in materials and methods. (A) Newman pEmpty, Newman DsecDF pEmpty and the complemented mutant Newman DsecDF pSecDF. FnBPs in the Newman background are truncated due to a point mutation leading to a stop codon before the sortase motif, which is required for cell wall anchoring, and therefore are secreted [56]. Hence, Cowan I was used as a functional control strain in the fibronectin binding assay. (B) CHE482 harbouring empty plasmid pEmpty (pCN34), the mutant strain CHE482 DsecDF harbouring empty plasmid pEmpty (pCN34) and the complemented mutant CHE482 DsecDF pSecDF. Mean of three independent experiments are shown with their standard deviation. doi:10.1371/journal.pone.0063513.g003 published MTT assay [31]. This assay measures MTT reduction by living cells and is inversely related to cell death. We previously showed that MTT reduction by HUVECs internalized S. aureus is negligible [31]. To evaluate cytotoxicity of strain Newman, we used similar conditions as previously described using a MOI of 50 [33]. In this assay, endothelial cells are infected with bacteria for 3 h. Then, extracellular bacteria are killed and damage is assessed 24 h after the addition of the bacteria. Strain Newman did not induce substantial damage under these conditions (15.8610.9% damage) and was not suited to study the effect of secDF inactivation on S. aureus cytotoxicity. We therefore decided to use strain CHE482, which caused a significant higher damage (69615%) and the corresponding secDF mutant. However, CHE482 DsecDF exhibited a growth defect in invasion medium during the 3 h invasion step ( Figure 4B). We therefore decreased the invasion time to 1 h, while bacteria appeared to be in the lag-phase showing no measurable difference in OD 600 or CFUs between the strains ( Figure 4B). Because these conditions have not been previously used to study S. aureus-induced endothelial cell damage, it was essential to identify an optimal MOI. Figure 4C shows the increasing cytotoxicity of strain CHE482 with increasing MOIs ranging from 25 to 2500. Again, damage induced by strain Newman under these conditions was very low. Using an MOI of 2500, we found that secDF inactivation in strain CHE482 led to a substantial reduction in cytotoxicity ( Figure 4D). This phenotype could be complemented using strain CHE482 DsecDF pSecDF, which contains the secDF locus on a plasmid. To determine whether the reduced cytotoxicity of CHE482 DsecDF was due to a reduction in endothelial cell invasion, invasion was assessed under similar conditions as for the MTT assay. The mutant strain CHE482 DsecDF exhibited significantly reduced invasion as compared to the wild type strain ( Figure 4E). This effect was restored in the complemented mutant CHE482 DsecDF pSecDF. In accordance with the low adherence and cytotoxicity of strain Newman, invasion was only 1.661.0% of strain CHE482.

Inactivation of secDF leads to reduced pathogenicity in an insect infection model
To assess the influence of SecDF on virulence, Newman strains were injected into last instar larval stage G. mellonella and survival of the larvae was monitored over time. Resuspension buffer for bacterial cultures was used as a negative control. The dimensions and masses of larvae in one group spanned representative and comparable ranges, as preparatory experiments did not show any correlation concerning larval weight and survival (data not shown). Pathogenicity of Newman DsecDF pEmpty in G. mellonella was significantly reduced (P,0.001) in comparison to the wild type Newman pEmpty and the complemented mutant Newman DsecDF pSecDF ( Figure 5A). To ensure that the attenuated virulence was not due to growth deficiencies of the mutant, the bacterial burden per larvae was measured after 24, 48 and 72 hours. Variation of CFU per larvae was rather high within the strains, for instance at 24 hours post infection (hpi) the wild type showed a bacterial burden ranging from 1.54610 5 to 4.62610 7 ( Figure 5B). However, all three strains were able to multiply in the larvae within a similar range during the first 48 hpi, with CFUs being even higher in the secDF mutant after 72 hpi than in the wild type and the complemented mutant. The different peak time points of CFU/ larvae in the secDF mutant (72 hpi) compared to the wild type or complemented mutant (48 hpi) was not reflected in the survival of G. mellonella, further indicating a strongly reduced virulence in Newman DsecDF pEmpty, that was not compensated by increased CFU numbers.

Discussion
Secretion of numerous virulence factors such as adhesins, proteases, autolysins and toxins rely on a functional Sec secretion system for the export across the cytoplasmic membrane [38]. So far, most studies of the Sec machinery were performed in the Gram-negative and Gram-positive model organisms E. coli and B. subtilis, respectively. This study focused on the impact of the auxiliary SecDF protein on virulence in the human pathogen S. aureus by analysing the exoproteome of Newman DsecDF in a gel-free approach, followed by in vitro and in vivo virulence studies.
Our secretomics results were in good agreement with previous data [23]. A relatively high number of cytoplasmic proteins was found in the SN, which has already been observed by other research groups and in different microorganisms [57,58,59,60]. This occurrence has been assigned to cell wall turnover, proteolytic processing, shedding, natural lysis and lysis due to handling [61,62,63]. Nevertheless, we identified several new proteins to be influenced by SecDF. This was confirmed by Western blot analysis and could be complemented for the proteins CHIPS, Eap, FnBPA, LytM and SEA.
Eight out of 11 proteins, identified in both the previously characterized S. aureus RN4220 secG mutant [7] and Newman DsecDF, showed a similar trend and were found to be reduced in both mutants in comparison to the wild types (Sle1, Geh, Hlb, Hla, HlgB, HlgC, NWMN_1927 and YfnI). Three proteins (IsaA, Spa and SsaA) were found to be increased in the secG mutant, but reduced in Newman DsecDF. Different methods, sampling time points and strain backgrounds could have contributed to those divergent findings. Another possibility is that SecDF is required for a subset of proteins deviating from the ones that require SecG. In addition, since they have different functions in the Sec pathway, the absence of SecG or SecDF might lead to only partially overlapping phenotypes.
Two Sec-dependent proteins were found to be increased in Newman DsecDF; the autolysin LytM and the transglycosylase SceD. Kouwen et al. had observed, that a B. subtilis LipA hyperproducing strain could export the normally Sec-dependent LipA, which also contains a potential Tat RR-SP via both Tat pathways [64]. S. aureus LytM and SceD do not contain a potential Tat RR-SP; whether they are secreted by another transport system in Newman DsecDF remains to be determined.
In the secG mutant only one Sec-independent protein was identified to be reduced; YfnI/LtaS [7], which was also found reduced in Newman DsecDF (Table S1). Additionally, the two Secindependent proteins Hld and SEA were found to be significantly reduced upon deletion of secDF. Hld belongs to the phenol-soluble modulins and has recently been shown to be exported by the ABC transporter Pmt [65]. The fact that the amounts of certain Secindependent proteins were changed in both the secG and the secDF mutant, points towards indirect effects caused by the deletion of these Sec translocase constituents. Slightly reduced levels for the regulatory and Hld-encoding mRNA RNAIII have been shown previously in Newman DsecDF [23]. Because RNAIII transcription is regulated by the two-component system accessory gene regulator (agr) the reduction of RNAIII levels suggest that agr is indirectly affected by secDF deletion [23,66]. Clearly, more work is required to confirm this hypothesis and to determine whether other regulatory processes are influenced directly or indirectly by the absence of SecDF.
Secretome analysis showed decreased SpA levels in Newman DsecDF consistent with a previously reported reduction of spa transcription [23]. In contrast, previous and present Western blot analysis showed similar levels of full-length SpA in the wild type and Newman DsecDF. A possible explanation for these seemingly contradictory findings could be the reduced amounts of SspA protease in the Newman DsecDF (Table S1). SspA is the main player of SpA degradation [67]. Higher SpA and SspA levels in the wild type would lead to more SpA degradation fragments, which can be detected by secretome analysis and increase the total protein amounts attributed to SpA, but would be too small to be detected by Western blot analysis. Whether this is the case still needs to be investigated.
Several well known virulence factors playing a central role during the initial steps of an infection were decreased in Newman DsecDF. These changes led to reduced binding to the two important human host factors fibrinogen and fibronectin in vitro. Although the effect of the secDF deletion on the adhesion intensity varied between the two strains tested and was smaller in CHE482, we also found significantly reduced adhesion of CHE482 DsecDF to HUVECs. This was accompanied by a significantly decreased invasion and must be attributed to the sum of impaired factors in the mutant and the multitude of additional host matrix factors S. aureus can bind to. In this study, the two important invasion factors Eap and FnBPA were found to be reduced [68,69]. While the receptor of Eap is still unknown, the FnBPs bind via the bridging factor fibronectin to the host receptor integrins a 5 b 1 , which is sufficient to induce the uptake of staphylococci [68]. Thus, the low cytotoxicity in HUVECs displayed by CHE482 DsecDF must be assumed to be at least partially caused by its defective adhesion and invasion, in addition to a reduced production of toxins. The decreased levels of numerous virulence factors and the reduced cytotoxicity, indicative for virulence in mice and rabbits [33,70,71], suggested the virulence of the secDF mutant to be attenuated. We therefore determined the pathogenicity of Newman DsecDF pEmpty in the invertebrate model host G. mellonella, sharing several features of the innate immune response with mammalians [72,73]. A positive correlation has been shown between the pathogenicity of microorganisms in insects and in mice [74,75] and in recent years this model has been increasingly used for assessing the virulence of S. aureus [36,76,77,78,79]. As expected, virulence was significantly attenuated in Newman DsecDF pEmtpy. The similar bacterial load per larvae indicated that Newman DsecDF pEmtpy has the ability to multiply as well as the parent strain and that the generally diminished virulence factor expression Newman DsecDF pEmtpy is the cause for its reduced virulence.
Taken together, our data provide new insights on the relevance of SecDF in S. aureus pathogenicity. We showed that deletion of secDF affects an important part of the extracellular proteome leading to reduced adhesion, invasion and cytotoxicity, as well as reduced virulence in G. mellonella. Because both MSSA and MRSA secDF mutants are less resistant to well established antibiotics [23], SecDF is an interesting target for the development of novel antimicrobial substances.

Supporting Information
Table S1 Proteins found in the supernatant of S. aureus Newman and Newman DsecDF and their predicted localization. (PDF)