SecDF as Part of the Sec-Translocase Facilitates Efficient Secretion of Bacillus cereus Toxins and Cell Wall-Associated Proteins

The aim of this study was to explore the role of SecDF in protein secretion in Bacillus cereus ATCC 14579 by in-depth characterization of a markerless secDF knock out mutant. Deletion of secDF resulted in pleiotropic effects characterized by a moderately slower growth rate, aberrant cell morphology, enhanced susceptibility to xenobiotics, reduced virulence and motility. Most toxins, including food poisoning-associated enterotoxins Nhe, Hbl, and cytotoxin K, as well as phospholipase C were less abundant in the secretome of the ΔsecDF mutant as determined by label-free mass spectrometry. Global transcriptome studies revealed profound transcriptional changes upon deletion of secDF indicating cell envelope stress. Interestingly, the addition of glucose enhanced the described phenotypes. This study shows that SecDF is an important part of the Sec-translocase mediating efficient secretion of virulence factors in the Gram-positive opportunistic pathogen B. cereus, and further supports the notion that B. cereus enterotoxins are secreted by the Sec-system.


Introduction
Bacillus cereus sensu stricto is a Gram-positive spore-forming bacterium producing several toxins associated with food-borne disease.While cereulide has been shown to cause the emetic syndrome [1], the pore-forming toxins cytotoxin K (CytK), haemolysin BL (Hbl) and nonhaemolytic enterotoxin (Nhe) inflict diarrhea [2][3][4].Fagerlund and co-workers have advocated that secretion of CytK and Nhe-and Hbl-components in B. cereus is directed via the Sec-translocase system [5].SecDF is widely conserved across bacterial genera but is believed to be an accessory, non-essential protein component of the Sec-complex, the main protein secretion machinery in bacteria [6][7][8].SecDF deletion has been shown to result in low-temperature sensitivity, aberrant cell division and impaired protein secretion in Escherichia coli, Staphylococcus aureus and Bacillus subtilis [9][10][11][12].SecDF exhibits the typical structure of RND-type (Resistance-Nodulation-Cell Division) transporters with 12 transmembrane helices and two large extracytoplasmatic loops.However, tertiary and quarternary structures differ from the well described drug efflux-mediating RND transporters.Members of the RND transporter family are generally required for effective efflux of potentially cytotoxic compounds from the cell [13], and their overexpression can confer multi-drug resistance in human pathogens [14].However, drug efflux is not necessarily the major function of most of the exporters, and their involvement in processes such as metal-ion homeostasis, quorum sensing, maintenance of cell homeostasis, interaction with plant or animal hosts, or efflux of toxic metabolic intermediates, fatty acids or other substances produced by the bacteria themselves, has been reported [15][16][17][18][19].
The exact role of SecDF during the protein translocation process has not yet been elucidated in detail.Based on SecDF crystal structures and in vitro experiments Tsukazaki and coworkers presented a model describing the proton motive forcedependent role of SecDF during later stage of protein translocation, where efficient protein translocation by SecDF is facilitated by preventing the emerging preprotein from backsliding into the SecYEG channel [20].Indeed, the charged residues shown to be important for H + translocation by other RND-type transporters are conserved in the SecDF proteins [8].Interestingly, in an early work Schiebel et al. estimated that in the absence of the PMF the costs of protein translocation increase from under 200 ATP units to several thousand ATP molecules per protein [21].
Previous reports suggested that SecDF is not an essential part of the Sec-translocase and fulfills only a noticeable function in secretion under protein hyper-expression and/or low temperature conditions.However, since the protein is ubiquitous, a more profound biological function is plausible.An important role in protein secretion has recently been acknowledged by Quiblier and co-workers ( [11,12], and indeed, a Staphylococcus aureus secDF knock out strain displays less virulence in an insect model, and less cytotoxicity to human umbilical vein endothelial cells, than its isogenic wild type strain [12].In this study we report that SecDF exhibits a substantial function in protein secretion in the sporeforming opportunistic pathogen B. cereus, severely affecting cellular export of major toxins and other virulence factors and resulting in reduced virulence of the DsecDF mutant in insect larvae, thus providing additional evidence for Sec-dependent secretion of the B. cereus enterotoxins.

Results
The DsecDF knock out mutant is affected in growth, shape and motility A markerless secDF deletion mutant was investigated for phenotypic alterations relative to the isogenic wild type strain B. cereus ATCC 14579.Bolhuis et al. reported a strong activation of the B. subtilis secDF promoter by the addition of glucose to the growth medium [9].There was a small but consistent lag in growth during the exponential phase of DsecDF mutant compared to the wild type in LB medium at 30uC as well as at 37uC (Fig. 1A and data not shown).In LB medium supplemented with 1% glucose (from now on referred to as LBG) growth of the DsecDF mutant was slightly slower than the wild type, and the DsecDF mutant did not reach the culture densities of the wild type at either 20uC, 30uC or 37uC, during the time window investigated (Fig. 1A and data not shown).After 24 h growth, microscopy showed that most DsecDF mutant cells appeared in uncharacteristically crooked chains (Fig. 1B).These growth-related effects of the secDF deletion could be circumvented by complementation with SecDF (Fig. S3, left).The mutant displayed a smaller colony size compared to the wild type on LB and LBG agar plates, and this was more pronounced in the presence of glucose (Fig. 1C) and at lower temperatures (data not shown).Growth of the wild type and mutant strains on B. cereus agar containing bromothymol blue as pH indicator did not indicate differential production of acidic byproducts as a result of glucose fermentation (data not shown).
Microscopy analyses of LBG liquid cultures had clearly showed a decreased motility of the DsecDF mutant compared to the wild type after 4 h of growth.When analyzed on 0.3% LB agar plates, motility of the DsecDF mutant was approximately half of the wild type, whereas following addition of glucose, maltose or sucrose, the corresponding relative motility was below 10% (Fig. 2A).Severe reduction in motility was also observed on 0.7% LBG agar (Fig. 2B).In B. subtilis secretion of the surface-tension reducing compound surfactin enables flagellum-independent motility [22].To test if differences in surface tension could explain the mutant motility phenotype, Tween 80 was added to the medium [23].This resulted in partly restored motility of the secDF mutant to almost 80% of wild type movement on medium supplemented with Tween 80. Simultaneous addition of Tween 80 and glucose resulted in 75% inhibition of motility relative to wild type under the same conditions (Fig. 2A), showing that a missing surfactant was not the only cause of reduced motility in the DsecDF mutant.Atomic force microscopy (AFM) amplitude images of DsecDF and wild type cells grown for 4 h in LBG showed that the mutant displayed about five times reduced number of flagella per cell in two independent experiments (Fig. 2C), which may explain its decreased motility (Fig. 2A and B).In addition, AFM amplitude images revealed a higher number of extracellular structures in the wild type compared to the DsecDF mutant samples (Fig. 2C), possibly representing extracellular vesicles [24].

SecDF deletion reduces resistance of B. cereus to xenobiotics
The 12-transmembrane secondary structure of SecDF is shared by other RND-type transporters known to mediate the efflux of a wide range of xenobiotics.In order to test if SecDF displays similar functions in addition to its role in protein translocation, the effect of SecDF expression in E. coli DacrB on the susceptibility towards various compounds relative to an empty vector control was tested (Tables S1 and S2).Deletion of acrB in E. coli, coding for the main xenobiotic efflux transporter in this organism, leads to hypersusceptibility to various toxic compounds [25].Furthermore, in search for additional phenotypic traits resulting from secDF deletion in B. cereus, minimal inhibitory concentration (MIC) and disk diffusion assays of several xenobiotics were conducted with the B. cereus DsecDF mutant and wild type strains.The DsecDF strain exhibited reduced tolerance to SDS and to the aminoglycoside antibiotic gentamicin, and the reduction in tolerance was amplified in the presence of glucose.We also observed a fourfold decrease in the resistance towards the widely used food preservative sodium benzoate, and a two-fold decreased resistance towards the antimicrobial polymyxin B in LBG medium.Strong effects on growth of the mutant were observed with alcoholic plant extracts of peppermint, calabash plant, and tea tree (Fig. S1).While expression of SecDF from the vector pHT304-pXyl in the wild type B. cereus strain did not result in modified resistance to any of the seven compounds tested, heterologous expression of SecDF in E. coli DacrB produced increased sodium benzoate resistance (Table S2 and data not shown), in accordance with the results from the B. cereus secDF deletion mutant.

The secDF deletion mutant exhibits a reduced level of secreted proteins
To test the effect of deleting secDF on the secretome of B. cereus, we compared the amount of proteins in the growth medium of the wild type and mutant.Since the phenotypic alterations of the DsecDF mutant seemed to be stronger when grown in glucose-containing medium, secretome analyses were carried out in the presence of 1% glucose.Silver staining following SDS-PAGE revealed a substantial overall reduction of total protein in the growth medium of the DsecDF mutant relative to wild type at different stages of growth (Fig. 3).In addition, an increase of small proteins in the DsecDF mutant secretome was observed.This, however did not seem to be due to an exacerbated proteolytic activity or autolysis rate of the mutant (see method section).

Label-free mass spectrometry reveals an important role for SecDF in secretion of virulence determinants and putative cell wall modulating enzymes
To further identify which proteins are specifically dependent on SecDF for transport, a label-free proteomic analysis was conducted on three biological replicates of sterile filtered culture supernatants from the DsecDF strain and its respective wild type, after 4 h of growth in LBG medium at 30uC.At this time point, motility of the deletion mutant was visibly reduced and the density of the culture was typically about 65% of the wild type strain (Fig. 1A).In total, 96 proteins were confidently identified in the secretome samples (Table S3).According to the PSORTb algorithm (version 3.0.2;[26]) 29 of these proteins (30%) were either extracellular or cell wall-associated, six (6%) were anticipated to be located within the cytoplasmic membrane, while the majority (55) of the proteins were of cytoplasmic origin (57%).For the remaining 6 proteins (6%) no convincing localization prediction could be made based on sequence similarities with known proteins.However, two of the six contained a putative signal peptide, suggesting an extracellular localization.
Using a paired Students T-test on normalized spectral abundance factors (NSAF, [27]) 34 of the 96 identified proteins were shown to be present at significantly different levels when comparing growth supernatants of the DsecDF mutant and the B. cereus wild type (Table 1), indicating fundamental differences in protein secretion between the strains.All the proteins present at reduced levels in the culture supernatant of the DsecDF strain compared to the wild type, were predicted or are known to be extracellular or cell wall-associated (Table 1).Phospholipase C and sphingomyelinase were major protein components in the growth medium of the wild type cells, while they were absent or nearly absent in the secDF mutant (Table 1).In addition, the Hbl and Nhe enterotoxin components and cytotoxin K were highly abundant in the extracellular environment of the wild type, while being present at low levels or absent in the mutant secretome.The M9A/M9B -type collagenase C (ColC, BC0556) was 18-fold reduced in the supernatant of the mutant.Another putative collagenase, Sfp (BC3762; also annotated as S-layer protein A), belonging to the intracellular subtilisin-related peptidase S8 group, was identified only in the wild type supernatant, in moderate amounts.
Due to frequent flagellar turnover, flagellum structural components are common constituents of bacterial secretomes [28][29][30][31].In agreement with the observed motility deficiency of the mutant and the highly reduced number of flagella seen in AFM experiments, levels of several flagellum structural proteins were reduced in the DsecDF secretome.B. cereus ATCC 14579 encodes three highly similar flagellin proteins (Q81FD3, Q81FD4, Q81FD5), whose peptide fragments could not be distinguished from each other by the applied analysis method and were therefore analyzed together.In total, slightly less flagellin was detected in the DsecDF mutant growth medium (60% of wild type level, p = 0.049).Furthermore, the cell-wall associated hook protein FlgE was detected in one of three mutant replicates only (4% of wild type level, p-value 0.12, table S3), and the three structural flagellar hook-associated proteins 1, 2 and 3 were present on average 18%, 40% and 52% of the wild type levels, respectively (p-values = 0.06, 0.35, 0.08, table S3).
Several cell wall-associated proteins were also found to be differentially secreted in the DsecDF mutant, most prominently the putative murein hydrolases BC0991 and BC1991 which were absent in the medium of the DsecDF mutant (Table 1).Furthermore, EntB (BC2952), annotated as enterotoxin/cell-wall binding protein, was present at 33-fold lower levels (p,0.001) and was, in fact, not detectable in two out of three biological replicates.The similar proteins EntA (BC5239) and EntC (BC0813) did not show this trend, as the abundances varied across the samples.
In contrast to the less abundant proteins in the DsecDF secretome, most of the 18 proteins found at higher levels in the growth medium of the mutant relative to wild type typically had intracellular functions, including ten ribosomal proteins, with sizes ranging between 11 and 25 kDa.Finally, it is also worth noticing that the so far uncharacterized putative enterotoxin BC1953 was among the most abundant proteins in the wild type secretome at the time of sampling (Table S3).

Toxin translocation is reduced in the DsecDF mutant
Mass spectrometry analysis of the DsecDF secretome indicated a potentially important function for the SecDF moiety in translocation of B. cereus proteins, including toxins and other virulence factors.To further characterize this phenomenon, Western blot analyzes were conducted on both the growth medium and cell lysates using monoclonal antibodies against the Hbl toxin components L1 and L2 as well as against NheA and NheB [32,33] (Fig. 4).In the absence of added glucose, the level of toxin components in the growth medium was reduced in the DsecDF mutant compared to the wild type after 3 h, 4 h and 6 h in LBG medium, but reached wild type levels after 6 h incubation in LB medium (Fig. 4).Over the same time period, cell-associated toxin components accumulated to a higher level in the DsecDF mutant compared to the wild type (Fig. 4).Complementation assays in the DsecDF mutant restored its ability to translocate and averted the cellular accumulation of the indicated toxin components (Fig. S3).Thus, differences in protein abundances in the DsecDF mutant secretome are most likely due to inhibition of toxin translocation across the plasma membrane rather than downregulation of toxin gene transcription or translation.In general, extracellular toxin levels decreased both in wild type and DsecDF mutant cells when grown in the presence of added glucose (Fig. 4), however, the difference in cellular accumulation of toxin components in the DsecDF mutant relative to the wild type was most prominent in cultures grown in LBG rather than in LB (Fig. 4).
While PC-PLC was the second most abundant protein in the wild type secretome, levels in the DsecDF mutant culture medium were below the detection limit (Table 1).To confirm the proteome data, both strains were grown in LB and LBG, and culture medium was collected periodically.The PC-PLC activity of sterilefiltered medium on egg yolk agar indicated reduced PC-PLC secretion by the DsecDF mutant (Fig. 5B).Notably, growth in LBG resulted in hardly any visible PC-PLC activity in the mutant culture.Simultaneously, secretion of PC-PLC into the agar by actively growing cells was not detected on LBG agar containing egg yolk (Fig. S2A).In a more sensitive approach to determine PC-PLC activity, culture supernatants were incubated with egg yolk suspension and the substrate degradation measured photometrically (Fig. 5A).These experiments showed that extracellular PC-PLC activity from the secDF mutant grown in LBG remains at about 35% of the wild type activity over the studied time course.While the presence of glucose reduced PC-PLC activity in both strains, the effect was more pronounced for the mutant (Fig. 5).

Deletion of SecDF affects virulence in Galleria mellonella
In order to test if the observed reduction of secreted virulence factors in batch cultures is mirrored by diminished virulence of the DsecDF mutant, in vivo infection assays using Galleria mellonella (G.mellonella) larvae were conducted [34].Survival of larvae 24 h and 72 h post infection either administered by oral feeding or by injection into the insect blood hemocoel was monitored and the LD 50 s of the wild type strain and of the DsecDF mutant were evaluated by Probit analysis (Table 2).The confidence limits at the 95% interval of lower (LDL) and upper level (UDL) doses of mutant and wild type strains were not overlapping.Thus, the about 4-fold and 3-fold differences in the dose killing 50% of the exposed larvae at 24 h and 72 h post infection by direct injection into the hemocoel of various doses (2x10 3 to <1x10 5 ) of vegetative bacteria, respectively, are significant (p-values #0.05).Estimation of virulence at 24 h and 72 h post oral infection revealed about 17-fold and 13-fold LD 50 difference, respectively.These results clearly indicate a reduced virulence of the DsecDF mutant strain towards the insect model which is more pronounced if larvae are infected orally.

Transcriptional profiling of the DsecDF mutant reveals the induction of multiple cellular stress responses
With the purpose of revealing molecular mechanisms linking the protein secretion defect and potential underlying processes to the observed phenotypic changes of the DsecDF mutant, a global transcriptional profiling experiment was conducted.Custom-made microarray slides were hybridized with reverse transcribed RNA extracted from wild type and DsecDF mutant cells at 3 h (two biological replicates) and 4 h (six biological replicates) of cultivation in LBG, on the basis that these time points mark the onset of morphological changes in the mutant compared to wild type.Significant differential expression was observed in more than 400 genes (.2-fold differential expression) during the transition phase (4 h).Table 3 lists 70 genes that exhibited confidently more than 5-fold differences in transcription levels at the 4 h time point.Quantitative RT-PCR confirmed the expression trend for 17 out of 18 selected genes (Fig. S4).In general, genes involved in metabolism and energy conversion processes, membrane transport, resistance and detoxification mechanisms, and motility, as well as several hypothetical genes, were most strongly affected.Furthermore, genes indicative of a cell wall stress response were stimulated in the DsecDF mutant (Table 3), including a phage shock response (pspA-like BC1436) gene, and an operon encoding a putative sigma W-type extracytoplasmatic function (ECF) sigma factor (BC5361-BC5363).Transcription of the genes entA (BC5239) and entC (BC0813) coding for putative cell-wall binding proteins were also upregulated.The entB gene (BC2952) showed a lower transcription level as well as a reduced amount of the EntB protein in the extracellular medium of the DsecDF mutant.
At the 4 h time point the DsecDF mutant showed a highly activated sigma B stress response regulon compared to the wild type (Table 3, Table S4).To confirm this, expression of sigB was followed over time by real-time quantitative PCR.While there were no significant changes (p,0.5) between the DsecDF mutant and the wild type strain at early and mid-exponential growth phase, sigB was 3-to 17-fold induced in the mutant compared to the wild type at late-exponential and transition phase (p,0.01;data not shown).In total 14 out of 26 previously described heatshock activated, sigma B-dependent genes [35], were more than two-fold upregulated in the DsecDF mutant (Table S4).The most strongly induced genes in the DsecDF mutant were also among the highest heat-shock induced genes (e.g.those encoding KatE and Protease I).In order to test if these transcriptional changes translated into a cellular phenotype, the catalase activity of cultures grown for 6 h was measured.In support of the activation of the SigB operon, the DsecDF mutant exhibited approximately 20% increased catalase activity (data not shown), however only when grown in the presence of glucose.Secretome analysis of the B. cereus DsecDF mutant had revealed strongly reduced levels of virulence factors in the supernatant, which was confirmed by Western blot analyses of cell-accumulated and extracellular Hbl and Nhe toxin component levels, thus indicating SecDF-mediated export.Transcriptional levels were also altered for several (plcB, smase, colC, BC2552, nprB), but not all (cytK, nhe, hbl) PlcR-regulated virulence determinants (Table S4, S5).PlcR plays a key role in pathogenicity as it acts as a transcriptional regulator of many extracellular virulence factors.plcR transcription is autoregulated and the activity of the protein depends on the signaling peptide PapR [36,37].However, PlcR was not differentially expressed over the course of 4 h growth (data not shown).It is nevertheless noteworthy, that the oligopeptide permease system, BC1179-BC1183, which is responsible for re-import of the PapR pheromone after extracellular cleavage [38], was transcriptionally downregulated in the DsecDF mutant (Table S5).
Interestingly, almost all motility-associated genes (BC1625-BC1671) were consistently downregulated two-fold or more in the mutant at the 4 h time point.Thus, the observed reduced flagellation and motility of the mutant was possibly due to reduced transcription of motility-associated genes encoding flagellar components and chemotaxis proteins.
The genome of B. cereus ATCC 14579 also contains a cryptic, linear plasmid pBClin15, encoding what appears to be a dormant prophage [43].Most of the pBClin15 genes were found to be downregulated in the DsecDF mutant (Fig. S5).This was not a result of loss of the pBClin15 plasmid, since (i) the presence of ORF 1-3 was detected via PCR using genomic DNA isolated from the bacterial culture used for the microarray analysis, and (ii) mRNA transcripts of BC_p0006 and BC_p0007 were detected by real-time qPCR from an independent culture.

Discussion
In the present study, deletion of secDF in B. cereus ATCC 14579 results in a pleiotropic phenotype which includes premature growth arrest and smaller colony size, aberrant cell morphology, reduced motility and reduced total protein in the bacterial secretome, consistent with previous reports on secDF mutants in other bacterial species [6,9,11,44,45].In addition, our experiments demonstrated more pronounced pleiotropic effects in the presence of glucose.
Nhe, Hbl and Cytotoxin K are well-studied toxins from B. cereus, causing the diarrheal syndrome after ingestion of contaminated food [2][3][4].The nhe and hbl operons in the wild type and mutant strains were not found to be differentially transcribed, while Western blotting experiments using monoclonal antibodies showed accumulation of Hbl and Nhe toxin components in the DsecDF mutant cells (Fig. 4, Fig. S3).A Sec-translocase -mediated Table 1.Cont. 2 Normalized Spectral Abundance Factor, mean average of three biological replicates; the NSAF normalizes across samples and takes protein sizes into account; values range between 0 and 1, increasing values indicate higher abundance [27], stdev standard deviation of the means of three biological replicates; probability ranges associated with Students t-test (Scaffold 4.0.5).

3
ND not detected (NSAF 0 in at least two biological replicates and ,0.005).*due to high sequence similarity all peptide hits for ''flagellin'' (Q81FD3, Q81FD4, Q81FD5) were combined.doi:10.1371/journal.pone.0103326.t001 export of these toxins has been advocated by Fagerlund et al. [5].However, it has also been indicated that the Hbl enterotoxin as well as the PC-PLC may be secreted via the flagellar apparatus [46,47], similar to what is known in C. jejuni [48] and C. difficile Knowing that the DsecDF mutant has such a strong impact on secretion of known virulence factors and that the respective S. aureus and L. monocytogenes SecDF null mutants were affected in virulence [12,50], we sought to evaluate the role of B. cereus SecDF in its capacity to kill the insect larvae Galleria mellonella, which is currently used for infection studies of B. cereus or B. thuringiensis strains [34,[51][52][53].Virulence tests were performed by two routes of infection and the strongest effect was recorded following oral infection with about 17-fold more DsecDF bacteria needed to kill 50% of the larvae at 24 hs compared to the wild type (Table 2).In addition, the mutant strain was also 4.5-fold less virulent 24 h post infection when the bacteria were injected into the hemocoel.This indicates that the DsecDF mutant is definitely affected in virulence but it is difficult to appoint the effect to a particular gene set because of the pleiotrophic effect of the mutation.Notably, the differences in virulence decreased after 74 h in both infection model experiments.This indicates that the reduced virulence of the DsecDF mutant might only be of transient nature, a notion supported by Western Blot experiments showing Nhe and Hbl components adapting comparable extracellular levels in both strains over time.Meanwhile the results are in line with former work on the non-motile mutant B. thuringiensis 407 cry 2 DflhA, where a defective flagellar machinery assembly led to a decrease in virulence [51].This was found to be partly due to a reduction of virulence gene expression, rather than direct involvement of the flagellar apparatus in virulence factor secretion [5,51].Since flagellar gene expression is reduced in the DsecDF mutant, the extent to which Hbl is transported via the Sectranslocase and the flagellar mechanism, respectively remains to be determined.
Out of the 96 proteins that could be identified in the B. cereus ATCC 14579 and isogenic DsecDF mutant secretomes (Table S3), the majority (57%) were predicted to be of cytoplasmic origin.Other studies also frequently report a high percentage of nonsecretory proteins in the medium [31,54,55], and cell lysis has been determined to be of only minor contribution [56][57][58].In LBG medium the DsecDF mutant did not exhibit increased autolysis compared to the wild type (data not shown).Cytoplasmic proteins like enolase and pyruvate dehydrogenase were detected in the growth medium of B. cereus (Table S3 and [29]), and these and other intracellular proteins have been reported to be secreted in B. subtilis during stationary phase by a non-classical translocation mechanism where protein domain structure appears to contribute [59].Although we did not find any indication of a stronger autolysis in the DsecDF mutant compared to the wild type, an increased amount of small sized ribosomal proteins was identified in the growth medium of the mutant (Table 1).No difference was seen at the transcriptional level of these genes between the wild type and the mutant.During co-translational insertion of proteins into the cell membrane the translocation channel protein SecY is bound to the ribosomal machinery [60,61] and in fact, it has been shown recently that this interaction opens the internal plug of SecY [62].Based on current knowledge, we cannot rule out the possibility that loss of SecDF could potentially result in a less specific translocation mechanism through a leaky SecYEG complex, feasibly affecting translocation of small sized proteins.
The DsecDF mutant presented an aberrant cell morphology combined with an earlier growth arrest during cultivation (Fig. 1), phenotypes potentially caused by atypical activity of peptidoglycan remodeling enzymes.Murein hydrolases function during cell wall growth, peptidoglycan turnover, cell separation, and autolysis [63].Two uncharacterized putative murein hydrolases (BC0991 and BC1991) were absent in the DsecDF mutant growth medium (Table 1).Both contain a transglutaminase domain, known to facilitate intra-and interprotein crosslinks and to potentially play an important role in cell wall maturation [64].In addition the putative cell wall binding proteins EntA (BC5239), EntB (BC2952), and EntC (BC0813), identified in the secretome of B. cereus [28], were affected at the transcriptional level (Table S5) and, in the case of EntB, also in the extracellular proteome in the  secDF deletion mutant (Table S3).Secretome analyses for EntA and EntC were, however, not conclusive.EntA, EntB and EntC all contain two copies of the cell wall-binding SH3 domain, and are members of the resuscitation-promoting factor/stationary-phase survival (Rpf/Sps)-family identified in actinobacteria and firmicutes [65].The B. subtilis muralytic enzyme YocH, which is a homolog of EntA, EntB and EntC, was induced by cell wallturnover peptidoglycan fragments of growing cells and a null mutant displayed reduced survival after post-exponential phase [66].Crucial residues for enzyme activity in YocH [66] are conserved in the three putative cell wall-binding proteins EntA, EntB and EntC.Clearly further analysis is required to understand the regulation and involvement of these and other muralytic enzymes in the phenotypic changes of the DsecDF mutant (Fig 1).AFM images clearly showed a reduction in cellular flagellation in the DsecDF mutant (Fig. 2C), probably as a result of transcriptional deactivation of genes coding for flagella components (Table 3).While intramembranous constituents of the flagellar body are generally believed to be inserted in a Sectranslocase dependent manner, the outer components are secreted via a flagellum-specific type III secretion system [67][68][69].It is known from E. coli and S. enterica that the expression of flagellar genes is dependent on the state of assembly, in a step-wise manner (see reviews [70,71]).Assuming a similar, energy-saving feedback loop in Bacillus, it is possible that the transcriptional downregulation of flagellar genes results from incomplete insertion and assembly of intramembrane flagellum body proteins.Thus, one could hypothesize that SecDF plays a role in early flagellum construction in B. cereus grown in the presence of glucose (Fig. 2).
A global transcriptional profiling experiment revealed profound transcriptional changes in the DsecDF mutant, a phenomenon seen previously for selected genes in a S. aureus secDF mutant [11].Among the genes most highly upregulated by secDF deletion were a range of genes thought to respond to disturbances in cell envelope structures: the phage shock response system, the sigma B regulon, an extracytoplasmatic function (ECF) sigma factor and the putative murein hydrolase BC1991 (Table 3).The PspA-like gene (BC1436) is similar to liaH of B. subtilis.The Lia operon (LiaIHFSR) is highly conserved in Firmicutes, and the system is a cell envelope stress response activated by peptide antibiotics [72,73].PspA is particularly well studied in E. coli and is induced by a wide range of cell envelope stress conditions and thought to maintain the energetic state of cells under stress (for review see [74]).In E. coli it has been shown that single gene deletions of Sectranslocase components such as SecA, SecD and SecF, lead to PspA overexpression [75], and that PspA supports the efficient translocation of Sec-and TAT-dependent proteins [76].In our study of the B. cereus DsecDF mutant, the strong induction of the pspA-like gene may be a result of sensing the secretion defect as well as of an internal accumulation of proteins.In addition, the sigma B regulon known to provide a non-specific stress response to a range of different stress signals affecting cell envelope integrity [77][78][79] is moderately upregulated.Among the ten ECF-type sigma factors identified in B. cereus [80], recognizing environmental signals [81], the so far uncharacterized BC5363 exhibits   similarity to the B. subtilis SigW sigma factor (34% identity at the protein level).Interestingly, sigW is induced by cell envelope stress factors (for review see [82]).While the addition of glucose to the growth medium resulted in general in more pronounced phenotypes (Fig. 1, Fig. 4, Fig. 5), it is noteworthy that only sugars consisting of at least one glucose component profoundly inhibited motility of the DsecDF mutant (Fig. 2A).Although the rationale and mechanism behind the effects of glucose on the phenotype of the DsecDF mutant remain to be elucidated, this study confirms previous reports showing that glucose exerts more functions than only being an important nutrient.Recent research indicates for instance a direct involvement of glucose in expression of the toxin hemolysin II in B. cereus by activation of HlyIIR by glucose 6P which resulted in repression of hlyII gene expression [83].
The present study shows that some toxins and other virulence factors produced by the pathogenic Gram-positive, spore-forming bacterium B. cereus are dependent on SecDF for proper translocation across the cell membrane, confirming a role for SecDF in protein secretion in general and efflux of some toxins, directly or indirectly, in particular.It could be assumed that the ubiquitous SecDF protein fills similar functions also in other bacteria, as it has been reported for S. aureus and L. monocytogenes [12,50].Finally, although we cannot explain the phenomenon at this moment, this study shows clearly an exacerbating effect of glucose on the phenotype of the DsecDF mutant.

Growth conditions
Unless otherwise stated, B. cereus and E. coli strains were streaked on LB agar plates and incubated at 30uC and 37uC, respectively.Liquid cultures were inoculated from a single colony, incubated overnight and then diluted 1:100 in LB medium.These starter cultures were grown at 30uC or 37uC, respectively, at 200 rpm.After reaching an OD 600nm of approximately 0.5, experimental cultures were inoculated from the starter culture to an initial OD 600nm of 0.02, and grown as above.If applicable, 1% glucose was added to LB (LBG).When relevant, erythromycin 5 mg/ml (with pHT304 plasmid) or ampicillin 100 mg/ml (with pTTQ18 plasmid) was added to the culture.For assessment of glucose fermentation the strains were streaked on Bacillus cereus agar (Oxoid) supplemented with 1% glucose.Acidic by-products of glucose fermentation were monitored by color change of the pH indicator bromothymol blue.

Construction of the DsecDF-mutant
The markerless DsecDF mutant of the type strain B. cereus ATCC 14579 was constructed by the method of Janes & Stibitz [84].A deletion construct consisting of overlapping flanking regions of the target gene is cloned into a temperature-sensitive shuttle vector carrying the homing endonuclease restriction site I-SceI.Under replication non-permissive temperatures and selection pressure the vector integrates either up-or downstream of the target gene.To enforce a double-strand break of the chromosomal DNA, a second plasmid encoding I-SceI is introduced into the organism.Repair of the break by cross-over leads to either wild type or knock-out genotypes.Mutants are then selected by PCR and the vector sporadically lost during non-selection.Oligonucleotides used for making the gene deletion construct, substituted the BC4405 ORF in frame with ATGGTCGACTAA and thus introduced a SalI restriction site (supplemental information S1).After cloning of the gene deletion construct with about 500 bp flanking regions into the suicide shuttle vector pBKJ236 and electroporation into B. cereus ATCC 14579, the protocol was followed as previously described [84].Successful gene deletion was confirmed by PCR using genomic DNA as template and oligonucleotides binding outside of the deleted region, and by DNA sequencing.The presence of the plasmid pBClin15 was confirmed by PCR as reported previously [85].

Assessment of phospholipase C activity
The activity of secreted Phospholipase C (PC-PLC) was measured for cells growing on agar and in liquid cultures.For the first test, bacteria were grown in LB medium for 16 h at 30uC and 220 rpm, washed in 0.9% NaCl and resuspended to an OD 600nm of 8.5.Five ml of the bacterial suspension was spotted onto LB and LBG agar plates supplemented with 5% egg yolk suspension (Oxoid).The phospholipase C activity was analyzed by visual inspection after 7 hours incubation at 30uC.PC-PLC activity of filter-sterilized supernatant sampled at different time points, from cultures grown in LB and LBG, was measured by spotting 5 ml on 1% egg yolk agar plates, and incubating them at 30uC for 24 h.In addition, 100 ml of these supernatants were incubated with 900 ml 2% egg yolk saline suspension at room temperature for 75 min after which the OD 600nm was measured.Variations in growth between the wild type and the mutant strains were accounted for when necessary by diluting the wild type supernatant with fresh LB after filter-sterilization.

Light microscopy and atomic force microscopy (AFM)
Micrographs were made using 3 ml sample of a fresh culture with 400-fold magnification.Pictures were obtained with a Nikon Labophot-2 microscope coupled to a Leica DFC320 camera and assessed with the LAS v3.6 program.For AFM, B. cereus ATCC 14579 wild type and DsecDF mutant strains were grown in LBG as detailed under ''growth conditions'' and one ml samples were collected after 4 h growth.Following 3 min centrifugation at 2400xg the cells were washed and resuspended in 1 ml 0.9% saline.Ten ml of the suspension was diluted to a final volume of 50 ml in 10 mM magnesium/Tris buffer, pH 7.5, ten ml of which was applied to a freshly cleaved muscovite mica (Agar Scientific, Norway) mounted on a glass slide, and incubated for 10 min at room temperature.After ten washing steps with 100 ml sterile filtered MQ water, the samples were dried under a gentle N 2 stream.AFM images were recorded in intermittent contact mode in air using a NanoWizard I atomic force microscope (JPK, Berlin, Germany).To quantify the number of flagella, a total of 103 cells for the DsecDF mutant and 26 cells of the wild type were analyzed, from two independent cultures.

Motility assays
To assess motility, 0.3% and 0.7% LB agar plates were used.Five ml of overnight cultures (OD 600nm between 7 and 10) of the wild type and mutant strains grown in 5 ml LB at 30uC at 220 rpm were spotted on the agar surface of the same plate, with two technical replicates per biological sample.The diameter of the culture was measured after 7-9 h incubation, the start diameter of the drop was subtracted and the ratio of the recorded motility for wild type and mutant was calculated.Every experiment was done at least four times, and the motility of the wild type strain in each condition was set to 100% (unpaired, two-tailed Student's t-Test for wild type vs. mutant, P,0.05).Statistical significance of differences between the mutant's motility compared to the wild type in pure LB and LB + additives was evaluated using the MS Office Excel unpaired t-test function with a two-tailed distribution.

Expression of SecDF
For expression of SecDF in B. cereus ATCC 14579, the native gene was cloned into the low-copy number E. coli/Bacillus plasmid shuttle vector pHT304-Pxyl [86].pHT304-Pxyl contains the xylR and xylA promoters from B. subtilis, allowing xyloseinducible expression of SecDF fused with a C-terminal 6x histidine tag.For heterologous overexpression of SecDF in E. coli, the secDF gene from B. cereus ATCC 14579 was cloned into a modified version of the high copy number, IPTG inducible vector pTTQ18 [87].Expression of secDF from this plasmid resulted in a recombinant protein carrying a C-terminal 6x histidine tag.The plasmid was introduced into E. coli BW25112 DacrB.This strain lacks the RND-type transporter AcrB, which has been shown to be the major xenobiotic efflux transporter in E. coli (for recent reviews see [88,89]).Correct cloning of the gene was in both cases confirmed by sequencing, and protein expression in both host organisms was measured using the histidine tags for detection by specific antibodies.Induction of protein expression by 20 mM xylose and 0.05 mM IPTG, respectively, resulted in a protein band of approximately 82 kDa on a Western blot, in both cases (data not shown).

Determination of minimum inhibitory concentrations (MICs)
To identify the susceptibility of B. cereus and E. coli DacrB to a range of xenobiotics, bacterial suspensions were incubated in LB and LBG, respectively, with 2-fold serial dilutions of the tested compounds.Pre-cultures grown in LB were diluted to an OD 600nm value of 0.02 and aliquoted into 96-well plates (final volume 150 ml).The plates were incubated in a humidified chamber at 30uC, 200 rpm for 22 h.The lowest concentration of xenobiotics that resulted in no visual growth was considered as the MIC.Experiments were done in technical duplicates and with at least two biological replicates.If protein overexpression strains were tested, xylose (20 mM) or IPTG (0.05 mM) was added to the medium for pHT304-Pxyl and pTTQ18 vector constructs, respectively.Alternatively, the susceptibility of B. cereus strains was examined by disk diffusion on LB or LBG agar plates.Midlogarithmic precultures were diluted to an OD 600nm of 0.05 in 0.9% NaCl, and 1 ml of this cell suspension was spread out on agar plates and air-dried.Thereafter, 6 mm paper disks applied on the surface were impregnated with 10 ml of each tested compound.Inhibition zones were examined after 16 h incubation at 30uC, for the following compounds: ethanol 100%, spectinomycin 100 mg/ ml, phosphomycin 25 mg/ml, ciprofloxacin 10 mg/ml, norfloxacin 10 mg/ml, chloramphenicol 25 mg/ml, tetracycline 10 mg/ ml, oxytetracycline 0.8 mg/ml, gentamicin 50 mg/ml, ampicillin 50 mg/ml, oxacillin-5 (BD), SDS 20%, DOC 80 mg/ml, chlorhexidin 1.6 mg/ml, ethidium bromide 5 mg/ml, CCCP 7.5 mM, sodium lactate 50%, polymyxin B 25 mg/ml, sodium benzoate 0.5 g/ml, erythromycin 100 mg/ml, kanamycin 10 mg/ ml, plant extracts: tea tree (Melaleuca alternifolia); steam distillates of peppermint leaves (Mentha piperita) and calabash (Melaleuca leucadendron var.cajaputi) (Primavera Life).

Microarray analysis
Cells were grown in LBG in 50 ml cultures in 500 ml nonbaffled Erlenmeyer flasks at 30uC, 220 rpm for 3 h (two biological replicates) and 4 h (six biological replicates), respectively.Five ml culture was then mixed with equal amounts of ice-cold methanol, followed by harvesting by a short centrifugation.Cells were lysed by beadbeating and the RNA was isolated using the RNA Mini Kit (Qiagen), including the on-column DNase treatment step.cDNA conversion and labelling, microarray hybridization and data analysis using Bayesian linear modelling (Limma-package [90] was basically performed as described previously by Gohar et al. [91] and detailed procedures and raw data were deposited according to MIAME guidelines in the Arrayexpress database (accession number E-MTAB-1759).

Analysis of secreted proteins
For the analysis of secreted proteins, mid-logarithmic cultures of the B. cereus ATCC 14579 wild type and the isogenic DsecDF mutant strains grown in LB were transferred into fresh LB or LBG medium.Following 3 h, 4 h and 6 h aerated growth at 30uC and 220 rpm, PBS-adjusted volumes (by dilution according to 1 ml culture with lowest OD 600nm ) of each culture were harvested by centrifugation.Sterile-filtered (0.2 mm) culture supernatant was mixed 1:4 with ice-cold methanol:acetone (1:1) and proteins were precipitated overnight at 220uC.Proteins were harvested by centrifugation at 12,000xg for 30 min at 4uC.For gel electrophoresis, 8 ml of duplicate, independent and normalized supernatants of cultures grown in LBG were concentrated 40-fold by methanol:acetone precipitation, resuspended in 250 ml TES (20 mM Tris pH 7.5, 0.8% NaCl, 1 mM EDTA), and 12 ml of each sample was analyzed on a 4-20% SDS-polyacrylamide gel (Pierce) by silver staining (Sigma-Aldrich).For label-free mass spectrometry analyses, triplicate independent cultures (from individual colonies) were grown for 4 h in LB added 1% glucose, as described above.Using acid-cleaned glassware, the PBSadjusted culture supernatants (according to the culture with lowest OD 600nm ; final volume of 2.5 ml) were subjected to methanol:acetone precipitation at 220uC overnight.After centrifugation at 12.000xg for 30 min at 4uC, proteins were resuspended in 50 mM ammoniumbicarbonate/1 M urea.After protein concentration determination using the Bradford Assay with BSA as a standard, 20 mg of each sample was used for analyses.
Sample Preparation.Proteins were reduced with 10 mM DTT (1 h at 70uC, pH 9), alkylated for 1 h using 25 mM iodoacetamide, and digested with trypsin (1 mg) at 37uC for 16 h.Digested protein samples were analysed using a TripleTOF 5600 mass spectrometer (AB SCIEX Foster City, CA, USA) coupled to an Eksigent NanoLC-Ultra 2Dplus system (Eksigent Technologies, Dublin, CA, USA).Peptides were separated as described previously [94], and the LC eluent subjected to positive ion nanoflow analysis using an ion spray voltage, heater interface temperature, curtain gas flow and nebulizing gas flow of 2.5 kV, 150uC, 20uC and 16uC, respectively.Information dependent acquisition-experiments utilized a survey scan (350-1500 amu) with an accumulation time of 100 ms, followed by 15 MS/MS product ion scans (350-1600 amu) with an accumulation time of 100 ms each.
Protein Identification.Proteins were identified using the Paragon search algorithm [95,96] in ProteinPilot Version 4.0.8085(AB SCIEX Foster City, CA, USA).Searches were carried out against the reference proteome of B. cereus ATCC 14579, extracted from the Universal Protein Resource (UniProt) (4) using the thorough search mode and included biological modifications, trypsin-cleaved peptides and iodoacetamide-modification of cysteine residues.False discovery rates were determined in Protein-Pilot using a detected protein threshold of 0.05 and the decoy database searching strategy, and only proteins at 1% global FDR and distinct peptides at 5% local FDR were reported.For further data analysis of all three biological replicates, Scaffold (version Scaffold_4.0.5, Proteome Software Inc., Portland, OR) was used to validate MS/MS based peptide and protein identifications.Equal amount of total protein was used for tryptic digestion and comparative analyses were conducted after normalization of the data sets accordingly to the Normalized Spectral Abundance Factor (NSAF) approach using total spectral counts [27].A Student's T-test comparing total spectral counts was performed to determine statistical significances of protein abundances in wild type and mutant strain samples.

Toxin detection
Cell lysates were prepared by harvesting 2 ml of B. cereus wild type and DsecDF mutant cultures by centrifugation at 4500xg for 5 min.The pellets were washed once in cold PBS and stored over night at 220uC.Cell pellets were then resuspended in TES containing 2 mg/ml lysozyme and the volume was adjusted according to the original culture OD.The bacterial suspensions were incubated at 37uC for 1 h.After partial cell wall degradation, cell lysis was achieved by six rounds of freezing and thawing in liquid nitrogen and a 37uC water bath respectively.Cell debris was removed by centrifugation and the supernatant was stored on ice for no more than 4 h.Twenty ml of normalized, sterile-filtered supernatants and 2 ml of cell lysates were separated on 10% SDS polyacrylamide gels and blotted onto a nitrocellulose membrane.Toxin components were detected using 1:20 dilutions of the following monoclonal antibodies: 1A8 and 1E11, against NheA and NheB, respectively [33]; and 1E9 and 8B12, specific for the L1 and L2-subunits of Hbl [32].1:10,000 dilution of HRPconjugated anti-mouse antibody (Sigma) was used for chemiluminescent signal development.

Analyses of proteolytic activity
Experiments analyzing milk and gelatin proteolytic activities in the secretomes of the wild type and DsecDF mutant strains did not reveal significant differences (data not shown).Skim milk agar plates were prepared by dissolving skim milk powder and agar separately in Milli-Q water, to a concentration of 75 mg/ml and 15 mg/ml, respectively.Following autoclaving for 15 min at 110uC and cooling to 50uC, the skim milk and agar solutions were mixed (1:1).Overnight cultures grown in LB broth were normalized to an OD 600nm of 1 with 0.9% NaCl and 50 ml was added into punched holes (5 mm) in skim milk plates and incubated at 37uC, 30uC and 20uC, respectively.Proteolytic activity was visible as change in opacity of the milk around the bacterial spots.Total gelatinase activity was carried out as described in Millipores technical publication on gelatin zymography (http:// www.millipore.com/userguides/tech1/mcproto009).Ten ml of sterile-filtered and normalized culture of B. cereus strains grown for 4 h in LBG at 220 rpm were loaded on an 8% acrylamide gel co-polymerized with 0.1% gelatin, using non-reductive SDS sample buffer.Gelatinase activity appeared as clear bands in the turbid gel background.

Autolysis
In order to determine if the DsecDF mutant displayed a higher autolysis rate than the wild type strain, two different tests were conducted: (i) cell lysis activity of B. cereus cell lysates was investigated by performing zymograms using whole B. cereus cells as substrate, according to the method of Raddadi et.al. [97]; (ii) spontaneous autolysis was determined as described by Quiblier et al. [11].Cells grown in LB or LBG, were harvested 4 h after inoculation, washed in 0.9% NaCl, and resuspended in 0.01 M Na-phosphate buffer, pH 7.4 to a final OD 600nm of 1.The resulting bacterial suspensions were incubated at 30uC, 200 rpm for 90 min and the decrease in optical density (600 nm) was measured at regular intervals.Neither of the experiments supported a higher autolysis rate in the secDF deletion strain compared to wild type when grown in LBG (data not shown).

Catalase test
The catalase test, based on a stable yellow complex-formation of hydrogen peroxide with molybdate, was carried out basically as described by Go ´th 1991 [98].Briefly, B. cereus strains were grown in LB or LBG medium at 30uC and 220 rpm.After 3 h and 4 h growth, OD 600nm was measured in duplicate, and a volume corresponding to an optical density of 14 per ml was pelleted and resuspended in 100 ml of 6 mM phosphate buffer, pH 7.4.Samples were mixed with 500 ml preheated substrate solution (65 mM H 2 O 2 in 6 mM phosphate buffer) and incubated at 37uC for 120 sec.The reaction was stopped by adding 500 ml of 32.4 mM ammonium molybdate in 6 mM phosphate buffer.After pelleting the cells, the color change was measured spectrophotometrically in a microplate reader at 405 nm.Each sample was analyzed in triplicate.Absorbance values were subtracted by values of the blank non-reactive wells containing 100 ml of 6 mM phosphate buffer and no bacterial cells.As a loading control, pelleted cells were lysed as described above, and equal volumes of cell lysates were applied on a 12% SDS-polyacrylamide gel.Proteins were stained with Bradford reagent.

Insect infection experiments
The virulence-related properties of DsecDF were assessed by comparing the killing effect of the B. cereus wild type and the DsecDF mutant strains by both oral infection and direct injection into the hemocoel of 5 th instar Galleria mellonella larvae [34,99].G. mellonella eggs were hatched at 25uC and the larvae reared on beeswax and pollen.In each experiment, groups of 20 to 30 G. mellonella larvae, weighing about 200 mg, were used.For oral infection, the larvae were force-fed with 10 ml of a mixture containing various doses (1.5x10 5 to 2.5x10 7 ) of vegetative bacteria (exponential growth OD<1in LB medium) and 3 mg of activated Cry1C toxin, prepared as previously described [100].For injection experiments, the larvae were also infected with vegetative bacteria at various doses, from <2,000 to <100,000 cfu (colony forming units).Experiments were repeated at least twice.Infected larvae were kept at 37uC and mortality was recorded at 24 h and 72 h post infection.The larvae in the control group were fed PBS buffer.The 50% lethal doses (LD 50s ) values, as estimated using the Probit analysis StatPlus program, corresponds to the cfu killing 50% of the treated larvae.

Figure 1 .
Figure 1.Growth characteristics of the DsecDF mutant in LB with and without glucose.Growth of B. cereus ATCC 14579 (WT) and DsecDF mutant in A) LB (no glucose) and B) LBG (1% glucose) in shaking cultures at 30uC.The graphs show average OD 600nm values with standard deviations of two independent cultures for each strain.C) light micrographs of cultures after 24 h growth.D) growth of WT (left) and the DsecDF mutant (right) at 30uC for 16 h on LB and LBG agar.All pictures represent results of at least two independent experiments.doi:10.1371/journal.pone.0103326.g001

Figure 2 .
Figure 2. Diminished motility of the DsecDF mutant.A) Motility of the DsecDF mutant compared to the wild type strain on 0.3% LB only -, or LB agar plates supplemented with: gluc, 0.4% glucose; malt, 1% maltose; sucr, 1% sucrose; xyl, 1% xylose; sorb, 1% sorbitol; galac, 1% galactose; tw, 0.02% Tween80, tw+gluc, 0.02% Tween80 and 0.4% glucose.The graph shows averages of four to ten independent experiments, error bars represent standard errors and an unpaired Students t-test with two-tailed distribution was performed between wild type and DsecDF mutant (all conditions P, 0.05).A nonparametric ANOVA with Dunn's multiple comparison post hoc test was performed for ''LB only'' and each of the conditions using additives (*P,0.01;**P,0.001).No movement of the DsecDF mutant was recorded in LB+sucrose in four experiments.B) Comparison of motility on 0.7% LBG after 7 h incubation at 30uC; top: wild type; bottom: DsecDF mutant.C) AFM amplitude images representative of two independent experiments of cells grown in LBG for 4 h show the grade of flagellation and secretion of putative membrane vesicles.Bars: 1 mm in whole cell images; 0.2 mm in the wild type detail image indicating putative vesicles (arrows).doi:10.1371/journal.pone.0103326.g002

Figure 3 .
Figure 3. Decreased protein secretion in the DsecDF mutant.The DsecDF mutant secretes less protein than the B. cereus ATCC 14579 wild type strain.Equal volumes of normalized and 40-fold concentrated supernatants collected after 3 h (exponential phase), 4 h (transition phase) and 6 h (stationary phase) of growth were applied on 4-20% gradient SDS-PAGE gels and silver stained.The gel represents results of two independent experiments.doi:10.1371/journal.pone.0103326.g003 prediction of PSORTb algorithm (version 3.0.2;[26]):EC extracellular, C cytoplasmic, U unknown, M membrane; references for experimentally defined locations are given for proteins with predicted unknown localization.
[49].Since the transcription of the flagellar machinery is downregulated in the DsecDF mutant we cannot state explicitly whether the translocation defect of Hbl components is due to secondary effects on the flagellar system or to direct inhibition of the Sec-translocase pathway.For virulence factors other than Hbl and Nhe, such as cytotoxin K, PLC, SMase and collagenase C, a weak to moderate, yet statistically significant, transcriptional downregulation was observed.PlcR is a key transcriptional regulator involved in integration of a range of environmental signals such as cell-density and nutrient deprivation, and controls the expression of a range of extracellular B. cereus virulence factors, including Nhe, Hbl, CytK, PC-PLC and SMase.Interestingly, CytK, PLC, SMase, BC0991 and ColC were among the highest differentially detected proteins in the culture supernatants.

Figure 4 .
Figure 4. Comparison of NheA, NheB, Hbl_L1 and Hbl_L2 secretion by western-blot analysis.Western-blot assay of secreted (left) and cell associated (right) toxin components NheA, NheB, Hbl_L1 and Hbl_L2.Samples of the growth medium were taken from the wild type (WT) and the DsecDF mutant (D) from 3 h (exponential phase), 4 h (transition phase) and 6 h (stationary phase) cultures with and without added glucose.The blots are representative of at least two biological replicates.To visualize size differences between pre-and mature proteins, a supernatant wild type sample (SN6H) has also been applied to the blot showing cell associated protein.doi:10.1371/journal.pone.0103326.g004

Figure 5 .
Figure 5. Reduced PC-PLC activity in the DsecDF mutant.A PC-PLC activity assay indicates reduced enzyme activity in the DsecDF mutant compared to the wild type strain.A) Filter-sterilized supernatant of cultures grown in LB (no glucose) or LBG (1% glucose) were assayed in a 2% egg-yolk solution.The results are the mean values of two independent experiments, and error bars represent standard deviations.B) Five ml of filter-sterilized supernatant of cultures grown in LB or LBG were spotted on 1% egg yolk agar plates.t0 marks the transition point of growth into stationary phase, and tn is the number of hours before (-) or after t0.The pictures represent one of two independent experiments.doi:10.1371/journal.pone.0103326.g005

2 FC
fold change of transcriptional expression in B. cereus DsecDF compared to wild type.

3 P
-values were computed using false discovery rate correction of 0.05 by an Bayesisn linear model as integrated in the Limma-package[90]; data represent six independent cultures.1 purine operon under the control of PurA.doi:10.1371/journal.pone.0103326.t003

Table 2 .
Role of SecDF in virulence against Galleria mellonella insect larvae.
a Infections with mid log phase (OD 600nm = 1) vegetative bacteria cultured in LB medium.bFor oral infection the bacteria are mixed with 3 mg/10 ml of Cry1C toxin.The toxin alone results in <10% mortality.cLC 50 , correspond to the bacterial dose (cfu, colony forming units) per larvae, killing 50% of the treated larvae.dconfidence interval at 95% level.*(P-value#0.05).LDL (lower dose limit), UDL (upper dose limit).Mortality was estimated by Probit analysis (StatPlus).based on at least two independent experiments.Control experiments were run with buffer (PBS) and no

Table 3 .
Genes with at least a five-fold differential transcription level in the DsecDF mutant compared to the isogenic wild type strain B. cereus ATCC 14579.
1 data on the linear plasmid pBClin15 can be found in the supplementary file.