Overnight cultures of E. feacalis V583 5 were diluted in fresh prewarmed brain heart infusion (BHI) medium (Oxoid Ltd., Hampshire, England) to an OD₆₀₀~0.2 and incubated at 37°C without agitation. The bacteria were harvested on the transition between late exponential and stationary phase (OD₆₀₀~1.7) by centrifugation (3000 × g, 10 min, 4°C). The cell pellet from 100 ml of culture was washed three times with 10 ml PBS by centrifugation (3000 × g, 10 min, 4°C) and subsequently resuspended in PBS containing 40% sucrose. Three different shaving reactions were set-up, all containing 5 mM DTT and all with the same final concentration of cells: (1) addition of 20 μg trypsin (Promega, Mannheim, Germany), (2) addition of trypsin-agarose (100 units; Invitrogen, Karlsruhe, Germany), or (3) no addition of trypsin (untreated). The samples were incubated for 1 or 2 hours at 37°C with shaking at 300 rpm. After incubation the cells were pelleted by centrifugation (3000 × g, 10 min) and the supernatants were collected for further protein digestion with 1 μg freshly added trypsin over night (16-18 h) at 37°C with agitation at 400 rpm. Cell samples taken before (i.e. after resuspending in PBS) and after the different enzymatic treatments were used to test cell viability by plating appropriate dilutions on BHI agar plates and counting of colony forming units (CFU). The overnight trypsin digestion of the supernatants was stopped by adding formic acid to a final concentration of 0.1% (v/v). Prior to nanoLC-MS/MS analysis, peptides were concentrated and purified in two steps using C₁₈ Dynabeads (Invitrogen) in the first step and C18 StageTips 20 in the second step. For each treatment, samples from four biological replicates were analysed.

Reverse phase (C18) nano online liquid chromatographic MS/MS analyses of tryptic peptides were performed using a HPLC system consisting of two Agilent 1200 HPLC binary pumps (nano and capillary) with corresponding autosampler, column heater and integrated switching valve. This LC system was coupled via a nanoelectrospray ion source to a LTQ-Orbitrap mass spectrometer (Thermo Fisher Scientific, Bremen, Germany). For the analyses, the peptide solution was injected onto the 5 × 0.3-mm extraction column filled with Zorbax 300 SB-C18 of 5-μm particle size (Agilent, Waldbronn, Germany). Samples were washed with a mobile phase consisting of 97% 0.1% formic acid & 3% acetonitrile. The flow rate of 4 μl/min provided by the capillary pump. After 7 min, the switching valve of the integrated switching valve was activated, and the peptides were eluted in the back-flush mode from the extraction column onto a 150 × 0.075-mm C₁₈, 3-μm resin, column (GlycproSIL C18-80Å, Glycpromass, Stove, Germany). The mobile phase consisted of acetonitrile and MS grade water, both containing 0.1% formic acid. Chromatographic separation was achieved using a binary gradient from 5 to 55% of acetonitrile in 120 min. The flow rate of 0.2 μl min^-1 was provided by the nanoflow pump.

Mass spectra were acquired in the positive ion mode applying a data-dependent automatic switch between survey scan and tandem mass spectra (MS/MS) acquisition. Peptide samples were analyzed by collision induced dissociation (CID) in the LTQ ion trap by acquiring one Orbitrap survey scan in the mass range of m/z 380-2000 followed by CID of the six most intense ions in the ion trap. The target value in the LTQ-Orbitrap was 1,000,000 for survey scan at a resolution of 60,000 at m/z 400 using lock masses for recalibration to improve the mass accuracy of precursor ions. Fragmentation was performed with a target value of 5,000 ions. The ion selection threshold was 500 counts. Selected sequenced ions were dynamically excluded for 180 s.

Mass spectrometric data were first analyzed by generating msf files from raw MS and MS/MS spectra using the Proteome Discoverer 1.0 software (Thermo Fisher Scientific) and the database searches were then performed with an in house maintained E. faecalis V583 protein sequence database, using the SEQUEST search engine. The following criteria were applied; database decoy, true; Enzyme name, trypsin (full); Missed cleavage sites, 2; Precursor mass tolerance, 10 ppm; fragment mass tolerance: 0.6 Da; dynamic modifications: N-term acetyl (any N-terminus), oxidation (M), carboxymethyl (C), deamidated (N, Q). Proteins were considered as significant hits if the following conditions were met: XCorr higher than 2.0; false discovery rate less than 5%; identified by at least two different peptides; identified in at least two of the independent parallels by at least one peptide in each.

To visualize proteins or protein fragments that were resistant to trypsin, 20 μl of the supernatant from the over-night trypsination were applied to 10% NuPAGE Novex Bis-Tris gels (Invitrogen). Only samples after two-hour incubation were studied. The gels were stained using SilverSNAP Stain for Mass Spectrometry (Pierce, Rockford, IL) following the manufacturer's procedure. After the silver staining, the gel-lane was sliced into 12 pieces, and destained using the protocol included in SilverSNAP Stain for Mass Spectrometry kit. Each gel piece was then incubated with 0.1 μg trypsin in 25 μl 25 mM ammonium bicarbonate, over night at 37°C and 400 rpm. The trypsin reactions were stopped by adding 0.1% formic acid. The supernatants were transferred to new tubes, and the rest of the peptides were extracted from the gel pieces by incubating with 0.1% (v/v) trifluoroacetic acid in 60% (v/v) acetonitrile, at 37°C, 400 rpm for 10 min. The extracts from three gel-pieces were pooled together (giving four samples from each treatment). The peptides were dried in a speed-vac, and rehydrated in 30 μl 0.1% (v/v) TFA. The peptide samples were desalted using C18 stage tips 20 prior to nanoLC-MS/MS. Proteins were considered as significant hits if the following conditions were met: XCorr higher than 2.0; false discovery rate less than 5%; identified by at least two different peptides; identified in at least one of the three samples from each treatment.

Protein sequences used for in computo analysis of the localization of the identified proteins were extracted from the LocateP database 12 and analyzed using several bioinformatic tools. Putative N-terminal signal sequences and cleavage sites were predicted using the Signal P 3.0 server 21 and LipoP v 1.0 22 . The TMHMM Server v. 2.0 23 was used to predict proteins with multiple transmembrane helices or N-terminal transmembrane anchors. Proteins with features indicating non-classical secretion were predicted using the SecretomeP 2.0 Server 24 . Domain annotations were done using Pfam 25 . After these verifications, the predicted localizations for 62 of the 69 proteins discussed below correspond to those given in the LocateP database (updated March 10, 2010). For seven proteins, we reached a different conclusion than LocateP, as described in results and discussion.

Before carrying out the experiments, we performed extensive tests to find optimal conditions for the trypsin treatment. Most importantly, we checked the effect of incubation time (30 min to 24 hours) on cell viability. Incubation times of 2 hours or less did not lead to significant reductions in the CFU number (Additional file 1), whereas longer incubation times led to decreased viability (data not shown). Based on these observations, incubation times were set to one or two hours. Two hour incubations led to a higher number of identified proteins (Additional file 2). Generally, the longer incubations did not lead to an increase in the fraction of cytoplasmic proteins, confirming the absence of cell lysis during the enzymatic treatment (Additional file 2).

Intact bacterial cells were harvested and treated with either trypsin or trypsin beads (trypsin bound to agarose beads) for one or two hours. Because trypsin beads are less likely to penetrate the cell wall than free trypsin, they are more likely to only act on proteins that protrude from the cell wall. As a control, cells were incubated for one or two hours without adding trypsin. Direct analysis of released tryptic peptides led to the identification of 57 unique proteins (Table 1). Subsequent analysis of solubilized proteins and large protein fragments using SDS-PAGE followed by MS-based identification (Figure 1; see Materials and Methods) led to the identification of another 12 unique proteins (Table 2). The sequences of all identified proteins were analysed using a variety of bioinformatic tools (LocateP, SignalP, Pfam, LipoP, pSORT and TMHMM) to verify or (for EF2860, EF0071, EF0123, EF0164, EF0394, EF0417 & EF_B0004) to adjust the localization given in the LocateP database. The results are incorporated in Table 1 & 2 and are discussed in appropriate sections, below.

Analysis with Signal P 3.0 21 suggested the presence of a signal peptidase I (SPase I) cleavage site in ten of the identified proteins. Four of these proteins contained a putative C-terminal LPxTG motif, whereas one additional protein (EF2860) is likely to be cell-wall anchored because it contains a putative peptidoglycan binding domain (Pfam PF12229). It should be noted that the sequences deposited in the GenBank database for two of the four LPxTG-containing proteins (EF1033 and EF2713) lack a predicted N-terminal signal sequence. A closer look at the upstream sequences showed that the start codons probably are located 72 and 96 nucleotides upstream of the start codon suggested in the GenBank entries, for EF1033 and EF2713, respectively (Additional file 3). After this N-terminal "extension" SignalP and LipoP detected a putative SPase I cleavage site in both sequences.

Table 3 gives an overview over the predicted localizations of the 69 identified unique proteins and shows that the methods yielded a strong bias towards identifying proteins that are predicted to be covalently anchored to the cell wall or to carry lipid anchors. Several proteins were identified in more than one experiment and an overview is provided in Figure 2. Treatment with free trypsin yielded 58 proteins and treatment with trypsin beads yielded 29 proteins. Analysis of samples from untreated cells yielded 16 proteins. More detailed information concerning the numbers of proteins identified after the various treatments is provided in additional file 2. Details of the proteomic analysis are provided in Additional file 4 and Additional file 5 containing Tables S5 and S6, respectively.

The number of proteins only identified after a "shaving" treatment amounted to 53. Nine of these were only found after treatment with trypsin beads, 38 were only found after treatment with free trypsin and six were found after both treatments (Figure 2). The 15 proteins identified with only trypsin-beads or with both trypsin and trypsin beads are likely to be exposed on the surface of the cell wall, whereas the 38 unique proteins found in the free trypsin samples fraction are probably localized deeper in the cell wall.

Of the 16 proteins identified from untreated cells, 12 were identified in all three treatments (Figure 2). While three of these are predicted to be secreted and one (EF0201) is probably cytoplasmic, the others are predicted to be attached to the bacterial surface through an anchor (five lipo-anchors and one LPxTG anchor) or cell wall binding domain (one, EF2860) or even as integral membrane protein (one, EF1264). The trypsin-independent release of these proteins may be a result of natural shedding, a phenomenon that indeed has been observed previously, in particular for lipoproteins 15 26 . According to a TMHMM topology prediction EF1264, annotated as membrane protein, contains five N-terminal transmembrane helices and a huge extracellular domain with putative sulfatase activity of 523 residues (starting at amino acid 179). EF1264 was identified by many significant peptide hits spread over all treatments. Figure 3 shows that all identified peptides stem from the extracellular domain and that there is a 115 amino acid gap between the predicted integral membrane domain and the first identified tryptic fragment. Perhaps the extracellular domain is shedded after natural cleavage of EF1264. It is conceivable that such (apparently rather abundant) shedding is a physiologically relevant phenomenon since the sulfatase may remove sulphate from mucin, which would allow more easy degradation of mucin by glycosidases 27 and perhaps also could facilitate bacterial adhesion.

According to the LocateP database 34 of the 69 identified proteins lack an N-terminal signal sequence and are therefore predicted to be cytoplasmic proteins. All these proteins were analysed using PSORTb v.3.0 28 and the SecretomeP (SecP) 2.0 Server 24 29 . SecP predict proteins that are putatively secreted without having a detectable N-terminal signal sequence, i.e. by "non-classical secretion". Analysis with SecP indicated that four of the proteins predicted to be cytoplasmic (EF0970, EF1523, EF2718 and EF_B0004) may follow non-classical secretion. PSORTb predicted EF_B0004 (TraC protein) to be a cell wall protein and Pfam gave a significant hit against "Bacterial extracellular solute-binding proteins, family 5" (PF00496). It has been shown that TraC proteins play a role as surface pheromone receptor and are thereby involved in the regulation of the conjugation process 30 31 . Therefore, EF_B0004 was classified as a cell wall protein in this study. The other three proteins were retained as cytoplasmic, despite the fact that PSORTb predicted EF0970 (ribosomal protein L27) to have an extracellular location.

Identification of cytoplasmic proteins at extracellular locations is not unusual and at least 20 of the 33 proteins found in this study have been identified in previous studies of the secretomes or surface proteomes of Gram-positive bacteria (Tables 1 and 2) 13 18 26 32 33 34 35 36 . The majority of the identified cytoplasmic proteins were unique to the trypsin fraction and/or the beads fraction, indicating that these proteins bind to the cell envelope and need to be "shaved" from the surface, despite the lack of known binding motifs or domains. Many of the identified cytoplasmic proteins are highly abundant proteins like ribosomal proteins (Rbps; more than 20% of all identified proteins), EF-Tu/G, GADPH and chaperones, which suggests that cell lysis rather than an unknown active secretion process determines their extracellular presence. While the cell viability checks described above indicate that cell lysis due to the trypsin treatment is unlikely, it is conceivable that cell lysis in the cell culture prior to the trypsin treatment may have released intracellular proteins that somehow have re-associated with the cell envelope and escaped proteolytic degradation 26 37 .

While cytoplasmic proteins found in studies such as the present generally must be considered contaminants, there have been speculations in the literature that some of these actually may have extracellular functions. One example is Rbp L7/L12 (EF2715) which has been identified at the surface of several Gram positive bacteria 19 26 34 35 38 . We found Rbp L7/L12 in the beads fraction only and this implies an exposed localization, similar to the localization suggested in B. subtilis 26 . Bacterial Rbp L7/L12 has immunogenic properties in humans 38 39 and is being explored as candidate antigen for vaccine purposes 40 41 . Recent data showed that exposure of bacterial Rbp L7/L12 is a risk factor for colorectal cancer and stimulates progression of adenomas into carcinomas 42 . There has also been some speculation about possible adhesive roles of extracellular EF-Tu, DnaK, enolase and GAPDH, all identified in the present study and in a recent study of the laboratory strain Enterococcus faecalis JH2-2 13 , since these proteins bind strongly to human plasminogen 32 .

Proteins were annotated as being secreted to the culture medium if the bioinformatic analyses showed the presence of a SPaseI cleavage site and did not reveal any sequence or domain known to be involved in covalent or non-covalent binding to the cell wall. Only five such proteins were identified indicating that few secreted proteins are closely associated to the microbe. Three of the secreted proteins (EF0123, EF0394, EF0417) are annotated as N-terminally anchored in the LocateP database. It is, however, not easy to differentiate between secreted proteins with processed signal peptides and proteins that retain their signal peptides as N-terminal membrane anchors 12 . The Signal P server predicted all three proteins to contain a unique cleavage site, when using both algorithms in the program. All three were detected even without trypsin treatment, suggesting a loose association with the cell envelope. Taken together, we conclude that EF0123, EF0394 and EF0417 are secreted proteins.

EF2174 was detected after trypsin treatment only and putatively encodes a glycoside hydrolase belonging to family GH25 43 . This family comprises enzymes with lysozyme activity that are thought to be involved in peptidoglycan remodelling during cell division 44 .

The well known secreted metalloprotease coccolysin, a gelatinase (EF1818; GelE) was identified after all treatments (Table 1), indicating that GelE is loosely associated to the cell surface. Coccolysin is known to be associated with virulence and is capable of degrading cellular tissues during infection, cleaving substrates such as haemoglobin, collagen and fibrin 45 46 . There are also indications that GelE is required for biofilm formation 47 .

Among the 31 non-secreted non-cytoplasmic proteins found in this study (Table 3), three are annotated as integral membrane proteins. We identified only 0.5% of the predicted transmembrane proteins in the genome of E. faecalis V583 (Table 3). Such low numbers of identified transmembrane proteins are not unusual 19 26 37 48 and are likely to be due to limited accessibility of the proteins and/or a limited ability of trypsin to penetrate the cell wall. One of these transmembrane proteins, the sulfatase (EF1264), was detected by many peptides after all treatments (see above and Figure 3). The other two were only detected after trypsin treatment and only by a minimal number of peptides. EF0502 is a 781-residue hypothetical protein which according to topology predictions contains several extracellular domains. Both detected peptides are predicted to have an extracellular location. EF3257 is a 648-residue oxidoreductase belonging to the pyridine nucleotide-disulfide family with probably only two trans-membrane helices and a large extracellular domain. Both detected peptides stem from extracellular domains.

Five identified proteins are thought to be N-terminally anchored to the cell membrane via a Sec-type signal peptide that is not cleaved off during secretion. Relatively few such proteins were identified (Table 3), suggesting that they are expressed at low levels or that they generally have low accessibility for trypsin. Among these proteins are two penicillin-binding proteins (EF2857 & EF0991), one amidase (EF0737) and one protein of unknown function (EF1319). EF2857 and EF0991 are class B penicillin binding proteins (PBP), which are transpeptidases involved in the final stage of cell wall synthesis 49 . E. faecalis has three class B PBPs that have low affinity for β-lactams and can take over the transpeptidase activity of more high affinity PBPs when these are inhibited by antibiotics 50 . We also identified a L,D-transpeptidase (EF2860; YkuD domain), in all treatments and with a high number of peptide hits, indicating that this protein is abundant at the surface of V583. Studies on E. faecium have indicated that L,D-transpeptidase activity may represent another way to bypass inhibition of PBPs 51 52 .

The fifth protein, DltD (EF2746), is also involved in the biosynthesis of surface structures. The dltD gene is part of an operon consisting four genes (dltA-dltD) whose gene products are all necessary to incorporate D-alanyl residues into lipoteichoic acids (LTA) 53 54 . It has previously been shown that disruption of genes in the dlt operon of E. faecalis lead to diminished adhesion to eukaryotic cells and less biofilm formation, indicating that the dlt operon is involved in pathogenicity 55 . Interestingly, it is not fully established whether the N-terminal anchor tethers DltD to the inner or outer leaflet of the membrane 53 56 . The fact that DltD was detected after all types of treatments may be taken to suggest that the protein is attached to the outer leaflet, as one would expect for proteins using a Sec-type signal peptide as membrane anchor.

Of the four proteins containing a putative LPxTG anchor, two (EF2224 & EF2713) have unknown functions. EF2713 is up-regulated when E. faecalis V583 is grown in the presence of blood 57 , indicating a putative role of this protein in infection processes. The LPxTG anchor protein EF2224 contains five copies of a so-called DUF11 repeat with unknown function and is a putative member of the MSCRAMM (mirobial surface component recognizing adhesive matrix molecules) family of proteins 58 . These proteins contain tandemly repeated immunoglobulin-like folds as observed for staphylococcal adhesins. The E. faecalis V583 genome contains seventeen proteins belonging to the MSCRAMM family 59 . Interestingly, EF2224 is highly expressed during the infection process in humans 59 .

The LPxTG anchor protein EF1033 is a lipoamidase (Lpa) cleaving lipoic acids from lipoylated molecules 60 61 . EF1033 was only detected after trypsin treatment, indicative of covalent cell wall attachment. Lipoic acid is an essential sulphur containing cofactor of several enzymes. Interestingly, so far, E. faecalis is the only bacterial species in which Lpa activity has been detected 61 . Jiang and Cronan 61 speculate that the Lpa is a cytoplasmic salvage enzyme, but our experimental and bioinformatic results indicate that the enzyme is a cell wall anchored protein. Most likely, Lpa recruits its substrates from the environment, such as the GI tract.

The fourth LPxTG protein is the plasmid encoded surface exclusion factor Sea1 (EFA0052) which is involved in the regulation of sex pheromone-controlled conjugation 62 .

EF2860 is linked to the cell wall by a peptidoglucan binding domain and encodes a cell wall modifying transpeptidase homologous with YkuD from Bacillus subtilis. This protein was found after all three treatments and identified with relatively many peptide hits (Table 1), indicating that EF2860 is abundant on the surface and may show a relatively large extent of shedding. This protein may contribute to the antibiotic resistance of E. faecalis, as discussed above.

The most populated group of proteins identified in this study are the lipoproteins, of which 17 were detected, representing 23% of the lipoproteins putatively encoded on the E. faecalis genome (Table 3). Twelve of these were detected only after a "shaving" treatment. Seven of the detected lipoproteins are proteins with no predicted function. Two of these unknown lipoproteins (EF0176 & EF0177) are located on the same operon, share 70% sequence identity, and contain a "Basic membrane protein" domain (PF02608) belonging to Clan CL0144 in the Pfam database. This clan consists of proteins that are involved in chemotaxis and membrane transport of sugars as well as outer membrane proteins that are known for their antigenicity in pathogenic bacteria. Both proteins are homologous with a CD4+ T-cell-stimulating antigen in Listeria 63 . One of the other proteins with unknown function, EF0164, is annotated as N-terminally anchored in the LocateP database, but is classified as a lipoanchored protein on the basis of our analyses with LipoP.

Of the ten lipoproteins with predicted functions, four lipoproteins resemble the substrate-binding domains of multi-component ABC transporters for the import of peptides (EF0907, EF3106) or sugars (EF2221, EF2903). Three proteins (EF3041, EF3256, EFA0003) are involved in pheromone-regulated processes that include conjugation 31 , adding to the two cell wall associated proteins involved in these processes that are discussed above (EFB0004 & EFA0052). EF0071 seems to encode a glycoside hydrolase belonging to clan GH-G in the CAZy database. The protein is classified as N-terminally anchored in the LocateP database, but our analyses with the LipoP program clearly indicated that EF0071 is a lipoanchored protein. EF0685 belongs to the rotamase family and is thus likely to be involved in extracellular protein folding, possibly by exerting prolyl-peptidyl isomerase activity.

The final lipo-anchored protein is EF2556, fumarate reductase, which was detected by remarkably large numbers of peptides after all treatments (Table 1). E. faecalis is one of few bacteria that produces substantial amounts of extracellular superoxide. Fumarate reductase is likely to be involved in superoxide production and may thus be an important source of oxidative stress for the host 64 . It has been demonstrated that the superoxide from E. faecalis promotes chromosomal instability in mammalian cells and that this can lead to colorectal cancer 65 66 .