Surface-Affinity Profiling To Identify Host-Pathogen Interactions

Abstract

Proteolytic treatment of intact bacterial cells has proven to be a convenient approach for the identification of surface-exposed proteins. This class of proteins directly interacts with the outside world, for instance, during adherence to human epithelial cells. Here, we aimed to identify host receptor proteins by introducing a preincubation step in which bacterial cells were first allowed to capture human proteins from epithelial cell lysates. Using Streptococcus gallolyticus as a model bacterium, liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis of proteolytically released peptides yielded the identification of a selective number of human epithelial proteins that were retained by the bacterial surface. Of these potential receptors for bacterial interference, (cyto)keratin-8 (CK8) was verified as the most significant hit, and its surface localization was investigated by subcellular fractionation and confocal microscopy. Interestingly, bacterial enolase could be assigned as an interaction partner of CK8 by MS/MS analysis of cross-linked protein complexes and complementary immunoblotting experiments. As surface-exposed enolase has a proposed role in epithelial adherence of several Gram-positive pathogens, its interaction with CK8 seems to point toward a more general virulence mechanism. In conclusion, our study shows that surface-affinity profiling is a valuable tool to identify novel adhesin-receptor pairs, which advocates its application in other hybrid biological systems.

INTRODUCTION

The key to bacterial infection of host tissue is the establishment of a dependable connection between the bacterium and host surface structures. This is essential for the bacteria to withstand mechanical cleansing processes and to compete with other bacterial strains for microbial succession (16). After initial adherence, several pathogens can invade host cells using intracellular structures, e.g., the cytoskeleton, to sustain growth and prolong their survival times (8, 12). Ultimately, both adhesion and internalization of pathogenic bacteria will directly or indirectly (via induction of host responses) cause damage to the infected tissue. It is therefore important to fully understand the mechanisms underlying pathogenic interference so that new methods to prevent pathogenic bacteria from initiating an infectious process can be developed. In addition, knowledge about pathogen-specific interactions and subsequent responses may aid in the diagnosis of corresponding diseases.

Current advances in proteomic technologies provide opportunities to compare the protein content from different biologic systems, making it possible to characterize host-pathogen interactions in a global view. Therefore, the aim of this study was to explore the use of a proteolytic shaving approach coupled to liquid chromatography-tandem mass spectrometry (LC-MS/MS) to identify potential host proteins for bacterial interference. For this purpose, intact bacterial cells were first allowed to selectively bind host proteins from epithelial cell lysates, after which their surfaces were proteolytically shaved to generate small polypeptides that could be directly identified by LC-MS/MS (36, 37). Importantly, peptides of host proteins can be effectively recognized and discriminated from the bulk of bacterial peptides by computer-assisted analysis of the identified peptide sequences.

To obtain proof of concept for this approach, the interaction between the Gram-positive bacterium Streptococcus gallolyticus subsp. gallolyticus and human colonocytes was used as a model system. S. gallolyticus is an inefficient colonizer of the healthy human large intestinal tract but has long been associated with colorectal cancer (CRC) and endocarditis (4a, 7, 41). Our recent work has indicated that malignant epithelial sites may provide a route of infection for this bacterium via CRC-specific adhesion and translocation mechanisms (5, 6, 21). Therefore, knowledge of specific epithelial receptors for either invasion or adhesion of S. gallolyticus could provide novel insight into the association of S. gallolyticus with colonic malignancy.

MATERIALS AND METHODS

Bacterial strains and medium.

The strains used in this study were S. gallolyticus subsp. gallolyticus UCN34 (here, S. gallolyticus), which was previously isolated from a CRC patient with a concurrent endocarditis (31), and the intestinal strain Enterococcus faecalis (ATCC 19433). Strains were cultured in brain heart infusion (BHI) broth (Difco Laboratories) supplemented with 1% glucose at 37°C in 5% CO₂.

Cell lines.

Adherent monolayers of HT-29, Caco-2, and HCT116 colon adenocarcinoma cells were grown in Dulbecco's modified Eagle's medium (DMEM; Lonza) supplemented with 10% fetal calf serum (FCS), 20 mM HEPES, 100 nM nonessential amino acids, and 2 mM l-glutamine (Gibco) at 37°C in 5% CO₂. Cells were maintained at logarithmic growth by subculturing them every 3 to 5 days.

Cell affinity profiling.

For cell affinity profiling, HT-29 cells were washed three times with phosphate-buffered saline (PBS; pH 7.4) and lysed by a 5-min incubation in 2.5 ml of mammalian protein extraction reagent (M-PER) (Pierce) at room temperature. Soluble colonocyte proteins in the supernatant were harvested (fraction P2), whereas insoluble proteins present in the pellet after centrifugation were solubilized by PBS containing 2% Triton X-100 (fraction P1). Both protein fractions were aliquoted and stored at −80°C until use.

S. gallolyticus bacteria were grown overnight, resuspended at 1:20 in fresh medium, and grown for another 5 h. Next, 10⁹ intact bacterial cells were washed with sterile PBS and incubated with the above described HT-29-derived protein fractions, diluted to 100 μg/ml in sterile PBS containing 0.1% Triton X-100 (soluble protein fraction; P2) or 2% Triton X-100 (insoluble proteins; P1), for 1 h at room temperature. As control experiments, the same procedure was performed with the soluble (control fraction C2) and insoluble (control fraction C1) fractions incubated without bacteria, while S. gallolyticus cells were also incubated in the absence of HT-29 proteins (control fraction C3). Since our main goal was to screen if this assay could be used to identify host proteins with affinity for bacterial surface components, we used only one biological replicate per condition, which was compensated by a thorough biochemical validation of a selected candidate. After incubation, bacteria were collected by centrifugation and washed twice with the corresponding incubation buffers containing 0.1% or 2% Triton X-100 and three times with PBS without detergent. Bacterial cells were resuspended in PBS containing 10 units of trypsin (sequencing grade; Sigma-Aldrich) and incubated for another hour at 37°C under gentle shaking. Next, cells were pelleted by centrifugation, after which the supernatant containing released peptides and protein fragments was treated with 1 mM dithiothreitol (DTT) and 1 mM iodoacetamide in two successive steps of 30 min at room temperature. Fresh trypsin was added, and tryptic cleavage was continued for 18 h at 37°C.

LC-MS/MS.

Prior to nano-LC-MS/MS analysis, all peptide samples were purified and desalted using C18 spin columns (Pierce), and Triton X-100 was removed using anti-Triton beads (Calbiochem). Nano-LC-MS/MS analysis was performed on an Agilent Nanoflow 1100 liquid chromatograph coupled online to a 7T linear quadrupole ion trap Fourier transform (LTQ-FT) ion cyclotron resonance mass spectrometer (Thermo Fisher, Bremen, Germany). Peptides were loaded directly onto the analytical column using buffer A and eluted using buffer B (see Table S1 in the supplemental material) at a flow rate of 300 nl/min. Peptides eluting from the column tip were injected into the mass spectrometer via a nano-spray ion source using a spray voltage of 2.1 kV. The mass spectrometer was operated in positive ion mode using data-dependent fragmentation to analyze the top four most abundant ions from each precursor scan. For detailed information on nano-LC-MS/MS analysis and MS settings, see Table S1.

Fragmentation and precursor data were parsed from the RAW files by ExtractMSN (ThermoFisher) and converted into a Mascot generic peak list. Peptides and proteins were identified using the Mascot algorithm (Mascot, version 2.2; Matrix Science) to search a local copy of human RefSeq database (release 33) combined with the S. gallolyticus protein database (provided by P. Glaser) (see Table S1 in the supplemental material for data analysis settings). The following modifications were allowed in the search: carbamidomethylation of cysteines (fixed), oxidation of methionine (variable), and deamidation of asparagine and glutamine (variable). After a search, the peptides matching to human proteins (see Table S2 in the supplemental material) were validated using the criteria that are summarized in Table S1.

CK8 binding assay, SDS-PAGE, and Western blot analysis.

To monitor affinity of cytokeratin-8 (CK8), intact bacteria (S. gallolyticus and E. faecalis) were incubated with colonocyte fractions as indicated before. After three washes with PBS, bacterial cells were directly incubated in SDS sample buffer (50 mM Tris-HCl, pH 6.0, 2% SDS, 5% β-mercaptoethanol, and 10% glycerol) and incubated for 15 min at 95°C prior to 10% glycine SDS-polyacrylamide (PAA) gel electrophoresis (SDS-PAGE) (38).

To determine subcellular localization of CK8, the colon adenocarcinoma cell lines were fragmented, using an S-PEK kit (subcellular proteome extraction kit; Calbiochem). into four protein fractions: cytosolic (FI), membrane and organelle (FII), nuclear (FIII), and cytoskeletal (FIV). Subcellular protein fractions (20 μg) were separated by SDS-PAGE. For detection of CK8, proteins were transferred to polyvinylidene difluoride (PVDF) membranes (Amersham) by Western blotting (38). Blots were incubated with monoclonal rabbit anti-keratin-8 antibodies (Abcam) diluted 1:25,000 or with polyclonal rabbit anti-CYP1A1 antibody diluted 1:5,000 (Millipore). CYP1A1, a typical membrane protein with an amino-terminal membrane anchor that resides in the endoplasmic reticulum, was used to determine the effectiveness of subcellular proteome fractionation. Bound antibodies were visualized with an enhanced chemiluminescence (ECL) detection system (Amersham) using anti-rabbit IgG-horseradish peroxidase (HRP) conjugates diluted 1:25,000 (Jackson ImmunoResearch).

In-cell Western (ICW) analysis.

Ninety-six-well plates were coated with intact S. gallolyticus or E. faecalis cells (1 × 10⁹ cells/ml) that were harvested during the exponential growth phase (optical density at 600 nm [OD₆₀₀] of 0.6) to investigate the binding of CK8. Briefly, bacteria were incubated with 10 μg of protein fraction FII of HT-29, HCT116, and Caco-2 cells in PBS at room temperature for 1 h with gentle mixing. Next, bacteria were pelleted and resuspended in PBS and allowed to attach to a poly-l-lysine-coated 96-well plate at 4°C overnight. The coated plates were washed with PBS and fixed with 3.7% formaldehyde before being blocked with 5% BSA. CK8 bound to bacterial cells was detected with monoclonal rabbit anti-CK8 antibody and visualized with an Odyssey instrument (Li-Cor Biosciences) using a secondary anti-rabbit antibody conjugated with Alexa Fluor 800 (Molecular Probes). The DNA stain DRAQ5 (Invitrogen) was used as a reference to control for the amount of cells.

Confocal microscopy.

HT-29, Caco-2, and HCT116 cells were seeded onto coverslips in six-well plates for 24 h at 37°C. Then, cells were fixed with 3.7% formaldehyde and either left intact or made permeable by four washes with PBS-0.3% Triton. Next, cells were blocked with 5% BSA in PBS and subsequently incubated with rabbit anti-cytokeratin-8 antibodies (1:500) overnight at 4°C. Cells were washed three times with PBS-0.3% Triton and stained with goat anti-rabbit antibodies conjugated with Alexa Fluor 488. The DNA stain DRAQ5 was used to stain the nucleus. CK8 was visualized with laser scanning confocal microscopy (Olympus FV1000).

Cross-linking of recombinant CK8 to bacterial surface proteins.

S. gallolyticus and E. faecalis were grown overnight, diluted 1:10 in fresh culture medium, and incubated for 4 h at 37°C in 5% CO₂. After centrifugation at 4,000 × g for 10 min, the bacterial pellet was washed three times with PBS to remove loosely bound proteins. The cells were resuspended in 500 μl of PBS containing 8 μg of recombinant CK8 (Progen Biotechnik) for 1 h at room temperature (RT) with gentle shaking. Next, bacterial cells were washed three times with incubation buffer to remove loosely bound proteins. Bound recombinant CK8 protein was cross-linked to the bacterial cell wall with 2 mM 3,3′-dithiobis(sulfosuccinimidylpropionate) (DTSSP; Pierce), washed three times with incubation buffer, and subsequently boiled in SDS sample buffer with or without DTT to either disrupt or conserve the cross-linked protein complexes. Cross-linked recombinant CK8 was visualized with SDS-PAGE and Western blotting.

Immunoprecipitation of cross-linked protein complexes containing CK8.

For the immunoprecipitation of CK8 containing cross-linked protein complexes, bacteria were lysed in 10 mM Tris-buffer containing 2 mM MgCl₂, 3 mg/ml lysozyme (Fluka), and 100 μg/ml mutanolysin (Sigma-Aldrich) for 60 min at 37°C. Then, 1% Triton and protease inhibitors were added to the lysate, and samples were sonicated on ice with intervals of 0.3 s for 20 s (two runs of 30 pulses each). To remove bacterial DNA, samples were incubated with Benzonase for 30 min on ice and subsequently spun at maximum speed for 30 min at 4°C. Supernatants were incubated with anti-rabbit CK8 antibody (1/100) overnight at 4°C with shaking. Next, antibodies were precipitated with protein G magnetic beads (Thermo Scientific) for 2 h at RT. Beads were washed three times with PBS-0.1% Triton, after which immunoprecipitated cross-linked protein complexes were boiled in electrophoresis buffer at 100°C for 5 min and analyzed by 7.5% SDS-PAGE after staining with blue-silver Coomassie (9). Alternatively, proteins were blotted on PVDF membrane to control the cross-linking with monoclonal rabbit anti-keratin-8 antibodies (Abcam) diluted 1:25,000 or with polyclonal rabbit anti-enolase antibodies (generously provided by V. Pancholi) diluted 1:10,000. Antibodies were visualized using an ECL detection system as described previously.

Analysis of protein complexes by vMALDI-LTQ MS/MS.

Cross-linked protein complexes were excised from SDS-PAA gels for in-gel trypsin digestion as described previously (15). After digestion, peptides were extracted by sonication in a water bath for 15 to 20 s, concentrated using a centrifugal evaporator, and diluted 1:1 with 2% trifluoroacetic acid (TFA). Next, samples were spotted on a stainless steel matrix-assisted laser desorption ionization (MALDI) plate. Sample analysis was performed on a linear ion trap fitted with an intermediate-pressure MALDI source (vMALDI-LTQ; Thermo Fisher Scientific) (18). In total, five full MS runs were analyzed, each resulting in a selection of the 10 highest peaks, which were further analyzed by collision-induced dissociation fragmentation analysis. Generated RAW data files were analyzed with SEQUEST (version 28, revision 12), and identifications were considered significant with a peptide probability of >1E−002 and a protein probability of >1E−003. For further details on MALDI-MS analysis, data analysis settings, and validation criteria see Table S1 in the supplemental material.

RESULTS

Profiling eukaryotic proteins with affinity for the bacterial cell surface.

The main aim of this study was to provide proof of concept for the use of a surface-shaving LC-MS/MS approach to identify host proteins with affinity for intact bacterial cells. To this purpose, eukaryotic proteins from HT-29 intestinal epithelial cells were first allowed to interact with intact S. gallolyticus cells, after which firmly attached host proteins were released by tryptic digestion of surface-exposed proteins and protein complexes (Fig. 1). As expected, shotgun MS/MS analysis of the released peptides resulted in the identification of both bacterial and eukaryotic proteins. Candidate targets for bacterial interference (samples P1 and P2) were discriminated from protein precipitates originating from eukaryotic cell lysates that could cofractionate with intact bacterial cells during centrifugation (samples C1 and C2) and “contaminating” proteins (sample C3) (see “Cell affinity profiling” above for details on these samples). A total of 27 eukaryotic proteins fulfilled the protein identification criteria (see Tables S1 and S2 in the supplemental material). Proteins were ranked according to the change in exponentially modified protein abundance index (delta emPAI) values that were generated by subtracting the emPAI value of control sample C1 or C2 from the corresponding experimental sample P1 or P2 (Table 1). Next, proteins that were identified by at least two peptides in sample P1 or P2, but not in sample C3, were assigned as potential bacterial interference factors. These stringent identification criteria narrowed the results to a total of eight identified proteins from the soluble fraction (P2) and three proteins in the insoluble fraction (P1) (Table 1). Among these proteins were several cytoskeleton proteins as well as ribonucleoproteins. Interestingly, cytokeratin-8 (CK8) and CK18 received the most significant emPAI values and peptide hits. Importantly, literature supports a role for CK8 as a potential bacterial adherence factor (17, 20) but also indicates that it is expressed at increased levels by colon tumor cells (2), which could be relevant for the increased incidence of infection of S. gallolyticus in CRC patients (4a). For these reasons CK8 was selected as a target for further evaluation in this study.

Fig. 1.

Schematic representation of the bacterial surface-affinity profiling approach by proteolytic shaving coupled to LC-MS/MS. S. gallolyticus cells were incubated with host protein P1/P2 fractions or without eukaryotic host proteins. Control conditions concerned the incubation of the respective host protein fractions without S. gallolyticus cells (C1/C2) to check for protein precipitates and a control for contaminating proteins that are not derived from the added host protein pool (C3). After a washing step, the bacterial surface was shaved with trypsin to release peptides from accessible proteins. Protein fragments in the supernatants were further cleaved with trypsin for 18 h and purified for LC-MS/MS analysis (A). Subsequent database searches resulted in the identification of 27 proteins, of which 11 proteins were assigned as candidate receptors for bacterial interference (B) (Table 1).

Table 1.

Eukaryotic proteins identified by LC-MS/MS^a

Protein group and identification no. (accession no.)	Protein name and/or description	Delta emPAI		Delta peptide no. (no. of peptide hits)
		P1 − C1	P2 − C2	P1 − C1	P2 − C2
High detergent concn (insoluble proteins)
gi|148352329 (NP_001011553.2)	Cell division cycle 10 isoform 2	0.233		2
gi|157412270 (NP_112480.2)	Heterogeneous nuclear ribonucleoprotein M isoform b	0.160		2
gi|4506675 (NP_002941.1)	Ribophorin I precursor	0.150		2
Low detergent concn (soluble proteins)
gi|4557888 (NP_000215.1)	Cytokeratin 18		1.228		7
gi|4504919 (NP_002264.1)	Cytokeratin 8		1.212		9
gi|14389309 (NP_116093.1)	Tubulin alpha 6		0.369		3
gi|57013276 (NP_006073.2)	Tubulin alpha, ubiquitous		0.369		3
gi|15718687 (NP_000996.2)	Ribosomal protein S3		0.359		2
gi|24234699 (NP_002267.2)	Cytokeratin 19		0.318		3
gi|4501885 (NP_001092.1)	Beta-actin		0.274		2
gi|116284394 (NP_079005.3)	Myosin, heavy chain 14 isoform 2		0.097		4

^aEukaryotic proteins identified by LC-MS/MS with corresponding reference number, emPAI values, and the number of peptide hits per fraction (see Table S2 in the supplemental material for peak lists and peptide sequences). Proteins that were identified in fraction P1 and P2 but not in the control fraction C3 were evaluated based on delta emPAI values above 0 and a delta peptide number of at least 2 with the respective control fractions C1 and C2. The 11 proteins matching these criteria were marked as potential candidate target proteins for bacterial interference. All other identified proteins can be found in Table S2. P1, S. gallolyticus plus the HT-29 insoluble fraction; P2, S. gallolyticus plus the HT-29 soluble fraction; C1, HT-29 insoluble fraction without S. gallolyticus; C2, HT-29 soluble fraction without S. gallolyticus; C3, S. gallolyticus without HT-29 proteins.

Presence of CK8 at the eukaryotic cell surface.

In general, keratins are intracellular cytoskeleton proteins. To validate a possible role of CK8 as an adherence receptor for S. gallolyticus, it was important to know whether this protein is also present at the cell surface of human epithelial cells. To investigate this, HT-29 cells were fractionated and evaluated by Western blotting. As expected, these experiments showed that CK8 is most abundant in the cytoskeleton fraction; however, this protein was also detectably present in the cytosolic, membrane, and nuclear fractions (Fig. 2A). Antibodies against the membrane protein CYP1A1 were used to control for correct fractionation, which confirmed that this protein was detected only in the membrane fraction (Fig. 2B). However, as this fraction also includes intracellular membranes, confocal microscopy was used to further investigate the cell surface expression of CK8 (Fig. 2C). This revealed a speckle-like membrane staining of CK8 in nonpermeabilized epithelial cells. In contrast, cell permeabilization results in a more equal staining that mainly presents the contours of the cell. Overall, these experiments indicate that CK8 is primarily an intracellular protein that also has an exposed component at the epithelial cell surface of colorectal cancer cells.

Fig. 2.

Localization of CK8 in epithelial cells. The presence of CK8 (A) and CYP1A1 (B) in HT-29 cell fractions was evaluated by immunoblotting. Fractions are as follows: FI, cytosol; FII, membrane; FIII, nucleus; FIV, cytoskeleton. (C) Subcellular localization of CK8 was visualized in permeable cells and cells that were nonpermeabilized by confocal microscopy, showing a speckle-like membrane staining in nonpermeabilized Caco-2 and HT-29 cells (Membrane), whereas a more equal staining along the cell contours is observed in permeable cells (Total).

CK8 binding to the bacterial cell surface.

After showing that CK8 also has a possible surface-exposed component in addition to the intracellular component, we confirmed the binding of CK8 to the bacterial cell surface by Western blotting and ICW analysis. After incubation with the epithelial membrane fraction, S. gallolyticus cells were washed and boiled in SDS-PAGE sample buffer to release extracellular bound proteins. Western blot analysis confirmed that CK8 from HT-29 and Caco-2 cells was detectably retained by the bacterial cell surface under the applied conditions (Fig. 3A). To evaluate whether this feature is restricted to S. gallolyticus, the same procedure was performed with E. faecalis cells. As shown in the lower panel of Fig. 3B, E. faecalis cells retain CK8 to a much higher extent than observed for S. gallolyticus. The latter observation was confirmed with ICW analysis (Fig. 3B), suggesting that bacterial binding to CK8 is a rather general phenomenon.

Fig. 3.

CK8 binding to the bacterial surface. (A) S. gallolyticus and E. faecalis cells were incubated with HT-29 and Caco-2 cell lysates (membrane fraction) to allow binding of CK8, which was evaluated by SDS-PAGE and immunoblotting. (B) ICW analysis was used to monitor binding of CK8 to immobilized bacteria in 96-well plates. Results are expressed as the number of fluorescent counts/mm² relative to the amount of bacterial cells as measured by DNA staining with DRAQ5 (see “In-cell Western analysis”).

CK8 interacts with bacterial enolase.

Evaluation of the bacterial cell surface confirmed that CK8 binds to S. gallolyticus and E. faecalis cell surfaces. To reveal which bacterial adhesin is involved in CK8 binding, recombinant CK8 was incubated with intact bacterial cells and subsequently cross-linked to the bacterial cell surface using the membrane-impermeable agent DTSSP. Next, protein complexes containing CK8 were isolated through immunocapture experiments, after which potential adhesion-receptor pairs were visualized by SDS-PAGE (Fig. 4A). The cross-linked adhesion-receptor pairs that were either left intact or disrupted with DTT were evaluated with Western blotting for their containment of CK8. Fig. 4B clearly shows that the cross-linked complex (CL) is retained at approximately 155 to 160 kDa. Disrupting the cross-linked proteins with DTT reduced CK8 to its non-cross-linked form at 58 kDa (reduced cross-linked complex [RCL]). In-gel trypsin digestion and subsequent vMALDI-LTQ MS/MS analysis of six possible complexes resulted in the identification of bacterial enolase (47 kDa) in relatively faint bands with estimated sizes of 155 (two peptides) and 160 (one peptide) kDa (Fig. 4C). LC-MS/MS analysis of this peptide sample confirmed the identification of enolase in this 155-kDa protein band (data not shown). The positions of other identified bacterial proteins on SDS-PAGE were in accordance with their theoretical native molecular weights, and these proteins were therefore considered not to be in complex with CK8 (see Table S3 in the supplemental material).

Fig. 4.

Identification of bacterial interaction partners of CK8. (A) Cross-linked adhesin-receptor pairs were immunoprecipitated with anti-CK8 antibodies and visualized by SDS-PAGE. The position of native CK8 at approximately 58 kDa is indicated. After in-gel tryptic digestion and MS/MS analysis, bacterial enolase could be identified in the protein complex at 155 kDa. The other protein products at 120, 110, 85, and 75 kDa were all identified as bacterial proteins with matching protein sizes, indicating that these proteins are not part of a complex (see Table S3 in the supplemental material). (B) Immunoprecipitated proteins were analyzed by Western blotting either in cross-linked (CL) or reduced (RCL) form; recombinant CK8 was loaded in lane 1. The positions of recombinant CK8 at 58 kDa (native) and cross-linked products between 110 and 160 kDa are indicated. (C) Protein sequence and peptides of S. gallolyticus enolase that were identified by MS/MS in the ∼155-kDa complex. (D) Visualization of enolase by Western blot analysis of the same samples that were loaded in panel B along with the enolase control. The positions of enolase at 50 kDa (native) and an enolase-containing cross-linked complex at ∼155 kDa are indicated. Importantly, this complex had the same mobility in the SDS-PAA gel as the CK8-containing product with the highest molecular mass, as indicated in the CL lane of panel B. Note that the cross-linked product containing CK8 and enolase (CL lanes) can be disrupted by DTT (RCL lanes), indicating that it concerns a genuine DTSSP-induced complex.

To further investigate the interaction between enolase and CK8, cross-linked complexes were analyzed by Western blotting. As already shown in Fig. 4B, CK8 is present in multiple protein complexes, of which the ∼155-kDa complex also reacts with anti-enolase antibodies (Fig. 4D). Importantly, the enolase-CK8 complex is detected only in the cross-linked nonreduced sample (CL) and not in the reduced sample in which cross-links have been disrupted (RCL). Taken together, these experiments reveal that epithelial CK8 and bacterial enolase have the potential to form an adhesin-receptor pair, which may contribute to the interaction between bacteria and host epithelial cells.

DISCUSSION

Advances in proteomic tools provide the opportunity to explore interactions between pathogens and the host in a broad manner. The fact that bacterial shaving has previously been proven to be a successful method to identify bacterial proteins at the cell surface of Gram-positive bacteria (14, 30, 33, 34, 36, 40) inspired us to explore the ability to use a modified approach to profile potential host-pathogen interactions. Here, we report that it is indeed possible to identify host factors that are candidate receptors for bacterial cells by proteolytically shaving the bacterial surface after preincubation with complex mixtures of eukaryotic proteins. Surprisingly, the most significant hits were limited to soluble proteins, whereas no classical membrane proteins were identified. However, in addition to the difficulties of solubilizing transmembrane proteins, the identification of this class of membrane proteins is more challenging. Transmembrane proteins naturally contain fewer cleavage sites, which results in larger peptides upon proteolytic cleavage. Despite these apparent limitations, we have shown that this approach allows the identification of cytoskeletal proteins, such as CK8 and CK18, that may be important mediators of bacterial infection by either functioning as surface receptors or by hijacking the protein during invasion of epithelial cells.

Keratins are part of the cytoskeleton of eukaryotic cells that is composed of three different types of morphologically distinct filamentous structures: microfilaments, intermediate filaments, and microtubules (3). This cytoskeletal network is responsible for the mechanical integrity of the cell. CK8 and CK18 are the major constituents of intermediate filaments of simple epithelia, such as those of the intestine. In addition, both keratin-18 and -19 can form heterodimers with CK8. Interestingly, CK8 and CK18 have continuous and increased expression in tumor cells, while the expression of other cytokeratins is lost (2, 29). CK8, CK18, and CK19 are the most abundant intermediate filaments expressed in malignancy and can be detected in a number of body fluids of cancer patients, while the level of cytokeratins in the circulation of healthy individuals is low (25). Our experiments suggest that CK8 has a surface-exposed component in colon adenocarcinoma cells that allows its interaction with intestinal pathogens. This theory is supported by several reports on CK8/CK18 membrane localization in colon tumor cells (10, 13, 17, 20, 24). Furthermore, Pankov et al. found that this surface localization was present only in tumor cells and not in healthy tissue (29). Since our results have identified CK8 and CK18 as possible receptors for S. gallolyticus adherence, this might reflect the association between CRC and infections with this bacterium. However, our data also suggest that binding to these keratins is not a very specific feature of S. gallolyticus, which is in line with our recent finding that this association is more likely to be dominated by the specific ability of S. gallolyticus to adhere to (intestinal) collagens (5).

Cytosolic proteins and, in particular, cytoskeleton proteins have been described to be important for adhesion, invasion, and transcytosis of several bacteria, e.g., Clostridium difficile, Neisseria gonorrhoeae, and Staphylococcus aureus (26, 32, 39). Interestingly, our study is not the first to provide evidence for the role of CK8 as a mediator of bacterial adherence. Previously, Tamura and Nittayajarn (35) showed a role of CK8 as an adherence factor for different serotypes of group B streptococcus (GBS) and concluded that this represents a shared mechanism among these strains. Our results underscore a more general role for CK8 as a bacterial interference factor since S. gallolyticus as well as E. faecalis cells were able to interact with CK8. Also, the observation that other bacteria like Streptococcus macedonicus, Lactobacillus plantarum (our unpublished observations), Staphylococcus aureus, and Streptococcus pyogenes (35) were able to interact with CK8 strengthens the notion that CK8 is a more general adherence mediator for Gram-positive bacteria. Notably, the observed high-CK8-binding capacity of E. faecalis cells compared to that of S. gallolyticus is consistent with their relative adherence characteristics to HT-29 monolayers (5).

The previous observation of Tamura and Nittayajarn (35), who showed that proteolytic treatment of the GBS surface abolished its interaction with CK8, is of particular importance for our findings as this bacterial proteinaceous compound remained previously unidentified. Our current investigations on CK8 interaction show for the first time that bacterial enolase is a bacterial counterpart of this eukaryotic protein. Notably, S. gallolyticus enolase was among the abundant bacterial proteins released by the proteolytic shaving procedure (our unpublished data). These findings fit with a common mechanism for bacterial adherence to CK8 as enolase is a conserved bacterial protein with great sequence conservation between different bacterial species (22). Furthermore, enolase has already been shown to be expressed at the bacterial cell surface and to be of importance for the binding of, e.g., S. pneumoniae to airway epithelial cells (1, 4, 11, 27, 28). Our current data suggest that surface-exposed bacterial enolase can, in addition to plasminogen activator (27), also bind to epithelial CK8. It should be kept in mind, however, that our data suggest that besides enolase, CK8 could also interact with other presently unidentified bacterial adhesins. In addition, our profiling approach also resolved other candidate eukaryotic receptors for bacterial adherence, like CK18, the role of which needs further investigations.

In conclusion, our study shows that surface-affinity profiling is a promising approach to discover eukaryotic mediators of bacterial interference. However, further refinement of this approach is necessary to apply this procedure also to eukaryotic membrane proteins. Furthermore, bacterial adherence to host tissue might also involve human or bacterial proteins, such as fimbriae, that are not cleaved by trypsin (19). Thus, it cannot be excluded that some important host-bacteria interactions are being missed by this procedure. In future investigations, the number of detected interactions might be enhanced by using complementary proteases that cleave surface proteins at other sites (34).

It goes without saying that future follow-up experiments should further confirm the biological relevance of the CK8-enolase interaction for bacterial infection. Such experiments could consist of, but not be limited to, an examination of the influence of CK8 downregulation on adherence to epithelial cells and of the virulence characteristics of an S. gallolyticus enolase knockout in vitro and in vivo. Our first experiments to reduce the expression of CK8 by okadaic acid treatment (23) appeared not to be compatible with bacterial adherence assays due to a general loss of cell attachment (our unpublished observations). It might be anticipated that an RNA-silencing approach would turn out to be more successful in this matter. Cork et al. have previously shown that enolase mutants of group A streptococci display a lower adherence capacity than cells of the wild-type strain, stressing the importance of enolase for bacterial adherence. Unfortunately, however, it was not possible until now to generate an S. gallolyticus enolase mutant strain to perform similar experiments in our model systems. Nevertheless, our present data clearly illustrate that a straightforward surface-affinity profiling procedure can provide leads for the identification of novel adhesin-receptor pairs. Thus, this approach demonstrates its potential to improve our understanding of host-pathogen interactions at human epithelial surfaces.