A strategy for enrichment of claudins based on their affinity to Clostridium perfringens enterotoxin

Background Claudins, a family of protein localized in tight junctions, are essential for the control of paracellular permeation in epithelia and endothelia. The interaction of several claudins with Clostridium perfringens enterotoxin (CPE) has been exploited for an affinity-based enrichment of CPE-binding claudins from lysates of normal rat cholangiocytes. Results Immunoblotting and mass spectrometry (MS) experiments demonstrate strong enrichment of the CPE-binding claudins -3, -4 and -7, indicating specific association with glutathione-S-transferase (GST)-CPE116–319 fusion protein. In parallel, the co-elution of (non-CPE-binding) claudin-1 and claudin-5 was observed. The complete set of co-enriched proteins was identified by MS after electrophoretic separation. Relative mass spectrometric protein quantification with stable isotope labeling with amino acids in cell culture (SILAC) made it possible to discriminate specific binding from non-specific association to GST and/or matrix material. Conclusion CPE116–319 provides an efficient tool for single step enrichment of different claudins from cell lysates. Numerous proteins were shown to be co-enriched with the CPE-binding claudins, but there are no indications (except for claudins -1 and -5) for an association with tight junctions.


Background
The permeation of small molecules through the paracellular pathway in epithelial and endothelial cells is controlled by tight junctions, which constitute a continuous intercellular contact. The tight junction network, organized in strands, is formed by different proteins, among them Zonula occludens (ZO) proteins, claudins, occludin and junctional adhesion molecules. Although the regulation and the molecular composition of the tight junctions are far from being understood, it is generally accepted that claudins play a crucial role in tightening cell-cell contacts. The expression of claudins varies in different cell types and organs, and their protein-protein interactions which seal the paracellular space have been intensively discussed [1,2]. Different members of the claudin family, in particular claudin-3 and -4, are known to be receptor molecules of Clostridium perfringens enterotoxin (CPE) [3]; the COOH-terminal part of this 35 kDa protein has been shown to bind to the second extracellular loop of claudin-3 [4]. The minimum binding region of CPE has been nar-rowed down to the last 30 amino acids [5] and deletion of the six C-terminal amino acids abolishes claudin-binding activity [6].
The enrichment of proteins from a complex mixture is still a challenge, in particular when the target structures are at low abundance and/or are integral membrane proteins. Moreover, one of the major concerns in affinity-based enrichment strategies is non-specific binding, which must be diminished or at least identified [7][8][9]. In addition to co-immunoprecipitation techniques [10], there is increasing interest in applying recombinant proteins [11] or protein domains [12] for selectively enriching proteins; there are, however, only a few examples directed at tight junction proteins. A co-immunoprecipitation approach (based on the binding of tight junction proteins to atypical protein kinase C zeta) has been applied for purification of the tight junction complex [13]. The interaction of ZO-1 with a-actinin-4 has been demonstrated by a pulldown assay utilizing immobilization of recombinant PDZ domains of this tight junction protein [14].
The investigations presented here are aimed at extracting members of the claudin protein family from cell lysates, using a pull-down assay based on the affinity of CPE 116-319 to the second extracellular loop of several claudins. The binding proteins are eluted, electrophoretically separated and identified by mass spectrometry (MS). Additionally, in combination with stable isotope labeling with amino acids in cell culture (SILAC) [15], it is investigated whether this approach allows the identification of proteins specifically associated with claudins and, consequently, with the tight junction complex.
GST-CPE 116-319 fusion protein and GST were expressed in E. coli (BL21). After induction, bacteria were grown to an optical density of 0.7 ± 0.1 (at 600 nm) and were harvested by centrifugation for 10 min at 3200 × g. The pellets were resuspended in lysis buffer containing 1% Triton X-100, 0.1 mM phenylmethylsulphonyl fluoride, 1 mM ethylenediaminetetraacetic acid (EDTA) and protease inhibitor cocktail (Sigma-Aldrich, Taufkirchen, Germany) in phosphate-buffered saline (PBS) and underwent two passages in an EmulsiFlex-C3 homogenizer (Avestin Europe GmbH, Mannheim, Germany). The insoluble cell debris was removed by centrifugation for 1 h at 40000 × g. For purification of the recombinant proteins, the clari-fied supernatants were loaded onto columns containing glutathione-agarose under gravity flow. The resin was rinsed twice with washing buffer (1% Triton X-100 in PBS). GST-CPE 116-319 and GST were eluted from the resin with elution buffer (10 mM reduced glutathione, 50 mM Tris/HCl, pH 9.5). The samples were dialyzed against PBS. Protein concentrations were determined using the 2-D Quant Kit (GE Healthcare, Freiburg, Germany).

Preparation of epithelial cell lysates
Normal rat cholangiocytes (NRC), a gift of N.F. LaRusso, Rochester, MN/USA, were cultured in 75 cm 2 rat tail collagen-coated cell culture flasks in DMEM/HAM's F12 medium (Biochrom, Berlin, Germany) with 5% CO 2 in air [16]. Arginine-and lysine-deficient medium (Biochrom) was used for SILAC experiments, where one cell population was supplemented with [ 12 C 6 ]arginine and [ 12 C 6 ]lysine, whereas another cell population was grown in medium containing [ 13 C 6 ]arginine and [ 13 C 6 ]lysine. Cells were grown to confluence, washed twice with icecold PBS to remove serum proteins and scraped in icecold PBS. The cell suspensions were centrifuged for 10 min at 300 × g; the pellets were frozen in liquid nitrogen and stored at -80°C.
NRC cell pellets corresponding to four 75 cm 2 cell culture flasks, prepared under either labeled or unlabeled conditions, were resuspended in 1 ml lysis buffer (1% Triton X-100, 50 mM Tris/HCl, pH 7.4, 150 mM NaCl, and Complete protease inhibitors, EDTA-free, Roche Diagnostics GmbH, Mannheim, Germany). Cells were homogenized using a syringe and a 24-gauge needle. After incubation on ice for 30 min, the cell extract was cleared by centrifugation for 10 min at 10000 × g.

Affinity purification
Glutathione-sepharose™ 4B beads (GE Healthcare, Freiburg, Germany) were washed with cold PBS before use. Equal amounts of GST-CPE 116-319 (this construct was found in preceding experiments to exert a stronger association to the second extracellular loop of claudin-3 as compared to CPE 290-319 ) or GST were rotated with beads for 1 h at 4°C. After three washing steps with PBS containing 1% Triton X-100, the beads loaded with GST-CPE 116-319 were incubated with the lysate of labeled cells for 2 h at 4°C. As a control, beads were loaded with GST and rotated in the lysate of unlabeled cells for the same period of time. In addition, a corresponding inverse experiment ( 13 C-labeling of GST fraction) was carried out. Non-specifically bound proteins were diminished by rinsing the beads three times with a washing buffer (twice with 1% and once with 0.1% Triton X-100 in PBS). Bound proteins were eluted with 200 ml of elution buffer (10 mM reduced glutathione, 50 mM Tris/HCl, pH 9.5) in two steps. Beads were incubated in elution buffer for 10 min at 4°C. Both eluates were pooled.

Immunoblotting
Proteins were separated by 15% sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and transferred to a nitrocellulose membrane. After blocking with 5% nonfat milk in Tris-buffered saline with Tween-20 (TBST, containing 10 mM Tris/HCl, pH 7.4, 150 mM NaCl, 0.05% Tween-20), the membranes were washed three times in TBST. Membranes were incubated at 4°C overnight with monoclonal antibodies to either claudin-1, -3, -4, -5, or -7 (Invitrogen/Zymed Laboratories, Karlsruhe, Germany), diluted in TBST. After washing three times in TBST, the immunoblots were incubated with horseradish peroxidase-conjugated anti-rabbit (claudin-4, anti-mouse) secondary antibody (Invitrogen) for 1 h at room temperature. The membranes were washed for at least 30 min, exchanging TBST every 10 min, and signals were visualized by horseradish peroxidase-dependent chemiluminescence (GE Healthcare, Freiburg, Germany). For sphingosine kinase 2 (SphK2) immunoprecipitation, 2 mg of the primary monoclonal antibody (Santa Cruz Biotechnology, Heidelberg, Germany) were added to 50 ml of NRC lysate (500 mg total protein) and incubated for 1 h at 4°C. Subsequently, 10 ml of resuspended protein A/ G PLUS agarose (Santa Cruz Biotechnology, Heidelberg, Germany) were added and incubated for 1 h at 4°C. As a control, 10 ml of protein A/G PLUS agarose were incubated with 50 ml of NRC lysate for 1 h at 4°C. Immunoprecipitates were collected by centrifugation at 300 × g for 1 min at 4°C. The pellet was washed twice with 1% and once with 0.1% Triton X-100 in PBS, each time repeating the centrifugation step. After the final wash, the pellet was resuspended in 20 ml electrophoresis sample buffer (2% SDS, 10% glycerol, 0.01% bromophenol blue and 0.1 M dithiothreitol in 50 mM Tris/HCl, pH 6.8). Samples were heated for 5 min at 40°C and analyzed by SDS-PAGE and immunoblotting.

Identification and quantitation of proteins by mass spectrometry
The eluates from the GST-CPE 116-319 and the control GST pull-down were combined at a ratio of 1:1, resolved by SDS-PAGE (4%-18%) and stained with Coomassie Brilliant Blue G-250. Excised gel slices were washed with 50% (v/v) acetonitrile in 50 mM ammonium bicarbonate, shrunk by dehydration in acetonitrile, and dried in a vacuum centrifuge. The dried pieces were reswollen in 15 ml of 50 mM ammonium bicarbonate containing 60 ng trypsin (sequencing grade, Promega, Mannheim, Germany). After 16 h incubation at 37°C, 15 ml of 0.3% trifluoroacetic acid (TFA) in acetonitrile were added, and the separated supernatant was dried under vacuum. For nanoLC-MS/MS, the samples were dissolved in 6 ml of 0.1% (v/v) TFA, 5% (v/v) acetonitrile in water.
Tandem MS experiments were performed on a quadrupole orthogonal acceleration time-of-flight mass spec-trometer Q-Tof Ultima (Micromass, Manchester, UK), equipped with a liquid chromatography system (CapLC, Waters GmbH, Eschborn, Germany), as previously described [17]. In brief, LC-separations were performed on a capillary column (Atlantis dC 18 , 3 mm, 100 Å, 150 mm × 75 mm i.d., Waters) at an eluent flow rate of 200 nl/ min, using an acetonitrile gradient in 0.1% formic acid (v/ v). MS/MS data were acquired in a data-dependent mode, using MS survey scanning followed by MS/MS of the most abundant peak. Data analysis was performed with MassLynx version 4.0 software (Micromass-Waters).
The processed MS/MS spectra and MASCOT server (version 2.0, Matrix Science Ltd, London, UK) were used to search in-house against the SwissProt protein database. The maximum of two missed cleavages was allowed and the mass tolerance of precursor and sequence ions was set to 100 ppm and 0.1 Da, respectively. Acrylamide modification of cysteine, methionine oxidation, 13 C 6 -arginine and 13 C 6 -lysine were considered as possible modifica-Strategy for affinity purification using GST-CPE 116-319 as "bait" tions. A protein was accepted as identified if the total MASCOT score was greater than the significance threshold and at least two peptides appeared for the first time in the report and these peptides were first ranking. Protein identifications were performed at a < 1% false positive rate as established by decoy database search strategy.
Quantitation was based on calculations of isotope intensity ratios of at least two arginine-or lysine-containing tryptic peptides that were identified by MS/MS with a score above the MASCOT identity threshold. Additional criteria were that no interfering mass peaks were observed, that the peptide appeared for the first time in the report and that it was a first ranking peptide.

Results and Discussion
Tight junction proteins, their homophilic and heterophilic interactions and their regulation, are of outstanding importance for the function of many organs. The present study is devoted to the claudin protein family, as its members play a decisive role in providing paracellular tightness. Based on their affinity to CPE 116-319 , enrichment of different claudins from cell lysates has been accomplished. Co-enriched proteins have been identified and their potential interaction with claudins assessed.
Several epithelial cell lines were tested initially, and normal rat cholangiocytes were selected for subsequent experiments, as higher levels of CPE-binding claudins -3, -4 and -7 were detected by immunoblotting when compared to the human colon carcinoma cell line Caco-2. However, it should be mentioned that the MS identification of the proteins eluted from GST-CPE 116-319 after incubation with Caco-2 lysates indicated that claudin-2 and claudin-6 also accumulate in this fraction (cf. additional file 1, table_a.pdf -Identification of claudins in Caco-2 and NRC cells).
In preceding experiments, different protocols were tested with respect to the enrichment of claudins. A huge number of non-specifically bound proteins were identified by mass spectrometry in a protein fraction obtained Claudin enrichment in GST-CPE 116-319 eluate fraction by a co-immunoprecipitation experiment using a monoclonal claudin-3 antibody (data not shown). Moreover, the data indicated lower relative enrichment of claudin-3 than with the CPE-based approach and the lack of any claudin different from claudin-3. Thus, another advantage of enrichment using CPE is its high affinity to several members of the claudin family.
The scheme shown in Fig. 1 specifies the protocol applied for affinity chromatographic enrichment of claudins and subsequent MS identification of co-eluted proteins. As initial experiments revealed relatively strong non-specific binding to various carrier materials, a protein labeling strategy (SILAC) was employed in order to define proteins non-specifically bound to GST and/or matrix. Cell lysates containing labeled proteins were added to CPE 116-319loaded GST fusion protein, whereas the non-labeled protein fraction was added to the unloaded control (or vice versa in an inverse labeling experiment). In NRC cells, the degree of 13 C-Arg and 13 C-Lys labeling was 90% (which was corrected for in the values given below).
The experimental procedures were optimized with respect to the amount of claudins enriched in the eluate fractions. Different detergents were tested to improve the solubilization of claudins in the lysate and eluate fractions; Fig. 2 gives a comparison of the recovery of claudin-1 and claudin-3 in lysis buffers containing Triton X-100 and 3-[(3cholamidopropyl)dimethylammonio]-1-propanesulfonate (CHAPS). Triton X-100 was selected for further experiments, as higher yields of the claudins were found both in the lysate-and in the GST-CPE 116-319 eluate fraction. Corresponding experimental conditions have also been successfully used in a study on the self-association of tight junction proteins [18]. Immunoblots against claudins of the different flowthrough and eluate fractions clearly demonstrate the strong enrichment of claudins -1, -3, -4, -5 and -7 in the GST-CPE 116-319 eluate fraction (Fig. 3). The blots of the eluate fractions of the GST controls show a faint signal originating from claudin-1 and a stronger signal of claudin-4, possibly indicating weak association of these claudins with GST and/or matrix material.
The C-terminal part of CPE (amino acids 290-319) is sufficient for interactions with claudin-3 and claudin-4 [5]. Neither claudin-3 nor claudin-4 were detected by immu-  [22]); # , protein fragment noblotting in a control affinity enrichment applying a truncated (CPE 194-309 ) construct (cf. additional file 2, figure_a.pdf -Association of claudin -3 and claudin-4 with truncated CPE constructs). It has been reported [19] that tight junction strands gradually disintegrate and disappear from the cell surface after treatment with the toxin. However, despite its high affinity interaction with claudins, opening of the tight junctions by CPE treatment needs hours [20] which corresponds with our observation (data not shown) that a significantly lower amount of claudin-3 is precipited with CPE 116-319 when the construct is incubated with intact cell monolayers instead of lysates.
For MS analysis, the eluted GST and GST-CPE 116-319 fractions were mixed at a 1:1 ratio, separated by SDS-PAGE, and in-gel digested. For the identified proteins, the degree of binding specificity is given by the value R, defined here as the mean of the MS peak intensity ratios ( 13 C-labeled/ unlabeled or vice versa for inverse labeling) of peptides originating from CPE 116-319 -and control fraction. Table 1 lists the proteins identified in at least two independent experiments with R ³ 3 (additional information based on [21] and [22] is included). In accordance with the immunoblotting data, the highest ratios were obtained for the claudins. High association constants (K a > 5 × 10 7 M -1 ) with CPE have been published for claudin-3 and -4 [19], as well as for claudin-7 [4]. Claudin-1 and claudin-5 (expressed at lower levels in NRC cells) do not bind to CPE according to the literature [4,19]. The unambiguous identification of these claudins by immunoblotting (claudin-1 also by MS) could be explained either by binding to CPE 116-319 or by a linkage to other CPE-binding proteins.
There are indeed literature data reporting interaction of claudins -1 and -3 in co-expression studies [23,24], and the association of claudin-3 and claudin-5 was demonstrated in transfected cell lines and human airway epithelium [24]. Additional experiments performed in either claudin-3-or claudin-5-transfected HEK293 cells revealed an association of claudin-3, but not claudin-5, to CPE 116-In addition to heterologous claudin-claudin interactions, claudins associate with ZO-1 [25] and possibly tetraspanins [26]. However, neither of these proteins was found by MS (ZO-1 also immunoblot) in the GST-CPE 116-319 fraction. Except for the claudins, the vast majority of proteins found to be enriched in the GST-CPE 116-319 fraction are of nuclear or mitochondrial origin. Among the proteins identified to co-elute with the claudins, the most specific association with the CPE construct was observed for DNA-directed RNA polymerase II. Interestingly, at least one other protein was identified which can be bound to this polymerase (RNA polymerase IIassociated protein 2 [27]). However, the interaction of this nuclear protein with the CPE complex is, although specific, thought to be a false positive (not claudinrelated), since it is likely to be favored by the disintegration of cellular structures given in cell lysates. This type of interaction must also be assumed for other proteins, e.g., mitochondrial proteins, which are strictly separated in vivo from the plasma membrane. The function(s) known so far for Ran-binding protein 2 (RBP2) and RanGTPase-activating protein 1 (RanGAP1) are related to nucleocytoplasmic transport, also suggesting a false positive interaction. Both RBP2 and RanGAP1 are substrates of post-translational modification by small ubiquitin-like modifier 1 (SUMO-1) [28], which was also identified. Moreover, RanGAP1 and SUMO-1 were found at an apparent molecular mass of 85 kDa in the same gel slice as the only proteins with R > 3, indicating that RanGAP1 is the protein sumoylated here.
The cytosolic localization of sphingosine kinase 2 could facilitate association with intracellular parts of claudins (although this protein has recently been found to localize Sphingosine kinase enrichment in GST-CPE 116-319 eluate frac-tion and sphingosine kinase immunoprecipitation to the nucleus too [29]). Moreover, a connection exists between sphingosine kinase 2 and the junctional complex: sphingosine 1-phosphate (the product of sphingosine phosphorylation by sphingosine kinases) induces in endothelial cells the formation of ZO-1 complexes with cortactin and a-catenin, which regulate chemotactic response and barrier integrity, respectively [30]. Additional experiments have therefore been performed to check for a possible interaction of claudins with sphingosine kinase 2. The immunoblot given in Fig. 4 (upper panel) demonstrates enrichment of SphK2 in the GST-CPE 116-319 fraction, although a lower amount of the protein was also found in the respective GST control. Coimmunoprecipation using a monoclonal antibody directed against SphK2 revealed strong enrichment of SphK2 in the eluate fraction (Fig. 4, middle panel). However, co-immunoprecipitation of claudins with SphK2 was not detectable, neither for claudin-3 (Fig. 4, lower panel) nor for claudins -1, -4 or -7 (data not shown). Moreover, immunofluorescence staining did not provide any evidence for co-localization of SphK2 and claudins in NRC cells (data not shown).

Conclusion
The present study introduces a new approach directed at enriching the CPE-binding claudins. The proteins interacting specifically with GST-CPE 116-319 include not only claudins -3, -4 and -7 (which are known to bind to the toxin), but also claudins -1 and -5, pointing at a possible indirect heterologous association of these tight junction proteins. Different proteins (and potentially also protein complexes) were shown to co-elute from the GST-CPE 116-319 complex, but there is no indication that any of these proteins plays a role with respect to tight junctions or cellcell contacts. Simultaneous enrichment of numerous claudins may nevertheless provide a useful tool applicable to further investigations devoted to these tight junctional proteins.