Interaction of the Clostridium difficile Binary Toxin CDT and Its Host Cell Receptor, Lipolysis-stimulated Lipoprotein Receptor (LSR)*

Background: Clostridium difficile binary toxin CDT enters host cells via its receptor LSR. Results: Regions responsible for the CDT-LSR interaction were identified. Conclusion: An immunoglobulin-like domain of LSR and amino acids 757–866 of the binding component of CDT determine the toxin-receptor interaction. Significance: Knowledge of the toxin-receptor interaction is key to the development of anti-toxin strategies. CDT (Clostridium difficile transferase) is a binary, actin ADP-ribosylating toxin frequently associated with hypervirulent strains of the human enteric pathogen C. difficile, the most serious cause of antibiotic-associated diarrhea and pseudomembranous colitis. CDT leads to the collapse of the actin cytoskeleton and, eventually, to cell death. Low doses of CDT result in the formation of microtubule-based protrusions on the cell surface that increase the adherence and colonization of C. difficile. The lipolysis-stimulated lipoprotein receptor (LSR) is the host cell receptor for CDT, and our aim was to gain a deeper insight into the interplay between both proteins. We show that CDT interacts with the extracellular, Ig-like domain of LSR with an affinity in the nanomolar range. We identified LSR splice variants in the colon carcinoma cell line HCT116 and disrupted the LSR gene in these cells by applying the CRISPR-Cas9 technology. LSR truncations ectopically expressed in LSR knock-out cells indicated that intracellular parts of LSR are not essential for plasma membrane targeting of the receptor and cellular uptake of CDT. By generating a series of N- and C-terminal truncations of the binding component of CDT (CDTb), we found that amino acids 757–866 of CDTb are sufficient for binding to LSR. With a transposon-based, random mutagenesis approach, we identified potential LSR-interacting epitopes in CDTb. This study increases our understanding about the interaction between CDT and its receptor LSR, which is key to the development of anti-toxin strategies for preventing cell entry of the toxin.

domembranous colitis (1)(2)(3). In recent years, hypervirulent strains of the pathogen have caused epidemics in developed countries that were associated with more severe disease and increased death rates (4 -7). One hallmark of hypervirulent C. difficile strains, such as the BI/Nap1/027, is the production of a third toxin, the C. difficile transferase (CDT), 3 in addition to the Rho-glucosylating toxins A and B (8,9).
CDT belongs to the family of clostridial ADP-ribosylating toxins, exhibiting the highest similarity to members of the subfamily of iota-like toxins, which includes the eponym Clostridium perfringens iota toxin and the Clostridium spiroforme toxin (9,10). CDT consists of two components: a biologically active enzyme component (CDTa) and a separate binding component (CDTb). At first, CDTb interacts with a host receptor on the cell surface. Before or after this step, CDTb undergoes proteolytic activation to form a heptameric prepore already in solution or at the cell surface, respectively. After binding of CDTa to the prepore-receptor complex, the toxin reaches endosomal compartments via receptor-mediated endocytosis. Acidification of the endosomal lumen triggers conformational changes within the prepore, resulting in membrane insertion and subsequent pore formation (11)(12)(13)(14). By the guidance of cytosolic chaperones, CDTa then translocates through the pore into the cytosol, where it ADP-ribosylates G-actin at arginine 177, thereby causing actin depolymerization and, eventually, cell death (9,(15)(16)(17)(18). However, at low doses of the toxin, destruction of the cortical actin results in the formation of microtubule-based protrusions on epithelial cells that increase the adherence and colonization of C. difficile (19).
Using a haploid genetic screen, we recently identified the lipolysis-stimulated lipoprotein receptor (LSR) as the host cell receptor for all members of the iota toxin family, including CDT (20 -22). LSR is a type I, single-pass, transmembrane protein, featuring an Ig-like V-type domain in the extracellular part of the protein. The receptor is mainly expressed in the liver, but also in the intestine and various other tissues (23)(24)(25). Initially, LSR was identified as a hepatic receptor for triglyceride-rich lipoproteins (26,27). More recently, LSR was found to act in the recruitment of tricellulin to tricellular contacts that are important for the integrity of epithelial barriers (28,29).
In the present study, we obtained binding kinetics of CDT and LSR and identified regions in both proteins that determine their interaction. We show with recombinant proteins that the extracellular, Ig-like domain of LSR provides the platform for the binding of CDT. Furthermore, we generated an HCT116 LSR knock-out cell line and ectopically expressed LSR truncations to clarify whether intracellular parts of LSR are required for plasma membrane targeting of the receptor and endocytic uptake of CDT. The receptor-binding domain (RBD) of CDT has been only roughly localized at the C terminus of CDTb. To narrow down this region, we generated a series of N-and C-terminal truncations of the RBD of CDTb and tested them for binding to LSR. Additionally, we applied transposon-based, random mutagenesis to the RBD of CDTb to identify epitopes potentially involved in the interaction with LSR.

Experimental Procedures
Generation, Transfection, and Cultivation of Mammalian Cells-All mammalian cells used in this study were grown in DMEM (12 mM L-glutamine) supplemented with 10% (v/v) FCS, 1% (v/v) nonessential amino acids, 1% (v/v) sodium pyruvate, and 1% (v/v) penicillin/streptomycin and incubated at 37°C with 5% (v/v) CO 2 under humidified conditions. Transfections were performed either with polyethylenimine or with peqFECT transfection reagent (PeqLab) according to the manufacturer's protocol. If necessary, cells were co-transfected with an EGFP-expressing plasmid (pEGFP-N1; Clontech) to visualize the transfected cells with fluorescence microscopy.
HCT116 LSR knock-out cells were generated via the CRISPR (clustered regularly interspaced short palindromic repeats)-Cas9 technology and by following the recently established protocol from the Church laboratory (30). Briefly, a guide RNA expression fragment was ordered as a gBlock (Integrated DNA Technologies, Leuven, Belgium) that consisted of an U6 promoter followed by a 20-bp protospacer, a protospacer adjacent motif sequence, and a scaffold and terminator sequence. The protospacer sequence (5Ј-GGACAGCGTGCGCACCGTCA-3Ј) was complementary to a sequence in exon 2 of the LSR gene. The gBlock was then inserted into the pCR-Blunt II-TOPO vector and the resulting plasmid transfected into HCT116 cells together with the human codon-optimized Cas9 expression plasmid pcDNA3.3-TOPO/hCas9 (Addgene plasmid 41815). Transfected cells were first grown under antibiotic selection pressure with neomycin and then treated with CDT for the selection of toxin-resistant HCT116 LSR knock-out cells. Finally, a single HCT116 LSR knock-out clone was isolated via limiting dilution cloning and named HCT116 ⌬LSR . A base pair insertion in exon 2 of the LSR gene of HCT116 ⌬LSR cells, resulting in a frameshift mutation, could be confirmed by PCR and Sanger sequencing of the genomic region flanking the mutation site. H1-HeLa(ϩLSR) cells were obtained by transducing H1-HeLa cells with a FLAG-LSR-expressing retrovirus as described previously (20), and the same procedure was applied for the transduction of HCT116 ⌬LSR cells with a FLAG-or FLAG-LSR-expressing retrovirus to generate HCT116 ⌬LSR (ϩFLAG) and HCT116 ⌬LSR (ϩFLAG-LSR) cells.
Cloning, Mutagenesis, Recombinant Proteins, and Source of Toxins-For the recombinant expression and purification of the Ig-like domain of human LSR, a DNA fragment encoding amino acids 39 -192 of the LSR protein (referring to the LSR variant with the accession number AAH04381.2) was amplified by PCR with oligonucleotides generating a 5Ј-NcoI and a 3Ј-HindIII restriction site and by using the previously described plasmid pMXs-IRES-Blasticidin/FLAG-LSR as template (20). The PCR product was then ligated into an NcoI/HindIIIopened pET28a vector to generate pET28a/LSR Ig -His 6 . The resulting plasmid was then transformed into Escherichia coli strain BL21 (DE3). Transformants were grown in LB medium at 37°C until an A 600 of 0.4 -0.6 was reached, and then protein expression was induced by the addition of 0.2 mM isopropyl-␤-D-thiogalactopyranoside and overnight incubation at 21°C. The next day, the cells were harvested by centrifugation (at 5,000 ϫ g for 20 min at 4°C) and resuspended in lysis buffer (20 mM Tris-HCl, pH 8, 300 mM NaCl, and 0.5 mM EDTA) complemented with 1 mM PMSF and 5 g/ml DNase, prior to lysis of the cells by the use of a microfluidizer at 15,000 p.s.i. Following centrifugation (at 14,000 ϫ g for 30 min at 4°C) of the cell lysate, the supernatant was discarded, and the pellet further was processed for purification of LSR Ig -His 6 from inclusion bodies. To obtain purified inclusion bodies, the pellet was resuspended in washing buffer (lysis buffer including 1% (w/v) Triton X-100 and 10 mM DTT) and then briefly sonicated (three times for 30 s at 80% power, 70% duty cycle; Bandelin Sonopuls HD 60), prior to centrifugation (at 14,000 ϫ g for 20 min at 4°C) to precipitate again the inclusion bodies. This washing step was performed twice and then once again with washing buffer without Triton X-100. Next, the inclusion bodies were resuspended in solubilization buffer (20 mM Tris-HCl, pH 8, 500 mM NaCl, and 5 mM DTT) complemented with 6 M urea, sonicated (three times for 30 s at 80% power, 70% duty cycle) and incubated overnight at 4°C under gentle rotation. Nonsolubilized material was removed by centrifugation (at 14,000 ϫ g for 30 min at 4°C), and the supernatant was dialyzed at 4°C overnight first against the 75-fold volume of refolding buffer (20 mM Tris-HCl, pH 8, 150 mM NaCl, 0.3 M L-arginine, 10 mM reduced glutathione, and 5 mM oxidized glutathione) and a second time against the 125fold volume of storage buffer (150 mM Tris-HCl, pH 8, and 20 mM NaCl).
The plasmid pEGFP-N1/LSR was used for the ectopic expression of LSR-EGFP in mammalian cells and is described elsewhere (31). The pEGFP-N1 vector backbone was also used for the ectopic expression of LSR 1-240 -EGFP and LSR ⌬CR -EGFP. Cloning of these two constructs was performed as follows: a DNA fragment encoding LSR 1-240 was generated by PCR amplification of the respective region, whereas LSR with deletion of the cysteine-rich region was generated via overlap extension PCR; in both cases primers were used that generated a 5Ј-NheI and a 3Ј-HindIII restriction site for ligation into a NheI/HindIII-opened pEGFP-N1 vector to finally obtain pEGFP-N1/LSR 1-240 and pEGFP-N1/LSR ⌬CR , respectively. Eventually, all LSR constructs were subcloned with identical restriction sites into the ptdTomato-N1 backbone for ectopic expression of LSR-, LSR  , and LSR ⌬CR -Tomato in mammalian cells.
Plasmids for the expression of mLSR, mILDR1, and mILDR2 in mammalian cells were kindly provided by Mikio Furuse (Kobe University, Japan) and are described elsewhere (29,32). The plasmid pcDNA3-dsRed-ER (contains the N-terminal ER targeting sequence of calreticulin and a C-terminal KDEL ER retention sequence) was kindly shared by Llewelyn Roderick (Cambridge, UK) and was used for ER co-localization studies.
CDTa and the precursor form of CDTb (pCDTb) were purified as recombinant proteins from the expression host Bacillus megaterium with protocols described earlier (33). CDTb was obtained by proteolytic activation of pCDTb (11,34). If not otherwise stated, 1 nM CDTa together with 0.2 nM CDTb were added to cultured cells for intoxication experiments.
The cloning, expression, and purification of the RBD of CDTb as a GST fusion protein (GST-RBD) has been described elsewhere (31). N-and C-terminal truncations of RBD were expressed and purified from E. coli strain BL21 (DE3) as GST fusion proteins in the same manner as GST-RBD. The expression plasmids for the RBD truncations were generated by amplifying the corresponding sequence of the RBD by PCR with primers generating a 5Ј-EcoRI and a 3Ј-XhoI restriction site and by using pGEX-4T3/RBD (31) as template. The PCR products were then ligated into an EcoRI/XhoI-opened pGEX-4T3 vector. The pGEX-4T3/RBD plasmid was also used as template to generate transposon-based RBD mutants with the MGS TM mutation generation system (Thermo Scientific) following the manufacturer's protocol.
Analysis of Cell Surface-binding of DyLight-labeled Proteins by Flow Cytometry-DyLight 488 N-hydroxysulfosuccinimide ester (Thermo Scientific) was coupled to purified proteins in PBS including 10% (w/v) glycerol following the manufacturer's instructions. Micro Bio-Spin 6 columns (Bio-Rad) were used to remove excess dye after the labeling reaction.
To analyze cell surface-binding of DyLight-labeled proteins, cells from the mammalian cell line of interest were first detached from culture plates with PBS complemented with 10 mM EDTA, then washed twice with PBS, and kept on ice. Typically, 3 ϫ 10 5 cells were incubated with the indicated amounts of DyLight-labeled proteins for 10 min in 0.1 ml of DMEM on ice. For competition studies, nonlabeled proteins were preincubated with cells for 10 min on ice, prior to the addition of the DyLight-labeled protein. The cells were then washed twice with PBS and subjected to flow cytometric analysis using the BD FACSCalibur platform. Cell surface-bound fluorescence was detected with an argon ion laser (488 nm) and a 530-nm bandpass filter (FITC). For multicolor FACS measurements, the BD FACSCANTO II platform was used. Flowing software (version 2.5.0) was used for the computational analysis of the FACS measurements.
Surface Plasmon Resonance Measurements-Dynamic interactions of GST-RBD with the immunoglobulin-like domain of LSR were analyzed using a Biacore X100 system (Biacore). LSR Ig -His 6 (25 g/ml in 10 mM sodium acetate, pH 4.0) was amine-coupled on a CM5 sensor chip according to the manufacturer's instructions, achieving ϳ3,000 response units. Sensograms were performed in binding buffer (10 mM HEPES, 150 mM NaCl, and 0.05% Tween 20, pH 7.4) at 25°C with a flow rate of 30 l/min. Up to 5 increasing concentrations of putative protein ligand were injected per cycle for 120 s, and dissociation was allowed for 600 s. The chip surface was regenerated at the beginning of each binding cycle by sequential flux of two buffers, containing (a) 1 M NaCl and 10 mM sodium acetate (pH 4.0) and (b) 1 M NaCl, and 10 mM NaOH, with each flux lasting 90 s. Kinetic constants were calculated using the Biacore X100 Evaluation Software (version 2.0).
Microscopy-Morphological changes of cultured cells after intoxication with CDT were analyzed directly in wells by using an inverted light microscope (Axiovert 25; Carl Zeiss). Images were obtained with an inbuilt digital camera (AxioCam HRc). Fluorescence microscopy was performed with an inverted stage manual fluorescent microscope (Axiovert 35 M; Carl Zeiss) equipped with standard filters and an AxioCam HRm (Carl Zeiss) digital camera. Confocal fluorescence microscopy was performed with an inverted microscope (Axiovert 200M; Carl Zeiss) equipped with Plan-Apochromat objectives, a spinning disk head (Yokogawa) with emission filters, and solid state laser lines (488 and 561 nm). Confocal fluorescence and differential interference contrast images were collected with a CoolSNAP-HQ2 digital camera (Roper Scientific) followed by processing with Metamorph imaging software (Universal Imaging). For actin staining with TRITC-conjugated phalloidin, cells were first washed with PBS, then fixed with 4% formaldehyde in PBS, washed again, and permeabilized with 0.15% Triton X-100 in PBS. Following washing of the permeabilized cells with PBS, actin was stained for 1 h at room temperature with TRITCconjugated phalloidin (diluted 1:200 from a 50 g/ml stock solution; Invitrogen). Subsequently, cells were washed with PBS, incubated for 10 min with 0.05% Tween 20 in PBS, washed successively with 70% EtOH and 100% EtOH, and finally embedded in Mowiol supplemented with DABCO (Sigma).

Preparation of Whole Cell Lysates and Cell Membrane
Fractions-Whole cell lysates were prepared by washing cells directly in wells with ice-cold PBS, followed by lysis on ice in radioimmune precipitation assay buffer (50 mM Tris-HCl, pH 7.5, 1% (w/v) Triton X-100, 1% (w/v) glycerol, 0.5% sodium deoxycholate, 0.1% sodium dodecylsulfate, 137 mM sodium chloride, 1 mM sodium orthovanadate, and 0.5 mM EDTA) complemented with Complete TM protease inhibitor mixture. Cell debris was removed by centrifugation (at 21,000 ϫ g for 10 min at 4°C), and the protein concentration of the supernatant was measured with a bicinchoninic acid protein assay kit (Uptima, Montluçon, France).
For the preparation of cell membrane fractions, the cells were first washed in wells with ice-cold PBS and then scraped off into PBS including Complete TM protease inhibitor mixture. Following sonication (three times for 30 s at 80% power, 70% duty cycle) to disrupt the cell membrane, intact cells, nuclei, and cell debris were removed by centrifugation (at 2,000 ϫ g for 5 min at 4°C), and the supernatant was centrifuged again (at 21,000 ϫ g for 2 h at 4°C) to obtain a pellet with the crude membrane fraction.
GST-based Pulldown Assay-Glutathione-Sepharose 4B beads (20 l per sample) were washed two times with GST-fish buffer (50 mM Tris-HCl, pH 7.5, 100 mM NaCl, 2 mM MgCl 2 , 10% (w/v) glycerol, 1% (w/v) Nonidet P-40), before incubation for 60 min at 4°C in 500 l GST-fish buffer with 2 to 6 M GST, GST-RBD, or the GST-tagged N-and C-terminal RBD truncations, respectively. Beads were then washed two times with GST-fish buffer, incubated with GST-fish buffer containing 1% BSA, washed again with GST-fish buffer, and incubated with 20 g of LSR Ig -His 6 in 500 l GST-fish buffer. The beads were then washed five times with GST-fish buffer. Bound proteins were eluted by the addition of preheated (95°C) Laemmli sample buffer and incubation for 10 min. Samples were then subjected to SDS-PAGE, followed by immunoblotting to detect the bait and prey proteins with specific antibodies.
Generation of a HCT116 cDNA Library for the Identification of LSR Isoforms-Total RNA was isolated from HCT116 cells with a RNeasy mini kit (Qiagen) followed by cDNA synthesis with a first strand cDNA synthesis kit (Thermo Scientific) using an oligo(dT) 18 primer according to the manufacturers' instructions. LSR-specific cDNA transcripts were amplified by PCR using the forward primers 5Ј-GCAACAGGACGGACTTG-GAGTAG-3Ј (long N terminus) or 5Ј-ATATCCATGGCGCT-GTTGGCCGGCGG-3Ј (short N terminus) and the reverse primer 5Ј-TCAGACGACTAAACTTTCCCGACTCAGGGC-C-3Ј. PCR products were finally cloned into the pCR-Blunt II-TOPO vector and analyzed by Sanger sequencing. Optionally, LSR isoforms were subcloned into the mammalian expression vectors ptdTomato-N1 or pEGFP-N1.

CDT Interacts with the Extracellular, Ig-like Domain of LSR-
According to the protein database UniProt, human LSR (accession number Q86X29) exists in six isoforms varying in length because of alternative splicing or differential translation initiation. Isoform 1 of LSR represents the canonical, nonspliced version of the protein with the length of 649 amino acids. The other isoforms are predicted to lack either small extracellular and/or intracellular protein segments or even the putative transmembrane region and a cysteine-rich region.
In a previous study, we cloned an LSR variant into a mammalian expression vector that derived from a human pancreatic adenocarcinoma cDNA library (20). Compared with isoform 1, this LSR variant (GenBank TM accession number AAH04381.2) lacks the first 48 amino acids of the N terminus and a short segment (amino acids 240 -258) between the Ig-like domain and the putative transmembrane region (Fig. 1, A and B). However, as known from our previous study, the N-terminally shortened LSR variant efficiently promotes cell entry of CDT. Unless otherwise stated, this LSR variant is also used in the current study for functional analyses. Intriguingly, isoform 6 of human LSR (UniProt accession number Q86X29-6), as well as all isoforms of murine LSR (UniProt accession number Q99KG5), also lack the N-terminal extension upstream of the Ig-like domain, which is present in human LSR (Fig. 1B).
We therefore wanted to test whether murine LSR can promote cell entry of CDT into HeLa cells that do not express detectable amounts of LSR and are highly resistant to the toxin. To this end, murine LSR (mLSR) was ectopically expressed in HeLa cells that were co-transfected with an EGFP-expressing plasmid to visualize the transfected cells. In addition, we also ectopically expressed the murine orthologs of ILDR1 (mILDR1) and ILDR2 (mILDR2), two members of the angulin family of tricellular tight junction-associated proteins with homology to LSR in the Ig-like domain (32). The addition of CDT caused cell rounding because of the destruction of the actin cytoskeleton only in mLSR-expressing HeLa cells, but not in cells that expressed either mILDR1 or mILDR2 or that expressed EGFP only (Fig. 1C). In addition, a FACS-based analysis of the binding of fluorescence-labeled CDTb to the cell surface of HeLa cells was performed. As expected, upon addition of labeled CDTb to HeLa cell suspensions, a strong increase in cell surface-bound fluorescence over background was only observed in a fraction of HeLa cells that expressed the mLSR protein (Fig. 1D). These results confirmed that the N-terminal extension found upstream of the Ig-like domain in human LSR isoforms is dispensable for binding and uptake of CDT and that the LSRrelated proteins ILDR1 and ILDR2 do not serve as cell entry receptors for the toxin.
Therefore, we aimed to investigate whether CDT interacts with the Ig-like domain of LSR. To this end, we recombinantly expressed a DNA fragment encoding the Ig-like domain of human LSR as a C-terminally His-tagged protein (denoted as LSR Ig -His 6 ) with the pET expression system in E. coli. The protein was mainly found in the pellet fraction after cell lysis and could be purified from inclusion bodies by urea-induced denaturation, followed by dialysis to remove the denaturation component for the refolding of the protein (Fig. 2, A and B). We then used LSR Ig -His 6 as prey protein in a GST pulldown assay and a GST fusion of the RBD from CDTb (amino acids 677-876) as bait. LSR-dependent cell surface-binding of GST-RBD was shown previously (31). As shown in Fig. 2C, the LSR Ig -His 6 protein co-precipitated with GST-RBD-loaded glutathione beads but not with beads that were loaded with GST only. To confirm the direct interaction between the immunoglobulin-like domain of LSR and the receptor-binding domain of CDTb, we performed surface plasmon resonance spectroscopy measurements. The LSR Ig -His 6 protein was immobilized on a sensor chip, and the binding kinetics of GST or GST-RBD were measured. A biphasic association and dissociation profile was only obvious for GST-RBD but not for the GST protein (Fig.  2D). Thereby, we obtained a dissociation constant (K D ) of ϳ110 nM for the interaction between GST-RBD and LSR Ig -His 6 .
Next, we wanted to determine the binding kinetics of the receptor-binding domain of CDTb when it binds to LSR in its native environment in the plasma membrane. For this purpose, we chose H1-HeLa(ϩLSR) cells that ectopically express the LSR variant AAH04381.2 and two human colon carcinomaderived cell lines (CaCo-2, HCT116), because CDT is a toxin from an intestinal pathogen. Cell suspensions were incubated with increasing concentrations of fluorescence-labeled GST-RBD at 4°C to prevent endocytic uptake, followed by FACS analysis to measure the cell surface-bound fluorescence of the cells. GST-RBD revealed saturable binding to all three cell lines (Fig. 3), thereby allowing the calculation of the following K D values: 40.4 Ϯ 9.3 nM for H1-HeLa(ϩLSR), 42.6 Ϯ 13.9 nM for CaCo-2, and 32.6 Ϯ 4.7 nM for HCT116 cells. As a comparison, for the binding of CDTb to HCT116 cells, we obtained by the same approach a K D value of 47.6 Ϯ 2.8 nM. These K D values for the interaction of GST-RBD with LSR at the cell surface were in good agreement with the binding constant that was obtained by surface plasmon resonance measurements for the interaction between GST-RBD and the immunoglobulin-like domain of LSR.
Generation and Evaluation of an HCT116 LSR Knock-out Cell Line for Functional Studies-We next decided to generate a defined LSR knock-out cell line to be able to ectopically express various LSR mutants for functional studies. We chose the human, colon carcinoma-derived cell line HCT116 for this purpose and applied the CRISPR-Cas9 technology to disrupt the LSR gene. Hence, we were able to isolate an HCT116 clone with a base pair insertion between positions 518 and 519 of the LSR coding region (termed HCT116 ⌬LSR in this study), which leads to a frameshift mutation in exon 2 and to incomplete translation of the LSR protein (Fig. 4A). Immunoblot analysis with an LSR-specific antibody confirmed that LSR is not expressed in HCT116 ⌬LSR cells (Fig. 4B). We then transduced the HCT116 ⌬LSR cells with a FLAG-or FLAG-LSR-encoding retrovirus for functional reconstitution studies. The ectopic expression of FLAG-LSR could be confirmed in FLAG-LSRtransduced HCT116 ⌬LSR cells by an anti-LSR immunoblot analysis of the membrane fraction (Fig. 4C). HCT116 wild-type and LSR knock-out cells, as well as FLAG-and FLAG-LSRtransduced HCT116 ⌬LSR cells, were incubated with CDT for 24 h, followed by microscopic analysis of the cell morphology. As expected, HCT116 ⌬LSR and FLAG-transduced HCT116 ⌬LSR cells remained fully resistant toward CDT, and toxin-induced cell rounding was obvious only in wild-type HCT116 cells and FLAG-LSR-expressing HCT116 ⌬LSR cells (Fig. 4D). FACS analysis with DyLight488-labeled GST-RBD additionally proved the cell surface expression of LSR only in wild-type HCT116 cells and FLAG-LSR-expressing HCT116 ⌬LSR cells, but not in non-or FLAG-only-transduced HCT116 ⌬LSR cells (Fig. 4E). These results confirmed that the HCT116 ⌬LSR cells are suitable for the task to express different LSR mutants or isoforms for functional studies.
Identification of Multiple Isoforms of LSR in HCT116 Cells-We were also interested in the identification of LSR isoforms that are expressed in HCT116 cells. To this end we isolated the total RNA of HCT116 cells for cDNA synthesis using an oligo(dT) 18 primer. The cDNA mixture was then used to amplify possible LSR isoforms via PCR with LSR-specific oligonucleotides. PCR products were finally cloned into the pCR-Blunt II-TOPO vector and analyzed by Sanger sequencing. Surprisingly, we never obtained any PCR product when the LSR reverse primer was used in combination with an LSR forward primer that was complementary to the starting sequence of the canonical LSR isoform (isoform 1). However, PCR products were obtained when the LSR reverse primer was combined with an LSR forward primer that was complementary to the starting sequence of the shortened LSR isoforms (e.g. LSR isoform 6 and LSR variant AAH04381.2). We found at least four different isoforms in HCT116 cells, all with potentially shortened N terminus. Three isoforms represent the same splice variants as LSR isoforms 3, 5, and 6 (UniProt accession numbers Q86X29-3, Q86X29-5, and Q86X29-6). Interestingly, we found a previously unknown splice variant in HCT116 cells that is not annotated in the UniProt database. This LSR splice variant is lacking a large central part of the protein. We designate this novel LSR splice variant as LSR isoform 7. The LSR isoforms that we iden-  tified in HCT116 cells are schematically represented in Fig. 5A, in direct comparison to already known LSR isoforms from the UniProt database. It is also of note that we found single amino acid deletions or insertions in the cytoplasmic tail of the LSR isoforms from HCT116 cells that differ from the canonical sequence ( Fig. 5A; indicated with Greek symbols).
Obviously, some LSR isoforms found in HCT116 cells lack the putative transmembrane helix (TH) and the cysteine-rich region (CR) in the central part of the protein, and we wanted to clarify whether this part of LSR is required for integration of the protein into the plasma membrane. To this end, isoform 3-like (with TH/CR) and isoform 5-like (without TH/CR) LSR proteins were co-expressed ectopically in HCT116 ⌬LSR cells as C-terminally Tomato-and EGFP-fused proteins, respectively, prior to confocal fluorescence microscopy. Isoform 3-like LSR clearly localized to the cell periphery of HCT116 ⌬LSR cells, whereas in contrast, isoform 5-like LSR exhibited an intracellular staining pattern (Fig. 5B). Interestingly, isoform 5-like LSR clearly co-localized with a co-expressed endoplasmic reticulum (ER) marker protein (cal-dsRed2) (Fig. 5C). These observations point to the fact that the putative transmembrane helix and cysteine-rich region are required for targeting and insertion of LSR into the plasma membrane. LSR isoforms lacking the transmembrane helix and cysteine-rich region most likely represent ER-resident proteins.
The Cytoplasmic Tail and the Cysteine-rich Region of LSR Are Not Essential for Targeting of the Receptor to the Plasma Membrane and for Cell Entry of CDT-We next took advantage of the HCT116 ⌬LSR cells to transiently express EGFP or Tomato fusion proteins of full-length LSR (LSR-EGFP/Tomato) and LSR truncations that lacked either the cytoplasmic tail (LSR 1-240 -EGFP/Tomato) or the cysteine-rich region (LSR ⌬CR -EGFP/Tomato) (Fig. 6A). Following the transfection of the LSR knockout cells with the respective plasmids, we first aimed to validate whether the LSR truncations are expressed and properly targeted to the plasma membrane. As shown in Fig. 6B, 1 day after transfection, expression of LSR-EGFP, LSR 1-240 -EGFP, and LSR ⌬CR -EGFP could be confirmed in HCT116 ⌬LSR cells by performing an anti-EGFP immunoblot analysis with the solubilized membrane fractions. As expected, no protein could be detected by the anti-EGFP antibody in membrane fractions of nontransfected HCT116 and HCT116 ⌬LSR cells. Transient expression of LSR-EGFP and LSR ⌬CR -EGFP in HCT116 ⌬LSR cells, as well as native expression of LSR in wild-type HCT116 cells, could be further confirmed with an anti-LSR antibody that recognizes a region at the C-terminal end of LSR. The anti-LSR antibody did not recognize LSR 1-240 -EGFP, because the antibody recognizes an epitope at the C-terminal end of LSR, which is missing in this truncated version of LSR (Fig. 6B).
To further validate that the LSR truncations are properly targeted to the plasma membrane with extracellular exposure of the Ig-like domain, a FACS-based analysis of the binding of fluorescence-labeled GST-RBD to the cell surface was performed. For this purpose, the HCT116 ⌬LSR cells were transiently transfected with plasmids encoding the Tomato fusion proteins of LSR and its truncated variants, prior to incubation of the cells with green fluorescence-labeled GST-RBD. Following a gating procedure to discriminate between the nontransfected (region 1; R1) and the transfected cell fraction expressing the Tomato constructs (region 2; R2), binding of green fluorescence-labeled GST-RBD to the cell surface of the R1 or the R2 fraction was analyzed in two-dimensional dot plots (green fluorescence versus side scatter). As expected, the nontransfected fraction of HCT116 ⌬LSR cells (R1) did not show any binding of green-labeled GST-RBD. However, an increase of cell surfacebound green fluorescence was observed in the R2 fractions, representing the LSR-Tomato-, LSR 1-240 -Tomato-, and LSR ⌬CR -Tomato-expressing HCT116 ⌬LSR cells (Fig. 6C). These results indicated that full-length and both truncated LSR variants were properly targeted to the plasma membrane and that the Ig-like domain of LSR variants is correctly folded and exposed at the cell surface to allow the binding of the receptorbinding domain of CDTb. The data allow the conclusion that the cysteine-rich region and the cytoplasmic tail of LSR are not required for plasma membrane integration of the receptor. Greek symbols indicate amino acid insertions or deletions: ␣, deletion of glutamate 386; ␤, insertion of an arginine residue between amino acids 519 and 520; ␥, deletion of histidine 607; ␦, deletion of lysine 637 (amino acid numbering refers to LSR Q86X29 isoform 1). B, isoform 3-and isoform 5-like LSR variants obtained from a HCT116 cDNA library were co-expressed as Tomato-(LSR isoform 3-like -Tomato) or EGFP-fused (LSR isoform 5-like -EGFP) proteins in HCT116 LSR knock-out cells, prior to analysis of their subcellular localization via confocal fluorescence microscopy. Fluorescence images were obtained 48 h after co-transfection of the cells with the corresponding plasmids. C, LSR isoform 5-like -EGFP shown in B was co-expressed in HCT116 LSR knock-out cells together with cal-dsRed2, an endoplasmic reticulum marker. Confocal images were obtained 2 days after transfection of the cells with the respective plasmids. Scale bar, 10 m. mag., magnified area.
We then wanted to analyze whether the cytoplasmic tail or the cysteine-rich region of LSR is essential for endocytic uptake of CDT. For that purpose, HCT116 ⌬LSR cells were transfected with plasmids for the expression of LSR-EGFP, LSR 1-240 -EGFP, or LSR ⌬CR -EGFP followed by incubation with CDT for 3 h at 37°C. TRITC-phalloidin staining and confocal fluorescence microscopy was then applied to visualize the toxin-induced disruption of the actin cytoskeleton. As shown in Fig. 7, LSR-EGFP-expressing cells were susceptible to CDT, because the actin cytoskeleton in these cells could not be stained with TRITC-phalloidin after incubation with the toxin. Moreover, toxin-induced cell rounding was clearly recognizable only in LSR-EGFP-expressing cells (Fig. 7, differential interference contrast image). As expected, nontransfected HCT116 ⌬LSR cells (cells without green fluorescence) remained resistant to the toxin (Fig. 7). However, CDT-induced depolymerization of the actin cytoskeleton was also observed in LSR 1-240 -EGFPand LSR ⌬CR -EGFP-expressing cells, indicating that also in these cells CDT was endocytosed and active (Fig. 7). Taken together, our findings indicate that binding and uptake of CDT does not essentially require the cytoplasmic tail or the cysteinerich region of LSR.
Amino Acids 757-866 of CDTb Are Sufficient for the Interaction with the Ig-like Domain of LSR-Marvaud et al. (35) mapped the receptor-binding domain of the iota toxin to the C terminus of its binding component (Ib). CDT is a member of the iota-like toxin family, and we recently confirmed this finding with a GST fusion of the C terminus of CDTb (GST-RBD) that bound directly to LSR-expressing cells and competitively inhibited the binding of CDTb to the cell surface (31).
Our next aim was to narrow down the region within the RBD (amino acids 677-876) of CDTb that is sufficient for the interaction with the Ig-like domain of LSR. Accordingly, we generated a series of GST fusion proteins with N-or C-terminal truncations of RBD (Fig. 8A) and tested whether LSR Ig -His 6 co-precipitates with the GST fusion proteins in a pulldown approach. As shown in Fig. 8B, deletion of the N-terminal 80 amino acids (RBD/⌬80N; GST-RBD aa757-876 ) or the C-terminal 10 amino acids of RBD (RBD/⌬10C; GST-RBD aa677-866 ) did not abolish the binding of LSR Ig -His 6 (Fig. 8B, upper panel). However, LSR Ig -His 6 did not bind to RBD truncations that lacked 120 N-terminal (RBD/⌬120N; GST-RBD aa797-876 ) or 20 C-terminal amino acids (RBD/⌬20C; GST-RBD aa677-856 ), respectively (Fig. 8B, lower panel).
To substantiate these findings, we performed FACS competition experiments, in which fluorescence-labeled GST-RBD (RBD DL488 ) was incubated with H1-HeLa(ϩLSR) cell suspensions either alone or together with 1-, 10-, or 100-fold molar excess of nonlabeled N-or C-terminal truncations of RBD. Concomitantly, FACS analyses revealed that RBD/⌬80N and RBD/⌬10C (Fig. 8, C and D, white bars), but not RBD/⌬120N and RBD/⌬20C (Fig. 8, C and D, black bars), competitively inhibited the binding of fluorescence-labeled RBD to H1-HeLa(ϩLSR) cells. Notably, a 100-fold excess of RBD/⌬10C was necessary to strongly decrease the binding of RBD DL488 , indicating that the C-terminal 10 amino acids of RBD might enhance the interaction with the toxin but are not essential. From these results, it can be concluded that a relatively small region between amino acids 757 and 866 of CDTb is already sufficient for the interaction with LSR. It seems that CDTb 757-866 represents the precise receptor-binding domain of the toxin.
To identify epitopes within the receptor-binding domain of CDTb that are crucial for the interaction with LSR, we carried out a transposon-based mutagenesis approach with a plasmid encoding GST-RBD. We obtained eight independent GST-RBD mutants (T1-T8), each containing a 5-amino acid in-frame insertion at different positions within the amino acid range 757-876 of the RBD (Fig. 9, A and B). To test whether the 5-amino acid linker disturbs an important epitope required for the interaction with LSR, we again performed a FACS competition experiment as described above. Now, fluorescence-labeled GST-RBD (RBD DL488 ) was incubated with H1-HeLa(ϩLSR) cell suspensions either alone or together with 1-, 10-, or 100-fold molar excess of nonlabeled GST-RBD or the nonlabeled GST-RBD mutants (T1-T8). Then the amount of cell surface bound fluorescence was analyzed by FACS analysis.
As shown in Fig. 9C, nonlabeled GST-RBD, as well as the GST-RBD mutants T1, T3, T6, and T7, notably inhibited the binding of RBD DL488 to H1-HeLa(ϩLSR) cells already at a molar ratio of 1:1 (RBD DL488 :competing protein). Clear inhibition of RBD DL488 binding to H1-HeLa(ϩLSR) cells was observed for the T4, T5, and T8 mutants only when the proteins were co-incubated in a 10-fold molar excess together with RBD DL488 . Interestingly, the GST-RBD mutant T2 decreased binding of RBD DL488 only when co-incubated in a 100-fold excess together with RBD DL488 . To validate the correct folding of the GST-RBD mutant T2, we performed limited proteolysis of RBD and T2 by incubating the proteins with increasing concentrations of trypsin for 10 min at 4°C. As shown in Fig. 9D, RBD and T2 exhibited a similar pattern of proteolytical fragments after incubation with increasing concentrations of trypsin, indicating that the 5-amino acid insertion in the T2 mutant did not alter the overall structure of the protein. Taken together, the FACS competition analyses indicated that the 5-amino acid linkers that are present in the GST-RBD mutants T4, T5, and T8 and especially in the GST-RBD mutant T2 may   disturb an important LSR-interacting epitope and, thereby, decrease the affinity of the RBD for binding to LSR.

Discussion
LSR is a type I transmembrane protein exposing its N-terminal part at the cell surface. The extracellular part of LSR can be subdivided into two regions, an uncharacterized region at the N terminus followed by an Ig-like, V-type domain (36). Both regions could serve independently or together as a platform for the interaction with CDT. In reference to the UniProt protein entries for LSR, the N terminus of the mouse LSR ortholog is much shorter (ϳ33 amino acids) when compared with human LSR (ϳ86 amino acids). However, human LSR variants exist that also feature a shortened N terminus (e.g. LSR isoform 6). Furthermore, we previously studied the uptake of CDT into cultured cells with an LSR variant that was cloned from a human pancreatic adenocarcinoma cDNA library (20). This LSR variant exhibits a shortened N terminus (ϳ38 amino acids), identical to isoform 6 of human LSR and similar to the mouse LSR ortholog, and was fully functional in terms of mediating cellular uptake of the toxin. Consistently, as shown in the current study, ectopically expressed mouse LSR efficiently promoted cell entry of CDT in HeLa cells.
From these findings, we hypothesized that CDT interacts most likely with the Ig-like domain of LSR. By performing coprecipitation analyses, we confirmed the direct binding of the RBD of CDTb to the Ig-like domain of LSR and obtained by surface plasmon resonance measurements a K D value of ϳ110 nM for this interaction. Similar dissociation constants were calculated for the binding of RBD to the surface of various cultured cells via FACS analysis (K D ϭ ϳ30 -40 nM). An Ig-like domain is also present in anthrax toxin receptors 1 (tumor endothelial marker 8) and 2 (capillary morphogenesis gene 2 protein), two known receptors of anthrax toxin. In contrast to CDT, anthrax toxin does not interact with the Ig-like domain of its receptors but binds with a very high affinity (K D ϭ ϳ0.2 nM for CMG2) to a von Willenbrand factor A domain that is located in front of the Ig-like domain within the extracellular domain (37)(38)(39). However, it has been shown that the Ig-like domain of CMG2 is essential for proper functioning of the receptor-bound protective antigen pore (40). Furthermore, the crystal structure of the protective antigen-CMG2 complex (41), as well as a study from Pilpa et al. (42), confirmed the involvement of the CMG2 receptor in protective antigen pore formation. Thus, it remains to be studied whether the Ig-like domain of LSR participates in the pore formation of the B component of CDT.
The intracellular domain of LSR contains a cysteine-rich region directly after the transmembrane helix followed by a cytoplasmic tail. So far, no functional roles have been attributed to the cysteine-rich region or the cytoplasmic tail of LSR. LSR mutants that lack either the cysteine-rich region or the cytoplasmic tail were properly targeted to the plasma membrane of HCT116 LSR knock-out cells. To our surprise, intoxication of HCT116 LSR knock-out cells with CDT still occurred with both mutants. These results raise the important question about how the endocytic uptake of LSR is triggered because the cysteinerich region and the cytoplasmic tail are not required for the uptake of CDT. It is conceivable that the endocytosis of LSR is triggered also by other membrane proteins, which might form a complex with LSR. We found recently that LSR accumulates in lipid rafts upon binding of CDTb (31). Interestingly, Wigelsworth et al. (43) found that the lipid raft-associated protein CD44 (cluster of differentiation 44 protein) is crucial for CDT uptake into host cells. CD44 is therefore a suitable candidate that could bind to LSR and trigger the endocytic uptake of the protein complex. However, a physical interaction between LSR and CD44 has not been described until now. It remains to be addressed in the future, whether the cysteine-rich region or the cytoplasmic tail of LSR contributes to other physiological functions of LSR in addition to the uptake of CDT.
All isoforms of LSR can be subdivided into two groups, depending on the presence or absence of the transmembrane segment plus the cysteine-rich region in the central part of the receptor. As we show in this study, the absence of the transmembrane segment and cysteine-rich region of LSR renders the protein unable to integrate into the plasma membrane, and such isoforms are most likely retained in the ER with unknown physiological consequences. It might also be possible that LSR variants that are not capable of integrating into the plasma membrane represent partially secreted proteins. One can speculate that secreted LSR variants might block LSR-dependent intracellular signaling pathways by acting as decoy receptors for extracellular ligands. Secreted LSR variants could also serve as ligands for other cell surface receptors. Because LSR is involved in the formation of tricellular tight junctions, it would be of particular interest to analyze whether the expression levels of different LSR isoforms differ between confluent and nonconfluent cultured cells.
Another aim of this study was to narrow down the region and to identify possible epitopes within the receptor-binding domain of CDTb that are essential for interaction with LSR. Based on truncated variants of the B component of iota toxin (Ib), Marvaud et al. (35) confirmed the presence of the RBD at the C-terminal part of Ib. The RBD of Ib corresponds to amino acids 677-876 of CDTb, and we recently confirmed that this region of CDTb binds to LSR exposed at the cell surface (31). The current study suggests that amino acids 757-866 of CDTb are sufficient for the interaction with LSR. In a previous study, a truncated Ib variant lacking only the C-terminal 10 residues was not able to bind to Vero cells, which is in contrast to our results that we obtained with a CDTb RBD mutant lacking also 10 C-terminal amino acids (35). The authors of this study used Ib truncations with a cleaved His tag in their experiments rather than GST-fused constructs as we did for our CDTb truncations. However, we also noticed a decreased binding of the RBD truncation lacking the 10 C-terminal amino acids to the Ig-like domain of LSR. The C-terminal end of the B components of iota-like toxins might be important for the conformational integrity of the proteins.
Our transposon-based mutagenesis approach to identify important epitopes in the RBD of CDTb that are crucial for the interaction with LSR yielded four mutants with decreased ability to compete with the cell surface-binding of the RBD. Interestingly, the T2 mutant, which showed the weakest competitive potential, contained a 5-amino acid in-frame insertion between amino acids 853 and 854. Strikingly, this position lies in between the amino acids 843 and 866 of CDTb, a region corresponding to a linear epitope in Ib, which is recognized by a monoclonal antibody that inhibits in vitro cytotoxicity of iota toxin (35,44). The 5-amino acid insertions of the T4 and T8 mutants were also found within the neutralizing epitope, confirming that this region most likely interacts with the LSR protein. However, the T5 mutant contained a 5-amino acid insertion in between amino acids 781 and 782, indicating an additional epitope in the RBD of CDTb that might be important for the interaction with LSR.
Taken together, our findings provide novel insights into the interaction between CDT and its receptor LSR. A precise knowledge about specific binding regions enables the development of anti-toxin strategies that might help to prevent cell entry of the toxin and, potentially, to prevent the outcome of the toxin-associated disease symptoms.