Binding of Clostridium botulinum C2 Toxin to Asparagine-linked Complex and Hybrid Carbohydrates*

Clostridium botulinum C2 toxin is a binary toxin composed of an enzymatic subunit (C2I) capable of ADP-ribosylating actin and a binding subunit (C2II) that is responsible for interaction with receptors on eukaryotic cells. Here we show that binding of C2 toxin depends on the presence of asparagine-linked carbohydrates. A recently identified Chinese hamster ovary cell mutant (Fritz, G., Schroeder, P., and Aktories, K. (1995) Infect. Immun. 63, 2334–2340) was found to be deficient inN-acetylglucosaminyltransferase I. C2 sensitivity of this mutant was restored by transfection of anN-acetylglucosaminyltransferase I cDNA. C2 toxin sensitivity was reduced after inhibition of α-mannosidase II. In contrast, Chinese hamster ovary cell mutants deficient in sialylated (Lec2) or galactosylated (Lec8) glycoconjugates showed an increase in toxin sensitivity compared with wild-type cells. Our results show that the GlcNAc residue linked β-1,2 to the α-1,3-mannose of the asparagine-linked core structure is essential for C2II binding to Chinese hamster ovary cells.

ture, modifying G-actin, although they show some variations with respect to their substrate specificity. While C2 only ADP-ribosylates ␤and ␥-actin, iota toxin can modify all actin isoforms (12). The binding components of these toxins exhibit about 40% sequence homology to C2II (13), although the former three are more similar to each other than to C2II. The receptor structure for none of these toxins has been defined yet. Recently, a Chinese hamster ovary (CHO) 1 cell mutant (RK14) resistant to C2 toxin has been isolated (14). The resistance of this mutant could be attributed to a lack of toxin binding rather than mutation of an intracellular factor that could inhibit the toxic action of C2 (14). In this study, we show that inability of RK14 cells to bind C2 toxin is based on deficiency in N-acetylglucosaminyltransferase I (GlcNAc-TI) activity, thus providing evidence that an N-linked complex (or hybrid) carbohydrate is essential for C2 toxin binding.

EXPERIMENTAL PROCEDURES
Materials-Lec2 and Lec8 cells (15) were obtained from the American Type Culture Collection (Manassas, VA). Lec13 (16), ldlD.Lec1 (17), and 15B cells (18) were kindly provided by Dr. Pamela Stanley (Albert Einstein College of Medicine, Yeshiva University, New York) and Dr. Felix Wieland (Biochemie-Zentrum, Universitä t Heidelberg, Germany), respectively. Monoclonal antibody 735 directed against polysialic acid (19) was a kind gift of Dr. Rita Gerardy-Schahn (Medizinische Hochschule, Hannover, Germany). An antiserum against C2II was generated by immunization of rabbits with a peptide (ANANRDTDRDGIPDE) corresponding to the amino terminus of trypsin-activated C2II. Lectins, swainsonine, and tetramethylrhodamine isothiocyanate (TRITC)-phalloidin were obtained from Sigma. Taq polymerase was from Roche Molecular Biochemicals. Recombinant C. botulinum toxin C2I and C2II were expressed in Escherichia coli as glutathione S-transferase fusion proteins, and the toxin components were liberated using thrombin as described recently (20). C2II was activated by incubation with 0.2 g of trypsin/g of C2II for 30 min at 37°C. The reaction was stopped by adding trypsin inhibitor (2 g/g of trypsin). Toxin components were stored at Ϫ20°C.
Cell Culture-CHO mutants Lec2, Lec8, Lec13, 15B, and ldlD.Lec1 were maintained in ␣-modified Eagle's medium supplemented with 10% fetal calf serum, 2 mM L-glutamine, 100 units/ml penicillin, and 100 g/ml streptomycin. All cells were incubated at 37°C in 5% CO 2 . CHO-K1 wild-type and RK14 cells were maintained in Dulbecco's modified Eagle's medium/Ham's F-12 (1:1) supplemented with 5% fetal calf serum and antibiotics described above. In experiments where cytotoxic effects of C2 toxin were compared among CHO-K1, RK14, and the other CHO mutants, all cells were maintained in ␣-modified Eagle's medium to exclude differences in C2 cytotoxicity due to different medium components or pH values.
Cytotoxicity Assays-Semiconfluent monolayer cultures were treated with increasing concentrations of equimolar amounts of C2I and trypsin-activated C2II. After appropriate incubation at 37°C, cells were * This work was supported by Deutsche Forschungsgemeinschaft Sonderforschungsbereich 388. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ To whom correspondence should be addressed. Tel.: 49-761-203-5301; Fax: 49-761-203-5311; E-mail: aktories@uni-freiburg.de. 1 The abbreviations used are: CHO, Chinese hamster ovary; GlcNAc-TI, UDP-GlcNAc:␣-D-mannoside-␤1,2-N-acetylglucosaminyltransferase I (EC 2.4.1.101); LCA, lens culinaris agglutinin; PCR, polymerase chain reaction; TRITC, tetramethylrhodamine isothiocyanate; PBS, phosphate-buffered saline. fixed in 4% paraformaldehyde/PBS, and the percentage of rounded cells was determined. In a second approach, cells were seeded at low density into 96-well cell culture plates (2 ϫ 10 3 cells/well) in the presence of increasing amounts of C2 toxin. When the untreated control wells became confluent, the C2 concentration that led to more than 90% cell death was determined. Lectin sensitivity of CHO mutants was determined in the same way, using increasing concentrations of the ricin B chain, lens culinaris agglutinin (LCA), and concanavalin A. To determine the protective effect of ␣-mannosidase II inhibition, cells were first incubated with 10 M swainsonine for 48 h, before treatment with C2 toxin. The influence of toxin treatment on the colony forming ability was determined by incubating CHO cells at a density of 500 cells/6-cm cell culture dish with increasing amounts of C2 toxin for 4 h. Thereafter, cells were washed three times with PBS, medium was added, and the appearing colonies were counted 12-14 days later.
Actin Staining with TRITC-Phalloidin-Toxin-treated or untreated cells grown on glass coverslips were fixed in 4% paraformaldehyde, 0.1% Triton X-100 in PBS for 20 min at room temperature. After washing three times with PBS, cells were incubated with 0.75 g/ml TRITC-phalloidin for 30 min at room temperature. Thereafter, cells were washed four times for 10 min each with PBS followed by a short rinse in water. Specimens were embedded in Kaiser's gelatin.
Complementation Analysis-1:1 mixtures of different CHO mutants were seeded in 35-mm culture dishes (approximately 5 ϫ 10 5 cells in total). The next day, cell fusion was induced using 50% polyethylene glycol 4000 in 75 mM Hepes (pH 8.0). Another 24 h later, cells were harvested and analyzed by SDS-polyacrylamide gel electrophoresis and Western blotting. After transfer onto nitrocellulose membranes, nonspecific binding sites were blocked in 5% nonfat dry milk in PBS. Polysialic acid was detected using monoclonal antibody 735 and antimouse peroxidase conjugate. Bound secondary antibodies were revealed using ECL chemiluminescence detection.
Cloning and Expression of Hamster GlcNAc-TI-Hamster GlcNAc transferase I was cloned from genomic DNA from CHO-K1 wild-type cells and mutants RK14 and 15B by PCR using Taq polymerase and primers 5Ј-GCGGTACCATGCTGAAGAAGCAGTC-3Ј and 5Ј-GCGAAT-TCAGTCCAGCTAGGATCATAG-3Ј (the KpnI and EcoRI restriction sites introduced to facilitate subcloning are underlined). The PCR product was subcloned into the KpnI and EcoRI sites of the vector pcDNA3 (Invitrogen), resulting in the plasmid pGnT1. CHO-RK14 cells were transfected with pGnT1 using LipofectAMINE, following the instructions of the manufacturer. G418-resistant colonies were isolated and examined for C2 toxin sensitivity.
Binding of C2II to CHO Cells-Confluent monolayers were prechilled at 4°C and than incubated with different amounts of C2II in Dulbecco's modified Eagle's medium for 2 h at 4°C. Thereafter, cells were washed six times with PBS and lysed in 20 mM Tris-HCl (pH 8.0), 150 mM NaCl, 5 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, 1% Triton X-100. Protein concentrations were determined according to Bradford (23). Equal amounts were subjected to SDS-polyacrylamide gel electrophoresis and Western blotting as described above. C2II was detected by sequential incubation with anti-C2II antiserum and anti-rabbit peroxidase-conjugate followed by ECL detection, as described (3).

C2 Toxin Binding Depends on N-Linked Complex or Hybrid
Carbohydrates-Binding of several bacterial toxins to target cells has been shown to be carbohydrate-dependent (24), and it has been suggested that cell binding of C2 toxin might be carbohydrate-dependent (25). To characterize the receptor for   Table I). Lec2 and Lec8 mutants lack sialylated and galactosylated glycoproteins and glycolipids, due to genetic defects in the CMP-sialic acid and UDP-galactose transporter, respectively (26,27). Lec13 cells cannot produce GDP-fucose and therefore lack fucosylated glycoconjugates (16). 15B cells are deficient in GlcNAc-TI (18). As a result, they produce only high-mannose and hybrid N-glycans, whereas O-linked glycans and glycolipids are normal. The cytotoxic effect of these mutants was compared with that of CHO-K1 wild-type cells and the C2 toxinresistant CHO mutant RK14, described previously (14). To FIG. 2. Influence of C2 toxin on the colony-forming ability of CHO mutant and wild-type cells. Cells were seeded at low density and treated with increasing concentrations of C2 toxin as described under "Experimental Procedures." Thereafter, cells were grown for 12-14 days, and the number of colonies appearing was determined. Data shown are from one out of two experiments done in duplicate.
FIG. 3. Cell binding and ADP-ribosylation activity of C2 toxin. a, binding of C2II to CHO cells. Cells were incubated with the indicated amounts of C2II for 2 h at 4°C in Dulbecco's modified Eagle's medium. Thereafter, cells were washed five times with PBS, and the amount of bound C2II was determined by Western blotting using an C2II antiserum as described under "Experimental Procedures." b, ADP-ribosylation of actin after treatment of cells with C2 toxin. Cells were treated with 200 ng/ml C2I and 400 ng/ml C2II (ϩC2II) for 16 h or left untreated (ϪC2II). Lysates were incubated with C2I and [ 32 P]NAD. Thereby, actin that had not been modified by the toxin pretreatment was ADP-ribosylated and detected by PhosphorImager analysis.

TABLE II
C2 toxin sensitivity of CHO mutants Cells were seeded into 96-well plates at a density of 10 3 per well in the presence of increasing concentrations of C2 toxin. Cells were then cultivated until the untreated control wells reached confluency, and the lowest C2 concentration that causes more than 90% cell death was determined. Pretreatment with 10 M swainsonine was for 48 h. -Fold resistance (R) and sensitivity (S) are indicated with respect to CHO wild-type cells (CHO-wt). The highest toxin concentration tested did not affect growth of RK14 and 15B cells (R Ͼ 16). Data are the mean from three experiments. n.d., not done. measure the cytotoxic effect of C2 toxin, the cell rounding activity of increasing C2 toxin concentrations was determined for all cell lines (Fig. 1). CHO wild-type cells as well as the mutants Lec8 and Lec13 were sensitive against C2 toxin. However, significant differences in the sensitivity were observed. Lec8 cells were approximately 4-fold more sensitive than wildtype cells (Fig. 1j), and the same was true for Lec2 cells (data not shown). In contrast, Lec13 cells were found to be slightly (2-fold) more resistant than wild-type cells. As reported previously (14), RK14 cells were strongly resistant to C2 toxin. Moreover, the CHO mutant 15B, which lacks complex and hybrid N-linked carbohydrates, was C2-resistant. This was also true for an independent mutant belonging to the same complementation group (ldlD.Lec1; data not shown). The above results could be confirmed by determination of the colony forming ability of toxin-treated cells (Fig. 2). Again, Lec13 cells were found to be more resistant, whereas Lec2 and Lec8 cells exhibited increased sensitivity to C2 toxin compared with wild-type cells. In a different approach, cells were seeded at low density in 96-well plates and treated with increasing concentrations of C2 toxin. When untreated control wells were confluent, the lowest C2 concentration that causes more than 90% cell death was determined. The results obtained further confirmed the above results (Table II).
To show the specificity of the cytotoxic effects and to exclude the possibility that differences in the organization of the cytoskeleton are responsible for the differences observed, mutants were treated with other clostridial toxins known to affect actin fibers by glucosylating Rho GTPases (28,29). All mutants tested (Lec13, Lec2, Lec8, 15B, and RK14) were equally sensitive to the action of C. difficile toxins A and B (data not shown).
Binding of C2II to CHO cells was examined by Western blotting using a specific antiserum directed against the carboxyl terminus of the toxin (Fig. 3a). Wild-type, Lec13, and Lec8 cells bound C2II in a concentration-dependent manner. However, significant differences in C2II binding were not evident. Only a faint, nonspecific binding to RK14 cells was observed. In order to correlate the rounding activity of C2 toxin with the ADP-ribosylation of actin, we determined the amount of modified actin in toxin-treated cells. Cell lysates from toxintreated cells were incubated with C2I to [ 32 P]ADP-ribosylate the unmodified actin fraction. As expected, actin from untreated cells was effectively ADP-ribosylated (Fig. 3b, ϪC2II). By contrast, after toxin treatment (200 ng/ml C2I and 400 ng/ml C2II for 6 h) most actin in wild-type, Lec8, and Lec13 cells was no longer ADP-ribosylated (Fig. 3b, ϩC2II). On the other hand, most of the actin from RK14 cells was still modified by C2I, indicating that it has not been ADP-ribosylated before. The slight reduction in the amount of ADP-ribosylated actin compared with untreated RK14 cells is probably due to the presence of cells that regained the ability to synthesize complex N-glycans. Such revertants, which are C2 toxin-sensitive, appeared after several weeks of culture.
C2 Toxin-resistant RK14 Cells Are Deficient in GlcNAc-TI-Since both 15B and RK14 cells were resistant to C2 toxin, we next investigated whether they belong to the same (Lec1) complementation group. A phenotypic characterization of 15B and RK14 cells revealed that both exhibit nearly identical sensitivity against the lectins ricin, LCA, and concanavalin A (Table III). A complementation assay using polysialic acid as a marker for expression of complex N-linked carbohydrates (30) showed that RK14 belongs to the 15B/ Lec1 complementation group (Fig. 4). While fusions between RK14 and 15B, both of which are negative for polysialic acid, did not correct the phenotype, control fusions with the mutant Lec8 led to reexpression of the marker polysialic acid. Thus, RK14 and 15B cells most probably have a defect in the same gene. This conclusion was confirmed by expressing a cDNA encoding hamster GlcNAc-TI in RK14 cells. RK14 cells stably transfected with pGnT1, encoding hamster GlcNAc-TI, were as sensitive to C2 toxin as CHO wild-type cells (Fig. 5). Taken together, these results prove that C2 toxin resistance of RK14 cells is a result of GlcNAc-TI deficiency and not of an unrelated genetic defect. The hypothesis of a genetic defect in the GlcNAc-TI gene was confirmed by PCR cloning of the GlcNAc-TI gene from RK14 cells. Sequencing of the PCR product revealed a nonsense mutation (G 3 A) at nucleotide position 287 resulting in a stop codon instead of Trp-96 (Fig.  6a). To exclude PCR errors, two independent clones were sequenced. We also sequenced the GlcNAc-TI gene from 15B cells and found a missense mutation (A730G), which changes Asn-244 to Ser-244 (Fig. 6b). Notably, Asn-244 is conserved among all GlcNAc-TI genes cloned to date, including that of C. elegans and Arabidopsis thaliana.
Inhibition of ␣-Mannosidase II Partially Inhibits C2 Cytotoxicity-The next step in N-glycan processing following the condensation of GlcNAc by GlcNAc-TI is the removal of the two terminal mannose residues by ␣-mannosidase II. To further characterize the carbohydrate structure necessary for C2 toxin uptake, CHO cells were treated with the ␣-mannosidase II inhibitor swainsonine. Pretreatment with 10 M swainsonine for 48 h resulted in a weak protection (4-fold increase in resistance) of CHO wild-type cells against C2 toxin (Table II and Fig.  7, a-d). The effect was much stronger in Lec8 cells, where pretreatment with swainsonine resulted in an approximately 16-fold decrease in sensitivity (Table II and Fig. 7, e-h). While treatment of Lec8 cells with 100 ng/ml C2II and 200 ng/ml C2I for 4 h caused rounding up of more than 80%, pretreatment with swainsonine did not result in a significant increase in the number of rounded cells, and there was apparently no breakdown of F-actin filaments (Fig. 7h). In contrast, swainsonine treatment of Lec13 cells affected toxin sensitivity only slightly (Table II). In fact, swainsonine-treated Lec8, Lec13, and wildtype cells were found to be similarly sensitive to C2 toxin. Increasing the toxin concentration (or longer incubations; data not shown) resulted in strong morphological changes, indicating that swainsonine-treated cells are still C2 toxin-sensitive (Fig. 7j). DISCUSSION RK14 cells have been isolated from mutagenized CHO cells by selection for C2 toxin resistance and were found to be deficient in a functional receptor for this toxin (14). We show here that the inability of RK14 cells to bind C2II is the result of a mutation in the GlcNAc-TI gene, which prevents the cells from producing N-linked complex and hybrid carbohydrates. Correcting this defect by transfecting a cDNA encoding an active GlcNAc-TI made the cells C2-sensitive again. Furthermore, independent mutants of the same complementation group (15B, ldlD.Lec1) are also C2-resistant. Thus, the possibility can be excluded that a secondary genetic defect is responsible for C2 resistance in RK14 cells, showing that binding of C2II depends on the presence of N-linked complex or hybrid carbohydrates. Since synthesis of glycolipids is unaffected in Glc-NAc-TI-deficient cells, our results also indicate that glycolipids do not serve as receptors for C2 toxin. Treatment of CHO cells with the ␣-mannosidase II inhibitor swainsonine reduced the toxin sensitivity but did not abolish toxin binding completely. The only difference in the N-glycan structures of RK14 or 15B cells and swainsonine-treated Lec8 cells is the absence of the ␤1,2-linked GlcNAc to the ␣-1,3-linked mannose of the core structure (Fig. 8) in the former cell lines (note that CHO cells did not express the bisecting enzyme N-acetylglucosaminyltransferase III) (31). Our results therefore provide strong evidence that this GlcNAc residue is an essential determinant of the C2 toxin receptor.
Partial inhibition of the cytotoxic effect by swainsonine suggests that complex carbohydrates are more efficient in C2II binding than hybrid structures. Alternatively, the two terminal mannose residues of the GlcNAc 1 -Man 5 -GlcNAc 2 hybrid structure might interfere with toxin binding. However, the observation that Lec8 and Lec2 cells are highly C2-sensitive clearly demonstrate that galactose and sialic acid are not required for toxin binding. In contrast, the higher sensitivity to C2 toxin of Lec2 and Lec8 cells suggests that sialic acid partially blocks the interaction of the toxin with its receptor. Masking of cell surface receptor structures by sialic acid is not uncommon due to their exposed position and negative charge (32). In line with our results, Sugii and Kozaki (25) showed that hemagglutination by C2II is not inhibited by sialic acid and that neuraminidase-treated erythrocytes are more sensitive to the hemolytic activity of C2II.
Fucose-lacking Lec13 cells are slightly less sensitive to C2 toxin. One possible explanation is that fucose is part of the receptor structure. Alternatively, the fucose deficiency of Lec13 cells might reduce the cell surface expression level of the receptor. At present, it is not possible to decide between these possibilities. However, a direct involvement of fucose in toxin binding would be in agreement with observations by Sugii and Kozaki (25) showing that hemagglutination by C2II can be inhibited by GlcNAc, mannose, and fucose. In this context, it is noteworthy that the sensitivity of CHO glycosylation mutants to C2 toxin is similar to their sensitivity to LCA, which specifically binds to ␣-1,6-fucosylated complex bi-and triantennary N-glycan core structures (33). Lec1 cells are highly resistant to LCA and Lec2, and Lec8 cells are more sensitive than wild-type cells (Ref. 34; see also Table III). Furthermore, Lec13 cells are more resistant to LCA than CHO wild-type cells (34).
C2II exhibits significant sequence homology to the binding components of other actin ADP-ribosylating toxins, namely C. difficile CD196 toxin (11), C. perfringens iota toxin (35), and C. spiroforme toxin (9). In addition, sequence comparison showed that anthrax protective antigen is structurally related to C2II. Considering the fact that anthrax toxin protective antigen binds its receptor via the C terminus (36), it seems reasonable to assume that the other toxins mentioned above recognize their receptors also by their C termini. Interestingly, sequence similarity between C2II and the other toxins is completely absent from the C terminus. Thus, it seems very likely that the C2-related toxins have different structural requirements for binding. At least this holds true for iota toxin, which is able to enter C2 toxin-resistant RK14 cells (14). At present, it remains unclear whether GlcNAc-ylated Nglycans are sufficient for C2II binding or whether C2II binding requires also the presence of a specific receptor protein. A C2-specific receptor has to be ubiquitously expressed, because no naturally C2 toxin-resistant cell line has been found, and even erythrocytes from different species are agglutinated by C2II (25). C2 toxin-resistant CHO mutants expressing complex carbohydrates have not been found in the screen reported by Fritz et al. (14), which might be due to the very high frequency of GlcNAc-TI-deficient CHO mutants obtained after chemical mutagenesis (34). FIG. 8. Schematic representation of the N-glycan structures of RK14, Lec8, and swainsonine-treated Lec8 cells. Note that the complex core structure of Lec8 cells can be further modified by GlcNAc transferases IV to VI to generate tri-, tetra-, and pentaantennary glycans. Whether all of these structures can contribute to C2II binding remains to be determined.