- Email: [email protected]
The complete receptor-binding domain of Clostridium difficile toxin A is required for endocytosis
BBRC Biochemical and Biophysical Research Communications 300 (2003) 706–711 www.elsevier.com/locate/ybbrc
The complete receptor-binding domain of Clostridium difficile toxin A is required for endocytosis Cornelia Frisch,a Ralf Gerhard,b Klaus Aktories,a Fred Hofmann,b and Ingo Justb,* a
Institut f€ ur Experimentelle und Klinische Pharmakologie und Toxikologie der Universit€at Freiburg, Albertstr. 25, 79104 Freiberg, Germany b Institut f€ur Toxikologie der Medizinischen Hochschule Hannover, Carl-Neuberg-Str. 1, 30625 Hannover, Germany Received 31 October 2002
Abstract Clostridium difficile toxin A, the chief pathogenicity factor of the antibiotic-associated pseudomembranous colitis, is an intracellular acting cytotoxin that reaches its targets, the Rho GTPases, after receptor-mediated endocytosis. The C-terminal part, constructed of repetitive peptide elements, is thought to bind to a lot of carbohydrate containing receptor molecules to induce clustering and endocytosis. To study which part of the receptor-binding domain is in charge of addressing toxin A into the target cells, we studied the functional, i.e., endocytosis-inducing, binding of toxin A. By a competition assay between various receptorbinding fragments of toxin A and the holotoxin A we found that the complete receptor-binding domain, encompassing the entire repetitive elements, but not parts of it, is necessary for binding-induced endocytosis. The receptor binding domain itself shows weaker competition with holotoxin A than the fragment consisting of receptor-binding domain plus intermediary part of the toxin. All toxin A fragments that compete with holotoxin A are capable of inducing their own endocytosis. Thus, the entire receptorbinding domain, covering the C-terminal third of the toxin A molecule, is responsible for cell uptake of toxin A and the intermediary part contributes to the correct folding and assembly of the repetitive domains. Ó 2002 Elsevier Science (USA). All rights reserved. Keywords: Endocytosis; Toxin A; Receptor; Receptor-binding domain
Toxin A and toxin B from Clostridium difficile are the causative agents of the antibiotic-associated pseudomembranous colitis [1,2]. In addition to their in vivo effects characterized by secretory diarrhea and inflammation of the colonic mucosa, both toxins act cytotoxically upon cultivated cells to induce disaggregation of the actin cytoskeleton [3,4]. They are intracellular acting toxins that reach their targets after receptormediated endocytosis and passing an acidic endosomal compartment from which the toxins are translocated to the cytoplasm [5,6]. In the cytoplasm, both toxins execute their toxicity by intracellular catalytic activity to mono-glucosylate the key regulators of the actin cytoskeleton—the Rho GTPases—resulting in their inactivation [3,7,8]. Toxin A is a single-chained protein toxin (MM 308 kDa) with three functional domains. The N-terminal * Corresponding author. Fax: +49-511-532-2879. E-mail address: [email protected] (I. Just).
domain harbors the catalytic activity, the small putative transmembrane segment in the intermediary domain is thought to mediate translocation into the cytoplasm, and the C-terminal part, covering about one third of the toxin molecule, is the receptor-binding domain [9,10]. The receptor-binding domain comprises repetitive structures, so-called CROPS (combined repetitive oligopeptides) that show homology to the glycan binding domains of glycosyltransferases [11–13]. This homology and the findings that toxin A shows binding to sugar structures such as Gala1-3Galb1-4GlcNAc [14,15] and GalNAcb1-3Galb1-4GlcNAc [16] led to the notion that toxin A binds through its C-terminal part to a carbohydrate structure of the cell receptor. Furthermore, this notion is supported by the finding that monoclonal antibodies directed to the receptor-binding domain are capable of blocking toxin A toxicity [9,17]. Although there are some reports on the identity of the cellular toxin A receptor, the relevant receptor in human colonic cells has not been identified so far.
0006-291X/02/$ - see front matter Ó 2002 Elsevier Science (USA). All rights reserved. doi:10.1016/S0006-291X(02)02919-4
C. Frisch et al. / Biochemical and Biophysical Research Communications 300 (2003) 706–711
Research on toxin A interaction with its receptor has been done so far by studying mere binding of the toxin or toxin fragments, based on the notion that saturable binding is identical with specific binding [14,15,18,19]. If the toxins possess the suggested properties of lectins, it cannot be excluded that the toxins bind to several structurally related carbohydrates in a saturable manner, but only one is the specific receptor that mediates endocytosis of the toxins. This notion is supported by the findings that toxin A binds to carbohydrate structures that do not exist in humans [14,15,20] and to immunoglobulin and non-immunoglobulin components of milk [21]. Therefore, we tested the functional binding of toxin A, i.e., its binding to intact cells followed by cellular up-take and cytotoxic effects, to study which part of the receptor-binding domain of toxin A is in fact responsible for endocytosis.
Materials and methods Materials and chemicals. C. difficile toxin A and B from strain VPI 10463 were purified as described [22]. Toxin A was additionally affinity-purified using bovine thyreoglobulin (Sigma) immobilized to Sepharose. Antibodies against toxin A were produced as described elsewhere [23]. Amplification of toxin fragments from genomic DNA. Genomic DNA from C. difficile strain VPI 10463 was prepared by standard methods. PCR of the toxin fragments (Fig. 1) was performed under the following conditions: 250 ng template, 2.5 mM of each didesoxynucleotide, different concentrations of MgCl2 , and variant units of TaqPolymerase (New England Biolabs). The primers were: A3-N: 50 GGA
Fig. 1. Toxin A fragments. Toxin A was subdivided into three fragments representing the putative functional domains: A1, catalytic domain; A2, translocation domain; and A3, receptor-binding domain, encompassing the repetitive peptide repeats. The numbers at the top give the amino acid position in toxin A.
707
TCC TTA GCC ATA TAT CCC AGG 30 , and A2-C: 50 AGA TCT AGA TTT ATT AAC AAA AGT AAT GG 30 ; A3-N1: 50 GGA TCC AGT AAT ATA TCC ATT TTG AAG 30 , A3-C: 50 AGA TCT TAC TTA GAA GAA AGT AAT AAA AA 30 ; and A1446-C: 50 GGA TCC GAG AAA ATC AAT ACT TTA GG 30 The PCR products were separated by gel electrophoresis, purified, ligated into the TOPO-TAvector, and sequenced. The primers were flanked by a 50 BglI and a 30 BamHI site that were used for the subcloning of the toxin fragment genes into the pGEX-2T vector. Preparation of recombinant proteins. Fragments of C. difficile toxin A (A2–3, AD2–3, A3, AD2–D3, AD3, and A2) were expressed as GSTfusion proteins in Escherichia coli TG1. At an OD600 of 0.6, constructs A2–3 and AD2–3 were induced with 1 lM IPTG at 29 °C for 9 h, A3 with 1 lM IPTG at 29 °C for 4.5 h, and AD2–D3 with 1 lM IPTG at 29 °C for 6 h and AD3 with 1 mM IPTG at 29 °C for 5 h. E. coli were broken with a French Press and the lysates were centrifuged at 8000g for 20 min. The supernatants were used for affinity purification on glutathione–Sepharose beads (Pharmacia). GST-fusion proteins were eluted with glutathione in PBS supplemented with 0.1 mM PMSF. The GST toxin fragments were further purified using thyreoglobulin immobilized to Sepharose. Binding of the A3 containing GST fragments was at 4 °C. After extensive washing with ice-cold PBS, elution was performed with PBS warmed to 37 °C. GST-A2 was given on Centricons (30 kDa cut off) to remove glutathione and to concentrate. The concentration of the fragments was estimated from SDS–PAGE with albumin as reference. The concentration of the fragments was in the range of 10–30 lg/ml. The toxin fragments were stored in the presence of 0.1 mg/ml bovine serum albumin. Cell intoxication. HT29-cells were grown to subconfluency in DMEM/F12 (1:1) supplemented with 10% FCS, 4 mM L -glutamine, 100 lg/ml penicillin, and 100 U/ml streptomycin. Toxin A or its fragments were diluted in PBS supplemented with 0.2 mg/ml bovine serum albumin and the indicated concentrations were added in appropriate volumes to sub-confluently grown HT29 cells. For determination of cell sensitivity towards toxin A and toxin B, the time until full cell rounding was recorded. For determination of the time course of intoxication and the competition experiments, photographs were taken at the indicated time points. Thousands cells were counted and the percentage of completely rounded cells was calculated. For the competition assays toxin A fragments were added to cells directly followed by addition of toxin A or toxin B. The surplus was calculated on the basis of molecular masses of toxin A/B and those of the GST-fusion proteins of the toxin fragments. Cell uptake of the toxin fragments. The basis of this assay is the observation that endocytosed proteins are insensitive to trypsin extracellularly added. Subconfluently grown HT29 cells in dishes (diameter 3 cm) were incubated with toxin A and the GST-fusion proteins of AD2–3, A3, AD3, and A2. Incubation was for 2 h either on ice to prevent or at 37 °C to allow endocytosis. In the case of toxin A, bafilomycin (0.1 lg/ml) was present to allow endocytosis but to prevent cell rounding and detachment. Thereafter, the cells were washed once with ice-cold medium, followed by two washes with ice-cold HankÕsbuffered saline solution. One of the two dishes, previously incubated on ice or at 37 °C, was further incubated with FCS-free medium on ice. The other dish was incubated with trypsin (100 lg/ml) dissolved in FCS-free medium on ice. This trypsin concentration did not lead to rounding and detachment of cells within the observation time. The incubation was under agitation for 20 min. Thereafter, the cells were washed once with ice-cold medium, followed by two washes with icecold HankÕs-buffered saline solution. The cells were mechanically removed and centrifuged and the pellet was dissolved in 50 ll of Laemmli sample buffer. Twenty microliters of the samples were resolved on 7% SDS– PAGE. The lanes were scanned and the protein content was adjusted. Based on this adjustment, samples were run on 7% SDS–PAGE, transferred to PVDF membranes, and analyzed by immunoblotting. Toxin A and the A3 containing fragments were detected with poly-
708
C. Frisch et al. / Biochemical and Biophysical Research Communications 300 (2003) 706–711
clonal anti-toxin A IgG, whereas A2 was detected using anti-GST. The proteins were visualized by enhanced chemiluminescence (ECL).
Results Since the study of mere binding of toxin A or its fragments to intact cells or membrane fractions does not correctly characterize the functional toxin receptor, we applied a competition assay, to check which part of the toxin A molecule is involved in true receptor interaction. Competition means inhibition or delay of holotoxin Ainduced cytotoxicity. This approach ensures that in fact the interaction with the functional receptor was studied and not mere membrane binding. To this end, toxin A was genetically divided into three parts. Two parts cover the functional domains involved in cell entry of toxin A (Fig. 1): the putative translocation domain (fragment A2 covering aa 901–1749) and the receptor-binding domain (fragment A3 covering aa 1750–2750). Competition experiments were performed upon the colonic cell line HT29 that is more sensitive to toxin A than to toxin B (unpublished data and [24]). As expected, only the fragment A3 competed at an about 100-fold surplus
with holotoxin A whereas A2 did not compete even at a 10,000-fold surplus (Fig. 2A). Then, we tested whether extension or deletion of A3 altered the property to compete. The following constructs were created and tested: A2–3 covering aa 901–2710, AD2–3 covering aa 1300–2710 as well as a deletion of A3 (AD3 covering only aa 1750–2257) (Fig. 1). AD2–3 strongly competed at a 30-fold surplus with holotoxin A, thereby inducing a delay of 3 h in the onset of toxin A-induced cytotoxicity (Fig. 2A). Thus, AD2–3 did compete more efficiently with toxin A at the cellular toxin receptor than A3. A further extension of fragment of AD2–3 resulting in A2–3 (that covered about two third of the toxin molecule) did further increase the competition potency (Fig. 2C). Fig. 2C gives the concentration-dependent competition of the three constructs and thus summarizes the competition potencies of A3, AD2–3, and A2–3. A3 together with the intermediary part A2 was more potent to compete with holotoxin A than the mere A3. Because A2 did not compete at all it seems that it indirectly contributes to the competition potency of AD2–3 and A2–3, respectively. Surprisingly, AD3 encompassing half of the putative receptor-binding domain and AD2–D3 (aa 1300–2257) did not compete with holotoxin A, even
Fig. 2. Competition of the receptor-binding domain fragment A3 with toxin A for cytotoxicity. (A) Sub-confluent HT29 cells were incubated with toxin A (3 pM) in the presence of the indicated toxin A fragments. The concentration is given as surplus calculated on the mol mass basis. The percentage of rounded cells was determined from photographs taken at the indicated time points. (.) control; ( ) A3 (100-fold); (j) AD2–3 (30fold); (M) A2 (10,000-fold); ( ) AD3 (10,000-fold). (B) HT29 cells were treated with toxin A (3 pM) ( , ) and toxin B (30 pM) (, j) in the absence ( , ) or the presence ( , j) of AD2–3 (3 lM). (C) HT29 cells were incubated with toxin A (3 pM) in the absence or presence of increasing concentrations (as indicated) of A3, AD2–3, and A2–3. The percentage of rounded cells (SD) was determined from 1000 cells.
C. Frisch et al. / Biochemical and Biophysical Research Communications 300 (2003) 706–711
not at a 10,000-fold surplus (Fig. 2A and data not shown). The time course of holotoxin B-induced cell rounding was not altered by A3 neither by AD2–3, a finding corroborating the notion that toxin A and toxin B recruit different cell receptors (Fig. 2B). Next we wanted to address the question whether fragments A3 and AD2–3 were able to induce their own endocytosis. To this end we made use of the fact that endocytosed toxin A fragments were protected from degradation by extracellularly added proteases. Only extracellular bound but not endocytosed fragments were degraded. Trypsin was found to be the optimal protease because toxin A fragments were sensitive and the cells were not lysed under the experimental conditions. After binding for two hours, the cells were washed, followed by treatment with trypsin on ice. Toxin A fragments in the cell lysates were detected by immunoblot analysis. Binding on ice prevented cellular uptake and was used as control for complete tryptic degradation. As shown in Fig. 3 (left panel) holo-toxin A, AD2–3, A3 and of AD3 bound to HT29 on ice but subsequent trypsin treatment of the intact cells resulted in complete degradation of the toxin fragments, indicating absence of their endocytosis. Then, HT29 cells were incubated with holotoxin A and the toxin fragments at 37 °C, a temperature permissive for endocytosis. In the case of holotoxin A, bafilomycin was present to prevent cell rounding and detachment of
Fig. 3. Endocytosis of toxin A fragments AD2–3 and A3. HT29 cells were incubated with toxin A (1.3 nM), AD2–3 (7 nM), A3 (5 nM), and AD3 (12 nM) on ice (inhibition of endocytosis) or at 37 °C (permissive for endocytosis). After intensive washing, half of the samples were treated with medium only (£) the other half with trypsin (+) for 20 min on ice. After washing, the cells were lysed and the cell lysates were subjected to immunoblot analysis with anti-toxin A (shown). A representative of three experiments is shown.
709
cells. As demonstrated in Fig. 3 (right panel), toxin A, AD2–3, and A3 were detected after trypsin treatment, indicating intracellular presence and thus endocytosis. Fragment AD3, that did not compete with holotoxin A, was completely degraded reflecting lack of endocytosis. The intermediary fragment of toxin A, A2, that was unable to compete, did not show any binding to HT29 cells (data not shown).
Discussion Toxin A is thought to bind lectin-like to target cells. Thus, mere binding is to be distinguished from functional binding, i.e., binding that leads to subsequent cell uptake. To study functional binding of toxin A, competition of the putative receptor binding domain with the holotoxin A was performed at 37 °C, conditions allowing endocytosis in addition to cell binding. Because endocytosed toxin A escapes the competition process and only some toxin molecules are sufficient for cell intoxication, an effective competition can result only in a delay in the onset of cytotoxic features but not in a complete prevention of cytotoxicity. These functional assays in the present study confirm the generally accepted notion that the C-terminal part of toxin A is the receptor binding domain. The binding domain is reported to be constructed of repetitive peptide elements, called ‘‘combined repetitive oligopeptide’’, CROPs, that show homology to streptococcal glycosyltransferases [10,13,25]. It is thought that each or some of the repetitive regions together bind to receptors, so that one toxin A molecule can interact with several receptor molecules leading to clustering followed by endocytosis [10]. However, this concept is not reconcilable with our finding that half of the receptor-binding domain covering about 15 from 30 repetitive elements is completely incapable of binding and competing at the toxin receptor. This finding together with the result that the intermediary part A2 contributes indirectly to receptor binding does not argue against multiple receptor-binding subunits, but supports the view that the correct assembly and the correct folding of all CROPs are prerequisite for receptor binding. Alternatively, it is conceivable, that toxin A interacts with a single receptor molecule, that means, that despite the repetitive structure, only one unique cell receptor is recognized. The latter notion is supported by the finding that a monoclonal antibody that recognizes only two small epitopes in the 109 kDa receptor-binding domain is able to block receptor binding and toxic effects [17,26]. Fragment A2 itself, the intermediary part of toxin A, that is thought to mediate translocation through membranes, did not exhibit any competition or binding to cells, but it increases the competition potency of A2–3 compared to A3. The basis for this observation may be
710
C. Frisch et al. / Biochemical and Biophysical Research Communications 300 (2003) 706–711
the contribution to a correct folding of the receptorbinding domain. The notion that the intermediary part A2 participates in the correct three dimensional structure of the receptor-binding domain A3 is supported by the finding that the antigenicity of native toxin A is restricted to A3 [23]. It seems that the A3 covers the rest of the toxin surface, so that A3 resides on the not accessible A2 core. Thus, the correct folding of the complete receptor-binding domain is important for receptor interaction. The correctly folded receptor-binding domain is also fully functional, i.e., it induces endocytosis. Cellular uptake of the receptor-binding fragment was proven by resistance to extracellular tryptic degradation. Only those toxin fragments that sufficiently compete with holotoxin A are endocytosed. Thus, the signal for cell uptake resides exclusively in the receptor-binding domain A3. The endocytosis experiments result in a quite strange finding that AD3—half of the receptor-binding domain A3—binds at 4 °C but not at 37 °C. This finding is not unique, because toxin A is reported to bind stronger at 4 °C than at 37 °C [27]. Furthermore, the established technique of affinity purification of toxin A is based on the property of toxin A to bind at 4 °C to thyreoglobulin and be released at 37 °C [28].
Acknowledgments This work was supported by Deutsche Forschungsgemeinschaft (Project Ju231/3) and Sonderforschungsbereich 621, project B5. We thank J€ urgen Dumbach for expert technical assistance.
References [1] C.P. Kelly, C. Pothoulakis, J.T. LaMont, Clostridium difficile colitis, New Engl. J. Med. 330 (1994) 257–262. [2] E. Mylonakis, E.T. Ryan, S.B. Calderwood, Clostridium difficileassociated diarrhea, Arch. Intern. Med. 161 (2001) 525–533. [3] I. Just, F. Hofmann, K. Aktories, Molecular mode of action of the large clostridial cytotoxins, Curr. Top. Microbiol. Immunol. 250 (2000) 55–83. [4] M. Thelestam, E. Chaves-Olarte, Cytotoxic effects of the Clostridium difficile toxins, Curr. Top. Microbiol. Immunol. 250 (2000) 85–96. [5] I. Florin, M. Thelestam, Internalization of Clostridium difficile cytotoxin into cultured human lung fibroblasts, Biochim. Biophys. Acta 763 (1983) 383–392. [6] C. Fiorentini, M. Thelestam, Clostridium difficile toxin A and its effects on cells, Toxicon 29 (1991) 543–567. [7] A. Hall, Rho GTPases and the actin cytoskeleton, Science 279 (1998) 509–514. [8] I. Just, F. Hofmann, H. Genth, R. Gerhard, Bacterial protein toxins inhibiting low-molecular-mass GTP-binding proteins, Int. J. Med. Microbiol. 291 (2001) 243–250. [9] C. Von Eichel-Streiber, Molecular biology of the Clostridium difficile toxins, in: M. Sebald (Ed.), Genetics and Molecular Biology of Anaerobic Bacteria, Springer-Verlag, New York, 1993, pp. 264–289.
[10] C. Von Eichel-Streiber, P. Boquet, M. Sauerborn, M. Thelestam, Large clostridial cytotoxins—a family of glycosyltransferases modifying small GTP-binding proteins, Trends Microbiol. 4 (1996) 375–382. [11] C. Von Eichel-Streiber, M. Sauerborn, H.K. Kuramitsu, Evidence for a modular structure of the homologous repetitive C-terminal carbohydrate-binding sites of Clostridium difficile toxins and Streptcoccus mutans glucosyltransferases, J. Bacteriol. 174 (1992) 6707–6710. [12] B.W. Wren, A family of clostridial and streptococcal ligandbinding proteins with conserved C-terminal repeat sequences, Mol. Microbiol. 5 (1991) 797–803. [13] C. Von Eichel-Streiber, M. Sauerborn, Clostridium difficile toxin A carries a C-terminal structure homologous to the carbohydratebinding region of streptococcal glycosyltransferase, Gene 96 (1990) 107–113. [14] H.C. Krivan, G.F. Clark, D.F. Smith, T.D. Wilkins, Cell surface binding site for Clostridium difficile enterotoxin: evidence for a glycoconjugate containing the sequence Gala1–3Galb 1–4GlcNAc, Infect. Immun. 53 (1986) 573–581. [15] K.D. Tucker, T.D. Wilkins, Toxin A of Clostridium difficile binds to the human carbohydrate antigens I, X, and Y, Infect. Immun. 59 (1991) 73–78. [16] K.A. Karlsson, Microbial recognition of target-cell glycoconjugates, Curr. Opin. Struct. Biol. 5 (1995) 622–635. [17] S.M. Frey, T.D. Wilkins, Localization of two epitopes recognized by monoclonal antibody PCG-4 on Clostridium difficile toxin A, Infect. Immun. 60 (1992) 2488–2492. [18] E. Chaves-Olarte, M. Weidmann, C. Von Eichel-Streiber, M. Thelestam, Toxins A and B from Clostridium difficile differ with respect to enzymatic potencies cellular substrate specificities and surface binding to cultured cells, J. Clin. Invest. 100 (1997) 1734– 1741. [19] C. Pothoulakis, R.J. Gilbert, C. Cladaras, I. Castagliuolo, G. Semenza, Y. Hitti, J.S. Montcrief, J. Linevsky, C.P. Kelly, S. Nikulasson, H.P. Desai, T.D. Wilkins, J.T. LaMont, Rabbit sucrase-isomaltase contains a functional intestinal receptor for Clostridium difficile toxin A, J. Clin. Invest. 98 (1996) 641– 649. [20] R.D. Larsen, C.A. Rivera-Marrero, L.K. Ernst, R.D. Cummings, J.B. Lowe, Frameshift and nonsense mutations in a human genomic sequence homologous to a murine UDP-Gal:b-DGal(1,4)-D-GlcNAc a(1,3)-galactosyltransferase cDNA, J. Biol. Chem. 265 (1990) 7055–7061. [21] R.D. Rolfe, W. Song, Immunoglobulin and non-immunoglobulin components of human milk inhibit Clostridium difficile toxin A-receptor binding, J. Med. Microbiol. 42 (1995) 10– 19. [22] I. Just, J. Selzer, F. Hofmann, K. Aktories, Clostridium difficile toxin B as a probe for Rho GTPases, in: K. Aktories (Ed.), Bacterial Toxins—Tools in Cell Biology and Pharmacology, Chapman & Hall, Weinheim, 1997, pp. 159–168. [23] H. Genth, J. Selzer, C. Busch, J. Dumbach, F. Hofmann, K. Aktories, I. Just, New method to generate enzymatically deficient Clostridium difficile toxin B as an antigen for immunization, Infect. Immun. 68 (2000) 1094–1101. [24] J. Torres, M. Camorlinga-Ponce, O. Munoz, Sensitivity in culture of epithelial cells from rhesus monkey kidney and human colon carcinoma to toxins A and B from Clostridium difficile, Toxicon 30 (1992) 419–426. [25] D.M. Lyerly, T.D. Wilkins, Clostridium difficile, in: M.J. Blaser, P.D. Smith, J.I. Ravdin, H.B. Greenberg, R.L. Guerrant (Eds.), Infections of the Gastrointestinal Tract, Raven Press, Ltd., New York, 1995, pp. 867–891. [26] D.M. Lyerly, C.J. Phelps, J. Toth, T.D. Wilkins, Characterization of toxins A and B of Clostridium difficile with monoclonal antibodies, Infect. Immun. 54 (1986) 70–76.
C. Frisch et al. / Biochemical and Biophysical Research Communications 300 (2003) 706–711 [27] J.A. Smith, D.L. Cooke, S. Hyde, S.P. Borriello, R.G. Long, Clostridium difficile toxin A binding to human intestinal epithelial cells, J. Med. Microbiol. 46 (1997) 953–958.
711
[28] H.C. Krivan, T.D. Wilkins, Purification of Clostridium difficile toxin A by affinity chromatography on immobilized thyroglobulin, Infect. Immun. 55 (8) (1987) 1873–1877.
The complete receptor-binding domain of Clostridium difficile toxin A is required for endocytosis Cornelia Frisch,a Ralf Gerhard,b Klaus Aktories,a Fred Hofmann,b and Ingo Justb,* a
Institut f€ ur Experimentelle und Klinische Pharmakologie und Toxikologie der Universit€at Freiburg, Albertstr. 25, 79104 Freiberg, Germany b Institut f€ur Toxikologie der Medizinischen Hochschule Hannover, Carl-Neuberg-Str. 1, 30625 Hannover, Germany Received 31 October 2002
Abstract Clostridium difficile toxin A, the chief pathogenicity factor of the antibiotic-associated pseudomembranous colitis, is an intracellular acting cytotoxin that reaches its targets, the Rho GTPases, after receptor-mediated endocytosis. The C-terminal part, constructed of repetitive peptide elements, is thought to bind to a lot of carbohydrate containing receptor molecules to induce clustering and endocytosis. To study which part of the receptor-binding domain is in charge of addressing toxin A into the target cells, we studied the functional, i.e., endocytosis-inducing, binding of toxin A. By a competition assay between various receptorbinding fragments of toxin A and the holotoxin A we found that the complete receptor-binding domain, encompassing the entire repetitive elements, but not parts of it, is necessary for binding-induced endocytosis. The receptor binding domain itself shows weaker competition with holotoxin A than the fragment consisting of receptor-binding domain plus intermediary part of the toxin. All toxin A fragments that compete with holotoxin A are capable of inducing their own endocytosis. Thus, the entire receptorbinding domain, covering the C-terminal third of the toxin A molecule, is responsible for cell uptake of toxin A and the intermediary part contributes to the correct folding and assembly of the repetitive domains. Ó 2002 Elsevier Science (USA). All rights reserved. Keywords: Endocytosis; Toxin A; Receptor; Receptor-binding domain
Toxin A and toxin B from Clostridium difficile are the causative agents of the antibiotic-associated pseudomembranous colitis [1,2]. In addition to their in vivo effects characterized by secretory diarrhea and inflammation of the colonic mucosa, both toxins act cytotoxically upon cultivated cells to induce disaggregation of the actin cytoskeleton [3,4]. They are intracellular acting toxins that reach their targets after receptormediated endocytosis and passing an acidic endosomal compartment from which the toxins are translocated to the cytoplasm [5,6]. In the cytoplasm, both toxins execute their toxicity by intracellular catalytic activity to mono-glucosylate the key regulators of the actin cytoskeleton—the Rho GTPases—resulting in their inactivation [3,7,8]. Toxin A is a single-chained protein toxin (MM 308 kDa) with three functional domains. The N-terminal * Corresponding author. Fax: +49-511-532-2879. E-mail address: [email protected] (I. Just).
domain harbors the catalytic activity, the small putative transmembrane segment in the intermediary domain is thought to mediate translocation into the cytoplasm, and the C-terminal part, covering about one third of the toxin molecule, is the receptor-binding domain [9,10]. The receptor-binding domain comprises repetitive structures, so-called CROPS (combined repetitive oligopeptides) that show homology to the glycan binding domains of glycosyltransferases [11–13]. This homology and the findings that toxin A shows binding to sugar structures such as Gala1-3Galb1-4GlcNAc [14,15] and GalNAcb1-3Galb1-4GlcNAc [16] led to the notion that toxin A binds through its C-terminal part to a carbohydrate structure of the cell receptor. Furthermore, this notion is supported by the finding that monoclonal antibodies directed to the receptor-binding domain are capable of blocking toxin A toxicity [9,17]. Although there are some reports on the identity of the cellular toxin A receptor, the relevant receptor in human colonic cells has not been identified so far.
0006-291X/02/$ - see front matter Ó 2002 Elsevier Science (USA). All rights reserved. doi:10.1016/S0006-291X(02)02919-4
C. Frisch et al. / Biochemical and Biophysical Research Communications 300 (2003) 706–711
Research on toxin A interaction with its receptor has been done so far by studying mere binding of the toxin or toxin fragments, based on the notion that saturable binding is identical with specific binding [14,15,18,19]. If the toxins possess the suggested properties of lectins, it cannot be excluded that the toxins bind to several structurally related carbohydrates in a saturable manner, but only one is the specific receptor that mediates endocytosis of the toxins. This notion is supported by the findings that toxin A binds to carbohydrate structures that do not exist in humans [14,15,20] and to immunoglobulin and non-immunoglobulin components of milk [21]. Therefore, we tested the functional binding of toxin A, i.e., its binding to intact cells followed by cellular up-take and cytotoxic effects, to study which part of the receptor-binding domain of toxin A is in fact responsible for endocytosis.
Materials and methods Materials and chemicals. C. difficile toxin A and B from strain VPI 10463 were purified as described [22]. Toxin A was additionally affinity-purified using bovine thyreoglobulin (Sigma) immobilized to Sepharose. Antibodies against toxin A were produced as described elsewhere [23]. Amplification of toxin fragments from genomic DNA. Genomic DNA from C. difficile strain VPI 10463 was prepared by standard methods. PCR of the toxin fragments (Fig. 1) was performed under the following conditions: 250 ng template, 2.5 mM of each didesoxynucleotide, different concentrations of MgCl2 , and variant units of TaqPolymerase (New England Biolabs). The primers were: A3-N: 50 GGA
Fig. 1. Toxin A fragments. Toxin A was subdivided into three fragments representing the putative functional domains: A1, catalytic domain; A2, translocation domain; and A3, receptor-binding domain, encompassing the repetitive peptide repeats. The numbers at the top give the amino acid position in toxin A.
707
TCC TTA GCC ATA TAT CCC AGG 30 , and A2-C: 50 AGA TCT AGA TTT ATT AAC AAA AGT AAT GG 30 ; A3-N1: 50 GGA TCC AGT AAT ATA TCC ATT TTG AAG 30 , A3-C: 50 AGA TCT TAC TTA GAA GAA AGT AAT AAA AA 30 ; and A1446-C: 50 GGA TCC GAG AAA ATC AAT ACT TTA GG 30 The PCR products were separated by gel electrophoresis, purified, ligated into the TOPO-TAvector, and sequenced. The primers were flanked by a 50 BglI and a 30 BamHI site that were used for the subcloning of the toxin fragment genes into the pGEX-2T vector. Preparation of recombinant proteins. Fragments of C. difficile toxin A (A2–3, AD2–3, A3, AD2–D3, AD3, and A2) were expressed as GSTfusion proteins in Escherichia coli TG1. At an OD600 of 0.6, constructs A2–3 and AD2–3 were induced with 1 lM IPTG at 29 °C for 9 h, A3 with 1 lM IPTG at 29 °C for 4.5 h, and AD2–D3 with 1 lM IPTG at 29 °C for 6 h and AD3 with 1 mM IPTG at 29 °C for 5 h. E. coli were broken with a French Press and the lysates were centrifuged at 8000g for 20 min. The supernatants were used for affinity purification on glutathione–Sepharose beads (Pharmacia). GST-fusion proteins were eluted with glutathione in PBS supplemented with 0.1 mM PMSF. The GST toxin fragments were further purified using thyreoglobulin immobilized to Sepharose. Binding of the A3 containing GST fragments was at 4 °C. After extensive washing with ice-cold PBS, elution was performed with PBS warmed to 37 °C. GST-A2 was given on Centricons (30 kDa cut off) to remove glutathione and to concentrate. The concentration of the fragments was estimated from SDS–PAGE with albumin as reference. The concentration of the fragments was in the range of 10–30 lg/ml. The toxin fragments were stored in the presence of 0.1 mg/ml bovine serum albumin. Cell intoxication. HT29-cells were grown to subconfluency in DMEM/F12 (1:1) supplemented with 10% FCS, 4 mM L -glutamine, 100 lg/ml penicillin, and 100 U/ml streptomycin. Toxin A or its fragments were diluted in PBS supplemented with 0.2 mg/ml bovine serum albumin and the indicated concentrations were added in appropriate volumes to sub-confluently grown HT29 cells. For determination of cell sensitivity towards toxin A and toxin B, the time until full cell rounding was recorded. For determination of the time course of intoxication and the competition experiments, photographs were taken at the indicated time points. Thousands cells were counted and the percentage of completely rounded cells was calculated. For the competition assays toxin A fragments were added to cells directly followed by addition of toxin A or toxin B. The surplus was calculated on the basis of molecular masses of toxin A/B and those of the GST-fusion proteins of the toxin fragments. Cell uptake of the toxin fragments. The basis of this assay is the observation that endocytosed proteins are insensitive to trypsin extracellularly added. Subconfluently grown HT29 cells in dishes (diameter 3 cm) were incubated with toxin A and the GST-fusion proteins of AD2–3, A3, AD3, and A2. Incubation was for 2 h either on ice to prevent or at 37 °C to allow endocytosis. In the case of toxin A, bafilomycin (0.1 lg/ml) was present to allow endocytosis but to prevent cell rounding and detachment. Thereafter, the cells were washed once with ice-cold medium, followed by two washes with ice-cold HankÕsbuffered saline solution. One of the two dishes, previously incubated on ice or at 37 °C, was further incubated with FCS-free medium on ice. The other dish was incubated with trypsin (100 lg/ml) dissolved in FCS-free medium on ice. This trypsin concentration did not lead to rounding and detachment of cells within the observation time. The incubation was under agitation for 20 min. Thereafter, the cells were washed once with ice-cold medium, followed by two washes with icecold HankÕs-buffered saline solution. The cells were mechanically removed and centrifuged and the pellet was dissolved in 50 ll of Laemmli sample buffer. Twenty microliters of the samples were resolved on 7% SDS– PAGE. The lanes were scanned and the protein content was adjusted. Based on this adjustment, samples were run on 7% SDS–PAGE, transferred to PVDF membranes, and analyzed by immunoblotting. Toxin A and the A3 containing fragments were detected with poly-
708
C. Frisch et al. / Biochemical and Biophysical Research Communications 300 (2003) 706–711
clonal anti-toxin A IgG, whereas A2 was detected using anti-GST. The proteins were visualized by enhanced chemiluminescence (ECL).
Results Since the study of mere binding of toxin A or its fragments to intact cells or membrane fractions does not correctly characterize the functional toxin receptor, we applied a competition assay, to check which part of the toxin A molecule is involved in true receptor interaction. Competition means inhibition or delay of holotoxin Ainduced cytotoxicity. This approach ensures that in fact the interaction with the functional receptor was studied and not mere membrane binding. To this end, toxin A was genetically divided into three parts. Two parts cover the functional domains involved in cell entry of toxin A (Fig. 1): the putative translocation domain (fragment A2 covering aa 901–1749) and the receptor-binding domain (fragment A3 covering aa 1750–2750). Competition experiments were performed upon the colonic cell line HT29 that is more sensitive to toxin A than to toxin B (unpublished data and [24]). As expected, only the fragment A3 competed at an about 100-fold surplus
with holotoxin A whereas A2 did not compete even at a 10,000-fold surplus (Fig. 2A). Then, we tested whether extension or deletion of A3 altered the property to compete. The following constructs were created and tested: A2–3 covering aa 901–2710, AD2–3 covering aa 1300–2710 as well as a deletion of A3 (AD3 covering only aa 1750–2257) (Fig. 1). AD2–3 strongly competed at a 30-fold surplus with holotoxin A, thereby inducing a delay of 3 h in the onset of toxin A-induced cytotoxicity (Fig. 2A). Thus, AD2–3 did compete more efficiently with toxin A at the cellular toxin receptor than A3. A further extension of fragment of AD2–3 resulting in A2–3 (that covered about two third of the toxin molecule) did further increase the competition potency (Fig. 2C). Fig. 2C gives the concentration-dependent competition of the three constructs and thus summarizes the competition potencies of A3, AD2–3, and A2–3. A3 together with the intermediary part A2 was more potent to compete with holotoxin A than the mere A3. Because A2 did not compete at all it seems that it indirectly contributes to the competition potency of AD2–3 and A2–3, respectively. Surprisingly, AD3 encompassing half of the putative receptor-binding domain and AD2–D3 (aa 1300–2257) did not compete with holotoxin A, even
Fig. 2. Competition of the receptor-binding domain fragment A3 with toxin A for cytotoxicity. (A) Sub-confluent HT29 cells were incubated with toxin A (3 pM) in the presence of the indicated toxin A fragments. The concentration is given as surplus calculated on the mol mass basis. The percentage of rounded cells was determined from photographs taken at the indicated time points. (.) control; ( ) A3 (100-fold); (j) AD2–3 (30fold); (M) A2 (10,000-fold); ( ) AD3 (10,000-fold). (B) HT29 cells were treated with toxin A (3 pM) ( , ) and toxin B (30 pM) (, j) in the absence ( , ) or the presence ( , j) of AD2–3 (3 lM). (C) HT29 cells were incubated with toxin A (3 pM) in the absence or presence of increasing concentrations (as indicated) of A3, AD2–3, and A2–3. The percentage of rounded cells (SD) was determined from 1000 cells.
C. Frisch et al. / Biochemical and Biophysical Research Communications 300 (2003) 706–711
not at a 10,000-fold surplus (Fig. 2A and data not shown). The time course of holotoxin B-induced cell rounding was not altered by A3 neither by AD2–3, a finding corroborating the notion that toxin A and toxin B recruit different cell receptors (Fig. 2B). Next we wanted to address the question whether fragments A3 and AD2–3 were able to induce their own endocytosis. To this end we made use of the fact that endocytosed toxin A fragments were protected from degradation by extracellularly added proteases. Only extracellular bound but not endocytosed fragments were degraded. Trypsin was found to be the optimal protease because toxin A fragments were sensitive and the cells were not lysed under the experimental conditions. After binding for two hours, the cells were washed, followed by treatment with trypsin on ice. Toxin A fragments in the cell lysates were detected by immunoblot analysis. Binding on ice prevented cellular uptake and was used as control for complete tryptic degradation. As shown in Fig. 3 (left panel) holo-toxin A, AD2–3, A3 and of AD3 bound to HT29 on ice but subsequent trypsin treatment of the intact cells resulted in complete degradation of the toxin fragments, indicating absence of their endocytosis. Then, HT29 cells were incubated with holotoxin A and the toxin fragments at 37 °C, a temperature permissive for endocytosis. In the case of holotoxin A, bafilomycin was present to prevent cell rounding and detachment of
Fig. 3. Endocytosis of toxin A fragments AD2–3 and A3. HT29 cells were incubated with toxin A (1.3 nM), AD2–3 (7 nM), A3 (5 nM), and AD3 (12 nM) on ice (inhibition of endocytosis) or at 37 °C (permissive for endocytosis). After intensive washing, half of the samples were treated with medium only (£) the other half with trypsin (+) for 20 min on ice. After washing, the cells were lysed and the cell lysates were subjected to immunoblot analysis with anti-toxin A (shown). A representative of three experiments is shown.
709
cells. As demonstrated in Fig. 3 (right panel), toxin A, AD2–3, and A3 were detected after trypsin treatment, indicating intracellular presence and thus endocytosis. Fragment AD3, that did not compete with holotoxin A, was completely degraded reflecting lack of endocytosis. The intermediary fragment of toxin A, A2, that was unable to compete, did not show any binding to HT29 cells (data not shown).
Discussion Toxin A is thought to bind lectin-like to target cells. Thus, mere binding is to be distinguished from functional binding, i.e., binding that leads to subsequent cell uptake. To study functional binding of toxin A, competition of the putative receptor binding domain with the holotoxin A was performed at 37 °C, conditions allowing endocytosis in addition to cell binding. Because endocytosed toxin A escapes the competition process and only some toxin molecules are sufficient for cell intoxication, an effective competition can result only in a delay in the onset of cytotoxic features but not in a complete prevention of cytotoxicity. These functional assays in the present study confirm the generally accepted notion that the C-terminal part of toxin A is the receptor binding domain. The binding domain is reported to be constructed of repetitive peptide elements, called ‘‘combined repetitive oligopeptide’’, CROPs, that show homology to streptococcal glycosyltransferases [10,13,25]. It is thought that each or some of the repetitive regions together bind to receptors, so that one toxin A molecule can interact with several receptor molecules leading to clustering followed by endocytosis [10]. However, this concept is not reconcilable with our finding that half of the receptor-binding domain covering about 15 from 30 repetitive elements is completely incapable of binding and competing at the toxin receptor. This finding together with the result that the intermediary part A2 contributes indirectly to receptor binding does not argue against multiple receptor-binding subunits, but supports the view that the correct assembly and the correct folding of all CROPs are prerequisite for receptor binding. Alternatively, it is conceivable, that toxin A interacts with a single receptor molecule, that means, that despite the repetitive structure, only one unique cell receptor is recognized. The latter notion is supported by the finding that a monoclonal antibody that recognizes only two small epitopes in the 109 kDa receptor-binding domain is able to block receptor binding and toxic effects [17,26]. Fragment A2 itself, the intermediary part of toxin A, that is thought to mediate translocation through membranes, did not exhibit any competition or binding to cells, but it increases the competition potency of A2–3 compared to A3. The basis for this observation may be
710
C. Frisch et al. / Biochemical and Biophysical Research Communications 300 (2003) 706–711
the contribution to a correct folding of the receptorbinding domain. The notion that the intermediary part A2 participates in the correct three dimensional structure of the receptor-binding domain A3 is supported by the finding that the antigenicity of native toxin A is restricted to A3 [23]. It seems that the A3 covers the rest of the toxin surface, so that A3 resides on the not accessible A2 core. Thus, the correct folding of the complete receptor-binding domain is important for receptor interaction. The correctly folded receptor-binding domain is also fully functional, i.e., it induces endocytosis. Cellular uptake of the receptor-binding fragment was proven by resistance to extracellular tryptic degradation. Only those toxin fragments that sufficiently compete with holotoxin A are endocytosed. Thus, the signal for cell uptake resides exclusively in the receptor-binding domain A3. The endocytosis experiments result in a quite strange finding that AD3—half of the receptor-binding domain A3—binds at 4 °C but not at 37 °C. This finding is not unique, because toxin A is reported to bind stronger at 4 °C than at 37 °C [27]. Furthermore, the established technique of affinity purification of toxin A is based on the property of toxin A to bind at 4 °C to thyreoglobulin and be released at 37 °C [28].
Acknowledgments This work was supported by Deutsche Forschungsgemeinschaft (Project Ju231/3) and Sonderforschungsbereich 621, project B5. We thank J€ urgen Dumbach for expert technical assistance.
References [1] C.P. Kelly, C. Pothoulakis, J.T. LaMont, Clostridium difficile colitis, New Engl. J. Med. 330 (1994) 257–262. [2] E. Mylonakis, E.T. Ryan, S.B. Calderwood, Clostridium difficileassociated diarrhea, Arch. Intern. Med. 161 (2001) 525–533. [3] I. Just, F. Hofmann, K. Aktories, Molecular mode of action of the large clostridial cytotoxins, Curr. Top. Microbiol. Immunol. 250 (2000) 55–83. [4] M. Thelestam, E. Chaves-Olarte, Cytotoxic effects of the Clostridium difficile toxins, Curr. Top. Microbiol. Immunol. 250 (2000) 85–96. [5] I. Florin, M. Thelestam, Internalization of Clostridium difficile cytotoxin into cultured human lung fibroblasts, Biochim. Biophys. Acta 763 (1983) 383–392. [6] C. Fiorentini, M. Thelestam, Clostridium difficile toxin A and its effects on cells, Toxicon 29 (1991) 543–567. [7] A. Hall, Rho GTPases and the actin cytoskeleton, Science 279 (1998) 509–514. [8] I. Just, F. Hofmann, H. Genth, R. Gerhard, Bacterial protein toxins inhibiting low-molecular-mass GTP-binding proteins, Int. J. Med. Microbiol. 291 (2001) 243–250. [9] C. Von Eichel-Streiber, Molecular biology of the Clostridium difficile toxins, in: M. Sebald (Ed.), Genetics and Molecular Biology of Anaerobic Bacteria, Springer-Verlag, New York, 1993, pp. 264–289.
[10] C. Von Eichel-Streiber, P. Boquet, M. Sauerborn, M. Thelestam, Large clostridial cytotoxins—a family of glycosyltransferases modifying small GTP-binding proteins, Trends Microbiol. 4 (1996) 375–382. [11] C. Von Eichel-Streiber, M. Sauerborn, H.K. Kuramitsu, Evidence for a modular structure of the homologous repetitive C-terminal carbohydrate-binding sites of Clostridium difficile toxins and Streptcoccus mutans glucosyltransferases, J. Bacteriol. 174 (1992) 6707–6710. [12] B.W. Wren, A family of clostridial and streptococcal ligandbinding proteins with conserved C-terminal repeat sequences, Mol. Microbiol. 5 (1991) 797–803. [13] C. Von Eichel-Streiber, M. Sauerborn, Clostridium difficile toxin A carries a C-terminal structure homologous to the carbohydratebinding region of streptococcal glycosyltransferase, Gene 96 (1990) 107–113. [14] H.C. Krivan, G.F. Clark, D.F. Smith, T.D. Wilkins, Cell surface binding site for Clostridium difficile enterotoxin: evidence for a glycoconjugate containing the sequence Gala1–3Galb 1–4GlcNAc, Infect. Immun. 53 (1986) 573–581. [15] K.D. Tucker, T.D. Wilkins, Toxin A of Clostridium difficile binds to the human carbohydrate antigens I, X, and Y, Infect. Immun. 59 (1991) 73–78. [16] K.A. Karlsson, Microbial recognition of target-cell glycoconjugates, Curr. Opin. Struct. Biol. 5 (1995) 622–635. [17] S.M. Frey, T.D. Wilkins, Localization of two epitopes recognized by monoclonal antibody PCG-4 on Clostridium difficile toxin A, Infect. Immun. 60 (1992) 2488–2492. [18] E. Chaves-Olarte, M. Weidmann, C. Von Eichel-Streiber, M. Thelestam, Toxins A and B from Clostridium difficile differ with respect to enzymatic potencies cellular substrate specificities and surface binding to cultured cells, J. Clin. Invest. 100 (1997) 1734– 1741. [19] C. Pothoulakis, R.J. Gilbert, C. Cladaras, I. Castagliuolo, G. Semenza, Y. Hitti, J.S. Montcrief, J. Linevsky, C.P. Kelly, S. Nikulasson, H.P. Desai, T.D. Wilkins, J.T. LaMont, Rabbit sucrase-isomaltase contains a functional intestinal receptor for Clostridium difficile toxin A, J. Clin. Invest. 98 (1996) 641– 649. [20] R.D. Larsen, C.A. Rivera-Marrero, L.K. Ernst, R.D. Cummings, J.B. Lowe, Frameshift and nonsense mutations in a human genomic sequence homologous to a murine UDP-Gal:b-DGal(1,4)-D-GlcNAc a(1,3)-galactosyltransferase cDNA, J. Biol. Chem. 265 (1990) 7055–7061. [21] R.D. Rolfe, W. Song, Immunoglobulin and non-immunoglobulin components of human milk inhibit Clostridium difficile toxin A-receptor binding, J. Med. Microbiol. 42 (1995) 10– 19. [22] I. Just, J. Selzer, F. Hofmann, K. Aktories, Clostridium difficile toxin B as a probe for Rho GTPases, in: K. Aktories (Ed.), Bacterial Toxins—Tools in Cell Biology and Pharmacology, Chapman & Hall, Weinheim, 1997, pp. 159–168. [23] H. Genth, J. Selzer, C. Busch, J. Dumbach, F. Hofmann, K. Aktories, I. Just, New method to generate enzymatically deficient Clostridium difficile toxin B as an antigen for immunization, Infect. Immun. 68 (2000) 1094–1101. [24] J. Torres, M. Camorlinga-Ponce, O. Munoz, Sensitivity in culture of epithelial cells from rhesus monkey kidney and human colon carcinoma to toxins A and B from Clostridium difficile, Toxicon 30 (1992) 419–426. [25] D.M. Lyerly, T.D. Wilkins, Clostridium difficile, in: M.J. Blaser, P.D. Smith, J.I. Ravdin, H.B. Greenberg, R.L. Guerrant (Eds.), Infections of the Gastrointestinal Tract, Raven Press, Ltd., New York, 1995, pp. 867–891. [26] D.M. Lyerly, C.J. Phelps, J. Toth, T.D. Wilkins, Characterization of toxins A and B of Clostridium difficile with monoclonal antibodies, Infect. Immun. 54 (1986) 70–76.
C. Frisch et al. / Biochemical and Biophysical Research Communications 300 (2003) 706–711 [27] J.A. Smith, D.L. Cooke, S. Hyde, S.P. Borriello, R.G. Long, Clostridium difficile toxin A binding to human intestinal epithelial cells, J. Med. Microbiol. 46 (1997) 953–958.
711
[28] H.C. Krivan, T.D. Wilkins, Purification of Clostridium difficile toxin A by affinity chromatography on immobilized thyroglobulin, Infect. Immun. 55 (8) (1987) 1873–1877.