Structural basis for the differential toxicity of cholera toxin and Escherichia coli heat-labile enterotoxin. Construction of hybrid toxins identifies the A2-domain as the determinant of differential toxicity.

Cholera toxin (Ctx) and E. coli heat-labile enterotoxin (Etx) are structurally and functionally similar AB5 toxins with over 80% sequence identity. When their action in polarized human epithelial (T84) cells was monitored by measuring toxin-induced Cl- ion secretion, Ctx was found to be the more potent of the two toxins. Here, we examine the structural basis for this difference in toxicity by engineering a set of mutant and hybrid toxins and testing their activity in T84 cells. This revealed that the differential toxicity of Ctx and Etx was (i) not due to differences in the A-subunit's C-terminal KDEL targeting motif (which is RDEL in Etx), as a KDEL to RDEL substitution had no effect on cholera toxin activity; (ii) not attributable to the enzymatically active A1-fragment, as hybrid toxins in which the A1-fragment in Ctx was substituted for that of Etx (and vice versa) did not alter relative toxicity; and (iii) not due to the B-subunit, as the replacement of the B-subunit in Ctx for that of Etx caused no alteration in toxicity, thus excluding the possibility that the broader receptor specificity of EtxB is responsible for reduced activity. Remarkably, the difference in toxicity could be mapped to a 10-amino acid segment of the A2-fragment that penetrates the central pore of the B-subunit pentamer. A comparison of the in vitro stability of two hybrid toxins, differing only in this 10-amino acid segment, revealed that the Ctx A2-segment conferred a greater stability to the interaction between the A- and B-subunits than the corresponding segment from Etx A2. This suggests that the reason for the relative potency of Ctx compared with Etx stems from the increased ability of the A2-fragment of Ctx to maintain holotoxin stability during uptake and transport into intestinal epithelia.

The severe, and at times fatal, diarrheal disease caused by Vibrio cholerae is due to the potent action of cholera toxin (Ctx) 1 (for a review, see Ref. 1). A structurally and functionally similar toxin, heat-labile enterotoxin (Etx), is produced by certain strains of enterotoxigenic Escherichia coli that are responsible for causing the generally milder "traveler's" diarrhea of humans and scouring in farm animals (2,3). Both Ctx and Etx are hetero-oligomeric proteins comprised of a single A-subunit (M r 28,000) and five B-subunits (M r 12,000 each) (4 -6). The A-subunit contains two distinct structural domains linked by a disulfide bridge: an A1-fragment (residues ) that displays ADP-ribosyl transferase activity and an A2-fragment (residues 193-240) that mediates interaction with the B-subunit pentamer (4,6). The B-subunits of Ctx and Etx (CtxB and EtxB, respectively) bind to cell surface receptors, principally G M1ganglioside, found ubiquitously on the plasma membranes of eukaryotic cells (7). Although Ctx and Etx exhibit a remarkable degree of structural homology, with 81.6% sequence identity between CtxA and EtxA and 82.5% sequence identity between CtxB and EtxB (8 -10), there are a number of subtle physicochemical and functional differences between the two toxins. For example, although the B-subunit pentamers of both toxins are resistant to environmental conditions that normally lead to protein denaturation, the pH stability of EtxB is almost 2 orders of magnitude greater than that of CtxB (11). Studies have also shown that the receptor binding specificities CtxB and EtxB are slightly different; whereas both bind with high affinity to ganglioside G M1 , and to a lesser extent to G D1b , EtxB also binds significantly to glycoprotein receptors and polyglycosylceramides, and with lower affinity to G M2 , asialo-G M1 , and paragloboside (12)(13)(14)(15)(16)(17). Sequence differences between CtxA and EtxA are spread throughout the polypeptides with the lowest identity surrounding the A1/A2 cleavage site and near the C terminus of the A2-fragment (10). The influence on toxin function of most of the primary amino acid sequence divergence has not yet been fully explored.
The action of cholera toxin and related enterotoxins on eukaryotic cells depends on a complex sequence of events that eventually leads to alterations in ion fluxes and a concomitant loss of water, characteristic of cholera and related diarrheal diseases. The use of polarized human colonic epithelial T84 cells has greatly facilitated studies of Ctx and Etx, since toxicity can be readily monitored as the induction of electrogenic Cl Ϫ secretion (18). Toxin action is initiated by binding of the B-subunit moiety to cell surface receptors. Recently, it was demonstrated that, following binding of Ctx to T84 cells, Ctx-G M1 complexes cluster in caveolae-like detergent insoluble subdomains of the plasma membrane (19). Invagination and internalization of these membrane domains results in the formation of smooth endocytic vesicles that enter vesicular trafficking pathways leading to transport of the toxin to the trans-Golgi network (TGN) (18, 20 -25). The observation that brefeldin A inhibits cholera toxin action (20,21,26), and the presence of a KDEL sequence at that the C terminus of CtxA (RDEL in EtxA) have suggested that the toxin is transported from the TGN to the endoplasmic reticulum (ER). Indeed, mutations in the K(R)DEL sequence of CtxA and EtxA reduce the efficiency of toxin-induced Cl Ϫ secretion in T84 cells (22). It has been speculated that the A-subunit may detach from the B-subunits in the TGN, since only the A-subunit (A1/A2-fragments) has been detected in the ER (23)(24)(25), although transport of the holotoxin from the Golgi to the ER, followed by rapid dissociation and anterograde transport of the B-subunit back to the Golgi, has not been excluded. Reduction of the disulfide bond between the A1-and A2-fragments is thought to be catalyzed by protein disulfide isomerase resident in the ER (27,28), followed by translocation of the A1-fragment across the ER membrane to the cytosolic compartment. Because of its hydrophobicity, the A1-fragment may remain associated with the cytosolic face of the ER membrane, rather than being released free into the cytosol. The subsequent trafficking and presumed delivery of the A1-fragment to the basolateral membrane are less well understood, although the finding that the A-subunit interacts with "so-called" ADP-ribosylation factors (ARFs) (29), involved in vesicular transport, may facilitate its anterograde targeting to the basolateral membrane. ARFs have also been demonstrated directly to increase toxin activity in vitro (30), a finding that may be important in determining the magnitude of toxin action in vivo. The A1-fragments of Ctx and Etx accomplish their toxic effects by ADP-ribosylating G S␣ , a component of the trimeric GTP-binding protein that activates adenylate cyclase. In T84 cells, toxin-induced elevations in cAMP levels lead to electrogenic Cl Ϫ secretion: the primary ion transport event responsible for secretory diarrhea in humans.
For Ctx and Etx to exhibit full toxicity, the A-subunits must undergo proteolytic cleavage or "nicking" at Arg-192 to give separate A1-and A2-fragments (31). In the case of cholera toxin, extracellular proteases, such as HA protease produced by V. cholerae, can efficiently nick and activate the A-subunit (32). By contrast, Etx from enterotoxinogenic E. coli, as well as recombinant Ctx produced in E. coli, are normally isolated with their A-subunits intact; trypsin or other gut-associated proteases have been postulated to accomplish toxin activation in such cases (33). Recent studies on T84 cells have shown that a serine protease, which efficiently nicks and activates both CtxA and EtxA, is present on either the apical surface or in apicallyderived transport vesicles (31). In this respect, no difference was found in toxicity of commercial (nicked) preparations of Ctx and recombinant Ctx that was either unnicked or nicked with trypsin.
The time course and magnitude of electrogenic Cl Ϫ secretion in T84 cells demonstrated that Ctx (hereafter referred to as CtxAB) was significantly more active than Etx (hereafter referred to as EtxAB); CtxAB exhibited a shorter apparent lag period and generated a higher short circuit current (22). In this paper, we describe the structural basis for this difference in toxicity. The data reveal that it is not due to structural or functional differences in the ADP-ribosylating A1-fragment or in the receptor-binding B-subunits, but to the A2-adapter fragment, which mediates A/B-subunit interactions. These findings highlight the importance of a region in the toxin that has hitherto been generally overlooked, and provides a possible contributory explanation for the difference in severity of cholera and traveler's diarrheal disease.

Materials
All reagents were purchased from Sigma or BDH unless otherwise stated. Monoclonal antibodies CT-17, which recognizes CtxA was the gift of Professor J. Holmgren (University of Gothenburg, Gothenburg, Sweden), and 118-8, which recognizes the CtxB and EtxB, was provided by Dr. H. Person (University of Umeå, Umeå, Sweden). Recombinant human ADP-ribosylation factor-6 (rhARF 6) was provided by Dr. W. Patton (NHLBI, National Institutes of Health, Bethesda, MD)

ctx and etx Operons
The ctxAB operon encoding the A-and B-subunits of cholera toxin was amplified by PCR from the template plasmid, pJBK33 (34) using the following primers; 5Ј-TTGGGCCCGATATCTTTCTGTTAAACAAA-G-3Ј (CROL1) and 5Ј-GCTCTAGACTAGTTTGCCATACTAATTGCGG-C-3Ј (CROL2) and Expand TM high fidelity polymerase (Boehringer Mannheim). The resulting fragment was cleaved with ApaI and SpeI and then ligated into pBluescript II KSϩ (Stratagene) that had been cut with the same restriction enzymes to yield pRC1 ( Fig. 1). 2 To facilitate controlled, high level expression of CtxAB, the EcoRV-SpeI fragment of pRC1 was subcloned into pTTQ18 (35) at the SmaI and XbaI sites to yield pRC9. Plasmid pTRH29 was described previously and is a pBluescript derivative containing the etxAB operon of E. coli enterotoxin flanked by EcoRV and SpeI restriction sites (36). A pTTQ18 derivative, harboring the EcoRV-SpeI fragment from pTRH29 and designated pAM29, has been described (37).

Construction of Mutant and Hybrid Toxin Operons
Mutant and hybrid operons were constructed by ligating two separate fragments derived by PCR amplification of upstream and downstream segments of the ctxAB or etxAB operons.
Construction of a Mutant ctx Operon Encoding CtxA (RDEL) CtxB-pJBK33 was used as the template for amplification of a 5Ј-fragment encoding CtxA using primers CROL1 and 5Ј-TCATCCCGAATTCTAT-TATGTG-3Ј, and a 3Ј-fragment encoding CtxB using primers 5Ј-ATA-GAATTCGGGATGAA-3Јand CROL2. The 3Ј-fragment was cloned into pBluescript at the EcoRI-SpeI sites, yielding pATA8. This plasmid was subsequently cut with ApaI and EcoRI, and the 5Ј PCR fragment inserted to reconstruct the ctxAB operon, yielding plasmid pATA14 ( Fig. 1). 2 The EcoRI site in pATA14 at the junction between the genes encoding the A-and B-subunit resulted in replacement of Lys-237 by Arg in CtxA. The EcoRV-SpeI fragment of pATA14 was subcloned into pTTQ18, yielding pCDR3.
Construction of a Hybrid Operon Encoding EtxA  CtxA (225)(226)(227)(228)(229)(230)(231)(232)(233)(234)(235)(236)(237)(238)(239)(240) CtxB-pTRH29 was used as the template for amplification of a 5Ј-fragment from the etxA gene using the universal reverse primer (Stratagene) 5Ј-AACAGCTATGACCATG-3Ј and oligonucleotide 5Ј-CCTCTGACTGGTACCCTGAA-3Ј. The DNA fragment was cleaved with EcoRV and KpnI and inserted into these sites in pBluescript to yield pRC7. pRC1 was used as the template for amplification of a 3Ј-fragment of the ctxAB operon from codon 226 of ctxA using the oligonucleotide 5Ј-TTTCAGGGTACCAATCTG-3Ј and the Ϫ40 primer (Stratagene) 5-TTTTCCCAGTCACGAC-3Ј. The fragment was then cleaved with KpnI and SpeI and inserted into the corresponding sites of pBluescript, yielding pRC2. The hybrid operon was constructed using an intermediate vector pRCK18, a derivative of pK18 (38) in which the polylinker sequence had been replaced by one with sites for BamHI-XbaI-SpeI-ClaI-KpnI-NdeI-EcoRV-ApaI-PstI. 3 The EcoRV-KpnI fragment from pRC7 was inserted into pRCK18, followed by insertion of the KpnI-SpeI fragment from pRC2. The resultant plasmid was cleaved with EcoRV and SpeI and the hybrid operon cloned into the correspond- 2 The ApaI site is immediately adjacent to the EcoRV site shown in Fig. 1. 3 Difficulties were encountered in the construction of the hybrid operons, possibly due to the toxicity of gene products encoded by fragments of the toxin genes; and necessitated the design and use of an intermediate vector.
Construction of a Hybrid Operon Encoding CtxA  EtxA (226 -240) EtxB-The hybrid operon was constructed in a similar manner to that in pRC18. The 5Ј-fragment from the ctxA gene was amplified from pRC1 using the universal reverse primer and oligonucleotide 5Ј-CA-GATTGGTACCCTGAAA-3Ј, and the 3Ј-fragment of the etxAB operon from codon 226 of etxA was amplified from pTRH29 using the oligonucleotide 5Ј-TTTTCAGGGTACCAGTCAGA-3Ј and the Ϫ40 primer. These were cloned separately into pBluescript to yield pRC3 and pRC6, respectively. The subsequent subcloning of these fragments to generate a hybrid operon was carried out in a manner identical to that described above for the construction of pRC18. The resultant plasmid encoding CtxA  EtxA (226 -240) EtxB was designated pRC19 (Fig. 1). The EcoRV-SpeI fragment of pRC19 was subcloned into pTTQ18, yielding pRC21.
The DNA sequences of all of the cloned operons and constructs were verified by dideoxy nucleotide sequencing.

Toxin Purification
V. cholerae 0395NT (9) with a chromosomal ctxAB deletion and harboring pTTQ-plasmid derivatives encoding wild-type, mutant, or hybrid toxins were cultured overnight at 37°C on Luria-Bertani agar plates supplemented with 30 g/ml streptomycin, 50 g/ml kanamycin, and 200 g/ml ampicillin. Bacteria were inoculated into NZCYM medium (1% (w/v) NZ amine (ICN Biochemicals), 0.5% (w/v) NaCl, 0.1% (w/v) casamino acids (Difco), 0.5% (w/v) yeast extract (Betalab), 0.2% (w/v) MgSO 4 ⅐7H 2 O) plus 200 g/ml ampicillin and grown at 37°C in a rotary shaker. When cultures reached an absorbance of 0.6 ( ϭ 600 nm), isopropyl-1-thio-␤-D-galactopyranoside was added to a final concentration of 0.5 mM and cells were cultured for an additional 12 h. A total of 3 liters of culture was centrifuged at 8,000 ϫ g in a Sorvall RC-5B centrifuge at 4°C. Sodium hexametaphosphate (39) was added to the culture supernatant to a final concentration of 0.25% (w/v), the pH adjusted to 4.5 with concentrated HCl, and the mixture stirred for 2 h at room temperature. Precipitated material was recovered by centrifugation at 8,000 ϫ g in a Sorvall RC-5B centrifuge at 4°C and then dissolved in 150 ml of 0.1 M sodium phosphate, pH 8.0. The sample was dialyzed against two changes of 20 mM Tris-HCl, pH 7.0, at 4°C in the presence of 0.02% (w/v) sodium azide and then centrifuged in a Sigma 4K10 centrifuge at 15,000 ϫ g for 1 h at 4°C. The supernatant was adjusted to 1.0 M (NH 4 ) 2 SO 4 , stirred for 1 h at room temperature, and then centrifuged at 15,000 ϫ g for 1 h at 4°C. The supernatant was filtered through a 0.2-m filter (Millipore) and toxin purified by FPLC (Amersham Pharmacia Biotech) using a modification of the method described by Amin and Hirst (42). Briefly, the sample was applied to a hydrophobic interaction column (2 ϫ 11 cm, 35-ml bed volume, Source 15PHE, Amersham Pharmacia Biotech) and eluted with a linear gradient (1.0 to 0 M) of (NH 4 ) 2 SO 4 containing 20 mM Tris-HCl, pH 7.5. Fractions containing the toxin were pooled, desalted by overnight dialysis against 20 mM Tris base, 20 mM NaCl, pH 8.75, at 4°C and then applied to a Resource Q column (6 ml, Amersham Pharmacia Biotech). Samples were eluted with a linear gradient of NaCl (20 mM to 1.0 M) in 20 mM Tris base, pH 8.75. Fractions containing the toxin were added to a NAP-10 column (Amersham Pharmacia Biotech) to replace the buffer with phosphate-buffered saline (PBS: 0.15 M NaCl, 10 mM sodium phosphate, pH 7.2) and stored at -80°C. Toxin concentration was determined by UV absorbance at ϭ 280 nm using the theoretical molar extinction coefficients for each holotoxin calculated according to Gill and von Hippel (40).

Toxin Assays
Electrophysiology-T84 cells (from passages 75-100) obtained from ATCC were grown and passaged as described (18,41). Toxins were diluted in prewarmed Hanks' balanced salt solution (HBSS) containing (per liter): 0.185 g of CaCl 2 , 0.098 g of MgSO 4 , 0.4 g of KCl, 0.06 g of KH 2 PO 4 , 8 g of NaCl, 0.048 g of Na 2 HPO 4 , 1.0 g of glucose, and 10 mM HEPES at pH 7.4, and applied to the apical surface of confluent T84 cell monolayers in Transwell inserts (Costar, Cambridge, MA), followed by incubation at 37°C. Alternatively, T84 cells were placed at 4°C; the toxins diluted in ice-cold HBSS were then applied apically for 30 min at 4°C. Cells were washed to remove unbound toxin, fresh HBSS was added, and the cells were incubated at 37°C. Measurements of short circuit current (I sc ) and resistance (R) were performed as reported elsewhere (18,41).

Assessment of Holotoxin Stability
The stability of A-and B-subunit interaction in hybrid toxins was determined using an adaptation of a G M1 -based enzyme-linked immunosorbent assay (42). 96-well microtiter plates were coated with G M1ganglioside (1. Absorbance was measured at 450 nm in a microtiter plate reader (Anthos).

RESULTS AND DISCUSSION
Ctx and Etx Exhibit Differential Toxicity in T84 Monolayers-Previously, when samples of periplasmic fractions from E. coli that contained equivalent amounts of CtxAB or EtxAB were applied to the apical surfaces of polarized T84 cells, the time course of Cl Ϫ efflux elicited by the toxins were markedly different; a higher rate of Cl Ϫ secretion was induced by CtxAB than by EtxAB (22). This suggested that structural or functional attributes in either the A-or B-subunits of cholera toxin conferred enhanced toxicity. To investigate this further, highly purified CtxAB and EtxAB were prepared as described under "Experimental Procedures." Briefly, recombinant plasmids encoding CtxAB or EtxAB were introduced into the non-toxinogenic V. cholerae O395NT strain, and the toxins purified from the culture media. The CtxAB preparation eluted as a single homogeneous peak on anion-exchange chromatography. When analyzed by SDS-polyacrylamide gel electrophoresis, the Asubunit of CtxAB migrated with an apparent M r of 21 kDa, corresponding to the A1-fragment, indicating that it had been proteolytically nicked by V. cholerae, as expected (Fig. 2, lane  1). The EtxAB preparation contained a mixture of both unnicked and nicked toxin, which eluted at slightly different salt concentrations, enabling separation of the two populations (Fig. 2, lanes 2 and 3). When the toxicity of purified CtxAB and EtxAB (100 nM, nicked) was tested on T84 cells the toxins exhibited clear differences in time course and magnitude of Cl Ϫ secretion, with CtxAB being more potent than EtxAB (Fig. 3). This confirmed the earlier observations on CtxAB and EtxAB derived from periplasmic fractions of E. coli (22).
Comparison of the toxicity of 100 nM amounts of the purified nicked and unnicked preparations of EtxAB revealed that they were equally active in inducing electrogenic Cl Ϫ secretion by T84 cells (Fig. 3). This is consistent with our previous finding that T84 cells contain a serine protease that can nick and activate the A-subunits of cholera toxin and E. coli enterotoxin, and it further demonstrates that nicking per se is not the rate-determining step defining the magnitude and time course of Cl Ϫ secretion.

Is the K(R)DEL Motif Responsible for Differential Toxicity?-
The Lys-Asp-Glu-Leu-COOH (KDEL) motif at the C terminus of the A2-fragment of cholera toxin has been demonstrated to contribute to the efficiency of toxin action in T84 cells (22). In EtxAB, the same motif contains a conservative amino acid replacement of Arg for Lys (Fig. 4). Although the RDEL motif is known to serve as a functional ER retention signal (44,45) and enhances EtxAB activity in T84 cells (22), the possibility remained that the differential toxicity of cholera toxin and E. coli enterotoxin might be attributable to the subtle alteration in this C-terminal motif. To investigate this, the KDEL motif of cholera toxin was replaced by an RDEL sequence, and resultant toxin, CtxA (RDEL) CtxB, was expressed and purified as described above (Fig. 5, lane 3). When the effects of 100 nM CtxA (RDEL) CtxB and wild-type CtxAB on T84 cells were compared, both toxins elicited similar anamnestic responses, with lag periods of ϳ30 min (Fig. 6). We therefore conclude that the different ER retention signals of cholera toxin and E. coli enterotoxin are not the cause of their differential toxicity. These findings are consistent with those of Kreitman and Pastan (46), who found that KDEL or RDEL attached to the C terminus of Pseudomonas exotoxin A were equally effective in enhancing toxicity (46).
Construction of Hybrid Ctx/Etx Toxins to Determine the Basis for Differential Toxicity-To define whether the A-or Bsubunits of the two toxins are responsible for their different activities in T84 cells, a set of hybrid toxins was constructed. A schematic representation of the holotoxin structure of EtxAB is shown in Fig. 4i. It illustrates that the A2-fragment can be considered as two discrete structural segments primarily responsible for interacting with the A1-and the B-subunit pentamer; a long ␣-helix in A2-(residues 197-224) abuts the A1fragment, while the remainder (A2 residues 225-236) winds through the pore of the toroidal B-pentamer. Because of the relatively large number of non-conservative differences in amino acids in the A2-fragments of CtxA and EtxA (Fig. 4ii), hybrid toxins were designed with the fusion site located at the end of the long ␣-helix of the A2-fragment, corresponding to amino acid 225. By so doing, the interactions between the A1and A2-fragments and between the A2-fragment and B-subunits in the hybrids would involve homologous toxin domains, which we considered would favor toxin assembly and stability.
Construction of the hybrid toxins, with a fusion joint at codon 225 of the A-subunit gene, involved a series of PCR amplifications of the ctxAB and etxAB operons. The 5Ј-portion of the ctxA and etxA genes (to codon 226) were amplified and cloned separately as EcoRV/KpnI fragments, while the downstream 3Ј-portions (from codon 226 of ctxA and etxA, together with the entire ctxB and etxB gene, respectively) were cloned as KpnI/ SpeI fragments. Consequently, the ligation of the 5Ј and 3Ј portions at the KpnI site (Fig. 1) enabled the generation of two chimeric operons, one comprising codons specifying the signal peptide plus residues 1-225 of ctxA, 226 -240 of etxA, and the entire etxB gene (to yield a hybrid toxin designated CtxA  EtxA (226 -240) EtxB), and the other codons specifying the signal peptide and 1-224 of etxA, 225-240 of ctxA, and the entire ctxB gene (to yield a hybrid toxin designated EtxA  CtxA (225-240) CtxB). The generation of these chimeric operons resulted in the introduction of a KpnI site, spanning codons 224 -226 of the A-subunit genes; this had no effect on the encoded amino acid sequence in CtxA, but changed codon 225 of EtxA to that normally found in CtxA, i.e. from an Asp to Gly, accounting for the designation of EtxA  CtxA (225-240) CtxB.
The two hybrid toxins were expressed and purified (Fig. 5,  lanes 4 and 5) using conditions identical to those employed for the wild-type and RDEL mutant toxins, above, then assayed for their capacity to induce electrogenic Cl Ϫ secretion by T84 cells. The hybrid toxin CtxA  EtxA (226 -240) EtxB was less potent than wild-type CtxAB at inducing Cl Ϫ secretion (Fig. 7A). Indeed the delay in the time course of Cl Ϫ secretion with the hybrid more closely resembled that of wild-type EtxAB (Fig.  7B). By contrast, the activity of the other hybrid, EtxA  CtxA (225-240) CtxB, more closely resembled that of wild-type CtxAB (Fig. 7, compare panels A and B). These findings demonstrated that a potent electrogenic response could be elicited by toxins in which the first 224 amino acids were derived from either CtxA or EtxA. Since this region contains the entire ADP-ribosylating A1-fragment (i.e. residues 1-192), we conclude that the differential toxicity of cholera toxin and E. coli enterotoxin cannot be due to differences in the enzymatic activities of the A-subunits. This conclusion was further substantiated by analysis of the intrinsic ADP-ribosyltransferase activity of both the wild-type and hybrid toxins (Table I). Agmatine was used as the substrate for ADP-ribosylation, and the ADP-ribosylating activity measured with or without recombinant hARF6. Comparison of wild-type cholera toxin with the various toxin constructs containing the CtxA1-fragment, revealed that they all had similar ADP-ribosyltransferase activity. The activities of wild-type E. coli enterotoxin and the hybrid toxin containing the Etx A1-fragment were found to be approximately 1.6 -2.8-fold higher than that exhibited by cholera toxin (Table I). Thus, the lower potency of E. coli entero-toxin in T84 cells cannot be attributed to a lower level of intrinsic enzymatic activity. Lee et al. (47) also noted that the activity of E. coli enterotoxin, in the agmatine assay, was slightly higher than that of cholera toxin (47), but the reasons for this remain unknown.
Taken together, the findings implicate either the C-terminal portion of the A2-fragment (from residues 226 -240) or the B-subunits of the two toxins, as important determinants of the difference in toxin activity. Given that the B-subunits of E. coli enterotoxin are more promiscuous in receptor binding than the B-subunits of cholera toxin, one possible explanation for the lower potency of EtxAB was that differential binding to receptors might influence the efficiency of toxin delivery into T84 cells.
To address this question, a third hybrid toxin was constructed. The EcoRV-EcoRI fragment from plasmid pATA14 was substituted for the same fragment in pRC19 to generate a hybrid composed of codons specifying the signal peptide and residues 1-236 of ctxA, 237-240 of etxA, and the entire etxB gene (Fig. 1). The resultant toxin comprising the entire A-subunit of cholera toxin with a C-terminal RDEL sequence and the entire B-subunit of E. coli enterotoxin, was designated Ctx-A (RDEL) EtxB. Since the presence of a KDEL or RDEL sequence at the C terminus of the A-subunit had no effect on the relative potency of the toxin in T84 cells (see above), the Ctx-A (RDEL) EtxB hybrid allowed us to determine whether the basis of differential toxicity is due to the C-terminal portion of the A2-fragment (from residue 226) or to the B-subunit.
b ADP-ribosyltransferase activity in the absence or presence of recombinant human ADP-ribosylation factor 6 (rhARF 6). The data points represent the mean Ϯ S.E., for two independent monolayers. Four independent experiments gave similar results. but are equally effective in triggering Cl Ϫ secretion in T84 cells, the differential toxicity of cholera toxin and E. coli enterotoxin cannot be associated with their respective B-subunits. By contrast, the data shown in Fig. 8 demonstrate that Ctx-A (RDEL) EtxB and CtxA  EtxA (226 -240) EtxB, which only differ in the C-terminal portion of the A-fragment, have toxicities characteristic of cholera toxin and E. coli enterotoxin, respectively. We therefore conclude that the structural basis for differential toxicity resides in the A2-fragment.
The sequences of CtxA (RDEL) EtxB and CtxA  EtxA (226 -240) EtxB only differ by 4 amino acids (Fig. 4ii): Asp-229, Ile-230, Thr-232, and His-233 (in CtxA) and Glu-229, Val-230, Ile-232, and Tyr-233 (in EtxA). The residues at these positions in Ctx-A (RDEL) EtxB are those normally found in the A-subunit of cholera toxin, whereas those present in CtxA  EtxA (226 -240) EtxB are normally found in the A-subunit of E. coli enterotoxin. We conclude that one or more of these residues influence toxin action in T84 cells. Given that the amino acid differences are in the A2-fragment that winds through the central pore of the B-pentamer (see Fig. 4i), a number of possible explanations for the difference in toxicity can be considered.
The A2-fragment from residues 225 to 240 in cholera and E. coli enterotoxin is the most structurally different region of the two toxins, as revealed by x-ray crystallography (4,6). In E. coli enterotoxin, this segment comprises an extended chain (residues 227-231), a small ␣-helix (232-236), and the last four RDEL residues. The extended chain and ␣-helix are located within the central pore of the B-pentamer, while the exact location and structure of the RDEL sequence remain uncertain due to lack of definition in the electron density. Many of the residues in this segment are involved in intersubunit salt bridge, hydrophobic, and hydrogen bond interactions with the B-subunits. These include Glu-229, Ile-232, and Tyr-233, which interact with specific residues in the B-subunit pentamer (4). The x-ray crystal structure of cholera toxin revealed that the A2-segment from residue 227 enters the central pore of the B-pentamer as an ␣-helix rather than as an extended chain (6). The last four (KDEL) residues were also visible in the electron density of cholera toxin and may reflect the more compact structure of the Ctx A2-than the Etx A2-fragment. Given these structural differences, a possible explanation for differential toxicity would be that amino acid changes between cholera toxin and E. coli enterotoxin in the A2-fragment alter the stability of A/B-subunit interaction, which may in turn influence maintenance of the holotoxin structure during its transport along the endocytic pathway or potentiate liberation of the A-subunit at a stage when translocation occurs.
To test the possibility that the four amino acid differences between CtxA (RDEL) EtxB and CtxA  EtxA (226 -240) EtxB had resulted in a change in holotoxin stability, each hybrid was subjected to a range of in vitro conditions that might normally be expected to cause protein denaturation. The toxins were bound to G M1 -ganglioside-coated microtiter plates, treated with denaturant, washed, and probed with monoclonal antibodies specific for CtxA or EtxB as described under "Experimental Procedures." The hybrids were found to be remarkably resistant to denaturants, maintaining their holotoxin structure in the presence of 8 M urea and at pHs as low as 4.0. These findings are not unexpected, given the exceptional stability of the cholera toxin and E. coli enterotoxin B-pentamers (11). When the hybrids were incubated in the presence of 0.5% (w/v) SDS at pH 7.2 for 30 min, the A-subunits were released from G M1 -bound B-subunits, and this was greater at pH 5.5 (Fig. 9). Of particular interest was the finding that CtxA (RDEL) EtxB (see H1 in Fig. 9) was more resistant than CtxA  EtxA (226 -240) EtxB (see H2 in Fig. 9) to A/B-subunit dissociation, under these conditions. Given that the toxins are likely to encounter a lower pH in endosomal compartments, and the A-subunits may well be subjected to conditions that promote unfolding during translocation, these in vitro conditions provide an important insight into holotoxin stability. Since Ctx-A (RDEL) EtxB is the more stable of the two toxins under these in vitro conditions, as well as being more active in inducing electrogenic Cl Ϫ secretion in T84 cells, we hypothesize that the difference in toxicity of cholera toxin and E. coli enterotoxin is due to the greater ability of cholera toxin to maintain its holotoxin structure during transit through the endocytic pathway and thus to deliver a greater proportion of internalized toxin to the site of translocation and action.
An alternative explanation for the difference in cholera toxin FIG. 9. Comparison of the stability of hybrid toxins which differ only in the C-terminal segment of the A2 polypeptide. The maintenance of A-subunit/B-subunit interaction in toxins bound to G M1 was determined for CtxA (RDEL) EtxB (designated H1) and CtxA-  EtxA (226 -240) EtxB (designated H2) after incubation for 30 min at room temperature in PBS, PBS containing 0.5% SDS, or McIlvaine buffer, pH 5.5, containing 0.5% SDS. The amount of pentameric EtxB present following such treatment was assessed by enzyme-linked immunosorbent assay using an anti-EtxB monoclonal antibody (118-8), whereas the amount of A-subunit associated with EtxB was assessed using an anti-CtxA monoclonal antibody (CT-17). Each sample was analyzed in duplicate (with the data representing the mean Ϯ S.E., where n ϭ 2). The figure is a representative data set from three independent experiments. and E. coli enterotoxin activity in T84 cells may be that the structural differences in the A2-fragments, highlighted above, cause a subtle but significant change in conformation of the K(R)DEL sequence. Since it is known that this sequence plays a role in the efficiency of toxin action in T84 cells, which presumably stems from its ability to engage the KDEL receptor involved in retrograde transport to the ER, it remains possible that the function of the KDEL sequence will be influenced by its position at the opening of the central pore of the B-pentamer. The absence of structural data on the RDEL sequence of E. coli enterotoxin does not allow a comparison to be made with that of the KDEL sequence of cholera toxin. It will clearly be important to determine whether the conformation of RDEL differs in CtxA (RDEL) EtxB and CtxA  EtxA (226 -240) EtxB. Crystallization of these two hybrids is in progress.
Finally, it has been proposed that the amphipathic nature of the A2-fragment may endow it with an ability to insert into eukaryotic cell membranes, presumably the ER membrane during A-subunit translocation (24), the occurrence of which may depend on unfolding and release of the A-subunit from the B-pentamer. Thus, the nature of interactions between the A2fragment and the B-pentamer may affect this step in the toxicity pathway.
Concluding Remarks-This study provides definitive evidence that the difference in activity of cholera toxin and E. coli heat-labile enterotoxin in T84 cells is due to structural elements in the C terminus of the A2-fragment. Preliminary x-ray crystallographic data on the CtxA (RDEL) EtxB hybrid have confirmed that the A2-tail penetrating the pore of the EtxB pentamer resembles that found in CtxAB rather than EtxAB. 4 This would be consistent with the view that the structural differences in the A2-fragments in CtxA (RDEL) EtxB versus EtxAB would alter the stability of A/B-subunit interactions and/or change the spatial location of the K(R)DEL motif. These studies also highlight the importance of a region in the toxin molecule that has hitherto been generally overlooked with respect to a role in toxin action. In this regard, the difference in toxicities of Ctx and Etx is likely to be of considerable significance in the pathobiology of cholera and traveler's diarrhea.