Interaction of ADP-ribosylation Factor with Escherichia coli Enterotoxin That Contains an Inactivating Lysine 112 Substitution*

Cholera toxin and Escherichia coli heat-labile enterotoxin (LT) exert their effects on cells through ADP- ribosylation of guanine nucleotide-binding proteins. Both toxins consist of one A subunit, which is an ADP-ribosyltransferase, and five B (or binding) subunits. Their enzymatic activities are latent; activation re- quires reduction and proteolysis, resulting in a catalytically active AI protein and a much smaller A, pro- tein. These ADP-ribosyltransferases are activated by GTP-dependent 20-kDa ADP-ribosylation factors or ARFs. To determine if proteolysis plus reduction is required for appearance of the ARF allosteric site as well as for catalytic activity, an inactive mutant of LT, LT(E112K), with replacement of glutamate by lysine at position A utilized a competitor in cholera toxin ADP-ribosyltransferase assays containing limiting amounts of ARF. and an inhibitory studies are consistent with the conclusion that the ARF site is not expressed in the latent Both

of transferase activity requires proteolysis near the carboxyl terminus and reduction of the single disulfide linking the two proteolytic fragments, resulting in the generation of an -23-kDa enzymatically active A1 protein (CTA1) and a 6-kDa Az protein (CTA,) derived from the carboxyl terminus (Mekalanos et al., 1979(Mekalanos et al., , 1983). E. coli heat-labile enterotoxin (LT), like cholera toxin, possesses A and B subunits and has similar enzymatic, immunological, and structural properties .
The A subunits of cholera toxin and LT and the S1 subunit of pertussis toxin share considerable amino acid identity (Rappuoli and Pizza, 1991). Similarities in the functional amino acids have been noted with other bacterial ADPribosyltransferases (Rappuoli and Pizza, 1991). For example, a glutamate residue participates in the pertussis toxin-, diphtheria toxin-, and Pseudomonas exotoxin A-catalyzed reactions and is covalently linked to the nicotinamide residue of NAD following irradiation Collier, 1984, 1987;Carroll et al., 1985;Barbieri et al., 1989;Cockle, 1989). Mutant forms of the toxins with replacements of the respective glutamate residues are inactive (Tweten et al., 1985;Barbieri et al., 1989;Tsuji et al., 1990). The critical glutamate in LT is believed to be at position 112, corresponding to 148 in diphtheria toxin, 533 in Pseudomonas exotoxin A, and 129 in pertussis toxin Collier, 1984, 1987;Barbieri et al., 1989;Tsuji et al., 1990). ADP-ribosylation by cholera toxin and E. coli enterotoxin is stimulated in vitro by a multigene family of -20-kDa guanine nucleotide-binding proteins, known as ADP-ribosylation factors or ARFs (for review, see Serventi et al. (1992)). ARFs are highly conserved proteins and have been found in all eukaryotic cells, from Giardia to mammals (Kahn et ai., 1988;Price et al., 1988;Sewell and Kahn, 1988;Tsai et al., 1991a;Tsuchiya et al., 1991;Murtagh et al., 1992). In the presence of GTP or its analogues, but not GDP, ARFs stimulate the ADP-ribosyltransferase activity of cholera toxin and E. coli heat-labile enterotoxin in a reaction enhanced by certain phospholipids and detergents (Bobak et al., 1990;Lee et al., 1991;Price et al., 1992). At least six mammalian ARFs, which have been identified by cDNA cloning, fall into three classes based on similarities in deduced amino acid sequence, size, and gene structure (Price et ai., 1988;Sewell and Kahn, 1988;Bobak et al., 1989;Monaco et al., 1990;Tsai et al., 1991b;Tsuchiya et al., 1991;Lee et al., 1992). ARFs from all three classes, when expressed as recombinant proteins in E. coli or insect cells, stimulate cholera toxin activity (Weiss et al., 1989;Kunz et al., 1990;Price et al., 1992). ARFs are allosteric activators of the active CTA, protein (Noda et al., 1989(Noda et al., , 1990. It is not known whether the ARF interaction site is present in the latent, inactive CTA or LTA 6383 species. In the studies reported here, we used an enzymatically inactive E. coli heat-labile enterotoxin containing a lysine for glutamate replacement at position 112 to determine whether latent forms of toxin were capable of interaction with ARF, specifically examining the effects of proteolysis and reduction on generation of an ARF binding site.

RESULTS AND DISCUSSION
Trypsinization of LT increased basal and ARF-stimulated NAD:agmatine ADP-ribosyltransferase activity ( Table I). As noted previously, glutamate 112 of LTA appears to be necessary for activity (Tsuji et al., 1990). Substitution of lysine for glutamate LT(E112K) resulted in a loss of transferase activity. As shown with LT and CT, trypsinization and reduction of LT(E112K) were associated with the generation of an A, protein ( Fig. 1) (Mekalanos et al., 1979;Moss et al., 1981).
To determine whether LT(E112K) interacted with ARF, its ability to inhibit ARF stimulation of CTA-catalyzed ADPribosylation was examined in assays in which concentrations TABLE I Effect of trypsinization of L T and LT(EI 12K) on basal and rARF6stimulated ADP-ribosyltransferase activity LT (10 pg) or LT(E112K) (60.8 pg) was incubated in 50 mM glycine (pH 8.0), 20 mM dithiothreitol for 30 min a t 30 "C with the indicated amount of trypsin in a total volume of 90 pl for LT or 190 pl for LT(E112K) before addition of 10 p1 of trypsin inhibitor (2 or 4 mg/ ml, respectively) following which samples (10 or 20 pl, respectively) were assayed as described under "Materials and Methods." Activity of CTA in this experiment was 3.2 and 25.8 pmol. min" without and with rARF6 (A6), respectively. LT(El12K) was inactive without or with rARF6. ND, not determined.

TABLE I1
Effect of trypsinization of LT(EI12K) on inhibition of rARF6stimulated cholera toxin ADP-ribosyltransferase activity L T (30 pg) or LT(E112K) (91.2 pg) was incubated in 50 mM glycine (pH 8.0). 20 mM dithiothreitol for 30 min a t 30 "C with or without trypsin (6 pg) (final volume of 285 pl) before addition of 15 pl of water or trypsin inhibitor (30 pg). Bovine serum albumin (BSA) and ovalbumin (OVAL) (96 pg each) were similarly treated. Samples (20 pl) were assayed with CTA without and with ARF. rA6, rARFG; T, trypsin; TI, trypsin inhibitor. Activity of trypsinized LT was 7.4 and 37.0 pmol.min", without and with rARF6, respectively. ND, not determined. of rARF6 were limiting. Native LT(E112K) was a relatively weak inhibitor of rARF6-stimulated CTA-catalyzed ADPribosylagmatine formation (Tables I and I1 and Fig. 2). Incubation with trypsin under reducing conditions, however, dramatically increased the inhibitory activity of LT(E112K) ( Table I). Reproducible inhibition of basal CTA activity was also observed (Table 11). Inhibition of CTA was not due to the presence of trypsin and/or trypsin inhibitor. Unrelated proteins (e.g. bovine serum albumin and ovalbumin) when substituted for LT(E112K) did not block rARF6-stimulated activity (Table 11) nor did cholera toxin B subunit when subjected to the same trypsin/trypsin inhibitor incubation procedure (data not shown). No inhibition was observed when the order of addition of trypsin and trypsin inhibitor was reversed (Table 111)
Inhibition by trypsinized LT(E112K) of rARF6-stimulated and basal activities was concentration-dependent ( Fig. 2). At equal protein concentrations, the inhibitory effects on rARF6stimulated activity were much more marked than those on basal activity; at concentrations of trypsinized LT(E112K) that completely blocked rARF6-stimulated activity, basal activity was slightly inhibited (Fig. 2). Increasing the concentrations of rARF6 overcame inhibition by trypsinized LT(E112K) (Fig. 3), as might be expected if the inhibitory effect resulted from competition between CTA and LT(El12K) for limiting amounts of rARF6.
Activation of LT requires both trypsinization and reduction and leads to the generation of LTA1; similar treatment of the inactive mutant resulted in the formation of LTA1(E112K), visualized after electrophoresis in sodium dodecyl sulfatepolyacrylamide gels run under reducing conditions (Fig. 1). These data suggest that trypsinization and reduction generate an LTAI(E112K) protein, which although enzymatically inactive due to replacement of a critical glutamate residue, still possesses a rARF6 binding site. Since thiol was present during trypsinization and in the control preparation, reduction of LT(E112K) by thiol was alone insufficient to generate an inhibitory molecule, consistent with the conclusion that proteolysis is necessary. Similarly, activation of LTA requires proteolysis and reduction.
To examine the thiol requirements, trypsinization of LT(E112K) was conducted with and without dithiothreitol (Table IV). Reactions were terminated with trypsin inhibitor, and the ability of the products to inhibit rARF6-stimulated CTA ADP-ribosyltransferase activity was assessed (Table  LT(E112K). A , no additions; B, trypsin; C, trypsin inhibitor following incubation; D, trypsin during incubation followed by trypsin inhibitor; E, no additions; F, trypsin inhibitor following incubation; G, trypsin during incubation; H , trypsin during incubation followed by trypsin inhibitor; I , trypsin inhibitor during incubation followed by trypsin. Samples (10 pl of LT and 20 pl of LT(E112K)) were assayed in the standard NAD:agmatine ADP-ribosyltransferase assay to verify that trypsinolysis had occurred (i.e. that LT was activated and LT(E112K) was inhibitory of ARF) and also to monitor the ADP-ribosylation of proteins. For the latter, assays (total volume of 300 p l ) contained 50 mM potassium phosphate (pH 7.5), 10 mM MgCl,, 0.1 mg/ml ovalbumin, 100 p~ [32P]NAD (2 pCi), 100 pM GTP, 20 mM dithiothreitol, and 0.003% SDS with CTA (0.25 pg) and/or rARF6 (1 pg), as indicated on the figure. After 90 min at 30 "C, 0.7 ml of 10.7% trichloroacetic acid was added, and the samples were refrigerated overnight. After centrifugation (1 h, -2000 X g), the supernatant was removed, and the pellet was dispersed in 30 p1 of water plus 30 pl of SDS sample mix (see legend to Fig. 1) and heated (75 "C, 10 rnin). A sample (50 pl) was subjected to electrophoresis in 18% polyacrylamide gels. Gels were stained in Coomassie Blue and exposed to x-ray film.

IV).
Thiol appeared not to be necessary for trypsinization of LT(E112K) ( Table IV). Trypsinized LT(E112K) interfered with activation of CTA by rARF6 but appeared also to inhibit basal activity, an effect that was not solely the result of trypsin and trypsin inhibitor in the preparation (Table IV). In these studies, the CTA assays were conducted with and without dithiothreitol to determine whether there was an endogenous source of thiol responsible for activation. A low level of activity was observed in the presence of rARF6 (Table IV), consistent with a source of thiol in that preparation but an amount insufficient to activate CTA significantly. LT activated by trypsin in the presence of thiol had slightly higher activity than LT activated in its absence (35.9 pmol. min" uers'sus 28.2 pmol. rnin"), but, again, thiol was not necessary for trypsinization under these conditions, consistent with the above result.
To determine whether the unreduced but nicked LT(E112K) containing the intrachain disulfide was active and possessed a rARF6 binding site, CT ADP-ribosyltransferase activity was assayed under conditions in which it was not dependent on thiol. Commercially available, nicked CTA was reduced and alkylated with iodoacetamide; the alkylated CTA, was purified and assayed. LT(E112K) was trypsinized in the absence of dithiothreitol, under conditions similar to those shown in Table IV. In contrast to nicked CTA, which required thiol for activity, presumably to reduce the intrachain disulfide, alkylated CTA, was thiol-independent and was still stimulated by rARF6 in a thiol-independent fashion ( Table V). Dithiothreitol appeared to slightly inhibit ADPribosyltransferase activity. LT(E112K) that was trypsinized but not reduced did not inhibit rARF6-stimulated activity ( Table V), consistent with the view that trypsinization and reduction are required to generate an ARF binding site. The requirements for inhibition of rARF6-stimulated activity by LT(E112K) are thus identical to those needed to generate an active A, protein, i.e. the appearance of an ARF binding site appears to coincide with formation of an active catalytic site (proteolysis and reduction of the single disulfide).
Without trypsinization, the toxin thus seems to be inactive, whether or not it is reduced, when assayed both catalytically and by its ability to interact with ARF. Proteolysis in the absence of reduction is insufficient to create a functional ARF-binding protein. Both proteolysis and reduction are  therefore necessary for activation and to generate an ARF binding site. The fact that basal ADP-ribosylagmatine formation was inhibited, but to a much lesser extent, suggests that toxin-toxin interactions may also be a factor in expression of basal ADP-ribosyltransferase activity.