Site-directed mutagenic alteration of potential active-site residues of the A subunit of Escherichia coli heat-labile enterotoxin. Evidence for a catalytic role for glutamic acid 112.

Escherichia coli heat-labile enterotoxin (LT) and the related cholera toxin exert their effects on eukaryotic cells through the ADP-ribosylation of guanine nucleotide-binding proteins of the adenylate cyclase complex. The availability of the crystal structure for LT has permitted the tentative identification of residues that lie within or are vicinal to a presumptive NAD(+)-binding site and thus may play a role in substrate binding or catalysis. Using a plasmid clone encoding the A subunit of LT, we have introduced substitutions at such potential active-site residues and analyzed the enzymatic properties of the resultant mutant analogs. Enzymatic analyses, employing both transducin and agmatine as acceptor substrates, revealed that substitutions at serine 61, glutamic acid 110, and glutamic acid 112 resulted in reduction of enzyme activity to < 10% of wild-type levels. Kinetic analyses indicated that alteration of these sites affected the catalytic rate of the enzyme and had little or no effect on the binding of either NAD+ or agmatine. Of the mutant analogs analyzed, only glutamic acid 112 appeared to represent an essential catalytic residue as judged by the relative effects on kcat and kcat/Km. The results provide formal evidence that glutamic acid 112 of the A subunit of LT represents a functional homolog or analog of catalytic glutamic acid residues that have been identified in several other bacterial ADP-ribosylating toxins and that it may play an essential role in rendering NAD+ susceptible to nucleophilic attack by an incoming acceptor substrate.

Escherichia coli heat-labile enterotoxin (LT) and the related cholera toxin exert their effects on eukaryotic cells through the ADP-ribosylation of guanine nucleotide-binding proteins of the adenylate cyclase complex. The availability of the crystal structure for LT has permitted the tentative identification of residues that lie within or are vicinal to a presumptive NAD ؉ -binding site and thus may play a role in substrate binding or catalysis. Using a plasmid clone encoding the A subunit of LT, we have introduced substitutions at such potential active-site residues and analyzed the enzymatic properties of the resultant mutant analogs. Enzymatic analyses, employing both transducin and agmatine as acceptor substrates, revealed that substitutions at serine 61, glutamic acid 110, and glutamic acid 112 resulted in reduction of enzyme activity to <10% of wild-type levels. Kinetic analyses indicated that alteration of these sites affected the catalytic rate of the enzyme and had little or no effect on the binding of either NAD ؉ or agmatine. Of the mutant analogs analyzed, only glutamic acid 112 appeared to represent an essential catalytic residue as judged by the relative effects on k cat and k cat /K m . The results provide formal evidence that glutamic acid 112 of the A subunit of LT represents a functional homolog or analog of catalytic glutamic acid residues that have been identified in several other bacterial ADP-ribosylating toxins and that it may play an essential role in rendering NAD ؉ susceptible to nucleophilic attack by an incoming acceptor substrate.
Escherichia coli heat-labile enterotoxin (LT) 1 is a member of a family of bacterial ADP-ribosylating toxins that possess an A-B-type structure. LT is structurally very similar to cholera toxin (CT) and is composed of an enzymatically active A subunit (M r 28,000) and five B subunits that form a homopentameric B oligomer (M r 55,000) (reviewed in Refs. 1 and 2). The B pentamer is responsible for binding of the toxins to gangliosides on the surface of eukaryotic cells (3), while the A subunit catalyzes the transfer of ADP-ribose from NAD ϩ to regulatory guanine nucleotide-binding proteins (G s␣ ) of the adenylate cyclase complex. ADP-ribosylation of G s␣ results in constitutive activation of adenylate cyclase, elevation of intracellular cAMP levels, and, ultimately, the disruption of the normal electrolytic balance in intestinal cells characteristic of diarrheal disease (4).
Apart from their role as important virulence determinants, both LT and CT have been the subject of considerable study in recent years because they are also potent enhancers of mucosal immune responses (5)(6)(7). Therefore, there has been considerable interest in elucidating the structure-function relationships of LT and CT with respect to enzymatic and toxic activity. Such efforts have been predicated on the notion that identification of residues or motifs involved in the ADP-ribosyltransferase activity will provide suitable targets for the construction of genetically detoxified analogs for inclusion in acellular vaccines and will further permit systematic investigation of precise relationships between enzymatic activity, toxicity, and the immunopotentiating activity. Several studies, using either random or site-directed mutagenesis of the LT gene, have identified residues that appear to be important to the ADPribosyltransferase activity of the A subunit (8 -13). Among these residues are included Arg-7, Glu-110, and Glu-112. Relatively conservative substitutions of these residues (e.g. R7K, E110D, and E112D) result in marked attenuation of ADPribosyltransferase activity and in the formation of holotoxin molecules that are largely devoid of toxic activity when assayed on eukaryotic cells in vitro (11,12). Either Glu-110 or Glu-112 of LT likely represents the functional equivalent of active-site Glu residues that have been identified in other ADP-ribosylating toxins including diphtheria toxin (DT), Pseudomonas aeruginosa exotoxin A (ETA), and pertussis toxin (PT) (14 -16). Specific catalytic roles for such Glu residues in the enzymatic activities of DT, ETA, and PT were originally indicated by studies using UV-induced photolabeling with NAD ϩ (17)(18)(19)(20). The crystal structures of DT, ETA, and PT have also shown that these glutamic acid residues reside in grooves or clefts that appear to constitute the NAD ϩ -binding sites in this class of toxins (21)(22)(23). Kinetic analyses of mutant analogs have demonstrated that such Glu residues possess a catalytic role in DT, ETA, and PT (14 -16). A presumptive NAD ϩ -binding site has also been identified in the refined crystal structure of the A subunit of LT, and Arg-7, Glu-110, and Glu-112 are found to reside within or very near this site (2,24).
Recently, Pizza et al. (11) reported the effects of a number of site-directed mutations in the A subunit of LT on the toxic and enzymatic activities of the holotoxin. The mutations were designed to affect residues in or vicinal to the NAD ϩ -binding site identified in the crystal structure. The results confirmed that substitutions at Arg-7, Glu-110, and Glu-112 markedly atten-uate enzymatic and toxic activities. A model was proposed in which Arg-7 is directly involved in NAD ϩ binding, while Glu-112 is involved in binding to and/or orienting the incoming acceptor substrate for subsequent catalysis. However, the mechanistic basis for this model has not been investigated by kinetic analyses or other means. Using a rationale based on primary and secondary structural comparisons to other toxins and analysis of the proposed NAD ϩ -binding site in the crystal structure, we have constructed and analyzed the enzymatic properties of various mutant analogs of the A subunit of LT. The results provide evidence that Glu-112 has a catalytic role in ADP-ribosyltransferase activity and is likely involved in the hydrolysis of the glycosidic linkage between the ribose and nicotinamide groups of NAD ϩ .

EXPERIMENTAL PROCEDURES
Materials-DNA-modifying enzymes were purchased from New England Biolabs Inc. and Life Technologies, Inc. and were used according to the recommendations of the supplier. [adenylate-32 P]NAD ϩ (20 -30 Ci/mmol) and [carbonyl-14 C]NAD ϩ (41-54 mCi/mmol) were obtained from DuPont NEN and Amersham Corp., respectively. Recombinant bovine ADP-ribosylation factor I (rARF-I) was purified from E. coli BL21(DE3) harboring plasmid pOW12 (25) as described (13). Plasmid pOW12 was kindly provided by Dr. Richard Kahn (NCI, Bethesda, MD). Rod outer segment (ROS) membranes were prepared from frozen bovine retinas as described previously (13). The monoclonal antibodies CP7-3003F7 and CP7-3004GX1 were provided by Dr. James Kenimer (Food and Drug Administration, Bethesda, MD). A polyclonal anti-A subunit antiserum was provided by Dr. Scott Manning (University of Montana, Missoula, MT). All other reagents were obtained from Calbiochem, Boehringer Mannheim, or Sigma Bacterial Strains and Expression Plasmids-E. coli DH5␣ was obtained from Life Technologies, Inc., and phagemid pTZ18R was obtained from Pharmacia Biotech Inc. The construction of recombinant expression plasmid pTZrLTA has been described in detail (13). pTZr-LTA encodes the entire mature A subunit of LT with a 9-amino acid amino-terminal fusion peptide derived from vector and signal peptide coding sequences.
Site-directed Mutagenesis of the A Subunit Gene in pTZrLTA-The DNA sequence of the LT operon from pEWD299 (26) as reported by Yamamoto et al. (27) was used as the basis for all sequence manipulations. All plasmid vectors were treated with calf intestinal alkaline phosphatase following digestion with restriction enzymes and prior to insertion of heterologous DNA. Construction of the plasmids encoding A subunits with substitution of Asp for Glu at positions 110 (rLTA/ E110D) and 112 (rLTA/E112D) has been described previously (12). Other mutations were introduced into the coding sequence for rLTA by either the procedure of Kunkel (28) or by DNA amplification procedures. DNA amplifications were performed using Pfu DNA polymerase (Stratagene) and conditions recommended by the supplier. Oligonucleotide primer pairs encoding the desired mutations were used in separate reactions with primers 5Ј and 3Ј of the target mutation site. The resultant fragments, which contained sequences overlapping the target site, were purified from agarose gels. The purified fragments were mixed, denatured, and subjected to one to five rounds of amplification in the absence of any other primers, thereby allowing the amplification of the full-length fragment containing the desired mutation. Primers 5Ј and 3Ј of the target mutation site were then added, and amplification was allowed to proceed for 30 cycles. The resultant amplified products were then purified from agarose gels prior to digestion with appropriate restriction endonucleases and insertion into pTZrLTA that had been digested with the identical enzymes. The plasmids were then transformed into E. coli DH5␣ by electroporation (29). Plasmid DNA was extracted from 5 to 10 electrotransformants and subjected to DNA sequence analysis to confirm the presence of the desired mutations and the fidelity of the amplifications. For expression of mutant genes, the plasmids were transformed into E. coli BL21(DE3) by electroporation.
Purification of rLTA and Mutant Analogs-Genes encoding rLTA and mutant analogs were expressed under control of the T7 promoter in E. coli BL21(DE3) by induction with isopropyl-␤-D-thiogalactopyranoside. Briefly, bacterial cultures were grown to an absorbance at 600 nm of 0.6. Isopropyl-␤-D-thiogalactopyranoside was added to a final concentration of 1 mM, and incubation was continued for 2-4 h. The cells were harvested by centrifugation; resuspended in 1 mM phenylmethylsulfonyl fluoride, 20 mM Tris-HCl, pH 7.5; and then sonicated and extracted with 8 M urea as described previously (13). The urea extracts were dialyzed exhaustively against 20 mM Tris-HCl, pH 7.5, and samples containing 15-20 mg of protein were fractionated by anion-exchange chromatography using a fast protein liquid chromatography system equipped with a Mono-Q HR 5/5 column (Pharmacia Biotech Inc.). The column was eluted with a gradient of 0 -0.4 M NaCl in 20 mM Tris-HCl, pH 7.5, at a flow rate of 1 ml/min. Column fractions (1 ml) were analyzed for the presence of recombinant subunits by SDS-polyacrylamide gel electrophoresis (PAGE) using the buffer system described by Laemmli (30). Fractions containing the recombinant subunits were pooled, dialyzed against 50 mM ammonium bicarbonate, and dried under vacuum in aliquots. The relative purity and amount of each of the recombinant subunits contained in the preparations were determined by quantitative densitometric scanning of proteins separated by SDS-PAGE and stained with Coomassie Brilliant Blue R-250.
ADP-ribosyltransferase Assays-NAD ϩ :agmatine ADP-ribosyltransferase activity was measured as the ability to catalyze the release of nicotinamide from [carbonyl-14 C]NAD ϩ in the presence of rARF-I. Reaction mixtures (50 l) containing 10 -20 g/ml recombinant A subunit, 10 mM dithiothreitol, 100 M GTP, 80 g/ml rARF-I, 4 mM dimyristoylphosphatidylcholine, 4 mM MgCl 2 , 30 mM agmatine, and 50 mM potassium phosphate, pH 7.5, were incubated at 30°C for 1 h. The amount of nicotinamide released was determined using Dowex AG 1-X8 resin as described previously (12,13). The ability to ADP-ribosylate the ␣-subunit of transducin in bovine ROS membranes was assayed using the recombinant subunits at a final concentration of 25 g/ml as described (13).
Initial rate data for the NAD ϩ :agmatine ADP-ribosyltransferase activity were collected under conditions where either the NAD ϩ or agmatine concentration was held constant at saturating conditions and the concentration of the other substrate was varied from ϳ0.1 to 2.0 estimated K m . Kinetic parameters were determined using Lineweaver-Burk plots of the initial velocity data. Kinetic analyses of mutant analogs exhibiting very low levels of activity were facilitated by simultaneously increasing the amount of subunit analyzed, the time of the assay, and the specific radioactivity of the NAD ϩ employed as substrate all by 3-4-fold. Similar alterations in assay conditions were used to measure the rates of NAD ϩ glycohydrolysis in the absence of acceptor substrate.
Other Analytical Methods-Protein determinations were done using the bicinchoninic acid method and bovine serum albumin as the standard (31). Plasmid DNA sequencing was done using the primer extension dideoxy chain termination method (32) using modified T7 DNA polymerase (Sequenase, U. S. Biochemical Corp.). Immunoblotting, using a combination of monoclonal and rabbit polyclonal anti-A subunit antibodies and horseradish peroxidase-linked anti-mouse IgG and antirabbit IgG, was done as described previously (12).

Construction of A Subunit Mutants-Various mutations
were introduced into the A subunit gene to assess the contribution of specific residues to enzymatic activity and conformational integrity. The rationale for the selection of the mutations to be introduced and that are listed in Table I follows. In the crystal structure, the side chain of Arg-54 extends into the NAD ϩ -binding cleft and closely interacts with other residues shown to be important to enzymatic activity (24). Arg-54 makes a salt bridge with both Glu-110 and Glu-112 and is also involved in hydrogen bonding with Arg-7. Ser-61 lies within a region of sequence similarity with the S1 subunit of PT and also lies within the presumptive NAD ϩ -binding cleft of the A subunit (16,24). Substitution of this residue with Phe has been found to abrogate the toxicity of LT (10). Ala-69 is positioned in a loop that is in close proximity to the NAD ϩ -binding cleft, and His-70 lies within an ␣-helix that forms part of the NAD ϩbinding site (24). The other His residue that may be of importance is His-44, which is located in a ␤-sheet that is in close proximity to the NAD ϩ -binding site and has been proposed to be functionally equivalent to a His residue (His-35) that appears to play a role in NAD ϩ binding in PT (2,24). Substitutions at both Glu-110 and Glu-112 have been previously shown to markedly reduce ADP-ribosyltransferase activity, and both residues appear to be located in or near the entrance to the NAD ϩ -binding site (8,12,24). Tryptophans appear to play an important role in NAD ϩ binding by DT and PT as judged by studies using both NAD ϩ -induced quenching of intrinsic tryptophan fluorescence and site-directed mutagenesis (33,34). Like the important Trp residues of DT and ETA, Trp-127 and Trp-174 of the A subunit of LT both occur in ␤-turns (24).
The mutagenic substitutions were introduced into the coding sequence for the protein designated rLTA, which possesses a 9-amino acid amino-terminal fusion peptide. We have previously shown that the presence of this peptide has little or no influence on enzymatic activity as judged by comparison to the activity exhibited by a purified recombinant subunit that lacks the fusion peptide (13).
Partial Purification and Characterization of rLTA and Mutant Analogs-After extraction with urea, dialyzed crude extracts containing rLTA and mutant derivatives were fractionated by anion-exchange chromatography. The fractions containing the recombinant A subunits were pooled and analyzed for purity by SDS-PAGE (Fig. 1). The majority of mutant analogs could be at least partially purified by this method; however, several analogs, including those containing substitutions at His-44, could not be purified by anion-exchange chromatography (data not shown). The purity of the subunits ranged from 70 to 95% as judged by densitometric scanning of the gels. To screen for conformational integrity of the mutant analogs, samples were subjected to limited trypsinolysis and analyzed by SDS-PAGE and immunoblotting. As shown in Fig.  2, the majority of the mutants yielded a fragmentation pattern that was qualitatively similar to the pattern exhibited by wildtype rLTA (i.e. production of the M r 23,000 A1 subunit). However, the subunits containing substitutions at Trp-174 exhibited enhanced sensitivity to trypsin and were largely degraded by limited proteolytic treatment, suggesting that the alterations imparted a significant conformational change to the molecule. Only those mutants that appeared to retain the wildtype tryptic fragmentation pattern were subjected to further analyses.
ADP-ribosyltransferase Activity of Mutant Subunits-To provide an initial measure of the relative activity of the mutant analogs, we assayed the ability of the proteins to catalyze the ADP-ribosylation of the ␣-subunit of transducin in bovine ROS membranes (Fig. 3). We have previously shown that trypsin treatment of rLTA, in contrast to LT, resulted in only a 2-3-fold increase in specific activity (13). To avoid the potential variation due to minor differences in sensitivity to trypsinolysis, all subunits were assayed without trypsin pretreatment. With the exception of the Glu to Asp substitutions at positions 110 and 112 (rLTA/E110D and rLTA/E112D) and substitution of Ser-61   (30) and heating at 95°C for 5 min. Aliquots containing 7.5 g of protein were then electrophoresed on 12.5% resolving gels and electroblotted to polyvinylidene fluoride membranes. The A and A1 peptides were then detected using anti-A subunit antibodies as described under "Experimental Procedures." The positions of the molecular mass standards are shown on the left and are given in kilodaltons.

FIG. 3. ADP-ribosyltransferase activity of rLTA and mutant analogs.
Purified rLTA and mutant analogs were incubated with bovine ROS membranes, [adenylate-32 P]NAD ϩ and other additions as described (12,13). The indicated recombinant subunits were assayed at a final concentration of 25 g/ml for 2 h at 30°C. The labeled reaction products were then separated by SDS-PAGE using 12.5% resolving gels. The gels were stained with Coomassie Brilliant Blue R-250 and then dried onto Whatman No. 3MM filter paper prior to autoradiography using Kodak X-Omat radiographic film. The positions of the molecular mass standards are shown on the left and are given in kilodaltons. The position of the ␣-subunit of transducin (T␣) is also shown. The autoradiographs shown were obtained after 2.5-h exposures.
with Thr (rLTA/S61T), the majority of alterations did not appear to diminish the activity of the A subunits appreciably as judged by the relative intensities of the autoradiographic bands corresponding to ADP-ribosylated transducin-␣ (Fig. 3). However, rLTA/A69G and rLTA/A69V exhibited decreased levels of auto-or self-ADP-ribosylation as indicated by the reduction in relative intensity of the autoradiographic band migrating at M r ϳ29,000 (which represents the A subunit), suggesting that the mutations resulted in an alteration in substrate specificity. In contrast, rLTA/E110D, rLTA/E112D, and rLTA/S61T were impaired in both ADP-ribosylation of transducin and auto-ADP-ribosylation.
To provide a more quantitative estimate of the relative activity of the mutants, they were assayed for their ability to ADP-ribosylate a small guanidino compound, agmatine, in the presence of rARF-I (Table I). The results generally confirmed the findings using ROS membranes and further revealed that some of the mutants, notably rLTA/R54G, rLTA/R54K, and rLTA/H70N, exhibited enhanced specific NAD ϩ :agmatine ADP-ribosyltransferase activity. Initial rate kinetic measurements using variable NAD ϩ concentrations indicated that, in each of the cases in which enhanced activity was observed, the primary effect was on k cat (data not shown). The only mutant analogs that appeared to possess significantly reduced enzymatic activity were rLTA/S61T, rLTA/E110D, and rLTA/ E112D. All three of these mutant analogs exhibited Ͻ10% of the wild-type level of NAD ϩ :agmatine ADP-ribosyltransferase activity.
Initial Rate Kinetics-Kinetic analyses of the three mutant analogs that possessed Ͻ10% of the wild-type NAD ϩ :agmatine ADP-ribosyltransferase activity (rLTA/S61T, rLTA/E110D, and rLTA/E112D) were performed to help determine the mechanistic basis for the observed reductions (Table II). Examination of initial rates with increasing NAD ϩ concentrations revealed that the reactions catalyzed by all three mutants were associated with decreased k cat values, with little or no effect on the K m for NAD ϩ . The catalytic efficiency of each of the mutant analogs, as evidenced by the second-order rate constant (k cat / K m ), was correspondingly decreased for each mutant, but was most affected in the case of rLTA/E112D (by a factor of ϳ100). The k cat /K m values associated with rLTA/S61T and rLTA/ E110D were reduced by factors of 8.4 and 23, respectively. In the presence of increasing agmatine concentrations, the activities of rLTA/S61T and rLTA/E112D were again associated with diminished k cat values and exhibited little or no alteration in the K m for agmatine. Interestingly, rLTA/E110D exhibited a significantly reduced K m for agmatine (by a factor of ϳ30). The value of k cat /K m associated with rLTA/E112D was reduced by a factor of 110, while the k cat /K m values for rLTA/S61T and rLTA/E110D were only reduced by factors of 10.5 and 2.6, respectively.
NAD ϩ Glycohydrolase Activity of rLTA and rLTA/E112D-rLTA also displays measurable NAD ϩ glycohydrolase activity in the absence of added acceptor substrate (agmatine) ( Table  III). This assay revealed that the k cat and k cat /K m values for this activity are only 10-fold less than those for the NAD ϩ : agmatine ADP-ribosyltransferase activity (compare Tables III  and II). As with the NAD ϩ :agmatine ADP-ribosyltransferase activity, the NAD ϩ glycohydrolase activity is enhanced by the addition of rARF-I (15-20-fold) and is almost entirely dependent upon the presence of a reducing agent (data not shown). The magnitude of the velocity of the NAD ϩ glycohydrolase activity and similarity in cofactor requirements to the NAD ϩ : agmatine ADP-ribosyltransferase activity suggest that the glycohydrolase reaction is likely to be relevant to the mechanism of ADP-ribosylation of substrates other than water. As shown in Table III, measurement of the NAD ϩ glycohydrolase activity of rLTA/E112 revealed that the glycohydrolase activity associated with this mutant was decreased by a factor of ϳ35, suggesting an important role for Glu-112 in catalyzing the hydrolysis of the N-glycosidic bond of NAD ϩ .

DISCUSSION
Several studies have now been published that have attempted to identify residues that are important to the ADPribosyltransferase activity or toxicity of LT by mutagenesis. Using random mutagenesis, Tsuji et al. (8,9) and Harford et al. (10) reported that substitutions at Glu-112 (with Lys) and Ser-61 (with Phe), respectively, could abrogate toxicity. We have previously reported that crude extracts containing recombinant A subunit mutant analogs possessing substitutions at Arg-7, Glu-110, and Glu-112 exhibited markedly reduced ADPribosyltransferase activity (12), and more recently, Pizza et al. (11) have confirmed these findings by examining the toxicity of holotoxins containing alterations at these and other positions. While substitution of Glu-110 and/or Glu-112 has been shown to attenuate enzymatic or toxic activity in several experimental systems, the mechanistic bases for the reductions using kinetic or other biophysical methods have not been investigated. We therefore sought to characterize in detail the enzymatic characteristics associated with alterations at residues that have been predicted by crystallographic analysis to be potentially important to the enzymatic activity of LT. Since ARF is known to enhance both the catalytic rate and substrate affinity of CT and, by inference, LT (35), the availability of rARF-I in purified form (25) has permitted more detailed and accurate quantitative enzymatic analyses than had been performed in prior studies of mutant analogs (9,11,12). The enzymatically active fragments of DT, ETA, and PT all contain a catalytically important Glu residue, and in all three cases, the essential Glu was initially shown to be at or near the NAD ϩ -binding site by direct photolabeling with radiolabeled NAD ϩ (17)(18)(19)(20). A role in catalysis was established by kinetic analyses of mutant analogs (14 -16). We have not yet been able to efficiently photolabel a specific residue in the A subunit of LT using unmodified NAD ϩ and ultraviolet light, presumably owing to the low affinity of the enzyme for NAD ϩ (millimolar range) under presently used conditions. As noted, previous studies showed that substitution of both Glu-110 and Glu-112 in the A subunit resulted in marked reductions in enzymatic activity (11,12). The results of our kinetic analyses strongly support a catalytic role for Glu-112 in the mechanism of the ADP-ribosyltransferase reaction and indicate that Glu-110 is unlikely to play a specific role in the reaction mechanism. Substitution of Glu-112 with Asp results in a marked reduction of k cat (by a factor of 100), with little or no effect on the K m for NAD ϩ or agmatine. These results are similar to those of Wilson et al. (14) using the enzymatically active A fragment of DT. In the case of the DT A fragment, the K m value for neither NAD ϩ nor elongation factor 2 was affected by substituting Asp for Glu-148, but the k cat of the reaction was reduced by a factor of 100. However, the NAD ϩ glycohydrolase reaction catalyzed by the DT A fragment in the absence of acceptor substrate was relatively unaffected by substitutions at Glu-148. Coupled with the observation that the glycohydrolase reaction catalyzed by DT proceeds at a rate that is ϳ3 orders of magnitude less than the rate of ADP-ribosylation of elongation factor 2, this finding has been interpreted to indicate that the NAD ϩ glycohydrolase reaction may occur through a different mechanism than that of ADP-ribosylation of elongation factor 2 and perhaps involves strain or distortion effects on the NAD ϩ molecule that result in scission of the N-glycosidic bond (14). Based on these findings, Wilson et al. (14) proposed that the carboxyl group of Glu-148 participates in catalysis both by acting as a general base in the abstraction of a proton from the incoming diphthamide residue in a displacement reaction and by maintaining the geometry of the active site through hydrogen bonding interactions.
In contrast to DT, the catalytic rate of NAD ϩ glycohydrolysis catalyzed by the S1 subunit of PT is only 10-fold less than that of the ADP-ribosylation of the ␣-subunit of transducin. Substitution of Glu-129 with Asp results in a marked (by a factor of Ͼ200) decrease in NAD ϩ glycohydrolase activity (16). These findings have been interpreted to indicate a catalytic role for Glu-129 in the glycohydrolase reaction catalyzed by the S1 subunit, perhaps acting as a general base in the stabilization of a developing oxycarbonium-like intermediate in an S n 2-type mechanism as proposed by Locht and colleagues (16,36). Our finding that the catalytic rate of the NAD ϩ glycohydrolase activity of the A subunit of LT is only 10 -20-fold less than that of the NAD ϩ :agmatine ADP-ribosyltransferase activity suggests that the reaction mechanism of LT is similar to that associated with the S1 subunit of PT. The finding that substitution of Glu-112 also results in a marked decrease in NAD ϩ glycohydrolase activity suggests that this residue and glutamic acid 129 of the S1 subunit have similar roles in the reaction mechanism. However, the observation that the NAD ϩ glycohydrolase activity of the A subunit is less affected (by 3-fold) when compared with the NAD ϩ :agmatine ADP-ribosyltransferase activity by substitution of Glu-112 might indicate a contribution of alternative mechanisms, like that cited above for DT (14), in the hydrolysis of NAD ϩ .
The similarity between the enzymatic mechanisms of the A subunit of LT and the S1 subunit of PT is also supported by the effects of mutations on other residues that appear to be conserved among the two toxins. The active subunits of LT, PT, and the mosquitocidal toxin from Bacillus sphaericus (37) can be aligned to reveal the apparent conservation of several residues that have been shown by mutagenesis studies to be important for the retention of enzymatic activity (38). In the A subunit of LT, these include His-44, Ser-61, and Glu-112. Various substitutions at the positionally equivalent histidine (His-35) in the S1 subunit of PT markedly reduce the k cat in both the NAD ϩ glycohydrolase and ADP-ribosyltransferase reactions and have little or no effect on the K m for either NAD ϩ or acceptor substrates (38,39). We have found that substitution of His-44 with Arg, Gln, or Asn in the A subunit of LT results in substantial loss of activity when the mutant subunits are assayed in unpurified or crude form; however, these results must be interpreted with caution since these mutants also exhibit enhanced sensitivity to limited trypsinolysis. 2 Accordingly, the potential role of His-44 will await development of alternative purification schemes and detailed kinetic analyses. Our results do, however, appear to formally exclude any important role for His-70 in the enzymatic mechanism since none of the three substitutions at this position decreased activity. The current analyses also suggest that Ser-61 of the LT A subunit does not play an essential role in the catalytic mechanism per se since the k cat and k cat /K m values for the NAD ϩ :agmatine ADP-ribosyltransferase reaction catalyzed by rLTA/S61T were only reduced by a factor of 10, and the K m values for NAD ϩ and agmatine were relatively unaffected. A similar finding has been made with respect to the role of the positionally equivalent Ser residue (Ser-52) in the NAD ϩ glycohydrolase reaction catalyzed by the S1 subunit of PT (16). Furthermore, we have recently isolated a mutant analog containing an Ala substitution at this position and have found that it retains ϳ10% of the wild-type NAD ϩ :agmatine ADP-ribosyltransferase activity. 3 Therefore, Ser-61 likely plays a role in maintaining the overall geometry of the active site as this residue is observed to participate in hydrogen bonding to Arg-9 in the crystal structure (24). Domenighini et al. (40) have proposed a tentative model of the enzymatic mechanism of LT based on the crystal structure of LT. In this model, Arg-7 is proposed to participate directly in the binding of NAD ϩ at the active site. The role of Arg-7 in binding is supported by the observation that alteration of the positionally equivalent residue in the S1 subunit of PT (Arg-9) results in abrogation of UV-induced photolabeling with NAD ϩ , supporting the presumption that Arg-7 of the LT A subunit is involved in the productive binding of NAD ϩ rather than in catalysis per se (41). Mutagenic substitution of His-21 of DT, which is positionally equivalent to Arg-7 of LT in the two crystal structures, results in marked increases in the K m for NAD ϩ , with little or no effect on k cat (42,43). Although we have not been able to directly examine Arg-7 mutant analogs by photolabeling or other means, we favor the interpretation that Arg-7, like His-21 of DT, participates directly in NAD ϩ binding. However, it should be noted that substitution of His-21 of DT with Arg reduces the affinity for NAD ϩ substantially, suggesting potential mechanistic differences in the roles of the positionally equivalent His and Arg residues in the two toxins (42,43). In addition, substitution of Arg-7 with Lys results in conformational perturbation of the A subunit as judged by alteration of trypsin sensitivity and ability to support complete assembly of the holotoxin (11,12), and unlike Pizza et al. (11), we have been unable to isolate intact holotoxin molecules containing the Arg-7 to Lys substitution. 4 Domenighini et al. (40) also proposed that Glu-112 of the A subunit of LT, based on its position within the NAD ϩ -binding cavity, is involved in the interaction with the incoming acceptor substrate, either through binding or by stabilizing the formation of a nucleophilic transition state, similar to one proposed role for Glu-148 of DT (14). As noted above, our results support the notion that Glu-112 is a catalytic residue that likely participates in the formation of a transitional form (oxycarbonium-like intermediate) of NAD ϩ that is capable of reacting with an incoming nucleophile.
An important caveat concerning the identification of potential active-site residues resides in the fact that the published crystal structure of LT is that of a molecule that is essentially enzymatically inactive (not proteolytically nicked, unreduced, without bound ARF), although a more recent analysis indicates that the A subunit of trypsin-cleaved LT has essentially the same structure of the untreated molecule (44). Therefore, the precise location and orientation of various active-site amino acids in the enzymatically active conformation are not likely to be accurately reflected in the current structure. A more precise and informative picture of the geometry of the active site will likely be revealed by the crystal structure of the isolated A subunit since, as shown here and elsewhere (13), the A subunit does not require proteolysis for expression of activity, and it can be maintained in an active conformation by reduction and alkylation (35). The specific activities and kinetic parameters we have obtained for rLTA in the NAD ϩ :agmatine ADP-ribosyltransferase reaction compare favorably with those reported for LT, CT, and the conventionally isolated CT A subunit (13,35,45). Accordingly, efforts to crystallize the purified recombinant A subunit in reduced form are currently underway. Such efforts may also permit characterization of the interaction of ARF with the A subunit since expression of the ARF-binding site in LT requires prior activation by proteolysis and reduction (46).