A catalytic loop within Pseudomonas aeruginosa exotoxin A modulates its transferase activity.

Mutagenesis techniques were used to replace two loop regions within the catalytic domain of Pseudomonas aeruginosa exotoxin A (ETA) with functionally silent polyglycine loops. The loop mutant proteins, designated polyglycine Loops N and C, were both less active than the wild-type enzyme. However, the polyglycine Loop C mutant protein, replaced with the Gly(483)-Gly(490) loop, showed a much greater loss of enzymatic activity than the polyglycine Loop N protein. The former mutant enzyme exhibited an 18,000-fold decrease in catalytic turnover number (k(cat)), with only a marginal effect on the K(m) value for NAD(+) and the eukaryotic elongation factor-2 binding constant. Furthermore, alanine-scanning mutagenesis of this active-site loop region revealed the specific pattern of a critical region for enzymatic activity. Binding and kinetic data suggest that this loop modulates the transferase activity between ETA and eukaryotic elongation factor-2 and may be responsible for stabilization of the transition state for the reaction. Sequence alignment and molecular modeling also identified a similar loop within diphtheria toxin, a functionally and structurally related class A-B toxin. Based on these results and the similarities between ETA and diphtheria toxin, we propose that this catalytic subregion represents the first report of a diphthamide-specific ribosyltransferase structural motif. We expect these findings to further the development of pharmaceuticals designed to prevent ETA toxicity by disrupting the stabilization of the transition state during the ADP-ribose transfer event.

People suffering from AIDS, major burns, cancer, or cystic fibrosis often succumb to an infection of Pseudomonas aeruginosa. This aerobic, motile, Gram-negative bacterium survives in most environments, including water, soil, vegetation, and hospitals. P. aeruginosa secretes the highly virulent protein exotoxin A (ETA), 1 which is largely responsible for the local tissue damage, bacterial invasion, and possibly immunosuppression within the eukaryotic host (1). This toxin is extremely lethal, possessing an LD 50 of 0.2 g/animal upon intraperitoneal injection into mice (2). For these reasons, it is often referred to as the most potent virulence factor of this organism (1).
ETA is a 66-kDa extracellular protein (6) that, upon infection, is internalized into the eukaryotic cell so that it can catalyze the ADP-ribosylation of eukaryotic elongation factor-2 (eEF-2) (7). This ADP-ribosylation reaction inactivates eEF-2 through the covalent attachment of the ADP-ribosyl moiety to the diphthamide residue of eEF-2, which results in cessation of protein synthesis and eventually cell death (6). Although the NAD ϩ substrate-docking site within ETA is well characterized (8,9), the eEF-2 substrate site is virtually unknown (10 -14). Only a few residues have been implicated to associate with eEF-2: His 426 , Tyr 481 , and His 440 ; however, these have not been confirmed. The crystal structures (8,9) show that His 440 is located at the base of the active site distant from the scissile N-glycosidic bond; therefore, contact with NAD ϩ is unlikely at this position. The three-dimensional structure shows Tyr 481 stacking against the nicotinamide ring of NAD ϩ (8,9). Substitution of phenylalanine for Tyr 481 alters the rate of reaction, possibly by changing the orientation of the diphthamide on eEF-2 (10). When mutated to tyrosine, His 426 (located in helix I and distant from the active site) renders the protein nontoxic (12). An antibody that specifically binds to helix I within ETA shows that NAD ϩ binding is not affected, thereby implicating eEF-2 binding as the defect in enzymatic activity for the H426Y mutant protein (13). However, further mutagenic studies involving His 426 indicated that its role is not eEF-2 binding, but rather maintaining the structure of the active site through the proper alignment of the active-site residues (11). Therefore, those residues involved with eEF-2 or the transferase step remain in question.
Loop regions are common in forming binding and active sites (15,16). Therefore, loop regions were the first targets for investigating the inner workings of the active-site machinery of exotoxin A. In the crystal structures of the catalytic domain of ETA, two surface-exposed loops near the known active site were found (8,9). It was proposed that either or both of these loops act as clamps holding eEF-2 near the active site, or perhaps these loops work to stabilize the transition state structure during the ADP-ribosylation reaction. In this study, to elucidate the importance of these loops, the entire region was replaced with glycine; thus, the loop was left intact and flexible, but acted as a functionally silent region. ADP-ribosyltransferase activity and NAD ϩ binding were determined for all mutant proteins, and the structural stability was evaluated in those mutant proteins that showed reduced activity. One loop region showed a significant decrease in activity compared with the wild-type protein and thus was further analyzed by alanine-scanning mutagenesis (17). This allowed mapping of the proposed ADP-ribosyltransferase motif.

EXPERIMENTAL PROCEDURES
Purification of PE24H-The catalytic fragment of ETA with a Cterminal polyhistidine tag (PE24H) encoded by the plasmid pPH1 was overexpressed in Escherichia coli strain BB101 (DE3) and purified as described (18) with slight modifications. The cells were lysed in an osmotic shock lysate solution (19) containing 10 mM EDTA. The osmotic shock lysate was loaded onto a 10-ml Q-Sepharose Fast Flow anionexchange column (Amersham Pharmacia Biotech, Baie D'Urfe, Quebec, Canada). The Q-Sepharose column was washed with 50 mM NaCl and 20 mM Tris-HCl (pH 7.9), and the protein was eluted using 300 mM NaCl and 20 mM Tris-HCl (pH 7.9). This eluant was loaded onto a 1-ml chelate-agarose affinity column (Amersham Pharmacia Biotech) charged with 50 mM NiSO 4 as previously described (18), except for the purification of the S585C and S585C-pG (where pG is polyglycine)-Loop C mutant proteins, which required the use of 100 mM ZnSO 4 to charge the column. Fractions containing purified protein were pooled and dialyzed against 1 mM EDTA, 50 mM NaCl, and 20 mM Tris-HCl (pH 7.9) and subsequently concentrated to 1 ml using a Centriprep-30 concentrator (Amicon Inc., Beverly, MA).
Purification of eEF-2-eEF-2 was purified from wheat germ as previously described (20) with some modifications.
Site-directed Mutagenesis-The pG-Loop N mutation was created using the Kunkel mutagenesis method as described (21). Two-stage QuikChange TM mutagenesis was used to create the polyglycine Loop C and double-loop mutants (DLM; both loops replaced with glycine) (22). The S585C mutation was prepared as previously described (23). Alanine-scanning mutants and the S585C-pG-Loop C mutant were prepared by standard QuikChange TM mutagenesis (Stratagene, La Jolla, CA). The DNA template was the plasmid pPH1 containing the gene for wild-type PE24H (unless otherwise indicated) using several sets of primers (Table I). The presence of the desired mutation was determined by dideoxy sequencing with an ABI Prism Model 377 DNA sequencer using dye termination and cycle sequencing, and the DNA samples were analyzed on a 4.5% acrylamide gel that was 36 cm in length.
Fluorescence Measurements-All steady-state fluorescence measurements (except for the eEF-2 binding assay) were obtained using a PTI Alphascan spectrophotometer interfaced with a computer using Felix TM Version 1.21 software (Photon Technology International) and equipped with a water-jacketed sample chamber set to 25°C.
ADPRT Assay (NAD ϩ -dependent)-The ADPRT activity of the various enzyme samples was tested as described previously (20). Briefly, the excitation and emission monochromators were set to 305 nm and zero order (no diffraction), respectively, and a 309-nm cutoff filter (Oriel Corp., Stratford, CT) was included on the sample chamber side of the emission monochromator to eliminate scattered excitation light and to maximize the signal. Buffer (20 mM Tris-HCl (pH 7.9)), ⑀-NAD ϩ (Sigma; various concentrations were used up to 500 M from an ⑀-NAD ϩ stock solution prepared in distilled water; ⑀ M 265nm ϭ 6000 M Ϫ1 cm Ϫ1 ), and 14 M eEF-2 (at saturating levels) were combined in an ultramicrocuvette (3-mm path length; Helma Inc., Concord, Ontario, Canada). The cuvette was equilibrated at 25°C for 10 min. Toxin dilutions were performed in siliconized tubes with the buffer described above. The reaction was started by rapidly adding PE24H (final concentrations for toxin ranged from 5 nM to 19 M depending upon the mutant protein), and the reaction progress was monitored by an increase in fluorescence during the production of ⑀-ADP-ribose.
ADPRT Assay (eEF-2-dependent)-The assay was performed as described above with the following modifications. Necessary adjustments to the eEF-2 protein solution were made to maintain KCl concentrations at 85 mM since it was stored in 300 mM KCl-containing buffer.  ⑀-AMP Standard Curve and Assay Calibration-A stock solution of ⑀-AMP (Sigma; prepared in distilled water; ⑀ M 265nm ϭ 10,000 M Ϫ1 cm Ϫ1 ) was used to prepare a series of standards in 20 mM Tris-HCl (pH 7.9). The fluorescence of the ⑀-AMP standards (0 -10 M) was recorded to generate a standard curve having a slope with units of fluorescence intensity per micromolar ⑀-AMP. This slope from the standard curve was used to convert the slopes obtained for the ADPRT measurements to catalytic rates with units of micromolar ⑀-NAD ϩ /s.
Quenching of Intrinsic Protein Fluorescence-The NAD ϩ -dependent quenching of the intrinsic protein fluorescence of the tryptophans in PE24H was used to determine the binding constant (K D ) for NAD ϩ and was observed by exciting the protein at 295 nm (4-nm slit width) and measuring the fluorescence intensity with the emission wavelength set to 340 nm (4-nm band pass) as previously described (6). Briefly, the protein solution was titrated with concentrations of NAD ϩ ranging from 0 to 981.5 M in the presence of toxin in an initial volume of 600 l in 50 mM NaCl and 20 mM Tris-HCl (pH 7.9). The final concentration of toxin used in this experiment was 1.25 M, except for D484A/D488A and Q483A/D484A/D488A, which were at 0.625 M.
Fluorescence Labeling of PE24H-Purified PE24H protein (0.5 mg of protein) was incubated in 150 mM NaCl and 200 mM Tris-HCl (pH 8.1) with an excess of DTT (5 mol of DTT/mol of PE24H) at room temperature for 60 min. A concentrated solution of IAEDANS (in Me 2 SO) was added to the reaction mixture (5% Me 2 SO final concentration) to give a molar ratio of 20:1 IAEDANS/PE24H, and the reaction was gently mixed on a nutator for 15 min at room temperature. The reaction was quenched by the addition of an excess of DTT (100 mM final concentration). The quenched sample was loaded onto a prepacked Econo-Pac10DG column (Bio-Rad, Mississauga, Ontario), and the protein adduct was eluted with 150 mM NaCl and 20 mM Tris-HCl (pH 7.9). The protein adduct peak was well separated from the free dye peak, with the latter peak appearing in later fractions. The labeled PE24H protein was further characterized by UV and fluorescence spectroscopy using the extinction coefficients for PE24H (⑀ M 280nm ϭ 27,310 M Ϫ1 cm Ϫ1 ) and for AEDANS (⑀ M 337 NM ϭ 6000 M Ϫ1 cm Ϫ1 ) (24). The labeling efficiency, based on the appropriate absorbances and molar extinction coefficients of the bound fluorophore and the measured protein, was nearly 100% (1 mol of AEDANS/mol of PE24H).
Fluorescence Labeling of eEF-2-Purified wheat germ eEF-2 protein was labeled with 5-iodoacetamide fluorescein (5-IAF; Molecular Probes, Inc., Eugene, OR) by incubation of 1 mg of protein in reaction buffer (200 mM Tris-HCl (pH 8.1) containing DTT (3 mol of DTT/mol of protein) for 30 min, followed by the addition of a concentrated stock of 5-IAF (20 mg/ml in Me 2 SO at 10 mol of 5-IAF/mol of protein). The reaction solution was gently mixed on a nutator, and the sulfhydryl labeling reaction was stopped after 10 min at 25°C by the addition of a 100-fold excess of DTT. The reaction mixture (415 l) was loaded onto a 10DG column, equilibrated in 50 mM NaCl and 20 mM Tris-HCl (pH 7.9); and 0.6-ml fractions were collected. The protein-5-AF adduct eluted from the column in a peak that was completely separated from the free fluorophore peak. The stoichiometry of the labeling reaction was ascertained by determining the eEF-2 concentration using the BCA protein assay with bovine serum albumin as the standard (20) and by absorption spectroscopy using the molar extinction coefficient for 5-AF (⑀ M 492NM ϭ 70,000 M Ϫ1 cm Ϫ1 ). The stoichiometry of labeling of eEF-2 with 5-AF indicated that only one of the eight sulfhydryl groups within the protein was modified.
Fluorescence-based eEF-2 Binding Assay-Aliquots of 5-AF-labeled eEF-2 protein were added to 1.0 M AEDANS-labeled S585C-PE24H or to AEDANS-labeled S585C-pG-Loop C mutant in 50 mM NaCl and 20 mM Tris-HCl (pH 7.9). The fluorescence intensity of AEDANS was monitored with a Cary Eclipse fluorometer (Varian, Mississauga, Ontario, Canada) equipped with a Peltier thermostatted multicell holder at 25°C with excitation and emission wavelengths at 337 and 460 nm, respectively (5-nm band passes). The raw fluorescence data were corrected for the dilution factor, and the dissociation constant for eEF-2 binding with PE24H was determined using the following equation (two-site binding model): where ⌬F i is the change in fluorescence intensity for each ligand (eEF-2) concentration upon macromolecular association, ⌬F max is the maximum change in fluorescence intensity at saturation of the ligand-binding site within PE24H, and K D1 and K D2 are the dissociation constants for the binding of eEF-2 with PE24H.
Unfolding with Urea-The wild-type and pG-Loop C proteins in 20 mM Tris-HCl (pH 7.9) were mixed with an appropriate amount of 8 M urea (Pierce) to provide solutions from 0 to 5.5 M urea and 3.6 g/ml toxin (final concentration). Solutions were incubated at room temperature for a minimum of 15 min before spectroscopic analysis. For tryptophan fluorescence, excitation was at 295 nm, and emission was scanned from 325 to 355 nm in ultramicrocuvettes. The excitation and emission slit widths were both 4 nm. Unfolding profiles were determined as previously described (21,24).

Location and Nomenclature of the Loop Regions-
The sequences of two loop regions within the catalytic domain of ETA with proposed involvement either in catalysis or through an association with eEF-2 are shown and aligned with the corresponding sequences within DT (Fig. 1). The nomenclature (ETA numbering) for these loops is such that the loop closest to the N terminus of the protein is termed Loop N (Arg 458 -Ala 464 ), and the loop closest to the C terminus is Loop C (Gln 483 -Arg 490 ). The alignment of these two loops within ETA did not produce a good match between Loop N and the corresponding loop within DT; however, the alignment of Loop C showed good alignment (sequence identity and homology) with the corresponding loop within DT (Fig. 1). Notably, there is sequence conservation at Asp 484 and Glu 486 within the Loop C region of these two diphthamide-specific toxins. Interestingly, Loop C did not appear in the original crystal structure of intact whole ETA (25); however, it was resolved in the structure of the catalytic domain of ETA (8,9). The three-dimensional structure of Loop N was not fully resolved in the catalytic domain structure that bound a less hydrolyzable NAD ϩ analog, ␤-methylene-thiazole-4-carboxamide adenine dinucleotide (␤-TAD ϩ ), as it was disordered (9), but was evident in the structures of whole ETA and the catalytic domain complexed with nicotinamide and AMP (8). It has been shown that, upon proteolytic cleavage within a helix in the translocation domain of the whole toxin, Loop N moves to allow greater access to the active site; and consequently, it has been proposed to interact with eEF-2 due to its solvent exposure upon shifting (9).
Polyglycine Loop Replacement Investigation-One possibility in the study of the importance of either Loop N or C would be to delete the entire region. However, analysis of the crystal structure (9) showed that this might not be advisable. The distance between the beginning and end of the loop is significant, and its removal would impose tight constraints on the protein structure, thereby altering it. The distance between the beginning and end of Loop C was determined to be 14.6 Å (between the C-␣ atoms of Gln 483 and Arg 490 ), and this distance is 12.4 Å (between the C-␣ atoms of Ala 457 and Ala 464 ) for Loop N. An alternative to loop deletion mutants was to replace each loop with a glycine segment. Glycine is a very unreactive residue that allows much flexibility in terms of possible main chain conformations. Therefore, in the alternative approach, the loop region is present, but acts essentially as a functionally silent region.
These polyglycine loop replacement mutants were constructed in the catalytic domain of ETA (PE24H) (18), and their importance was initially screened by their ADPRT activity to decide if either or both of these regions were involved in catalysis. The kinetic parameters for NAD ϩ as substrate for these polyglycine mutants are shown in Table II and were determined with eEF-2 at saturating levels (14 M). The pG-Loop N mutant protein had a lower activity than the wild-type protein (18-fold decrease). Although the turnover number was affected by this mutation, the K m(⑀-NAD ϩ ) actually decreased (showing a modest increase in the enzyme's ability to interact with the ⑀-NAD ϩ substrate), implying that the effect of Loop N replacement is at the level of catalysis and not substrate docking and binding. However, the effect of Loop N replacement was small in comparison with the effect of Loop C replacement; consequently, we chose to focus our efforts on the Loop C mutant protein. In contrast, kinetic measurements for pG-Loop C showed a drastic loss of activity upon its replacement with glycine residues. The data in Table II indicate that only 0.0056% of the wild-type activity (18,000-fold decrease) was retained in this mutant. However, despite this large activity loss, the effect of Loop C replacement on the K m(⑀-NAD ϩ ) was opposite to that anticipated if Loop C was involved in substrate binding (5-fold decrease in K m ) (Table II). Thus, from a consideration of the specificity constant (Table II), it was evident that this parameter reflected the change in k cat and not K m(⑀-NAD ϩ ) . This indicates that the ability of the mutant toxin to stabilize the transition state species for the ADPRT reaction is likely more affected than the ability to bind the substrate, ⑀-NAD ϩ . NAD ϩ binding data for the two polyglycine loop mutants confirmed that NAD ϩ bound normally to both mutant proteins (Fig. 2), with K D values of 215 Ϯ 16 and 191 Ϯ 19 M for the pG-Loop N and pG-Loop C mutants, respectively. DLM (both loops replaced with glycine) was also assessed; however, because its activity was at base-line levels, reliable kinetic parameters could not be obtained. As an alternative, the rate of reaction at an ⑀-NAD ϩ concentration of 200 M was used to compare the mutant and wild-type proteins. This comparison indicated that the turnover number is in the range of 65,000-fold less than that of the wild-type protein (Table II). However, despite its low activity, DLM also was capable of binding NAD ϩ (149 Ϯ 12 M) (Fig. 2). Therefore, the effect on catalysis when both loops are replaced with glycine was further attenuated; however, this decrease was not strictly cumulative. In essence, the major contributing region affecting catalysis is Loop C, with Loop N playing a minor role.
An investigation into the structural integrity of the loop mutants was undertaken. pG-Loops N and C are expressed as soluble proteins at levels similar to those of the wild-type protein. The sensitivity of these mutant proteins to trypsin proteolysis compared with the wild-type protein was also investigated. Examination of the resultant proteolytic pattern and time required for complete digestion indicated that all three loop replacement mutants (pG-Loop N, pG-Loop C, and DLM) were slightly less stable, but were structurally similar to the wild-type protein (data not shown). Furthermore, the chemical unfolding of pG-Loop C was studied using urea as the denaturant; the data were fit to a two-state folding model as previously described (21); and the results for the free energy change associated with this unfolding (⌬G (F3 U) ) and the transition midpoints (D1 ⁄2 ) were estimated (data not shown). ⌬G (F3 U) represents the free energy change for the conversion of the native to unfolded state of a protein in the absence of denaturant. Comparison of the values of ⌬G (F3 U) for the wildtype and pG-Loop C proteins showed that less free energy was needed to unfold the mutant protein. ⌬G (F3 U) for the wild-type protein is 8.8 Ϯ 2.9 kJ mol Ϫ1 , whereas that for pG-Loop C is 5.5 Ϯ 2.5 kJ mol Ϫ1 . This instability at lower urea concentrations is reflected in the transition midpoints, which were 1.1 and 1.7 M for the mutant and wild-type proteins, respectively. The instability of the mutant protein at lower concentrations of urea may be a reflection of some instability in the region of the loop; however, as the protein was titrated with urea, unfolding like that of the wild-type protein was observed. Additional support for pG-Loop C having a similar structure as the wildtype protein is given by the fluorescence wavelength emission maximum (fluorescence em(max) ). In the absence of denaturant, both of these proteins have a fluorescence em(max) at 334 nm, illustrating that initially the tryptophan residues in these proteins are in a similar environment and therefore similarly localized in the tertiary structure of the proteins. For the fully denatured proteins, the fluorescence em(max) is 350 nm. Since the three tryptophans are situated within or near the active site of the enzyme, these data suggest that the active site in this mutant protein has a similar structure to that in the wild-type protein.
Further evidence for the native-like fold of the pG-Loop C mutant protein is shown in Fig. 3, for which an assay was devised for the binding of toxin with its eEF-2 substrate based on fluorescence resonance energy transfer between donor-labeled PE24H (AEDANS) and acceptor-labeled eEF-2 (5-AF). The binding data were similar for the pG-Loop C mutant compared with the wild-type protein (K D1 ϭ 1526 Ϯ 76 and 1471 Ϯ 76 nM, respectively). 2 The binding constants for the polyglycine loop mutants with the NAD ϩ substrate were similar to the wild-type protein (see text). The pG-Loop C mutant protein showed a modest increase in the K D value compared with the wild-type protein (nearly 4-fold). These data indicate that the polyglycine loop mutant proteins show near wild-type substrate-binding properties.
Alanine-scanning Mutagenesis-The role of each residue within Loop C was probed using alanine-scanning mutagenesis based on the results obtained from the study of the polyglycine replacement mutant proteins (17). Like the earlier study involving the polyglycine mutant proteins, these alanine mutant proteins were assessed for their ADPRT activity. As shown in Table II, none of the single alanine replacement mutant proteins displayed a significant alteration in K m(⑀-NAD ϩ ) compared with the wild-type protein. However, three mutant proteins had reduced k cat values. The activities of the mutant proteins Q483A, D484A, and D488A were 7.2, 4.0, and 7.7%, respectively, of that of the wild-type protein. Table III indicates that the dissociation constants for NAD ϩ binding were not adversely affected by the alanine replacements. Although D484A and D488A did show a decreased affinity for this substrate (2.8-and 2.4-fold, respectively), the effect was not large in comparison with the observed effect on the k cat values for these mutant proteins. Therefore, these data indicate that the NAD ϩ substrate-binding site was not impaired by any of the alanine replacements within Loop C. Trypsin proteolysis studies on the three alanine mutant proteins that exhibited the most reduced ADPRT activity (Q483A, D484A, and D488A) showed band patterns and time courses similar to those of the wild-type protein (data not shown). Therefore, to resolve whether or not the cause of the defect in Loop C was related to an alteration in the eEF-2 substrate site, the kinetic parameters for the eEF-2 substrate were determined (variable eEF-2 concentrations and [⑀-NAD ϩ ] fixed at 500 M) (Table IV) on those alanine mutant proteins that had previously showed altered ADPRT activity. for AEDANS-labeled S585C-pG-Loop C PE24H (F) proteins was determined from the titration of these protein adducts (1.0 M) with 5-AF-labeled eEF-2 (0 -4790 nM) as described under "Experimental Procedures." The fluorescence excitation was at 337 nm, and emission was at 460 nm (5-nm band passes) at 25°C. The raw fluorescence quenching data were converted to fractional saturation values (⌬F/⌬F max ) and are plotted against the 5-AF-labeled eEF-2 concentration. These data were fit to a two-site binding equation as described under "Experimental Procedures." As dictated by the turnover numbers (Table IV), Q483A, D484A, and D488A had significantly reduced activity. The K m values for these alanine mutant proteins were only marginally increased (1.02-2.19-fold), indicating that Loop C plays a major role in the catalytic process within ETA and not in substrate association. The pG-Loop C protein was also examined; however, its activity was only marginally above background in this assay (eEF-2 as limiting substrate); and therefore, a K m for eEF-2 for this loop mutant could not be determined. It was noted that the magnitudes of the effects on the relative k cat values showed some differences, as shown in Tables II and IV (relative k cat(⑀-NAD ϩ ) and k cat(eEF-2) values as follows: Q483A, 0.07 and 0.11; Q484A, 0.04 and 0.07; Q485A, 0.19 and 0.69; and D488A, 0.08 and 0.17, respectively). However, these differences are not large and likely reflect the use of separate batches of the protein substrate (eEF-2) for each data set. Notably, although the magnitudes of the changes in the k cat values showed some differences, the trends remained consistent.
Two additional mutant proteins (multiple replacements) were made based on the initial activity measurements of the alanine mutant proteins. These were D484A/D488A and Q483A/D484A/D488A, which were used to determine if these residues are primarily responsible for the reduced catalysis of the entire loop region. Both D484A/D488A and Q483A/D484A/ D488A had reduced activity (relative k cat values for both were Ͻ0.5% of the wild-type value), with a minimal effect on K m (⑀-NAD ϩ ) (Table II). It is notable that when both aspartate residues were mutated to alanine, the resulting k cat equaled the accumulation of the individual alanine mutants, suggesting that Asp 484 and Asp 488 cooperate in an event linked to catalysis. However, the addition of the mutation Q483A to the D484A/D488A mutant (Q483A/D484A/D488A) did not further significantly reduce the k cat . Therefore, not all three residues act in concert. The NAD ϩ binding studies on D484A/D488A and Q483A/D484A/D488A showed near wild-type protein behavior (Table III), providing further support for the idea that Loop C participates in an important event in catalysis. DISCUSSION In the initial polyglycine loop investigation, pG-Loop C had a significant loss of activity (18,000-fold decrease), yet was able to retain its ability to interact with both NAD ϩ and eEF-2 substrates. However, the structural integrity of this mutant became in question because of this activity defect. As was shown, the NAD ϩ binding data illustrated that pG-Loop C is capable of binding NAD ϩ with a similar affinity as the wildtype protein. Hence, the pG-Loop C mutant protein has a folded structure with an active site resembling that of the wild-type protein. Earlier work has shown that N-acetyltryptophanamide, a control for free tryptophan, does not bind NAD ϩ since no tryptophan quenching was observed (6). Therefore, it is unlikely that the quenching of the intrinsic protein fluorescence from this mutant protein would be observed if the protein was significantly unfolded. Moreover, it has been demonstrated that NAD ϩ quenching of tryptophan fluorescence is protein-specific (20). Furthermore, the tryptophan fluorescence em(max) for pG-Loop C is identical to that for the wild-type protein, indicating that these proteins are folded properly. Chemical denaturation (although suggesting slightly reduced stability), combined with the proteolysis study, demonstrated that pG-Loop C is not misfolded and has a similar structure as the wild-type protein. Therefore, the experimental data indicate that this loss of activity for pG-Loop C must correlate with the replacement of critical residue(s) involved in the ADPribosyltransferase step within the catalytic mechanism of the toxin enzyme. Unfortunately, this study alone could not fully decipher the significance of Loop C. As an extension of this work, alanine-scanning mutagenesis on Loop C was used to determine if specific residues within the loop or if an inherent feature of the loop is responsible for what was observed in the pG-Loop C study.
The other loop region in this study, Loop N, does not appear to be an essential catalytic factor since its replacement with polyglycine resulted in a significant, but relatively small decrease in enzymatic activity. This loop region was previously investigated when the crystal structure for the catalytic domain of ETA was solved in the presence of an NAD ϩ analog. Results showed that the binding of NAD ϩ does not stabilize this loop, indicating that this region is not needed for NAD ϩ binding (8,9). Li et al. (8,9) also proposed that this loop may be able to interact with eEF-2 since it becomes solvent-accessible once the toxin is activated and cleaved. This process shifts the loop from protecting the active site in the proenzyme state to a conformation that exposes the active site. When compared with other toxins, this Loop N region shows no structural similarities. The corresponding loop region in diphtheria toxin, which also ADP-ribosylates eEF-2, is much longer. The x-ray structures of both catalytic domains of ETA and DT superimpose (9,25,26); therefore, if a catalytic motif exists within this enzyme, it is likely not contained within this region.
The kinetics of the alanine replacement mutant proteins within Loop C suggested the identity of a small subregion within this loop that is responsible for the decrease in ADPRT activity, namely Gln 483 -Asp 484 and Asp 488 . Since NAD ϩ and eEF-2 binding data and the K m(⑀-NAD ϩ ) and K m  values are similar to those of the wild-type protein, it suggests that Loop C is an important catalytic element within ETA. In the structure of the catalytic domain of ETA, the phenol ring of Tyr 481 stacks with the nicotinamide ring of NAD ϩ (8,9) near the site of cleavage where the ADP-ribosyl group of NAD ϩ is transferred to eEF-2. Therefore, residues involved in this transfer event must be spatially situated near Tyr 481 . Examination of   (Fig. 4) correlates with this notion since they are on one face of the loop in close proximity to Tyr 481 and are located close to the NAD ϩ substrate. However, the remaining residues within Loop C are situated on the opposite side of the loop and are more distant from the site of the reaction. Asp 488 is important for activity; however, it is not situated as closely to Tyr 481 as the other catalytically important residues in question. However, the kinetic data for the D484A/D488A mutant enzyme showed that Asp 488 acts in concert with Asp 484 . Unfortunately, the distance between these two residues is too large for any direct interaction, but these residues could be linked through a bridged water molecule (however, it was not resolved as a heteroatom in the structure (9)) since the x-ray structure shows Asp 488 participating in several hydrogen bonds. Therefore, Asp 488 may play a structural role within the loop region by properly aligning those residues (in particular, Gln 483 , Asp 484 , and Gln 485 ) that are perhaps involved in the stabilization of the transition state for the ADPRT reaction, which would involve parts of both the NAD ϩ and eEF-2 substrates as the kinetic data suggest (Tables II and IV).
To investigate if a possible ADP-ribosyltransferase motif of the eEF-2-specific ADP-ribosylation reaction had been uncovered, a comparison of the catalytic domains of ETA and DT was undertaken. Both of these proteins recognize eEF-2 and have structurally similar active sites (8,25,26). The structure of the NAD ϩ -bound catalytic domain of DT was superimposed with the catalytic domain of ETA, and a sequence alignment was proposed (25). The alignment illustrated that Tyr 481 and Tyr 470 in ETA correspond to Tyr 65 and Tyr 54 in DT, respectively, in which the nicotinamide ring is positioned within a groove created by these tyrosines. In addition, those residues in DT that align with Loop C from ETA form a loop with notable sequence identity and residue conservation. Importantly, Asp 484 , shown by the alanine-scanning experiment as the most critical catalytic residue, is conserved between these two diphthamidespecific toxins. At Gln 485 (a residue important for catalysis) in Loop C, a conservative substitution of Asn 69 in DT is found. The E486A mutation within Loop C gave no indication that this glutamate residue is necessary for catalysis; however, it is conserved within DT (Glu 70 ) (Fig. 1). However, analysis of the entire Loop C region (Val 67 -Ser 74 ) shows that Gln 70 in DT may be required to maintain the overall net charge of this catalytic motif. As discussed, Gln 483 and Gln 485 , in addition to Asp 484 , represent the region proposed to stabilize the transition state during the transferase step. These residues are polar and have the ability to hydrogen bond. The residues analogous to the glutamines in DT are Val 67 and Asn 69 , with only the latter having this property. Although a hydrophobic residue corresponds to Gln 483 in the sequence alignment, a serine residue positioned next to this valine may participate in the hydrogen bonding in the DT structure. The remaining residues in Loop C of ETA have little similarity to the corresponding residues in DT and are located on the opposite face of the loop away from the active site (Fig. 4). Asp 488 in Loop C is proposed to play a role in stabilizing Loop C; yet in DT, at this position is found a proline, which cannot stabilize the loop through hydrogen bond formation. Therefore, other residues in DT must be responsible for stabilization of the active-site loop structure. In conclusion, an active-site structure exists within DT that is analogous to the predicted ADP-ribosyltransferase structure within Loop C and is also near the site of ADP-ribose transfer ( Fig. 4 and Tables II and IV). It is proposed that this catalytic region within these two diphthamide-specific ADPRT enzymes represents a catalytic structural motif that appears to be specific to only this subclass of the ADPRT family (27).
In summary, a working model is proposed that illustrates how these data may relate to catalysis. Initially, a binary complex between ETA and NAD ϩ forms (14,28), exposing the transferase site. The eEF-2 substrate then docks onto the surface of the enzyme, and the transition state structure is stabilized by favorable interaction with Loop C (Gln 483 , Asp 484 , and Gln 485 ) within the catalytic domain of the toxin. This alignment and stabilization of the transition state structure within the active site by the catalytic residues Gln 483 , Asp 484 , and Gln 485 may occur through the formation of hydrogen bonds. The completion of the transferase event then induces a conformational change within the complex, weakening the association between eEF-2 and the enzyme, thereby releasing ADP-ribosyl-eEF-2.