NAD Binding Induces Conformational Changes in Rho ADP-ribosylating Clostridium botulinum C3 Exoenzyme*

We have solved the crystal structures of Clostridium botulinum C3 exoenzyme free and complexed to NAD in the same crystal form, at 2.7 and 1.95 Å, respectively. The asymmetric unit contains four molecules, which, in the free form, share the same conformation. Upon NAD binding, C3 underwent various conformational changes, whose amplitudes were differentially limited in the four molecules of the crystal unit. A major rearrangement concerns the loop that contains the functionally important ARTT motif (ADP-ribosyltransferase toxin turn-turn). The ARTT loop undergoes an ample swinging motion to adopt a conformation that covers

Many bacterial toxins ADP-ribosylate nucleotide-binding proteins that are involved in essential cell functions (for review see Ref. 1). The molecular basis of their action consists of the binding of NAD, its glycohydrolysis into ADP-ribose and nicotinamide, and the transfer of the ADP-ribose moiety to a specific residue on the eukaryotic protein substrate. All these toxins share a highly conserved catalytic glutamate, which is critical for the NAD-glycohydrolase activity. Although they exhibit a similar global mode of action, ADP-ribosyltransferase toxins are quite distinct regarding their substrate and their pathophysiological properties. Thus, these toxins can be divided into four subfamilies: diphtheria-like toxins, cholera-like toxins, binary toxins and C3-like exoenzymes.
C3-like exoenzymes are distinct from other ADP-ribosyltransferase toxins in that they lack specific cell-surface binding and translocation components to facilitate their cell entry. Also, they are unique because of their high specificity for the small GTP-binding proteins RhoA, RhoB, and RhoC on an asparagine residue. This specificity makes them particularly useful to switch off selectively the cellular function of Rho proteins. Thus, C3-like exoenzymes are used to study the Rho-dependent processes, including cytoskeleton organization, endocytosis, phagocytosis, nucleus signaling, and regulation of gene transcription (for review see Ref. 2). However, the molecular basis of C3 action including NAD binding, its glycohydrolysis, Rho binding, and Rho ADP-ribosylation, is not well understood.
The structure of Clostridium botulinum C3 exoenzyme free of NAD was recently determined, revealing that C3-like exoenzymes possess the main features that characterize the NADbinding site of other ADP-ribosyltransferase toxins (3). These authors have proposed that C3-like exoenzymes and binary toxins should be included in the same subfamily on the basis of structural and sequence similarities in their catalytic domain. A new motif of this subfamily has been described and termed the ARTT 1 (ADP-ribosylating toxin turn-turn) motif (3). This motif of two short amino acid stretches (or turns) encompasses residues 207-210 and 211-214 and may be important for catalytic activity and substrate recognition (3,4). The highly conserved catalytic glutamate (Glu-214) belongs to this motif. Another similarity with the C3-like exoenzymes and binary toxins is the absence of the "active site" loop that has been described for other ADP-ribosyltransferase toxins as a substrate recognition site (5,6).
To approach the molecular basis of C3 action, we carried out a structural study of the C. botulinum C3 exoenzyme complexed to NAD. Crystals of C3 free of NAD were initially obtained with four independent molecules in the asymmetric unit and diffracted to 2.7 Å resolution. To prevent NAD hydrolysis during crystal growth, we briefly soaked these crystals in NAD solution. Hence, the structure of C3 complexed to NAD was determined at 1.95 Å resolution, preventing crystallographic bias in a structural comparison with the free form.
These structures showed that upon NAD binding, C3 underwent various conformational changes whose amplitudes were differentially limited in the four molecules of the crystal unit. Our data emphasized a remarkable NAD-induced plasticity of the NAD binding pocket and suggest that a novel ARTT conformation may be privileged by the C3-NAD complex to bind to Rho.

EXPERIMENTAL PROCEDURES
Protein Expression-The mature form of C. botulinum C3 exoenzyme (residues 41-251) was subcloned into the pET-28a vector, expressed in the BL21 (DE3) strain of Escherichia coli, and purified to homogeneity in one step by cation exchange chromatography (CM-Sepharose Fast Flow). Mutants were constructed by site-directed mutagenesis with the pET-28a C3 wild-type as templates and the respective oligonucleotides using the Stratagene QuikChange kit according to the manufacturer's instructions. The different C3 mutants (S174A, L177C, Q182A, R186E, and Q212A) were expressed and purified as the wild-type. RhoA was produced as a pure protein as previously described (7) using the pGEX-2T RhoA-F25N vector (gift of A. Hall) expressed in the XLI Blue strain of E. coli after cleavage of the GST moiety from the fusion protein RhoA-GST with 3 units of thrombin for 18 h at 4°C.
Crystallization-Crystals of wild-type and mutant C. botulinum C3 exoenzyme were grown in sitting drops by the vapor diffusion method at 18°C. Crystals were obtained with 12 mg/ml protein against a reservoir containing 22.5% PEG (polyethylene glycol) 3350 w/w, 100 mM Li 2 SO 4 , and 80 mM citric acid at pH 3. Seeding techniques were applied to obtain large single crystals. A cryoprotectant consisting of 10% propanol, 15% ethylene glycol, 6% xylitol, 22.5% PEG 3350 w/w, and 100 mM citric acid at pH 3 was used for flash-freeze crystals in liquid ethane. The crystals belong to the monoclinic space group C2 with four molecules per asymmetric unit named, hereafter and in the PDB coordinates, as A, B, C, and D.
Structure of Free C3-To solve the structure of C3 by the isomorphous replacement method, we crystallized a cysteine mutant of C3 (L177C). The mutant was co-crystallized with 0.1 mM mercury acetate added to the precipitant. Crystals were later soaked in 97% of the crystallized solution and 3% of saturated methyl mercuric chloride for 3 days. An x-ray diffraction data set was collected at 17°C to 2.7 Å resolution on a MAR image plate detector mounted on a Rigaku x-ray source. All the data were processed with DENZO and scaled with SCALEPACK (8) (see Table I). Meanwhile, the crystal structure of the wild-type C3 exoenzyme of C. botulinum was solved (PDB accession code 1G24) (3). Therefore, this structure was used as search model to solve the C3-L177C structure by molecular replacement with AMoRe (9). The structure was refined after one round of rigid-body refinements by simulated annealing, energy minimization, and isotropic B-factor refinements through maximum-likelihood as implemented in CNS (10). A model was built using TURBO (11) with SIGMAA-weighted 2F o Ϫ F c and F o Ϫ F c maps calculated with the CCP4 suite (12) through alternating cycles of CNS refinement. No non-crystallographic symmetry restraints were used during the refinement. The final structure has an R-factor (R-free) of 24.0% (29.0%). The average r.m.s. deviation between the C␣ atoms of the C3-L177C models versus those of the C3 wild-type models reported by Han et al. (3) at 1.7 Å in a different crystal form and pH condition is 0.65 Å, confirming the original polypeptide trace and that the cysteine mutation and mercury atom cause no major change. Consequently, the C3-L177C structure represents well the structure of the wild-type C3 in the nucleotide-free state.
Structure of C3 Complexed to NAD-Crystals of wild-type C3 complexed to NAD were obtained by soaking experiments rather than co-crystallization to limit NAD hydrolysis during crystal growth. Crystals of C3 were soaked for 30 -60 min in the crystallized solution with the addition of 20 mM NAD. Then, the crystals were flash-frozen to prevent the NAD hydrolysis in the crystals. An x-ray diffraction data set was collected at Ϫ170°C to 1.95 Å resolution on the European Synchrotron Radiation Facility beamline BM30. Because the soaked crystals were isomorphous with those of C3-free, the structure of C3-NAD was directly refined starting from the coordinates of C3-free. Then, ARP/WARP 5 (13) was used to automatically position water molecules in uninterpreted electron density (higher than 3.5 ) corresponding to NAD. REFMAC from the CCP4 suite was used in ARP/ WARP 5 to carry out structural refinement (12). The C3-NAD structure was refined as the C3-free structure. The final structure has an R-factor (R-free) of 23.4% (27.7%). The final statistics for the models are summarized in Table I. The coordinates and structure factors have been deposited in the Protein Data Bank with PDB accession codes 1gze and r1gzesf, respectively, for C3 L177C -free and 1gzf and r1gzfsf, respectively, for C3 WT -NAD.
NAD-glycohydrolase and ADP-ribosyltransferase Activity Measurements-To measure the NAD-glycohydrolase activity of C3 wild-type or its mutants, 0.1 nmol of the exoenzyme and 1 mmol of NAD enriched with 50 mol of [ 32 P]NAD in 10 l of phosphate buffer 10 mM, pH 7.5, were incubated for 1, 20, 50, 100, and 120 min at 37°C. 1 l of each sample was then loaded on a cellulose MN 300 sheet (Polygram CL 300 PEI/UV 254 , Machery-Nagel) and eluted with 100% ethanol:H 2 O:sodium acetate 1 M at pH 5 20:40:20. Radioactivity was detected and quantified with an imaging system (Storm 840, Molecular Dynamics). Results are representative of three independent experiments and were subjected to an analysis of variance (StatView).
To measure ADP-ribosyltransferase activity of C3 wild-type or its mutants, 0.1 nmol of the exoenzyme diluted in 10 l of Tris 50 mM, pH 7.5, containing 100 mM NaCl, 10 mM MgCl 2 , 20 mM dithiothreitol and 1.3 M [ 32 P]NAD was incubated for 1 h at 37°C with 0.05 nmol of RhoA. This mixture was then loaded on a 12% SDS polyacrylamide gel, and radioactive RhoA was directly detected and quantified on the dried gel by an imaging system (Storm 840, Molecular Dynamics) NAD Binding Measurement by Plasmon Resonance-NAD binding measurements on C3 wild-type and mutants were performed using a BIAcore 2000 apparatus (BIAcore, Uppsala, Sweden). C3 proteins (1 g/ml in 10 mM sodium acetate, pH 5, (pH Ͻ Ͻ pI)) were immobilized on CM5 sensor chip surfaces using the standard amine coupling procedure and a flow rate set at 5 l/min. The blank surface was constructed in the same way except that the C3 protein was omitted to serve as a reference surface. Kinetics assays were conducted at 25°C using 10 mM HEPES buffer, pH 7.4, and 150 mM NaCl containing 0.005% Surfactant P20 as running buffer. Various concentrations of NAD (10 -250 M in 10 mM HEPES, 150 mM NaCl, 0.005% Tween, pH 7.4) were injected at a flow rate of 50 l/min to reduce the mass transport effect. All assays of C3 and NAD interactions were performed in each case in duplicate and in three separate experiments. The blank run was subtracted from each sensorgram prior to data processing using BIAevaluation software 3. The data were fitted using the steady state affinity (Req versus C) option available within BIAevaluation 3 software.

RESULTS
Overall Description-The structures of C. botulinum C3 exoenzyme free (C3-free) and complexed to NAD (C3-NAD) were solved by molecular replacement at 2.7 and 1.95 Å resolution, respectively. Both structures were obtained in the same crystal form, which includes four independent molecules in the asymmetric unit, named hereafter molecules A, B, C, and D. In the C3-free form, each of the four molecules adopts an overall structure that is similar to the three others or to the one previously described (3) (r.m.s.d. Ͻ 0.67 Å; Table II). The overall structure of the four NAD-bound C3 molecules is also conserved, and each of them is similar to its equivalent unbound form (r.m.s.d. Ͻ 0.88 Å; Table II). The overall structure of C3 is a mixed ␣/␤ fold with a ␤-sandwich core formed by perpendicular packing of a five-stranded mixed ␤-sheet (␤1, ␤4, ␤8, ␤7, and ␤2) against a three-stranded antiparallel ␤-sheet (␤3, ␤6, and ␤5). The three-stranded sheet is flanked by four consecu- tive ␣-helices (␣1, ␣2, ␣3, and ␣4), and the five-stranded sheet is flanked by an additional ␣-helix (␣5) (Fig. 1A).
A more detailed analysis revealed that the four unbound molecules adopt the same conformation, whereas the four NAD-bound molecules display various structural differences. Two of them are quite striking (Fig. 1). The first one concerns the loop comprising residues 204 -212 that overlaps the socalled ARTT motif (residues 207-214), which was previously proposed to be implicated in substrate specificity and recognition (14). This "ARTT loop" is orientated differently in the NAD-bound molecule A (Fig. 1A), as compared with both the three other NAD-bound molecules (Fig. 1, B, C, and D) and the four unbound molecules, which all possess an ARTT loop in the same conformation. More precisely, the ARTT loop conformational difference between the NAD-bound molecule A and the other unbound or bound molecules is characterized by an average r.m.s.d. of 2.6 Å on the main chain atoms. In molecule A, the ARTT loop is not constrained by the crystalpacking environment, whereas in molecules B, C, and D it makes major crystal-packing contacts with helix ␣5 from a neighboring C3 molecule (Fig. 2). In addition, a superimposition of molecule A on molecule B reveals that the ARTT loop conformation adopted by molecule A would be hindered by the local crystal-packing environment of molecule B (Fig. 2). Therefore, due to its unique unconstrained crystal-packing environment, only molecule A could adopt this NAD-induced ARTT loop conformation. A second striking difference that characterizes the four NAD-bound molecules concerns the NAD itself ( Figs. 1 and 3). In molecule A, the electron density of NAD is well defined and its B-factor is low (33.5 Å 2 ) as compared with the B-factor of the overall structure (38.2 Å 2 ) (Fig. 3A). In molecule B, the electron density for the NAD is also well defined but its B-factor is higher (46.2 Å 2 for the AMP moiety and 67.2 Å 2 for the nicotinamide mononucleotide (NMN) moiety) (Fig. 3B). In molecule C, the electron density for NAD is weaker and broken between the C-5Ј of the ribose-nicotinamide moiety and the O-5Ј of the second phosphate group of ADP (Fig. 3C). The B-factor of NAD bound to molecule C is also higher than in molecule A. However, unlike NAD bound to molecule B, the B-factor of the ADP-ribose moiety (76.7 Å 2 ) is higher than that of the nicotinamide moiety (49.8 Å 2 ) (Fig. 3C), showing that two parts of NAD can be stabilized differentially in the binding site. In molecule D, there is no electron density for the ribosenicotinamide moiety and the density for ADP moiety is fragmented (Fig. 3D). The presence of a poorly bound ADP molecule in the active site may be due to the NAD cleavage similar to that observed for NAD bound to molecule C, followed by the release of the ribose-nicotinamide. This NAD cleavage into ADP and ribose-nicotinamide is not the one that is expected by the NAD-glycohydrolase activity of the toxin, which hydrolyzes NAD into ADP-ribose and nicotinamide (Fig. 3E, yellow bond). A similar situation was observed in the crystal structure of Pseudomonas aeruginosa exotoxin A, a diphtheria-like toxin in complex with ␤-TAD, an NAD analog (PDB accession code 1AER) (15). In this structure, two molecules are present in the asymmetric unit with one that binds an intact ␤-TAD and the other that binds a ␤-TAD cleaved at this unexpected position. The origin of this cleavage is not understood. Therefore, of the four molecules seen in the asymmetric unit, molecules A and B appear to be the most interesting for investigating how NAD may bind to C3. Because molecule A binds an intact NAD and its functionally important ARTT loop is not constrained by crystal-packing contacts, we will now focus on this particular molecule to understand tentatively how NAD binds to C3 and may induce conformational changes.
Overall Binding of NAD to C3-The NAD lies in a prominent cleft formed by the junction of two antiparallel ␤-sheets (␤1-␤3; Fig. 1A). One end of this cleft is open to the solvent, while the other is closed by both the ARTT loop and an additional loop that links strands ␤3-␤4 (Fig. 1A). The latter, so-called "Phosphate-Nicotinamide" (PN) loop stabilizes one phosphate group and the nicotinamide moiety of NAD. The interactions between the toxin and the NAD are reported in Table III The aliphatic chain of Arg-91 lies against the adenine aromatic ring, whose N-6 atom forms a hydrogen bond with the carboxylate group of Glu-169 (␤2 strand). The NAD adopts an unusual highly folded compact structure, where the AMP and NMN moieties are both in a closed conformation (16). This results in both a short intramolecular distance of 4.8 Å between the PA phosphate and the N-9 atom of AMP moiety and an intramolecular hydrogen bond between the N-7 atom of nicotinamide and the PN phosphate group of the NMN moiety. Therefore, NAD binds to C3, like to other ADP-ribosyltransferase toxins (5,6,15,17,18).
The ARTT Loop Changes Its Conformation upon NAD Binding-We have made a detailed structural comparison of the ARTT loop between the unbound and NAD-bound forms of C3. Because the C3-free and C3-NAD structures have been solved in the same crystal form, this analysis is prevented from crystal-packing bias. The NAD-induced ARTT loop movement causes a swinging rearrangement of some of its side chains from a solvent-exposed environment to a buried conformation (Fig. 4). In particular, the side chain of Gln-212 makes a large switch (7 Å C␦-C␦ atoms) toward the interior of the NADbinding cleft, where it interacts with the O-2Ј-hydroxyl of the nicotinamide ribose. Also, but to a lesser extent, Ser-207 and Phe-209 move closer to NAD (Fig. 4, A and B). In this novel conformation, the ARTT loop and the three side chains covered the ribose-nicotinamide moiety in the NAD-binding site. The ribose-nicotinamide moiety is further buried by stacking with the aromatic side chain of Phe-183 from the PN loop (Fig. 4, A  and B). All these residue movements lead to the formation of an adjusted pocket that surrounds the ribose-nicotinamide moiety in the NAD-binding site (Fig. 4C). Gln-212 contributes to the formation of this pocket by forming a network of interactions (Fig. 4B). It interacts with Gln-182 from the PN loop, which in turn lies along the aromatic ring of Phe-183 and with a water molecule that links Ser-207 and the catalytic Glu-214. This water molecule takes the same position as the Gly-211 amide from the ARTT loop, which has moved by 6 Å upon NAD binding (Fig. 4, A and B). The catalytic Glu-214 is maintained in the same position in the free and NAD-bound forms for two reasons. First, the interaction between Glu-214 and Gly-211 in the free form is no longer possible in the NAD form due to the glycine movement but is replaced by indirect interactions with Gln-212 and Ser-207 through a water molecule. Second, the interaction of Glu-214 with Ser-174 from the STS motif, which forms the bottom of the site, is preserved. These two sets of interactions lock Glu-214, which thus interacts with the O-2Јhydroxyl of the nicotinamide ribose (Fig. 4B), in agreement with the proposed mechanism of catalysis (18,19). Therefore, upon NAD binding, the ARTT loop undergoes a major move- ment, accompanied by a rearrangement of two side chains of the PN loop, which lock the ribose-nicotinamide moiety in the binding site, without disturbing the critical position of the catalytic Glu-214. Upon this NAD-induced conformational change, the ARTT loop is further stabilized. Thus, whereas in the free form the B-factor of the ARTT loop is higher (76.9 Å 2 ) than the B-factor of the overall molecule (62.8 Å 2 ), in the NAD-bound form it becomes much lower (26.9 Å 2 ) than the B-factor of the overall molecule (33.9 Å 2 ). The ARTT loop conformational change, however, is not an absolute prerequisite for NAD to bind to C3. Fig. 1B reveals that NAD binds C3 in the crystal, whereas the conformation of the ARTT loop remains similar to that observed in the free form. However, inspection of B-factors revealed that NAD bound in molecule B (B ϭ 56 Å 2 ) is less stabilized than in molecule A (B ϭ 33.5 Å 2 ). The conformational change of the ARTT loop, as seen in molecule A, may be responsible for the stabilization of bound NAD. Whether or not the NAD-induced network of interactions contributes to the stabilization process of NAD has been investigated by the mutational experiments described hereafter.
Other Conformational Changes-Another important change that occurs in C3 upon NAD binding is a "crab-claw" movement, which closes and opens the ADP moiety-binding site. It concerns a large subdomain consisting of three strands and a helix (␤8-␤7-␤2 and ␣5), which undergoes a hinge movement of ϳ9°with respect to the rest of the molecule (Fig. 5). Structural superimposition on helices ␣2, ␣3, and ␣4 between the unbound and NAD-bound forms for each of the four C3 molecules in the asymmetric unit showed that in molecules A, B, and C this movement closes the ADP moiety-binding site by 5°, 5.8°and 4°, respectively. The major consequence is the shift of the C-terminal extremity of the strand ␤1 by ϳ1.7 Å toward the ADP moiety of NAD and consequently the formation of interactions of residues Asp-130 and Asp-131 with the hydroxyl of the adenine ribose. In contrast in molecule D, where only ADP is bound, the ADP moiety-binding site is more open by 3°than when empty (Fig. 5). The opening of the ADP moiety-binding site together with the weak electron density observed for ADP (Fig. 3D) suggests that the co-substrate is less stabilized in this conformation. Therefore, the crab-claw closure may correlate with NAD stabilization, whereas the opening might reflect the local conformation associated with product release. This crabclaw movement is independent of the ARTT loop conformation  A "serine-threonine-serine" sequence, called the "STS motif " (strand ␤3; Fig. 1) stabilizes the NAD-binding cleft by connect-ing the two perpendicular ␤-strands, which form the bottom of the cleft. The first serine of the STS motif, Ser-174, which is important for catalytic function in other toxins (20,21), adopts two different orientations in our structure. In the first orienta- tion, Ser-174 interacts with the carboxylate of the catalytic Glu-214 and the hydroxyl of Tyr-79 (Fig. 4B), as observed in the binary toxin, VIP2, which suggests a role in properly positioning the catalytic glutamate (18). In the second orientation, Ser-174 interacts with the phosphate group of the NMN moiety, close to the NAD cleavage observed in molecule C. In this position, it may play a role in NAD hydrolysis and could be responsible for the hydrolysis observed in our crystal structure.
Mutational Experiments-Structural data have shown that Gln-212 and to a lesser extent Gln-182 are involved in the conformational changes induced by NAD binding. We therefore probed their contribution to NAD binding and to both NADglycohydrolase and ADP-ribosyltransferase activities by substituting them individually into alanine. Plasmon resonance experiments showed that NAD binds to the Q182A and Q212A mutants with approximately the same and 2-fold lower affinity as compared with the wild-type enzyme, respectively (Table  IV). These data show that neither of these two residues plays a major role in the binding energy of NAD to C3. The NADglycohydrolase activity of these mutants was probed by following the [ 32 P]NAD hydrolysis by chromatography. Fig. 6a shows that the Q212A mutant has virtually the same activity as the wild-type enzyme and that the Q182A mutant even has a slightly increased activity. Neither of these two residues is a major contributor to the NAD-glycohydrolase activity of C3. A different result was obtained for the ADP-ribosyltransferase activity as deduced from a qualitative monitoring of the transfer of the [ 32 P]ADP-ribose moiety to RhoA by electrophoresis. Thus, we found that the Q212A mutant was virtually inactive, whereas the Q182A mutant was not affected in its capacity to ADP-ribosylate Rho (Fig. 6b). Clearly, Gln-212 plays no major role in NAD hydrolysis but is critical for the ADP-ribose transfer to RhoA, whereas Gln-182 does not have a predominant function in either activity.
The contribution of Arg-186 from the PN loop was also probed by mutagenesis. The reason for this experiment was that this residue interacts with the phosphate group of the NMN moiety of NAD and is conserved in most C3-like exoenzymes and binary toxins. To evaluate the role of Arg-186 in NAD binding and ADP-ribosyltransferase activity, we generated the mutant R186E, with a charge reversal at this position. This mutant was unable to bind NAD (Table IV) and hence had no catalytic activity (Fig. 6b). Therefore, Arg-186 is involved in the interaction with NAD.
Our structural observations have revealed that in the presence of NAD, Ser-174 from the STS motif can adopt two different orientations; one where it interacts with the catalytic Glu-214 and another where it interacts with the phosphate group of the NMN moiety of NAD. To investigate the role of Ser-174 in the catalytic mechanism, we prepared the S174A mutant and tested its ADP-ribosyltransferase activity. This mutant retains the wild-type ADP-ribosyltransferase activity (Fig. 6b), suggesting that Ser-174 plays no major role in C. botulinum C3 exoenzyme activity. DISCUSSION This structural study has shown that binding of NAD to the C3 exoenzyme has induced various conformational changes, differentially limited by the crystal-packing environment. Two major regions of the toxin are involved in these movements. The first one concerns the functionally important ARTT loop (14), which undergoes an ample motion that occurs only in the absence of local crystal-packing constraints. Such a movement is not unique to C3 because upon NAD analogs binding to P. aeruginosa exotoxin A the region equivalent to the ARTT loop also undergoes a conformational change (22). The NAD-induced ARTT loop movement observed in C3 causes a large swinging rearrangement of some of its side chains from a solvent-exposed environment to a buried conformation in the NAD-binding site. In this latter conformation, these side chains cover the nicotinamide ribose moiety and hence contribute to the formation of a network of interactions that also involves residues from the adjacent PN loop. Surprisingly, mutational experiments on Gln-212 from the ARTT loop and Gln-182 from the PN loop that are both involved in this network of interaction revealed that these residues make no predominant energetic contributions to NAD binding. Therefore, the NAD-in-  Because of its lower polarity, the migration distance of NAD is longer than that of the hydrolyzed product. b, 12% SDS-PAGE of RhoA after incubation with [ 32 P]NAD for 1 h at 37°C in the presence of wild-type or mutant C3. Radioactivity of the ADP-ribosylated RhoA was directly detected on the dried gel by an imaging system. duced ARTT loop conformational change may not be required for NAD to bind to C3. A second major NAD-induced movement of C3 concerns a large subdomain that undergoes a crab-clawlike motion and thus closes the binding site of the AMP moiety of NAD. As a result of this movement, several residues of this subdomain establish direct contacts with NAD, including Arg-128, Gly-129, and the two aspartate 130 and 131 residues. The equivalent subdomain of the binary toxin of B. cereus vegetative insecticidal protein VIP2 also undergoes such a movement (18), suggesting that this motion may be a common means for ADP-ribosyltransferase toxins to bind NAD. Additional though less ample changes were noticed in the vicinity of the NAD binding pocket as a result of NAD binding. They concern residues of the PN loop, which include Gln-182 and Phe-183, whose side chains tend to stack above the nicotinamide ring, and Arg-186 whose side chain interacts with the phosphate group of the NMN moiety of NAD. Therefore, upon NAD binding several regions of the NAD binding pocket undergo structural rearrangements of variable amplitudes.
What could be the biological implications, if any, of these NAD-induced conformational changes in C3? Binding of NAD to C3 is a prerequisite for the Rho ADP-ribosylation reaction to occur (23). Also, evidence has been accumulated to indicate that the ARTT loop, and in particular Gln-212 are critical factors of the ADP-ribosylation process (Refs. 4, 14 and this study). We found that the ARTT loop may exist in two distinct conformations when NAD is bound to C3. In one of them, the loop adopts a solvent-exposed side chain conformation as a result of local crystal-packing constraints that prevent it from moving freely. In the other, the loop, which has undergone a large swinging movement upon NAD binding, adopts a side chain-buried conformation. Therefore, we propose that this latter conformation, which is unconstrained by crystal-packing contacts may predominantly reflect the inherent conformational trend of the ARTT loop in the presence of NAD. This might be a privileged form for the C3-NAD complex to bind to Rho. In this structure Gln-212 is buried in the NAD binding pocket, and hence it is inaccessible to Rho, a situation that is not incompatible with the previous observation that this residue is not an energetically important contributor of Rho binding (4).
The role of Gln-212 in the ADP-ribosyl transfer to Rho (4, 24) is presumably specific because this residue neither contributes to NAD binding (this work and Ref. 4) nor to Rho binding (4,24), nor to NAD hydrolysis (this work and Ref. 4). On the basis of structural comparison between binary toxins and C3-like exoenzymes, a model has been proposed that suggests that Gln-212 is solvent-exposed to interact specifically at position Rho Asn-41 for ADP-ribosylation. As discussed above, our data suggest a plausible Rho-binding conformation of the C3-NAD complex in which Gln-212 is buried and, hence, unlikely to be directly accessible to Rho Asn-41. Would the binding of Rho be able to trigger a further conformational change of the ARTT loop to make Gln-212 switching from a buried to a solventexposed conformation? This is not inconceivable because in the presence of NAD and a particular crystal-packing environment (Fig. 1B), the side chain of Gln-212 exists in an exposed conformation.
We found it intriguing that Ser-174 interacts with the catalytic Glu-214 in both the unbound and NAD-bound molecule A, while it changes its orientation in the NAD-bound molecule B, where it makes contact with the phosphate group of the NMN moiety of NAD. However, the mutant S174A showed no difference in ADP-ribosyltransferase activity, discarding the possibility that Ser-174 may be a major actor in this activity, as also recently shown for the Staphylococcus aureus C3 exoenzyme (4). Along similar lines of thought, we were intrigued by the slight movement undergone by Arg-186, which switched from a flexible state to a well defined interacting conformation with the same phosphate group of the NMN moiety. We found that reversal of the charge of Arg-186 made C3 unable to bind to NAD and hence to express an ADP-ribosyltransferase activity. Because Arg-186 is exclusively conserved in most C3-like exoenzymes and binary toxins, we suggest that this residue is a functionally important feature of most C3-like exoenzymes and binary toxins.
Altogether, our observations suggest that the mechanism of Rho ADP-ribosylation by C3 may be more complex than originally anticipated. The plasticity of the critical ARTT loop evidenced in this work may have to be taken into account to understand this mechanism. We propose that the C3 activity may require several subtle rearrangements of the ARTT loop involving at least three steps: NAD binding, Rho binding, and Rho ADP-ribosylation. A similar multistep mechanism is also suggested by a mutational study in which the non-substrate Rac was progressively transformed into a C3-binding form and then into a true C3 substrate form (25).