Characterization of the Active Site of ADP-ribosyl Cyclase*

ADP-ribosyl cyclase synthesizes two Ca2+ messengers by cyclizing NAD to produce cyclic ADP-ribose and exchanging nicotinic acid with the nicotinamide group of NADP to produce nicotinic acid adenine dinucleotide phosphate. Recombinant Aplysia cyclase was expressed in yeast and co-crystallized with a substrate, nicotinamide. x-ray crystallography showed that the nicotinamide was bound in a pocket formed in part by a conserved segment and was near the central cleft of the cyclase. Glu98, Asn107 and Trp140 were within 3.5 Å of the bound nicotinamide and appeared to coordinate it. Substituting Glu98 with either Gln, Gly, Leu, or Asn reduced the cyclase activity by 16–222-fold, depending on the substitution. The mutant N107G exhibited only a 2-fold decrease in activity, while the activity of W140G was essentially eliminated. The base exchange activity of all mutants followed a similar pattern of reduction, suggesting that both reactions occur at the same active site. In addition to NAD, the wild-type cyclase also cyclizes nicotinamide guanine dinucleotide to cyclic GDP-ribose. All mutant enzymes had at least half of the GDP-ribosyl cyclase activity of the wild type, some even 2–3-fold higher, indicating that the three coordinating amino acids are responsible for positioning of the substrate but not absolutely critical for catalysis. To search for the catalytic residues, other amino acids in the binding pocket were mutagenized. E179G was totally devoid of GDP-ribosyl cyclase activity, and both its ADP-ribosyl cyclase and the base exchange activities were reduced by 10,000- and 18,000-fold, respectively. Substituting Glu179 with either Asn, Leu, Asp, or Gln produced similar inactive enzymes, and so was the conversion of Trp77 to Gly. However, both E179G and the double mutant E179G/W77G retained NAD-binding ability as shown by photoaffinity labeling with [32P]8-azido-NAD. These results indicate that both Glu179 and Trp77 are crucial for catalysis and that Glu179 may indeed be the catalytic residue.

(cADPR) 1 (1,2). Accumulating evidence indicates cADPR is a Ca 2ϩ messenger important in regulating various cellular functions in a variety of species from plant to human (reviewed in Refs. 3 and 4). Similarly, the cyclase has since been shown to be ubiquitous and is particularly abundant in Aplysia ovotestis (5)(6)(7)(8)(9). Sequence comparison reveals that the Aplysia cyclase is homologous to antigens CD38 on human lymphocytes and BST1/BP3 on bone marrow cells (10,11). In addition to sharing 25-30% sequence identity, there is also a perfect alignment of the cysteines of the homologs. Subsequent work shows that all members of the cyclase family are also functionally similar, since they all are enzymes catalyzing the cyclization of NAD to cADPR (Refs. 12 and 13; reviewed in Refs. 14 and 15). More surprising is the discovery that all these enzymes also can use NADP as a substrate and catalyze the exchange of its nicotinamide group with nicotinic acid, producing nicotinic acid adenine dinucleotide phosphate (16), which is capable of activating a novel Ca 2ϩ release mechanism that is totally independent of that activated by cADPR or inositol trisphosphate (17). This novel nicotinic acid adenine dinucleotide phosphate-dependent mechanism was first shown to be present in invertebrate eggs and oocytes (17,18) and has recently been found to be important in coordinating the hormonal Ca 2ϩ signaling functions in mouse pancreatic acinar cells (19), indicating that it is likely to be also a signaling mechanism of general relevance.
It is uncommon that a single enzyme can use two different substrates and produce two structurally and functionally distinct signaling molecules. A model has been proposed to account for this novel multiplicity in catalysis (15, 20 -23). The salient feature of this catalytic model is the proposal that a single intermediate is responsible for cyclization, hydrolysis, and base exchange chemistries of the cyclase family of enzymes. Experimental support for this model has recently been provided (23). x-ray crystallography reveals that the Aplysia cyclase is a bean-shaped molecule with a central cleft (24). In solution as well as in the crystalline state, the cyclase forms dimers from two monomers in a head-to-head fashion enveloping a central cavity the size of a molecule of cADPR (24,25). This structural feature suggests that the active site of the enzyme may be at the central region of the molecule. In this study, the Aplysia cyclase was co-crystallized with nicotinamide, a substrate for the base-exchange reaction. The active site of the enzyme was identified and characterized by x-ray crystallography and site-directed mutagenesis.

MATERIALS AND METHODS
Expression of the Aplysia ADP-ribosyl Cyclase in Yeast-The Pichia expression vector pPICZ␣A (Invitrogen, Carlsbad, CA) was used, and the construct consists of cloning the wild type or mutant cyclase in frame with the yeast mating ␣-factor signal sequence. The details of the * This work was supported by National Institutes of Health (NIH) Grants HD17484 (to H. C. L.) and DA11806 (to T. F. W.). CHESS is supported by National Science Foundation Grant DMR-9311772, and MacCHESS is supported by NIH Grant RR-01646. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  construct were as we have described previously (26 -28). The tightly controlled alcohol oxidase promotor was used to regulate the expression of the construct. Yeast (GS115 strain) transfected with the construct was grown in a fermentation bioreactor, and expression was induced by methanol. The expressed cyclase was secreted as a soluble protein in the fermentation medium, and as much as 0.3 g/liter could be obtained (26,27). Endogenous proteolytic enzymes that normally are responsible for processing the prepro-␣-mating factor also removed the ␣-factor signal sequence from the expressed cyclase.
Some of the mutant proteins were also expressed in the X-33 strain (Invitrogen) of Pichia pastoris. This strain is isogenic to the GS115 strain, except that it contains the native alcohol oxidase AOX1 and HIS4 genes. The AOX1 gene allows the yeast to metabolize methanol and grow to higher densities, particularly during fermentation. The native histidine gene, HIS4, facilitates production of endogenous histidine and allows rapid growth of the yeast without any external supplementation of histidine. The dominant selectable marker, Zeocin, also allows for use of positive selection vectors pPICZ␣A. With this vector the transformation efficiency was nearly 100%, and no revertants were detected. After a 5-day induction with methanol in a standard culture tube, the yeast produced and secreted up to 40 g/ml of the mutant cyclase in the culture medium. A 10-ml culture provides more than enough protein for all of the enzymatic characterizations, and a bioreactor is not needed.
Production of Antibodies against the Aplysia ADP-ribosyl Cyclase-Polyclonal antibodies were produced in rabbits by Genosys Biotechnologies, Inc. (The Woodlands, TX). The recombinant wild type cyclase produced as described above was used as antigen. Antibodies against the conserved region of the Aplysia cyclase were also raised using a synthetic peptide, TLEDTLLGYL. Western blots were visualized by chemiluminescence using a kit from Amersham Pharmacia Biotech.
Enzyme Assays-The wild type or mutant cyclase (50 -3,000 ng/ml) was incubated (2-120 min) at room temperature with various concentrations of NAD (25-1,000 M) or nicotinamide guanine dinucleotide (NGD) (10 -2,000 M) at pH 7.8 buffered by 40 mM Tris-HCl. The total volume of the reaction mixture was 0.12 ml, and the reaction was stopped by the addition of SDS (0.17% final concentration). The reaction products were analyzed by HPLC (BioCad) using either a Poros 10 HQ column (Roche Molecular Biochemicals) or an AG MP-1 column (6.6 ϫ 150 mm). The resin AG MP-1 was obtained from Bio-Rad. The elution was performed either using a gradient of NaCl (HQ column) or trifluoroacetic acid (AG MP1 column) at a flow rate of 5 ml/min. The base exchange reaction was similarly assayed at pH 4.5 in the presence of 50 mM nicotinic acid and various concentrations of NADP (20 -4,000 M). Enzyme activities of the wild type and mutant cyclases were measured, each using two or three different protein preparations. The results shown are mean Ϯ S.E. of 9 -12 determinations. The V max and K m values were obtained from double-reciprocal plots, and the results shown are mean Ϯ S.E. of at least five determinations.
Site-directed Mutagenesis of ADP-ribosyl Cyclase-Primers for the 5Ј-and 3Ј-ends of the cyclase cDNA were as follows: 1) CYC-F2, hybridizing to the 5Ј region of mature soluble cyclase starting with IVPTR . . . and also generating an EcoRI restriction site, 5Ј-CAT GAA TTC ATC GTC CCC ACT CGC-3Ј and 2) CYC-R, hybridizing to the 3Ј region of cyclase as well as generating a NotI site, 5Ј-AAT GCG GCC GCG AAA TTG GTA AAG AGC-3Ј. Amplification was carried out using 0.5 unit of Vent polymerase (New England Biolabs), 280 M dNTPs (New England Biolabs), and 0.4 M primers in a 50.0-l reaction. The mutations were carried out in two sequential amplification steps. During the first amplification, two fragments were generated. The primer CYC-F2 and an antisense primer for a specific mutation were used to generate one fragment, and the primer CYC-R and a sense primer for the same mutation were used to generate the other. For both of these polymerase chain reactions, the native cyclase sequence served as the template. The two fragments were gel-purified and used for the second amplification, during which the purified fragments served as the templates, and primers CYC-F2 and CYC-R were used to generate the 0.77-kilobase pair cyclase containing the mutation. The mutated cyclase was then gel-purified, digested with EcoRI and NotI, and ligated into a yeast expression vector that we had constructed previously (27).
Some mutations were done using the QuickChange site-directed mutagenesis kit obtained from Stratagene. The kit employs the high fidelity PfuTurbo DNA polymerase and mutant oligonucleotide primers to replicate both strands of the supercoiled plasmid containing the insert. A mutated plasmid containing staggered nicks is generated. Following thermal cycling, the reaction mixture is treated with DpnI endonuclease, which digests hemimethylated parent DNA leaving behind the nicked DNA containing the mutation of interest. Plasmid DNA from most Escherichia coli strains is dam-methylated and can be digested away with the DpnI enzyme. This method eliminates the use of various subcloning procedures necessary to generate mutants. The procedure only takes 7 days, starting from the mutation reaction, to transforming E. coli with the mutated DNA and purifying the resulting SacI-digested DNA for use in yeast transformation.
Co-Crystallization of the Recombinant ADP-ribosyl Cyclase and Nicotinamide-Crystals were grown using the hanging drop method. Crystallization conditions for the recombinant cyclase were initially determined and optimized. These conditions were subsequently used to cocrystallize the enzyme with nicotinamide. Co-crystals were obtained using a reservoir solution containing 17% (w/v) polyethylene glycol (molecular weight of 4,000) and 0.1 M imidazole buffer (pH 7.5). The crystallization drops contained 2.5 l of the reservoir solution and 2.5 l of concentrated protein (OD ϭ 12 at 280 nm). Crystallization trays were stored at 18°C, and crystals displaying thin rod-shaped morphology with a length typically of 0.5 mm generally appeared within 7 days.
X-ray Structure Determination-x-ray diffraction data were collected from a single co-crystal using the A-1 station ( ϭ 0.919 Å) at the Macromolecular Diffraction Facility at the Cornell High Energy Synchrotron Source (MacCHESS). To reduce radiation damage, the crystal was cooled to liquid nitrogen temperature. Prior to this, the crystal was placed in a cryosolvent containing 22% polyethylene glycol 4000 to prevent ice formation within the crystalline lattice.
The diffraction data were collected using the 82 ϫ 82-mm 2 K ϫ 2 K Princeton CCD-based x-ray detector (29). The data indexed as a triclinic lattice and extended to 2.4 Å, displaying a very high anisotropic mosaicity, which varied from 1.75 to 3.5°. Data reduction and scaling yielded a partially complete data set with a unit cell of 60.4 ϫ 75.3 ϫ 138.1 Å. The overall R sym for the data set was 6.8%, with the highest resolution shell of 2.6 -2.4 Å having a high R sym of 31%.
The x-ray structure of the nicotinamide-cyclase complex was solved by molecular replacement using the program AMORE (30). A search model was constructed from the previously determined dimeric structure of ADP-ribosyl cyclase. Because the two cyclase monomers of this dimer were built and refined independently, a modified dimer was constructed using only the chain A monomer truncated at residue 247. Side chain atoms were retained in the search model.
Four solutions to the rotation function were found. The translation search using these orientations resulted in the position of four individual dimers in the asymmetric unit of the unit cell. The resulting positions of the four cyclase dimers displayed reasonable crystal packing contacts, suggesting that the solutions were correct. The four molecules were found to be independently positioned within the asymmetric unit lacking noncrystallographic symmetry. This was verified by calculating self-rotation functions using the program GLRF (31). The best solution according to the AMORE statistics had a correlation coefficient of 42.9 and an R-factor of 41.3.
The presence of four dimers in the asymmetric unit results in a favorable Mathews coefficient of 2.59. The four dimers give a total molecular mass of 240 kDa. The solvent content of the crystals is calculated to be 53%.
Structure Refinement-Residues 127-129 were mutated to alanines in each of the eight monomers for subsequent check of the quality and bias of electron density maps to be calculated. X-ray structure factors were calculated using the position of the atoms resulting from the placement of the four dimers in the unit cell, and these resulting structure factors were refined against the measured diffraction data. 5% of the reflections were set aside for calculating the R-free parameter. Rigid body refinement was conducted using the program X-PLOR (32) and all data between 8 and 3 Å. Because the native monomeric structure is composed of two discrete domains (the N-terminal domain consisting of residues 1-66 and residues 100 -150, the C-terminal domain consisting of residues 74 -97 and residues 153-250), the refinement was carried out with each domain defined as a rigid body. In addition, rigid body refinement was conducted with each monomer defined as a rigid body. The approach where each monomer was refined as two discrete domains gave the lower R-factor. With an overall B-factor assigned as 20, the R-factor after rigid body refinement was 36.3%.
Although no proper NCS relations are present in the crystal structure, each domain of each monomer can be related to that of a reference molecule by translation and rotation. In this manner, strict NCS restraints (weight ϭ 500 kcal/mol-Å 2 in X-PLOR) were applied to the domains of all of the monomers, thereby reducing the number of parameters to be refined. Simulated annealing refinement was conducted against all data between 8 and 2.4 Å, resulting in an R-factor of 31.2% (R-free ϭ 34.0%). Including a bulk solvent model and using all data between 20 and 2.4 Å in another cycle of simulated annealing refine-ment resulted in an R-factor of 28.4% (R-free ϭ 31.6%). Finally, tightly restrained individual B-factor refinement gave an R-factor of 24.6% (R-free ϭ 28.0%).
Wild type or mutant cyclases (0.3 g) were incubated on ice in 50-l droplets containing 50 M [ 32 P]8-azido-NAD (ϳ530,000 cpm), 0 -5 mM ␤-NAD, and 0.5 mM ␣-NAD. The droplets were exposed to 160 mJ of ultraviolet light energy (ϳ45 s) in a Stratalinker chamber (Stratagene, La Jolla, CA). Pilot experiments showed that this amount of UV energy produced optimal labeling without detectable protein-protein crosslinking. ␣-NAD is not a substrate for the cyclase (5) and was included in all samples to reduce potential nonspecific labeling. Samples were analyzed by SDS-polyacrylamide gel electrophoresis, and the radioactivity associated with the proteins was measured using a phosphor imager (Packard, Meriden, CT). The phosphor images were collected at 16-bit resolution for quantitative analyses but reduced to 8-bit for printout.

RESULTS
X-ray Crystallography of the Aplysia ADP-ribosyl Cyclase-We have previously shown that nicotinamide, similar to nicotinic acid, can serve as a substrate for the base exchange reaction (16). In this case, the product is the same as the substrate (NAD or NADP), and the enzyme appeared inhibited. However, the rapid exchange can be readily demonstrated using radioactive substrate (16). That nicotinamide is a substrate indicates that it should have affinity for the active site of the cyclase. We thus crystallized the recombinant cyclase with nicotinamide. The structure of the complex was refined using various procedures as described under "Materials and Methods," giving an R-factor of 24.6% (R-free ϭ 28.0%). Electron density maps calculated from these refined coordinates suggest that there are additional atoms present in the crystal structure. None of the maps were biased by substrate assignments, yet in the initial unaveraged 2F o Ϫ F c and F o Ϫ F c maps and all subsequent maps, strong peaks appeared where the assignments had been made. Before map averaging, the peaks were around 2.75 in the F o Ϫ F c map. After 8-fold averaging using the RAVE programs (35), the peaks were as large as 6 . In particular, between Glu 98 and Trp 140 there was strong density of the shape of a six-member ring resembling nicotinamide. Fig. 1A shows the van der Waals contact surface of a cyclase dimer with one monomer colored light blue and the other dark blue. Each monomer was a bean-shaped molecule consisting of two domains separated by a central cleft. The two monomers associated with each other in a head-to-head fashion forming a donut-like molecule with a central cavity (24). We have previously shown by chemical cross-linking and dynamic laser light scattering that the cyclase forms dimers in solution as well (25).
The tentative assignment of the bound nicotinamide is shown in yellow in Fig. 1A. The binding site was in a pocket at the central cleft of the monomer and was very close to the central cavity of the dimer. Nicotinamide was bound to each monomer. The binding pocket of the monomer colored dark blue was not visible, because it was on the other side of the monomer. The cyclase has a stretch of six amino acids (TLEDTL) that are highly conserved with CD38, a homolog that also possesses similar enzymatic activities as the cyclase (10,12). This conserved region formed part of the binding pocket and is shown in orange in Fig. 1A. Fig. 1B shows the three amino acids in the binding pocket that were closest to the placement of the bound nicotinamide (yellow) and appeared to complex the bound substrate. These three residues were Trp 140 (2.8 Å), Glu 98 (3.5 Å), and Asn 107 (3.1 Å). The conserved sequence is shown as orange sticks, and Glu 98 is part of this sequence. Trp 140 , although not in the conserved region, is nevertheless conserved among two species of the cyclase and three species of CD38 (24). Asn 107 , on the other hand, is not a conserved residue. If these amino acids are indeed responsible for substrate binding, their alteration by site-directed mutagenesis should result in reduction or elimination of the enzymatic activities of the cyclase.
Site-directed Mutagenesis of the ADP-ribosyl Cyclase-Various mutations of the three amino acids closest to the bound nicotinamide were produced, and the resulting mutant cyclases were expressed in yeast as secretory proteins. The yeast media containing the mutant proteins were dialyzed and purified by treatment with an anion exchange resin (DEAE), which removed nondialyzable contaminants in the yeast media. SDSpolyacrylamide gel electrophoresis showed that the preparations of the mutant proteins after the DEAE treatment contained essentially a single protein with molecular weight of about 29,000, similar to that of the wild type protein (Fig. 2). Examples of Western analysis of some of the mutant proteins are shown in Fig. 2. All mutant proteins are recognized by the polyclonal antibody against the wild type cyclase. On the other hand, the antibody raised against the conserved peptide (anti-ConPep) recognized the wild type, N107G, and W140G proteins but not the mutant proteins E98L and E98G. Since Glu 98 is one of the amino acids in the conserved sequence, its change to leucine or glycine is expected to render the mutant proteins unrecognizable by the antipeptide antibody. The DNA from the regions containing the cyclase insert in all four mutants was amplified by the polymerase chain reaction using appropriate primers. The sequence results confirmed that they harbored the correct mutations (Fig. 2, lower panel). Fig. 3 shows that substitution of Glu 98 with either Gln, Gly, Leu, or Asn resulted in a 16 -222-fold decrease in the rate of cADPR synthesis as compared with the wild type cyclase, indicating that Glu 98 is important for catalysis. The large decrease in activity is likely to be due to the elimination of the negative charge on Glu 98 by all these mutations. The size of the substituting amino acids appears to have a secondary effect on the activity. The mutant with the conservative substitution by Gln, which is similar in size to Glu, had only about 6% of the activity of the wild type enzyme. Smaller side chains such as Leu produced a slightly larger decrease in activity. The conversion to glycine (E98G), which is equivalent to deleting the side chain of the glutamate, produced similar inhibition. Much more destructive is substituting with a residue only slightly smaller than Gln, such as Asn, which resulted in more than 200-fold decrease in activity. It is unlikely that the inhibitory effect of the mutation was due to nonspecific alterations of the structure of the enzyme, since similar decreases were seen with such conservative substitutions as Asn and Gln, which are generally considered to have minimal effect on protein structures. More likely is that the enzyme activity has an exquisitely specific requirement for both the charge and size of the side chain at the 98-position, and any alteration would result in a decrease in activity. Table I summarizes and compares the enzymatic parameters of several mutant cyclases with the wild type. Of the three amino acids that complex the nicotinamide, Trp 140 is the most critical to the cyclase activity. It is closest to the bound nicotinamide, and its conversion to glycine resulted in a 4,648-fold decrease in activity. Asn 107 is the least critical; only a 2-fold decrease in activity was measured for the mutant N107G. Asp 99 is next to Glu 98 and is similarly charged. It is also in the highly conserved sequence, TLEDTL. Nevertheless, its conversion to glycine produced much less pronounced effect on the cyclase activity, only a 10-fold decrease. These results indicate that the effect of the site-directed mutagenesis is residue-specific and that Trp 140 is the most critical residue for the cyclase activity.
In addition to comparing the maximal cyclase activity at high substrate concentration (V max values), Table I also lists V max /K m ratios of the wild type and mutants. The Michalis-Menten model predicts that at low substrate concentrations the enzyme activity is proportional to the substrate concentration and the proportionality constant is the V max /K m ratio. The pattern of changes of the ratios follows that of the V max values, indicating that the mutant cyclases had lower activity at both limits of high and low substrate concentration. The kinetic parameters for W140G were not determined because its activity was too low. The value shown in Table I was measured at 1 mM NAD, which should be approximately the maximal activity. Fig. 4 compares the cyclase activity with the base exchange activity of five of the mutants. The pattern of inhibition of both activities was similar. The mutant W140G had the least cyclase and exchange activity, while N107G had the most. The mutation thus affected both enzymatic activities similarly, suggesting that the same active site is responsible for catalyzing both reactions.
We have previously shown that the wild type cyclase can also use NGD, an analog of NAD, as substrate and cyclize it to cyclic GDP-ribose (cGDPR), a fluorescent analog of cADPR (21,36). Table II compares the GDP-ribosyl cyclase activities of the wild type and the mutants. It can be seen that, contrary to the large decrease in the ADP-ribosyl cyclase activity (cf. Table I), all mutant cyclases retained substantial GDP-ribosyl cyclase activity. Indeed, both Glu 98 mutants had activity 2-3-fold higher than the wild type. Even the W140G mutant, which only had residual ADP-ribosyl cyclase activity (Table I), had substantial GDP-ribosyl cyclase activity. Its K m value, however, was close to 40-fold higher. This disparity of the effect of the mutations on the two enzyme activities suggests that Glu 98 and Trp 140 may be important mainly in the positioning of the substrate and that the binding of NGD and NAD may involve different amino acid residues. The two amino acids thus may not be directly participating in catalysis.
To search for the catalytic residues, other amino acids that make up the binding pocket (cf. Fig. 1, A and B) were mutagenized. Table III shows that the mutant cyclase, E179G, was totally devoid of GDP-ribosyl cyclase activity, and both its ADP-ribosyl cyclase and the base exchange activities were reduced by more than 10,000-and 18,000-fold, respectively. Conversion of Trp 77 to glycine likewise resulted in large decreases in all three enzymatic activities as compared with the wild type cyclase. Furthermore, all enzymatic activities were eliminated in the double mutant, W77G/E179G. Indeed, the enzymatic activities of these mutant cyclases were so low that it was difficult to obtain reliable measurements at lower substrate concentrations. The results listed in Table III were thus measured only at a high substrate concentration so that the rates were close to the V max values of the respective enzymatic activities. These results indicate that Glu 179 and also Trp 77 are most likely the catalytic residues.
The critical importance of Glu 179 was further shown by substituting it with various other residues. As shown in Table IV, all substitutions resulted in inactivation of the enzyme. This was true also for the highly conservative substitution with residues such as Gln, Asn, or even Asp. Although the substitution with Asp maintained the charge, the size of the side chain was smaller. The importance of size was also seen in the large differences in activity from changing Glu 98 to Asn as compared with changing it to Gln (Fig. 3).
Although mutations at Glu 179 and Trp 77 produced inactive enzymes, the mutant cyclases did maintain the ability to bind NAD. Fig. 5 shows the photoaffinity of the wild type and the mutant cyclases. The labeling was competitively inhibited by increasing concentrations of NAD, indicating the labeling was specific. Even the double mutant, which had no observable enzyme activity (Table III), retained the NAD-binding ability. The half-maximal inhibition was about 0.1 mM NAD for the wild type and about 0.5 mM NAD for E179G and the double mutant. That NAD was less effective in competing in the case of the mutants suggests that the affinity of the mutant active sites for NAD is less than the wild type. As shown in the inset of Fig. 5 (lane 1), in the absence of NAD, the labeling intensity of E179G and the double mutant was actually higher than the wild type although the same amounts of protein were loaded onto the gel and the labeling was performed under identical conditions. In fact, quantification of the phosphor image shows that the labeling intensity of E179G and the double mutant was 2-and 1.5-fold higher, respectively, than that of the wild type. These results indicate that the inactivating effect of the mutations is not due to nonspecific alteration of the protein structure, which would expectedly destroy the substrate binding ability of the enzyme.
The Structure of the Active Site of ADP-ribosyl Cyclase- Fig.  1C shows a stereo view of the active site. The van der Waals contact surface of the site is shown with the three most critical amino acids, Trp 140 , Trp 77 , and Glu 179 , rendered as spacefilling models. The active site is in a pocket with Glu 179 deep inside the pocket, while Trp 140 and Trp 77 line the rim of the pocket, one on each side. The carboxyl oxygen atoms of Glu 179 and the carbonyl oxygen atom of Trp 77 , which are likely to participate directly in catalysis, are shown in red. The bound nicotinamide is shown in yellow. The aromatic ring of Trp 140 is close to, and appears to position, the bound nicotinamide, presumably through hydrophobic interaction. DISCUSSION In this study, the active site of the cyclase was first identified by co-crystallizing the cyclase with nicotinamide, a substrate of the base exchange reaction. X-ray crystallography allowed the unambiguous identification of the three amino acids in the cyclase that were closest to the bound nicotinamide. To determine if they are the catalytic residues, site-directed mutagenesis was used to alter these amino acids. Although large de-  creases in the cyclase and the base exchange activities were observed in the mutants, they nevertheless retained substantial GDP-ribosyl cyclase activity. The results suggest that the residues are most likely important in positioning of the substrate but may not be directly participate in catalysis. Indeed, our previous results have suggested that NAD and NGD may bind to the active site differently. Since the site of cyclization in cADPR is at the N1-position of the adenine, but is at the N7-position of the guanine in cGDPR (36,37), the purine ring must thus be in the anti-configuration during cyclization in the case of NGD but is in the syn-configuration in the case of NAD. These structural differences of the products are consistent with the idea that different amino acid residues may be involved in binding of the two substrates.
It is reasonable to expect that the catalytic residues should be closer to the substrate, NAD, instead of nicotinamide. It is technically difficult to co-crystallize the cyclase with NAD because it can quickly convert NAD to its product, cADPR, which may not have much affinity for the active site of the cyclase. Nevertheless, it is logical to assume that the NAD-binding site must be close to the nicotinamide-binding site. Asp 99 was a good candidate, since it was next to Glu 98 , one of the coordinating residues for the bound nicotinamide, and it was also in the highly conserved sequence. Conversion of Asp 99 to glycine produced, however, less inhibition of the enzymatic activities than the mutant, E98G. To search for the catalytic residues, other residues were explored. x-ray crystallography showed that the nicotinamide-binding site is in a pocket formed in part The recombinant wild type and mutant ADP-ribosyl cyclases were incubated with various concentrations of NAD. The rates of cADPR produced were measured by HPLC. The kinetic parameters, K m and V max , were determined by double-reciprocal plots of the data. The distances shown are the shortest distances between the indicated amino acid residues to the bound nicotinamide and were determined from x-ray crystallography data.  4. Comparison of the effects of site-directed mutagenesis on the ADP-ribosyl cyclase and the base exchange activities. The V max values of the two activities of the mutants were compared with those of the wild type. A similar pattern of inhibiton on the two enzyme activities is exhibited by the five different mutants. The conversion of Trp 140 to glycine (W140G) produced the most inhibition of both enzyme activities, while the conversion of Asn 107 to glycine (N107G) produced the least. The V max value of the ADP-ribosyl cyclase activity of the wild type (WT) was measured to be 0.95 Ϯ 0.19 mmol/mg/min (see Table I), and the V max value of the base exchange activity of WT was about 0.35 Ϯ 0.07 mmol/mg/min (see Table III).   by the conserved sequence (cf. Fig. 1, A and B). Deep inside this binding pocket is another acidic residue, Glu 179 , which is conserved among various species of cyclases, CD38, and BST-1 (24). Indeed, its conversion to glycine resulted in complete elimination of the GDP-ribosyl cyclase activity and an over 10,000-fold decrease in both the ADP-ribosyl cyclase and the base exchange activities. These results are consistent with Glu 179 being a catalytic residue. It has previously been proposed that a single intermediate is responsible for the cyclization and the base exchange activities of the cyclase (15, 20 -23). According to this model, the cyclase binds NAD in a folded conformation so that the ends can be brought together for cyclization (15). The next step is likely to be the release of nicotinamide and the formation of an intermediate, which is proposed to be an oxocarbonium ion. This ionic intermediate has been proposed also for a variety of other NAD-utilizing enzymes, such as diphtheria and Clostridium botulinum C2 toxins, and orotate phosphoribosyltransferase (38 -40). The cationic oxocarbonium intermediate is unstable and must be stabilized by anionic residues nearby. A glutamate, which is also the catalytic residue, was found to perform such a function in both diphtheria and C. botulinum C2 toxins. This is consistent with our results indicating that the catalytic residue of the cyclase is also a glutamate, Glu 179 . In the case of orotate phosphoribosyltransferase, an aspartate serves a similar function.
Conversion of Trp 77 , another residue in the nicotinamidebinding pocket of the cyclase, to glycine resulted in a similar degree of reduction of enzymatic activities as mutagenizing Glu 179 . X-ray crystallography shows that both Trp 77 and Trp 140 line the rim of the binding pocket, one on each side. Trp 140 is in close proximity to the bound nicotinamide, suggesting that it is important in positioning the substrate. Trp 77 may serve a similar function in positioning the substrate NAD through hydrophobic interaction. Another important function that the two tryptophan residues may serve is to curtail the access of water to the catalytic site through the hydrophobicity of their side chains. Inaccessibility of water to the active site would ensure that the nucleophilic attack of the oxocarbonium intermediate by the 1-nitrogen of the adenine is the dominant reaction, leading to the cyclization of NAD to produce cADPR. Access of water to the active site, on the other hand, would lead to the formation of ADP-ribose through hydrolysis, which is the dominant reaction catalyzed by CD38 (12,41).
Another notable structural feature of the active site shown in this study is its close proximity to the central cavity formed by the two monomer cyclase molecules (Fig. 1A). The size of the central cavity is similar to the dimensions of a molecule of cADPR. This suggests the possibility that the central cavity may serve as a channel, through which the product of catalysis, cADPR, can pass (4). A recent report has now provided strong experimental support that a homolog of the cyclase, CD38, may indeed function as such a novel transporter of cADPR (42). It is shown that when resealed right-side-out red cell ghosts that contain ecto-CD38 are incubated with NAD (or NGD), active transport of cADPR (or cGDPR) occurs, generating a concentration gradient of the product 10 -80-fold higher inside than outside (42). Inhibition of the enzymatic activity of CD38 by mercaptoethanol blocks the process. Similar active transport of the enzymatic products is seen with purified CD38 reconstituted into liposomes (42). That CD38 is a transporter capable of active translocation of its enzymatic products across membranes provides a novel explanation of how a surface antigen with its catalytic domain on the outside of the cells can participate in intracellular Ca 2ϩ signaling (43). The structural features of the active site of the cyclase shown in this study are consistent with such a proposal.