CD38 and ADP-ribosyl Cyclase Catalyze the Synthesis of a Dimeric ADP-ribose That Potentiates the Calcium-mobilizing Activity of Cyclic ADP-ribose*

CD38, a lymphocyte differentiation antigen, is also a bifunctional enzyme catalyzing the synthesis of cyclic ADP-ribose (cADPR) from NAD+ and its hydrolysis to ADP-ribose (ADPR). An additional enzymatic activity of CD38 shared by monofunctional ADP-ribosyl cyclase from Aplysia californicais the exchange of the base group of NAD+ (nicotinamide) with various nucleophiles. Both human CD38 (either recombinant or purified from erythrocyte membranes) and Aplysia cyclase were found to catalyze the exchange of ADPR with the nicotinamide group of NAD+ leading to the formation of a dimeric ADPR ((ADPR)2). The dimeric structure of the enzymatic product, which was generated by recombinant CD38 and by CD38+Namalwa cells from as low as 10 μm NAD+, was demonstrated using specific enzyme treatments (dinucleotide pyrophosphatase and 5′-nucleotidase) and mass spectrometry analyses of the resulting products. The linkage between the two ADPR units of (ADPR)2 was identified as that between the N1of the adenine nucleus of one ADPR unit and the anomeric carbon of the terminal ribose of the second ADPR molecule by enzymatic analyses and by comparison with patterns of cADPR cleavage with Me2SO:tert-butoxide. Although (ADPR)2 itself did not release Ca2+ from sea urchin egg microsomal vesicles, it specifically potentiated the Ca2+-releasing activity of subthreshold concentrations of cADPR. Therefore, (ADPR)2 is a new product of CD38 that amplifies the Ca2+-mobilizing activity of cADPR.

NAD ϩ can be metabolized by either symmetrical hydrolysis of the pyrophosphate bond by a number of pyrophosphatases (1,2) or by cleavage of the ␤-glycosyl linkage between nicotinamide and ribose (3). The latter linkage has been shown to be cleaved by an increasing number of enzymes belonging to diverse classes and having widely varying cellular distribution. For example, the classical NAD ϩ glycohydrolases, which catalyze the hydrolysis of NAD ϩ to nicotinamide and adenosine diphosphate ribose (ADPR), 1 are mostly ectoenzymes in mammalian cells (4 -7). Endogenous ADP-ribosyl transferases responsible for either mono-ADP-ribosylation (8 -12) or poly-(ADP-ribosylation) (13) reactions are localized at the plasma membrane or in the nucleus, respectively. Finally, ADP-ribosyl cyclase, which converts NAD ϩ to nicotinamide and cyclic ADPribose (cADPR) (14 -16), was first demonstrated as a membrane-bound enzyme in sea urchin egg extracts (14). Various bifunctional enzymes in many mammalian tissues can also catalyze both the synthesis and the hydrolysis of cADPR. The first such enzyme reported was purified from dog spleen (17). The ectocellular domain of CD38, a lymphocyte antigen (18,19), can also catalyze a two-step reaction involving transient formation of cADPR (cyclase) followed by its hydrolysis to ADPR (hydrolase), eventually generating nicotinamide and ADPR (20 -24). The overall reaction catalyzed by CD38 is thus identical to that catalyzed by the classical NAD ϩ glycohydrolases. The confusion between CD38-like enzymes and the NAD ϩ glycohydrolases has been recently resolved by the use of NGD ϩ , an NAD ϩ analog, as an alternative substrate. Only CD38-like bifunctional enzymes possessing a cyclase activity can convert NGD ϩ to cyclic GDP-ribose, a fluorescent product. NAD ϩ glycohydrolases, on the other hand, can only convert NGD ϩ to GDP-ribose, a non-fluorescent product (25).
In addition to CD38, another family of NAD ϩ -utilizing enzymes has been recently identified that includes BST-1 (26) and BP-3 (27). These enzymes are glycosyl-phosphatidylinositol-anchored ectoproteins having considerable sequence homology to CD38 and may be involved in the ectocellular metabolism of NAD ϩ , cADPR, and ADPR as well.
Various models have been proposed for the catalytic mechanism of cleavage of the nicotinamide-ribosyl linkage by NAD ϩutilizing enzymes. A common feature among these models is the postulated presence of an enzyme-ADP-ribose intermediate (28 -31). Depending on whether water is accessible to the active site, the intermediate can dissociate to yield either cADPR or ADPR, respectively, thus accounting for either a cyclase or an NAD ϩ glycohydrolase reaction (31). This catalytic scheme can also account for the base-exchange reaction actually observed. Thus, other nucleophiles including nicotinic acid can interact with the intermediate and generate products with the nicotinamide base exchanged for the nucleophiles. A physiologically relevant example of this type of base-exchange reaction is the synthesis of the new calcium-mobilizing molecule NAADP ϩ (32) from NADP ϩ and nicotinic acid, which is catalyzed by CD38 and ADP-ribosyl cyclase (31)(32)(33)(34).
In this paper, we report another example of the exchange reaction that can lead to the formation of a hitherto unknown biologically active product. Using CD38 purified from human erythrocyte membranes, we show that ADP-ribose itself can function as a nucleophile and can interact with the enzyme intermediate. The result is the formation of a dimeric ADPR molecule. Of particular relevance is the fact that this ADPR dimer possesses functional activity, as it potentiates in a specific way the calcium-releasing activity of cADPR from sea urchin egg microsomal vesicles.

EXPERIMENTAL PROCEDURES
Materials-CD38 was purified to homogeneity from human erythrocyte membranes as described (21). Its cyclase specific activity, which was measured with NGD ϩ as a substrate (35), was 2.8 mol cyclic GDP-ribose/min/mg. Its cADPR hydrolase-specific activity (35) was 2.8 mol ADPR/min/mg. Production and purification of recombinant human CD38 was carried out as described previously (31,36). Its cyclase specific activity on NGD ϩ was 18 mol cyclic GDP-ribose/min/mg. ADPribosyl cyclase purified from Aplysia californica ovotestes as previously reported (31) had a specific activity that was measured using NAD ϩ as a substrate (35) of 20 mol cADPR/min/mg. [ 3 H]NAD ϩ (30.7 Ci/mmol) was obtained from DuPont NEN (Florence, Italy). All other chemicals were from Sigma and were of the highest purity grade available.
Analytical HPLC-The conditions for HPLC analyses using a C18 reverse phase column on a Hewlett-Packard HP 1090 instrument were as described previously (35). The nucleotide peaks were identified by comparing their retention times and UV spectra with highly pure standard compounds.
Synthesis and Purification of Dimeric ADPR-Purified CD38 (0.75 g) was incubated for 15 h at 37°C with 1 mM NAD ϩ and 20 mM ADPR (HPLC-purified) in 5 mM Tris-HCl, pH 6.5, containing bovine serum albumin (0.1 mg/ml) and 0.05% Triton X-100. The final volume was 5.0 ml. The reaction was terminated by addition of trichloroacetic acid (5% final concentration), the mixture was centrifuged, and the excess trichloroacetic acid was removed with three successive extractions with diethylether. The excess diethylether was evaporated under an N 2 stream, and the solution was vacuum-dried. The resulting powder was dissolved in 0.5 ml of H 2 O, and 2 l of the solution was submitted to analytical HPLC (35). The yield of the dimeric ADPR ((ADPR) 2 ) eluted at 34 min was around 1% of the total area of all peaks. Therefore, to obtain sufficient amounts of (ADPR) 2 for analyses the incubation conditions described above were repeated nine times. Each of the ten 0.5-ml samples thus obtained was then submitted separately to preparative HPLC. This was carried out on a System Gold HPLC (Beckman Instruments) equipped with a diode-array spectrophotometric detector set at 260 nm. An anion exchange perfusion chromatography column (100 ϫ 4.6 mm, Poros QE/M, PerSeptive Biosystems, Cambridge, MA) was used. Solvent A was H 2 O, and solvent B was 0.15 M trifluoroacetic acid. The solvent program was a linear gradient starting at 100% solvent A and increasing to 100% solvent B in 15 min. The flow rate was 6 ml/min. (ADPR) 2 eluted at 7 min, well separated from all other nucleotides (NAD ϩ , ADPR, and cADPR) and nicotinamide, which eluted in the same peak between 2 and 5 min. The 7-min peak was lyophilized and stored at Ϫ20°C until used.
Enzymatic Digestion of (ADPR) 2 and HPLC Purification of the Reaction Products-Purified (ADPR) 2 (0.25 mg) was dissolved in H 2 O and incubated for 18 h at 37°C with 30 g of commercial dinucleotide pyrophosphatase (EC 3.6.1.9) from Crotalus adamanteus venom in 0.2 ml of 20 mM Tris-HCl, pH 8.3, containing 2 mM MgCl 2 . The reaction was terminated with trichloroacetic acid (5% final concentration), and the trichloroacetic acid was removed as described above. Part of the deproteinized mixture (2%) was analyzed by HPLC (35) to monitor the reaction products. Two products were formed, AMP (retention time, 15 min) and an unknown product that eluted at 25 min (hence designated P25). The remaining 98% of the deproteinized mixture was submitted to anion exchange perfusion chromatography as described above except that a flow rate of 4 ml/min was used instead. The unknown compound (P25) that eluted at 7 min was well separated from AMP, which had a retention time of 3 min in this system. Purified P25 was lyophilized and stored at Ϫ20°C.
Lyophilized P25 (0.1 mg) was dissolved in H 2 O and incubated for 15 h at 37°C with 15 g of 5Ј-nucleotidase (EC 3.1.3.5) from C. adamanteus venom in 0.25 ml of phosphate-buffered saline. The reaction was terminated with trichloroacetic acid (5% final concentration), the trichloroacetic acid was removed as described above, and the resulting solution was analyzed by HPLC (35), which showed only one peak with a retention time of 12 min (P12). This peak was purified by HPLC using C18 reverse phase chromatography (ODS Hypersil 3 m, 60 ϫ 4.6 mm) at a flow rate of 0.5 ml/min. Solvent A was deionized water, and solvent B was methanol (HPLC grade). The solvent program was a gradient starting at 100% Solvent A for 5 min, then linearly increasing to 100% Solvent B in 15 min. Under these HPLC conditions, the P12 peak showed a retention time of 12 min. This peak was collected, lyophilized, and stored at Ϫ20°C before HPLC/MS analyses.
MS Analyses-Various purified samples were resuspended at 5-50 g/ml in H 2 O:methanol:trifluoroacetic acid (49.5:49.5:1). The sample was pumped by an HPLC pump (Kontron Instruments, Milan, Italy) through a Valco valve into the mass spectrometer (Hewlett Packard 5989A Engine, Palo Alto, CA) electrospray source. Background spectra were obtained by injecting the solvent without the samples and were subtracted from the averaged spectra of various samples. Data were collected in the positive ion mode in a range including the expected molecular masses to avoid the influence of the sodium and potassium adducts usually found with phosphate-containing molecules.
Conversion of cADPR to N 1 -(5Ј-phosphoribosyl)-AMP-The procedure was as described by Gu and Sih (37) with minor modifications. Briefly, sodium tert-butoxide was prepared by dissolving 0.3 g of sodium in 13 ml of tert-butyl alcohol overnight at room temperature. The solution was vacuum-dried, and the resulting white powder was dissolved in an Me 2 SO:water (998:2) solution to give a final concentration of 0.4 mg/ml. cADPR (30 g) was then incubated for 60 min at 35°C in 1.0 ml of the Me 2 SO:tert-butoxide mixture. Under these conditions, the main product observed by analytical HPLC (35) had a retention time of 25 min, identical to that obtained upon incubation of (ADPR) 2 with dinucleotide pyrophosphatase (P25, see above). This product was purified by anion exchange perfusion chromatography (see above), lyophilized, and stored at Ϫ20°C before being submitted to MS analysis. Prolonging the incubation of cADPR in Me 2 SO:tert-butoxide to 2 h resulted in greatly reduced formation of the peak that eluted at 25 min and in the concomitant appearance of another compound that had a retention time of 12 min. This is identical to the retention time of the product obtained by sequential digestion of (ADPR) 2 with dinucleotide pyrophosphatase and 5Ј-nucleotidase (P12, see above).
Calcium Release Assay-Fractionation of sea urchin egg (Strongylocentrotus purpuratus) microsomes was performed as described in Ref. 38 except that calmodulin (10 g/ml) was added to the dilution medium. Ca 2ϩ release assays were performed at 17°C under constant stirring. Fig. 1 shows that native CD38 purified from human erythrocyte membranes catalyzes the base-exchange reaction resulting in the synthesis of NAAD ϩ from NAD ϩ and nicotinic acid. The reaction occurs even at pH 6.5, which is significantly higher than that previously demonstrated to be optimal for the base-exchange reaction catalyzed by a recombinant CD38 containing only the extracellular domain (31). The extent of NAAD ϩ formation was comparable with that catalyzed by Aplysia cyclase also shown in Fig. 1. In addition to NAAD ϩ , erythrocyte CD38 also results in the synthesis of a modest amount of cADPR and of a substantial amount of ADPR. In contrast, the Aplysia cyclase mainly catalyzes the formation of cADPR with minor production of NAAD ϩ and ADPR (Fig. 1).

ADP-ribosyl Cyclase and Native CD38 Catalyze a Base-exchange Reaction-
ADP-ribosyl Cyclase and CD38 Catalyze the Formation of an Unknown Product from NAD ϩ and ADPR-Whether ADPR can serve as a nucleophile in the base-exchange reaction was investigated by incubating the native CD38 with [adenine-3 H]NAD ϩ in the presence of excess ADPR. Fig. 2 shows that the majority of the 3 H label was converted to ADPR (eluted at 23 min) with a small portion of it going to cADPR (eluted at 5.5 min). In addition, a detectable amount of the label (around 1%) was present in an unknown peak eluted at 34 min (P34). This peak was also labeled when [␣-32 P]NAD ϩ was used as traced instead of [ 3 H]NAD ϩ (not shown). The UV spectrum of the unknown peak shown in the inset of Fig. 2 is clearly different from that of ADPR, with a maximum at 270 nm rather than at 260 nm. Fig. 3 shows that the formation of P34 from [ 3 H]NAD ϩ and ADPR is catalyzed not only by CD38 but also by ADP-ribosyl cyclase from A. californica. The time course of P34 production was similar with the two enzymes. The amount of P34 formed by CD38 at 18 h was only slightly lower than that of cADPR, indicating that it is a major product of CD38 (Fig. 3A). The long incubation time was chosen in these experiments because CD38 is known to be substantially inactivated due to the NAD ϩ -induced self-aggregation of the protein (35,39,40). This inactivating effect as well as the attendant oligomerization was also observed with a recombinant form of human CD38 corresponding to its ectocellular C-terminal region (35,36). The NAD ϩ -dependent inactivation of CD38 is different from that recently observed with NAD ϩ glycohydrolase purified from rabbit erythrocytes (41), which is due to auto-ADP-ribosylation and is reversible.
Since the two-step activity of CD38 leads to accumulation of ADPR as the major end product (Fig. 3A), we investigated whether P34 production also occurs without any initial addition of ADPR as co-substrate. As shown in Fig. 4, this proved to be the case, although formation of P34 became detectable only at 8 h of incubation, when production of ADPR peaked at approximately 5 mM and was comparably lower in extent than that recorded with both NAD ϩ and ADPR added. In these experiments, the NAD ϩ -induced inactivation of CD38 was counteracted by repeated additions of the soluble recombinant protein. At the same times, the incubation mixtures were also regularly supplemented with NAD ϩ to provide enough substrate for the reactions catalyzed by CD38 to take place. Indeed, under these conditions the disappearance of NAD ϩ over the extended incubation times went to completion differently from the experiment shown in Fig. 3A in which NAD ϩ accumulated because of enzyme inactivation (not shown).
Generation of P34 and cADPR was also observed at much lower initial concentrations of NAD ϩ . In these experiments also, no ADPR was added, and the starting concentration of NAD ϩ was 10 M (Fig. 5). The amount of CD38 was remarkably lower than in the experimental conditions of Fig. 4 to avoid fast exhaustion of NAD ϩ . Again, both 10 M NAD ϩ and fresh recombinant CD38 were added to the incubation mixtures at sequential time intervals to allow the enzymatic reactions to proceed continuously. As shown in Fig. 5, even at these physiological concentrations of NAD ϩ , the formation of P34 progressed slowly and was only slightly lower than that of cADPR.
In an attempt to address the physiological occurrence of P34 formation from NAD ϩ , experiments were carried out on intact Namalwa cells (a continuous B-cell-derived line from Burkitt's lymphoma), which have both enzymatic activities of CD38 at their outer surface (42) rather than on purified CD38. Using these cells at the same activity levels as purified recombinant CD38 on 10 M NAD ϩ (Fig. 5), disappearance of NAD ϩ and concomitant progressive formation of ADPR, cADPR, and P34 were apparently observed at short time intervals as low as 5-30 min (not shown). In these experiments, however, unequivocal identification of the nucleotide peak eluted at 34 min with authentic P34 was not possible on the basis of UV spectrum because amounts of this peak were too limited. Moreover, the picture of NAD ϩ metabolism in Namalwa cells was remarkably complicated by additional products of ADPR catabolism including AMP, adenosine, and hypoxanthine, which were generated partly at the cell surface and partly intracellularly.
Attempts were then made to establish optimal conditions of P34 formation. In these experiments we used the general setting shown in Fig. 3, i.e. a single addition of CD38 (either purified from erythrocyte membranes or recombinant) and one of NAD ϩ . However, even in these simplified conditions, variations of substrate concentrations (both NAD ϩ and ADPR) as well as of the pH of the reaction produced no further improvement. Likewise, P34 formation was not enhanced upon performing the incubations in more apolar media such as 6 M methanol to minimize hydrolytic reactions. Control experiments showed that the formation of P34 requires catalysis since it was undetectable in the absence of enzyme.
Structural Analysis of P34 -Various enzymes were used to convert P34 into known products (see "Experimental Procedures"). P34 proved not to be a substrate of commercial NAD ϩ glycohydrolase or Aplysia ADP-ribosyl cyclase. On the contrary, P34 was quantitatively converted to ADPR by purified CD38, and no other UV-absorbing or labeled compounds were produced in these conditions. This finding indicates P34 is likely a dimer or an oligomer of ADPR.
Mass spectrometry measurements showed that the molecular ion of P34, (MϩH) ϩ , has a mass of 1,101. This value corresponds to two individual ADPR molecules joined together with the loss of a water molecule (Fig. 6A). P34 can, therefore, be designated as (ADPR) 2 . Incubation of (ADPR) 2 with dinucleotide pyrophosphatase cleaved the pyrophosphate bond of ADPR, yielding two degradation products. One was identified as AMP by its retention time of 15 min. The other product was designated as P25 since it eluted at 25 min. The m/z value of P25 was 560, which is consistent with an adenine moiety linked to two ribose-phosphate units, i.e. A(R-5-P) 2 as shown in Fig.  6B. Further incubation of A(R-5-P) 2 with 5Ј-nucleotidase led to the formation of a single compound with a retention time of 12 min (indicated as P12 in Fig. 6C). P12 had an m/z value of 400 corresponding to an adenine bonded to two ribose moieties (AR 2 ).
Further insight into the structure of (ADPR) 2 was provided by parallel experiments in which the pyrophosphate linkage of authentic cADPR was cleaved under conditions not perturbing the integrity of its C-N 1 glycosyl bond. The pyrophosphate linkage of cADPR is markedly resistant to a number of pyrophosphatases, both from commercial sources and present in human plasma (not shown). However, it can be chemically broken by exposure to tert-butoxide in Me 2 SO. The resulting product is N 1 -(5Ј-phosphoribosyl)-AMP, a compound in which both N 1 and N 9 are bound to two anomeric carbons of ribose 5Ј-P (37). If the C-N 1 glycosyl bond were involved in joining the two ADPR moieties of (ADPR) 2 together, then the same molecule of N 1 -(5Ј-phosphoribosyl)-AMP (i.e. the compound referred to as A(R-5-P) 2 in Fig. 6B) should arise from the digestion of (ADPR) 2 by dinucleotide pyrophosphatase. Indeed, the degradation procedures employed for cADPR and (ADPR) 2 both yielded a compound with identical retention time (25 min using a reverse phase column as described under "Experimental Procedures"), the same UV spectrum, and also the same mass of 560 for its molecular ion (MϩH) ϩ .
Prolonging the incubation of cADPR in Me 2 SO:tert-butoxide for 2 h instead of 1 h further degraded the molecule into a product having a retention time of 12 min on a reverse phase column and a mass of 400 for its molecular ion, (MϩH) ϩ . These properties coincide with those of the product obtained upon digesting P34 sequentially with dinucleotide pyrophosphatase and 5Ј-nucleotidase (Fig. 6C). These results suggest the same type of N 1 -glycosyl linkage that occurs in cADPR is present also in (ADPR) 2 . Taken together, these data are therefore consistent for the molecular structure shown in Fig. 7, in which the same adenine moiety binds two anomeric carbons of the two ADP-ribosyl units at N 1 and N 9 , respectively.
Functional Activity of (ADPR) 2 - Fig. 8A shows that addition of (ADPR) 2 to reach a final concentration as high as 50 M produces no detectable Ca 2ϩ release from sea urchin egg microsomes. For comparison, as low as 20 nM cADPR elicits detectable Ca 2ϩ release in the same microsomal preparation (not shown). However, when increasing amounts of (ADPR) 2 were added to microsomes pretreated with 15 nM cADPR, which was too low to produce Ca 2ϩ release by itself, a concentration-dependent release of Ca 2ϩ was observed reaching its maximum at about 50 M. This synergistic effect was half-maximal at 25-30 M (ADPR) 2 (Fig. 8A). Ca 2ϩ stores were not emptied by the combination of 50 M (ADPR) 2 and 15 nM cADPR. Indeed, higher concentrations of cADPR alone (up to 200 nM) could induce a further Ca 2ϩ release from the microsomes. Fig. 8B shows the converse experiment when the egg microsomes were pretreated with a 50 M concentration of (ADPR) 2 , followed by various concentrations of cADPR. The pretreatment sensitized the microsomes such that only 18 nM of cADPR was sufficient to induce near-maximal Ca 2ϩ release from the microsomes. When this experiment was performed at low constant concentrations of (ADPR) 2 (10 and 25 M, respectively), a potentiating effect on subthreshold concentrations of cADPR was still observed, although the effect was less evident than with 50 M (ADPR) 2 . These results indicate that (ADPR) 2 can function as a sensitizer of cADPR.
The amplifying effect of (ADPR) 2 on cADPR activity proved to be specific. Thus, ADPR and its potential metabolites AMP (43) and adenosine, each at concentrations up to 100 M, failed to release Ca 2ϩ from sea urchin egg microsomes and to display any synergistic effect on cADPR (not shown).

DISCUSSION
The involvement of an enzyme-ADP-ribose intermediate in the catalytic mechanism of monofunctional ADP-ribosyl cyclase, bifunctional cyclases/hydrolases, or NAD ϩ glycohydrolases has been suggested by several lines of evidence including: (i) methanolysis of NAD ϩ (29,30,44), (ii) base-exchange reaction between nicotinamide and nicotinic acid (31), (iii) NAD ϩ synthesis from cADPR and nicotinamide (17), and (iv) the conformation of the N 1 -glycosyl linkage of cADPR being in the ␤ configuration in both its crystal form (45) and in aqueous solution (37,46). The data reported in this paper are also consonant with this mechanism because the dimeric structure of (ADPR) 2 implies the transfer of an enzyme-activated ADPribosyl moiety to a pre-existing ADPR molecule behaving as an acceptor substrate. Fig. 9 illustrates a working model accounting for the versatility of CD38 in terms of catalyzing the formation of multiple products. As already reported (17,31,47), the ADP-ribosylenzyme intermediate can react either with water, resulting in generation of ADPR, or with the N 1 of the adenine moiety, leading to production and release of cADPR. The third possible pathway is closely reminiscent of the previously reported mechanism of base-exchange leading to the synthesis of NAAD ϩ or of NAADP ϩ in the presence of nicotinic acid (31,34). In the scheme shown in Fig. 9, the exchange occurs between nicotinamide and the preformed ADPR, thereby producing the dimeric ADPR molecule identified in this study. (ADPR) 2 itself is a good substrate of the hydrolase activity of CD38. This fact may be responsible for the low amounts of (ADPR) 2 detected in our experiments. Also, it is possible that exceedingly high concentrations of ADPR in the proximity of the active site are needed to compete with the intramolecular cyclization that results in cADPR synthesis. This can account for the similarly low production of (ADPR) 2 by the Aplysia cyclase despite its weak hydrolase activity as compared with CD38. Although less extensive, formation of (ADPR) 2 was also observed upon incubating CD38 with NAD ϩ in the absence of added ADPR (Figs. 4 and 5). Nevertheless, since (ADPR) 2 became detectable only when ADPR reached its maximal concentrations, this finding provides further confirmation of the exchange mechanism shown in Fig. 9. Moreover, it demonstrates that the reaction of CD38 on NAD ϩ can ultimately generate cADPR, ADPR, and (ADPR) 2 and that the two last compounds can be reciprocally interconvertible.
The formation of (ADPR) 2 requires enzymatic catalysis since it is not detectable under our experimental conditions without enzyme. Nevertheless, a compound sharing the same mass as the enzymatically prepared (ADPR) 2 was observed as a 0.1% impurity of commercial ADPR preparations. This impurity also co-eluted with 3 H-labeled (ADPR) 2 that was synthesized by CD38. In all experiments reported in this study, the impurity of the commercial ADPR was removed by repurification using HPLC. A similar situation has been described for NAADP ϩ , which has also been shown to be present as an impurity in commercial NADP ϩ preparations. Also similar to (ADPR) 2 is that NAADP ϩ can be synthesized either chemically by alkaline treatment of NADP ϩ (32) or, as recently reported, by a baseexchange reaction catalyzed by CD38 using NADP ϩ and nicotinic acid as substrates (31).
In this study, specific patterns of enzymatic degradation combined with MS analyses of the products were used to unequivocally establish the dimeric structure of (ADPR) 2 . Evidence for the N 1 -ribosyl bond as the linkage responsible for joining the two ADPR units together is taken from the following two findings: (i) the susceptibility of (ADPR) 2 to be hydrolyzed by CD38; (ii) the analyses of the degradation products of the digestion of (ADPR) 2 with dinucleotide pyrophosphatase either alone or in combination with 5Ј-nucleotidase that show that they are the same as those produced following chemical cleavage of the pyrophosphate bond of cADPR. Since the cyclizing linkage in cADPR has been unambiguously shown to be the N 1 -ribosyl bond by x-ray crystallography (45), the identity of the degradation products of (ADPR) 2 and cADPR provides convincing evidence that the linkage in (ADPR) 2 is also the N 1ribosyl bond.
(ADPR) 2 itself is devoid of Ca 2ϩ -releasing activity on sea urchin egg microsomal vesicles. However, it synergizes with cADPR in producing an enhancement of the Ca 2ϩ -releasing activity of subthreshold concentrations of cADPR (Fig. 8B). This effect is specific for (ADPR) 2 as it was not observed with ADPR or with known metabolites of ADPR including AMP (43) and adenosine. The Ca 2ϩ -releasing activity of cADPR has been previously shown to be potentiated by calmodulin and divalent cations (38,48) and also by sensitizers of the Ca 2ϩ -induced Ca 2ϩ release such as caffeine (49). The potentiation effect of (ADPR) 2 represents yet another way of modulating the Ca 2ϩ release mechanism activated by cADPR.
The amount of (ADPR) 2 produced by incubation of CD38 with NAD ϩ , either with or without added ADPR, is low. It can be, nevertheless, physiologically significant since it is comparable with the amounts of cADPR produced under the same conditions (Figs. [3][4][5]. Moreover, generation of (ADPR) 2 was also detectable starting from the reaction of recombinant CD38 on NAD ϩ concentrations as low as 10 M (Fig. 5). Use of intact Namalwa B cells as a source of ectocellular CD38 (42) instead of purified CD38 also supported the notion that 10 M NAD ϩ acts as a precursor of (ADPR) 2 as well as of cADPR and ADPR.
Previously unrecognized formation of (ADPR) 2 might account for regulation of Ca 2ϩ -releasing activity of cADPR in biological systems. Recently, ADPR has been reported to potentiate cADPR synthesis from NAD ϩ in sea urchin egg homogenates (50), and this effect was ascribed to an inhibitory mechanism afforded by ADPR on the cADPR hydrolase activity resulting in elevated cADPR levels. Since cADPR concentrations were evaluated using a bioassay based on Ca 2ϩ -releasing activity from sea urchin egg homogenates, an alternative explanation could be the formation of (ADPR) 2 thus potentiating the Ca 2ϩ release induced by cADPR. Indeed both mechanisms,  i.e. inhibition of cADPR hydrolase and formation of (ADPR) 2 , could result in the observed effect of potentiation of Ca 2ϩreleasing activity in sea urchin egg homogenates (50). Accordingly, (ADPR) 2 might be involved in up-modulating the biological effect of cADPR in this system.
The ectocellular localization of the catalytic domain of CD38 suggests that its enzymatic products may have extracellular functions. It has been previously shown that cADPR applied extracellularly to B lymphocytes can enhance the cell proliferation stimulated by co-agonists, while ADPR has the opposite effect (20). Recently, transient exposure of rat cerebellar granule cells to extracellular cADPR has been observed to enhance the Ca 2ϩ -induced Ca 2ϩ release activity of these cells in approximately 50% of the experiments (51). Conversely, this functional effect was an almost constant finding when the same CD38 ϩ cerebellar granule neurons were exposed to extracellular NAD ϩ as a precursor of cADPR. Whether this difference is due to (ADPR) 2 formation from NAD ϩ potentiating the effect of cADPR is an attractive possibility that is currently under study.
(ADPR) 2 adds to a growing number of metabolites generated by the enzymes involved in the metabolism of cADPR. Depending on the presence of NAD ϩ or NADP ϩ as a substrate, cADPR, NAADP ϩ , and 2Ј-P-cADPR can be produced under different experimental conditions. All of these have been shown to have Ca 2ϩ -releasing activity (14,32,33,52). Degradation pathways for these active metabolites have also been described. Thus, cADPR is known to be hydrolyzed by CD38-like enzymes to ADPR, while alkaline phosphatase has been shown to be effective in hydrolyzing 2Ј-P-cADPR to cADPR (31) and NAADP ϩ to the functionally inactive NAAD ϩ (53). Importantly, we observed that CD38 is able to not only synthesize (ADPR) 2 but also to hydrolyze it, suggesting that the concentration of this nucleotide in cells may be regulated. It is conceivable that a small augmentation of the synthesizing activity of CD38 coupled with a corresponding decrease in the hydrolyzing activity can result in a larger increase of the steady state concentration of (ADPR) 2 . Indeed, it has been shown that Zn 2ϩ can modulate the cADPR synthesizing activity of CD38 in precisely this manner (21,54).
This proliferation of interacting metabolites of this signaling pathway is reminiscent of the phosphatidylinositide pathway, which began with the discovery of inositol trisphosphate but soon grew with the multiplicity of interacting inositol phosphates (reviewed in Ref. 55). In this context, (ADPR) 2 can be viewed as a product of the signaling pathway designed to amplify the physiological effects of cADPR. In any case, discovery of (ADPR) 2 as an end product of CD38 enzymatic activities does raise many intriguing possibilities demanding further investigation.