Structural determination of a cyclic metabolite of NAD+ with intracellular Ca2+-mobilizing activity.

Incubation of NAD+ with extracts from sea urchin eggs resulted in production of a metabolite which could mobilize intracellular Ca2+ stores of the eggs. In this study we present structural evidence indicating that the metabolite is a cyclized ADP-ribose having an N-glycosyl linkage between the anomeric carbon of the terminal ribose unit and the N6-amino group of the adenine moiety. In view of this structure we propose cyclic ADP-ribose as the common name for the metabolite. The purification procedure for the metabolite consisted of deproteinizing the incubated egg extracts and sequentially chromatographing the extracts through three different high pressure liquid chromatography (HPLC) columns. The homogeneity of the purified metabolite was further verified by HPLC on a Partisil 5 SAX column. Using radioactive precursor NAD+ with label at various positions it was demonstrated that the metabolite was indeed derived from NAD+ and that the adenine ring as well as the adenylate alpha-phosphate were retained in the metabolite whereas the nicotinamide group was removed. This was confirmed by 1H NMR and two-dimensional COSY experiments, which also allowed the identification of all 12 protons on the two ribosyl units as well as the two protons on the adenine ring. From the chemical shifts of the two anomeric protons it was concluded that the C-1 carbons of both ribosyl units were still bonded to nitrogen. The positive and negative ion fast atom bombardment mass spectra showed (M + Na)+, (M - H + 2Na)+, (M - H)-, and (M - 2H + Na)- peaks at m/z 564, 586, 540, and 562, respectively. Exact mass measurements indicated a molecular weight of 540.0526 for (M - H)-. This together with the constraints imposed by the results from NMR, radioactive labeling, and total phosphate determination uniquely specified a molecular composition of C15H21N5O13P2. Analysis by 1H NMR and mass spectroscopy of the only major breakdown product of the metabolite after prolonged incubation at room temperature established that it was ADP-ribose, thus providing strong support for the cyclic structure.

Incubation of NAD+ with extracts from sea urchin eggs resulted in production of a metabolite which could mobilize intracellular Ca2+ stores of the eggs. In this study we present structural evidence indicating that the metabolite is a cyclized ADP-ribose having an Nglycosyl linkage between the anomeric carbon of the terminal ribose unit and the AT'-amino group of the adenine moiety. In view of this structure we propose cyclic ADP-ribose as the common name for the metabolite. The purification procedure for the metabolite consisted of deproteinizing the incubated egg extracts and sequentially chromatographing the extracts through three different high pressure liquid chromatography (HPLC) columns. The homogeneity of the purified metabolite was further verified by HPLC on a Partisil5 SAX column. Using radioactive precursor NAD+ with label at various positions it was demonstrated that the metabolite was indeed derived from NAD+ and that the adenine ring as well as the adenylate a-phosphate were retained in the metabolite whereas the nicotinamide group was removed. This was confirmed by 'H NMR and two-dimensional COSY experiments, which also allowed the identification of all 1 2 protons on the two ribosyl units as well as the two protons on the adenine ring. From the chemical shifts of the two anomeric protons it was concluded that the C-1 carbons of both ribosyl units were still bonded to nitrogen. The positive and negative ion fast atom bombardment mass spectra showed (M + Na)+, (M -H + 2Na)+, (M -H)-, and (M -2H + Na)-peaks at mlz 564, 586, 540, and 562, respectively. Exact mass measurements indicated a molecular weight of 540.0526 for (M -H)-. This together with the constraints imposed by the results from NMR, radioactive labeling, and total phosphate determination uniquely specified a molecular composition of ClaHZ1NaOlsP2. Analysis by 'H NMR and mass spectroscopy of the only major breakdown product of the metabolite after prolonged incubation at room temperature established * This work was supported by National Institutes of Health Grant HD17484 and National Science Foundation Grant DCB8602499 (to H. C. L.), National Institutes of Health Grant HD18247 and the Minnesota Medical Foundation (to T. F. W.), and by the Midwest Center for Mass Spectrometry, a National Science Foundation instrumentation facility (Grant CHE-8620177). 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 U.S.C. Section 1734 solely to indicate this fact.
STo whom correspondence and reprint requests should be addressed Dept. of Physiology, 6- that it was ADP-ribose, thus providing strong support for the cyclic structure.
Mobilization of intracellular Ca2+ is a common response to activation of a variety of surface receptors. In many cases the second messenger involved in this signal transduction process is IP3.' However, IP3 may not be the only messenger molecule responsible for mobilizing intracellular Ca2+. For example, arachidonic acid has also been found to be as potent as IP3 in releasing Ca2+ from the endoplasmic reticulum in pancreatic islets (1). Similarly, cyclic IPS was shown to be even more active than IP3 in evoking a light response when microinjected into the Limulus photoreceptor (2). Both cyclic IP3 and arachidonic acid, on the other hand, are metabolites related to the polyphosphoinositide pathway and therefore could be considered as part of the same metabolic system which generates IP3. In a previous report we showed that a completely independent system involving pyridine nucleotide metabolites may also be involved in mobilization of intracellular Ca2+ in sea urchin eggs (3). We have developed a highly stable cellfree system using egg homogenates as a biological assay for Ca2+ release activators (4). With this screening assay we found a Ca2+-releasing metabolite of NAD+ which was produced by incubation with high speed supernatants of the egg. The metabolite which we originally called "enzyme-activated NAD" (E-NAD) was purified by HPLC. Microinjection of E-NAD into intact eggs elicited transient Ca2+ increases as well as cortical reactions showing that it was active both in vitro and in uivo. Structural evidence is presented in this article which indicates that the metabolite is a cyclized ADP-ribose having an N-glycosyl linkage between the anomeric carbon of the terminal ribose unit and the amino group at the 6-position of the adenine ring. In view of this novel structure we propose cyclic ADP-ribose as a more descriptive common name for the metabolite instead of E-NAD. For the sake of consistency we will retain the use of the name E-NAD in this article.
Calcium release from microsomes in the homogenates (0.8-1 ml) was measured using the Caz+ indicator, fura-2, as described previously (3). The 10% homogenate was first thawed and incubated at 17 "C for about 1 h before it was diluted 2-4 times (2.5-5% homogenate) with intracellular medium containing 1 p M fura-2, all of the protease inhibitors, and the ATP-regenerating system described above. The diluted homogenate was further incubated at 17 "C for an additional 1-2 h before use. The incubation periods allowed the microsomes to sequester the excess Ca2+ in the medium.
Preparation of Egg Extracts-Eggs of Strongylocentrotus purpuratus were dejellied and washed as described above, except that the last wash was with a medium containing 0.72 M glucose, 1 mM MgC12, and 20 mM HEPES, pH 7.2. Eggs were finally resuspended in 3 volumes of the same medium, and aprotinin (10 pg/ml), leupeptin (10 pg/ml), soybean trypsin inhibitor (50 pg/ml), benzamidine (2.5 mM), EDTA (0.1 mM), andEGTA (0.1 mM) were added. The egg suspension (25%) was then homogenized with a Dounce homogenizer and centrifuged for 30 min at 18,000 rpm (0-4 "C) in a J-21B centrifuge with a JA-20 rotor (Beckman Instruments). The supernatant was collected and was either used fresh or stored frozen at -70 "C. Egg extracts from both L. pictus and S. purpuratus were active in producing the metabolite. However, much larger amounts of eggs could be obtained from the latter species and were used in this study for the purpose of making egg extracts.
Purification of the NAD+ Metabolite-Egg extracts (25%) were incubated with 2 mM NAD+ for 8-12 h at 17 "C. The increase in Ca2+ release activity of the mixture was monitored periodically by adding small aliquots (1-5 pl) of the mixture to egg homogenates (2.5-5%, 0.8 ml), and the resultant Ca2+ release was measured fluorimetrically using the indicator fura-2 as described previously (3). The reaction was terminated at the end of the incubation period by the addition of an equal volume of acetone. The precipitated protein was removed by centrifugation, and the acetone was evaporated from the supernatant by a stream of Nz gas. The pH of the supernatant was then adjusted to 7.6 by the addition of NaOH or Tris base. The first step of the purification procedure employed an HPLC anion exchange column. The resin, AG MP-1 (Bio-Rad), was slurry-packed into a stainless steel column (1 X 10 cm) under pressure using an HPLC pump (model llOB, Beckman) operating at a flow rate of 8 ml/min. A 4.5-ml aliquot of the deproteinized supernatant was injected into the column either manually or by an autosampler (Gilson Co. Inc.). The solvent program used was a nonlinear trifluoroacetic acid gradient starting at 2% of 150 mM trifluoroacetic acid in water for 6 min, linearly increased to 4% in 5 min, linearly increased to 8% in another 5 min, then linearly increased to 11.6% in 2 min. The flow rate was 4 ml/min. The UV (254 nm)-absorbing peaks of the eluent were collected by a fraction collector operating at a peak-detecting mode and tested for Ca2+ release activity using the egg homogenate assay. The E-NAD eluted in a peak between 13 and 16 min. The column was cleaned afterward for 5 min with 150 mM trifluoroacetic acid and re-equilibrated for 6 min with 2% of 150 mM trifluoroacetic acid before the next injection. The eluent containing E-NAD was reduced in volume in a rotary evaporator and finally dried either in a Savant Speedvac or a Vortex evaporator (Haake Buchler Instruments, Inc.). The next step of purification was by reverse phase HPLC on a 0.46 X 25-cm column of either Ultrasphere ODS (Beckman) or Hypersil 5 C18 (Phenomenex). The separation was performed isocratically with 6.6 mM formic acid at a flow rate of 1 ml/min, and E-NAD eluted in a peak at about 6-7 min. Afterward, the column was cleaned with 100% methanol for 12 min and re-equilibrated with 6.6 mM formic acid for another 8 min before the next injection. The last step of purification was performed on a 0.46 X 25-cm mixed mode column (RPB/anion, 7 pm, Alltech Associates, Inc.); solvent A, water; solvent B, 10% acetonitrile, 90% 0.2 M formic acid with pH adjusted to 4.0 with ammonium hydroxide; solvent program, linear gradient from 0 to 100% B in 15 min, 100% B for another 9 min, re-equilibrated with water for 6 min before the next injection. E-NAD eluted in a sharp peak at about 15 min. The purified E-NAD was then desalted by passage through a 0.46 X 15-cm AG MP-1 column, eluted with a trifluoroacetic acid gradient similar to the one described above, vacuum-dried to remove trifluoroacetic acid, and stored at -70 "C. Usually, a preparation would consist of processing 200-300 ml of egg extracts to produce approximately 1 pmol of purified E-NAD.
Radioactive Labeling of E-NAD-E-NAD was double labeled by of [3H]NAD+ (5 pCi) together with 2.5 pCi of either ["C]NAD+ or incubating egg extracts (5 ml) with 1.9 mM NAD+ and a tracer amount [32P]NAD+. The positions of the label on the radioactive precursors used were adenir~-2,8-~H, adenine-U-14C, carbonyl-"C, aden~late-~'P, and ni~otinarnide-4-~H. The last two radioactively labeled forms of NAD+ were first purified by passage through an AG MP-1 HPLC column, and the others were used without further purification. The incubation was carried out at 17 "C for 4-6 h, and the labeled E-NAD was purified afterward as described above.
Determinution of the Extinction Coefficient-The amount of radioactively labeled E-NAD was determined by dividing the counts in the purified E-NAD by the specific activity of the precursor NAD+, which was determined from the radioactivity and the amount of NAD' in an aliquot of the reaction mixture. The extinction coefficient was then calculated from the absorbance at 254 nm and the concentration of the labeled E-NAD.
Analytical HPLC Procedures-The purity of the E-NAD was further verified by chromatographing it on a Whatman Partisil 5 SAX column (0.46 X 10 cm) and a Vydac 303NT405 column (0.46 X 5 cm). The SAX column was used with a saturator column (0.46 X 25 cm) packed with 30-40 pm of silica, and the solvent program used was a linear gradient of 0-0.6 M formic acid (pH 4.0 with NH4OH) in 30 min at a flow rate of 1 ml/min. The column was re-equilibrated with water for another 15 min before the next injection. The Vydac column consisted of a low capacity ion exchange material on a high efficiency protected silica substrate. It is specifically designed for nucleotides and can completely resolve the 12 major ribonucleotides in 10 min. Solvent A was 0.045 M NH4COOH, pH 4.6, with H3P04; solvent B was 0.5 M NaH2POr, pH 2.7, with HCOOH. The solvent program was 0-100% B in 10 min at a flow rate of 2 ml/min. Other analytical separations employed a 0.46 X 15-cm AG MP-1 column at a flow rate of 1 ml/min. Solvent A was water, and solvent B was 150 mM trifluoroacetic acid. The solvent program started with 1% B for 1 min and increased in steps to 2,4,8, 16, 32,64, and 100% B. In each step the increase was linear to the next higher step in 5 min, except the last step which took 10 min.
'H NMR Procedures-The one-dimensional 'H NMR spectra were obtained using a Bruker Instruments, Inc. WM-250 superconducting spectrometer equipped with a variable temperature control, a noisemodulated broadband decoupler, and a 5-mm selectively tuned 'H probe. The spectra were acquired by the application of 90" pulses with a quadrature pulse sequence at a frequency of 250.13 MHz with an acquisition time of 2.048 s and a relaxation delay of 10 s. Each spectrum was accumulated in 16,000 data points and transformed into 8,000 real data points. The number of transients was between 32 and 1000 depending on the concentration of the samples.
The two-dimensional COSY spectra were obtained with a Bruker AM-300 spectrometer. A 90'-~-90" pulse sequence was used with a magnitude calculation and contour plot to display the connectivities of juxtaposed 'Hs. The basic F2 matrix was constructed from 16 transients and 512 7 values.
Mass Spectrometry Procedures-All mass spectrometry experiments were conducted at the Midwest Center for Mass Spectrometry (Lincoln, NE) with a Kratos Analytical Instruments MS-50 triple analyzer equipped with a fast atom bombardment source, a high resolution MS-I of Nier-Johnson geometry followed by an electrostatic analyzer used as MS-11. Samples were dissolved in water at about 1 pg/pl, and a 1-p1 aliquot was added to a matrix of either dithiothreitol/dithioerythritol for positive ion spectra or triethanolamine for negative ion spectra. Fast atom bombardment by 7-keV argon atoms was used to desorb the preformed ions from the matrix which was supported on a gold probe held at +8 kV (positive ion mode) or -8 kV (negative ion mode).
The MS-MS experiments were performed using MS-I to select the sample ion, and the collisionally activated decomposition spectra were obtained by activating the ion in the third field-free region by collision with helium gas in a cell at a pressure which reduced the main beam transmission by 50%. The resulting spectra were obtained by scanning the electrostatic analyzer (MS-11). Detection was with a postacceleration detector at 15 kV with respect to ground. 10-30 scans were signal-averaged for each spectrum by using software developed at the Midwest Center for Mass Spectrometry.
Materiak-Sea urchins were purchased from Marinus, Inc. (Long Beach, CA). Radioactive NAD+ with labels on various positions was from either Du Pont-New England Nuclear or Amersham Corp. NAD+ was from either Sigma or Boehringer Mannheim. Fura-2 was from Molecular Probes.

RESULTS
Purification of the NAD+ Metabolite-Incubation of sea urchin egg extracts with NAD' produced a multitude of metabolites. The Ca2'-releasing metabolite, E-NAD, was previously purified from the mixture by anion exchange and ion pair reverse phase HPLC (3). This procedure, however, was not suitable for large scale purification of E-NAD, and the use of ion pair reagents in the reverse phase also complicated the subsequent structural characterization of the purified product. To overcome these disadvantages we introduced several modifications of the previous procedure. A semipreparative anion exchange (AG MP-1) column was used in the first step which allowed routine processing of 200-300 ml of deproteinized egg extracts. The large volume of eluent (0.5-1 liter) was then vacuum-dried. The active materials were redissolved in small volumes and applied to an analytical reverse phase column. The chromatography was performed isocratically with formic acid which is a volatile acid. The resolution of this reverse phase system was not as good as the ion pair system used earlier, and therefore an additional HPLC procedure employing the mixed mode column was introduced as the final step of purification.
As in the previous study, the induction of Ca2+ release from egg homogenates was used as an assay to monitor the metabolite through all of the purification procedures. In order to quantitate the purification in each step, the total and specific Ca2+ release activities were defined and determined as follows. A unit of Ca2+ release activity was defined as the amount sufficient to produce half-maximal Ca2+ release from 0.8 ml of a 2.5% egg homogenate. Therefore, each fraction containing activity was serially diluted, and 2-10 pl was added to 0.8-ml aliquots of homogenate to determine the dilution and volume required to produce half-maximal Ca2+ release. From the volume (pl) added to the homogenate, the serial dilution required, and the initial volume of the fraction, the total units of activity in the initial fraction were computed. The specific Ca2+ release activity was determined by dividing the total activity by the amount of UV (254 nm)-absorbing material in the fraction, which was estimated by the absorbance and the extinction coefficient measured for E-NAD (14.3 X lo3 M" cm") as described below. The Ca2+ release activity was found to elute as a single peak with u v 2 5 4 absorbance from all three HPLC columns.
Typical results of the purification procedure are shown in Table I. Passage through the AG MP-1 column increased the specific Ca2+ release activity by 15-fold with a yield of about 28%. The loss was likely due to the combination of reduced resolution of the semipreparative column and the requirement for large volume reduction of the eluent by rotary evaporation and vacuum drying before the next step. After the reverse phase and the mixed mode columns, the specific Ca2+ release activity increased by another 19-fold, and there was essentially no loss in activity. The overall purification was 284-fold with approximately a 25% yield. Typically, 4-8 nmol of E-NAD can be purified from each milliliter of egg extract incubated with 2 mM NAD' , so the percent conversion of NAD' to E-NAD is about 0.8-1.6% after correcting for the loss during purification. Total and specific Ca2+-release activities were defined and determined as described under "Experimental Procedures." Values shown are averages from two to three determinations except those for the deproteinized supernatant which are from a single determination.
The total activity value was from 18 ml of deproteinized supernatant. The homogeneity of the purified E-NAD was further verified by chromatographing it on a Whatman Partisil 5 SAX HPLC column as shown in Fig. 1. Essentially a single uv264absorbing peak was observed with a recovery of 91%. The first 30 fractions were tested for Ca2+ release activity, and only the two fractions coinciding with the UV peak contained activity. Indicated also on Fig. 1 are the elution times of 11 standard metabolites of NAD+, and none of them are found to coelute with E-NAD, suggesting its uniqueness.
A similar degree of homogeneity of the purified E-NAD was indicated when chromatographing it on yet another HPLC column, a Vydac 303NT405, which was capable of completely resolving all 12 major ribonucleotides (mono-, di-, and triribonucleotides of adenine, cytosine, guanine, and uracil). Only a single UVZs4 peak with all of the Ca2+ release activity was eluted (data not shown). The purified E-NAD was also eluted as a single peak when chromatographed on each of the three columns used in the purification procedure; therefore, it was judged homogeneous by five different HPLC columns.
Radioactive Labeling of E-NAD-To demonstrate that E-NAD was indeed derived from NAD' and to determine the structural changes brought about through conversion of NAD+ to E-NAD, radioactive NAD' labeled at various positions was used as a precursor. In each experiment, two precursor NAD's with label on different positions were used.
The ratio of the two labeled NAD+ precursors (RN.& was determined by dual isotope counting of aliquots of the reaction mixture. A similar ratio of the purified E-NAD (RE.NAD) was also determined. If both labels were conserved in E-NAD then the two ratios should be the same. That is, the isotope ratio of E-NAD divided by the isotope ratio of NAD' (RE.N,&&A~) should equal unity. On the other hand, if one label were removed during the reaction, the quantity (&.n~o/ ENAD) should be 0 (or infinite). Therefore, by measuring RE.

NAO/RNAD
with various combinations of radioactive NAD' precursors, it was possible to determine which part of the NAD' molecule was modified by the enzymatic reaction. Table II summarizes the results of a series of such labeling experiments.
By using NAD+ precursors with 3H labels on the 2-and 8-positions of the adenine ring and also uniform 14C labels on all of the carbons, the measured "C/'H ratio in E-NAD was found to be the same as in the precursors, giving an RE or carbonyl-'4C, the labels were lost from the E-NAD. This is shown in Table II for the last three precursor pairs, all of which gave RE.n.&RnAovalues of essentially 0 (0.01-0.04).
These results indicate that both the adenine ring and the adenylate a-phosphate were conserved in E-NAD while the nicotinamide group was removed. The radioactive labeling of E-NAD provided a means for measuring the extinction coefficient of the molecule as follows. First, the concentration of the radioactively labeled E-NAD (either on the adenine ring or on the adenylate (Yphosphate) was determined from its radioactivity and the known specific activity of the precursor NAD'. Then dividing the absorbance of E-NAD at 254 nm by the concentration gave a value of (14.3 f 1.5) X lo3 M-' cm-' (n= 8 + SE) for the extinction coefficient of E-NAD. This value was about 85% of the extinction coefficient of NAD' at the same wavelength. The lower extinction coefficient of E-NAD is consistent with the absence of the UV-absorbing nicotinamide group in E-NAD. The UV-absorption property of E-NAD was therefore determined only by the adenine group, and indeed we have previously shown that the UV spectrum of E-NAD was indistinguishable from that of ADP-ribose (3). The value for the extinction coefficient of E-NAD determined here was used in the rest of the study for the determination of E-NAD concentration.
'H NMR of E-NAD- Fig.  2 compares the 'H NMR spectrum of E-NAD with those of ADP-ribose and NAD'. The partial assignments of the peaks in the ADP-ribose and the NAD' spectra were based on published spectra (5,6). Comparison between the E-NAD and the NAD+ spectra reveals an immediately noticeable difference, namely all of the proton peaks associated with the nicotinamide group, HN2, -4, -5, and -6, were absent from the E-NAD spectrum. The two adenine protons, HA2 and -8, were still present, although the HAM was shifted quite a bit downfield. These interpretations are consistent with the radioactive labeling results indicating the presence of the adenine protons and the removal of the nicotinamide group. The positions of the two anomeric protons of the ribose units of E-NAD, HA~' and Hl', were in the same spectral region as in NAD', indicating that both C-l carbons still had nitrogen atoms bonded to them as in NAD'. This is in contrast to ADP-ribose which has an -OH group bonded to one of the anomeric carbons instead. As a result the position of that anomeric proton was shifted upfield to 5.3 and 5.2 ppm, respectively, for the (Y-and /3-anomers (HJ', H$').
In order to be able to identify the resonances of the rest of the ribosyl protons two-dimensional COSY experiments were performed.
The results are summarized in a contour plot shown in Fig. 3, which graphically displays the connectivities of the juxtaposed protons. Using the cross-correlation peaks and starting from the two anomeric protons, the rest of the six protons from each of the ribosyl units were sequentially traced and their positions assigned as shown.
Results presented so far indicate that except for the removal of the nicotinamide group, the rest of the NAD+ molecule including the adenine ring, the adenylate a-phosphate, and the two ribosyl units remained intact in E-NAD. Furthermore, from the NMR results we can conclude that both anomeric carbons of the two ribosyl units are bonded to nitrogen atoms.  was not observed. The exact mass of the m/z 540 ion was determined by peak matching using CsI as a reference and was measured to be 540.0526 with an accuracy of better than 2 ppm. Examination of all possible combinations of C, H, N, 0, and P (atoms present in E-NAD) which yield masses within 3 ppm of the observed m/z gave 23 possibilities. Evidence presented above established the presence of the adenine ring and the two ribosyl units in E-NAD. Therefore it must contain at least 15 carbons and 4 nitrogens. It should also contain at least one phosphate, the adenylate a-phosphate. Imposing these constraints reduces the possible combinations to seven which are listed in Table I11 together with the deviations from the observed mass of 540.0526. To distinguish among these seven possibilities the total phosphate content of E-NAD was measured according to (8). A plot of phosphate content versus amount of E-NAD was constructed with 15 data points, and the slope of the linear regression line (correlation coefficient was 0.99) was found to be 2.2. This value indicates 2 mol of phosphate for each mole of E-NAD, and from Table I11 the molecular composition of E-NAD was uniquely determined to be CI5H21N5013P2.
Structure of E-NAD-The only structure that was found to be consistent with all of the data presented above is a cyclized ADP-ribose having an N-glycosyl linkage between the anomeric carbon of the terminal ribose unit and the N6amino group of the adenine moiety. The structural formula as well as the CPK model are shown in Fig. 5. This structure gives the exact molecular composition as determined above. It also accounts for the NMR findings that both anomeric carbons of the two ribose units must have nitrogen atoms attached to them. The fact that the CPK model could be constructed indicates that there is no intrinsic steric hindrance prohibiting the cyclization. Both the structural formula and the CPK model in Fig. 5 show the conformation of the newly formed N-glycosyl bond in the a-configuration. However, a CPK model in the @-configuration can be constructed as easily, and we do not have definitive evidence for the actual conformation at the present time.
The strongest support for the cyclic structure comes from the fulfillment of an obvious prediction of the structure, namely that the hydrolysis of the N-glycosyl linkage should produce ADP-ribose as diagramed in Fig. 5. This was indeed proven to be the case by the analysis of the breakdown product of E-NAD as described in the following section.
Analysis of the Breakdown Product of E-NAD-Hydrolysis of E-NAD was induced by prolonged incubation in aqueous medium at room temperature. About 2 pmol of E-NAD was purified as described under "Experimental Procedures." The purity was confirmed by 'H NMR which gave a spectrum identical to that shown in Fig. 2. The sample was allowed to incubate in the unbuffered DzO in the NMR tube for about 40 h at room temperature and then rechromatographed on an AG MP-1 HPLC column. Two major UV-absorbing peaks were eluted. The first major peak was E-NAD since it contained Ca2+ release activity, and its elution time of about 17 min was characteristic of authentic E-NAD. The total amount of E-NAD in this peak was 0.87 pmol which was about half of the original amount. The breakdown product was in the second major peak which contained the other half (0.94 pmol) of the original material and was eluted at about 23 min. This peak did not have any Ca2+ release activity, and the retention time was characteristic of ADP-ribose (about 23 min). The peak was collected, and the 'H NMR spectrum was found to be essentially identical to ADP-ribose showing the characteristic e-and @-anomeric protons at 5.3 and 5.2 ppm (cf. the ADP-ribose spectrum in Fig. 2). Even the shape of the complex group of ribose protons at the spectral region between 4 and 4.7 ppm was essentially identical to that of authentic ADPribose. The two fragmentation patterns were virtually superimposable. Therefore, by criteria of HPLC, NMR, and mass spectrometry, it can be concluded unequivocally that the break- down product of E-NAD was ADP-ribose, a compound larger than E-NAD by one water molecule.

DISCUSSION
Four different approaches were used in this study to characterize structurally E-NAD, the Ca2+-releasing metabolite of NAD' .
1) The first approach used radioactive precursors, and the results indicated that the adenine ring was conserved in E-NAD while the nicotinamide group was not. That the pyrophosphate linkage was also unchanged was suggested by the conservation of the adenyZate-c~-~*P label and by the total phosphate determination showing 2 mol of phosphate/mol of 2) The second approach was 'H NMR, and four conclusions could be drawn. First, there were two, and only two, protons on the adenine ring. This rules out structures such as those in which the "amino group has been replaced by a proton. Second, both ribose units were intact with all 12 protons identified. Third, the nicotinamide group was removed which confirmed the results from radioactive labeling. Fourth, both anomeric carbons of the ribose units were bonded to nitrogen.
3) The third approach was mass spectrometry. The exact mass measurement allowed unique specification of the molecular composition of E-NAD to be C16H21NS013P2.

4)
The strongest evidence was provided by the fourth approach which was to analyze the only major breakdown product of E-NAD. Using HPLC, NMR, and mass spectrometry, the product was unequivocally identified as ADP-ribose.
The very fact that E-NAD differs from ADP-ribose by one water molecule and that E-NAD could be converted to ADPribose by simple hydrolysis strongly indicated that E-NAD was a cyclic compound. Theoretically there are several ways E-NAD.
ADP-ribose can be cyclized. For example, a phosphodiester linkage could be formed with either ribose unit in a manner similar to CAMP. However, these types of cyclic structures are contrary to the NMR results showing that both anomeric carbons of the ribose units were bonded to nitrogen. In fact the NMR results dictated that the cyclization must occur at the terminal anomeric carbon and must be linked to a nitrogen on the adenine moiety. There are four nitrogen atoms in the ring itself and one free amino group. If the linkage were to any of the ring nitrogens it would put a positive charge on the ring and would be expected to change the UV-absorption spectrum substantially as compared with ADP-ribose. Instead, it was observed that the UV spectrum of E-NAD was virtually identical to that of ADP-ribose (3). Furthermore, linkage to a ring nitrogen would result in a structure having a molecular weight less than the observed value by one proton.
The only structure that was found to be consistent with all available data was the one shown in Fig. 5. Analysis of the collisional activated decomposition spectrum of E-NAD (inset of Fig. 4) provided further support of this cyclic structure. For nucleotides such as ADP-ribose the major fragmentation pathway is the breakage of the pyrophosphate bond. As shown in Fig. 6, this gave rise to the major peak at m/z 345 corresponding to the loss of (ribose-HP03) from the molecule. However, this fragmentation pathway is not possible in a cyclized molecule such as E-NAD since the ribose phosphate is still attached to the molecule through the N-glycosyl linkage even after the pyrophosphate bond is broken. Indeed, as shown in the inset of Fig. 4, the major fragmentation pathway for E-NAD was a simple loss of phosphate and water (HP03 + HzO) giving rise to the m/z 426 peak. This was followed by loss of another water resulting in the peak at m/z 408. This pattern is consistent with the pyrophosphate being broken in a manner similar to that observed in ADP-ribose, but because of the N-glycosyl linkage the ribose unit remained attached, and only H P 0 3 + HzO was lost.
The secondary fragmentation pathway in ADP-ribose was the loss of the adenine base giving rise to the two smaller peaks at mlt 424 and 406 which corresponded to (Mbase H)and (Mbase H -H20)-, respectively. This pathway is not possible in E-NAD because the base is linked to both ribose units. Indeed, this facile loss of base, characteristic of nucleotides (7), was not observed. The collisional activated decomposition spectrum of E-NAD is therefore consistent with it being cyclic.
The cyclic structure of E-NAD can also account for the anomalous chemical shifts of the protons HA8 and HA^' seen in the NMR spectrum (cf. Fig. 2). The unusual chemical environment in which these two protons exist is apparent from the CPK model shown in Fig. 5 . In order for the molecule to cyclize, the sugar-phosphate backbone has to fold backward. Consequently, the HAS proton is surrounded by the pyrophosphate group while the HA^' proton is brought into close proximity with the adenine ring. The unusual environment in which these two protons exist could account for the downfield shifts observed for these protons.
The bulk of the structural characterizations presented in this study strongly indicates the cyclic structure for E-NAD. We therefore propose a descriptive common name, cyclic ADP-ribose, for the metabolite. The novelty of the structure itself warrants detailed investigation into the biochemistry of this NAD' metabolite. Moreover, cyclic ADP-ribose appears to be a commonly occurring metabolite since a variety of mammalian tissue extracts can also produce it upon incubation with NAD'.' Taken together, these results indicate that the importance of this metabolite could go beyond novelty and suggest that it may be a general second messenger like IP3 for mobilizing intracellular Ca2+. If so, the cyclic ADPribose would represent an addition to the already well known second messenger family of cyclic nucleotides.