Muscarinic Receptor-mediated Dual Regulation of ADP-ribosyl Cyclase in NG108-15 Neuronal Cell Membranes*

Cyclic ADP-ribose (cADP-ribose) is an endogenous modulator of ryanodine-sensitive Ca2+ release channels. An unsolved question is whether or not cADP-ribose mediates intracellular signals from hormone or neurotransmitter receptors. The first step in this study was to develop a TLC method to measure ADP-ribosyl cyclase, by which conversion of [3H]NAD+ to [3H]cADP-ribose was confirmed in COS-7 cells overexpressing human CD38. A membrane fraction of NG108-15 neuroblastoma × glioma hybrid cells possessed ADP-ribosyl cyclase activity measured by TLC. Carbamylcholine increased this activity by 2.6-fold in NG108-15 cells overexpressing m1 or m3 muscarinic acetylcholine receptors (mAChRs), but inhibited it by 30–52% in cells expressing m2 and/or m4 mAChRs. Both of these effects were mimicked by GTP. Pretreatment of cells with cholera toxin blocked the activation, whereas pertussis toxin blocked the inhibition. Application of carbamylcholine caused significant decreases in NAD+ concentrations in untreated m1-transformed NG108-15 cells, but an increase in cholera toxin-treated cells. These results suggest that mAChRs couple to ADP-ribosyl cyclase within cell membranes via trimeric G proteins and can thereby control cellular function by regulating cADP-ribose formation.

Recently, it has been shown that the formation of cADPribose is regulated by nitric oxide or cGMP (21)(22)(23) and that nitric oxide or cGMP is increased by stimulation with agonists (24,25). These findings suggest the hypothesis that the regulation of the cADP-ribose level is located far downstream in the signal transduction cascade from receptors (11). An alternative hypothesis is that the cADP-ribose formation is regulated by ADP-ribosyl cyclase through the direct action of G proteins activated by receptors within the surface membrane, as already shown for the formation of cyclic AMP, inositol 1,4,5trisphosphate, and diacylglycerol (26 -28). To test this hypothesis, we used NG108-15 neuroblastoma ϫ glioma hybrid cells (29), in which signal transduction from receptors to effectors has been extensively characterized (29,30). In particular, in NGPM1-27 cells (31), which overexpress muscarinic acetylcholine receptors (mAChRs), it has been shown that intracellular NAD ϩ or NAD ϩ metabolites are involved in signal transduction from m1 mAChRs to K ϩ channels (32,33). In this context, such neuronal cell lines have advantages for analyzing receptor-ADP-ribosyl cyclase coupling in detail.
For measurement of ADP-ribosyl cyclase, high pressure liquid chromatography (HPLC) is commonly used to separate cADP-ribose-related compounds (1,2,8,14,15,17,19,34,35). However, since it takes 30 -60 min to process one sample, it is essential to develop a much more rapid method that can allow processing of multiple samples at once. There are two papers that describe ADP-ribosyl cyclase assay by TLC (21,36), in which NAD ϩ migrates faster than cADP-ribose. The methods used in those reports seem to be affected by large amounts of radiolabeled substrates. We here developed a TLC method that overcomes this problem and allows separation of cADP-ribose in up to 19 samples within 40 -50 min. Our TLC method was first tested on COS-7 cells overexpressing human CD38 and was shown to be applicable for measuring ADP-ribosyl cyclase activity. We demonstrate that crude cell membranes of NG108-15 cells possess ADP-ribosyl cyclase activity and that such activity is activated or inhibited in a mAChR subtypespecific manner in NG108-15 cells overexpressing distinct mAChR subtypes. Furthermore, to ascertain the intracellular role of the catalytic activity of ADP-ribosyl cyclase in neuronal cell membranes, the time course of [NAD ϩ ] i after extracellular application of acetylcholine to these cells was investigated.
Thin-layer Chromatography-Two-l aliquots were spotted on silica gel plastic TLC sheets (20 ϫ 10 cm), and the layers were developed in the ascending direction for 40 -120 min at room temperature with four different solvents, as listed in Table I. A mixture of water/ethanol/ ammonium bicarbonate (in the ratio 30%:70%:0.2 M) was used most frequently.
Cell Membrane Preparation from COS-7 Cells-Mock-and CD38transfected COS-7 (COS-CD38) cells were washed once with Dulbecco's phosphate-buffered saline without Ca 2ϩ and Mg 2ϩ (PBS(Ϫ)), dissociated in 10 ml of PBS(Ϫ), and collected by centrifugation at 300 ϫ g for 5 min. The washed cells were frozen and stored at Ϫ80°C. Immediately before use, the cell pellet was thawed and suspended in 10 mM Tris-HCl solution, pH 7.4, with 5 mM MgCl 2 (1 ml for cells from a 75-cm 2 flask) at 4°C for 25 min (39). The suspension was homogenized in a Dounce glass homogenizer with 50 strokes. The resultant homogenate was centrifuged at 4°C for 5 min at 1000 ϫ g to remove unbroken cells and nuclei. Crude membrane fractions were prepared by centrifugation (twice) of homogenates at 105,000 ϫ g for 15 min. The supernatant was removed, and the final pellet (100,000 ϫ g) was dispersed in 10 mM Tris-HCl solution, pH 6.6. Protein was measured by the Bio-Rad protein assay dye reagent.
ADP-ribosyl Cyclase Assay-Each 20-l reaction mixture contained 50 mM Tris-HCl, pH 6.6, 100 mM KCl, 5 mM MgCl 2 , 0.1 mM EDTA, 2 M ␤-NAD ϩ , 0.11 M ␤-[2,8-adenine-3 H]NAD ϩ (0.06 Ci), and 1.2-6 g of membrane proteins according to a formula reported previously (34), with a slight modification. In experiments in which the activation effect by agonists was measured, MgCl 2 and EDTA was replaced with 10 M CaCl 2 . Reaction mixtures were incubated for 0.5-16 min at 37°C. The production of [ 3 H]cADP-ribose and [ 3 H]ADP-ribose was proportional to protein concentration within the range of 1-20 g of membrane fraction protein/reaction mixture. Reactions were stopped by adding 2 l of 48% trichloroacetic acid to the reaction mixture. Aliquots were centrifuged for 2.5 min at 14,000 ϫ g, and 2 l of the supernatant were used for analysis by TLC. The positions of authentic cADP-ribose, ADP-ribose, and NAD ϩ after UV detection were confirmed in each run. Corresponding positions (ϳ1 ϫ 0.7 cm) were cut, and the radioactivity was counted in a liquid scintillation counter.

Autoradiography of Thin-layer Chromatograms with [ 3 H]NAD ϩ -
The same reaction mixture used for the ADP-ribosyl cyclase assay containing 0.06 Ci of [2,8-adenine-3 H]NAD ϩ was incubated with membranes of COS-CD38 and NGPM1-27 cells. Two l of reaction mixture were spotted on TLC sheets and developed. Autoradiography was carried out after exposure on a Fuji BAS 1000 3 H imaging plate for 24 -36 h.
Mass Spectrographic Measurements-Silica gel spots of the enzyme reaction samples corresponding to the migration position of authentic cADP-ribose were removed by shaving. NAD ϩ metabolites were recovered in water or acidic solutions and concentrated by freeze-drying. The freeze-dried material was dissolved in 20 -200 l of H 2 O and subjected to fast atom bombardment mass spectral analysis (JMS-DX303, JOEL Inc., Tokyo).
NAD ϩ Content-NGPM1-27 cells were cultured on polyornithinecoated dishes (35 mm in diameter) for 4 days. The NAD ϩ content in the supernatant of the heat-inactivated cell homogenate was determined by a slight modification (32) of an enzyme cycling method (40).
Although 3 H count accumulation in the ADP-ribose fraction was below the detectable level during the first 1-2 min with COS-CD38 cell membranes, [ 3 H]ADP-ribose formation was gradually increased and exceeded the cADP-ribose level after 8 min (Fig. 3). These results seem to faithfully reflect the twostep enzyme reaction of human CD38, from NAD ϩ to ADPribose via cADP-ribose, and thereby indicate that our TLC method is sensitive enough to measure ADP-ribosyl cyclase activity.
ADP-ribosyl Cyclase Activity in NGPM1-27 Cells-Using the TLC method, ADP-ribosyl cyclase activity was examined with crude membranes (100,000 ϫ g) of NGPM1-27 cells. The 3 H count in the cADP-ribose fraction increased linearly during the first 0.5-6 min and thereafter was maintained at a steadystate level (Fig. 4B) or decreased (Fig. 3). The average activity measured by 3 H accumulation in the cADP-ribose fraction was 229.8 Ϯ 29.5 pmol/min/mg of protein (n ϭ 12). In contrast, the 3 H count in the ADP-ribose fraction increased linearly over 16 min (Figs. 3 and 4B). The average activity was 296.3 Ϯ 63.5 pmol/min/mg of protein (n ϭ 7). The total rate of NAD ϩ utilization was measured under the same conditions, except that 2 M NAD ϩ was mixed with nicotinamide-labeled [ 3 H]NAD ϩ (0.01 Ci/sample). The release of [ 3 H]nicotinamide catalyzed by NGPM1-27 cell membranes was 460.5 Ϯ 59.3 pmol/min/mg of protein (n ϭ 10). It thus appears that a large part of NAD ϩ is converted to either cADP-ribose or ADP-ribose with some other metabolites.
To verify that the above 3 H accumulation in cADP-ribose fractions is due mainly to accumulation of [ 3 H]cADP-ribose produced by the activity of the ADP-ribosyl cyclase of NGPM1-27 cells, the compounds collected in the cADP-ribose fraction by TLC were analyzed. The first test carried out was to show whether or not the product is converted to ADP-ribose by heat inactivation (1) or CD38. Extracts from the silica gel were inactivated by treating them for 30 min at 95°C or were metabolized by cADP-ribose hydrolase in maltose-binding protein-CD38 for 16 min before lyophilization and then rechromato-graphed. The radioactivity in the ADP-ribose fraction recovered increased with both treatments, and the amount of 3 H count in the cADP-ribose fraction was reduced correspondingly. Secondarily, lyophilized samples extracted from the ADP-ribose fraction in the rechromatographed thin-layer chromatogram were analyzed by mass spectrometry. A main peak of mass ions at m/z 558, a peak unique for ADP-ribose (1), was found, reflecting either the original ADP-ribose or ADP-ribose hydrolyzed from cADP-ribose during processing, or both. These two lines of evidence suggest that the majority, if not all, of the products separated in the cADP-ribose fraction are cADPribose, which is very hydrolyzable (1). Armed with this evidence, the receptor control mechanism of membrane-bound ADP-ribosyl cyclase was investigated in NG108-15 cells expressing various subtypes of mAChRs.
Carbamylcholine-induced Changes in ADP-ribosyl Cyclase Activity-After application of 10 M carbamylcholine (CCh), an increased cyclase activity in NGPM1-27 cell membranes over the control level was observed for 2-4 min (Fig. 5A) and was abolished in the presence of 0.1-1 M atropine. The average increase was 2.61 Ϯ 0.27-fold (n ϭ 6) (Fig. 6A), which was significantly larger (p Ͻ 0.001). The activation by CCh was dose-dependent, with an ED 50 value of 26 Ϯ 1.4 nM (n ϭ 3). Addition of 10 M GTP alone or together with CCh also caused increased activity (Fig. 5B). Membranes washed extensively with 10 mM Tris-buffered saline, pH 6.6, with or without 0.1 mM EDTA (n ϭ 4) retained the cyclase activity, which could be stimulated by 10 M GTP or GTP␥S (2.97-and 2.27-fold, respectively; n ϭ 2), although no activation by CCh was observed (102.0 Ϯ 5.8% of the control level; n ϭ 4). Cyclic GMP (10 M), however, had no effect on ADP-ribosyl cyclase activity (106.7 Ϯ 9.4%; n ϭ 7) in this membrane system.
We next examined which G proteins mediate these mAChR subtype-specific and GTP-dependent effects. When m1-and m3-transformed cells were pretreated with 100 ng/ml cholera toxin (CTx) for 5-8 h, CCh-induced activation of the cyclase activity was specifically eliminated, and instead, CCh induced a significant inhibition. The CCh-induced inhibition remained in a set of nontransfected, mock-transfected, and m2-or m4transformed cells that had been treated with CTx (Fig. 6B). Thus, CCh induced an inhibition to 49.7-70.8% of the control level in all types of CTx-treated cells (p Ͻ 0.001). In contrast, treating the above set of cells (except for m1-and m3-transformed cells) with 100 ng/ml pertussis toxin (PTx) for 12 h prevented the CCh-induced inhibition (Fig. 6C). The significant activation by CCh was retained in PTx-treated m1-and m3transformed cells (171 Ϯ 19 and 150 Ϯ 16% of the control level, respectively; n ϭ 6; p Ͻ 0.01), although the apparent activation was reduced. The toxin sensitivities suggest that the activation or inhibition signal from each mAChR subtype to ADP-ribosyl cyclase is conveyed by distinct G proteins.
Agonist-induced Decrease in [NAD ϩ ] i -To confirm the above agonist effects in vivo, we examined agonist-stimulated changes in substrate levels, as shown previously (32).

DISCUSSION
The chromatographic procedure used here resulted in a resolution sufficient to measure 3 H accumulation in the cADPribose fraction originating from adenine-labeled [ 3 H]NAD ϩ , and it was free from interference by the labeled substrate. One possible limitation could be that the cADP-ribose fraction was contaminated by ϳ15% of the newly synthesized [ 3 H]ADPribose ( Fig. 2A). However, the risk that we measured the ADPribose count in the cADP-ribose fraction seems to be negligible since different patterns of 3 H accumulation in ADP-ribose and cADP-ribose fractions were obtained (Fig. 4), reflecting that distinct materials are accumulated with different rates of formation and degradation. The autoradiograms for COS-CD38 and NGPM1-27 cells (Fig. 3) clearly demonstrated the overall fate of NAD ϩ in that the majority of NAD ϩ was converted to either ADP-ribose or cADP-ribose or both in two or more enzyme reaction steps, probably ADP-ribosyl cyclase and cADPribose hydrolase.
The NAD ϩ concentration used in our reaction mixture (2 M) was much lower than values (ϳ100 M) used in other studies (1,2,5,15). However, our concentration is very similar to the estimated K m value (3-5 M) for ADP-ribosyl cyclase in NG108-15 cell membranes. 2 Under such conditions, [ 3 H]cADPribose accumulated linearly for 2-4 min and [ 3 H]ADP-ribose for 16 min, suggesting that it is not an insufficient dose. Thus, our TLC method was used for measurement of ADP-ribosyl cyclase, at least during the first few minutes. We do not know whether or not this TLC method is applicable for measurement of other recently found enzyme reactions in ADP-ribosyl cyclase, such as production of cyclic 2Ј-phosphoadenosine diphosphoribose (41), nicotinic acid adenine 5Ј-diphophate ribose (42), and dimeric ADP-ribose (43). It remains to be tested how our TLC method differs in accuracy from HPLC and fluorometric measurement of cyclic GDP-ribose formation (20,44,45).
The results provide the first evidence that cholinergic signals are transduced from mAChRs to ADP-ribosyl cyclase within the membrane in a receptor subtype-specific fashion (Figs. 5 and 6). The finding that the inhibition of ADP-ribosyl cyclase via m2/m4 mAChRs is mediated through a PTx-sensitive G protein resembles the inhibitory signal transduction known for adenylate cyclase in NG108-15 cells (29 -31). Coupling of m1/m3 mAChRs to ADP-ribosyl cyclase is relatively resistant to PTx, but highly sensitive to CTx, which seems to be different from the signal pathway to phospholipase C␤ (30,31). Thus, the coupling from mAChRs to ADP-ribosyl cyclase via G proteins would be a new mode of signal transduction, in parallel with the known pathways to adenylate cyclase and phospholipase C in NG108-15 cells.
The simplest explanation for our [NAD ϩ ] i data in NGPM1-27 cells (Fig. 7) could be that NAD ϩ consumption is accelerated by the activation with the agonist of a membrane-bound form of ADP-ribosyl cyclase whose catalytic site resides in the cell interior. This speculation, however, would require a different topology from that proposed for the CD38 cell-surface antigen, whose catalytic sites of ADP-ribosyl cyclase (15,18,45), ADPribose hydrolase (15,46), and NAD ϩ glycohydrolase (19) are located at the extracellular side. Thus, ADP-ribosyl cyclase activity detected in the intracellular side in NG108-15 cells could not be the same as CD38, but a neuronal isoform of ADP-ribosyl cyclase. The degree of identity between ADP-ribosyl cyclase in NG108-15 neuronal cells and that in ovotestis of Aplysia (12) or CD38 should be further examined.
cADP-ribose formation was not affected by cGMP in FIG. 7. CCh-induced change in [NAD ؉ ] i in intact NGPM1-27 cells. Cells were grown in 35-mm culture dishes for 4 days. The growth medium was replaced with 2 ml of 10 mM Tris-buffered Dulbecco's modified Eagle's medium and incubated for 40 min at 37°C. The preincubated cells were then stimulated by gently adding 1 ml of Dulbecco's modified Eagle's medium alone (Con) or Dulbecco's modified Eagle's medium with 30 M CCh (to give a final concentration of 10 M) for the indicated periods of time. Incubation was stopped by replacing the medium with 1 ml of cold PBS(Ϫ) and washing again with 1 ml of cold PBS(Ϫ) with 10 mM nicotinamide. Cells were scraped, and the homogenates were heat-inactivated. [NAD ϩ ] i was measured as described previously (32). Each value represents the mean Ϯ S.E. of six dishes in triplicate cultures. The control value for the NAD ϩ level in NGPM1-27 cells was 4.84 Ϯ 0.43 nmol/10 6 cells (n ϭ 4). ૾, significantly different from the value at time 0 at p Ͻ 0.001; *, **, and ***, significantly different from the control values at each time at p Ͻ 0.05, p Ͻ 0.01, and p Ͻ 0.001, respectively. NG108-15 cell membranes. This conclusion, however, does not exclude the regulatory mechanism in the cytoplasm, where cytosolic ADP-ribosyl cyclase could be activated through the nitric oxide/cGMP-dependent cascade (11). In summary, our results suggest that stimulation of conventional neurotransmitter receptors can regulate cellular function through cADPribose as a second messenger, independently from or in concert with other second messengers.