Extracellular Application of Nicotinic Acid Adenine Dinucleotide Phosphate Induces Ca 2 (cid:1) Signaling in Astrocytes in Situ *

Nicotinic acid adenine dinucleotide phosphate (NAADP (cid:1) ) has been identified as a novel second messenger triggering Ca 2 (cid:1) release from intracellular stores. Here we report that murine cortical astrocytes in culture and in acute slices respond with transient intracellular Ca 2 (cid:1) increases to extracellularly applied NAADP (cid:1) and express the NAADP (cid:1) -producing enzyme CD38. The Ca 2 (cid:1) tran-sients triggered by NAADP (cid:1) occurred with an average delay of 35 s as compared with ATP-triggered Ca 2 (cid:1) signaling, suggesting that NAADP (cid:1) may have to enter the cell to act. Blockage of connexin hemichannels (a possible entry route for NAADP (cid:1) into the cell) reduced the number of astrocytes responding to NAADP (cid:1) . Disruption of lysosomes as the suggested site of NAADP (cid:1) receptors reduced the number of astrocytes responding to NAADP (cid:1) strongly. The NAADP (cid:1) -triggered Ca 2 (cid:1) signal also depended on intact endoplasmic reticulum Ca 2 (cid:1) stores linked to activation of inositol 1,4,5-trisphosphate receptors and on the activity of voltage-gated Ca 2 (cid:1) channels. Adenosine receptor-mediated signaling contributes to the NAADP

Nicotinic acid adenine dinucleotide phosphate (NAADP ؉ ) has been identified as a novel second messenger triggering Ca 2؉ release from intracellular stores. Here we report that murine cortical astrocytes in culture and in acute slices respond with transient intracellular Ca 2؉ increases to extracellularly applied NAADP ؉ and express the NAADP ؉ -producing enzyme CD38. The Ca 2؉ transients triggered by NAADP ؉ occurred with an average delay of 35 s as compared with ATP-triggered Ca 2؉ signaling, suggesting that NAADP ؉ may have to enter the cell to act. Blockage of connexin hemichannels (a possible entry route for NAADP ؉ into the cell) reduced the number of astrocytes responding to NAADP ؉ . Disruption of lysosomes as the suggested site of NAADP ؉ receptors reduced the number of astrocytes responding to NAADP ؉ strongly. The NAADP ؉ -triggered Ca 2؉ signal also depended on intact endoplasmic reticulum Ca 2؉ stores linked to activation of inositol 1,4,5trisphosphate receptors and on the activity of voltage-gated Ca 2؉ channels. Adenosine receptor-mediated signaling contributes to the NAADP ؉ -evoked signal, since it is strongly reduced by the adenosine receptor blocker CGS-15943. Moreover, NAADP ؉ triggered responses in all other cell types (cultured cerebellar neurons, microglia, and oligodendrocytes) of the central nervous system.
Astrocytes are in intimate contact with neurons, in particular with synapses, and are able to sense, react to, and even influence neuronal activity. The astrocytic response to neuronal activity can be most readily detected by observing changes in the intracellular Ca 2ϩ concentration mediated via transmitter receptors (e.g. glutamate, GABA (␥-aminobutyric acid), or acetylcholine) or via other messengers, such as nitric oxide. Ca 2ϩ increases can occur by Ca 2ϩ influx via the plasma membrane as found for the nitric oxide-mediated neuron-glia signaling in the cerebellum (1). Ca 2ϩ can furthermore be released from intracellular stores by activation of metabotropic receptors such as GABA B receptors, which mediate neuron-glia signaling in the hippocampus (2). Indeed, astrocytes express a variety of receptors linked to the Ca 2ϩ mobilization from intracellular stores (3). To produce specific signals with one and the same messenger (Ca 2ϩ ), several second messenger pathways and different sources of Ca 2ϩ are involved. It seems important, at the current stage, to decipher the astrocytic Ca 2ϩ code, since this would lead to further understanding of how these cells can react to and influence the neuronal network.
The best investigated second messenger linking activation of metabo-tropic receptors and Ca 2ϩ release is inositol 1,4,5-trisphosphate (IP 3 ). 2 The importance of other intracellular messengers, such as cyclic ADPribose (cADPR), is less well understood (4) or still hypothetical as for the two sphingomyelin metabolites sphingosine 1-phosphate (5) and sphingosylphosphorylcholine (6). Almost a decade ago, a novel intracellular Ca 2ϩ -releasing second messenger was identified in sea urchin eggs, nicotinic acid adenine dinucleotide phosphate (NAADP ؉ ) (7), which binds to an unknown receptor. Since its discovery, there is increasing evidence that NAADP ϩ also has a physiological role in Ca 2ϩ signal transduction in vertebrates, including mammalian cells (8). In the vertebrate nervous system, intracellular Ca 2ϩ release by NAADP ϩ has only been demonstrated from rat brain microsome preparations (9) and in frog motoneurons (10). NAADP ϩ -specific binding sites are found in gray and white matter of rat brain, but the cellular specificity of the binding site has not yet been determined (11). The major way of NAADP ϩ synthesis in mammalian tissue takes place under acidic conditions and in the presence of nicotinic acid from NADP ϩ by the type II transmembrane glycoprotein ADP-ribosyl cyclase CD38. In a neutral or alkaline environment, CD38 exclusively catalyzes the conversion of NAD ϩ into cADPR (12). Apart from cell membrane localization, CD38 has been located to various intracellular organelle membranes and can moreover be internalized by endocytosis (13). Therefore, NAADP ϩ synthesis most likely takes place in membranes of intracellular acidic organelles such as lysosomes or late endosomes, whereas CD38 in the plasma membrane synthesizes cADPR. Regarding the mammalian central nervous system, CD38 expression was demonstrated in the rat cerebral and cerebellar cortex for neurons and astrocytes by immunoelectron microscopy (14). Also, cultured rat hippocampal astrocytes are immunopositive for CD38, with plasma membrane staining and staining in the perinuclear Golgi region (15). Furthermore, CD38 activity could be detected in rat cortical astrocytic cultures and their membrane preparations, where activity was also found intracellularly (16), in rat hippocampal astrocytic cultures (15), and in mouse (17) and rat brain homogenates (18).
NAADP ϩ as an intracellular messenger acts at very low concentrations. The cytoplasmic concentration has been determined for human red blood cells to be as low as ϳ60 nM (19). In contrast to invertebrate cells, where NAADP ϩ binding is irreversible and the receptor is inactivated even at subthreshold concentrations (20), NAADP ϩ binding is reversible in mammalian cells, and the receptor is only inactivated at high concentrations (ϳ50 -100 M) (8). The intracellular Ca 2ϩ stores sensitive to NAADP ϩ were in most studies in cell-free systems different from the thapsigargin-sensitive stores, which respond to IP 3 and cADPR (21), although there is one report in which NAADP ϩ appears capable to release Ca 2ϩ directly from the nuclear envelope by acting on ryanodine receptors, a process that is thapsigargin-sensitive (22). NAADP ϩ -induced Ca 2ϩ signaling can be abolished by glycyl-L-phenylalanine 2-naphthylamide (GPN), a substrate of lysosomal cathepsin C whose hydrolysis leads to the specific osmotic lysis of lysosomes, which can also function as Ca 2ϩ stores (23). In intact cells, both types of stores interact, and it was suggested that the NAADP ϩ -activated stores serve to prime the other stores and are involved in initiating Ca 2ϩ oscillations then maintained by IP 3 and cADPR (24). In addition, there is evidence for Ca 2ϩ influx in response to NAADP ϩ (25).
The general aim of this study was to investigate whether extracellularly applied NAADP ϩ could elicit Ca 2ϩ signaling in intact cells of the mouse central nervous system, specifically in astrocytes. In most studies carried out with intact cells so far, NAADP ϩ was injected into single cells (7) or released intracellularly by flash photolysis of the caged compound (26). In one case, application of extracellular NAADP ϩ was tested on starfish oocytes, but no Ca 2ϩ rise was detected (27). We showed for the first time that all cell types of the murine central nervous system are able to respond to extracellular NAADP ϩ . Aware of the fact that there is no physiological evidence for an extracellular action of NAADP ϩ , it was applied extracellularly at relatively high concentration, assuming its uptake into the cells. Similarly, cADPR, the other product of CD38, or its precursor, NAD ϩ , could be applied extracellularly and was effective in evoking Ca 2ϩ responses in cultured astrocytes (15) and Müller cells (28).

MATERIALS AND METHODS
Animals, Preparation of Brain Slices-For slice preparation, 10 -14day old mice (NMRI mice) were decapitated, and their brains were removed. Cortical slices of 250 m thickness were cut in ice-cold bicarbonate buffer (134 mM NaCl, 2.5 mM KCl, 2 mM CaCl 2 , 1.3 mM MgCl 2 , 26 mM NaHCO 3 , 1.25 mM K 2 HPO 4 , and 10 mM glucose) using a vibratome (VT 1000 S; Leica, Heidelberg, Germany). By gassing with carbogen, the pH was adjusted to 7.4. Slices were stored at room temperature in gassed buffer for at least 45 min prior to staining.
Glial Culture Preparations-Cultures were prepared from the brains of newborn NMRI mice as described previously (29) with small modifications. For cortical astrocyte cultures, only the cortex was used, whereas for the preparation of microglial and oligodendrocyte cultures, the whole brain was taken. Briefly, brain tissue was freed from blood vessels and meninges, trypsinized, and gently triturated with a fire-polished pipette in the presence of 0.05% DNase (Worthington). After washing twice, cells were cultured in 75-cm 2 flasks and for astrocyte cultures also directly plated on poly-L-lysine (100 g/ml; Sigma)-coated glass coverslips (A 15 mm) at densities of 3-5 ϫ 10 4 cells/coverslip, kept in A-10-cm-dishes using Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum (FCS), 2 mM L-glutamine, 100 units/ml penicillin, and 100 g/ml streptomycin. One day later, cultures were washed twice with Hanks' balanced salt solution to remove cellular debris.
To obtain astrocytic cultures, the cells were maintained for 4 days, and after reaching a subconfluent state, microglial cells and oligodendrocytes as well as their early precursors were dislodged by manual shaking and removed by washing with Hanks' balanced salt solution. The purity of the astrocytes was routinely determined by immunofluorescence using an antibody against glial fibrillary acidic protein (Sigma), a specific astrocytic marker. The cultures typically showed more than 90% cells positive for glial fibrillary acidic protein.
To obtain microglial and oligodendrocytic cultures, mixed glial cells were maintained longer (9 -12 days) with medium changes every third day until the astrocytes were confluent. Microglial cells were then separated from the underlying astrocytic monolayer by gently shaking the bottles for 1 h at 37°C in a shaker-incubator (100 rpm). The microglia were seeded on glass coverslips at a density of 5 ϫ 10 4 cells/coverslip, kept in four-well Nunc culture plates. Cultures usually contained Ͼ95% microglial cells, as revealed by staining with Griffonia simplicifolia isolectin B4 (Sigma), a marker for microglia. Cultures were used for experiments 1-5 days after plating.
Oligodendrocytes and their early precursors were then dislodged from the astrocytic monolayer by strong manual shaking and plated on poly-L-lysine (20 g/ml)-coated glass coverslips at densities of 3-5 ϫ 10 4 cells/coverslip, kept in A-10-cm-dishes in SATO medium supplemented with 2% FCS, 100 units/ml penicillin, and 100 g/ml streptomycin. Without additional growth factors, most precursors develop into oligodendrocytes, which begin to mature within a few days.
Cerebellar Neuron Preparation-Cerebellar neurons were prepared from 5-day-old NMRI mice as described by Schnitzer and Schachner (30) with some modifications. Briefly, neurons were dissociated from freshly dissected cerebelli by mechanical disruption in the presence of trypsin (Roche Applied Science) and DNase in HANKS (Biochrom, Berlin, Germany). Afterwards, neurons were plated on poly-L-lysine-coated (100 g/ml) glass coverslips, kept in A-10-cm-dishes at a density of 3 ϫ 10 5 cells/coverslip in basal medium Eagle, 10% FCS, 25 mM KCl, 32 mM glucose. After 1 day, the culture medium was changed to SATO medium supplemented with 2% FCS, 100 units/ml penicillin, and 100 g/ml streptomycin, and cytosine arabinoside (10 M; Sigma) was added to prevent proliferation of glial cells. Experiments started a week after plating.
Cell Lines-The mouse glioma cell line GL261 was purchased from NCI-Frederick, and the 1321N1 human astrocytoma cell line was a kind gift of Georg Reiser (Institut für Neurobiochemie, Universität Magdeburg, Germany). Cell lines were kept in Dulbecco's modified Eagle's medium supplemented with 10% FCS, 2 mM L-glutamine, 100 units/ml penicillin, and 100 g/ml streptomycin in 75-cm 2 flasks. The cell line GL261 was stably transfected with the red fluorescent protein dsRedII (BD Bioscience, Heidelberg, Germany), and Geneticin (G-418 sulfate; Invitrogen) was added to their medium at a concentration of 0.15 mg/ml for selection of dsRed-positive cells. Medium was changed twice a week, and cells were split after reaching confluence, usually once a week. Cells were plated on coverslips, which were kept in A-10-cm dishes at a density of 4 ϫ 10 5 cells/coverslip. They were washed to remove cellular debris the day after plating.
All cultures were maintained at 37°C in a humidified atmosphere of 95% air, 5% CO 2 . Cell media and supplements were purchased from Invitrogen if not stated otherwise.
Ca 2ϩ Imaging-Slices were incubated with the Ca 2ϩ indicator dyes Fluo-4-acetoxymethylester (10 M; Molecular Probes, Inc., Eugene, OR) in bicarbonate buffer containing 0.02% Pluronic-127 (Molecular Probes) at room temperature for 40 -50 min followed by 10 -15 min at 37°C and 5% CO 2 . Slices were transferred to a perfusion chamber on an upright microscope (Axioskop FS, Zeiss, Oberkochen, Germany) equipped with a ϫ20 water immersion objective (UMPlanFl; numeric aperture 0.5; Olympus, Hamburg, Germany) and fixed in the chamber using a U-shaped platinum wire with a grid of nylon threads. Slices were superfused with gassed bicarbonate buffer at a flow rate of 4 -6 ml/min. Substances were applied by changing the perfusate. Intracellular Ca 2ϩ changes were detected using either a confocal laser-scanning microscope (Sarastro 2000; Molecular Dynamics) with an open pinhole or a conventional imaging system with a cooled CCD camera (SensiCam; PCO, Kelheim, Germany). As a light source for fluorophore excitation, we used the 488-nm band of an Ar ϩ /Kr ϩ laser for the laser-scanning microscope or a monochromator set to 488 nm for the imaging system (Till Photonics, München, Germany). Emission was detected with a 510-nm-long pass filter. Due to the scanning rate of the confocal microscope, the sampling rate was limited to 0.25 Hz, whereas with the CCD camera, 0.5 Hz was achieved. Images were stored on a PC and processed with conventional software (ImagePro; Media Cybernetics). The magnitudes of Ca 2ϩ concentration changes were detected by temporal analysis of single cells and expressed as fluorescence intensity ratios F/F 0 . The resting fluorescence value F 0 was determined at the beginning of each experiment.
To study the effect of drugs on NAADP ϩ signaling, we applied NAADP ϩ twice, separated by a washout and/or drug preincubation interval of at least 10 min. The response of a given slice during the second NAADP ϩ application in the presence of a drug was then compared with the response during the first "control" NAADP ϩ application with regard to the responding cell population and the amplitude of the Ca 2ϩ responses. All experiments were carried out at room temperature.
Cultured or acutely isolated cells were plated on glass coverslips, loaded with Fluo-4-acetoxymethylester (5 M for 30 min at room temperature in HEPES (150 mM NaCl, 5.4 mM KCl, 1.3 mM CaCl 2 , 0.83 mM MgSO 4 , 10 mM HEPES, 5 mM glucose adjusted to pH 7.4) containing 0.01% Pluronic-127 (Molecular Probes)). Subsequently, cells were washed and kept in bath solution (HEPES) for 15-20 min prior to experiments. All other procedures were identical with the description for brain slices.
Immunohistochemistry-Mice were deeply anesthetized with sodium pentobarbital (100 mg/kg body weight; Sanofi, Paris, France), and perfused intracardially with a solution of 4% paraformaldehyde in 0.1 M phosphate buffer (PB; pH 7.4). Brains were dissected out and postfixed for 24 h at 4°C in the same fixative. Subsequently, they were transferred to a 20% sucrose-containing PB and incubated for several days. Brains were quickly frozen in isopentane cooled by dry ice. Cryosections (16 m thick) were mounted on gelatin-coated slides, allowed to dry for at least 30 min at room temperature, and then stored at Ϫ80°C. After thawing, sections were postfixed for 15 min with 4% PA and then permeabilized with 0.1% Triton X-100 in PB for 15 min and incubated in blocking buffer (0.5% bovine serum albumin, 1% horse serum, 4% normal goat serum, 0.01% Triton X-100 in PB) for 45 min at room temperature. Rabbit anti S-100␤ antibodies (Swant, Bellinzona, Switzerland) were diluted 1:500 (in PB, 1% bovine serum albumin, 1% horse serum) and incubated with the sections for 5 h at room temperature. Sections were washed with PB and then incubated with 1:20 diluted purified rat anti-mouse CD38 monoclonal antibody (Cedarlane, Ontario, Canada) overnight. Sections were again washed with PB followed by incubation with the two secondary antibodies Cy2-conjugated goat anti-rabbit IgG (H ϩ L) (1:100 dilution; Jackson/Dianova, Hamburg, Germany) for ␣S-100 antibody and Alexa Fluor 594 goat anti-rat IgG (H ϩ L) (1:200 dilution; Molecular Probes) for ␣CD38 antibody detection for 5 h at room temperature. After three washes, sections were mounted with CitiFluor (Agar Scientific, Stansted, Essex, UK) and inspected with the confocal microscope described above. Specificity of immunoreactivity was controlled by incubation of tissue sections in dilution buffer instead of primary antibodies. In these control experiments, the immunocytochemical reactions were always negative.
Cells grown on glass coverslips were fixed with 4% paraformaldehyde for 15 min at room temperature, washed with TBS (150 mM NaCl, 100 mM Tris, pH 7.4), permeabilized with 0.05% Triton X-100 in TBS, and incubated in a blocking solution (TBS plus 5% bovine serum albumin and 5% normal goat serum) for 1 h at room temperature. Cells were incubated at 4°C with rabbit anti-glial fibrillary acidic protein antibodies (Sigma) for 1 h. Then the cells were washed with TBS plus 0.5% Tween 20 and incubated with the secondary antibody Alexa-568-conjugated goat anti-rabbit IgG (H ϩ L) (Molecular Probes) for 1 h and washed again. As a control, cells were incubated in buffer without primary antibodies followed by incubation in secondary antibody solution. Cells were mounted by the ProLong Antifade Kit (Molecular Probes).
Solutions M verapamil, purchased from Tocris. All ingredients of buffers were purchased from Sigma, except for K 2 HPO 4 and glucose, which were from Merck.
Statistical Analysis-The population of cells within a slice, which responded during application of NAADP ϩ , was determined for both the "control" administration and the administration in the presence of a drug, and both were compared; the number of cells responding the second time was expressed as percentage of the number of cells responding the first time. It was distinguished between cells that reacted both times or only the second. The results for a given condition obtained from the different slices were averaged.
As a second measure, an activity parameter was evaluated and compared for the first and the second response: the "Ca 2ϩ signaling activity." It was defined as the average change in the F/F 0 value of single cells between two consecutive images (equal to 4.5 s) within a set time frame of about 2 min during an NAADP ϩ application. The Ca 2ϩ signaling activity during the first and second NAADP ϩ application was then compared for cells that reacted both times; each cell's activity during the second NAADP ϩ application was expressed as a percentage of that during the first. Then it was averaged for cells within one slice, and subsequently results from the different slices were averaged for a given condition.
All values are expressed as mean Ϯ S.D. Means and S.D. were calculated with Excel (Microsoft, Silicon Valley, CA). Further statistical analysis of the data was performed with the SPSS software (SPSS Inc., Chicago, IL). The multiple Kruskal-Wallis Test was used to test for significant differences between all groups within one set of groups (four sets; cell numbers and average F/F 0 amplitude change or maximal amplitudes for both the NAADP ϩ and control experiments, respectively) and if it was positive, each group was separately compared with its respective control by the paired Mann-Whitney Test. Significance was defined as p Ͻ 0.05.

NAADP ϩ Induced Different Types of Ca 2ϩ Responses in Astrocytes-
We tested whether extracellularly applied NAADP ϩ induced Ca 2ϩ signaling in cortical astrocytes in acute brain slices of 10 -14-day-old mice. By an appropriate protocol, we preferentially loaded the Ca 2ϩ sensor Fluo-4-acetoxymethylester into astrocytes (31). Superfusion with NAADP ϩ (5 M, duration of application 2-3 min) triggered an increase in Fluo-4 fluorescence, reflecting the change in [Ca 2ϩ ] i in astrocytes (Fig. 1). The time course of the response showed considerable variations; single, transient responses, oscillations, a sustained elevated level, and oscillations followed by an elevated level were observed (n ϭ 74 slices, A cell number per slice ϭ 21 Ϯ 10; Fig. 1

, A (bottom) and B).
Generally, the Ca 2ϩ responses already declined with time of NAADP ϩ application, and we did not observe any specific effect of the washout. Oscillations had an average frequency of 1.9 Ϯ 0.7 transients/min (for analysis: 59 oscillating cells from 21 slices). There was a considerable delay of 64 Ϯ 17 s between the application of NAADP ϩ to the slice and the Ca 2ϩ response (n ϭ 11 slices, A cell number per slice ϭ 18 Ϯ 9). To estimate whether this delay exceeded the time required for the substance to simply reach the astrocytic membranes within the slice, we applied ATP (100 M) to activate cell membrane-located purinergic receptors. This led to a more rapid Ca 2ϩ response with a 20 Ϯ 9-s delay (n ϭ 11 slices, A cell number per slice ϭ 16 Ϯ 7). This difference was also confirmed with experiments in which NAADP ϩ and ATP were applied to the same slice, separated by a 10-min washout. The NAADP ϩ response occurred 32 Ϯ 3 s (n ϭ 5 slices, A cell number per slice ϭ 69 Ϯ 25) later as compared with the ATP response.
NAADP ϩ Repetitively Induced Ca 2ϩ Signaling in Astrocytes-In nonmammalian cells, the NAADP ϩ system desensitizes quickly, NAADP ϩbinding is irreversible, and only one response can be triggered (20,32), whereas for mammalian cells, binding is reversible and desensitization only seems to occur at high concentrations. Therefore, repetitive signaling is possible (24,33,34). To investigate the inactivation properties of astrocytic responses in slices, we applied NAADP ϩ twice with a 10-min interval (Fig. 2, A and B). We then compared the cell populations responding to the first and the second application. Almost all cells (98 Ϯ 3%) that had responded with a Ca 2ϩ increase to the first NAADP ϩ application also responded to the second (n ϭ 7, A cell number per slice ϭ 24 Ϯ 9). Moreover, we found that additional cells responded to the second application, so that the total number of responsive cells increased to 117 Ϯ 37% as compared with the first NAADP ϩ application (Fig. 3D). As a second parameter, we compared the "Ca 2ϩ signaling activity" of the cells that reacted twice for the first and the second response. The Ca 2ϩ signaling activity was defined as the differential F/F 0 amplitude between two consecutive images (sampling rate ϭ 1/4.5 s) averaged over 2 min during NAADP ϩ application (for details, see "Materials and Methods"). The Ca 2ϩ signaling activity was on average higher for the second reaction (136 Ϯ 37%; Fig. 3E). Thus, the induced signaling was comparable for both NAADP ϩ applications, even higher during the second one. The delay from the NAADP ϩ solution reaching the chamber to the onset of the reaction was dependent on the NAADP ϩ concentration, since the reaction to a higher concentration (25 M) occurred faster (11.8 Ϯ 6.4 s earlier; n ϭ 13, A cell number per slice ϭ 35 Ϯ 12).
NAADP ϩ Acted Directly on Astrocytes-To show the direct effect of NAADP ϩ on astrocytes, we used purified astrocyte cultures from mouse brain. In these cells, similar types of Ca 2ϩ responses were triggered by NAADP ϩ (n ϭ 7 coverslips; Fig. 3, A and B). Also in culture, NAADP ϩ -induced signaling could be repetitively elicited (n ϭ 3 cover-  slips; Fig. 3C). In a slice preparation, indirect effects mediated by other cell types cannot be excluded. The response of the astrocytes could therefore result from substances released by other cell types in response to NAADP ϩ . To eliminate at least one factor, we blocked neuronal action potentials by TTX during application of NAADP ϩ . Since there is a high variability of responses to NAADP ϩ between different slices, we compared the Ca 2ϩ responses to NAADP ϩ in the absence (internal control) and presence of TTX in the same area of one slice, separated by a 5-min washout and a 5-min TTX preincubation. TTX did not significantly reduce the number of responding cells (96 Ϯ 19%; n ϭ 5, A cell number per slice ϭ 23 Ϯ 6). Of the cells, which still responded, 90 Ϯ 16% had reacted again, whereas 6 Ϯ 8% reacted in addition (Fig. 3D). However, the average Ca 2ϩ signaling activity was reduced to 78 Ϯ 24% (p Ͻ 0.05; Fig. 3E).
NAADP ϩ -induced Ca 2ϩ Signaling Depended on Functional Connexin Hemichannels/Gap Junctions and Extracellular Ca 2ϩ -We investigated whether connexin hemichannels may serve as an entry route for NAADP ϩ into the cytoplasm. For this reason, we blocked connexin hemichannels (and gap junctions) during the application of NAADP ϩ to observe whether this would reduce the response to NAADP ϩ . After the first NAADP ϩ application, it was washed for 5 min followed by superfusion with the gap junction blocker carbenoxolone (100 M, 5 min). NAADP ϩ was then applied in the presence of carbenoxolone. Only 54 Ϯ 34% of the cells responded in comparison with a previous control application. 47 Ϯ 24% of the cells had reacted already before, and 7 Ϯ 18% reacted additionally (p Ͻ 0.005; n ϭ 12, A cell number per slice ϭ 20 Ϯ 7; Fig. 5A). The average activity was also reduced to 71 Ϯ 38% during blockage of connexin hemichannels (p Ͻ 0.05; n ϭ 6, A cell number per slice ϭ 20 Ϯ 9; Fig. 5B).
Ca 2ϩ Influx Is a Component of NAADP ϩ -induced Ca 2ϩ Signals-To analyze the mechanism of the NAADP ϩ -triggered response, we tried to distinguish between Ca 2ϩ influx and release from internal stores. We therefore compared the effect of NAADP ϩ application in normal and A shows Fluo-4-loaded cultured astrocytes. The first image was taken before and the second image was taken during perfusion with NAADP ϩ (5 M). B, F/F 0 traces are shown from cells marked in A. C, F/F 0 traces from five different astrocytes in culture superfused three times with NAADP ϩ (5 M); cultured astrocytes also respond to repetitive superfusion with NAADP ϩ . D, bar graph on the left compares the number of astrocytes responding with a [Ca 2ϩ ] i increase to two subsequent NAADP ϩ applications separated by 10 min (left). The number of reacting cells from the first application is defined as 100%, and the relative number of cells responding to the second application is given in the graph. The graph shows averages of several experiments, and n indicates the number of brain slices from which data were obtained for this and for the figures to follow. The right bar shows a similar type of experiment, in which TTX (1 M) was present during the second application. TTX was applied 5 min prior to the second NAADP ϩ application. E, bar graph comparing the average amplitude change of the fluorescence intensity of cells during two subsequent NAADP ϩ applications (for details, see "Materials and Methods"). Only those cells were selected that reacted to both NAADP ϩ applications. The left bar corresponds to two control NAADP ϩ applications, and the right bar corresponds to a control followed by a NAADP ϩ application in the presence of TTX. *, p Ͻ 0.05; **, p Ͻ 0.005.
nominally Ca 2ϩ -free bath solution. We first recorded a control response in normal bath solution and then washed out the NAADP ϩ with control solution for 5 min, switched to a nominally Ca 2ϩ -free buffer for 5 min, and applied NAADP ϩ still in the absence of Ca 2ϩ in the bath. The number of cells per slice responding to NAADP ϩ with a Ca 2ϩ increase was reduced to 48 Ϯ 38% as compared with the control. Almost all of   Fig. 3, the populations of reacting cells were compared between two control NAADP ϩ applications (2ϫ NAADP ϩ ), between a control NAADP ϩ application and a NAADP ϩ application combined with carbenoxolone (100 M, 5-min preincubation), nominal Ca 2ϩ free buffer (5-min preincubation), Ca 2ϩ -free buffer containing EGTA (1 mM, 2-3-min preincubation), or the L-type Ca 2ϩ channel blocker verapamil (100 M, 15-min preincubation). B, the average amplitude change of the fluorescence intensity was obtained as described in the legend to Fig. 3. C and D, DHPG (20 M) was applied using the same experimental paradigm as described for NAADP ϩ application in A and B. Cell populations were compared in C. The peak amplitudes of the DHPG responses are given in D. Control DHPG applications were compared with DHPG applications combined with nominal Ca 2ϩ -free buffer (5-min preincubation), and Ca 2ϩ -free buffer containing EGTA (1 mM, 2-3-min preincubation). *, p Ͻ 0.05; **, p Ͻ 0.005. these cells had responded in the control solution (47 Ϯ 37%), and nearly none (1 Ϯ 2%) responded only in the Ca 2ϩ -free solution (p Ͻ 0.005; n ϭ 8, A cell number per slice ϭ 24 Ϯ 10; Fig. 5A). The lack of extracellular Ca 2ϩ also reduced the Ca 2ϩ signaling activity to 61 Ϯ 23% compared with the original reaction (p Ͻ 0.005; Fig. 5B). Furthermore, we only observed transient Ca 2ϩ responses and never an increase to a sustained level.
To test whether residual extracellular Ca 2ϩ might have accounted for the remaining signals, we repeated the experiment with EGTA-containing Ca 2ϩ -free buffer (1 mM, 2-3-min preincubation after a 7-8-min washout with normal bath solution) and found far fewer cells responding to NAADP ϩ (14 Ϯ 17%, p Ͻ 0.05) than in a previous Ca 2ϩ -containing control reaction (n ϭ 4, A cell number per slice ϭ 11 Ϯ 3). The cells that still responded to NAADP ϩ had also responded in the control solution (Fig. 5A). To test whether the Ca 2ϩ -free or EGTA/Ca 2ϩ -free treatment would generally affect the filling state of the intracellular stores, we substituted NAADP ϩ with the metabotropic GluR group I agonist DHPG (20 M), which triggers release from intracellular stores using the same experimental paradigm. In the nominally Ca 2ϩ -free buffer, the average number of cells responding to DHPG was the same as in the presence of Ca 2ϩ (100 Ϯ 3%; n ϭ 3, A cell number per slice ϭ 36 Ϯ 14). 94 Ϯ 8% were the same cells, and 6 Ϯ 7% reacted additionally (Fig.  5C). However, their amplitude was on average reduced to 78 Ϯ 19% (Fig.  5D). In EGTA-containing Ca 2ϩ -free buffer, the number of DHPG-responsive cells was slightly reduced to 92 Ϯ 10% (Fig. 5C). The amplitude of the astrocytes, which still responded, was more strongly affected (57 Ϯ 16% of the internal control, p Ͻ 0.05; n ϭ 4, A cell number per slice ϭ 49 Ϯ 7; Fig. 5D). Two consecutive applications of DHPG in Ca 2ϩ -containing solution resulted in responses of the same population of astrocytes (Fig. 5C), and the average amplitude of the second response was only slightly reduced to 97 Ϯ 14% (n ϭ 5, A cell number per slice ϭ 33 Ϯ 8; Fig. 5D). When comparing responses to DHPG and NAADP ϩ , it is evident that the NAADP ϩ -triggered Ca 2ϩ signaling was more sensitive to the omission of extracellular Ca 2ϩ .
NAADP ϩ -induced Ca 2ϩ Signaling Depended on the Integrity of Intracellular Ca 2ϩ Stores-To test whether intracellular thapsigargin-sensitive stores are involved in the NAADP ϩ -induced Ca 2ϩ responses in astrocytes, a control NAADP ϩ application was compared with an application after a 12-20-min superfusion with thapsigargin (1 M), which leads to a depletion of endoplasmic reticulum Ca 2ϩ stores. This treatment nearly completely abolished Ca 2ϩ responses to NAADP ϩ application (5 Ϯ 11% cells of internal control, p Ͻ 0.005; n ϭ 5, A cell number per slice ϭ 17 Ϯ 9; Fig. 6A). Similarly, blocking IP 3 receptors with 2-aminoethoxydiphenylborane (2-APB; 100 M, 10-min preincubation) reduced the number of cells responding to NAADP ϩ very strongly, namely to 10 Ϯ 15% as compared with a control response in normal bath solution (p Ͻ 0.05; n ϭ 4, A cell number per slice ϭ 12 Ϯ 3; Fig. 6A).
NAADP ϩ is thought to act on receptors located on lysosomes and to release Ca 2ϩ from these organelles. Therefore, we disrupted lysosomes by incubation with GPN, a cathepsin C substrate. Cathepsin C is located exclusively in lysosomes, and the GPN degradation products lead to osmotic swelling followed by disruption of lysosomes. Perfusion with GPN (200 M) itself induced transient Ca 2ϩ signaling in astrocytes in brain slices (n ϭ 4, A cell number per slice ϭ 18 Ϯ 6; Fig. 6B). When NAADP ϩ was applied in the presence of GPN after a 4-min GPN preincubation, the average number of NAADP ϩ -responsive cells was reduced to 25 Ϯ 39% in comparison with the internal control NAADP ϩ application (p Ͻ 0.005; n ϭ 14, A cell number per slice ϭ 26 Ϯ 10; Fig.  6A). To explore a possible interference of GPN with intracellular Ca 2ϩ release from thapsigargin-sensitive stores, we used the metabotropic glutamate receptor type I agonist tACPD (50 M). In comparison with the effect on the response to NAADP ϩ , GPN had a modest effect on the response to tACPD. It reduced the average number of reacting cells to 65 Ϯ 34% (p Ͻ 0.05; n ϭ 4, A cell number per slice ϭ 22 Ϯ 8) as compared with the internal control (Fig. 6C). GPN furthermore reduced the Ca 2ϩ signaling amplitude of cells to 58 Ϯ 20% (p Ͻ 0.05; n ϭ 3, A cell number per slice ϭ 18 Ϯ 3; Fig. 6D). This effect was not due to run down of tACPD-triggered Ca 2ϩ signaling; 100% of cells responded to a second application of tACPD (n ϭ 6, A cell number per slice ϭ 35 Ϯ 10; Fig.  6C), and the amplitude of the second reaction was in average only slightly reduced (94 Ϯ 21% of the first response; Fig. 6D).
Purinergic and Adenosine-mediated Contribution to the NAADP ϩinduced Signal-Extracellularly applied NAADP ϩ could potentially activate purinergic signaling either by degradation products or by direct interaction with purinergic receptors. To address this question, we used two different purinergic receptor blockers (suramin and PPADS); however, in control experiments with ATP, both did not have a strong effect in brain slices. Therefore, we applied another approach to study the interaction between purinergic and NAADP ϩ -induced signaling.
We desensitized purinergic receptors by continuous ATP application (100 M), waited for 3-5 min until the Ca 2ϩ concentration returned to base-line level, and then applied NAADP ϩ in the presence of ATP. In most experiments, NAADP ϩ triggered a Ca 2ϩ response, albeit reduced (n ϭ 29; Fig. 7, B and C). As control, we used tACPD (50 M), whose signaling pathway is at least initially different from that of ATP, and applied it in the presence of ATP. The Ca 2ϩ signal to tACPD in the presence of ATP was similarly reduced as that of NAADP ϩ in comparison with a first control application without ATP (n ϭ 8; Fig. 7, A and C).
Astrocytes and Neurons Expressed the NAADP ϩ -synthesizing Enzyme CD38 in Situ-To gain information about the distribution of the NAADP ϩ -synthesizing enzyme CD38, we used a specific antibody to CD38 and immunolabeled neocortical cryosections (16 m) of 2-week-old mice. To identify astrocytes, we double-labeled them FIGURE 7. Possible contribution of the purinergic pathway. A, two average F/F 0 traces calculated from all responsive cells of one slice, representing the strongest and the weakest reduction of a tACPD-initiated response in the presence of ATP (100 M; 3-5-min preincubation) in comparison with a first control response to tACPD (50 M) are shown. B, as in A, two average F/F 0 traces are shown to demonstrate the most and the least strongly reduced second reaction to NAADP ϩ in the presence of ATP in comparison with a first control reaction to NAADP ϩ . C, the graph shows percentages of maximal F/F 0 amplitudes of average traces in response to tACPD or NAADP ϩ application in the presence of ATP compared with a control condition (equal to 100%). The numbers of slices (n) averaged are given in the figure. with an antibody against the astrocyte specific marker S-100␤ (secondary antibody coupled to Cy2; Fig. 9a). Neurons were identified based on their morphology.
Both neurons and a subset of astrocytes in the neocortex were labeled by the antibody against CD38 (Fig. 9b). In the image overlay (Fig. 9c), the arrows indicate double-labeled astrocytes. The neuronal staining for CD38 was much stronger than that of astrocytes, and the neuronal staining was especially prominent in the perinuclear region. The astrocytic staining was much weaker, more diffuse, and found in processes rather than in the cell body.
However, not all cells we tested react to NAADP ϩ ; the glioma cell line GL261 did not react to NAADP ϩ stimulation (n ϭ 7 coverslips, data not shown); nor did the 1321N1 human astrocytoma cell line (n ϭ 4 coverslips, data not shown).

DISCUSSION
NAADP ϩ Triggers Ca 2ϩ Signaling in All Major Brain Cell Types-Using cell cultures, we have shown that extracellular application of NAADP ϩ is able to induce diverse patterns of Ca 2ϩ responses in all major cell types of the mammalian central nervous system: neurons, astrocytes, oligodendrocytes, and microglial cells. We have used purified cultures, and we therefore can assume that NAADP ϩ acts directly on all these cell types. Cells in culture, however, may have properties distinct from those in situ. We have expanded our work to brain slices with the focus on cortical astrocytes to study the mechanism of action of NAADP ϩ . Since the astrocytic responses in culture are similar to those in situ, we assume that NAADP ϩ directly stimulated the astrocytes in situ as well and that their Ca 2ϩ response is not due to an indirect effect mediated by another cell type. Moreover, we could, at least partially, exclude effects mediated by neuronal activity, since we found comparable astrocytic responses after neuronal action potential propagation had been blocked by tetrodotoxin.
Question of Specificity and Contribution of the Purinergic and Adenosine-mediated Pathway-In contrast to other studies (7), we found that the NAADP ϩ precursors NADP ϩ and NAAD did elicit similar Ca 2ϩ responses as NAADP ϩ itself. Analytical HPLC (data not shown) excluded relevant contaminations for all three substances. One expla-  nation of our observation would be that astrocytes are expressing more active enzymes for NAADP ϩ synthesis than those cell types examined so far, allowing for immediate intracellular conversion to NAADP ϩ . In support of this, we found astrocytes in brain slices expressing the enzyme CD38, which is possibly mainly responsible for synthesis of NAADP ϩ from NADP ϩ (35). Furthermore, extracellular conversion to NAADP ϩ could be taken into consideration in our case, although CD38 is thought to synthesize NAADP ϩ only in acidic milieu (12). Whole rat brain extracts could produce NAADP ϩ from NADP ϩ and nicotinic acid; however, phosphorylation of NAAD could not be demonstrated (9). Other alternatives would include the same (unspecific) extracellular actions as for NAADP ϩ , including degradation processes.
In the continuous presence of ATP, NAADP ϩ can still trigger Ca 2ϩ signals. The response is reduced in comparison with control, but a similar reduction was found for the metabotropic glutamatergic agonist tACPD. This speaks against an action of NAADP ϩ via purinergic receptors. Moreover, GL 261 cells, a cell line with high affinity purinergic receptors (36), did not respond to NAADP ϩ , supporting the hypothesis that NAADP ϩ does not act via purinergic receptors.
We found that adenosine receptor activity is affecting NAADP ϩinduced signaling. NAADP ϩ -induced Ca 2ϩ signals were strongly reduced by the adenosine receptor inhibitor CGS-15943. Therefore, activation of adenosine receptors contributes strongly to the observed signal, whether by direct interaction of NAADP ϩ with adenosine receptors or by its degradation to adenosine and the action of adenosine. However, adenosine receptor-mediated signaling may not directly account for part of the observed signal but only modulate (enhance) it. Adenosine is a well known modulator of several signaling pathways and has, for instance, been shown to modulate purinergic signaling (37) or dinucleotide signaling (38).
NAADP ϩ Potentially Enters the Cell to Act-NAADP ϩ responds with a delay when compared with a plasma membrane receptor-activating ligand (ATP). This again supports the hypothesis that NAADP ϩ enters the cell and acts intracellularly. A similar observation was made for cADPR, the alternate product of the enzyme CD38. When added extracellularly, it elicited Ca 2ϩ signals, and it was suggested to reach the cytosol via the enzyme CD38 itself (15), which can function as a cADPR transporter (39). Moreover, nucleoside transporters were found to mediate cADPR transport across the cell membrane (40). So far, however, NAADP ϩ transporters have not been identified, and there may be no physiological requirement for its transport across the cell membrane, since it is produced in an intracellular acidic milieu. A potential entry route for NAADP ϩ is via connexin hemichannels, which can function as regulated transporters of NAD ϩ and related pyridine dinucleotides (41). Hemichannels also permit the entry of the fluorescent dye Lucifer yellow in low divalent cation solution (42). Our finding that a blocker of hemichannels (and gap junctions), carbenoxolone, reduced NAADP ϩ signaling is consistent with hemichannel-mediated NAADP ϩ entry, yet an additional entry route, which could account for the residual signaling during the block of hemichannels, could be pinocytosis followed by fusion of the NAADP ϩ -containing endosomes with lysosomes, the site of NAADP ϩ action.
NAADP ϩ Signaling Requires Normal Extracellular Ca 2ϩ Levels-We found that omission of Ca 2ϩ from the superfusion solution led to a decrease in the number of NAADP ϩ -responsive cells and their signaling activity. This effect was augmented by further decreasing extracellular Ca 2ϩ levels in the slice by adding EGTA. Under these conditions, only a few cells were able to respond. This indicates that either extracellular Ca 2ϩ levels control the transport of NAADP ϩ into the cell or that Ca 2ϩ influx is required for a full Ca 2ϩ response to NAADP ϩ . Similar findings were made in T-cells (33) or mature starfish oocytes (27), in which the response to NAADP ϩ was strongly reduced or not even detected in Ca 2ϩ -free buffer. In other cell types, Ca 2ϩ influx plays merely a partial role. In human ␤-cells, the initiation of the NAADP ϩ -induced Ca 2ϩ signal was not affected, but the second phase of the response was (34), in contrast to pancreatic acinar cells (24), where washout of Ca 2ϩ after initiation of the response did not prevent the sustained phase, and in contrast to sea urchin eggs, where the removal of extracellular Ca 2ϩ abolished the first Ca 2ϩ rise localized to the plasma membrane but not the global Ca 2ϩ response to NAADP ϩ (43).
In our study, the L-type Ca 2ϩ channel blocker verapamil reduced the response to NAADP ϩ . The involvement of voltage-gated Ca 2ϩ channels was also observed in other studies; the sperm-induced "cortical flash" at fertilization, which requires voltage-gated Ca 2ϩ channels (44), is abolished by desensitization of the NAADP ϩ system (43). Also in mature starfish oocytes, the amplitude of the response to NAADP ϩ could be strongly reduced by the L-type Ca 2ϩ channel blockers verapamil and nifedipine (27). FIGURE 11. Hypothetical mechanism of NAADP ؉ action in astrocytes. This scheme illustrates a model of NAADP ϩ action in cortical astrocytes, which is compatible with our experiments. NAADP ϩ may at least partly enter the cell via connexin hemichannels. Inside the cell, it first acts on receptors associated with lysosomes. This results in the activation of Ca 2ϩ release from thapsigargin-sensitive stores (by IP 3 receptors) and Ca 2ϩ influx (by L-type voltage-gated Ca 2ϩ channels), which together leads to cytosolic Ca 2ϩ increase. This signaling cascade could be modulated by adenosine, a potential degradation product of NAADP ϩ .
Moreover, NAADP ϩ induced a Ca 2ϩ influx, as measured by the patch clamp technique, which could be significantly reduced by L-type Ca 2ϩ channel blockers (25). There has even been speculation about a physical link between lysosomally located NAAPD ϩ receptors and Ca 2ϩ channels in the plasma membrane (45) in analogy to the potential link of IP 3 receptors and capacitative Ca 2ϩ entry channels (46). NAADP ϩ Signaling Requires both Lysosomes and Thapsigargin-sensitive Stores-The integrity and functionality of two intracellular compartments, the thapsigargin-sensitive and the GPN-sensitive ones, were essential for the ability of NAADP ϩ to trigger Ca 2ϩ signaling in astrocytes. GPN specifically destroys lysosomes, which are the suggested site of Ca 2ϩ release by NAADP ϩ and NAADP ϩ receptor localization (23). In addition, NAADP ϩ -triggered Ca 2ϩ signaling crucially required the recruitment of thapsigargin-sensitive stores. The Ca 2ϩ release from the endoplasmic reticulum is mediated by IP 3 receptors, since the Ca 2ϩ signal in response to NAADP ϩ was almost completely eliminated by 2-APB, an IP 3 receptor blocker. Similarly, in starfish oocytes, the Ca 2ϩ response was sensitive to thapsigargin (25), and thapsigargin also abolished the NAADP ϩ -stimulated secretion of insulin in human ␤-cells (34). The fact that no massive Ca 2ϩ release from the endoplasmic reticulum was detected after lysosomal disruption could be explained by the hypothesis that lysosomal activation by NAADP ϩ is a requirement for the subsequent activation of Ca 2ϩ release from thapsigargin-sensitive stores. The link between lysosomal activation and activation of the endoplasmic reticulum could yet be provided by very restricted Ca 2ϩ release from the lysosomes, too low or confined to too small compartments to be detected by our recording system. Perfusion of slices with GPN itself induced Ca 2ϩ responses in astrocytes, demonstrating that these organelles contain Ca 2ϩ or that at least their disruption can trigger Ca 2ϩ responses. Alternatively, the link could involve direct physical interaction of signaling cascade partners localized to lysosomes and to the endoplasmic reticulum.
Hypothetical Mechanism of Extracellularly Applied NAADP ϩ Action-In our experimental setup, extracellular NAADP ϩ -triggered Ca 2ϩ signaling in cortical astrocytes in slices is dependent on 1) connexin conductance, 2) extracellular Ca 2ϩ and Ca 2ϩ channels, 3) the integrity of lysosomal stores, and 4) Ca 2ϩ release from thapsigargin-sensitive stores. At present, we can only speculate how these different systems are interrelated. Our results would be consistent with an intracellular action of NAADP ϩ , which enters into the cytoplasm via connexin hemichannels. However, this presumably applies merely to our simplified way of extracellular administration and may be of no physiological relevance, since NAADP ϩ has so far only been recognized as an intracellular messenger. Moreover, there may be some purinergic but mainly adenosine receptor-mediated contribution. Therefore, the observed signal is most likely a superimposition or interplay of different components. Observed as a whole, activation of receptive sites on or linked to lysosomes seems to be an initial intracellular event prior to Ca 2ϩ release from thapsigarginsensitive stores and Ca 2ϩ channel activation, which are both crucial for the observed Ca 2ϩ response and may function as an amplifier. Under physiological conditions, lysosomes could be furthermore important as the site of NAADP ϩ generation. Our data indicate that Ca 2ϩ release from thapsigargin-sensitive stores and Ca 2ϩ influx by Ca 2ϩ channel activation do not occur independently from each other (Fig. 11).