Localization of the Cyclic ADP-ribose-dependent Calcium Signaling Pathway in Hepatocyte Nucleus*

CD38 is a type II transmembrane glycoprotein found on both hematopoietic and non-hematopoietic cells. It is known for its involvement in the metabolism of cyclic ADP-ribose (cADPR) and nicotinic acid adenine dinucleotide phosphate, two nucleotides with calcium mobilizing activity independent of inositol trisphosphate. It is generally believed that CD38 is an integral protein with ectoenzymatic activities found mainly on the plasma membrane. Here we show that enzymatically active CD38 is present intracellularly on the nuclear envelope of rat hepatocytes. CD38 isolated from rat liver nuclei possessed both ADP-ribosyl cyclase and NADase activity. Immunofluorescence studies on rat liver cryosections and isolated nuclei localized CD38 to the nuclear envelope of hepatocytes. Subcellular localization via immunoelectron microscopy showed that CD38 is located on the inner nuclear envelope. The isolated nuclei sequestered calcium in an ATP-dependent manner. cADPR elicited a rapid calcium release from the loaded nuclei, which was independent of inositol trisphosphate and was inhibited by 8-amino-cADPR, a specific antagonist of cADPR, and ryanodine. However, nicotinic acid adenine dinucleotide phosphate failed to elicit any calcium release from the nuclear calcium stores. The nuclear localization of CD38 shown in this study suggests a novel role of CD38 in intracellular calcium signaling for non-hematopoietic cells.

CD38 is a 42-45-kDa type II transmembrane glycoprotein (1) found in various mammalian tissues and cell types and is capable of converting NAD ϩ into cyclic ADP-ribose (cADPR) 1 (2). This ability is due to the ADP-ribosyl cyclase activity found on the extracellular carboxyl domain of the enzyme. The product, cADPR, possesses calcium mobilizing activity independent of inositol 1,4,5-trisphosphate (IP 3 ) (3). Instead, cADPR seems to regulate calcium mobilization through the calcium-inducedcalcium release mechanism (4,5). In addition to cyclizing NAD ϩ to cADPR, CD38 is also able to use NADP ϩ as a substrate and catalyze the exchange of its nicotinamide group with nicotinic acid to produce NAADP (5), another potent calciummobilizing metabolite.
CD38 is generally believed to be an important surface immunoregulatory molecule; its myriad of possible functions include the induction of B and T cell proliferation (6), regulation of the humoral immune response (7), apoptosis (8), tyrosine phosphorylation of various proteins (9), activation of certain kinases (10), and cytokine release (11). CD38 also displays adhesion properties and might possibly mediate a selectin-type adhesion between different blood populations and human vascular endothelial cells via its putative ligand, CD31 (12). In addition to its immuno-functions, CD38 has been shown to play an important role in insulin secretion via the cADPR-dependent calcium signaling pathway (13). Indeed, results suggest the appearance of autoantibodies to CD38 in patients may contribute to the development of noninsulin-dependent diabetes (14).
The calcium stores mobilized by cADPR and NAADP have been shown to be co-purified with that sensitive to IP 3 , which is believed to be the endoplasmic reticulum (3,15,16). Increasing evidence suggests that the nuclear envelope may also be an important source of Ca 2ϩ stores. Thus, calcium transients generated around the nucleus have been shown to be important in various cellular functions including the regulation of cell division (17), gene transcription (18), and nuclear envelope breakdown (19).
It has been shown previously that nuclear envelope isolated from mouse liver cells is responsive to cADPR (20). An important unresolved question is how an ectoenzyme like CD38 with its catalytic site facing the extracellular environment could transport its product, cADPR, into the cells to exert its calcium mobilizing properties. In this study, we show that CD38 is not exclusively an ectoenzyme but, instead, is also an integral protein of the inner nuclear envelope in hepatocytes. We further demonstrate that cADPR but not NAADP is able to release Ca 2ϩ from the nuclear stores independently of IP 3 . All the components of the Ca 2ϩ -signaling pathway mediated by cADPR are thus present and operational in the hepatocyte nucleus, suggesting an important functional role in the nuclear environment.
Isolation of Nuclear Fraction-The rat liver nuclear membrane fraction was isolated as described (23)(24)(25). Briefly, freshly removed rat liver was homogenized in 6 -8 volumes of a medium containing 1.3 M sucrose, 1.0 mM MgCl 2 , and 10 mM potassium phosphate (pH 7.2). After filtration, the homogenate was centrifuged for 15 min at 1000 ϫ g. The pellet was suspended in a minimum volume of the homogenization buffer with a final sucrose concentration of 2.2 M and centrifuged at 100,000 ϫ g for 1 h. The final nuclear pellet was suspended in a medium containing 0.25 M sucrose, 5.0 mM MgCl 2 , and 20 mM Tris-HCl (pH 7.2) and centrifuged for 10 min at 1,000 ϫ g. The resulting pellet was resuspended at a protein concentration of 10 mg/ml in a medium containing 50 mM Tris-HCl (pH 7.2), 0.3 M sucrose, 150 mM NaCl, 2 mM EGTA, 20% glycerol, 2% Triton X-100, 2 mM phenylmethylsulfonyl fluoride, leupeptin (10 g/ml), aprotinin (10 g/ml), and soybean trypsin inhibitor (50 g/ml). This was allowed to stand for 2 h on ice with gentle stirring and centrifuged for 30 min at 12,000 ϫ g. The supernatant containing solubilized nuclear proteins at the final step was then used as the starting material for its further purification.
Purification of CD38 from the Nuclear Fraction-The chromatography steps followed the method of Khoo and Chang (21,22). Briefly, the solubilized nuclear extract was subjected to a series of column chromatography, in the order of blue Sepharose CL-6B, hydroxyapatite, copper-iminodiacetic acid-agarose and concanavalin (Con) A-Sepharose columns. The extract was applied to a blue Sepharose CL-6B column that has been equilibrated with equilibration buffer (20 mM HEPES (pH 7.2), 0.2 M NaCl and 0.1% Triton X-100). The column was then washed with 2 volumes of washing buffer (20 mM HEPES (pH 7.2), 0.5 M NaCl and 0.1% Triton X-100) and eluted with 2 volumes of elution buffer (20 mM HEPES (pH 7.2), 0.5 M KSCN and 0.1% Triton X-100).
ADP-ribosyl Cyclase and NADase Assays-The ADP-ribosyl cyclase activity of CD38 was followed by the NGD ϩ assay first described by Graeff et al. (26). Samples were incubated at 37°C for various time intervals with 100 M NGD ϩ in 20 mM Tris-HCl (pH 7.2) containing 0.1% Triton X-100. CD38 cyclizes NGD ϩ , a non-fluorescent substrate, to a fluorescent product, cyclic GDP-ribose (cGDPR). The reactions were stopped by adding HCl to a final concentration of 20 mM. The product cGDPR was measured at an excitation and emission wavelength of 300 and 410 nm, respectively, in a LS 50B luminescence spectrophotometer (Perkin-Elmer). The fluorometric assay to measure NAD ϩ -glycohydrolase activity (27,28) was determined in the same manner as for the ADP-ribosyl cyclase assay except that the substrate NGD ϩ was replaced with 1,N 6 -etheno-NAD ϩ .
SDS-PAGE and Immunoblot Analysis-The partially purified enzyme was subjected to 10% (w/v) SDS-PAGE according to Laemmli (29). Immunoblotting was performed following the method of Towbin et al. (30). Briefly, the proteins resolved in the gel were electrophoretically transferred to a 0.2-m nitrocellulose membrane (Bio-Rad). The transferred membrane was treated with TBS (20 mM Tris-HCl, 150 mM NaCl (pH 7.5)) containing 5% (w/v) skim milk and 0.1% Tween 20 for 1 h. The membrane was then incubated overnight at 4°C with the appropriate dilutions of the primary antibody. This was followed with washing and incubation with the secondary antibody conjugated with alkaline phosphatase (Bio-Rad). After subsequent washing, the membrane was developed with 5-bromo-4-chloro-3-indoyl phosphate -toluidine and -nitroblue tetrazolium chloride. Alternatively, for the detection of low levels of protein, a horseradish peroxidase-conjugated secondary antibody was used, and the membrane was developed using the ECL TM system. A prestained SDS-PAGE standard protein marker (Sigma) containing myosin (205 kDa), ␤-galactosidase (116 kDa), albumin (66 kDa), ovalbumin (45 kDa), carbonic anhydrase (29 kDa), trypsin inhibitor (20.1 kDa), ␣-lactalbumin (14.2 kDa), and aprotinin (6.5 kDa) was used to calibrate the molecular mass.
Elution of Enzyme Activity from the SDS-PAGE Gel-The eluate from the final column chromatography step was run in a 10% (w/v) SDS-polyacrylamide gel. The sample buffer used was prepared following the method of Laemmli (29) except with the SDS concentration at 1.6%. The samples were not heated, and the gel was run at 4°C. After the run, the 42-kDa protein band was excised from the gel and transferred to a dialysis tubing containing 0.5 ml of 20 mM Tris-HCl (pH 7.4) with 0.1% Triton X-100 and dialyzed against 4 liters of the same buffer for 48 h at 4°C. The content of the tubing was recovered and concentrated using Centricon 30 (Amicon Inc.). An irrelevant band was excised and a parallel experiment was conducted as control.
Immunohistochemistry-Purified rat liver nuclei were isolated as described above. Before the solubilization step, the final nuclear preparation was then diluted 1:100 in PBS with 2% BSA and added to an optimized mixture of anti-CD38 antibody and goat anti-rabbit Alexa 488-labeled secondary antibody (Molecular Probes). The antibody mixture had been previously incubated for 1 h at room temperature (20°C) before addition to the nuclear fraction. The final suspension was then further incubated at room temperature for 1 h, washed twice with PBS, and counterstained with Hoechst 33342 (Sigma). The suspension was washed again with PBS and mounted with Vectashield mounting medium (Vector Laboratories, Burlingame, CA) on glass slides before analysis using a 40ϫ PlanApo objective on a Leitz Aristoplan Fluorescent Microscope (Leitz Wetzlar, Germany).
For immunofluorescence studies of liver tissues, rats at various stages of development (fetuses from the 20th day of pregnancy, 1 week postnatal rats, and 2-3-month-old adult rats) were perfused transcardially with 0.9% NaCl containing 500 IU of heparin followed by a periodate/lysine/paraformaldehyde fixative according to the method of McLean and Nakane (31). The tissues were further post-fixed for 1 h before infiltration with 30% sucrose in PBS overnight at 4°C. The rat liver tissues were then frozen at Ϫ45°C using isopentane cooled by liquid nitrogen and embedded in Tissue Tek (Miles). Cryostat sections of 5 m were cut and placed on gelatin-coated slides. An alternative procedure whereby the unfixed freshly removed organs were snapfrozen in liquid nitrogen-cooled isopentane before cryosectioning re-sulted in similar staining profiles but with a decrease in autofluorescence level, although the degree of morphological preservation was less than that of the fixed tissue. Sections were incubated with PBS buffer containing 10% non-immune goat serum (Sigma) for 1 h at room temperature. The goat serum had previously been decomplemented by heat inactivation for 30 min at 56°C. The sections were then incubated overnight at 4°C with the appropriate dilutions of the polyclonal antibody against CD38 in 10% goat serum, rinsed with PBS buffer, and incubated for 1 h with Alexa 488-conjugated goat anti-rabbit IgG diluted 1:100 -1:200 in 10% goat serum. For samples whereby the nuclei were highlighted, propidium iodide was used as a counterstain. For the double-labeling experiment, immunostaining was performed with an optimal dilution of both the rabbit polyclonal antibody against rat CD38 and the monoclonal antibody against Na ϩ -K ϩ -ATPase (Affinity Bioreagents Inc.) added concurrently, and the subsequent steps were identical to that of the single-labeling experiment. The immunofluorescence signal was visualized with goat anti-rabbit Alexa 488-conjugated antibody and goat anti-mouse Alexa 546-conjugated antibody. Both Alexaconjugated dyes were obtained from Molecular Probes.
The optimal incubation time and dilution of antibodies, defined as the highest dilution producing maximal staining and minimal background, were determined for all batches of antibodies and conjugates. All experiments were repeated at least twice, and slides were made in duplicates. Controls with preimmune serum, non-immune serum, and omission of the primary antibody were carried out.
Confocal Microscopy-Confocal microscopy was performed using a Carl Zeiss LSM 410 confocal microscope (Carl Zeiss, Germany) equipped with a ϫ 40 Fluar objective (NA ϭ 1.3), ϫ 63 Plan-Apochromatic objective (NA ϭ 1.4), and a ϫ 100 Fluar objective (NA ϭ 1.3), oil immersion. Excitation of the Alexa 488 fluorescent dye was performed at 488 nm, and the emission signal was collected with a BP515-525 emission filter while the Alexa 546 fluorescent dye was excited at 543 nm and the emission signal was collected with an LP590 filter.
Images were stored and processed with Zeiss software. Optical sections collected using the LSCM at 0.5-m intervals were subsequently processed with a graphic user interface that allowed the display of a nucleus of interest by cursor and adjustment of a three-dimensional box. Other three-dimensional reconstructions were produced using the program Velocity2 by Images3 (Salt Lake City, UT). The gray scale images of the red and green channels were processed first using a smoothing algorithm and were added separately into the program, which combined them and created the three-dimensional surfaces using gray level iso-surfacing and mask definitions. Background thresholding was used during the final reconstruction to enhance the three-dimensional images, which were then tilted by about 30°so that the membrane surfaces could be better appreciated.
Immunoelectron Microscopy-The rats were perfused in 0.5% glutaraldehyde and 4% paraformaldehyde before processing for electron microscopy (EM) as described previously (32). Briefly, post-osmicated samples were dehydrated in an ascending series of ethanol and embedded in araldite. Ultrathin sections were cut and mounted on Formvarcoated nickel grids. For immunoelectron microscopy, antigen unmasking was done following the procedure of Stirling and Graff (33). Grids for immunolabeling were incubated in large drops of saturated sodium metaperiodate solution for 1 h at room temperature in a humidified chamber followed by washing in distilled water for 45-60 s. The grids were heated by floating the sections in preheated retrieval medium (0.01 M sodium citrate buffer (pH 6)) maintained at 95-100°C for 10 min. The sections were then cooled in the retrieval medium for 15 min followed by washing in distilled water for 1 min. After antigen unmasking treatment, the grids were first incubated in 0.01 M PBS (pH 7.4) containing 1% immunoglobulin-free BSA for 10 min. Sections were then incubated overnight (18 h) at 4°C with the CD38 primary antibody at a dilution of 1:50 (optimization was previously done) in PBS containing 1% BSA and 1% Tween 20. Negative controls with both the preimmune sera and omission of the primary antibody were performed concurrently as well. The grids were washed by floating them on drops of PBS (3 changes, 5 min each), followed by 5 min incubation in PBS with 1% BSA. Bound antibodies were visualized by incubating the sections with EM-grade goat anti-rabbit immunoglobulin gold conjugate (15 nm) at 1:75 dilution with 1% BSA and 1% Tween 20 for 1 h at room temperature before washing with distilled water (3 changes, 5 min each). Grids were stained with uranyl acetate and lead citrate before viewing in a Philips BioTwin CM 120 transmission electron microscope (Phillips Electron Optics B.V., Eindhoven, Netherlands).
Marker Enzyme Assays-Succinate dehydrogenase activity was measured according to the method of Green et al. (34). Basically, the spectrophotometric change in the absorbance at 600 nm, which accom-panied the enzymatic reduction of 2,6-dichlorophenolindophenol by succinate, was monitored. The molar extinction coefficient was taken to be 16.1 ϫ 10 3 cm Ϫ1 M Ϫ1 .
NADPH-cytochrome c reductase activity was assayed according to the method of Dignam and Strobel (35). This assay is based on the spectrophotometric measurement of the increase in absorbance at 550 nm due to the reductase-catalyzed production of reduced cytochrome c using an extinction coefficient of 21 ϫ 10 3 cm Ϫ1 M Ϫ1 . Na ϩ -K ϩ -ATPase activity was assayed by the release of P i from ATP according to the method of Paul et al. (36). The incubation medium (1 ml final volume) contained 50 mM Tris-HCl (pH 7.4), 5 mM MgCl 2 , 130 mM NaCl, 20 mM KCl, 0.1 mM EGTA, 0.1 mM Tris-ATP with or without 1 mM ouabain at 37°C. Na ϩ -K ϩ -ATPase was defined as the ATPase activity that could be inhibited by ouabain.
Measurement of Ca 2ϩ Release-Ca 2ϩ loading was done by resuspending the purified nuclear fraction in a standard solution containing 125 mM KCl, 2 mM K 2 HPO 4 , 50 mM HEPES (pH 7.0), 4 mM MgCl 2 , 1 mM CaCl 2 and supplemented with an ATP-regenerating system, consisting of 2 mM MgATP, 10 mM phosphocreatine, and 20 units of creatine kinase per ml. The mixture was then incubated for 1 h at room temperature and washed with a buffer containing 125 mM KCl, 2 mM K 2 HPO 4 , 50 mM HEPES (pH 7.0), 4 mM MgCl 2 , 2 mM MgATP, 10 mM phosphocreatine, and 20 units/ml creatine kinase. The final nuclear suspension was adjusted to a protein concentration of 10 mg of protein/ml.
The suspension was then added with Fura-2 at a final concentration of 24 M. The fluorometric measurement of Ca 2ϩ release was done in a thermostatically controlled cell holder, and the nuclear fraction was stirred continuously for the duration of the experiment. Fluorescence was measured at 510 nm at two excitation wavelengths, 340 and 380 nm. The ratio of the fluorescence (R, 340:380) was calculated using the FeliX software in a RatioMaster fluorescence spectrophotometer (Photon Technology International, Brunswick, NJ). The maximum, R max , and minimum, R min , values of the fluorescence ratio were obtained by the addition of 10 M ionomycin and 4 mM of Ca 2ϩ or 4 mM of EGTA, respectively. Standard Ca 2ϩ -EGTA buffers (Molecular Probes) were used for calibration. All experiments were performed at 37°C. Calibration of the fluorescent signal in terms of calcium concentration was performed as described (37).

Isolation of a Purified Nuclear
Fraction-We adopted a well characterized method for purifying the nuclei of rat liver hepatocytes with minimal contamination from other cellular organelles (23)(24)(25). The isolation of nuclei based on the application of a high molarity sucrose concentration in the homogenization medium and followed by higher sucrose concentration in a subsequent centrifugation step has been shown to produce a nuclear preparation with minimal microsomal, mitochondrial, and plasma membrane contamination (23) as well as minimal microsomal nuclear association during the nuclear purification steps (24). Table I shows the results of the marker enzyme assays for the liver nuclear fraction compared with that of the liver homogenate. The activity of NADPH cytochrome c reductase, a microsomal marker enzyme, in the isolated nuclei was less than 5% of the activity in the total homogenate (Table I). Mitochondrial contamination was also minimal as succinate dehydrogenase activity was less than 1% of the total homogenate (Table I). We also observed the absence of any detectable Na ϩ -K ϩ -ATPase in the nuclear fraction, indicating a clean nuclear membrane fraction free from plasma membrane contamination. This is further demonstrated by immunoblot analyses shown in Fig. 1. Again, it can be seen that the nuclear fraction was devoid of Na ϩ -K ϩ -ATPase (Fig. 1A) and prohibitin (Fig. 1B), a mitochondrial marker. The fraction was, however, enriched with nucleoporin p62 (Fig. 1C), a constitutive nuclear protein, thus attesting to the fact that our nuclear fraction has been enriched for nuclear components.
The nuclear fractions were also probed with antibody against three dominant endoplasmic reticulum markers consisting of GRP78 (BiP), calreticulin, and calnexin (Fig. 1, D-F). Their presence in the nuclear fraction could be seen clearly, but the amount was considerably less than the homogenate. This is consistent with the minor presence of the NADPH cytochrome c reductase activity in the nuclear fraction (Table I). Since the nuclear envelope is a continuation of the endoplasmic reticulum, it is likely that a majority of markers found in the microsomal fraction will be found in the nuclear fraction as well (38,39). However, our results indicate that the microsomal contamination is very minimal.
Purification of CD38 from the Nuclear Fraction-The purified nuclei were solubilized and subjected to four sequential column chromatography steps to purify CD38 (Table II). The ADP-ribosyl cyclase activity of the final eluate from the ConA-Sepharose column was determined to be 255 nmol/min/mg protein, ϳ700-fold higher than in the starting extract (Table II). The purified preparation was analyzed by SDS-PAGE and Western blotting under non-reducing and reducing conditions with an anti-CD38 polyclonal antibody (21). Under both conditions, a single 42-45-kDa protein band was detected (Fig. 2). Elution and renaturation of the protein band from SDS-PAGE showed that the protein possessed ADP-ribosyl cyclase and NADase activities (Fig. 3), indicating that the CD38 in the nuclear fraction is enzymatically active.
Immunohistochemical Localization of CD38 to the Nuclei-The presence of CD38 in the hepatocyte nuclei was further demonstrated by immunohistochemical staining. The poly-clonal antibody used has been shown to be specific for rat (21,22) and murine CD38. 2 This was confirmed in controls using CD38 Ϫ/Ϫ mice (strain C57BL/6), which showed no staining when the anti-CD38 antibody was used under similar conditions (data not shown).
Phase-contrast imaging of the isolated nuclear fraction showed the preparation to be free of whole cells (Fig. 4A). Counterstaining with Hoechst 33342 (Fig. 4B) further showed that the preparation consisted mainly of intact nuclei and has minimal DNA fragments. The nuclei were immunolabeled with the anti-CD38 antibody (Fig. 4C), indicating the presence of CD38 in the nuclei.
The nuclear localization of CD38 was confirmed in both fixed and unfixed cryostat sections of rat liver. Confocal optical sections showed that Ͼ90% of the hepatocytes were immunopositive for CD38 (Fig. 5A). In Fig. 5B, the hepatocyte nuclei were counterstained with propidium iodide to show that CD38 was confined mainly to the periphery of the nucleus, indicating its localization on the nuclear envelope or regions of the endoplasmic reticulum adjacent to it. Very low or no immunoreactivity was observed on the plasma membrane. A low level of granular-patterned autofluorescence could be detected in the hepatocytes, which was also observed in unlabeled control slides (data not shown). An individual nucleus in a series of confocal optical sections was graphically selected by a three-dimensional box and digitally analyzed to show that CD38 was primarily located on the nuclear envelope but not in the nucleoplasm (Fig. 5C). A stereo image constructed from the same series of optical sections clearly showed that CD38 was localized exclusively to the nuclear envelope (Fig. 5D).
Electron microscopic examination of the rat hepatocyte showed that the colloidal gold immunoreactivity was localized to the nucleoplasmic side of the inner nuclear envelope (Fig. 6,  A and B). Since the primary antibody was raised against a polypeptide of 87 amino acid residues consisting of the carboxyl-terminal region of rat CD38 (21,22), it is likely that the enzymatic site of CD38 that is located in the carboxyl-terminal region is facing the nucleoplasm whereby it would have access to NAD ϩ located therein. No immunoreactivity was observed on the outer nuclear envelope or the endoplasmic reticulum regions adjacent to it. Negative CD38 immunoreactivity observed with incubation in either the preimmune sera or omission of the primary antibody in the incubation buffer (Fig. 6C) confirmed the specificity of our immunoelectron microscopy results. The results here further corroborate our immunofluorescence data showing that CD38 is localized exclusively to the nuclear envelope.
Developmental Expression of CD38 -In 2-3-month-old rats, expression of CD38 in the nuclei of the hepatocytes was consistently seen (Fig. 7C). However, in fetuses from the 20th day of pregnancy there was minimal or no expression of CD38 in the hepatocytes (Fig. 7A). In the 1-week-old postnatal rats 2 C. F. Chang, unpublished observations.

TABLE I
Marker enzyme activities in the homogenate and the purified nuclear fraction Numbers in parentheses represent the percentage. Na 1 -K ϩ -ATPase activity was not detected (N/D) in the nuclear fraction. These data are means of two independent preparations with replicates varying by Ͻ10%. ( Fig. 7B), not all the nuclei were immunopositive for CD38, and the staining intensity was less compared with the adult rats (Fig. 7C). Again, no staining was observed on the plasma membrane or in the nucleoplasm. This can be clearly seen from Fig.  7D whereby the sections were double-labeled with antibodies against CD38 and Na ϩ -K ϩ -ATPase. Clear demarcation with regard to the spatial distribution of the two enzymes was seen. In comparing Fig. 7, A and C, we could observe a relative increase in the autofluorescence of the liver tissue, which could be due to a developmental increase in levels of autofluorescent substances including that of NADH, flavins, and lipofuscin. In the 1-week-old postnatal rats (Fig. 7B), we could also observe several sinusoid lining cells, in addition to the hepatocytes, showing immunopositivity toward CD38. The immunopositive signal for sinusoid lining cells was also observed in 2-3-monthold adult rats (data not shown) but was not seen in fetal rats. Our result is in agreement with a previous report showing the absence of CD38 expression in human fetal organs such as the kidney, liver, lung, and lymphoid tissue (40). These results indicate that the expression of CD38 in liver is developmentally regulated. It has been suggested that the marked differences of developmentally regulated expression of a number of hepatocytic protein markers that occur in fetal and adult hepatocytes might be responsible for the immaturity of some liver functions in the neonate (41). A previous study showing the ineffectiveness of cADPR in mobilizing calcium from internal stores in digitonin-permeabilized chick embryo retina cells (42) gives credence to the possibility that CD38 might have a role in the development process. Ca2 2ϩ Release from Nuclear Stores-Liver nuclei were first loaded with Ca 2ϩ as described under "Experimental Procedures," which was found to be ATP-dependent (20, 43). Consistent with that reported by Gerasimenko et al. (20), the nuclei responded rapidly to cADPR (Fig. 8A). Subsequent addition of IP 3 , but not ryanodine, elicited a smaller calcium response. Ryanodine added prior to cADPR induced rapid Ca 2ϩ release and blocked cADPR from releasing more Ca 2ϩ (Fig. 8B). A further addition of IP 3 again managed to elicit a smaller calcium response. These results are consistent with cADPR and ryanodine acting on the same target, which is distinct from the target of IP 3 . This is further demonstrated in Fig. 7C. The primary addition of IP 3 elicited a Ca 2ϩ response but failed to block a larger Ca 2ϩ release activated by subsequent addition of ryanodine. The nuclei became refractory after the addition of ryanodine and failed to respond to cADPR (Fig. 8C). A normal response was seen, however, if cADPR was added before ryanodine, even in the presence of IP 3 (data not shown). In all the experiments, the amount of Ca 2ϩ released by either 10 M cADPR or 5 M ryanodine was consistently measured to be approximately twice that released by 5 M IP 3 (Fig. 8, A-C).
The dose-response experiment of cADPR-dependent Ca 2ϩ mobilization showed that the addition of 2 M cADPR and 10 M cADPR resulted in a graded increase of Ca 2ϩ release from TABLE II Purification of CD38/ADP-ribosyl cyclase from the rat liver nuclear fraction CD38/ADP-ribosyl cyclase activity of each fraction was determined by measuring the formation of the fluorescent product cGDPR from the substrate NGD ϩ . The blue Sepharose eluate containing KSCN was dialyzed overnight in 10 mM Tris-HCl (pH 7.2) containing 0.1% Triton X-100 before cyclase activity was measured, because KSCN interferes with the fluorescent determination of the cGDPR formed. IDA indicates iminodiacetic acid-agarose.  the nuclear stores (Fig. 8D). 8-NH 2 -cADPR has been shown to be a potent antagonist of cADPR (44), and Fig. 8E shows that prior addition of 100 M 8-NH 2 -cADPR effectively blocked the subsequent release of calcium by cADPR. Addition of ADPribose as a control does not result in Ca 2ϩ mobilization from the nuclear stores (data not shown). However, unlike cADPR, NAADP has no effect on Ca 2ϩ mobilization from the nuclear envelope stores (Fig. 8, F and G). In Fig. 8F, the addition of cADPR caused a Ca 2ϩ response, but further addition of NAADP (100 nM to 100 M) in the final concentration failed to elicit any further response. Conversely, the primary addition of NAADP failed to elicit any Ca 2ϩ mobilization from the nuclear Ca 2ϩ stores, but the subsequent addition of cADPR again caused the Ca 2ϩ mobilization response as observed previously (Fig. 8G).
Addition of 1 mM NAD ϩ resulted in a slow and steady release of Ca 2ϩ from the nuclear stores (Fig. 8H). This result suggests that cADPR could be produced from NAD ϩ in the nucleus, and the cADPR, in turn, was able to act on the cADPR-sensitive Ca 2ϩ stores. Taken as a whole, our results indicate the existence of a cADPR-sensitive Ca 2ϩ release mechanism in the nuclear envelope of rat hepatocytes. The sensitivity of the mechanism to ryanodine suggests the existence of ryanodinesensitive Ca 2ϩ release channels, which could be regulated by cADPR in the nuclear envelope of rat hepatocytes. DISCUSSION CD38 is a lymphocyte surface antigen originally defined by monoclonal antibody typing (45). Its expression in lymphocytes is stage-specific and has thus been widely used as a marker for lymphocyte differentiation (46,47). Subsequent studies (5,48) demonstrate that it is a novel multifunctional ectoenzyme capable of not only synthesizing cADPR but also hydrolyzing it. The ecto location of CD38, however, raises an unresolved issue of how the product, cADPR, synthesized by the ecto-CD38 can exert its calcium mobilizing activity intracellularly (49).
One model that has been proposed to answer this conundrum involved the internalization of surface CD38, which has been shown to occur upon the addition of various stimuli including that of NAD ϩ (50) and NADP ϩ (51). However, this mechanism would be a relatively slow and inefficient method for causing calcium release from intracellular cADPR-sensitive stores. It is possible that internalization represents an alternative mechanism of intracellular signaling unrelated to its enzymatic properties and the Ca 2ϩ releasing properties of cADPR. The study by Funaro et al. (52) where they showed that the internalization step is likely to be a negative feedback control mechanism that interrupts signal transduction processes mediated by the surface membrane CD38 gives credence to the above-mentioned possibility.
Recently, it was postulated that transmembrane juxtaposition of two or four CD38 monomers is able to generate a catalytically active channel for selective formation and influx of cADPR to reach cADPR-responsive intracellular calcium stores (53). However, da Silva et al. (54) have shown that there was no direct involvement of ectocellular synthesis of cADPR on the regulation of the cADPR-mediated intracellular Ca 2ϩ signaling in T-lymphocytes. In addition, the same study has shown that there was no increase of intracellular cADPR when the intact cells were incubated with NAD ϩ .
The results described in this study provide a possible resolution to this topological issue. Here we show that CD38 can also be an intracellular enzyme. Its localization in the nucleus of hepatocytes represents the first demonstration of selective association of CD38 with an intracellular organelle. Over 20 years ago, Tamulevicius et al. (55) and Fukushima et al. (56) had independently shown the presence of an unidentified NA-Dase in the nuclear envelope of mouse and rat hepatocytes, respectively. The fact that we identified this NADase in the nuclear envelope as likely to be CD38 suggests that the subcellular distribution of this molecule is more complex than originally thought.
In addition to CD38, we also demonstrated that the isolated liver nuclei also possess the cADPR-sensitive calcium release mechanism. Therefore, all the components of the Ca 2ϩ -signaling pathway mediated by cADPR are present and functional in the nucleus of the hepatocytes. Our results are in agreement with those reported by Gerasimenko et al. (20) but contradict those by Lilly and Gollan (57), which suggested that the hepatocytes are not responsive to cADPR. The present study and that by Gerasimenko et al. (20), both focused on the hepatocyte nuclei, whereas Lilly and Gollan (57) used mainly liver microsomes. In view of the selective localization of the cADPR system in the nucleus, the negative result observed with the micro- somes indicates the possibility of a cADPR-sensitive calcium pool located exclusively in the nuclei.
Another important aspect of nuclear Ca 2ϩ signaling that might be regulated by CD38 is the fact that, apart from cADPR, it also produces NAADP that has previously been shown to possess Ca 2ϩ mobilizing properties as well (5). Recently, it was shown that NAADP is able to mobilize calcium from pancreatic acinar cells and rat brain microsomes in an IP 3 -and cADPRindependent manner (16,58). However, our results show that NAADP is not able to mobilize Ca 2ϩ from rat hepatocyte nuclear calcium stores. This observation suggests the possibility that both the nuclear and endoplasmic reticulum calcium stores are physiologically regulated in a different manner. However, we also cannot rule out that different tissue or cell types might have different Ca 2ϩ -signaling pathways in play that are not only regulated in a spatial but in a temporal manner as well.
CD38 has been shown to be present on the surface of many cell types (5,48). Indeed, we have found that the plasma membrane fraction isolated from hepatocytes contained abundant GDP-ribosyl cyclase activity (63). Upon using a different polyclonal antibody against CD38, we observed high immunoreactivity on the sinusoidal domain of rat hepatocytes but, surprisingly, the same antibody did not cross-react with the nuclear CD38 (63). It is possible that a different isoform or post-translationally modified form of CD38 is present on the nuclear envelope of the rat hepatocyte as compared with the plasma membrane CD38. This is not without precedence. For example, multiple IP 3 receptor isoforms have been shown to be present both on the plasma membrane and internal membranes (64,65).
It is also interesting to note that from the data of our purification steps (Table II) and the size of the protein calculated from the result of the immunoblot (Fig. 2), the nuclear CD38 appears to be glycosylated as it apparently binds to ConA-Sepharose, whereas its molecular weight is similar to that of glycosylated surface membrane CD38. ConA is a lectin, which binds reversibly to molecules with ␣-D-mannopyranosyl, ␣-D- Unfixed snap-frozen rat liver sections were immunostained with anti-CD38. A, CD38 is shown to be localized to the nuclear region. No signal was observed on the plasma membrane. Scale bar, 50 m. B, a liver section was counterstained with propidium iodide to reveal the nuclei. Scale bar, 25 m. C, an individual nucleus (indicated by the cross-hair) in a series of 0.5-m confocal optical sections was graphically selected by a three-dimensional box. Orthogonal representation of the same nucleus is shown. The nuclear envelope could be seen to be immunopositive for CD38, but no signal was observed in the nucleoplasm. Scale bar, 10 m. D, a stereo image constructed from a series of optical sections obtained via confocal microscopy. CD38 (green signal) is clearly localized on the nuclear envelope that surrounded the nucleoplasm stained by propidium iodide (red signal).
glucopyranosyl, and sterically related residues. Our results indicate that nuclear CD38 contains similar glycoconjugates that are able to bind ConA. The presence of glycoproteins localized to the nuclear envelope is not surprising as Willemer et al. (66) have previously shown that the nuclear envelope of rat acinar cells was selectively labeled for ConA, thereby indicating the presence of specific glycoproteins with an affinity for ConA in the nuclear envelope.
The fact that CD38 is transported to the surface membrane with its carboxyl-terminal domain containing the enzymatic site facing the extracellular region suggests that the transport to the nuclear envelope would result in the luminal localization of the CD38 enzymatic site. However, our immunoelectron microscopy studies allowed us to identify the localization of CD38 to the inner nuclear envelope with its catalytic site facing the nucleoplasm. Currently, the precise mechanism of protein transport to the inner nuclear envelope is not known. Compounded with the fact that the outer membrane of the nuclear envelope has different characteristics including different protein composition from the inner nuclear membrane, it is possible that the pathway for the distribution of CD38 to the nucleus is distinct from that of the pathway to the surface membrane.
Our results indicate that the Ca 2ϩ release mechanism is probably located on the inner nuclear membrane. This would allow the release of calcium into the nucleus where it can act directly to activate Ca 2ϩ -sensitive nuclear events. In fact, this appears to be the case in starfish oocyte where it was shown that the introduction of cADPR directly into the nucleus is able to activate calcium changes in the nucleoplasm (67). Furthermore, Gerasimenko et al. (20) have shown that both IP 3 and cADPR can cause the release of calcium from the nuclear envelope into the nucleoplasm where it can also diffuse out through the nuclear pores.
NAD-pyrophosphorylase, an enzyme that catalyzes NAD ϩ formation from nicotinamide mononucleotide and ATP, is found predominantly with an intranuclear localization (68). Thus, NAD ϩ produced locally could be a source of substrate for the ADP-ribosyl cyclase activity of CD38. In any case, there should be a mechanism to regulate the cADPR production. It is likely that there is a mechanism whereby CD38 is inactive until stimuli trigger it to an active state, and cADPR is produced. This is supported by the fact that a previous study has shown that the NADase in the nuclear envelope is present in a latent form (55). In addition, Sato et al. (69) have recently discovered a novel peptide inhibitor to human BST-1/CD157. This cADPR-synthesizing enzyme and CD38 are believed to have evolved from a common ancestor by gene duplication (70). It is not inconceivable that there exists similar inhibitor peptides or proteins to regulate the intracellular activity of CD38 as well.
It is generally believed that small molecules such as cADPR or its precursor, NAD ϩ , can freely diffuse through the nuclear pore complex due to the fact that the nuclear pore complex supposedly allows the permeability of the nuclear envelope to all molecules with masses as large as 30 -60 kDa (71). In addition, Oishi and Yamaguchi (72) have shown that liver cytosolic NAD ϩ is a factor in the regulation of the nuclear Ca 2ϩ concentration. Thus, the NAD ϩ present either in the cytosol or in the nucleus should be accessible to the CD38-mediated Ca 2ϩ release mechanism in the inner nuclear membrane.
Another point of interest is the localization of the calcium channels regulated by cADPR. Although cADPR has been shown to be an activator of ryanodine receptors, curiously until now, there have been conflicting reports regarding the nature of ryanodine receptors in the liver. The presence of a ryanodinesensitive Ca 2ϩ -induced Ca 2ϩ release pool in the hepatocyte cell has been shown in the study by Osada et al. (73), and another study by Martinez-Merlos et al. (74) has shown via a [ 3 H]ryanodine binding assay that the rat liver contained high levels of ryanodine-binding sites when compared with the liver of five other rodent species. The physiological importance of ryano-FIG. 6. Ultrastructural localization of CD38 to the nuclear envelope in rat hepatocyte nucleus. Ultrathin sections after antigen unmasking were immunostained with anti-CD38 primary antibody and secondary antibody conjugated with 15-nm gold particles. A and B, CD38 was localized to the inner membrane (indicated by arrows) of the nuclear envelope. C, control section with no gold particles observed. C and N denote the cytoplasm and nucleus, respectively. Scale bar, 100 nm.

FIG. 7. CD38 in the rat liver at different stages of development.
Confocal images of fixed liver sections were obtained by exciting at 488 nm and collecting from two channels, with the green filter (BP515-525) to obtain the Alexa 488 signal (anti-CD38) and the red filter (LP590) to obtain the autofluorescence in order to highlight the liver morphology. Liver section of fetal rat at the 20th day of pregnancy shows that the hepatocytes did not express CD38 (A), whereas the liver section of 1-week postnatal rat shows that CD38 was expressed in the nuclear region of the hepatocytes (B). In addition, several sinusoid lining cells (white arrows) were also immunopositive for CD38. C, liver section of a 3-month-old adult rat shows an increase in staining intensity for CD38 in the nuclear region. An increase in the autofluorescence of the hepatocytes is also seen. D, liver sections were doubly labeled with anti-CD38 (green) and anti-Na ϩ -K ϩ -ATPase (red). The figure on the left shows a single optical section. The figure on the right is a representation of a three-dimensional reconstruction based on a series of optical sections. The level of autofluorescence was minimized by lowering the laser intensity and using unfixed snap-frozen liver. A-C, scale bar, 50 m. D, scale bar, 10 m. dine receptors in the liver is clearly shown in the work of Komazaki et al. (75) whereby they observed that mutant mice lacking both RyR1 and RyR3 showed hypertrophy of the liver and an excessive accumulation of glycogen granules in the hepatic cells.
In contrast, Giannini et al. (76) have reported the absence of messenger RNAs encoding the three known RyR isoforms in liver extracts, and Guihard et al. (77) reported that the liver nuclei does not possess ryanodine receptors. In our study, we also failed to detect the presence of ryanodine receptors in the rat hepatocyte using two different antibodies against all the known isoforms of ryanodine receptors (data not shown). Immunostaining and Western blot with the anti-ryanodine receptor antibodies failed to yield any positive result while immunoreactivity was observed in the positive control tissues including the heart and skeletal muscle (data not shown). Our observation raised three possible corollaries. One possibility is that there exists a novel isoform of RyR in liver cells, which has yet to be characterized. On the other hand, it is possible that ryanodine and, in turn, cADPR might be mobilizing Ca 2ϩ release from the nuclear stores in a novel manner other than is usually associated with the Ca 2ϩ -induced Ca 2ϩ release response mediated by both these metabolites. The fact that ryanodine receptors are high conductance Ca 2ϩ channels with a higher conductivity than IP 3 receptors (78) also gives rise to a third corollary that implies the possible presence of ryanodine receptors at low densities, which are below the limit of our detection system.
Based on our results, we proposed a model whereby the presence of functional CD38 localized to the inner nuclear envelope is playing a central role in regulating intracellular calcium signaling via its ability to form the potent Ca 2ϩ -mobilizing nucleotide, cADPR (Fig. 9). The diagram illustrates the presence of the two known Ca 2ϩ -signaling pathways on the inner nuclear envelope as indicated by past studies (20,24) as well as by the present study that IP 3 and cADPR can both cause the transient rise of calcium concentration in the nucleus.
Our study did not allow us to reach a definite conclusion with regard to the precise localization of the cADPR-sensitive Ca 2ϩ channels. Although we cannot exclude the possibility that the cADPR-sensitive Ca 2ϩ channels are located on the outer nuclear membrane, it is more likely that they are situated in the inner nuclear membrane where they can be directly acted upon by cADPR produced by CD38 in the nuclear environment adjacent to the inner nuclear membrane. Indeed, as a precedent, the IP 3 -sensitive Ca 2ϩ channels have been previously shown to have an inner nuclear membrane localization as well (23). In any case, the specific localization of both CD38 and the cADPRdependent Ca 2ϩ -signaling pathway in the hepatocyte nucleus suggests an important involvement with the various biological and genetic events occurring in the nucleus.
While this paper was under consideration for publication, Adebanjo et al. (79) reported the presence of functionally active CD38 in the inner nuclear envelope of MC3T3.E1 cells. In a parallel finding, they also showed that the nuclear CD38 has its catalytic site within the nucleoplasm. However, unlike our study with hepatocytes, they reported the presence of ryanodine receptors in the nucleus of MC3T3.E1 cells. This disparity in observation could be due to cell-specific differences as the osteoblast-like MC3T3.E1 cells are distinct from hepatocytes. FIG. 9. Schematic diagram of the cADPR-dependent Ca 2؉ -signaling pathway present in the nucleus according to the results presented in this paper. The IP 3 -dependent Ca 2ϩsignaling pathway known to be present in the nucleus (23) is also illustrated to highlight the fact that multiple Ca 2ϩ -signaling pathways can be found in the nuclear region. The possibility that cADPR might be diffusing through the nuclear pores to act on distant target sites is also shown in the diagram.