Identification and Characterization of a Novel Hepatic Canalicular ATP Diphosphohydrolase*

We have identified and characterized a novel ATP diphosphohydrolase (ATPDase) with features of E-type ATPases from porcine liver. Immunoblotting with a specific monoclonal antibody to this ectoenzyme revealed high expression in liver with lesser amounts in kidney and duodenum. This ATPDase was localized by immu-nohistochemistry to the bile canalicular domain of hepatocytes and to the luminal side of the renal ductu-lar epithelium. In contrast, ATPDase/cd39 was detected in vascular endothelium and smooth muscle in these organs. We purified the putative ATPDase from liver by immunoaffinity techniques and obtained a heavily gly-cosylated protein with a molecular mass estimated at 75 kDa. This enzyme hydrolyzed all tri- and diphosphonucleosides but not AMP or diadenosine polyphos-phates. There was an absolute requirement for divalent cations (Ca 2 1 > Mg 2 1 ). Biochemical activity was unaf-fected by sodium azide or other inhibitors of ATPases. Kinetic parameters derived from purified preparations of hepatic ATPDase indicated V max of 8.5 units/mg of protein with apparent K m of 100 m M for both ATP or ADP as substrates. This was for the evaluation of the kinetic parameters of the purified hATPDase. To determine specific activities, the protein of measured using protein assay (43). pooled fractions, obtained after immunoaffinity chromatography, containing 20 m g of protein, were loaded on a 10% SDS-polyacrylamide gel. After electrophoresis under reducing conditions, proteins were transferred to a ProBlott polyvinylidene difluoride membrane. Prior to NH 2 -terminal sequencing, the ProBlott membrane was subjected to three washing cycles with 50% HPLC grade methanol and HPLC grade water and then the band of interest excised. NH 2 -terminal sequencing carried out on a Procise 492 sequenator (Perkin-Elmer/Applied Biosystems), operating in a pulsed-liquid

The biological function of these enzymes may relate to the regulation of extracellular concentrations of nucleotides (e.g. ATP, ADP, UTP) and nucleosides (e.g. adenosine) (2,3,6). Extracellular nucleotides are known to be released by distinct mechanisms such as exocytosis from platelets (7). They may also originate from damaged or dead cells as well as from apparently uninjured cells under such physiological stimuli as shear stress and hypoxia (6,7). ATP, ADP, adenosine, and UTP act as extracellular signaling substances in essentially all tissues (8,9). For example, extracellular ATP and ADP are involved in neurotransmission, nociception, control of secretion from a variety of endocrine and exocrine glands, smooth muscle tone, and vascular hemostasis (8). Such nucleotides and nucleosides act via purinoceptors. Moreover, the action triggered by ATP on P2-receptors is often opposite of the action triggered by the dephosphorylated derivative on adenosine receptors (10). Ectonucleotidases play a critical modulatory role in these signaling pathways. For example, in cd39-deficient mice, a reversible form of platelet P2Y1 receptor desensitization can be observed. We have proposed that ATPDase/cd39 has a role in regulating thrombotic reactions induced at sites of vessel inflammation (11).
ATP is also an indispensable intracellular molecule for all cells, and the liver appears to be the major source of purines for several extrahepatic tissues, such as brain and muscle, that are incapable of de novo synthesis (12)(13)(14)(15). Ectonucleotidases at the hepatic sinusoid and bile canalicular membrane may therefore play a key role in the critical regulation of nucleotide and nucleoside trafficking and turnover in the liver, in certain extrahepatic tissues and in bile or plasma (15).
Over 4 decades ago, functional ATPases were shown to be associated with bile canalicular plasma membranes by histochemical techniques (16). This activity was subsequently shown to be distinct from the classical ATPase pumps (15). However, the relationship of putative canalicular ecto-ATPases and their functional identification as ATPDases remains unclear. The canalicular ecto-ATPase was previously considered to be identical to cCAM105 and closely related proteins like pp120/HA4 and gp110 (17,18). However, cCAM has low ATPase biochemical activities and was later identified as a cell adhesion molecule related to carcinoembryonic antigen rather than an ecto-ATPase (19). Other types of ectonucleotidases are also present in hepatocyte membranes. For example, the ectophosphodiesterase family member PC-1 is expressed on the basolateral membrane of hepatocytes, while a closely related protein termed B10 is predominantly canalicular in distribution (20,21).
Recently, CD39 was identified as a mammalian ATPDase (22)(23)(24). The cloning of CD39 was followed by the identification of a series of related genes denoted as CD39L1 to L4 (25)(26)(27)(28) and of a UDPase gene (29). Northern blot analysis of various human tissues with the four cloned members of the CD39-like family revealed that two of these genes were expressed in liver, namely CD39L2 and CD39L4 (27). CD39L4 has been recently characterized as a soluble ATPDase expressed by macrophages (30), while CD39L2 has not yet been studied at the protein/ biochemical level. Northern blot and immunoblot show that the prototype CD39 is expressed in liver, but at low levels when compared with other tissues (11,23,31). CD39L1, a preferential ATPase (26), has been shown to be expressed in mouse hepatoma cells induced by the environmental contaminant dioxin (32). CD39L3 is an ATPDase originally found in the brain (28), while the UDPase is localized to the Golgi apparatus (29). To date, none of these CD39 family members have been localized to the hepatic canaliculus. The goal of this study was to identify the source of this as yet uncharacterized ATPDase activity in liver and to examine the putative enzyme functional properties and distribution.
Preparation and Solubilization of Microsomal Pellets-Plasma membrane-enriched particulate fractions were prepared as described previously (33). Pig liver particulate fractions used for either capture assay or purification of the hepatic ATPDase (hATPDase) were prepared using a slightly different procedure as described previously. 2 Particulate fractions were suspended in 25 mM Tris-HCl (pH 7.6) to about 5 mg of protein/ml and sonicated (MSE Instruments, Leicester, United Kingdom) at 75% of maximal intensity. An equal volume of the same buffer, containing 0.6% Triton X-100 was added and the mixture was centrifuged for 1 h at 100,000 ϫ g. The pellet was discarded, and the supernatant containing the solubilized membrane proteins was used for further purification.
N-Glycosidase F Treatment-Proteins were deglycosylated with peptide N-glycosidase F (PNGase F), as described previously (35). Samples were analyzed by Western blotting as detailed below.
Anti-ATPDase Monoclonal Antibodies-Partially purified nucleotidase fractions from porcine kidney cortex were injected into mice. 2 Hybridomas and monoclonal antibodies were prepared as described previ-ously. 2 Hybridomas secreting monoclonal antibody to nucleotidases were selected using an ATPDase capture assay (36).
Immunoblotting Procedures-Proteins were fractionated by SDSpolyacrylamide gel electrophoresis (PAGE) according to Laemmli (37). Protein samples were boiled in sample buffer (2% (w/v) SDS, 10% (v/v) glycerin, 0.001% bromphenol blue in 65 mM Tris, pH 6.8) under nonreducing conditions. The proteins were separated on a 10% acrylamide SDS-gel and transferred to Immobilon-P membrane by semidry electroblotting (38). After incubation with mouse anti-ATPDase monoclonal antibodies, the bands were visualized using horseradish peroxidaseconjugated goat anti-mouse IgG, at a dilution of 1:4,000, and the Renaissance Chemiluminescence Reagent Plus, according to the manufacturer's instructions.
Immunohistochemistry-Liver, kidney, and other tissues were embedded in freezing medium, snap-frozen in isopentane in liquid nitrogen, and stored at Ϫ80°C until used. Sections of 5 m were obtained, air-dried, fixed in acetone, and stained using murine monoclonal antibodies to pig hATPDase (3A1) or ATPDase/cd39 (4D8 or 4D11), as described previously (39). Briefly, sections were incubated overnight at 4°C with primary antibodies, washed in phosphate-buffered saline, and blocked with 5% normal horse serum. Primary antibodies were detected using biotin-labeled horse anti-mouse secondary antibodies. Endogenous peroxidase activity was previously blocked by an incubation of 9 min in 0.03% hydrogen peroxide in methanol. After several washes with phosphate-buffered saline, the sections were stained with horseradish peroxidase complex as specified by manufacturer's recommendations. Peroxidase activity was revealed using 3,3Ј-diaminobenzidine as a substrate, as specified by the manufacturer's instructions. Isotype-matched monoclonal antibodies (mouse IgG1) and a control for residual endogenous peroxidase activity were included in each experiment.
Chromatography Column Purification of hATPDase-Typically, 160 mg of solubilized hepatic membrane proteins were incubated with 5 ml of 1/1 (v/v) Sepharose 4B in Tris buffer (25 mM Tris-HCl, pH 7.6) for 1 h to exclude nonspecific interactions. The Sepharose was removed by passage through a 0.22-m membrane filter (Gelman), and the filtrate was applied to an immunoaffinity column prepared with purified monoclonal anti-hATPDase antibody (3A1) conjugated to CNBr-activated Sepharose 4B (Amersham Pharmacia Biotech). The column was washed consecutively with Tris buffer, the same buffer containing 0.5 M NaCl and 0.3% Triton X-100, Tris buffer with 1 M LiCl and 0.3% Triton X-100 and finally H 2 O, all at flow rates of 1 ml/min. The ATPDase activity was recovered by elution with 3 M NaSCN, 0.3% Triton X-100 in Tris buffer. The eluted fractions with the highest ATPase activity were pooled and dialyzed overnight against Tris buffer. The dialyzed fractions were applied to a 7.5-ml Q Sepharose Fast Flow column (diameter 9 mm 2 ) at 0.5 ml/min. After washing the column with 0.01% Triton X-100, 25 mM Tris-HCl, pH 8.2, the activity was eluted by a gradient of 0 -500 mM NaCl in 0.3% Triton X-100 in 25 mM Tris-HCl, pH 8.2. Fractions with peak ATPase levels were pooled and dialyzed.
Enzyme Activity Measurement-Reaction mixtures contained 2 mM substrate (as indicated), 2 mM CaCl 2 , and 0.1 mM EDTA in 30 mM Tris-HCl (or 100 mM where indicated) at pH 7.4. In a number of experiments, the combination of CaCl 2 with EDTA was replaced by the same concentration of MgCl 2 and EGTA in order to test for enzymes specifically activated by magnesium (40). For the determination of optimum pH of activity, the buffer used was 50 mM Pipes for pH 6 -7, 100 mM Tris for pH 6.5-9.5, and 50 mM borate for pH 8.5-11. After 30 min of incubation at 37°C the reaction was stopped by the addition of 150 l of 0.25% (w/v) CuSO 4 , 1% (w/v) SDS, and 4.6% (w/v) sodium acetate, pH 4.0. The inorganic phosphate, produced from exogenous nucleotides during the incubation, was determined according to the method of LeBel et al. (41) as described earlier. 2 One unit of enzyme activity corresponded to the release of 1 mol of P i /min at 37°C. The ATPDase capture assay described by Strobel et al. (36) and slightly modified by Lemmens et al. 2 was used for general characterization experiments. With the capture assay, the enzymatic activity was linear for at least 30 min with an R 2 of 0.98 (linear regression).
Activity of the purified enzyme was tested in 1.5 mM CaCl 2 , 0.5 mM content of nucleotide, 50 mM Tris, and 50 mM imidazole, pH 7.4. After the addition of the purified protein sample to the buffer, it was preincubated at 37°C for exactly 2.5 min and reactions were started by the addition of the substrate (ATP or ADP), then terminated at 5 min with 0.25 ml of the Malachite Green reagent (42). The technique of LeBel et al. (41) described above, allowed the detection of phosphate derived from high concentrations of nucleotides, and was not affected by the enzyme inhibitors used in some of our assays. The technique of quantification of phosphate described by Baykov et al. (42), on the other hand, was more sensitive and allowed measurements of nanomoles of P i /ml. This latter technique was used for the evaluation of the kinetic parameters of the purified hATPDase. To determine specific activities, the protein content of the enzyme preparations was measured using the BCA protein assay (43).
5Ј-p-Fluorosulfonylbenzoyladenosine (FSBA) Treatment-FSBA was incubated with hATPDase in conditions closed to what was described previously (44). FSBA was dissolved in dimethyl sulfoxide (Me 2 SO) and added to a final concentration of 0.125-2 mM FSBA and 5% Me 2 SO in 40 mM KCl, 2 mM CaCl 2 , 20 mM NaAc, 30 mM Tris, pH 7.4, and 700 ng of purified hATPDase in 50 l/assay. After a 25-min incubation at room temperature, 1 ml of reaction buffer was added and samples were immediately tested for enzyme activity as described above.
Peptide Sequencing-Aliquots of the pooled fractions, obtained after immunoaffinity chromatography, containing 20 g of protein, were loaded on a 10% SDS-polyacrylamide gel. After electrophoresis under reducing conditions, proteins were transferred to a ProBlott polyvinylidene difluoride membrane. Prior to NH 2 -terminal sequencing, the ProBlott membrane was subjected to three washing cycles with 50% HPLC grade methanol and HPLC grade water and then the band of interest was excised. NH 2 -terminal sequencing was carried out on a Procise 492 sequenator (Perkin-Elmer/Applied Biosystems), operating in a pulsed-liquid mode.

Generation and Characterization of a Monoclonal Antibody
to hATPDase-Monoclonal antibodies were developed against a partially purified fraction of porcine kidney nucleotidases. 2 Antibodies that bound ATPase activity from a pig kidney solubilized protein fraction were further tested by Western blotting against a particulate fraction of pig liver (Fig. 1). Monoclonal antibodies to cd39 were used as control. The antibody 3A1 detected a protein of about 75 kDa on Western blots after non-reducing SDS-PAGE that is distinct from the 78-kDa cd39 protein (Fig. 1). This signal disappeared on Western blot when protein samples were treated with ␤-mercaptoethanol prior to SDS-PAGE (data not shown). After PNGase F treatment, most of the signal was also lost resulting in a diffuse doublet of bands of lower molecular mass at about 60 kDa (Fig. 1). Assuming a M r contribution of 2,000 -4,000/oligosaccharide (45), the decrease in the apparent molecular mass observed would indicate that the protein core detected by 3A1 (the hATPDase) had 4 -8 N-linked sugars. The observed decrease in signal did not seem to be due to protein degradation, since controls carried out in the same conditions or otherwise tested with anti-cd39 antibodies after deglycosylation revealed the appearance of bands at the expected size on Western blot ( Fig. 1 and data not shown). The fact that 3A1 only recognized the fully glycosylated antigen with intact disulfide bridges suggests that this antibody detects a complex epitope.
Distribution of the hATPDase-To determine the distribution of the protein bound by 3A1 monoclonal antibody, particulate fractions enriched in plasma membranes from a series of tissues were separated by SDS-PAGE under non-reducing conditions and analyzed by immunoblotting. A very strong reaction was observed in liver at about 75 kDa with weaker reactions in kidney and duodenum (Fig. 2). Other tissues appeared either negative or very weakly positive. The upper band seen in few tissues at about 160 kDa was most likely due to crossreacting pig immunoglobulins, since it was detected with the secondary antibody goat anti-mouse antibody alone (data not shown).
We then investigated the immunohistological localization of the hATPDase in liver and kidney (Fig. 3). The monoclonal antibody 3A1 specifically reacted with the bile canaliculi of the hepatocytes. Other tissue components were negative, including the basolateral membrane of the hepatocytes and intracellular compartments, the bile duct epithelium, the sinusoids, and the hepatic capsule. In the kidney cortex, staining was detected mainly at the luminal surface of tubules incorporating the brush border membrane. Bowman's capsule, glomeruli, and blood vessels were consistently negative.
Comparisons of the localization of the hATPDase with the one of ATPDase/cd39 in liver and kidney revealed distinct differences. The 4D11 anti-cd39 antibody mainly reacted with the blood vessel walls of both organs, on endothelial and smooth muscle cells (Fig. 3). In liver, cd39 was found on the peribiliary vascular plexus while the epithelial cells of the bile ducts were negative. Endothelial cells of the centrolobular vein and the portal vein were stained but not as strongly as the endothelium of arteries. Sinusoids were weakly positive with staining more pronounced near the centrolobular vein. Hepatocytes were negative. In kidney, the 4D11 antibody reacted with peritubular capillaries and with the vascular and mesangial cells in the glomeruli.
Characterization and Purification of hATPDase-Since the ATPase detected by 3A1 appeared distinct from cd39, we further characterized the enzyme by capture assays. The effect of pH on the ATPase enzymatic activity was evaluated from pH 6.0 to 11.0. At pH 6.0 ATPase activity was about 50% of the FIG. 1. Immunoblot of pig liver particulate fraction with antibodies to ATPDases. Proteins from a pig liver particulate fraction (10 g/well) were separated by SDS-PAGE under non-reducing conditions, transferred to Immobilon-P membrane and then probed with two monoclonal antibodies to cd39 (2C7 and 4D11) and one monoclonal antibody to putative hATPDase (3A1). Note that the band detected with 3A1 had a slightly lower apparent molecular mass than the 78-kDa cd39 protein.
Monoclonal antibody 3A1 did not fully recognize the deglycosylated protein.
FIG. 2. Western blot analysis of particulate fractions from selected porcine organs with 3A1 monoclonal antibodies. Samples of 10 g (5 g for pancreatic zymogen granule membranes (ZGM)) were fractionated on a 10% acrylamide SDS-PAGE gel under non-reducing conditions, transferred to an Immobilon-P membrane, and incubated with the monoclonal antibody 3A1. Note that the liver expresses the highest levels of hATPDase. The upper band seen particularly in lymph node, thymus, and spleen was nonspecific and was also detected with the secondary anti-mouse IgG's alone (data not shown). maximum, while at or above pH 9.5 activity was barely detectable. Optimum pH ranges were from 6.8 to 8.0 (data not shown). Consequently, all further tests were carried out at physiological pH of 7.4.
Both calcium-and magnesium-dependent nucleotidase activities were tested with multiple nucleotides (Fig. 4). Although all triphospho-and diphosphonucleosides were hydrolyzed, a slight preference for pyrimidines over purines was observed in the presence of calcium. The hATPDase showed a marked preference for calcium over magnesium for all nucleotides tested. AMP, AP 3 A, AP 4 A, and AP 5 A were not hydrolyzed by the enzyme (Fig. 4 and data not shown). When 0.5 mM EDTA plus 0.5 mM EGTA was added to remove traces of divalent cations, no activity was detected with any of the substrates shown in Fig. 4 (data not shown). These results are largely comparable to those reported for ATPDase/cd39, albeit with minor differences in substrate and divalent cation preferences (see Ref. 3 and references therein). Addition of ATP and ADP together did not increase the enzyme activity (data not shown; see below for the purified enzyme data).
We next compared directly the effect of calcium and azide on ATPase and ADPase activity of ATPDase/cd39 and hATPDase using capture assays (Fig. 5). Clear differences were observed between these two ATPDases. The enzyme bound by 3A1 needed much less calcium for optimal activity than cd39. This effect was more evident with ATP than with ADP (Fig. 5A). Similar results were obtained with other ATP and ADP concentrations (0.5, 1, 1.5, 2, and 4.5 mM), with the caveat that less calcium was needed for optimal activity for lower concentration of substrates, for both ATPDases (data not shown). At higher calcium concentrations, from 4 to 15 mM CaCl 2 , the hATPDase activity decreased slightly, when tested with ATP as substrate (data not shown). Unlike cd39, the hATPDase was quite resist-ant to sodium azide, showing no or very little inhibition at all concentrations tested (Fig. 5B). The results obtained for cd39 by this capture assay are in agreement with previous observations for this enzyme (see Ref. 3, and references therein), while the data for 3A1 are in agreement with our previous determinations of ATPDase activity from crude pig liver particulate fractions (46).
The effects of a series of inhibitors of other ATPases and/or FIG. 4. Substrate specificity of the hATPDase. The ability of the hATPDase from liver solubilized membrane proteins to hydrolyze tri-or diphosphonucleosides was evaluated by the capture assay using 3A1 monoclonal antibody (see "Experimental Procedures"). Results are expressed as the mean Ϯ standard deviations (n ϭ 8) of the relative activity reported for ATP as substrate in the presence of calcium corresponding to 0.8 milliunits/well. Calcium-dependent activity was measured in a reaction mixture containing 2 mM CaCl 2 plus 0.1 mM EDTA (solid bars) and magnesium-dependent activity in a reaction mixture including 2 mM MgCl 2 plus 0.1 mM EGTA (open bars) (see "Experimental Procedures").

FIG. 3. Comparative immunohistological localization of two ATPDases in hepatic and renal tissues.
The monoclonal antibody 4D11 was used to localize cd39 and 3A1 the novel ATPDase, in liver (A-E) and kidney (F and G). Sections were counterstained with hematoxylin. A-C, a strong reaction was observed with 4D11 on endothelium and smooth muscle of arteries (arrows). Weaker staining was detected on the branches of the portal vein (pv), centrolobular vein (cv), sinusoidal endothelium (s), and the peribiliary vascular plexus surrounding bile ducts (*). D and E, the antibody 3A1 specifically stained the canalicular domain of the hepatocytes (arrowheads). F, in kidney, anti-cd39 stained the vascular elements and mesangial cells of the glomeruli (gl) and peritubular capillaries surrounding tubules (tub). G, antibody 3A1 stained the renal tubular epithelial cells predominantly on the luminal side. Magnifications: A and D, ϫ200; B, F, and G, ϫ150; C and E, ϫ600.
ectonucleotidases on hATPDase activity were also tested. Only two of these compounds partially inhibited ATPase activity. Sodium vanadate, an inhibitor of plasma membrane Ca 2ϩ -ATPase, inhibited by less then 30%, at a concentration of about 10-fold that generally used. An effective inhibitor of the hATPDase was FSBA, an agent that is frequently used to interact with ATP binding sites of proteins (47). About 50% inhibition was obtained after an incubation of the protein for 25 min with 2 mM FSBA. Data from Table I suggest that classical P, V, and F type ATPases were not candidates for the activity of the enzyme bound by 3A1. Adenosine, Ap 4 A, and p-chloromercuriphenylsulfonate also failed to inhibit hATPDase activity (Table I and data not shown). With the striking exception of sodium azide, the data presented in Table I are similar to what  was previously observed for cd39 (see Ref. 3, and references therein). The protein bound by 3A1 is then also distinct from other characterized ATPDases, cd39 and HB6, that are both inhibited by azide.
Purification of the hATPDase was done by differential centrifugation, immunoaffinity (Fig. 6) and anion exchange chromatography. These steps increased the specific activity by about 900-fold to 8.4 units/mg of protein using ATP as substrate (Table II). The actual enrichment achieved would be higher than what was observed because of levels of inhibition inherently caused by Triton X-100. Additionally, other enzymes present in the homogenate that could hydrolyze ATP, e.g. ATPDase/cd39, were absent in the purified hATPDase fraction (data not shown). The purity of the purified hATPDase preparation was evaluated by SDS-PAGE and silver staining. The hATPDase band comprised about 20% of the total protein loaded (not shown; see below).
Biochemical properties of the purified enzyme were tested and found similar to the results obtained with the capture assay. A time-course analysis of ADP hydrolysis by the purified hATPDase revealed that the reaction was initially linear (regression analysis: R 2 of 0.93 at 10 min, 0.85 at 34 min; Fig. 7A). All subsequent enzymatic tests were carried out for 5 min with a preincubation of exactly 2.5 min. The activities with 0.5 mM ATP or ADP as substrate were very similar with about 8 units/mg of protein of activity for both substrates. As expected for an ATPDase, the activity was not doubled by the addition of these substrates together (Fig. 7B). Kinetic parameters were evaluated for both ATP and ADP as substrate. Initial velocities plotted as a function of ADP concentration revealed normal Michaelis-Menten kinetics (Fig. 7C). Plotting the data according to the method of Woolf-Augustinsson-Hofstee gave a calculated V max of 8.5 units/mg protein for both ATP and ADP as substrates with similar K m values of 100 M (data not shown for ATP).
The purified hATPDase fraction was separated by SDS-PAGE and the NH 2 -terminal part of the protein band at 75 kDa sequenced. The obtained sequence of the hATPDase had limited homology (47% identity) to a putative and as yet uncloned rat liver lysosomal (Ca 2ϩ -Mg 2ϩ )-ATPase (Table III). No homology to pig cd39 was observed. 3 Other potential bands evident on the gel were also sequenced and could be readily eliminated as candidates for the hATPDase; one was identified as a mouse immunoglobulin heavy chain and another as complement C3 (bound nonspecifically by the immunoaffinity column). DISCUSSION In this study we have identified the hATPDase as a 75 kDa glycoprotein decorated with several N-linked oligosaccharides. Biochemical characteristics of hATPDase classify this enzyme as a member of the E-type ATPase/NTPDase family (1,4). Indeed, hATPDase hydrolyzed all tri-and diphosphonucleosides tested equally well but had no effect on AMP, AP 3 A, AP 4 A, or AP 5 A. Addition of both ATP and ADP did not augment the enzyme activity seen with ATP alone, suggesting the presence of only one active site, as observed previously for ATPDase/cd39 (3). The K m of the hATPDase of about 100 M for both ATP and ADP as substrate was higher than cd39, which is about 10 M (see Ref. 3, and references therein). The enzyme FIG. 5. Comparison of the effect of calcium and sodium azide titration on hATPDase and ATPDase/cd39 enzyme activity. Enzyme activity was tested by capture assay using the monoclonal antibody 4D8 to bind cd39 and 3A1 to capture hATPDase in 0.3% Triton X-100-solubilized liver membrane preparations, as described under "Experimental Procedures." Results are expressed as the mean Ϯ standard deviations of three different experiments, each with five data points. A, effect of calcium on ATPDase activity. Maximal activity was obtained at 2 mM CaCl 2 or less and remained stable for up to 4 mM CaCl 2 for both enzymes. Data points to 2 mM CaCl 2 are presented to show the difference in enzyme activity between both ATPDases in the low calcium ranges. Maximal activity was reached at a lower concentration of calcium for the hATPDase (f) when compared with ATPDase/ cd39 (Ⅺ). B, titration of azide from 1 to 20 mM with 2 mM ATP or ADP as substrate in the presence of 2 mM CaCl 2 showed that the hATPDase is not sensitive to azide while ATPDase/cd39 is inhibited. The activity of the hATPDase with ATP (ࡗ) and ADP (f) as substrate without sodium azide corresponded, respectively, to 0.73 Ϯ 0.11 milliunits/well and 0.68 Ϯ 0.15 milliunits/well. For ATPDase/cd39 with ATP (छ) and ADP (Ⅺ), the corresponding values were 0.45 Ϯ 0.05 milliunits/well and 0.38 Ϯ 0.02 milliunits/well, respectively. was active over a broad range of pH values around neutrality and required divalent cations. Inhibitors of classical ATPases had no effect on the hATPDase, and only FSBA partially inhibited enzymatic activity. FSBA has been frequently used to bind ATP binding sites of proteins (47). Interestingly, the hATPDase could be easily differentiated from cd39 by relative insensitivity to azide. There were also some differences between these two ATPDases with respect to calcium requirements as hATPDase required less calcium for maximal activity than cd39. It has been shown that calcium does not bind ATPDase/cd39 (48); in fact, the divalent cation forms a complex with the nucleotide that is the true substrate of cd39. Tetrabasic ATP 4appears to be an inhibitor of the latter ATPDase, since enzymatic activity decreases when the substrate concentration exceeds that of calcium.
The specific activity of the affinity-purified hATPDase preparation was evaluated at 8.4 units/mg, which appears comparable to ATPDase/cd39. Mammalian ATPDase/cd39 has been highly purified from various sources (human, porcine, bovine) and specific activities ranging from 20 to 115 units/mg has been reported (see Ref. 49, and references therein). We measured an apparent V max of 8.5 units/mg of protein for a fraction evaluated at 20% purity. Extrapolation suggests that pure hATP-Dase would have an adjusted specific activity of about 40 units/mg of protein similar to that found for the ATPDase/cd39 described above.
Western blot analysis showed that the hATPDase is ex-pressed at very high levels in liver and to a much lower extent in kidney and duodenum. In liver, the enzyme was specifically detected in the apical canalicular membrane but was absent from the basolateral membrane and intracellular compartments of the hepatocytes. In kidney, it was predominantly localized on the luminal side of the proximal tubular cells. As expected, we noted localization of ATPDase/cd39 to be confined to the vasculature of these organs. Immunohistological localization of cd39 in the liver showed expression to the vascular endothelium and smooth muscle with a minimal reaction with the sinusoidal endothelium. No reaction was observed on the bile duct epithelium or on the hepatocytes. In kidney, cd39 was detected in the peritubular capillaries and the vascular and mesangial cells of the glomeruli. The localization profile of hATPDase and cd39 combined is concordant with general ATPase histochemistry, in both of these organs (16). 2 Our data suggest that cd39 and hATPDase together would account for the major portion of this activity in these organs, under the conditions tested (see below). The 17 NH 2 -terminal amino acids were identified, but no homology could be found with any of the known mammalian members of the E-NTPDase family of proteins: CD39, CD39L1-L4, and UDPase (22,23,(25)(26)(27)(28)(29). However, the NH 2terminal amino acid sequences of these family members are quite heterogeneous. Although some of these proteins are expressed in liver, they all show biophysical and/or biochemical differences with the hATPDase. CD39L1 has been detected in mouse hepatoma cells induced by dioxin (32), but this NTP-Dase prefers ATP as substrate over ADP by a factor of about 40 (26). CD39L1 is therefore distinct from the enzyme described in this study. Northern blot analysis carried out on various human tissues revealed that CD39L2 and CD39L4 are expressed in liver (27). These Northern blots also indicated that CD39L4 was expressed in kidney and intestine and would appear a good candidate for the hATPDase. However, this protein has been recently cloned and expressed and again has quite different characteristics to hATPDase (30). The molecular mass of human CD39L4 was evaluated at about 50 kDa with three potential N-glycosylation sites. It preferentially hydrolyzed diphosphonucleosides over triphosphonucleosides by a factor of about 20, was associated only with macrophages, and secreted in a soluble form by transfected cells (30). The putative CD39L2 may also be a soluble NTPDase and has a different distribution to the hATPDase (27). The biochemical parameters of the latter protein have not yet been reported. CD39L3, also termed HB6, has been shown to be poorly expressed in liver at the mRNA level, and is sensitive to azide inhibition (27,28).  CD39 and the hATPDase are the only other NTPDases known to date that can hydrolyze equally well di-and triphosphonucleoside. CD39 was the first member identified of the mammalian NTPDase family of enzymes (22,23). Our prior observations suggested that ATPDase/CD39 could not account for the high levels of nucleotidase activity in liver. Indeed, Northern and immunoblot analysis indicated low levels of CD39 expression in human, pig, and mouse hepatic tissues (11,23,27,31), while high levels of ATPase and ADPase activities were found in pig liver that were resistant to sodium azide (46). Mutant mice deficient in cd39 have no apparent hepatic abnormality. In the experiments reported herein, the combined activity of cd39 and predominantly the hATPDase comprise more then 95% of plasma membrane E-NTPDase activity.
Our data suggest that the hATPDase represents a new member of the NTPDase family. We observed a NH 2 -terminal sequence identity of 47% with a rat liver lysosomal (Ca 2ϩ -Mg 2ϩ )-ATPase (50), suggesting that the hATPDase is related to this enzyme. This putative lysosomal enzyme has been purified, and found to be a 85-kDa protein that forms tetramers and differs from the H ϩ -ATPase in lysosomal membranes with respect to resistance to inhibition by N-ethylmaleimide and its sensitivity to inhibition by vanadate (50). Lysosomal localization has not been observed for the hATPDase by immunohistochemical techniques.
Two ecto-ATPases previously reported in human hepatoma Li-7A cells could be differentiated through their biochemical sensitivity to mercury (51). In rat liver, the mercurial insensitive ATPase appears to be the dominant form (52). Bovine ATPDase/cd39 activity was inhibited by about 65% with 10 M mercuric chloride (44). Therefore, CD39 expression could potentially account for the mercurial-sensitive ATPase, while the hATPDase would represent the insensitive ATPase form (see Table I).
The function of hATPDase, as other ectonucleotidases, would be expected to be the modulation of P2 receptor-mediated signaling. In liver, extracellular nucleotides and nucleosides exert a variety of physiological functions that could be modulated by the hATPDase described in this study. There is evidence that ATP, ADP, and AMP are released from hepatocytes and bile duct cells. Concentrations of canalicular adenine nucleotides in bile samples and effluents from two hepatic cell lines can be estimated at around 5 M (53). Furthermore AMPCP, a competitive inhibitor of 5Ј-nucleotidase, can increase AMP concentrations by 2-3-fold (53). The hATPDase and the ecto-5Ј-nucle- FIG. 7. Kinetics of the purified hATPDase. Enzyme activity of the purified hATPDase has been carried out in the buffer described under "Experimental Procedures." In all assays, less than 5% of the substrate was hydrolyzed. A, time course of ADP hydrolysis. A sample of 510 ng of the purified hATPDase was incubated in 7.5 ml of the reaction buffer. Reaction was started by the addition of 500 M ADP, and aliquots of 0.6 ml were taken at indicated time point and tested for P i concentrations in triplicate. B, activity of the purified enzyme was tested with ATP and ADP as substrate. Note that the addition of both substrates together at the concentrations indicated did not increase enzyme activity. C, Michaelis-Menten data representation of the influence of ADP concentration from 25 M to 1 mM. Data presented are the mean of two separate experiments each carried out four times. Inset, Woolf-Augustinsson-Hofstee plot was used to evaluate K m and V max with ADP concentration from 25 to 500 M.  otidase may in tandem regulate the concentration of nucleotides in the canaliculus. Since the K m of the ecto-5Јnucleotidase is in the low micromolar range (54), we could anticipate that hATPDase would be rate-limiting in the formation of adenosine in the canalicular area. The apparent K m of about 100 M for the hATPDase would permit the presence of micromolar concentration of nucleotides, as reported of 5 M, but would prevent the development of elevated nucleotide concentrations. Canalicular nucleotides may well serve as signaling molecules on subsets of purinoceptors expressed by hepatocytes and bile duct cells. Extracellular nucleotides regulate glycogenolysis through activation of glycogen phosphorylase and inactivation of glycogen synthase by inhibition of the glucagon effect on cAMP and by the activation of phospholipase D (55,56). Extracellular ATP has been shown to profoundly damage rat hepatocytes in vitro (57), and may be involved in the regulation of canalicular contraction (14,53,58). Hepatocytes and bile duct cells have been shown to interact and communicate via local ATP release in vitro (59). Some of these effects triggered by nucleotides could well be modulated by the hAT-PDase, considering the localization and K m of this enzyme. The hATPDase, as does cd39, also efficiently hydrolyzes pyrimidine nucleotides (e.g. UTP, UDP). This could be of physiological importance since this implies that these enzymes have the potential to hydrolyze agonists of all type of P2 receptors. For example, the purinoceptors P2Y 2 , P2Y 4 , and P2Y 6 are equally or preferentially activated by UTP over ATP (10). The ultimate generation of extracellular adenosine from the dephosphorylation of ATP not only activates adenosine receptors but also produces the key molecule for purine salvage and consequent replenishment of ATP stores within many cell types (1,15). Nucleotides do not appear to be taken up by cells, but their nucleoside derivatives interact with several specific transporters to enable membrane passage (14,15,60,61). Adenosine transporters are of major importance to organs and cells incapable of de novo nucleotide synthesis such as brain, muscles, intestinal mucosae, and bone marrow (14,34). As the liver appears to be a major source of purines for these tissues, curtailment of nucleotide loss into the bile may be important to maintain appropriate nucleotide/nucleoside concentrations within hepatocytes (15). Thus, dephosphorylation of nucleotides by ectonucleotidases may be critical for appropriate systemic purine homeostasis (14). The presence of the hATPDase in the canalicular domain of the hepatocytes is consistent with an important role of the enzyme in salvage and our observation that hATPDase was also found in the brush borders of renal tubules also supports this idea.
In summary, our biochemical, immunological, and histological analysis demonstrates the presence in liver and kidney of a distinct ATPDase with a different localization to cd39, suggesting different pathophysiological functions of these enzymes.