Expression and Characterization of Soluble and Membrane-bound Human Nucleoside Triphosphate Diphosphohydrolase 6 (CD39L2)*

Ecto-nucleoside-triphosphate diphosphohydrolase-6 (eNTPDase61, also known as CD39L2) cDNA was expressed in mammalian COS-1 cells and characterized using nucleotidase assays as well as size exclusion, anion exchange, and cation exchange chromatography. The deduced amino acid sequence of eNTPDase6 is more homologous with the soluble E-type ATPase, eNTPDase5, than other E-type ATPases, suggesting it may also be soluble. To test this possibility, both the cell membranes and the growth media from eNTPDase6-transfected COS-1 cells were assayed for nucleotidase activities. Activity was found in both the membranes and the media. Soluble eNTPDase6 preferentially exhibits nucleoside diphosphatase activity, which is dependent on the presence of divalent cations. Western blot analysis of eNTPDase6 treated with PNGase-F indicated both soluble and membrane-bound forms are glycosylated. However, unlike some membrane-bound ecto-nucleotidases, the eNTPDase6 activity was not specifically inhibited by deglycosylation with peptideN-glycosidase F. Soluble eNTPDase6 hydrolyzed nucleoside triphosphates poorly and nucleoside monophosphates not at all. Analysis of the relative rates of hydrolysis of nucleoside diphosphates (GDP = IDP > UDP > CDP ≫ ADP) suggests that soluble eNTPDase6 is a diphosphatase most likely not involved in regulation of ADP levels important for circulatory hemostasis.

Nucleoside triphosphate diphosphohydrolases (NTPDases) 1 are enzymes characterized by their ability to hydrolyze nucleoside tri-and diphosphates. Although nucleotide specificity and specific activities vary among the NTPDases, other characteristics such as insensitivity to typical ATPase inhibitors and cation dependence for activity are invariable (1). The NTP-Dases all have their active sites outside the cytoplasm of the cell and exist as either ecto-, endo-(luminal), or soluble apyrases whose activities have been speculated to be involved in processes as diverse as neurotransmission, N-and O-glycosylation, cardiac function, platelet aggregation, cell adhesion, muscle contraction and relaxation, vascular tone, secretion of hormones, immune responses, and cell growth (2). The ectoapyrases hydrolyze nucleoside diphosphates at rates comparable to nucleoside triphosphates and are perhaps the best studied of all the NTPDases; they include eNTPDase1 (CD39) (3) and eNTPDase3 (HB6) (4). These are integral membrane proteins with two membrane spanning regions, one at each end of the primary structure, as well as a large extracellular domain, which includes the active site of these enzymes. Intracellular luminal endo-NTPDase activities have been found associated with both the endoplasmic reticulum (ER) and the Golgi apparatus. At least two Golgi membrane-associated NTPDases have been described, a GDPase (5-7) and an UDPase, both with luminal active sites (8). The ER-UDPase lacks transmembrane domains and is believed to be a soluble luminal protein (9). In addition to the soluble ER apyrase, a soluble apyrase from potato has been characterized (10) and sequenced (8). Soluble apyrases have also been studied from mosquitoes and other hematophagus insects (11,12), as well as from the single-celled Tetrahymena (13). In addition, human soluble eNTPDase5, also known as CD39L4 (14), has recently been expressed and characterized (15). In the present study we expressed and characterized the putative soluble human eNTPDase6, also known as CD39L2 (14). This ecto-nucleotidase has properties distinct from eNTPDase5 (CD39L4), as well as from the other ecto-, endo-, and soluble NTPDases. In clarified media from eNTPDase-expressing COS-1 cells, its specific activity is 89 mol of P i /mg/h (using GDP as substrate) with nucleotide preference for GDP ϭ IDP Ͼ GTP Ͼ ITP Ͼ UDP Ͼ CDP Ͼ UTP Ͼ CTP Ͼ ADP Ͼ ATP. eNTPDase6 exists predominantly as an approximately 50-kDa soluble, monomeric protein. It is postulated that the membrane-bound form contains a 78-amino acid signal peptide, which is cleaved after insertion into and through the cell membrane resulting in release of the soluble ecto-nucleotidase into the extracellular space.

EXPERIMENTAL PROCEDURES
Materials-Epicurian coli ultracompetent bacteria were purchased from Stratagene. Plasmid purification kits were purchased from Qiagen, Inc. LipofectAMINE Plus reagent, Dulbecco's modified Eagle's medium, calf serum, goat anti-rabbit horseradish peroxidase-conjugated secondary antibody, and antibiotics/antimycotics were all obtained from Life Technologies. Falcon tissue culture-treated plates were from Becton-Dickinson. The mammalian expression vector pcDNA3 was obtained from Invitrogen. Restriction endonucleases and T4 DNA ligase were purchased from Promega. Ampicillin, nucleotides, diethylpyrocarbonate (DEPC), and Sephacryl S-200 were purchased from Sigma.
Human CD39L2 (eNTPDase6) cDNA in Mammalian Expression Vector pcDNA3 and Expression in COS-1 Cells-The human eNTPDase6 cDNA was isolated and sequenced and has been described previously (14). The 2762-bp eNTPDase6 cDNA insert was excised from a pLXPIH retroviral expression vector and inserted into the 5.4-kb pcDNA3 mammalian expression vector using EcoRI and XbaI restriction endonucle-ases. This construct was used to transiently transfect COS-1 cells using LipofectAMINE Plus reagent as described previously (16 -18), to characterize the expressed enzyme.
Cell Membrane and Media Preparation and Characterization-Crude (total) cell membrane preparations were harvested as described previously (16 -18). Briefly, COS-1 membranes from cells transfected with eNTPDase6 cDNA were harvested 72 h after transfection. Cell membranes were harvested by scraping the culture plates in tissue homogenization buffer (THB) consisting of 30 mM MOPS/NaOH (pH 7.4), 250 mM sucrose, and 2 mM EDTA, followed by homogenization in a glass homogenizer with a Teflon pestle (Thomas Scientific, Swedesboro, NJ). The cell homogenate was centrifuged at 48,000 rpm in a Ti-50 rotor (150,000 ϫ g) for 60 min at 4°C. The pellet was resuspended in a small volume of THB, homogenized again, and used for biochemical assay.
When culture media was to be collected for harvesting of soluble nucleotidase, the COS-1 cells were transferred to incomplete media 24 h after transfection (48 h before harvesting the media). Culture media was collected and centrifuged at 48,000 rpm for 1 h to remove particulate material. The clarified media, approximately 30 ml, was concentrated to 1.5 ml using a Centriprep-30 concentrator, as recommended by the manufacturer (Amicon). Nucleotidase activity was determined by measuring the amount of inorganic phosphate released from nucleotide substrates at 37°C using a modification of the technique of Fiske and SubbaRow (19), as described previously (4). Protein concentrations were determined using the Bio-Rad CB-250 dye binding technique according to the modifications of Stoscheck (20), using bovine serum albumin as the standard. SDS-polyacrylamide gel electrophoresis was performed according to Laemmli (21), and blotted onto polyvinylidene difluoride membrane (22) for Western analysis as described previously (18).
Size Exclusion and Ion Exchange Chromatography-Concentrated culture media from transfected COS cells was applied to a Sephacryl S-200 column equilibrated in 20 mM MOPS, 5 mM MgCl 2 , pH 7.4. Fractions were collected and assayed for nucleotidase activity. Bovine ␥ globulin (158 kDa), bovine serum albumin (66 kDa), ovalbumin (44 kDa), myoglobin (17 kDa), and vitamin B-12 (1.35 kDa) were used to calibrate the Sephacryl S-200 column. Size exclusion column fractions found to contain the majority of GDPase activity were pooled and used for ion exchange chromatography. First, 300 l was diluted to 1.0 ml with 20 mM MOPS, 2 mM MgCl 2 , pH 7.6, and applied to a quaternary ammonium MemSep anion exchange column (Millipore) equilibrated with the same buffer. 5-ml samples were sequentially eluted in 1.0-ml aliquots with MOPS buffer containing no NaCl, 0.1 M NaCl, 0.2 M NaCl, and 1.0 M NaCl. Fractions identified as containing GDPase activity were applied to a sulfo-propyl MemSep cation exchanger (Millipore) equilibrated with the same MOPS buffer. Fractions were eluted and collected as described above.
Polyclonal Antibody Production and Immunoprecipitation-The anti-peptide polyclonal antisera used in this study was raised (by Lampire Biological Laboratories, Pipersville, PA) against a peptide corresponding to the extreme COOH terminus of the eNTPDase6-deduced protein sequence (acetyl-(C)FAYIDSLNRQKSPAS-COOH). The peptide was synthesized and conjugated to keyhole limpet agglutinin (KLH) by Quality Controlled Biochemicals, Inc., Hopkinton, MA, via the NH 2terminal cysteine residue. The crude antisera was affinity purified using the Pierce Chemical Company SulfoLink kit (catalogue number 44895), covalently attaching the peptide via the NH 2 -terminal cysteine on the peptide, as described by the manufacturer.
PNGase-F Treatment-For monitoring the effect of deglycosylation on enzyme activity under non-denaturing conditions, samples of crude media (0.2 mg/ml total protein), containing soluble eNTPDase6, and, separately, membrane-bound eNTPDase6 (1.6 mg/ml) were diluted to 0.1 mg/ml in deglycosylation buffer consisting of 30 mM MOPS, 2 mM EDTA, 0.02% saponin, pH 7.2. At 0 min, 4 l (80 units) of control (boiled) or active (non-boiled) PNGase-F deglycosylation enzyme (Roche Molecular Biochemicals) was added to the eNTPDase6 samples and incubated at 37°C. After various incubation times at 37°C, 10-l aliquots were removed from the reaction tubes and diluted with nucleotidase assay buffer (20 mM Tris, 1 mM EGTA, 7 mM MgCl 2 , 100 mM KCl, pH 7.0), and used to determine GDPase activity as described above.
For complete deglycosylation by PNGase-F under denaturing conditions, samples were acetone-precipitated with 4 volumes of acetone overnight at Ϫ20°C, pelleted, redissolved in 0.2% SDS, and heated before diluting with buffer and incubating with PNGase-F overnight at 37°C. Laemmli sample buffer was then added, and the samples were Western blotted (see Fig. 2).
DEPC Treatment-Modification of COS cell-expressed eNTPDase6 was performed using 1 mM DEPC in 20 mM MOPS, pH 7.4, containing 5 mM MgCl 2 at 22°C. Aliquots were taken at various times and stopped by dilution into ice cold MOPS buffer containing excess histidine. The DEPC reaction was reversed by adding hydroxylamine to a final concentration of 100 mM at 16 min.

RESULTS
NTPDase6 was expressed in COS cells and characterized as both soluble and membrane-bound forms. The results of nucleotidase assays from three separate transfections are summarized in Table I. The soluble form of eNTPDase6 demonstrated the highest specific activity with IDP (101 mol/mg/h) and GDP (89 mol/mg/h). Lower nucleotidase activities were observed using UDP and CDP; 33% and 23%, respectively, of that observed with IDP. Soluble eNTPDase6 exhibited very low activity with all nucleoside triphosphates used as well as with ADP (Ͻ10 mol/mg/h). Although 90% of the total recovered GDPase and IDPase activity was found to be associated with the soluble form of eNTPDase6, the membrane-bound form was also characterized. The membrane-bound form of eNTPDase6, from crude total membranes, also demonstrated the highest specific activity with inosine and guanosine nucleotides. Unlike soluble eNTPDase6, the membrane-bound form showed a slight preference for hydrolysis of most nucleotide triphosphates. One notable exception was a significant preference for GDP (10 mol/mg/h) over GTP (3 mol/mg/h). Nucleotidase activity of both the soluble and membrane-bound forms of eNTPDase6 GDPase activities were examined for cation dependence, because this is a defining characteristic for members of the eNTPDase family (1). The soluble form was found to be highly dependent on the presence of divalent cations for optimal activity as is seen in Table II. In the absence of cations, specific activity for GDP hydrolysis was 1.1 mol/mg/h, whereas optimal GDPase activity was 74 mol/mg/h in the presence of 2 mM MnCl 2 . All concentrations of MnCl 2 , MgCl 2 , and CaCl 2 tested resulted in significant increases of GDPase activity over that observed in the absence of cations. The membrane-bound form of eNTPDase6 also was found to be dependent on the presence of cations for optimal GDPase activity. In the absence of cations, GDPase activity was 0.3 mol/mg/h, whereas GDPase activity was optimal in the presence of 10 mM MgCl 2 (6 mol/mg/h).
The soluble form of eNTPDase6 was partially purified using size exclusion chromatography, anion exchange chromatography, and cation exchange chromatography. Serum-free media was collected from COS-1 cells expressing eNTPDase6 over 3 days, as described under "Experimental Procedures." The media was concentrated to 1 ml and loaded onto a Sephacryl S-200  Fig. 1. The fractions containing the majority of the GDPase activity corresponded to a molecular mass of ϳ50 kDa. This molecular mass agrees well with the theoretical weight for a soluble, monomeric eNTPDase6 protein as determined from translation of the cDNA (theoretical non-glycosylated molecular mass ϭ 45 kDa, assuming signal cleavage at amino acid 78, predicted by computer analysis). After size exclusion chromatography, both anion and cation exchange chromatography were used to further characterize and purify the soluble eNTP-Dase6 109-fold as is shown in Table III. After the final step of purification, the purified soluble eNTPDase6 had a average (n ϭ 2) specific activity with GDP of 9227 mol/mg/h. An anti-eNTPDase6 polyclonal antibody was developed, affinity purified, and used for Western blot analysis and immunoprecipitation. As seen in Fig. 2, the anti-eNTPDase6 antibody reacted specifically with an approximately 50-kDa protein in both eNTPDase6-expressing COS cell membranes and media (lanes 1 and 5). The 50-kDa protein band was not present when COS cell control membranes were probed with the anti-eNTPDase6 antibody. The-50 kDa band also was not present when membranes and media were probed with pre-immune sera (not shown). Also shown in Fig. 2 is the shift in molecular mass of both the soluble and membrane-bound forms after treatment with PNGase-F, indicating both forms are N-glycosylated (lanes 2 and 6). In addition, the anti-eNTPDase6 antibody was able to immunoprecipitate 84% of the GDPase activity from media obtained from COS cells expressing eNTPDase6.
The dependence of eNTPDase6 activity on glycosylation was determined by treating both the membrane-bound and soluble forms with PNGase-F under conditions, which previously resulted in significant loss of glycan chains and activity of eN-TPDase1 and eNTPDase3 (HB6 (16)). There was no significant difference in activity between those eNTPDase6 samples treated with active PNGase-F and those control samples treated with heat-inactivated PNGase-F (data not shown). However, Western blot analysis of eNTPDase6 treated with active PNGase-F indicated the presence of two glycosylation sites (Fig. 3), as is evidenced by the presence of three discrete protein bands corresponding to eNTPDase6 protein with 2, 1, or 0 glycan chains attached (see arrows in Fig. 3 between the "Ϫ" and "ϩ" lanes incubated for 3.5 h).
Diethylpyrocarbonate (DEPC) is an inhibitor of eNTPDases, presumably due to histidine modification (23,24). Both the soluble and membrane-bound forms of eNTPDase6 were treated with DEPC followed by hydroxylamine to reverse his-tidine modification. As shown in Fig. 4, DEPC treatment of the soluble form of eNTPDase6 inhibited GDPase activity 60%. However, hydroxylamine treatment was unable to reverse the inhibition caused by DEPC treatment. Similar results were obtained with the membrane-bound form of eNTPDase6 (Fig. 4). DISCUSSION This paper describes the first expression of the soluble human nucleotidase eNTPDase6 (CD39L2), and for the first time, the characterization of both soluble and membrane-bound forms of the eNTPDase6. To date there have been several reports of soluble eNTPDases from organisms as diverse as plants (10), Tetrahymena (13), mosquitoes (11), and humans (15); however, most studies have focused strictly on the soluble forms, and comparative studies with their respective membrane-bound forms have not been reported. In this study, we found notable differences in nucleotide preference and cation dependence between the two forms of eNTPDase6, yet similarities in size, DEPC response, and lack of importance of glycosylation.
From hydropathy plots of the deduced amino acid sequence, it was postulated that eNTPDase6 (CD39L2) has a single potential transmembrane domain at the NH 2 terminus, suggesting the possibility of membrane attachment via a signal se-

TABLE III Partial purification of soluble eNTPDase6
The summation for purification of soluble eNTPDase6 is shown in Table 3. Starting material for purification was 30 ml of clarified, serum free, growth media from eNTPDase6-expressing COS-1 cells, which had been concentrated to 1.5 ml as described under "Experimental Procedures." The average of two separate transfections and purifications is shown. quence (14). Moreover, computer analysis of the translated amino acid sequence of the eNTPDase6 cDNA revealed a putative signal sequence at amino acid 78 ( 76 TRA2APG 80 ) which, if cleaved, should result in a soluble form. To characterize eNTPDase6 and determine if it exists as a predominantly soluble or membrane-bound form, we transiently expressed eNTPDase6 in COS-1 cells. Low nucleotidase activity (0.6 -12 mol/mg/h) was found associated with the eNTPDase6expressing COS-1 cell membranes (Table I), suggesting the majority of the activity is secreted into the media. Approximately 90% of total GDPase activity was found associated with the soluble form of eNTPDase6 (Table I).
Both the soluble and membrane-bound forms of eNTPDase6 demonstrated preference for inosine and guanosine nucleotides ( Table I). The nucleotide hydrolysis ratio for GDP and GTP (i.e. GDPase:GTPase activity) was approximately 11 for the soluble form of eNTPDase6. The soluble potato (ADPase:ATPase ϭ 1.0) (10) and endoplasmic reticulum (UDPase:UTPase ϭ 5.0) (9) eNTPDases have lower nucleotide hydrolysis ratios, whereas the soluble eNTPDase5 (UDPase:UTPase ϭ 34) (15) has a higher nucleotide hydrolysis ratio making the soluble eNTP-Dase6 unique in its nucleotide preference among the soluble eNTPDases. Overall, soluble eNTPDase6 hydrolyzed nucleoside triphosphates very poorly and exhibited a strong preference for purine and pyrimidine nucleoside diphosphates with oxygen at positions 6 and 4, respectively. The membrane-bound form of eNTPDase6 differed in that it hydrolyzed some nucle-oside triphosphates slightly better than nucleoside diphosphates with a notable exception in the case of guanosine nucleosides, where the GDPase:GTPase activity ratio was 3.3:1. As shown in Table II, both forms of eNTPDase6 were dependent on the presence of cations for optimal activity, a characteristic universally found in other members of the eNTPDase family (1,7,8,25).
Soluble eNTPDase6 was partially purified using size exclusion, anion exchange, and cation exchange chromatography. Fractions from the S-200 size exclusion column containing the majority of the GDPase activity were found to correspond with a molecular size of approximately 50 kDa (Fig. 1), as is predicted from the cDNA sequence of eNTPDase6. This molecular size is consistent with a monomer, and not consistent with a homo-oligomer of soluble enzyme. This is concordant with previous findings that soluble eNTPDase enzymes are monomeric (10,26,27), whereas membrane-bound forms are tetrameric (8,28).
It has been speculated (14) that eNTPDase6 (CD39L2) would be released as a soluble enzyme after cleavage of the NH 2terminal signal sequence. The theoretical pI of the deduced amino acid sequence of eNTPDase6, including the hydrophobic NH 2 -terminal signal peptide sequence, is 9.91. If the soluble eNTPDase6 is generated by cleavage after Ala-78 as predicted by computer analysis of the sequence, then the resultant soluble form is predicted to have a theoretical, non-glycosylated pI ϭ 7.52, and therefore should not bind to either anion or cation exchange resins at physiological pH. Chromatography was therefore performed under conditions (pH 7.4) that would not promote binding to either cation or anion exchangers. Using both the anion and cation columns, soluble GDPase activity was found in fractions that did not bind to either column (see "Experimental Procedures"). This indicates that the prediction of cleavage of a signal sequence after Ala-78 is most likely correct, because this would eliminate many positively charged amino acids, resulting in a 45-kDa soluble protein with a greatly reduced pI, close to 7.4, consistent with the purification results.
A polyclonal anti-peptide antibody (against a COOH-terminal non-conserved region of eNTPDase6) was developed to further study both the soluble and membrane-bound forms of eNTPDase6. As shown in Fig. 2, a single protein at approximately 50 kDa was recognized in the eNTPDase6-expressing cells and media that was not observed in control cell membranes.
Computer sequence analysis predicts two potential N-glycosylation sites for eNTPDase6. PNGase-F was used to deglycosylate eNTPDase6 to determine the extent of glycosylation required for GDPase activity. From activity data we found no difference in GDP hydrolysis between those samples treated with active PNGase-F as compared with those treated with heat-inactivated PNGase-F. These experiments were done under identical conditions used to demonstrate both deglycosyla-  , lanes 3 and 4) and membrane proteins (40 g, lanes 7 and 8), with or without PNGase-F treatment, were probed with affinity purified anti-eNTPDase6 antibody. A single band was differentially present in the eNTPDase6-expressing media and membranes that had been treated with active PNGase-F, at approximately 50 kDa (lanes 2 and 6). The eNTPDase6 antibody detected a slightly larger, glycosylated eNTPDase6 protein band in both cell media and membranes expressing eNTPDase6 (lanes 1 and 5). The eNTPDase6 antibody did not react with any proteins present in the mock transfected COS-1 cell media (lanes 3 and 4) or membranes (lanes 7 and 8). tion and loss of activity of eNPTDase1 and eNTPDase3 (16). However, the presence of two glycosylation sites was confirmed by Western blot analysis of PNGase-F-treated eNTPDase6 (Fig. 3). These data are consistent with our hypothesis that deglycosylation of eNTPDases decreases nucleotidase activity by disrupting normal quaternary structure, which is not present in the monomeric, soluble eNTPDase6 (16).
DEPC is recognized as a general inhibitor of eNTPDase activity, presumably due to modification of histidine residues in the eNTPDases, because histidine modification is most consistent with the pH dependence of inactivation and the reversibility with NH 2 OH (23,29). DEPC-induced histidine modification and enzyme inhibition can be reversed by the addition of hydroxylamine to the reaction mixture, whereas tyrosine modification reverses more slowly, and lysine and cysteine modifications are not reversible (23). We found that both forms of eNTPDase6 were inhibited by DEPC approximately 60% (Fig.  3), which is comparable to what has been reported for eNTP-Dase3 (70%) 2 and eNTPDase1 (80% (30)). However, unlike other eNTPDases studied, this inhibition was most likely due to DEPC modification of non-histidine residues, because it was not reversed by hydroxylamine (even at higher concentrations than used in Fig. 3, data not shown).
In summary, this report describes the expression and characterization of both the membrane-bound and soluble forms of eNTPDase6 (CD39L2). The expressed protein possesses shared characteristics of the ecto-apyrases such as cation dependence and sensitivity to DEPC. eNTPDase6 is unique in that it is the first soluble eNTPDase that prefers GDP and IDP as substrates. Human eNTPDase6 is much more widespread in its tissue distribution than the only other soluble human ectonucleotidase described to date, eNTPDase5 (CD39L4 (15)). eNTPDase5 (CD39L4) was shown in two studies to be rather limited in its tissue distribution (14,15), unlike the eNTP-Dase6 described here (CD39L2), for which mRNA has been found for all tissues tested (14). The broad tissue distribution of eNTPDase6 (CD39L2) suggests the possibility that the range of functions or the importance of the physiological function(s) for this soluble nucleotidase might exceed that of the narrowly expressed, soluble eNTPDase5 (CD39L4).
Note Added in Proof-After acceptance of this work, N-terminal amino acid sequencing of soluble eNTPDase6 indicated that cleavage between Arg-76 and Ala-77 separates the N terminus signal peptide from the remainder of the soluble protein. . Experiments were performed as described under "Experimental Procedures." The average of three separate experiments is shown with standard deviations. Hydroxylamine was added, to a final concentration of 100 mM, at t ϭ 16 min to initiate recovery from DEPC inhibition due to histidine modification. The soluble and membrane-bound eNTPDase6 activities were inhibited approximately 60% each, followed by recovery of activity of only approximately 5% and 10%, respectively.