ATP-consuming and ATP-generating Enzymes Secreted by Pancreas*

Pancreatic acini release ATP in response to various stimuli, including cholecystokinin octapeptide (CCK-8), as we show in the present study. There were indications that pancreatic juice also contains enzymes that could hydrolyze ATP during its passage through the ductal system. The aim of this study was to determine which ATP-degrading and possibly ATP-generating enzymes were present in pancreatic secretion. For this purpose, pancreatic juice was collected from anesthetized rats stimulated with infusion of CCK-8. Purine-converting activities in juice samples were assayed by TLC using either [γ-32P]ATP or 14C/3H-labeled and unlabeled nucleotides as appropriate substrates. Data show that the juice contains the enzyme ecto-nucleoside triphosphate diphosphohydrolase that can hydrolyze both [14C]ATP and [3H]ADP about equally well, i.e. CD39. Reverse-phase high-performance liquid chromatography analysis additionally shows that this enzyme has broad substrate specificity toward other nucleotides, UTP, UDP, ITP, and IDP. In addition, secretion contains ecto-5′-nucleotidase, CD73, further converting [3H]AMP to adenosine. Along with highly active hydrolytic enzymes, there were also ATP-generating enzymes in pancreatic juice, adenylate kinase, and NDP kinase, capable of sequentially phosphorylating AMP via ADP to ATP. Activities of nonspecific phosphatases, nucleotide pyrophosphatase/phosphodiesterases, and adenosine deaminase were negligible. Taken together, CCK-8 stimulation of pancreas causes release of both ATP-consuming and ATP-generating enzymes into pancreatic juice. This newly discovered richness of secreted enzymes underscores the importance of purine signaling between acini and pancreatic ducts lumen and implies regulation of the purine-converting enzymes release.

ciliary clearance by stimulation of ciliary beat frequency, Cl Ϫ and fluid secretion, and mucin secretion from goblet cells. Nucleotide levels are regulated by a number of enzymes especially active on the apical surface of various airway epithelial cell lines. These include both ATP-degrading enzymes, NPP, CD73, and alkaline phosphatase, and ATP-generating enzymes, NDPK and adenylate kinase (10 -12).
In exocrine glands, ATP and other purines are thought to be important regulators of salt and fluid transport (13). It seems that pancreatic acini have relatively few functional P2 receptors (14). However, nucleotide-and nucleoside-selective receptors may be important regulators in pancreatic ducts, which secrete bicarbonate-rich fluid. In rat pancreas, ATP is released from acini into the series of excurrent ducts that are rich in P2 receptors. Close to acini, the ATP concentrations are in the high micromolar range; however, low amounts of ATP are detected in the final pancreatic juice collected from the main duct (15,16). CD39 is expressed in pancreatic acini and also pancreatic ducts of rats and pigs, as demonstrated by immunohistochemistry and histochemistry (16 -18). Our recent study revealed that pancreatic juice also contains CD39. Relatively low levels of ADP and AMP in the juice indicated that other enzymes may be present in the juice (16). To understand the acino-ductal paracrine regulation, it is important to determine which ecto-enzymes are present in pancreatic secretions and thus estimate the prevalence of nucleotides versus nucleosides. Therefore, the aim of the present study was to determine nucleotide/nucleoside-converting enzymes secreted in pancreatic juice collected under in vivo stimulation of rat pancreas with CCK-8. To determine the whole spectrum of purine-converting enzymes in pancreatic juice, we employed TLC assay with 3 H/ 14 C-labeled nucleotide substrates, autoradiography of direct 32 P i transfer from [␥-32 P]ATP, and reverse-phase high-performance liquid chromatography (HPLC) analysis of NTP metabolism. The obtained data clearly demonstrate that pancreatic juice contains ATP-consuming enzymes, CD39 and CD73, as well as ATP-generating enzymes, adenylate kinase and NDPK. We propose that they have a role in purinergic signaling between pancreatic acini and pancreatic ducts.

EXPERIMENTAL PROCEDURES
Materials-Hormones secretin and CCK-8 were obtained from Sigma. Tissue culture media and phosphate-buffered saline were from Invitrogen. Mebumal (27.5 Ci/mmol) were from PerkinElmer Life Sciences. TLC plates were Alugram SIL G/UV 254 , and Polygram CEL 300 polyethyleneimine types were supplied by Macherey-Nagel (Duren, Germany). Synergi Hydro-RP 80A HPLC column (4-m, 150 ϫ 4.6 mm) protected by a reverse-phase C18 guard cartridge were from Phenomenex (Torrance, CA). All other reagents and standard chemicals were purchased from Sigma.
Collection of Pancreatic Juice-The necessary permission for animal experiments was obtained from the Danish Animal Ethical Committee. In vivo collection of pancreatic juice was undertaken on female Wistar rats weighing 160 -300 g. The animals, fasted overnight, were anesthetized with mebumal (pentobarbital, 40 mg/kg, intraperitoneal). Anesthesia was maintained during the experiments by additional, intravenous injections of mebumal. The body temperature of animals was maintained at 38°C by means of a thermostatically controlled heating. The animals were tracheostomized, and the facial vein was cannulated for infusions. The abdomen was opened by a midline incision, and the pylorus and the proximal ends of the bile duct were ligated. The common pancreatic bile duct was cannulated with an ϳ2-cm-long polyethylene tube, and collection of pancreatic juice was started with a control period of 30 -60 min during which medium (Dulbecco's modified Eagle's medium 1000/Ham's F12 medium) was infused into the facial vein. Secretion was then stimulated by infusion CCK-8 (5.6 pmol/min/200 g of animal), for about 60 min. The infusion rate (0.03 ml/min/animal) was held constant with a syringe pump (Cole-Parmer). Pancreatic juice was collected on ice over 10 -15-min periods and stored at Ϫ80°C or on dry ice for transport. Blood was collected in heparinized syringes, and plasma was separated by centrifugation. After experiments, animals were killed by overdose of mebumal. For estimations of ATP in pancreatic juice, 5-10-l samples were quickly thawed and immediately assayed with an SL kit using internal and external standards according to the manufacturer's instructions (BioThema, Haninge, Sweden). Luminescence was detected in a FLUOstar Optima microtiter plate reader (BMG Labtech, Offenburg, Germany).
Preparation of Pancreatic Acini-Acini were prepared by collagenase digestion as described earlier (15). After filtering through nylon mesh, cells were gently washed in physiological HCO 3 Ϫ -free buffer (ϪBIC) of the following composition (in mmol/liter): 145 Na ϩ , 3.6 K ϩ , 1.5 Ca 2ϩ , 1 Mg 2ϩ , 145 Cl Ϫ , 2.0 phosphate, 5 glucose, and 10 HEPES. Finally, cells were suspended in ϪBIC solution, and 50-l aliquots were pipetted into 96-well microtiter plates followed by 50 l of luciferin/luciferase mix HSII (Roche Diagnostics, Manheim, Germany), which was dissolved in ϪBIC. Acini were allowed to rest for 45-60 min, but most of them did not attach to substrate. Subsequently, luminescence was monitored after injection of 5-l volumes of ϪBIC and CCK-8 made up in ϪBIC. Luminescence was monitored in 1-s intervals in the microtiter plate reader. Temperature was 25°C to slow down ATP hydrolysis by enzymes. ATP standards were treated as samples, and standard curves were constructed for each experiment. Under the given experimental conditions, ATP standards gave stable luminescence signals. ATP release monitored in arbitrary luminescence units was recalculated as ATP concentration and corrected for 1 million cells/ml. Cell numbers were estimated by cell counting and from cellular ATP freed following cell lysis.
Protein Measurement-Total protein concentration in pancreatic juice was determined by using BCA Protein Assay Kit (Pierce).
HPLC Analysis of Nucleotide Metabolism-Pancreatic juice (1.5-2 l, ϳ250 g of protein) was incubated with 20 M ITP/ UTP in a final volume of 300 l of phosphate-buffered saline supplemented with 0.4 mM MgCl 2 . Aliquots of the mixture (100 l) were collected at the beginning (zero point) and after a 60-min incubation at 37°C, and nucleotides were extracted by adding 20 l of 4 M perchloric acid. After centrifugation, the supernatant was adjusted to neutral pH by 4 N KOH (ϳ28 l), clarified again by centrifugation, and stored at Ϫ70°C. The samples (20 l) were then injected onto a Synergi Hydro-RP 80A column and separated by reverse-phase HPLC as described previously (4).

RESULTS
From isolated preparations of pancreatic acini and ducts, we have learned that in response to cholinergic stimulation, pancreatic acini secrete ATP (15). This ATP might be destined for the adjoining pancreatic ducts, which express functional P2 receptors (see the Introduction). In the present study, we wished to determine whether similar signaling functions with the classical acinar agonist CCK-8, both at the acinar cell level and at the whole organ level. In the first series of experiments, ATP release was monitored in situ in a suspension of pancreatic acini. In the first part of the recording in Fig. 1A, one can observe that mechanical disturbance caused by pump injection of ϪBIC results in a small ATP release. Repeated injection of ϪBIC has no further effect (results not shown). However, injection of CCK-8 (10 Ϫ11 M) resulted in a transient increase in ATP. The peak agonist-induced ATP release is expressed per 1 mil-lion cells/ml (Fig. 1B). The decay in the ATP signal was most likely due to the presence of enzymes hydrolyzing ATP since ATP standards gave sustained luminescence signal under our conditions. In addition, supernatant collected from acinar incubation medium was able to degrade ATP standards, similar to what we have also seen with whole pancreas juice (see below).
Next we studied pancreatic secretion of intact pancreas stimulated to secrete with CCK-8, which when infused at 5.6 pmol/ min/200 g of animal, would be 10 Ϫ11 to 10 Ϫ10 M in the interstitium. Fig. 1C shows the secretory response of rat pancreas with time and ATP concentrations in pancreatic juice. Pancreatic juice collected from cannulated common pancreatic bile duct was analyzed for ATP, and the concentrations were 4.5 Ϯ 1.2 nmol/liter (n ϭ 7) (Fig. 1D). These ATP concentrations are higher than detected earlier, most likely due to improved techniques for juice collection (16,19). Nevertheless, ATP concentrations in juice are significantly lower than estimated from acini preparations. In fact, if we assume that all acinar cells that comprise the bulk of pancreatic tissue would release ATP into the juice, then disparity between proximal and distal ATP concentrations would be even larger. We postulated that the diminished ATP levels as secretion progressed from acini through the whole duct tree were indicative that the pancreatic juice also contained ATP-hydrolyzing enzymes. Therefore, pancreatic juice was collected between 60 and 80 min of stimulation, when secretion was well established, not contaminated by bile and possible cellular debris from cannulation processes. The collected samples were used for enzyme analysis, which was the primary focus of this study. Fig. 2 shows the time courses of hydrolysis of 20 M [ 14 C]ATP by pancreatic juice (Fig. 2A) and formation of its dephosphorylated 14 C metabolites (Fig. 2B). Gradual decay in ATP concentration was accompanied by an immediate rise in AMP and, after a time lag, there was progressive production of adenosine. [ Fig. 6A).
Little ADP formation in the assay medium may be explained either by immediate breakdown of the ATP-derived ADP via high NTPDase activity or, alternatively, due to direct ATP conversion into AMP through NPP reaction. Nevertheless, the latter suggestion seems highly unlikely for the following reasons. Firstly, in the presence of 300 M unlabeled ADP, hydrolysis of 20 M [ 14 C]ATP by juice samples was markedly diminished (Fig. 2A), suggesting that ADP and ATP compete for the same catalytic site of NTPDase. Secondly, Fig. 3  The levels of nonspecific phosphatases were also negligible as similar patterns of [ 14 C]ATP hydrolyzes were observed with and without the excess of alternative phosphorylated substrate, ␤-glycerophosphate (data not shown). For further evaluation of the nucleotidase substrate specificity, pancreatic juice was incubated with 20 M ITP or GTP, and the reaction products were separated by reverse-phase HPLC. Clearly, significant portions of ITP (Fig. 4A) and UTP (Fig. 4B) were hydrolyzed into respective nucleoside di-and monophosphates after a 60-min incubation, showing the broad specificity of pancreatic nucleotidase for various NTPs and NDPs.
The pattern of subsequent nucleotide hydrolysis by juice samples was then evaluated with 3 H-labeled ADP and AMP as initial tracer substrates, and the results are shown in Fig. 5. Pancreatic juice progressively hydrolyzed [ 3 H]ADP into [ 3 H]AMP/adenosine, and this catalytic reaction was markedly attenuated in the presence of 300 M ATP (Fig. 5, A and B). The addition of [ 3 H]AMP to pancreatic juice was also accompanied  by its gradual hydrolysis to [ 3 H]adenosine through 5Ј-nucleotidase reaction (Fig. 5C). Unlabeled ATP efficiently blocks the [ 3 H]AMP hydrolysis, presumably due to feed-forward inhibition of 5Ј-nucleotidase activity and/or concurrent activation of backward phosphotransfer reactions (3).
Quantitative radio-TLC analysis of pancreatic purine-converting enzymes was then performed using saturated concentrations of radiolabeled and unlabeled nucleotides/adenosine, and these results are summarized in Fig. 6. The data show the ability of rat pancreatic juice to efficiently hydrolyze [ 3 H]ADP and to about a similar extent as [ 14 C]ATP. Pancreatic juice also efficiently hydrolyzes [ 3 H]AMP by means of the 5Ј-nucleotidase activity. In contrast, incubation of juice with [ 3 H]adenosine did not cause its significant deamination to [ 3 H]nucleosides (Fig. 6A), indicating negligible activity of adenosine deaminase (also see above). Pancreatic adenylate kinase and NDPK activities were also determined by incubating juice samples with [ 3 H]AMP or [ 3 H]ADP as corresponding phosphate-accepting substrates in the presence of unlabeled ␥-phosphate-donating ATP/NTP (Fig. 6B). In support of qualitative autoradiographic data shown in Fig. 3 H]ATP is characterized by a higher rate and can be activated not only by ATP but also by other NTPs (Fig. 6B).
For comparative analysis, the major pancreatic activities shown in Fig. 6 were also expressed as nmol/ml of juice/hour and further correlated with soluble activities determined for two plasma samples. Mean activities for the following enzymes, ATPase, 5Ј-nucleotidase, adenosine deaminase, and adenylate kinase, were 2.1 and 0.29, 1.42 and 2.22, 0.07 and 0.20, 0.45 and 0.86 nmol/ml/hour for rat pancreatic juice and plasma, respectively.

DISCUSSION
The present study shows that pancreatic juice contains both ATP-consuming and ATP-generating enzymes. The following groups of enzymes were identified: NTPDase with broad substrate specificity, 5Ј-nucleotidase, NDPK, and adenylate kinase. No significant activities of NPP, nonspecific phosphatases, or adenosine deaminase were detected. The fact that these purinergic enzymes are secreted into pancreatic juice is a new observation, which indicates that the release of these enzymes may be regulated.  Mammalian NTPDases are a family of enzymes that catalyze hydrolysis of ␥ and ␤ residues of nucleotides with different specificities. To date, there are eight members of this family. The catalytic site is facing the extracellular milieu as in NTDPases 1-3 and 8 and/or intracellular organelles as in NTDPases 4 -7, and further, NTPDases 5 and 6 can be proteolytically cleaved from the plasma membrane and secreted (1, 20 -22). The NTPDases 1, 2, 3, and 8 catalyze hydrolysis of triphospho-and diphosphonucleotides with ATP:ADP rates of hydrolysis of about 1:1, 1:0.03, 3:1, and 2:1. The enzyme that we find in pancreatic juice can clearly hydrolyze ATP and ADP efficiently (Figs. 2, 5, and 6); ADPase activity is slightly higher and similar to the endothelial enzyme (3). There is no significant buildup of intermediate product ADP during ATP degradation (Fig. 2B). ATP and ADP compete for the same site (Figs. 2A and 5A), the enzyme has relatively broad specificity for NTP (Fig. 4), and it requires divalent cations (Fig. 3). Thus, the present study shows that the most likely candidate for pancreatic ATPase/ADPase is NTPDase 1, that is, CD39. This conclusion supports our earlier study where we detected CD39 by Western blotting on pancreatic juice collected from rats stimulated with CCK-8 (16).
Although CD39 is mainly regarded as a vascular enzyme, it has been originally identified in rat pancreatic tissue (23)(see Ref. 16) and isolated from membranes of zymogen granule of pig pancreas (24,25). There is ongoing discussion about whether CD39 and/or other members of the CD39 family are present in pancreatic tissue and whether distribution is speciesdependent, and most importantly, about possible function of these enzymes. In human pancreas, immuno-and histochemistry indicate that active NTPDase 1 (CD39) is not detected in acini or ducts (26). In recent studies on pancreatic tissue from mice, ATPase and ADPase activities determined by inorganic phosphate assays were significantly lower in CD39Ϫ/Ϫ mice when compared with control mice (17). Immuno-and histochemistry revealed that acini and ducts express both CD39 and CD39L1, i.e. NTPDases 1 and 2, but zymogen granules were not stained consistently. Nevertheless, earlier biochemical studies showed NTPDase activity on granule membranes (18,27,28). In our study where rat pancreas was prestimulated with CCK-8 before fixation, CD39 was redistributed to the secretory pole of acini (16), clearly ready to be secreted into pancreatic juice, as demonstrated by enzymatic analyses in the present study.
In addition to CD39, the enzyme hydrolyzing AMP to adenosine, i.e. 5Ј-nucleotidase, appears in pancreatic juice (Figs. 5 and 6A). This finding casts a light on an older study on rat pancreas AMPase activity that was detected histochemically, which showed localization of enzyme activity to different regions of acinar cells during the 24-h period (29). During day time, when secretory granules were accumulated, activity was seen in luminal and basolateral plasma membrane, as well as in intracellular organelles. During feeding/secretory phases, only basolateral marking was retained, suggesting that the enzyme was secreted or shed. Our study indicates that 5Ј-nucleotidase distribution within pancreas could be regulated since CCK-8 stimulation leads to secretion of 5Ј-nucleotidase into pancreatic juice. We propose that pancreatic 5Ј-nucleotidase, most likely CD73, serves a physiological role in epithelial function by providing adenosine for pancreatic ducts.
In contrast to airway epithelia and the cardiovascular system (see the Introduction), no significant activities of nonspecific alkaline phosphatase, NPP, and adenosine deaminase were detected in pancreatic juice. This would indicate that pancreas has a relatively specific set of nucleotide-handling enzymes destined for secretion. Some enzymes may also localize within pancreatic tissue.
Pancreatic juice also contained moderate adenylate kinase and NDPK activities (Figs. 3 and 6B), potentially interconverting extracellular nucleotides via backward phosphotransfer reactions. Our finding on regulated secretion of these kinases is supported by findings in another exocrine gland. Namely, cholinergically stimulated submucosal glands in airway epithelia also secrete adenylate kinase and NDPK (30), and on apical surfaces their avid activities would counteract ATP hydrolysis and thus propagate purine-mediated mucociliary clearance. In pancreatic juice, these ATP-generating enzymes have lower activities than ATP-hydrolyzing enzymes (Fig. 6), suggesting that adenosine receptor-mediated signaling would be important (see below).
Where do the enzymes come from, and are they really active in situ? Immunohistochemical studies showed that CD39 is localized in acini, mainly in granular compartments, and also in ducts. Since CCK-8 is the major stimulant of acini, presumably it is secreted from acini. Secretin, the ductal agonist, did not cause release of CD39 (16). Very likely then, other enzymes determined in this study (NTDPase1/CD39, 5Ј-nucleotidase, adenylate kinase, and NDPK) could also have originated in the acini. Since at least CD39 is secreted as a full enzyme and not cleaved one, the question is whether we can call these enzymes "soluble enzymes" or whether they are associated with postulated microvesicles (16), such as those known particularly in immune and hemostatic systems (31,32). Both types of release could be associated with zymogen granules, where at least CD39 immunolocalization is strong (16). In any case, enzyme release, however it happens, seems to be regulated. In pancreatic juice, where normal concentrations of free Ca 2ϩ and Mg 2ϩ are submillimolar (33,34), the conditions for enzymes would be suboptimal when compared with in vitro enzymatic assays. Nevertheless, since the juice emerging out of the organ/cannula has relatively low ATP concentrations when compared with estimated concentrations released from acini (Fig. 1), some ATP is degraded either by secreted enzymes or possibly also by ecto-enzymes lining ducts. We estimate that in the rat pancreas, the ductal system plus cannula would occupy a volume of around 50 l. Thus, with secretion rates of about 3 l/min, there would be sufficient "contact time" for enzymes to modify secretion.
The physiological implications for pancreas are as follows. CCK-8 stimulation of pancreatic acini leads to secretion of ATP (Fig. 1). It is not excluded that there are also other sites for ATP release within pancreas. In addition, CCK-8-stimulated secretion also contains CD39, CD73, and kinases, as shown by the present study. On one hand, relatively large activities of CD39 and 5Ј-nucleotidase would favor ATP hydrolysis and production of adenosine. On the other hand, there is also a possibility to generate ATP by adenylate kinase and by trans-phosphorylation of ADP to ATP, where other NTPs can be used as phosphate donors. Thus, P2 receptors with a preference for NTP over NDP, and P1 receptors, would have possibilities to be stimulated. Indeed, ducts from adult rat pancreas express the ATP/ UTP-preferring receptors, P2X 4 , P2X 7 , P2Y 2 , and P2Y 4 , but not the ADP/UDP-preferring receptors, P2Y 1 or P2Y 6 (35). In addition, ducts also express a number of adenosine receptors (preliminary studies). The nucleotide-and nucleoside-selective receptors are most likely involved in regulation of bicarbonate and fluid secretion occurring in pancreatic ducts (13,16). This newly discovered richness in secreted purine-handling enzymes underscores the importance of acini-to-duct communication and of P2 and P1 receptor signaling along pancreatic duct lumen. Most likely, the secretory profiles of ATP-generating versus ATP-hydrolyzing enzymes, as well as ATP release, may depend on the extent and duration of stimulation in this complex organ. Accordingly, P2 and P1 receptor distribution may vary with the generation of ducts.
In conclusion, enzyme assays on pancreatic juice in this study, together with our previous Western blotting data (16), demonstrate the presence of specific Ca 2ϩ -Mg 2ϩ -dependent soluble enzyme with hallmark characteristics of NTPDase/ CD39, which has a broad substrate specificity toward various nucleoside tri-and diphosphates. We have shown the presence of yet another soluble pancreatic nucleotide-hydrolyzing enzyme, 5Ј-nucleotidase, and in addition, provided kinetic evidence for the existence of moderate adenylate kinase and NDPK activities potentially interconverting extracellular nucleotides via backward phosphotransfer reactions. Soluble adenosine deaminase, nucleotide pyrophosphatase, and nonspecific phosphatase do not seem to contribute to the purine metabolism in the rat pancreatic juice. The given complement of enzymes may be important in purine signaling within pancreas and thus coordination of pancreatic secretion on the whole organ level.