Rat pancreas secretes particulate ecto-nucleotidase CD39

Abstract

In exocrine pancreas, acini release ATP and the excurrent ducts express several types of purinergic P2 receptors. Thereby, ATP, or its hydrolytic products, might play a role as a paracrine regulator between acini and ducts. The aim of the present study was to elucidate whether this acinar-ductal signalling is regulated by nucleotidase(s), and to characterize and localize one of the nucleotidases within the rat pancreas. Using RT-PCR and Western blotting we show that pancreas expresses the full length ecto-nucleoside triphosphate diphosphohydrolase, CD39. Immunofluorescence shows CD39 localization on basolateral membranes of acini and intracellularly. In small intercalated/ interlobular ducts, CD39 immunofluorescence was localized on the luminal membranes, while in larger ducts it was localized on the basolateral membranes. Upon stimulation with cholecystokinin-octapeptide-8 (CCK-8), acinar CD39 relocalizes in clusters towards the lumen and is secreted. As a result, pancreatic juice collected from intact pancreas stimulated with CCK-8 contained nucleotidase activity, including that of CD39, and no detectable amounts of ATP. Anti-CD39 antibodies detected the full length (78 kDa) CD39 in pancreatic juice. This CD39 was confined only to the particulate and not to the soluble fraction of CCK-8-stimulated secretion. No CD39 activity was detected in secretion stimulated by secretin. The role of secreted particulate, possibly microsomal, CD39 would be to regulate intraluminal ATP concentrations within the ductal tree. In conclusion, we show a novel inducible release of full length particulate CD39, and propose its role in the physiological context of pancreatic secretion.

Exocrine glands are some of the first organs where ecto-nucleotidases and purinergic receptors were described (Gallacher, 1982; Plesner 1995; Novak, 2003). The common substrates/ligands for these proteins are extracellular nucleotides, but the link between these two proteins, as well as the physiological context for exocrine secretion, is missing. In our recent study we have shown that pancreatic ducts possess functional P2 receptors on the basolateral and luminal membranes (Christoffersen et al. 1998; Hede et al. 1999). Acini, on the other hand, show only a few functional P2 receptors (Novak et al. 2002). However, acini release ATP in response to several stimuli, most importantly to cholinergic stimulation (Sørensen & Novak, 2001). The ATP release may be transient, but in addition, ecto-nucleotidases may hydrolyse the released ATP, as indicated by our earlier study (Sørensen & Novak, 2001).

Extracellular nucleotides can be hydrolysed by many enzymes including ecto-alkaline phosphatases, ecto-5′-nucleotidases, and ecto-nucleotidases with more distinct characteristics that are further classified into two families. Ecto-nucleoside triphosphate diphosphohydrolases that hydrolyse nucleoside 5′ tri- and diphosphates, now named as NTPDases, but often described as the CD39 family, include ecto-apyrase, ecto-ATPase and others (Vanduffel & Lemmens, 2000; Zimmermann, 2000). From the membrane topography it is predicted that the NTPDases 1–4 have two transmembrane domains with intracellular N- and C-terminals, and a large extra-membranous (extracellular) catalytic region with five conserved apyrase regions and several glycosylation sites. NTPDases 5 and 6 have the N-terminal in the membrane, while the rest of the molecule is extra-membranous and can be cleaved to yield a soluble protein. NTPDases 1–3 are bound to the plasma membranes, while NTPDases 4–6 can be associated with intracellular organelles (Braun et al. 2000; Zimmermann, 2000). Another family of enzymes, ecto-nucleotide pyrophosphatase/ phosphodiesterases (NPP), has broad substrate specificity. They can hydrolyse phosphodiester bonds of nucleotides and nucleic acids and pyrophosphate bonds of nucleotides and nucleotide sugars. These enzymes, similar to NTPDase5 (CD39L4), have the N-terminal embedded in the membrane, while the catalytic site and the C-terminal face outwards, and they can be cleaved by proteolytic enzymes to yield soluble proteins (Mulero et al. 1999; Zimmermann, 2000).

In the 1980s, Ca²⁺,Mg²⁺-requiring ecto-nucleotidase activity was detected in rat pancreas and parotid glands (Harper et al. 1978; Lambert & Christophe, 1978; Martin & Senior, 1980; Hamlyn & Senior, 1983), and the ecto-nucleoside triphosphate diphosphohydrolase (NTPDase1) was purified from the pig pancreas (LeBel et al. 1980; Lalibertéet al. 1982; Sévigny et al. 1995). NTPDase1 was shown to be identical to the lymphocyte surface marker CD39 (Kaczmarek et al. 1996; Wang & Guidotti, 1996), which is mainly thought of as a vascular enzyme. Nevertheless, apart from the pancreas, the same or similar enzymes are also found in the pig liver, rat salivary and mammary glands, human placenta, and pig kidney and duodenum (Valenzuela et al. 1989; Kaczmarek et al. 1996; Leclerc et al. 2000; Lemmens et al. 2000; Sévigny et al. 2000). Concerning pancreas, the physiological function of CD39 and its localization in the pancreas remained elusive. In the initial studies the enzyme (then taken to be a Ca²⁺,Mg²⁺-requiring ATPase) was studied in connection with Ca²⁺ and HCO₃⁻ transport in pancreas (Lambert & Christophe, 1978; Martin & Senior, 1980; Hamlyn & Senior, 1983). In biochemical assays and/or cytochemical studies, the enzyme was found in the plasma membranes, where it had various levels of activity (Lambert & Christophe, 1978; LeBel et al. 1980; Hamlyn & Senior, 1983), and also in zymogen granule membranes (Harper et al. 1978; Martin & Senior, 1980; Sévigny et al. 1998; Lemmens et al. 2000), from where it has been isolated (Lalibertéet al. 1982). Yet immunohistochemically, using antibodies for CD39, the enzyme was found to be diffusely distributed on the basolateral membranes and very weakly localized in zymogen granules of pig acinar cells (Sévigny et al. 1998). Furthermore, the pancreatic CD39 was an apparently smaller, perhaps truncated or cleaved form of the enzyme found in other tissues (Sévigny et al. 1995, 1998; Lemmens et al. 2000).

The aim of the present study was to show whether ecto-nucleotidase activity is related to the secretory processes in exocrine pancreas, and to characterize it and localize it within the rat pancreas. For this purpose we used Western blot and RT-PCR analysis, immunohistochemistry and microscopy, as well as assays for ATPase and ADPase activity in pancreatic juice collected from intact pancreas. Our findings show that the full length CD39 has a distinct pattern of localization within pancreas; it translocates during stimulation of pancreas and appears in the particulate fraction of pancreatic juice, most probably associated with microvesicles. Hence, we propose that the marked nucleotidase activity in pancreatic juice, which we attribute to CD39, would contribute to regulation of the concentration and composition of nucleotides that reach the downstream pancreatic ducts equipped with both P2 and adenosine receptors. Parts of the work have been published in reports (Neumann et al. 1998; Novak et al. 2000).

METHODS

Materials

All standard chemicals including collagenase, sera, secretin and cholecystokinin-octapeptide-8 (CCK-8) were obtained from Sigma (Copenhagen, Denmark). Secondary antibodies conjugated to Cy2 and Cy3 fluorophores were from Jackson ImmunoResearch Laboratories (West Grove, PA, USA). Tissue culture media and PBS were from GIBCO (Invitrogen, Denmark). Luciferase and luciferin were from Sigma or Roche Diagnostics (Mannheim, Germany). Mebumal was from Nycomed (Roskilde, Denmark). Pyruvate kinase, myokinase, phosphoenolpyruvate, triethanolamine, ATP, ADP and AMP were from Roche Diagnostics. Total RNA Isolation Reagent and RT and PCR kits were obtained from Advanced Biotechnologies (Epsom, UK). PCR primers were obtained from MWG-Biotech (Ebersberg, Denmark).

Collection of pancreatic juice

For several different experiments in vivo preparation of pancreas was performed. Female Wistar rats weighing 160–300 g were used. The necessary permission was obtained from the Danish Animal Ethical Committee. The animals, fasted for 16–24 h, were anaesthetized with Mebumal (pentobarbital, 40 mg kg⁻¹ I.P.). The facial vein was cannulated for infusions. Anaesthesia was maintained during the experiments by additional intravenous injections of Mebumal. A thermostatically controlled heating table was used to maintain the animal's body temperature at 38 °C. The animals were tracheostomized and the abdomen was opened by a midline incision. The pylorus and the proximal end of the bile duct were ligated. A polyethylene tube was inserted into the common pancreatic bile duct. Collection of pancreatic juice was started with a control period of 30–60 min during which medium (DMEM1000/Ham's F12 medium) was infused. Secretion was then stimulated by either infusion of CCK-8 (1.5–3 pmol min⁻¹ per 200 g animal) or secretin (10 pmol min⁻¹ per 200 g animal) for about 60 min. The infusion rate for each animal (0.03 ml min⁻¹) was held constant with a syringe pump (Cole-Parmer). Pancreatic juice was collected on ice over 10–15 min periods and stored at −80 °C. For some experiments the juice was diluted 1:2 in bicarbonate-free Ringer (-BIC) containing (mM): Na⁺ 145, K⁺ 3.6, Ca²⁺ 1.5, Mg²⁺ 1, Cl⁻ 122, gluconate 25, phosphate 2, glucose 5, levamizole 1, ouabain 2, pH 7.4, and ultracentrifuged at 210 000 g for 1 h at 4 °C in a Beckman ultracentrifuge. After experiments animals were killed by overdose of Mebumal.

Measurement of ATP, ATPase activity and breakdown products in pancreatic juice

ATP in samples of pancreatic juice was analysed by the luciferase method performed by continuous registration of the luminescence signal upon addition of the luciferin-luciferase reagent to the sample (Rasmussen & Nielsen, 1968; Sørensen & Novak, 2001). The initial luminescence increment is linearly related to the sample content of ATP in the range of 10 fmol to 1 nmol. Some luminescence increment in the blank was registered due to the reagent luminescence; this increment was small and reproducible. The signal increments were calibrated by analysis of ATP standards. Kinetic experiments were performed by mixing pancreatic juice with a solution of ATP in 50 mM Tris-HCl, 5 mM MgCl₂, pH 8.0 and analysing ATP after different times. In some experiments supernatants or pellets from ultracentrifuged pancreatic juice were mixed with ATP (0.1 and 0.01 μM) in 50 mM Tris-HCl, 5 mM MgCl₂, 2 mM ouabain, 1 mM levamizole, pH 7.5. The incubation was stopped by boiling for 1 min and the ATP was analysed. The breakdown products, ADP and AMP, were estimated after enzymatic conversion to ATP. AMP + ADP + ATP was determined (as ATP) after 3–5 min preincubation of the sample which was mixed with an equal volume of 275 mM triethanolamine-HCl, 10 mM MgCl₂, 30 mM KCl, 0.25 mM phosphoenolpyruvate, 50 U ml⁻¹ pyruvate kinase, 70 U ml⁻¹ adenylate kinase, pH 7.6. The reagent for ADP + ATP was similar, except that adenylate kinase was omitted. All experiments were performed at 25 °C.

RT-PCR on whole pancreas and isolated pancreatic acinar cells

Pancreases were obtained from female rats killed by cervical dislocation. Acini cells were isolated by collagenase digestion as described earlier (Sørensen & Novak, 2001). To ensure that there was no contamination from ducts and non-exocrine tissue, single acini were collected one by one under a microscope. The purity of the acinar preparation was checked by using primers for amylase, carbonic anhydrase and thrombin as described earlier (Hede et al. 1999). Total RNA from whole pancreas and acinar cells was prepared using the Total RNA Isolation Reagent according to the manufacturer's instructions. An anchored oligo dT primer from Reverse-iT was used for synthesis of first strand complementary DNA (cDNA), which served as the template for PCR amplification. The primers used for PCR were designed based on the rat CD39 sequence (GenBank accession no. U81295). The forward primer sequence was: 5′-ACTGACCCACAACAAACC and the sequence of the reverse primer was: 5′-CCTGATCCTTCCCATAGC, giving a product of 670 bp. A volume (2 μl) of cDNA was added to a solution of 10 μM each of forward and reverse primers. It was then mixed with 1.2 mM MgCl₂, PCR buffer, 0.2 mM dNTPs and 1.25 U Taq polymerase, and dH₂O was added to a total volume of 50 μl. Amplification was performed for 25 cycles of 30 s at 94 °C, 60 s at 50 °C and 90 s at 72 °C, followed by 10 min at 72 °C. The PCR products were then examined on a 1 % agarose gel containing ethidium bromide.

Cell lysates and Western blotting

For Western blot analysis, pancreatic juice and acinar cells were collected as described above. Pieces of whole pancreas, brain, liver and adipose tissues were obtained from female Wistar rats. Acinar cells and tissues were cut into small pieces and homogenized in 500 ml lysis buffer (20 mM Tris-acetate pH 7.0, 0.25 M sucrose, 1 mM EDTA, 1 mM EGTA, 1 mM sodium orthovanadate, 50 mM NaF, 1 % Triton X-100, 1 mM DTT), followed by removal of debris by centrifugation. For immunoprecipitation of CD39, 2 μl of monoclonal antibody against CD39 (Zymed Laboratories, CA, USA) was added to 100 μl acinar cell lysates and 50 μl undiluted pancreatic juice, respectively, and incubated overnight at 4 °C. A 20 μl volume of a 50 % slurry of Protein A-Sepharose (Protein A-Sepharose CL-4B Amersham Biosciences, Uppsala, Sweden) and PBST (phosphate-buffered saline containing 40 mM NaH₂PO₄, 10 mM Na₂HPO₄, 0.1 % Tween 20) was added to a total volume of 1 ml. The samples were incubated for 60 min at 4 °C on rotation, followed by washing in with PBST. Then the samples were boiled for 5 min in 2 × Laemmli buffer and subjected to SDS-PAGE together with cell lysates from the other tissues. The proteins were separated on a 7.5 % gel and subsequently transferred to a nitrocellulose membrane (Schleicher and Schüll, Denmark). The membrane was blocked for 1 h in PBST containing 5 % skimmed milk powder, followed by washing with PBST. It was then incubated overnight at 4 °C with a mouse anti-CD39 antibody (Zymed). Immunoreactive proteins were visualized by the use of horseradish-peroxidase-conjugated goat anti-mouse IgG (BioRad, Hercules, CA, USA) and Enhanced ChemiLuminescence assay (Amersham Biosciences) according to the manufacturer's instructions. Multicolored Protein Markers (NEN, Boston, MA, USA) were used as a molecular weight marker. For the detection of CD39 in total pancreatic juice, pellet or supernatant, the juice or fractions were boiled directly in 2 × Laemmli buffer prior to SDS-PAGE. For these experiments a guinea-pig anti-rat CD39 antibody raised against residues 54–65 was used (Neuromics, MN, USA). For the detection of Na⁺,K⁺-ATPase α1 subunit in either whole pancreas or juice, total cell lysate from pancreas, total juice or fractions was pre-incubated in 2 × Laemmli buffer for 20 min at 37 °C before application to SDS-PAGE. An anti-rat Na⁺,K⁺-ATPase α1 subunit antibody raised against residues 338–518 was used for detection of Na⁺,K⁺-ATPase (Upstate Biotechnology, NY, USA).

Immunolocalization of CD39

Immunolocalization was performed on pancreatic tissue prepared by several different techniques: paraffin section of pancreas, isolated acini and ducts and cryosections of pancreas. In addition, some pancreases were pre-stimulated in vivo, as described above. Pancreas from in vivo preparations was transferred to -BIC, containing 0.25 mg ml⁻¹ trypsin inhibitor, cut into small pieces and fixed with 4 % paraformaldehyde for 24 h. Following dehydration through ethanol, it was stored overnight in methylbenzoate and then transferred to xylene for 2.5 h. Finally the tissue was embedded in paraffin for 1 h and sectioned into 5–7 μm slices. For immunostaining, the paraffin sections were rehydrated and non-specific binding was blocked with 10 % normal goat serum (NGS) for 15 min. Isolated acini and ducts were prepared by collagenase digestion as described (Sørensen & Novak, 2001). The cells were suspended in DMEM-1000/F12 medium containing 0.25 mg ml⁻¹ trypsin inhibitor and hexokinase (1.8 U ml⁻¹). Subsequently, single acini and ducts identified with a dissection microscope were collected and fixed for 1 h in PBS containing 4 % paraformaldehyde, followed by three washes in PBS and blocking with 20 % NGS for 1 h. Pancreas sections or isolated acini/ducts were then incubated with monoclonal antibody (mAb) anti-CD39 from Zymed (1:100) for 1 h, washed in PBS several times and incubated for 30–60 min with goat anti-mouse antibody, conjugated to the fluorophore Cy2 (1:50) or goat anti-mouse antibody (with minimal cross-reactivity to rat) conjugated to Cy3 (1:100). The preparations were then washed in PBS and mounted with DABCO mounting medium (Sigma, Copenhagen, Denmark). For preparation of cryosections, pancreas was fixed with PLP fixative (10 mM NaIO₄, 75 mM lysine, 30 mM sucrose and 4 % paraformaldehyde, without dextran). The organs were removed from the animals and washed in PBS. They were cut into small pieces (ca 3–4 mm³), transferred to PBS containing 30 % sucrose and 0.02 % NaN₃ and rinsed a further four times in the same buffer. After the last washing the tissues were stored overnight at 4 °C in PBS containing 30 % sucrose and then embedded in Cryoembed (AX-LAB A/S, Denmark). Tissues were frozen for a short time at −30 °C, then transferred to liquid nitrogen and sectioned into 6–7 μm slices. The slices were dried for 30 min at room temperature and then stored at −20 °C. The cryosections were rehydrated in PBS and the rest of the immunostaining procedure was the same as described above. Fluorescence was examined using a TCS NT/SP confocal laser scanning microscope equipped with Ar-Kr and × 40 1.25 NA, × 100 1.4 NA and × 63 1.2 NA PL APO objectives (Leica Microsystems Heidelberg GmbH, Germany). Images were analysed using Leica CLSM software or MetaMorph 5.0 software (Universal Imaging Corporation, West Chester, PA, USA).

RESULTS

Pancreatic juice contains no ATP but nucleotidase activity

Since the isolated pancreatic acini release some 10–20 μM ATP in response to agonists (Sørensen & Novak, 2001), the aim of the first set of experiments was to determine how much ATP appeared in pancreatic juice collected from the common pancreatic duct of intact pancreas. In anaesthetized rats, pancreatic secretion was evoked by CCK-8 or secretin (n = 4 and 4, respectively) and a secretory response for one series is shown in Fig. 1A. From each experiment, two samples from the period of maximal secretion were analysed for ATP. The luminescence increments with 10 μl samples were not different from those of the reagent blank (see Fig. 1B, left part, for an example). Expressed in terms of amounts of ATP, the sample increments corresponded to 28 ± 2 fmol (n = 16, mean ± S.D.) and the blank increments to 30 ± 3 fmol (n = 6). These numbers are not significantly different and a conservative estimate of the detection limit indicates that the ATP content in 10 μl samples was below ˜10 fmol. Thus, the ATP concentration of the pancreatic juice was below ˜1 nM, or ˜0.01 % of the estimated ATP secretion from acini.

Figure 1. ATP concentrations in pancreatic juice.

A, the secretory response of rat pancreas stimulated with cholecystokinin-octapeptide-8 (CCK-8) at 3 pmol min⁻¹ per 200 g rat (means ± S.E.M., n = 4). Similar secretory response was observed with secretin (data not shown). B, the analysis of ATP: records of the luminescence signal upon addition of the luciferin-luciferase reagent to the 10 μl samples. Reagent blank and pancreatic juice show luminescence increments corresponding to 29 and 27 fmol ATP, respectively. Buffered ATP solution gave an increment corresponding to 346 fmol. Pancreatic juice hydrolysed ATP in the buffered ATP solution to 195 fmol after 30 s and 27 fmol after 300 s. Downward deflections between traces are due to changing of cuvettes.

It seemed likely that very low amounts of ATP in the final juice could be due to the concurrent ATPase activity in the juice. This was supported by experiments where boiling of the juice eliminated its ability to break down the ATP standard (data not shown). Further, the pancreatic juice was tested for its potential ATPase activity by incubation with ATP (Fig. 1B). The ATP concentration was reduced depending on the duration of incubation. The time course of the hydrolysis is shown in more detail in another type of experiment depicted in Fig. 2A. Taken together, these experiments show that the pancreatic juice contains an ATP hydrolase activity.

Figure 2. The time course of the ATPase activity in pancreatic juice and estimation of the breakdown products.

A, ATP concentrations were determined by the luciferin-luciferase assay. Pancreatic juice (1 μl) from CCK-8-stimulated pancreas was added to the ATP standard (19 μl). Points were fitted with an integrated Michaelis-Menten equation and K_m and V_max were estimated to be 55 μM and 0.130 μM s⁻¹, respectively. B, nucleotidase activity in pancreatic juice. Breakdown products of ATP estimated after 0, 5 and 16 min incubation with pancreatic juice. ADP and AMP were enzymatically converted to ATP and estimated as above. The figure shows the percentage of different nucleotides in relation to the start concentration of ATP, which was 19.3 μM.

In the next series of experiments, buffered ATP of higher concentration was incubated with the pancreatic juice obtained from CCK-8-stimulated pancreas and the breakdown products ADP and AMP were determined at different times and compared with the initial concentration of the ATP standard (Fig. 2B). At the beginning of the experiment the ATP concentration was 19.3 μM. After 5 min of incubation with the juice, ATP was reduced to 56 %, and 24 % of ATP was converted to AMP. No measurable amounts of ADP were detected at this time. After 16 min incubation, the initial ATP was reduced to 13 %, there was 34 % of AMP and only 4 % of ADP was detected. Figure 2B also shows that AMP presumably disappeared from the mixture, as the total sum of nucleotides after 5 and 16 min incubation was 80 and 51 %, respectively, of the original amount ATP. These results indicate that pancreatic juice contains one (or more) nucleotidases that break down ATP and ADP. Since ADP was hardly detected, it is possible that the two phosphodiester bonds of ATP were sequentially hydrolysed and AMP was released, as found for the native and heterologously expressed membrane bound CD39, in contrast to soluble CD39 (LeBel et al. 1980; Laliberté & Beaudoin, 1983; Heine et al. 2001). Disappearance of AMP indicates that the juice also contains 5′-nucleotidase activity that hydrolyses AMP to adenosine.

Nucleotidase activity is associated with a particulate fraction of pancreatic juice

Most of the ecto-nucleotidase activities in various cells are due to enzymes embedded in the plasma membrane. The biochemistry of the NTPDase1 catalytic activity is well described for the pancreatic tissue (see Introduction), but our observation that a similar hydrolytic activity is found in pancreatic juice is new. Recently, several reports show that some nucleotidases (NTPDases 5 and 6; and NPP1 to NPP3) may be released into the surrounding medium of various cells (Mulero et al. 1999; Bollen et al. 2000; Braun et al. 2000; Chen & Guidotti, 2001). Also sympathetic nerve stimulation causes release of soluble NTPDase along with ATP (Todorov et al. 1997; Westfall et al. 2000). Since pancreatic juice contains nucleotidase activity, we investigated whether it was the soluble enzyme that was responsible for the hydrolytic activity. Pancreatic juice collected from animals stimulated with CCK-8 was ultracentrifuged and the supernatant and the pellet were assayed for the nucleotidase activity. Figure 3 shows that the supernatant had a very weak nucleotidase activity. In contrast, the pellet hydrolysed ATP in the given standard to less than 1 % within 5 min. Since ouabain and levamizole were included in the samples, the activity was not due to the Na⁺,K⁺-ATPase or alkaline phosphatase.

Figure 3. Nucleotidase activity in the pellet and the supernatant of centrifuged pancreatic juice.

After 5 min incubation, the supernatant (S) reduced the ATP standard to 81 %, while the pellet (P) reduced it to < 1 %. The supernatant had very weak time-dependent nucleotidase activity as depicted in columns 2 and 3. The pellet reduced ATP concentrations in the ATP standard depending on the incubation time (columns 4 and 5), but after 5 min all ATP was essentially hydrolysed. The control values were the ATP standards (0.1 or 0.01 μM) to which -BIC solution was added instead of the pancreatic juice.

CD39 in pancreas: Western blot and RT-PCR

In the next series of experiments, we investigated whether the rat pancreatic tissue expressed one type of ecto-nucleoside triphosphate diphosphohydrolase, NTPDase1, also known as CD39. We used the monoclonal antibody and immunoblotting. Figure 4A shows a Western blot of various tissues, including pancreas. Since the whole pancreas would also include non-exocrine tissue, such as blood vessels that express CD39, we also tested immunoprecipitated samples containing ‘clean’ acini (see Methods). Notably, the pancreatic tissue and isolated acini show a prominent band at about 78 kDa, which is consistent with the full length CD39. Such a prominent band was not previously detected in the pancreas. In addition, there are other bands between 78 and 55 kDa, which are probably the deglycosylated forms of the enzyme, or possibly the lowest band is the protease degradation product. Furthermore, there is a weak band at around 85 kDa (also see Fig. 6).

Figure 4. CD39 in pancreas and other tissues.

A, Western blot with monoclonal antibody (mAb) CD39 of various rat tissues including pancreas and pure pancreatic acini. Pancreatic acini were immunoprecipitated (IP). The full size monomeric CD39 is 78 kDa. B, RT-PCR analysis of the CD39 expression in the whole pancreas and pure pancreatic acini. The band at 670 bp is expected for the CD39 transcript. The gels are representative of at least 3 different experiments.

Figure 6. Pancreatic juice contains CD39 in the particulate fraction.

A, pancreatic juice collected from a pancreas stimulated with secretin and later with CCK-8 was analysed for CD39 with Western blot. Samples for analysis were collected 15–45 min after stimulation. Lanes 2 and 3 show the untreated total juice; lanes 5 and 6 show the ultracentrifuged pellet and lanes 7 and 8 show the supernatant. Agonists are indicated in each lane. In addition, lanes 1 and 4 show the Western blot of whole pancreas tissue. B, Western blot with an antibody against the α subunit of the Na⁺,K⁺-ATPase was performed on total juice from secretin and CCK-8 stimulated pancreas (lanes 10 and 11) and as a positive control on the whole pancreas (lane 9). The gel is the representative of at least 3 different experiments.

Using specific primers for rat CD39, we further tested whether rat pancreas contains transcripts for CD39 using RT-PCR. Figure 4B shows bands at the expected 670 bp for the CD39 transcript (see Methods). Most importantly, the clean acini, devoid of ducts and non-exocrine tissue, also show transcripts for CD39. The purity of the acinar preparation was routinely checked as described earlier using RT-PCR and primers for amylase (acinar marker), thrombin (blood vessel marker) and carbonic anhydrase (ductal marker) prior to testing of RNA for CD39 (data not shown) (Hede et al. 1999).

Immunolocalization of CD39

The immunolocalization of CD39 was carried out on various preparations of pancreas-paraffin sections, cryosections, and isolated acini and ducts from collagenase digests to ensure that localization was not influenced by the method of preparation. The secondary antibody was coupled to horseradish peroxidase (results not shown), Cy2 or Cy3 and viewed with a confocal laser scanning microscope (Fig. 5). Pancreatic ducts showed an immunofluorescence, which in some ducts was associated with the luminal membrane (Fig. 5A and B), and in other ducts with the basolateral membrane (Fig. 5C and D). Closer analysis indicated that CD39 was localized luminally in the smaller ducts (< 10 μm in outside diameter), belonging to intercalated and small intralobular generation of ducts, while it was localized basolaterally in larger ducts (10–30 μm in outside diameter), belonging to the intra-extralobular generation of ducts (n = 9). The blood vessels also showed the expected CD39 expression (data not shown). Pancreatic acini from unstimulated pancreas showed fluorescence on the basolateral membranes (Fig. 5I and K) and intracellularly, often in regions occupied by zymogen granules (Fig. 5I). In another series of experiments, pancreas was stimulated with CCK-8 prior to immunolocalization of CD39. Figure 5J and L shows that the pattern of fluorescence changed in acini; it became more prominent close to the apical regions, often in clusters close to the region occupied by zymogen granules. In some cases fluorescence was detected in the acinar lumen (not shown). In control preparations, where the primary antibody (Fig. 5Q) or the secondary antibody (Fig. 5S) was omitted, essentially no fluorescence was detected.

Figure 5. Immunolocalization of CD39.

Immunofluorescence images of the CD39 distribution in rat pancreatic ducts (A-D) and acini (I-L) with corresponding transmission or differential interference contrast (DIC) images (E-H, M-P) and controls (Q-T). In the small duct the immunofluorescence was localized luminally (arrows in A and B), whereas in the large ducts it was localized basolaterally (arrows in C and D). In the acini the immunofluorescence was on the basolateral membranes (arrows in I and K) and intracellularly (double arrow in I). The images (J and L) were from pancreas stimulated with CCK-8 and the immunofluorescence had moved towards the luminal pole of acini (double arrows in J and L). Preparations for staining were from paraffin sections of pancreas (A, B, I, J, Q and S), isolated acini and ducts from collagenase digests (C and D) or cryosections of pancreas (K and L). The secondary antibody was conjugated to Cy2 or Cy3 (green and red, respectively). In control preparations, where only primary (Q) or secondary antibody (S) was used, no fluorescence was detectable with the same scanning settings. Corresponding transmission images for controls are in R and T. In all images the bar is 20 μm. Data are representative images from 12 independent staining experiments.

CD39 in pancreatic juice is associated with the pellet fraction

Since the rat pancreas shows a distinct immunolocalization of CD39, and it appears that the enzyme moves towards the lumen upon stimulation, the next series of experiments were designed to localize the enzyme in pancreatic juice. Pancreatic juice was analysed for CD39 with an antibody directed against residues 54–65 of the rat enzyme (Neuromics), since the Zymed antibody gave only a weak reaction with pancreatic juice. Figure 6 shows that the antibody reacted positively with bands of 78 and 85 kDa in the whole pancreas. The same bands were detected in total pancreatic juice of animals stimulated with CCK-8, but not in those stimulated with secretin (Fig. 6A). The large part of the activity was associated with the pellet (lane 6) and not with the supernatant (lanes 7 and 8). Notably, in the pellet we detected the full CD39 enzyme form of 78 kDa. The band at 85 kDa might be due to variation in glycosylation or palmitoylation of the enzyme (Schulte am Esch et al. 1999; Koziak et al. 2000). If the soluble CD39 was released into the supernatant, the 54 kDa form would have been expected, as seen in pig adrenals (Lemmens et al. 2000). In order to check that the 78 kDa band was not due to the plasma membrane from damaged cells that were shed into the juice during collections, we performed a Western blot with the antibody against the α subunit of the Na⁺,K⁺-ATPase. This antibody gave no reaction in the juice. The pancreatic tissue was positive and showed the predicted band for the intact Na⁺,K⁺-ATPase at 110 kDa (Fig. 6B). Together, these results show that stimulation by CCK-8 induced secretion of the particulate full length CD39 that was not ‘cellular debris’.

DISCUSSION

Our earlier study showed that pancreatic acini secrete substantial amounts of ATP upon stimulation (Sørensen & Novak, 2001), yet in the present study we could not detect ATP in pancreatic juice collected from the common duct using similar methods for measurements of ATP (Fig. 1). This dilemma is resolved by the finding that pancreatic juice actually contains high nucleotidase activity. In the following sections we will argue that a large part of this nucleotidase activity is due to CD39 that is released in a particulate fraction into pancreatic juice upon stimulation with acinar agonist CCK-8, and speculate about the physiological relevance of CD39 in purinergic signalling in pancreas.

Where has all the ATP gone? Ecto-nucleotidase in pancreatic juice

Pancreatic juice contains one or more enzymes that can break down exogenously added ATP (Fig. 1A and Fig. 2). Since pancreas secretes many digestive enzymes, or rather pro-enzymes, there are probably several candidates, and one could immediately suggest phosphatases or glycoprotein 2 (GP-2). Nevertheless, phosphatases have K_m values in the millimolar range, and we included levamizole in our assay. GP-2 nucleotidase activity requires a glycosylphosphatidylinositol (GPI) anchor for its activity, which it loses when associated with the fibrillar network in pancreatic juice (Grondin et al. 1992; Soriani & Freiburghaus, 1996). The nucleotidase in pancreatic juice that we are seeking should have a K_m in the micromolar range (Fig. 2), which would match 10–20 μM ATP released from acini upon stimulation (Sørensen & Novak, 2001). Since ADP does not accumulate, possibly one enzyme hydrolysing both ATP and ADP could account for the reaction. Although it is likely that pancreas would express multiple ecto-nucleotidases as do other cells (Vollmayer et al. 2001; Dranoff et al. 2002), the simplest postulate is that all characteristics deduced from our experiments speak in favour of CD39, previously detected in pancreatic tissue, but never in secretion (see below and Introduction). Interestingly, the membrane-bound native or heterologously expressed CD39 does not release ADP but hydrolyses it, thus preventing transient accumulation of ADP (LeBel et al. 1980; Laliberté & Beaudoin, 1983). Similar lack of ADP accumulation is also seen by the pancreatic nucleotidase (Fig. 2B). In contrast, the soluble form of CD39 hydrolyses ATP and ADP sequentially and ADP would accumulate transiently (Chen & Guidotti, 2001). Another interesting observation about the pancreatic nucleotidase is the fast rate of ATP degradation (Figs 1B, 2A and 3). The passage of secretion through the duct system and collection of pancreatic juice samples take some minutes. From the measured secretory rates (Fig. 1) and the duct volume of about 5 % of the pancreas, we can calculate that it would take 5–10 min for the newly formed secretion from acini to reach the common duct. Thus, it is no surprise then that we could not detect ATP in the juice collected from the common duct. Furthermore, since AMP does not accumulate after ATP/ADP hydrolysis, it would imply that the secretion also contains 5′-nucleotidase activity (Fig. 2B).

CD39 in pancreatic acini and ducts

Rat pancreatic acini show the most prominent localization of the CD39 antibody on the basolateral membranes. Interestingly, the old study of Hamlyn & Senior (1983) showed that dispersed rat pancreatic acini (non-stimulated) had a major ATP/ADP hydrolysing enzyme fully consistent with the later described CD39. Some fluorescence is also detected diffusely in the cytoplasmic regions and in clusters coinciding with zymogen granules in acini (Fig. 5I and J). However, even if the enzyme is localized intracellularly or in zymogen granules or their membranes, it is unlikely that it is active until it reaches the plasma membrane (Zhong et al. 2001). Most interestingly, upon CCK-8 stimulation CD39 is redistributed towards the apical pole of the acinar cells, following zymogen granules, or zymogen granular membranes, which are reputed to have the CD39 activity (see below). Thus, CD39 distribution might be relatively mobile, depending on the overall pancreatic activity, stimulated directly by classical agonists or by ATP released in response to agonists or mechanical stimulation (Sørensen & Novak, 2001). This may explain why there was an old dilemma from biochemical and cytochemical studies as to where the CD39 was actually localized (see Introduction).

Apart from CD39 in acini and juice, CD39 is clearly localized in pancreatic ducts. It appears that the smallest ducts, i.e. intercalated ducts that are the first generation of ducts emerging from acini, express CD39 on the luminal side, and this CD39 could also contribute to hydrolysis of released ATP. On the other hand, in the next generations of larger ducts, the localization of CD39 appears to be confined to the basolateral membranes. Similar differential pattern of distribution in the ductal tree is also seen with other proteins such as the Na⁺-Ca²⁺ exchanger, cystic fibrosis transmembrane conductance regulator (CFTR) and carbonic anhydrase (Kumpulainen & Jalovaara, 1981; Hyde et al. 1997; Ankorina-Stark et al. 2002), and this would indicate that different parts of duct tree might have different function regarding Ca²⁺, HCO₃⁻, fluid secretion and also ATP signalling (also see below).

CD39 is a full size 78 kDa protein in pancreatic tissue and pancreatic juice

In rat pancreas and in pure pancreatic acini (devoid of blood vessels) we find a most prominent 78 kDa band in Western blotting, which is the full length CD39 (Fig. 4A and Fig. 6), and a weaker 85 kDa band, which might be due to variation in glycosylation or palmitoylation of the enzyme (Schulte am et al. 1999; Koziak et al. 2000). In the pig pancreas the most prominent band found for many years was the 54 kDa band, which was thought to be the truncated pancreatic form of the enzyme (Sévigny et al. 1995, 1998), or, as accepted recently, the cleaved C-terminal part of the 78 kDa protein (Lemmens et al. 2000). This cleavage is most likely to be due to trypsin and the usual high proteolytic activity of the pancreatic tissue. We have striven to obtain protein (and also RNA) of high integrity, and the smaller bands of 54–57 kDa were only minor component, if any, of our preparation. Also the RT-PCR analysis of the whole pancreas gave the transcript of rat CD39 of the expected length (Fig. 4B). Regarding the pancreatic juice, the ATP diphosphohydrolase activity is most dominant in the particulate fraction of the juice (Fig. 3) and this is where we also find the CD39 protein (Fig. 6). Again CD39 detected in juice is the full length and not the cleaved soluble form of 54 kDa (Chen & Guidotti, 2001). Another secreted soluble human CD39L4 (NTPDase5) is specific for nucleoside diphosphates (Mulero et al. 1999), which the pancreatic enzyme is not. How could it be then that the pancreatic CD39 secreted in juice is the full length enzyme?

Notably, the nucleotidase activity (Fig. 3) and secreted CD39 (Fig. 6) are associated with the particulate fraction of pancreatic juice, which would contain microvesicles. The electron microscopy pictures of CCK-8-stimulated pancreas show luminal microvesicles (C. E. Sørensen, N. Møbjerg & I. Novak, unpublished observations). Interestingly, it is only CCK-8 (predominantly an acinar agonist in rat pancreas) and not secretin (predominantly a ductal agonist in rat pancreas) that causes release of CD39 into the particulate fraction (Fig. 6). Stimulation also causes relocalization of CD39 from the basolateral/intracellular to the apical/ granular part of acini. Possibly palmitoylation of CD39 would make the protein ready for intracellular trafficking (Koziak et al. 2000). Our unifying hypothesis is that after stimulation some CD39 might associate with the zymogen granule membrane, which would give rise to microvesicles and CD39 would have the ecto-orientation after release. This could explain some of the older studies that seemed contradictory. For example, biochemical studies did show that the active ATP-binding site of CD39 was on the interior of granular membranes, which seemed puzzling at that time (Harper et al. 1978; LeBel & Beattie, 1986). In earlier studies on pancreas, it has been described that pancreatic juice collected from unstimulated rat pancreas or secretin-stimulated pig pancreas contained microvesicles (Beaudoin et al. 1986; Grondin et al. 1992), which, however, had lipid composition unlike zymogen granular membrane (Beaudoin et al. 1987). Since the ATP diphosphohydrolase activity was postulated, but could not be detected in secretin-stimulated juice (LeBel & Beaudoin, 1985), the microvesicle-CD39 story has not ‘crystallized’. Nevertheless, microvesicular secretion is not only reserved for pancreas, but also other secretory epithelia secrete various types of vesicles. Prostate glands secrete prostasomes with ATPase activity, but its function is unknown (Ronquist & Brody, 1985). In chicken oviducts, lipoprotein vesicles are associated with ecto-ATPDase (Rosenberg et al. 1977; Strobel et al. 1996). It will be interesting to see whether the microvesicles are connected with other CD39 type enzymes, and whether a similar system might also apply to other tissues and cells, such as endocrine cells and neurons.

Physiological implications for pancreas

Pancreatic acini, and intralobular ducts, express CD39 on the basolateral membrane that could terminate purinergic signalling via ATP released from nerve terminals or acini or by B-cells (Novak, 2003). Surprisingly, pancreatic acini have only a few functional P2 receptors, while the pancreatic ducts have P2Y receptors on the basolateral membrane that inhibit K⁺ channels and thus would not support secretion (Hede et al. 1999; Novak et al. 2002). Therefore, it might be appropriate to focus on ATP signalling on the luminal side of the acinar-ductal axis (Fig. 7).

Figure 7. Model for nucleotide hydrolysis and signalling in pancreas.

Pancreatic acini secrete CD39 upon stimulation, and CD39 is possibly associated with microsomes. In addition, acini also release ATP (Sørensen & Novak, 2001). Pancreatic ducts express CD39, but localization depends on the generation or size of ducts. Pancreatic acini have only a few functional P2 receptors (Novak et al. 2002), while pancreatic ducts have P2Y and P2X type receptors (Hede et al. 1999). While P2Y receptors would inhibit secretion, P2X would upregulate secretion (see Discussion). Adenosine receptors open Cl⁻ channels, thus initiating a part of the secretory process.

Pancreatic acini are the site of ATP release; they express CD39 that can be released into the lumen with a suitable stimulus. Also small ducts are lined with CD39 on the luminal side (Fig. 5). The question is how active this CD39 is in situ, as it will be influenced by pH and cation content of the pancreatic juice. The pH of pancreatic juice of 7.4–8.2 (depending on whether HCO₃⁻ secretion is stimulated) would be suitable for CD39. However, the Ca²⁺ content of the normal juice is lower than 1 mM, which would be sub-optimal judging from the biochemical estimations of CD39 activity. In any case, the fact that there is insignificant ATP in the final secretion indicates that the secreted CD39 and the CD39 on the luminal membrane of pancreatic ducts could provide a regulatory system for controlling ATP levels in the duct system, important for termination of signalling and preventing overstimulation of ducts (Fig. 7). Since pancreatic ducts posses P2X receptors, presumably on the luminal membrane (Hede et al. 1999; Novak, 2003), it might be important to protect the cells from the potential lytic effects of ATP, although in normal conditions, P2X₇/P2X₄ receptor effects on plasma membrane ion transporters or channels are reversible in pancreatic ducts (Christoffersen et al. 1998; Hede et al. 1999). P2X₇ and P2X₄ receptors cannot stimulate Cl⁻ channels or H⁺/HCO₃⁻ transporters and thus cannot be direct mediators of secretion (Hede et al. 1999; Henriksen & Novak, 2003). Yet these receptors allow Ca²⁺ (Na⁺) influx, which could up-regulate secretion evoked by other agonists acting on ducts, or have totally different effect on cell growth/differentiation as they stimulate mitogen activated protein kinases (Amstrup & Novak, 2002), in which case a short exposure to ATP would be sufficient.

In addition to CD39, pancreatic juice contains 5′-nucleotidase activity (Fig. 2) that would hydrolyse AMP to adenosine, which can act on adenosine receptors (Fig. 7). In these duct cells, adenosine leads to opening of Cl⁻ channels, probably CFTR-Cl⁻ channels (I. Novak, unpublished data; Novak et al. 1996). If simultaneously the basolateral K⁺ channels were activated, such as happens with secretin and β-adrenergic stimulation, the secretory process would be activated (Novak, 2003). Thus there is a scenario for a complex interaction between the acinar and ductal parts with respect to ATP secretion and hydrolysis, as well as to stimulation of P2 and adenosine receptors (Fig. 7). However, the exact details regarding the quantity of ATP release and hydrolysis, as well as the time schedule of each process, are to be clarified.

Can one generalize that where there is ATP signalling there is also ATPase release? This seems to be true for autonomic nerves innervating vas deferens and urinary bladder, which release soluble apyrase, but not the nerves innervating taenia coli (Westfall et al. 2000). Bile ducts, which have a HCO₃⁻ secreting mechanism similar to pancreatic ducts, release significant amounts of ATP into bile, indicating the presence of nucleotidases with lower activities (Chari et al. 1996). Interestingly, bile ducts also seem to have different populations of P2 receptors, i.e. P2Y₂ receptors on the lumen (Dranoff et al. 2001). Similarly, guinea-pig pancreatic ducts, which probably have luminal P2Y₂ receptors (Novak, 2003), might have a different ATP release/ hydrolysis pattern.

In conclusion, rat pancreatic acini express CD39 that following CCK-8 stimulation is released as a full length active enzyme into pancreatic juice. Since the CD39 activity is found in the particulate fraction of the secretion, it might be associated with microvesicles. The role of secreted CD39, and CD39 lining small ducts, would be to control ATP concentrations within the ductal tree. Differences in CD39 localization and distribution of P2 and adenosine receptors would imply functional differences in ATP signalling/hydrolysis in the ductal tree.