Nitric oxide signals are interlinked with calcium signals in normal pancreatic stellate cells upon oxidative stress and inflammation

The mammalian diffuse stellate cell system comprises retinoid-storing cells capable of remarkable transformations from a quiescent to an activated myofibroblast-like phenotype. Activated pancreatic stellate cells (PSCs) attract attention owing to the pivotal role they play in development of tissue fibrosis in chronic pancreatitis and pancreatic cancer. However, little is known about the actual role of PSCs in the normal pancreas. These enigmatic cells have recently been shown to respond to physiological stimuli in a manner that is markedly different from their neighbouring pancreatic acinar cells (PACs). Here, we demonstrate the capacity of PSCs to generate nitric oxide (NO), a free radical messenger mediating, for example, inflammation and vasodilatation. We show that production of cytosolic NO in PSCs is unambiguously related to cytosolic Ca2+ signals. Only stimuli that evoke Ca2+ signals in the PSCs elicit consequent NO generation. We provide fresh evidence for the striking difference between signalling pathways in PSCs and adjacent PACs, because PSCs, in contrast to PACs, generate substantial Ca2+-mediated and NOS-dependent NO signals. We also show that inhibition of NO generation protects both PSCs and PACs from necrosis. Our results highlight the interplay between Ca2+ and NO signalling pathways in cell–cell communication, and also identify a potential therapeutic target for anti-inflammatory therapies.

MAJ, 0000-0002-4899-2606; PEF, 0000-0001-5582-6588; OVG, 0000-0003-2573-8258; JVG, 0000-0002-2262-2543; OHP, 0000-0002-6998-0380 The mammalian diffuse stellate cell system comprises retinoid-storing cells capable of remarkable transformations from a quiescent to an activated myofibroblast-like phenotype. Activated pancreatic stellate cells (PSCs) attract attention owing to the pivotal role they play in development of tissue fibrosis in chronic pancreatitis and pancreatic cancer. However, little is known about the actual role of PSCs in the normal pancreas. These enigmatic cells have recently been shown to respond to physiological stimuli in a manner that is markedly different from their neighbouring pancreatic acinar cells (PACs). Here, we demonstrate the capacity of PSCs to generate nitric oxide (NO), a free radical messenger mediating, for example, inflammation and vasodilatation. We show that production of cytosolic NO in PSCs is unambiguously related to cytosolic Ca 2þ signals. Only stimuli that evoke Ca 2þ signals in the PSCs elicit consequent NO generation. We provide fresh evidence for the striking difference between signalling pathways in PSCs and adjacent PACs, because PSCs, in contrast to PACs, generate substantial Ca 2þ -mediated and NOS-dependent NO signals. We also show that inhibition of NO generation protects both PSCs and PACs from necrosis. Our results highlight the interplay between Ca 2þ and NO signalling pathways in cell -cell communication, and also identify a potential therapeutic target for anti-inflammatory therapies.

Background
Mammalian stellate cells (Latin stella-star) are retinoid-storing cells woven into the tissue of various organs [1] including the liver, pancreas, kidney, spleen, lung and vocal folds. Stellate cells are capable of transformations from a quiescent to an activated myofibroblast-like phenotype [2]. Activated stellate cells have attracted attention owing to the pivotal role they play in pathological fibrosis: they overproduce extracellular matrix proteins to repair the chronic stress-induced injuries in the tissue [1][2][3]. Nevertheless, the initial pathophysiological role of stellate cells-prior to activation-remains enigmatic. Here, we studied the primary signalling events, evoked by either oxidative stress or proinflammatory mediators, in stellate cells (PSCs) and neighbouring acinar cells (PACs) in the normal mouse pancreas, and identified a link between calcium and nitric oxide signalling pathways in PSCs.
Reactive oxygen/nitrogen species (ROS/RNS), such as NO, are highly chemically active radical and non-radical molecules that initiate and propagate reactions of oxidative stress, and thus act as second messengers in various inflammatory processes [13 -16]. Both endogenous and exogenous ROS can modulate store-operated Ca 2þ entry (SOCE) [17 -20] and release [21]. Excessive Ca 2þ influx into PACs, together with the sustained elevation of the cytosolic calcium ion concentration ([Ca 2þ ] C ) [22][23][24], underlies the mechanism of AP [25], and store-operated Ca 2þ entry channels are therefore potential therapeutic targets [24,[26][27][28]. Nevertheless, the roles of ROS/RNS (including NO) in the ( patho)physiology of PSCs remain unexplored. Generation of NO has not yet been reported in PSCs, although a plausible link between PSC activation and NO has been established: in cultured rat PSCs, expression of nitric oxide synthase 2 (NOS2) is increased after stimulation with pathogen-associated molecular patterns (PAMP) that activate Toll-like receptors (TLR) of innate immunity [29]. TLRs also mediate responses to damage-associated molecular patterns (DAMP) released from injured tissues (e.g. the necrotizing pancreas) [12].
The proinflammatory mediator bradykinin (BK) induces NO production in vascular endothelial cells [30,31]. BK was recently shown to elicit Ca 2þ signals in PSCs [32], at pathophysiologically relevant concentrations [33], and this was linked to AP via specific action on PSCs through bradykinin receptor B2 [33]. So far, it is unknown whether BK elicits NO generation in PSCs and, if so, what the role of this process might be. Bile acids (BA), employed extensively in cellular [11,22,34,35] and animal [16,36 -38] studies of the pancreas, induce abnormal Ca 2þ signals [34] that cause necrosis and AP [25]. In the light of the inflammatory background of AP, it seems likely that BK and BA would evoke not only Ca 2þ signals, but also NO signals with the potential for crosstalk between the two signalling pathways.
In order to explore such interactions, we inhibited either Ca 2þ or NO signal generation in PSCs. Caffeine is known to reduce Ca 2þ signals via inhibition of inositol 1,4,5-triphosphate receptors in PACs [39,40]; what is more, a caffeinedependent decline in severity of pancreatic injury was recently demonstrated in three animal models of AP [41]. Thus, caffeine was used here to test whether the blockade of BA-elicited Ca 2þ responses might attenuate NO signals. The latter were also blocked pharmacologically by inhibitors of enzymatic NO synthesis (NOS inhibitors), widely used in therapy of various inflammatory diseases [42], including AP [43,44]. Nevertheless, so far, the actual outcome of NOS inhibition in AP remains unclear. The aim of this study was to investigate the role of normal PSCs in the initial signalling events upon proinflammatory stimulation. We report here that PSCs, in contrast to PACs, generate substantial Ca 2þmediated and NOS-dependent NO signals, and that inhibition of NO generation protects both PSCs and PACs from necrosis.

Isolation of pancreatic lobules
Six-to eight-week-old male mice were sacrificed by cervical dislocation, the pancreases were dissected and the lobules were immediately isolated by collagenase digestion. Briefly, the pancreas was injected intraductally with NaHEPESbased collagenase solution and incubated (5 -6 min, 378C) to allow partial digestion of the tissue.

Primary human pancreatic stellate cell line
hPSCs were cultured (up to the fifth passage) at 378C, 5% CO 2 , in complete stellate cell medium and split once a week.

Cytosolic calcium or nitric oxide measurements
Unless otherwise indicated, NaHEPES-based media, containing (mM): NaCl, 140; KCl, 4.7; HEPES, 10; MgCl 2 , 1; glucose, 10; and pyruvate, 1, were supplemented with 1 mM Ca 2þ for calcium measurements and with 1 mM Ca 2þ together with 0.5 mM L-Arg for nitric oxide recordings. For Ca 2þ measurements, the lobules were loaded with 10 mM Fluo-4 (1 h, 308C), and hPSC with 1 mM Fluo-4 (30 min, 378C). For NO measurements, the lobules were loaded with 20 mM DAF-2 or DAF-FM (1 h, 308C), and hPSC with 0.1 mM DAF-2 or DAF-FM (1 h, 378C). The lobules were transferred to a flow chamber and allowed to adhere to the glass surface; and for hPSC imaging, the coverslips with growing cells were used for flow chamber assembly. Experiments were performed in continuous perfusion with extracellular buffer-based solution; and the cells were visualized using a TCS SP5 II twophoton confocal microscope (Leica) with a 63 Â 1.2 NA water objective. Fluo-4 or DAF dyes were excited with a 488 nm Ar laser, at 1-4% power, and emitted light was collected in the three-dimensional recording mode at 495-580 nm. The speed of recordings was approximately one image per 10 s, and varied dependent on thickness of the samples (up to 15 mm, z-axis resolution 1 mm). Images were captured at 512 Â 512, and series of images were recorded at 256 Â 256 pixel resolution, respectively, and analysed using Leica software. In order to reconstruct the three-dimensional signal, the individual signals from z-stacks were cropped, and the maximal projection was applied. Fluorescence signals were plotted as F/F 0 , where F 0 was an averaged signal from first ten baseline images, and normalized as previously described [45].

Simultaneous cytosolic calcium and nitric oxide measurements
For simultaneous Ca 2þ and NO measurements, the lobules were loaded with 10 mM Fura-2 and 10 mM DAF-2 (1 h, rsob.royalsocietypublishing.org Open Biol. 6: 160149 308C). After the loading, the lobules were transferred to the chamber, perfused and visualized as described above. Fura-2 fluorescence was excited with 355 nm and 405 nm lasers, at 8% and 16% power, respectively; and emitted light was collected in the three-dimensional recording mode at 500-600 nm; DAF-2 fluorescence was excited and collected as described above.

Measurements of necrosis level in the lobules
The lobules were treated with 5 mM cholate, 5 mM taurocholate or 0.2 mM TLC-S challenge (2 h, room temperature), and in some experiments, 0.6 mM L-NAME was present. The lobular PSCs were visualized using Fluo-4 (10 mM, 2 h); the lobules were co-stained with Hoechst 33342 (32 mM, 30 min), and dead cells were identified by PI staining (1.5 mM, 15 min) as described [33]. The cells were visualized, using the confocal microscope with a 63 Â 1.2 NA water objective. Fluo-4, Hoechst 33342 and PI were excited with 488 nm Ar (1%), 355 nm diode (10%) and 543 nm HeNe laser (10%), respectively; and corresponding emissions were collected at 505-535, 415-485 and 615 -720 nm. The fluorescence signal was collected sequentially between frames in the three-dimensional mode from 20 mm thick lobules and 512 Â 512 pixel resolution. Five pictures of independent lobules were taken in each of four experimental replicates (n ¼ 20), and live (PI-negative) and dead (PI-positive) cells were counted.

Immunohistochemistry
Unless otherwise indicated, the procedure was performed at room temperature, and double distilled water (ddH 2 O) was used for preparation of all solutions. 0.1% Tween 20 was used as a washing buffer and 1% BSA in PBS with 0.1% Tween 20 was a blocking buffer. Mouse pancreatic tissue samples were fixed in formalin, embedded in paraffin and cut into 4 mm sections. The sections were heated in a dry oven (30 min, 658C), then deparaffinized in xylene (2 Â 10 min) and graded ethanol, and then incubated in 50 mM NH 4 Cl (20 min). Antigen retrieval was achieved by autoclaving (20 min, 1208C) the samples in TAE buffer ( pH 8.1), followed by slow cooling to room temperature (30 min). Permeabilization was performed in 0.4% Triton X-100 (10 min). In order to quench autofluorescence, the sections were incubated in 0.2% Sudan black B [46]. The sections were then transferred to a humid chamber, and blocking of non-specific binding sites was performed (1 h), followed by incubation with primary anti-BDKRB2 and anti-NOS2 Abs (0.5 mg ml 21 ) for 1 h at room temperature, and then overnight at 48C. The negative controls were incubated in blocking solution with no primary Abs. The following day, the sections were incubated (1 h) with goat anti-rabbit secondary Ab (4 mg ml 21 ), washed, and then incubated (1 h) with goat anti-mouse secondary Ab (4 mg ml 21 ). The sections were embedded in antifade mounting medium with DAPI, and imaged immediately using the confocal microscope (excitation wavelengths: 355, 488 and 633 nm). The slides were stored at 48C.

Statistics
The quantitative results were expressed as means + s.d. or s.e.m. (see the text for details). Statistical analysis was performed using the Student's t-test or ANOVA, and the significance threshold was set at 0.05.

Oxidative stress elicits (patho)physiological calcium and nitric oxide signals
Hydrogen peroxide (H 2 O 2 ) was used as an initiator of oxidative stress in human pancreatic stellate cells (hPSCs; figure 1a-b) and mouse pancreatic tissue lobules (figure 1c-h). In hPSCs, a sustained increase in [Ca 2þ ] C was evoked by 0.5 mM H 2 O 2 (blue), which was markedly ( p , 0.001) attenuated by removal of external Ca 2þ (orange; figure 1a-b). In the lobules, oxidative stress elicited rises in [Ca 2þ ] C in both PSCs (red) and PACs (blue; figure 1c), although cytosolic NO signals (figure 1d; see also electronic supplementary material, figure S1 and video S1) were limited spatially to PSCs (red), manifest as a sharp increase and sustained plateau phase. Three-dimensional reconstruction of the DAF-FM fluorescence signal collected from the 1 mM

Nitric oxide synthase inhibitors modulate bile-acidevoked nitric oxide signals
Pharmacological NOS inhibitors, the non-specific L-NAME and the irreversible inhibitor of NOS2 aminoguanidine (AG),

Blockade of nitric oxide production reduces necrosis in the pancreatic lobules
The role of pharmacological blockade of NO production was analysed after 2 h incubation of the lobules with 5 mM cholate, 5 mM taurocholate or 0.2 mM TLC-S (figure 4e; see also electronic supplementary material, figure S3), and L-NAME was used as NOS inhibitor. Because the levels of necrosis in the lobules that received 0.6 mM L-NAME (and no BA) treatment did not change in comparison with NaHEPES controls (less than 5%), L-NAME at this concentration was tested as a protective agent against the BA challenge.

Discussion and conclusion
It has previously been shown that cytosolic calcium signals can be elicited in cultured [32] and in normal (lobular) [33] pancreatic stellate cells. This study now reveals that these signals are interlinked with cytosolic nitric oxide signals. We show that NO signals are evoked in PSCs upon induction of oxidative stress (figure 1) or application of inflammatory mediators (figures 2-4). Oxidative stress originates from the imbalance between the production and neutralization of ROS/RNS [13,47], and is implicated in the rsob.royalsocietypublishing.org Open Biol. 6: 160149 mechanisms of numerous inflammatory diseases [15,16,20], including AP [11,16,19]. However, for pancreatitis, there remains the 'chicken or the egg' question regarding the roles of ROS/RNS in the development of inflammation.
Here, we show that H 2 O 2 at concentrations that are ( patho)physiologically relevant [47,48] evoke large Ca 2þ signals with an oscillatory pattern in both PACs and PSCs ( figure 1a-c). Importantly, we demonstrate that it is only in the PSCs that these Ca 2þ signals are accompanied by detectable NO signals ( figure 1dh). Spatial separations of BK-and ACh-or CCK-elicited Ca 2þ signals in pancreatic lobules have previously been reported. Whereas ACh and CCK evoke Ca 2þ signals in PACs, which control normal acinar secretion [49], BK-evoked Ca 2þ signals are entirely confined to PSCs [33]. BK-elicited Ca 2þ signalling events in PSCs are mediated via BDKRB2 [33] and pharmacological blockade of this receptor with the B2 antagonist WIN64338 protected lobular PACs from the necrosis evoked by alcohol/fatty acid or bile acids [33], indicating a possible paracrine interaction between PSCs and PACs. Here, we show that BK elicits simultaneous Ca 2þ and NO signals in PSCs ( figure 2a, upper curves), but fails to evoke responses in adjacent PACs ( figure 2a, lower curves). Because large Ca 2þ signals in PSCs are not accompanied by Ca 2þ signals in PACs, and vice versa, it may be NO that mediates paracrine communication between stellate cells and other cell types in the pancreas.
Bile acids are natural compounds of the bile that facilitate enzymatic hydrolysis of lipids in the process of digestion [50]. In the case of gallstone-induced bile reflux into the pancreatic duct, bile acids in high concentrations will get in direct contact with pancreatic cells. Bile acids evoke large abnormal Ca 2þ signals in PACs [22,34,35], followed by intracellular enzyme activation, PAC necrosis [25], autodigestion of the pancreas and finally pancreatitis [11,12,25]. Cytoplasmic and mitochondrial Ca 2þ signals, elicited in isolated PACs by TLC-S, impair ATP synthesis and induce ROS production [11]. Nevertheless, TLC-S has little effect on [Ca 2þ ] C in PSCs (figure 3a), and no detectable role in NO signalling in these cells (figure 3b). In contrast, robust Ca 2þ signals elicited in PSCs by cholate and taurocholate (figure 3a) are accompanied by NO signals (figures 3b and 4a -d). A plausible explanation for the diverse sensitivity of pancreatic cells to bile acids might be the different pattern of bile-acid-transporting proteins in PSCs and PACs.
Our results demonstrating that even substantial Ca 2þ signals generated in PACs fail to elicit detectable NO signals are in agreement with a previous study in which the non-enzymatic NO signal generation was explored in isolated PACs [45]. In that study, even supramaximal ACh concentrations failed to evoke detectable NO signals in more than 70% of intact PACs.
The potential benefit of blocking NO generation in the therapy of various diseases, including AP, remains controversial [15]: several studies show beneficial effects of NOS inhibitors in the therapy of cancer [51], arthritis [52] and pancreatitis [43,44], whereas others demonstrate that blockade of NO production aggravates liver injury [53], and exacerbates inflammation in the kidney [54]. Here, we show that L-NAME (blocker of NOS 1-3; figure 4a-c) and AG (irreversible inhibitor of NOS2; figure 4d) significantly reduce bile-acid-evoked NO signals in PSCs. Furthermore, we demonstrate that there is substantially more necrosis in PSCs than in PACs in cholate-or taurocholate-challenged lobules (figure 4e), which correlates with the presence of NO signals together with large Ca 2þ signals in PSCs upon stimulation with these BAs  rsob.royalsocietypublishing.org Open Biol. 6: 160149 lobules, irrespective of bile type and cell type (figure 4e). In cholate-or taurocholate-stressed lobules the protective effect of L-NAME (blockade of NO signals ameliorates necrosis) is more prominent in the PSCs than in the PACs, which do not have detectable cytosolic NO signals. The protection of PACs might be the result of, for example, paracrine communication, PSCs -PACs, via intercellular messengers ( plausibly NO) released by PSCs. Expression of the NO-forming enzymes, NOS 1-3, in the exocrine pancreas has recently been confirmed using state-ofthe-art immunohistological methods [55]. The authors of the study, however, did not refer directly to PSCs, although they report the presence of 'cells of the morphology of ductal cells' in chronic pancreatitis (CP) tissue specimens [55]. These cells could well have been PSCs, which are known to induce fibrosis in CP [2,56]. TLR-mediated expression of NOS2 has been detected in isolated PSCs upon stimulation with PAMP [29]. Importantly, TLRs are also activated upon stimulation with DAMP-endogenous molecules released from injured cells [12] (e.g. autodigested PACs). Thus, DAMP-exposed and then NOS2-expressing PSCs might be present in the pancreatic tissue. Here, using the Ab anti-BDKRB2 (figure 4h; green), we identify PSCs in mouse pancreatic tissue specimens: the cells localize in the interacinar spaces, encircling the base of adjacent acini with fine cytoplasmic processes [57]. This is in line with other immunohistological studies on PSCs in the milieu of the tissue [1,57]. The pattern of anti-NOS2 Ab staining (figure 4g; red) resembles the network of discrete stellate-shaped cells and their cytoplasmic processes [57]. The staining of BDKRB2 and NOS2 overlay, yielding yellow to orange colour (figure 4i). The strong staining of PSC processes might be explained by the results of a study on polarized epithelial cells [58] in which NOS2 was shown to be compartmentalized in the submembrane areas (here: near the membrane surrounding the nuclear area and covering the very fine cytoplasmic processes). Such submembrane localization of NOS2 would ensure precise vertical delivery (across the cell membrane) of the second messenger NO to the target [58,59].
NO generated in PSCs may not only exert an effect on PACs, but also reach endothelial cells in peri-acinar capillaries. As pointed out recently [60], the PSCs are strategically localized in a niche between acinar cells and peri-acinar capillaries, and, in the intact pancreas in vivo, NO generated in PSCs could well have vascular effects.
In conclusion, our results reveal interplay between NO and Ca 2þ in situ in PSCs, induced by stress related to inflammation (figures 1-2) and disease (figures 3-4). PSCs might then resemble a sensor that downstream the stress signals by means of at least two second messengers.
Ethics. All the experimental procedures were carried out in accordance with the UK Home Office regulations.