Mechanisms of constitutive and ATP-evoked ATP release in neonatal mouse olfactory epithelium

Background ATP is an extracellular signaling molecule with many ascribed functions in sensory systems, including the olfactory epithelium. The mechanism(s) by which ATP is released in the olfactory epithelium has not been investigated. Quantitative luciferin-luciferase assays were used to monitor ATP release, and confocal imaging of the fluorescent ATP marker quinacrine was used to monitor ATP release via exocytosis in Swiss Webster mouse neonatal olfactory epithelial slices. Results Under control conditions, constitutive release of ATP occurs via exocytosis, hemichannels and ABC transporters and is inhibited by vesicular fusion inhibitor Clostridium difficile toxin A and hemichannel and ABC transporter inhibitor probenecid. Constitutive ATP release is negatively regulated by the ATP breakdown product ADP through activation of P2Y receptors, likely via the cAMP/PKA pathway. In vivo studies indicate that constitutive ATP may play a role in neuronal homeostasis as inhibition of exocytosis inhibited normal proliferation in the OE. ATP-evoked ATP release is also present in mouse neonatal OE, triggered by several ionotropic P2X purinergic receptor agonists (ATP, αβMeATP and Bz-ATP) and a G protein-coupled P2Y receptor agonist (UTP). Calcium imaging of P2X2-transfected HEK293 “biosensor” cells confirmed the presence of evoked ATP release. Following purinergic receptor stimulation, ATP is released via calcium-dependent exocytosis, activated P2X1,7 receptors, activated P2X7 receptors that form a complex with pannexin channels, or ABC transporters. The ATP-evoked ATP release is inhibited by the purinergic receptor inhibitor PPADS, Clostridium difficile toxin A and two inhibitors of pannexin channels: probenecid and carbenoxolone. Conclusions The constitutive release of ATP might be involved in normal cell turn-over or modulation of odorant sensitivity in physiological conditions. Given the growth-promoting effects of ATP, ATP-evoked ATP release following injury could lead to progenitor cell proliferation, differentiation and regeneration. Thus, understanding mechanisms of ATP release is of paramount importance to improve our knowledge about tissue homeostasis and post-injury neuroregeneration. It will lead to development of treatments to restore loss of smell and, when transposed to the central nervous system, improve recovery following central nervous system injury.


Background
Although once widely assumed that the only source for extracellular ATP was from damaged cells under pathophysiological conditions, it is now accepted that ATP is also released under normal physiological conditions in several organ systems via exocytosis [1], efflux through connexon or pannexin "hemichannels" [2,3], voltage-dependent anion channels [4], and/or activated P2X 7 receptors [1]. Extracellular ATP has numerous functions in sensory systems. It is released via pannexins to mediate intercellular communication in taste buds [2,3]. In the auditory system, spontaneous ATP release is required for activity in the developing auditory system [5], ATP release through connexin hemichannels propagates calcium signals in the inner ear [6] and purinergic receptor activation by ATP is essential in the cell-cell communication that accompanies cochlear injury [7]. In retinal glial cells, ATP release propagates spontaneous intercellular glial calcium waves that alter the diameter of arterioles in the retina, suggesting released ATP can affect retinal physiology [8].
ATP also mediates essential functions in the olfactory system: (1) ATP activation of purinergic receptors in Swiss Webster mouse olfactory epithelium (OE) induces neuroproliferation [9] (2) ATP has proliferative and protective effects following injury [10,11] and (3) ATP induces upregulation and/or release of various neurotrophic factors [12][13][14]. Given the multiple roles of ATP in the mouse OE, we hypothesized that ATP was released under both normal and pathophysiological conditions in mouse OE and likely involved multiple mechanisms. Vesicles containing ATP could be released via calcium-dependent exocytosis. Cytosolic ATP could also be released by efflux via ATP binding cassette (ABC) transporters [15,16] expressed in the OE [17]. Several connexin subtypes are expressed in the OE [18][19][20][21][22] and pannexins are expressed in the olfactory bulb [23], but as of yet, no reports have examined pannexin expression in the epithelium. P2X 7 receptors have been identified in the OE using immunohistochemistry [24]. Thus, there are multiple pathways of ATP release possible in the OE. Determination of ATP release mechanisms is of paramount importance to improve our understanding of the OE regenerative properties.
The aim of this study was first to investigate whether ATP release mechanisms were present in mouse OE, using Swiss Webster neonatal OE slices as a model. Using 3 techniques, we showed that ATP is released in neonatal OE through constitutive and evoked release. We then characterized the mechanisms underlying the two types of ATP release using the luciferin-luciferase assay and the fluorescent ATP marker quinacrine to monitor the loss of ATP fluorescence from endogenous vesicular stores. Our results show that the evoked release of ATP is mediated by purinergic receptor activation and can occur via calcium-dependent exocytosis, efflux of ATP through activated P2X 7 receptors or activated P2X 7 receptor/pannexin complexes, and ABC transporters. Confocal calcium imaging of P2X 2 -transfected HEK293 cells used as biosensors further confirmed the presence of the evoked release of ATP. The constitutive release of ATP does not require purinergic receptor activation, but is mediated by continuous exocytosis and efflux through hemichannels or ABC transporters. In vivo studies indicate that constitutive vesicular release may play a role in neuronal homeostasis as inhibition of exocytosis inhibited the normal proliferation in the OE. Overall, determination of the ATP release mechanisms in the OE has identified numerous pharmacological targets to pursue that could modulate neuronal homeostasis.

Constitutive and ATP-evoked ATP release occurs in the mouse olfactory epithelium
We assessed whether ATP is released in the olfactory epithelium under control physiological conditions and following ATP application. Endogenous ATP stores in neonatal OE slices were labeled with the fluorescent compound quinacrine as previously performed [27]. Quinacrine-labeled ATP stores appear as puncta of fluorescence located throughout the OE (Figure 2A). The individual punctum ranged in diameter between 200 and 250 nm, and sometimes multiple puncta appeared as clusters ranging up to 1.8 μm in diameter, suggesting that quinacrine labels ATP that is stored in large vesicles. We visualized the release of ATP from the fluorescent puncta over a period of 400 seconds under physiological conditions (Ringer's solution application; Control) and observed either a linear or an exponential decrease in fluorescence intensity ( Figure 2B and Table 1). ATP fluorescence in a region of a quinacrinelabeled slice that was devoid of punctate fluorescence ( Figure 2B) showed a minimal change in fluorescence over time, suggesting that the decrease in fluorescence of the puncta is not photobleaching of quinacrine, but rather due to ATP release. Application of 50 μM ATP changed the properties of the fluorescent puncta. While some puncta showed an exponential decrease of their fluorescence ( Figure 2B, Table 1), fluorescence from other puncta either dynamically fluctuated ( Figure 2C, Table 1) or exhibited decreasing fluorescence intensity while the puncta became mobile (Table 1, data not shown). We quantified both the rate of ATP release and the amount of released ATP from all puncta recorded following application of physiological Ringer's solution (control) or 50 μM ATP. The rate of ATP release was significantly faster following ATP application compared to control ( Figure 2D, 31 ± 4 s vs. 304 ± 27 s, n = 11 puncta from 3 slices and 43 puncta from 3 slices, respectively; p < 0.0001, unpaired Student's t-test). In addition, exogenous ATP (50 μM) significantly increased the amount of released ATP compared to vehicle control ( Figure 2E; 65 ± 3% v. 54 ± 2%, p < 0.0001, unpaired Student's t-test). These data indicate that both constitutive (Ringer's solution application) and evoked (50 μM ATP application) release of ATP occurs in the mouse olfactory epithelium.
To confirm that the decrease in fluorescence from the puncta is due to release of quinacrine-labeled ATP rather than moving out of the focal plane, we assessed both constitutive and evoked release of ATP after 400 s using confocal Z stack recordings (see Methods). ATP evoked a significant increase in ATP release compared to control ( Figure 2F; 73 ± 1% v. 60 ± 1%; n = 124 puncta from 5 slices and 118 puncta from 5 slices, p < 0.0001, unpaired Student's t-test). The amount of ATP released from quinacrine-labeled ATP stores under control or ATP-evoked conditions was statistically similar using both the Z stack and the time series recording method (Compare Figure 2E v. 2 F; p > 0.7, unpaired Student's T tests). These data support the use of Z stack recording to measure both constitutive and ATP-evoked release of ATP. As this method is devoid of putative artifacts related to a shift of focus over time, we used Z stack recordings in all subsequent recordings using quinacrine.
We next used the luciferin-luciferase assay to quantify the amount of ATP release from neonatal OE slices. Under physiological conditions 19 ± 4 nM ATP was released from OE tissue slices ( Figure 2G; n = 8 slices). Although there was a 3-fold decrease in the amount of ATP released in the presence of non-specific purinergic receptor inhibitor PPADS (25 μM), it was not significantly different from control ( Figure 2G; 6 ± 2 nM, n = 4 slices, p = 0.07, one-way ANOVA with Bonferroni's planned comparison). These data are in agreement with the quinacrine-generated data and suggest that constitutive release is not mediated by purinergic receptors. Application of ATP, a co-factor for luciferase, would  interfere with the ability to measure endogenously released ATP using the luciferin-luciferase assay. However, the P2X 1,7 receptor agonist Bz-ATP (50 μM) induced a low level of bioluminescence corresponding to 576 ± 424 nM ATP ( Figure 2G; No Tissue (+ Bz-ATP), n = 5 slices), allowing the quantification of evoked ATP-release. Bz-ATP (50 μM) evoked a significant 5fold increase in bioluminescence in the presence of OE tissue slices ( Figure 2G; Tissue (+ Bz-ATP): 3108 ± 914 nM ATP, n = 7 slices, p < 0.05 v. no tissue, one-way ANOVA with planned comparisons). Significantly more ATP was released from slices stimulated with 50 μM Bz-ATP than from control conditions ( Figure 2G; p < 0.01; one-way ANOVA with planned comparisons). The non-specific purinergic receptor inhibitor PPADS (25 μM) significantly inhibited the Bz-ATP-evoked release of ATP to control (no tissue) levels ( Figure 2G; 215 ± 150; n = 4 slices; p > 0.05, one-way ANOVA with planned comparisons). These data confirm the involvement of P2X 1,7 receptors in the evoked release of ATP. We next used stable P2X 2 -transfected HEK-293 cells as biosensors to further confirm that ATP is released from OE slices. Fluo-4 AM loaded P2X 2 -transfected HEK-293 cells were stimulated with ATP concentrations ranging from 20 nM, the concentration of ATP constitutively released from OE slices as measured by the luciferin-luciferase assay, to 5 μM, the concentration released from OE slices by Bz-ATP stimulation. ATP application (20 nM) was insufficient to elicit an increase in intracellular calcium (data not shown) and 50 nM ATP induced a very small calcium response in fluo-4 AM loaded P2X 2 -transfected HEK-293 cells ( Figure 2H, 0.05 ± 0.02 ΔF/F, n = 10 cells). The mean peak (± SEM) ATP response elicited by 5 μM (0.41 ± 0.07 ΔF/F, n = 15 cells) was significantly higher than the mean peak amplitudes elicited by all lower ATP concentrations (1 μM: 0.13 ± 0.02 ΔF/F, n = 13 cells; 500 nM: 0.12 ± 0.02 ΔF/F, n = 19 cells, p < 0.05, one-way ANOVA with Bonferroni's planned comparison tests). Pretreatment with PPADS (25 μM) totally abolished the ATP-induced calcium transients in the P2X 2 -transfected HEK-293 cells (0.01 ± 0.01 ΔF/F, n = 10 cells from 3 coverslips; data not shown), indicating the calcium increase was mediated by activated P2X 2 receptors. Neither control Ringer's solution nor Bz-ATP (50 μM) induced increases in intracellular calcium in the P2X 2 -transfected HEK-293 cells (data not shown). These data suggest that P2X 2 -transfected HEK-293 cells are not sensitive enough to detect constitutively released ATP (~20 nM, Figure 2F). Indeed, no calcium increases were detected in P2X 2 -transfected HEK-293 cells located adjacent to an OE slice in the recording chamber (n = 3 coverslips; data not shown). However, the P2X 2 -transfected HEK-293 cells are sensitive enough to detect the evoked release of ATP from OE slices which is~3 μM (see Figure 2G). In the presence of an OE slice, 50 μM Bz-ATP induced a mean (± SEM) calcium transient peak amplitude of 0.34 ± 0.04 ΔF/F in the P2X 2 -transfected HEK-293 cells ( Figure 2H, n = 70 cells from 3 coverslips). As Bz-ATP did not elicit calcium increases in P2X 2 -transfected HEK-293 cells in the absence of OE tissue, these data indicate that Bz-ATP evokes the release of ATP from OE slices. Note that the calcium response elicited by Bz-ATP stimulation of an OE slice was similar to that elicited by 5 μM ATP in the absence of an OE slice ( Figure 2H; p > 0.05, oneway ANOVA with Bonferroni's planned comparisons). This result suggests that approximately 5 μM ATP is released from a slice following stimulation with Bz-ATP and is consistent with the data obtained by the luciferinluciferase assays. Taken together, data obtained from fluo-4 AM loaded P2X 2 -transfected HEK-293 cells show that ATP is released from OE slices following P2X 1,7 purinergic receptor stimulation.

Mechanisms of ATP-evoked ATP release
We tested the hypothesis that ATP-evoked ATP release in the olfactory epithelium is similar to that in other organ systems and investigated the roles of purinergic receptors, vesicular exocytosis, pannexin/connexin hemi channels and ABC transporters (see Figure 3A for model). Quinacrine imaging was used to investigate vesicular exocytosis as this technique allows visualization of vesicular ATP release, and luminometry assays were used to quantify ATP release for all the proposed mechanisms. Our data support the hypothesis that ATP-evoked release of ATP is mediated predominantly by ionotropic P2X 1,3,7 receptors ( Figure 2F-H), but there is evidence that G-protein coupled P2Y 2,4,6 receptors may also be involved ( Figure 2F). Activation of either receptor subtype increases intracellular calcium which could induce calcium-dependent exocytosis. Removal of extracellular calcium significantly inhibited ATP-evoked release of ATP (Figure 3B-C; 48 ± 4%, n = 30 puncta from 4 slices; 315 ± 34 nM, n = 7 slices; p < 0.05 v. ATP or Bz-ATP, Student's t-test). These data further support the involvement of ionotropic P2X 1,3,7 receptors, but does not rule out the possibility of a P2Y 2,4,6 receptor contribution. We hypothesized that increases in intracellular calcium following purinergic receptor activation may lead to calcium-dependent exocytosis of ATP stored in vesicles. Pre-treatment with Clostridium difficile toxin A (15 min, 0.5 nM), known to impair the fusion of vesicles and potently inhibit exocytosis, significantly impaired the evoked release of ATP (Figure 3B-C; 57 ± 2%, n = 91 puncta from 3 slices; 138 ± 16 nM ATP; n = 3 slices; p < 0.01 v. ATP or Bz-ATP, one-way ANOVA with Bonferroni's planned comparison tests). In addition, N-ethylmaleimide, another vesicle fusion inhibitor, significantly inhibited the ATP-evoked release of ATP ( Figure 3B; 56 ± 1%, n = 226 puncta from 3 slices, p < 0.001 v. ATP, one-way ANOVA with Bonferroni's planned comparison test). Exocytosis requires formation of soluble N-ethylmaleimide-sensitive factor attachment protein receptor (SNARE) core complexes [28]. We examined the expression of the SNARE complex protein synaptosomal associated protein (SNAP)-23. SNAP-23 localized to the thin cytoplasmic extensions throughout the OE that could be either sustentacular cells or neurons ( Figure 3D-E). Taken together, our data indicate that purinergic receptor-mediated calcium-dependent exocytosis evokes the release of ATP.
ATP can also be released through a large non-selective pore formed either by the P2X 7 receptor [29] or by the P2X 7 receptor in a complex with pannexin [30]. To test if ATP or Bz-ATP evokes the release of ATP via a P2X 7pannexin complex, we used carbenoxolone (100 μM), a non-specific pannexin and activated P2X 7 receptor inhibitor [31,32]. Carbenoxolone significantly decreased the amount of ATP released ( Figure 3B-C; 46 ± 4%, n = 34 puncta from 3 slices; 56 ± 9 nM, n = 3 slices; p < 0.01 v. ATP or Bz ATP, one-way ANOVA with Bonferroni's planned comparisons). In addition, probenecid (500 μM), a selective pannexin 1 channel inhibitor [31,33] as well as an ABC transporter inhibitor [17], significantly decreased the amount of ATP released from slices stimulated by Bz-ATP measured by luminometry ( Figure 3C; 365 ± 106 nM, n = 4 slices; p < 0.01, one-way ANOVA with Bonferroni's planned comparisons). These data indicate that the P2X 7 receptor and pannexin 1 channels, alone or in a complex, and ABC transporters may also be involved in ATPinduced ATP release.

Mechanisms of constitutive ATP release
Our data indicated that purinergic receptors are not involved in the constitutive release of ATP based on the lack of an effect on ATP release in the presence of purinergic receptor antagonist PPADS ( Figure 2E). Thus, we tested the hypothesis that constitutive ATP release in the olfactory epithelium is similar to that in other organ systems and investigated the role of pannexin/connexin hemi-channels, ABC transporters and vesicular exocytosis (see Figure 3A for models) using both luminometry assays and quinacrine imaging. We measured ATP release in the presence of the non-specific pannexin inhibitor carbenoxolone (100 μM). Carbenoxolone decreased the amount of constitutive ATP released by half ( Figure 3G, 9 ± 1 nM, n = 5 slices, p < 0.05 v. control, one-way ANOVA with Bonferroni's planned comparison). In addition, the selective pannexin 1 and ABC transporter inhibitor probenecid (500 μM) significantly decreased the amount of ATP released from OE slices ( Figure 3G; 9 ± 3 nM ATP, n = 4 slices; p < 0.05 v. control, one-way ANOVA with Bonferroni's planned comparison). Use of vesicular fusion inhibitor Clostridium difficile toxin A (1 nM) significantly impaired the release of ATP compared to control conditions ( Figure 3F; 49 ± 1%, n = 329 puncta from 3 slices, p < 0.001, 6 ± 3 nM ATP; n = 3 slices; p < 0.05, one-way ANOVA with Bonferroni's planned comparisons). We also used gadolinium chloride (Gd 3+ , 50 μM) that enhances spontaneous exocytosis in rat brain synaptosomes [34] as well as inhibits pannexin and connexin hemichannels [35]. Gd 3+ significantly increased constitutive exocytotic release of ATP ( Figure 3F; 78 ± 2%, n = 75 puncta from 3 slices; p < 0.001, one-way ANOVA with Bonferroni's planned comparisons). Note that in the presence of Gd 3+ , ATP did not significantly enhance the release of ATP compared to Gd 3+ alone or ATP alone ( Figure 3B, 74 ± 2% ATP, n = 49 puncta from 3 slices, p > 0.05 v. ATP and v. Gd 3+ , one-way ANOVA with Bonferroni's planned comparisons). These results suggest that exocytotic release of ATP may have a greater role in the constitutive release of ATP than release through pannexin/connexin hemichannels. In addition, Gd 3+ -evoked vesicular release of ATP may be operating at the maximum extent measurable under physiological conditions as ATP did not enhance the release of ATP in the presence of Gd 3+ . Alternatively, it may not be possible to observe a greater release of ATP under the quinacrine experimental conditions, and it is not possible to measure the effects using luminometry as the trivalent cation interferes with the luciferinluciferase assay. Overall, our data show that a constitutive release of ATP involves spontaneous exocytosis, ABC transporters and pannexins in neonatal mouse OE.
We observed that degradation of extracellular ATP by apyrase inhibited the release of ATP to levels even lower than control ( Figure 2E). In mouse neuromuscular junction the spontaneous release of neurotransmitters is inhibited by stimulation of the G protein-coupled P2Y purinergic receptors [36]. We hypothesized that in the presence of apyrase, the constitutively released ATP could be degraded to ADP that could inhibit the spontaneous release of ATP via P2Y 1,12,13 purinergic receptor activation. Application of ADP (50 μM) significantly impaired the release of ATP ( Figure 3F; 48 ±3% ATP, n = 80 puncta from 3 slices; p < 0.0001 v. control, unpaired Student's t-test). This result confirms that activation of P2Y 1,12,13 receptor subtypes inhibits the spontaneous vesicular release of ATP. Therefore, blockade of the P2Y 1,6,13 receptors with PPADS should prevent the inhibitory effect of ADP. As expected, following PPADS treatment (5 min; 25 μM), the amount of released ATP after ADP stimulation increased significantly and was back to control level ( Figure 3F; 59 ± 2%; n = 78 puncta from 3 slices; p < 0.001 vs. ADP and p > 0.05 v. control, one-way ANOVA, Bonferroni's planned comparison test). Activation of P2Y 1,12,13 receptors by ADP can either activate or inhibit PKA via adenylyl cyclase-mediated cAMP production. We therefore hypothesized that PKA might be involved in the ADPinduced inhibition of spontaneous ATP exocytosis. If ADP inhibits PKA, we would expect that PKA inhibition would have the same effect on spontaneous ATP exocytosis as ADP application alone. On the other hand, if ADP activates PKA, the PKA inhibitor should increase the amount of released ATP compared to ADP alone. Incubation with protein kinase inhibitor fragment [14][15][16][17][18][19][20][21][22] (1 hr, 10 μM) prevented ADP-mediated inhibition of the spontaneous ATP exocytosis ( Figure 3F; 57 ± 1%; n = 130 puncta from 3 slices; p < 0.001 v. ADP and p > 0.05 v. control, one-way ANOVA, Bonferroni's planned comparison test). Taken together, these data suggest that stimulation of P2Y 1,13 receptor subtypes that lead to PKA activation downregulates spontaneous exocytosis of ATP occurring during the constitutive ATP release in mouse neonate OE.

Constitutive ATP release induces proliferation
We previously observed that ATP induces neuroproliferation via purinergic receptor activation [9,10]. To determine if constitutive ATP release induces neuroproliferation, we examined the effect of Clostridium difficile toxin A, an inhibitor of constitutive release of ATP on cell proliferation in the OE. Neonatal mice were intranasally instilled with either saline vehicle or 1 or 10 nM toxin A. Three days post-instillation, mice were injected with BrdU and 6 h later tissue was collected to visualize proliferating cells. Clostridium difficile toxin A significantly decreased BrdU incorporation throughout the OE and in the apical region of the OE compared to vehicle control (Figure 4: 2.4 ± 0.6 v. 17.0 ± 0.7 BrdU + cells/mm OE and 0.8 ± 0.4 v. 8.0 ± 0.5 BrdU + cells/ mm apical OE; p < 0.0001, unpaired Student's t-test). While we cannot rule out that the toxin might damage the OE, we would expect that cell damage or cell death caused by exposure to Clostridium difficile toxin A should increase BrdU incorporation because of post-injury neuroregeneration. Thus, taken together, this data suggests that constitutive, toxin A sensitiveexocytosis of ATP or another substance induces increased proliferation and may have a role in the normal turnover of both the olfactory sensory neurons and the sustentacular cells.

Discussion
In this study, we demonstrated the presence of ATPevoked ATP release in neonatal mouse OE slices, and determined the mechanisms underlying this phenomenon. We show that activation of purinergic receptors increases intracellular calcium which could subsequently induce calcium-dependent exocytosis of ATP. Although multiple P2Y and P2X selective purinergic receptor agonists triggered release of ATP, our data suggest that P2X 1,7 receptor subtypes play key roles. Indeed, the selective P2X 1,7 agonist BzATP significantly increases the exocytotic release of ATP. Only small cations are permeable through the P2X 7 receptor under physiological conditions. However, in the maintained presence of high concentrations of ATP, when divalent cation levels are low, the P2X 7 receptor can either form a non-selective pore [29], or, after complexing with pannexin, form a large pore through which ATP can flow [30]. Bz-ATP is maintained for 30s -10 min in these studies, suggesting ample time to open the large pore of the P2X 7 receptor.
In the OE, ATP has been ascribed the functions of neuromodulation, neuroprotection and neuroproliferation by our laboratory and other research groups [9,10,12,13,26,37,38]. The evoked release of ATP could be triggered pathologically by the presence of increased extracellular ATP levels because of ATP leaking from damaged cells. Indeed, toxicant exposure depletes intracellular ATP [39], presumably due to released ATP from damaged cells [37]. The evoked release of ATP could play a role in any of these functions. The sustained release of ATP could either induce the neuroprotective expression of heat shock proteins [37], directly stimulate the progenitor cells to proliferate [9,38] and differentiate [9], or act indirectly by promoting the release of other neurotrophic factors [14]. Thus, the results in this study identify the release mechanisms involved in ATPmediated post-injury neuroregeneration.
We previously showed indirectly that constitutive release of ATP might modulate odorant responsiveness of the olfactory sensory neurons [26]. Here, using luciferin- luciferase assays, we directly confirm that constitutive release of ATP occurs via exocytosis and efflux through ABC transporters. We also demonstrate a negative feedback system for constitutive ATP release regulated by ADP, P2Y receptor stimulation and subsequent activation of PKA. The exact mechanism of constitutive ATP release inhibition is still unknown. To our knowledge, direct inhibition of exocytosis or ABC transporters does not involve PKA. P2Y receptor activation inhibits spontaneous vesicular release of acetylcholine in the mouse neuromuscular junction, but does not require PKA [36,40]. ADP is rapidly degraded to adenosine. Our previous work suggests that adenosine receptors are not expressed in the olfactory epithelium: we observed no responses to adenosine or AMP during confocal calcium recordings [26]. However, adenosine may modulate ADP's activation of P2Y receptors to regulate constitutive ATP release.
We also postulate that constitutively released ATP might play a role in neuronal homeostasis in the OE. Purinergic receptor activation induces cell proliferation [9], has post-injury proliferative and protective functions [10], and elicits upregulation and/or release of many different neurotrophic factors [12][13][14]. Here, we show constitutive exocytosis may induce cell proliferation. Neuronal homeostasis must be tightly regulated to prevent over-stimulation of cell proliferation. The P2Y receptor-mediated inhibition of the constitutive release of ATP would provide negative regulation. Further support for this negative feedback system is provided by the observation that basal progenitor cells of the OE express ectonucleotidases [41]. This provides a mechanism to degrade ATP to ADP, preventing excessive stimulation of P2Y 2 receptors present on the proliferative basal progenitor cells [26], and inhibiting the constitutive release of ATP. Thus, there is local regulation of constitutive ATP release.

Conclusions
We describe constitutive and purinergic receptor-mediated ATP-evoked ATP release via calcium-dependent vesicular exocytosis, pannexin/connexin hemichannels and ABC transporters. Overall, the demonstration that extracellular ATP can elicit ATP release provides a novel pathway with potential pharmacological targets to explore mechanisms of neuroregeneration and that have ramifications for the CNS. Likewise, the role of constitutive release of ATP in neurogenesis under physiological conditions may be utilized to promote neuroregeneration following cell injury in neurodegenerative disease and spinal cord injury.

Solutions
All chemicals were purchased from Sigma-Aldrich (St. Louis, MO, USA) unless noted.

Cell and tissue and culture
Olfactory epithelial slices (400-500 μm) were prepared from Swiss Webster neonatal mice (Charles River, Portage, MI) as previously described [26]. All animal procedures were approved by Michigan State University's IACUC, and all applicable guidelines from NIH were followed. P2X 2 -transfected human embryonic kidney (HEK-293) cells were generously provided by Dr. James Galligan from Michigan State University [43]. They were grown in Dulbecco's modified Eagle's medium (DMEM) F-12 containing 10% fetal bovine serum, 10% GluMax (Invitrogen, Carlsbad, CA, USA), and 100 units mL −1 penicillin and streptomycin. Cells were passaged once every 3 days when they reached 90% confluence. Afterwards, cells were plated on 35 mm coverslips and maintained at 37°C (5% CO 2 ) for 24 h before use in calcium imaging experiments. For each treatment group 3 coverslips were used.
Bioluminescence detection of ATP released from olfactory epithelium slices ATP was quantified using 0.5 mM D-luciferin and 4 μg/ml luciferase in Ringer's solution obtained from an ATP determination kit (Molecular Probes, Eugene, OR). Luminescence, in relative light units, was recorded continuously with 1-s photon collection intervals using a Turner TD20/20 n luminometer. Standard curves of ATP (Mg salt) ranging between 10 nM -5 μM and made using serial dilution from a 0.5 M ATP stock were recorded daily in 200 μl luciferin and luciferase solution. The ATP standard curves were linear in the range of 10 nM -5 μM (r 2 : 0.93-0.99). Individual OE slices (500 μm) were placed in a 200 μl bolus of luciferin and luciferase solution in a 35 mm petri dish and ATP release rates measured in presence or absence of inhibitors. In order to measure purinergic-evoked ATP release using the luciferinluciferase assay, the purinergic agonist should not act as a substrate or co-factor for luciferase. We tested the ability of Bz-ATP, UTP and CTP to produce luminescence in the luciferin-luciferase assay. Bz-ATP (50 μM) induced the smallest increase in luminescence when applied alone to the luciferin and luciferase solution, corresponding tõ 500 nM ATP, whereas UTP and CTP induced a response that corresponded to 5 μM ATP. In pilot studies, the luminescence baseline was stabilized at t = 10 minutes after a slice or Bz-ATP was put into the luciferin and luciferase solution. Therefore, all data were analyzed at 10 minutes by converting the raw relative light unit values to concentration of ATP using the standard curve (relative light units v. ATP concentration (μM)) and interpolation from a linear regression (Prism Software 5.01, Graphpad Software; San Diego, CA). Note that the estimated concentration of released ATP (into 200 μl) is likely to be an underestimate of the amount released into the microenvironment of the tissue. At the end of each experiment, as a positive control, a bolus of Triton X-100 was added (0.5% final concentration) to release ATP via lysis of the cells.

Confocal imaging of fluorescent ATP stores
OE slices were incubated for 30 min at room temperature (23°C) in Ringer's solution with quinacrine (5 μM). Note the diameter of fluorescent punctum (~200-250 nm) that resulted from quinacrine loading was within the optical resolution of confocal fluorescent microscopy. Recordings were performed at a depth of at least 100 μm to avoid damage using an Olympus Fluoview 1000 system (Olympus, Center Valley, PA) [44] and a superfusion flow rate of 1.0 ml/min. A static bath in which agonist or inhibitor were added directly to the bath was used in a few cases due to excessive movement of the slice (n = 6 slices). No statistical differences were observed between results recorded in a static or superfused bath (Table 2). Therefore, additionally, to minimize the application of select inhibitors in limited supply, experiments using the ectonucleotidase apyrase and Clostridium difficile toxin A were performed in a static bath. Quinacrine was excited at 488 nm and fluorescence was detected at 510-525 nm. Images ranged from 512 × 512 pixels to 1600 × 1600 pixels.
Time series experiments were performed at ≥ 1 Hz and data were collected for 400 s, the time required for the fluorescence decrease of most of the puncta recorded in control condition (Ringer's application) to reach a plateau. For each punctum, the percentage of released ATP was determined as 100*[1-(F/F 0 )], where F 0 is the mean fluorescence of the first 10 frames and F is the fluorescence remaining after 400 s. We also calculated the halftime of ATP fluorescence by fitting the change in fluorescence over time with a non-linear regression function (exponential decay) or linear regression function (linear decay). We analyzed only puncta with linear or exponential fluorescence due to the irregularity of the dynamic and mobile puncta.
Z-stack recordings were also performed to minimize changes in the fluorescence of quinacrine-labeled ATP inherent with imaging thick tissue. We recorded for a duration of 500 s by collecting 5 consecutive Z-stacks per slice, each with a duration of 100 s. Ringer's solution or pharmacological compounds were applied at the beginning of the second Z-stack (i.e., at 100 s). In Z-stack recordings, only one focal plane was selected for further analysis based on the ability to distinguish the same cell morphology in relationship to the same puncta of fluorescence within the one focal plane throughout the 500 s recording. For each punctum, the percentage of released ATP was determined as 100*[1-(F/F 0 )], where F 0 is the fluorescence of the punctum in the focal plane of the first Z stack (i.e., t = 0-100 s, depending on the focal plane selected) and F is the fluorescence remaining in the punctum of the same focal plane of the last Z stack, (i.e., t = 400-500 s, depending on the focal plane selected). Note the Z-stack F value is measured~400 s following agonist application to correspond as closely as possible to the time-series experiments. Baseline fluorescence intensity did not vary between slices immediately following quinacrine loading and following long incubations with control Ringer's solution or inhibitors. Therefore, there were no adjustments made to

Confocal calcium imaging
Live cell confocal imaging was performed as previously described on olfactory epithelial slices (400-500 μm) [26] and/or P2X 2 -transfected HEK 293 cells (n = at least 3 slices/coverslips) using an Olympus Fluoview 1000 system. OE slices and P2X 2 -transfected HEK 293 cells were loaded with fluo-4 AM (18 uM; Invitrogen, Carlsbad, CA, USA) for 90 and 30 minutes, respectively, as previously described [26]. Time series experiments were performed collecting 256 × 128 pixel images at 0.7 Hz for slices and ≥ 1 Hz for HEK-293 cells. HEK-293 cells were recorded for 10 minutes as pilot experiments showed that these cells always responded to ATP application within this time frame. When using OE slices, experiments were performed by sequentially superfusing purinergic receptor agonists in a randomized order that differed from slice to slice. The fluorometric signals obtained are expressed as relative fluorescence (F) change, ΔF/F = (F-F 0 )/F 0 , where F 0 is the basal fluorescence level (mean F of first 10 frames). Increases in fluorescence 10% above baseline fluorescence fluctuations were considered responses. Only cells that responded to a final ATP application were included in the data analysis and only responses were included in the analysis for peak height determination.

In vivo study
Neonatal mice (post-natal day 1) were instilled intranasally with 10 μl saline solution ± Clostridium difficile toxin A (1 nM). Each mouse received a re-instillation of saline ± Clostridium difficile toxin A every 24 hrs. Mice received an intraperitoneal injection of bromodeoxyuridine (BrdU; 180 mg/kg total) at 66, 68 and 70 h and tissue was collected at 72 hrs by decapitation. The retronasal passage was flushed with 4% paraformaldehyde (PFA) in 0.1 M phosphate-buffered saline (PBS) and tissue was stored in 4% PFA at 4°C overnight. Tissue was rinsed in 0.1 M PBS, cryoprotected in a 20% sucrose solution at 4°C overnight, placed in dry ice for 10 min, wrapped in parafilm and foil, and stored at −80°C until sectioning. Cryostat sections (20 μm) were collected on superfrost plus slides (Electron Microscopy Sciences, Hatfield, PA), and stored at −20°C.

Immunohistochemistry
To perform BrdU immunohistochemistry, tissue sections were warmed to room temperature, rehydrated with 0.1 M PBS, permeabilized with 0.3% Triton-X in 0. The total number of BrdU + cells was counted over a measured length of the olfactory epithelium lining the turbinates of each section. BrdU + cells found in the apical region of the turbinates were also counted. n = 15, 17, 19 sections from 3 and 4 animals for control and 1 nM toxin A, respectively. All other immunohistochemistry was performed by rehydrating paraformaldehyde-fixed tissue sections from adults (20 μm) collected as previously described [9] with 0.1 M PBS with 0.1% triton X-100 for 20 min, and blocked with bovine serum albumin (BSA) (3%, P2X7; 5%, SNAP 23) in 0.1 M PBS with triton X-100 for 60 min. Primary antibodies were added to tissue sections overnight. Goat-anti P2X 7 antibody (1:20; Santa Cruz, CA) and rabbit-anti SNAP 23 + goat-anti olfactory marker protein (OMP) (1:200; Abcam, Cambridge, MA and 1:250, Waco Chemicals, Plano, TX, respectively) antibodies were made in 1% BSA in 0.1 M PBS. FITC-conjugated donkey antirabbit or donkey anti-goat antibody (1:200; Jackson ImmunoResearch Labs, West Grove, PA) was applied for 1 hr at room temperature. Sections were then washed and mounted in Vectashield mounting medium for fluorescence (Vector Labs, Burlingame, CA) and visualized on an Olympus FluoView 1000 laser scanning confocal microscope. FITC dye was excited at 488 nm and low pass filtered at 505-525 nm. Antibody specificity was tested by omitting the primary antibodies. Additionally, a peptide neutralization protocol was performed by combining P2X 7 antibody (0.04 mg/ml) with a 10-20 fold excess of the immunizing peptide (0.4-0.8 mg/ml). No immunoreactivity was observed in any of the controls.

Statistical analysis
We employed either the Student's t-test or a one-way ANOVA followed by Bonferroni's planned comparison test without correction for multiple comparisons to compare data sets, as indicated in text, using Prism Software 5.01 (Graphpad Software; San Diego, CA). A value of p < 0.05 was considered statistically significant.
immunohistochemistry. CCH conceived of the study, participated in its design and coordination, carried out the confocal calcium imaging study using neonatal OE slices, and wrote the manuscript. All authors read and approved the final manuscript.