Increased Anion Channel Activity Is an Unavoidable Event in Ozone-Induced Programmed Cell Death

Background Ozone is a major secondary air pollutant often reaching high concentrations in urban areas under strong daylight, high temperature and stagnant high-pressure systems. Ozone in the troposphere is a pollutant that is harmful to the plant. Principal Findings By exposing cells to a strong pulse of ozonized air, an acute cell death was observed in suspension cells of Arabidopsis thaliana used as a model. We demonstrated that O3 treatment induced the activation of a plasma membrane anion channel that is an early prerequisite of O3-induced cell death in A. thaliana. Our data further suggest interplay of anion channel activation with well known plant responses to O3, Ca2+ influx and NADPH-oxidase generated reactive oxygen species (ROS) in mediating the oxidative cell death. This interplay might be fuelled by several mechanisms in addition to the direct ROS generation by O3; namely, H2O2 generation by salicylic and abscisic acids. Anion channel activation was also shown to promote the accumulation of transcripts encoding vacuolar processing enzymes, a family of proteases previously reported to contribute to the disruption of vacuole integrity observed during programmed cell death. Significance Collectively, our data indicate that anion efflux is an early key component of morphological and biochemical events leading to O3-induced programmed cell death. Because ion channels and more specifically anion channels assume a crucial position in cells, an understanding about the underlying role(s) for ion channels in the signalling pathway leading to programmed cell death is a subject that warrants future investigation.


Introduction
Ozone produced by a complex series of photochemical reactions from primary precursor emissions of nitrogen oxide and volatile organic compounds, is a major secondary air pollutant. Chronic exposures to low O 3 concentrations have a negative impact on crop yields by reducing photosynthesis and growth, and inducing premature leaf senescence in sensitive plants [1]. Acute transient O 3 exposures cause cell death, visible as O 3 lesions in the leaves [2] that resemble a programmed cell death (PCD) associated with the hypersensitive response (HR) in plantpathogen interactions [3,4,5]. The primary site of O 3 interaction within plant cells is the apoplast where O 3 challenges the antioxidant protection of the cell. Accordingly, O 3 sensitivity generally correlates with the ascorbate status of the apoplast, which is an important signal initiation point [6]. Ozone also breaks down into various reactive oxygen species (ROS) namely hydrogen peroxide, singlet oxygen and hydroxyl radicals [6]. A quick burst of ROS produced by plant cells is also induced by high concentrations of O 3 [3] and it resembles the oxidative burst of the HR in incompatible plant-pathogen interactions [7]. In plants, an exposure to O 3 , also leads to a rapid increase in cytosolic free calcium ([Ca 2+ ] cyt ) [8,9,10]. This increase in cytosolic Ca 2+ concentration is sensitive to Ca 2+ chelators, ion channel blockers, and ROS scavengers thus suggesting that calcium influx from the apoplast acts as a secondary messenger initiating oxidative cell death [4,10] in addition to rapid changes in the protein phosphorylation pattern [11]. The production of ROS finally leads to a PCD characterized by the early release of cytochrome c from mitochondria, activation of proteases, DNA fragmentation, electrolyte leakage and ultrastructural changes characteristic of PCD [3,4,5,10]. Responses that follow the O 3 challenge also include changes in gene expression and protein synthesis, and an accumulation of plant stress hormones, ethylene, salicylic acid (SA), abscisic acid (ABA) and jasmonic acid (JA) [4,12,13,14,15,16], that regulate the induction and spreading of the oxidative stress symptoms [3,14,17,18]. Treatment with O 3 also induces a rapid accumulation of NO, which could coincide with the formation of HR-like lesions, suggesting that NO is also an important signalling molecule in the plant response to O 3 [19].
Anion channels play fundamental roles in key biological processes including plant cell response to environmental stresses [20,21,22]. Several types of anion channel differing in their voltage dependence, kinetic properties and anion selectivity have been characterized, mostly by electrophysiological techniques. R-type anion channel activation is an essential step of the ROS-dependent innate immune response in Arabidopsis suspension cells [23]. In the same model system, ROS generation and change in [Ca 2+ ] cyt participate to ABA-induced anion channel regulation [24,25,26]. Recently, an anion channel SLAC1 [27,28,29,30,31] was shown to be essential for stomatal closure in response to ozone, Ca 2+ ions and H 2 O 2 in Arabidopsis [31], suggesting that anion efflux could be induced in guard cells by O 3 through an increase in anion channel activity. Anion effluxes are also amongst the earliest responses observed in plant cells following recognition of pathogenic signals. For instance, cryptogein induces a rapid and massive activation of anion channel mediated nitrate efflux regulated by Ca 2+ -dependent events [32,33]. This NO 3 2 efflux was shown to be necessary for the mediation of the cryptogeininduced oxidative burst, the induction of defence-related genes and the development of the HR. In this model, NO 3 2 efflux was also shown to promote the accumulation of transcripts encoding vacuolar processing enzymes (VPEs) [33], a family of proteases showing caspase-1 activity [34] and reported to contribute to the disruption of vacuole integrity observed during the HR [35]. Anion channel regulation was also shown to be an essential component of harpin-and oxalic acid-induced plant PCD [36,37]. The involvement of anion channels as a critical component of the cell death process in plants is similar to their key role in animal apoptosis. Indeed, several studies have reported that in various types of mammalian cells, the activation of a plasma membrane Cl 2 channel is an early prerequisite to apoptotic events including cell shrinkage (termed AVD for apoptotic volume decrease), cytochrome c release, the activation of proteases (including caspases) and nucleases and, ultimately, cell death [38,39]. Taken together, these data suggest that anion channel activation is not a passive secondary feature of plant responses to stress, but a driver of these processes.
Recent research has shown that the similarity of O 3 -induced cell death and hypersensitive cell death is not only external [2,40] since these phenomena also share many physiological and molecular features [3,11]. Therefore, we have analyzed the role of anion channels in O 3 -induced cell death signalling pathways using Arabidopsis thaliana cells as a model. It was found that an O 3 challenge induced the activation of a plasma membrane anion channel which was an early prerequisite of O 3 -induced cell death in A. thaliana. This oxidative cell death was mediated by the interplay of anion channel activation, Ca 2+ influx, ROS generation and included an increase in VPE transcripts. Furthermore, ABA and salicylic acid appear to participate to this anion channel activation leading to O 3 -induced cell death.

O 3 -induced cell death in A. thaliana suspension-cultured cells
We first checked if O 3 induced cell death in our model plant system by exposing Arabidopsis cell cultures to a pulse of ozonized air. Vacuole shrinkage was observed which led to completely collapsed cells and finally cell death as determined by Evans blue or neutral red staining ( Figure 1A). The percentage of dead cells reached a plateau about 2 h after O 3 treatment ( Figure 1B), the degree of cell death being dependent on O 3 treatment duration, with about 50% and 80% of dead cells detected 2 h after a 3 min and a 10 min O 3 treatment, respectively ( Figure 1B). The number of dead cells was similar when quantified by the fluorescein diacetate (FDA) spectrofluorimetric method ( Figure 1C). In order to check whether this O 3 -induced cell death was due to an active mechanism requiring active gene expression and cellular metabolism, A. thaliana cell suspensions were treated with actinomycin D (AD), an inhibitor of RNA synthesis, or with cycloheximide (Chx), an inhibitor of protein synthesis, at 20 mg.mL 21 each, 15 min prior to O 3 exposure. Actinomycin D and Chx significantly reduced the O 3 -induced cell death ( Figure 1D). These results indicated that this cell death required active cell metabolism, namely gene transcription and de novo protein synthesis. Fragmentation of nuclear DNA was observed by agarose gel analysis of DNA extracted from cell suspensions after a 10 min treatment with O 3 ( Figure 1E). This DNA fragmentation was also dependent on active gene expression and de novo protein synthesis since it was not detected after the addition of AD or Chx to the suspension cell cultures ( Figure 1E). Taken together, these data confirm that O 3 induced a PCD in A. thaliana cells.

Activation of anion channels is a crucial early event in O 3 -induced cell death
Cell shrinkage is a major hallmark of PCD. This process may be mediated by a net efflux of water resulting from the release of anions and K + . Indeed, anion efflux, detectable as a current increase, has been reported to be a necessary event to achieve cell death in suspension cells subjected to either cryptogein [33] or oxalic acid [37]. Since ozone induced cell shrinkage ( Figure 1A), an electrophysiological approach was undertaken to determine the role of O 3 on cell membrane potential and on anion currents. Two protocols were used to assess the O 3 effect on cell polarization: (i) free-running PM potential time-courses were recorded in cells exposed to ozonized air ( Figure 2A) and (ii) the mean PM potentials of cell populations exposed to ozonized air for 3 or 10 min were compared to those of cells pretreated by air alone during 10 min (Figure. 2B). The value of the resting membrane potential (Vm) of control cells (air treated or without treatment) were similar, 234.861.5 mV (n = 23) and 235,361.4 mV (n = 22) respectively and in the same range of previous studies [26,41,42,43,44]. A rapid depolarization of the cell PM in response to ozonized air was detected with the first protocol ( Figure 2A). Accordingly, 3 or 10 min O 3 -exposed cells showed a depolarized PM compared to cells treated with air. The amplitude of the observed PM depolarization depended on the duration of the O 3 treatment ( Figure 2B). Previous electrophysiological studies and pharmacological analyses identified a current displaying the characteristics of anion channels in the PM of A. thaliana cells [24,41]. This current was shown to be sensitive to structurally unrelated anion channel inhibitors, 9-anthracen carboxylic acid (9-AC) and glibenclamide (gli) [24,42]. It presented features of slow anion channels [41], slow activation upon hyperpolarization and slow deactivation upon depolarization [45], although part of the instantaneous current could have been carried out by fast activating anion channels as described for guard cells [46]. Since long hyperpolarizing or depolarizing voltages could artificially modify the ionic content of our living cells, we recorded the signature of this current using shorter voltage pulses [41]. In accordance to the PM depolarization, O 3 induced an increase of anion current after 4 min ( Figure 2C). A current showing the features of slow anion channels and a sensitivity to glibenclamide was also detected when cells were exposed to O 3 ( Figure 2D). When compared to air treated cells, the anion current amplitude increase depended on O 3 treatment duration ( Figure 2E), reaching 200% of the mean control value of 21.1760.05 nA (n = 21). The increase in anion current, that correlated to the PM depolarization amplitude ( Figure 2B), might explain the depolarization induced by O 3 since pretreatment of cells with gli or 9-AC (200 mM) drastically reduced the O 3 -induced depolarization and anion current increase ( Figure 2F-G). Therefore, the effect of anion channel blockers on the extent of O 3 -induced cell death was tested. Ozone (10 min treatment) induced around 80% of cell death within 6 h ( Figure 1B-C). When treated with anion channel blockers, gli or 9-AC (200 mM), cell death was reduced by approximately 50% ( Figure 2H). These results suggested that the anion current increase was a required upstream event in the signaling pathway leading to O 3 -induced cell death.

Ozone-induced calcium influx
Increases in [Ca 2+ ] cyt in response to O 3 have been reported previously in Arabidopsis seedlings [4,8,9] and in tobacco cells [10]. Moreover, addition of Ca 2+ chelators or Ca 2+ channel blockers resulted in a significant inhibition of an O 3 -induced [Ca 2+ ] cyt increase and subsequent cell death suggesting that the uptake of extracellular Ca 2+ via the activation of PM Ca 2+ channels is required for the induction of active cell death [4,10]. A change in [Ca 2+ ] cyt was detected in Arabidopis cells expressing aequorin in response to O 3 , as shown in Figure 3A. The observed changes in aequorin luminescence were biphasic, consisting of an immediate small increase in [Ca 2+ ] cyt immediately after the initiation of O 3 exposure that was followed by a larger increase when O 3 exposure was stopped ( Figure 3A). This increase is maintained and reached a plateau after 50 min (data not shown). The Ca 2+ channel blocker, La 3+ (500 mM) inhibited the second phase of [Ca 2+ ] cyt increase after O 3 exposure ( Figure 3B), indicating that the influx of extracellular Ca 2+ into the cells was via an activation of Ca 2+ channels. Addition of 500 mM La 3+ also resulted in the significant inhibition of O 3 -induced cell death ( Figure 3C). Similarly, addition of 1 mM BAPTA, a membraneimpermeable Ca 2+ chelator that is active at a physiological pH range, resulted in the inhibition of O 3 -induced cell death ( Figure 3C). These results suggested that the uptake of extracellular Ca 2+ was also required for cell death induction in A. thaliana cells. We thus tested the effect of BAPTA and La 3+ on the O 3induced increase in anion current and the subsequent cell depolarization. Pretreatment with BAPTA or La 3+ significantly lowered the O 3 -induced depolarization and increase in anion   Figure 3D-E). These results suggested that an increases in [Ca 2+ ] cyt was also an early upstream event in the signaling pathway leading to O 3 -induced cell death in A. thaliana cells, and also suggested the involvement of the anion current increase in this signaling pathway.

Effects of ROS scavengers
The generation of ROS, such as superoxide anions (O 2˙2 ) and hydrogen peroxide (H 2 O 2 ) by O 3 degradation in the apoplast, is a recognized mechanism involved in O 3 -induced damage [6,11]. Addition of ROS scavengers or chelators was shown to inhibit O 3 induced cell death, indicating that singlet oxygen ( 1 O 2 ), and H 2 O 2 generated from O 2˙2 by superoxide dismutase play central roles in acute O 3 -induced damage to tobacco cells [10]. We thus checked the effect of DABCO (a strong and selective scavenger of 1 O 2 ), tiron (a scavenger of O 2˙2 ) and diphenyleneiodonium chloride (DPI, an inhibitor of the NADPH-oxidase), on O 3 -induced ROS generation. ROS generation was monitored by Cypridina luciferin analog (CLA) chemiluminescence which reports the presence of both O 2˙2 and 1 O 2 (but to a lesser extent). Exposure of Arabidopis cells to O 3 resulted in a biphasic enhancement of the CLA-chemiluminescence yield ( Figure. 4A). An initial increase peaked at about 1 min while a second peak was observed at about 5 min. Since CLA is responsive to both O 2˙2 and 1 O 2 , tiron, DPI and DABCO were used to determine which ROS was involved [47]. Pretreament of the cells with DABCO (5 mM) allowed a significant decrease of the first peak while the 2nd peak remained unchanged ( Figure 4A). Pretreatment of the cells with DPI (50 mM) or tiron (5 mM) failed to effect the first peak but drastically decreased the second peak of CLA-chemiluminescence ( Figure 4A). Therefore, the first rapid chemiluminescence increase in the presence of CLA appeared to reflect the production of 1 O 2 (inhibited by DABCO), while a delayed production of O 2˙2 (possibly converted to H 2 O 2 ) is supported by the action of tiron and DPI (inhibition of the 2 nd peak). The effect of DABCO, tiron, and DPI on O 3 -induced A. thaliana cell death was also analysed after the cell suspensions were treated with DABCO (5 mM), tiron (5 mM) or DPI (50 mM) five minutes prior to O 3 exposure. The O 3 -induced cell death was significantly reduced in the presence of DABCO, tiron and DPI ( Figure 4B). Taken together, ROS generation is a likely component of the pathway leading to O 3dependent damage of A. thaliana cells, as previously suggested for tobacco and Arabidopsis seedlings [10,48]. Since (i) influx of Ca 2+ could act upstream of the O 3 -induced activation of anion channels ( Figure 4E) and (ii) H 2 O 2 was shown to activate PM Ca 2+ channels [49] and increase [Ca 2+ ] cyt in our model [25], we further tested whether H 2 O 2 could activate anion channels. Indeed, application of 500 mM H 2 O 2 induced an increase in anion current ( Figure 4C) suggesting that O 3 -induced  Figure 4D). The anion channel blockers gli and 9-AC (200 mM) were also found to be capable of decreasing the CLA luminescence increase observed upon O 3 challenge ( Figure 4E). Glibenclamide led to a decrease of only the second peak of O 3 -induced CLA-luminescence where as 9-AC decreased both peaks. Calcium channels blocker lanthanum or chelator BAPTA failed to decrease rapid ROS generation ( Figure 4E). As a whole, these data suggested a complex interplay between anion channel regulation, Ca 2+ influx and H 2 O 2 production in response to O 3 . We thus checked if incubation, prior O 3 exposure, with anion channel blokers (gli or 9-AC) or with the O 2˙2 scavenger tiron and DPI could impact on O 3 -induced [Ca 2+ ] cyt increase. Effectively, anion channel blockers as H 2 O 2 pharmacology (tiron and DPI) could reduced the late increase of cytosolic calcium ( Figure 4F) confirming thus the interplay between anion channel regulation, Ca 2+ influx and H 2 O 2 production in response to O 3 .

The O 3 -induced anion channel activation participates in vacuolar processing enzyme (VPE) gene expression
In mammalian cells undergoing AVD, the ion loss triggers activation of specific proteases [38]. Based on these data and on our finding that anion channel activity is involved in mediating vacuole shrinkage and O 3 -triggered cell death (Figure 1), we investigated whether the O 3 -induced anion efflux could be a key event in a signaling cascade leading to protease activation. Therefore, a putative role for the anion efflux in mediating the accumulation of mRNA encoding VPEs was investigated. This protease family has been shown to be essential for HR induction in tobacco challenged by pathogens [34,50], and their transcription is dependent on anion efflux in response to cryptogein [33]. Four VPE genes have been identified in Arabidopsis, namely VPEa, VPEb, VPEd, and VPEc [51]. To investigate the involvement of VPEs in O 3 -induced effects, we analyzed VPE mRNA accumu- lation in O 3 -treated cells by RT-PCR. Transcripts for VPEa, VPEb and VPEd were not detected in our model system (data not shown) however, the mRNA level of VPEc increased in response to O 3 with the changes rapidly occurring within 30 min ( Figure 5) thus supporting the idea that VPEs participate in O 3 -induced damage in Arabidopsis. When Arabidopsis cells were treated with gli, 9-AC or DPI before the O 3 challenge, accumulation of VPEc transcript was reduced ( Figure 5). Therefore, anion channel activation and H 2 O 2 generation could be involved in the pathway leading to a transcriptional regulation of VPEc transcripts.

Hormones and anion channels in response to O 3
Since the plant hormones SA, JA, ABA and ET as well as NO are involved in determining the duration and extent of O 3 -induced cell death and its propagation [3,11,19,52], their impact was checked in our model system. Ozone-induced cell death was analysed in suspension cells generated from NahG, cpr5 and npr1 plants, which are impaired in SA signalling, the sid2 mutant, which is impaired in SA synthesis through the isochorismate pathway, and the JA-resistant mutant jar1-1. For ethylene, ABA and NO, pharmacological approaches were undertaken with (i) aminooxyacétique acid (AOA), an inhibitor of ACC synthase [53], and alpha-aminoisobutyric acid (AIB), an inhibitor of ACC oxidase [54] for ET, (ii) fluridon an inhibitor of ABA synthesis [55] and (iii) PTIO, a scavenger of NO [56]. Ozone-induced cell death levels recorded after pretreament of the cells with AOA, AIB and PTIO or with the jar1-1 cell line were not significantly different from that after O 3 treatment alone in Col-0 background ( Figure 6A), indicating that JA, ET and NO are not major actors in the signalling pathways leading to O 3 -induced cell death in Arabidopsis cultured cells. On the other hand, ABA synthesis appeared to take place in response to O 3 since pretreatment of cells with fluridon counteracted the O 3 effect ( Figure 6A). In a similar manner, the O 3 -induced cell death extent in the NahG, cpr5 and npr1 cell lines showed significant decreases ( Figure 6B), revealing that O 3 -induced death depended on SA signalling in Arabidopsis cultured cells. Interestingly, the sid2 cell line showed the same degree of O 3 induced cell death as the wild-type Col-0 line ( Figure 6B) suggesting that SA synthesis via the isochorismate pathway was not required for cell death induction in our cells.
As relatively high concentration of SA could be rapidly released in the apoplast from the storage form salicylic acid-glucoside (SAG) [57,58], we tested the putative impact of SA on anion channel activation. Salycilic acid at 200 mM, a physiological concentration, induced a slight but rapid hyperpolarisation of the cells followed by a large depolarisation within a few minutes ( Figure 6C-D). This biphasic regulation of the PM potential was accompanied by the biphasic regulation of anion channel activity, the delayed depolarization being correlated with an increased anion channel activity ( Figure 6E-F). Thus, SA is not responsible for the early depolarization induced by O 3 but it could participate to the anion channel-mediated depolarization in a delayed manner.

Discussion
In this study we demonstrated that an acute exposure of Arabidopsis cells to ozone induced a controlled cell death displaying nuclear DNA fragmentation that required active gene expression and de novo protein synthesis. However, a typical laddering as previously reported in O 3 treated tobacco leaves [5] was not observed. Ozone-induced cell death was only partially decrease by AD or Chx treatment whereas no DNA fragmentation was detected when cells were pretreated with these chemicals.
Thus, we can not exclude that a small proportion of the cell death observed could be due to non active process. However, O 3 also induced cell shrinkage, another hallmark of the PCD process in both plant and animal cells [35,59]. These data fulfill the widely accepted criteria for PCD and confirm that our model responds to acute O 3 exposure in the same way as previously described for other plants [3,4,5,10].
We analyzed the putative role of anion channels in O 3 -induced cell death signalling pathways by using the microelectrode voltage clamp technique which allows working on living cells with their cell wall. An increase in anion current and a depolarization of the PM were observed after treatment of cell suspensions with O 3 . As previously discussed [24,41], the anion currents we recorded showed the characteristic kinetic features of S-type anion channels responsible for long-term anion efflux and depolarization in guard cells [60]. The addition, before an O 3 treatment, of 9-AC or glibenclamide, two structurally unrelated anion channel inhibitors was shown to be effective in A. thaliana suspension cells [37,42], strongly reduced the anion current and partially prevented the O 3 -induced PM depolarization thus suggesting the participation of slow anion channel currents. Such a causal link has been described already with anion channel inhibitors which block blue light-induced depolarization [61] or toxin-induced depolarization [33,44]. Activation of rapid-and/or S-type anion channels by ABA has been shown also to lead to the PM depolarization of guard cells in Vicia faba and Arabidopsis thaliana [60,62]. In guard cells the long-term anion efflux and sustained depolarization has mainly been attributed to the activity of S-type anion channels. This, together with our finding that O 3 induced anion currents displaying characteristic of slow type anion channels, suggests that the mechanism by which O 3 promotes PM depolarization resembles that occurring in guard cells. It is noteworthy that recently, SLAC1, which represents the slow type anion channel of guard cells [28,29], was shown to be essential for stomatal closure in response to O 3 , Ca 2+ and H 2 O 2 [31]. However, our data are the first direct evidence that O 3 can regulate anion channel activity and that this anion current increase is an early prerequisite to the morphological and biochemical events participating to PCD. The activation of anion currents appeared to be important in the O 3 response since 9-AC and gli decreased the induced cell death. Involvement of ion release via anion flux modulation is also considered to be essential among the earliest responses of plant cells to avirulent pathogens or elicitors capable of inducing PCD [63]. Our data pointed out a critical role for anion channels in the signaling pathways leading to cell death. In various mammalian cell types, AVD, which is mediated by water loss caused by activation of anion channels and a K + outward rectifying channel, is an early prerequisite to apoptotic events including cell shrinkage, cytochrome c release, activation of proteases (including caspases) and nucleases, and ultimately PCD [39,64]. Indeed, activation of outward K + channels was reported in response to O 3 in guard cell protoplasts [63,65]. Thus, efflux of anions and K + might drive water efflux leading to the observed cell shrinkage and death in response to O 3 .
The involvement of ROS generation, increases in [Ca 2+ ] cyt and the crosstalk between these events in response to O 3 is now widely accepted [3,4,8,9,10,11]. Furthermore, variations in both ROS and [Ca 2+ ] cyt levels are involved in the regulation of anion channels in response to ABA in Arabidopsis cells [24,25]. In our model system, although the first rapid increase in [Ca 2+ ] cyt did not resemble that reported in Arabidopis seedlings [9], the 2nd delayed increase was sensitive to La 3+ , thus suggesting that O 3 induced an influx through PM Ca 2+ channels [8,9]. ROS generation was also detected in our experimental system. The impact of tiron and DPI strongly suggested that O 2˙2 and subsequently H 2 O 2 were produced by NADPH-oxidase in response to O 3 as previously reported [16]. As expected from these data, a significant inhibition of O 3 -induced cell death was found after addition of BAPTA, La 3+ , tiron or DPI [4,10]. These events appear to be linked to O 3 -induced anion channel increase activity since we observed that H 2 O 2 , which induced an increase in [Ca 2+ ] cyt in our model [25] probably through PM Ca 2+ channel activation [49], was also capable of increasing anion channel activity. Accordingly, the O 2˙2 scavenger tiron and NADPHoxidase inhibitor DPI were able to decrease the late [Ca 2+ ] cyt variation. BAPTA and La 3+ led to a decrease of the O 3 -induced depolarization and anion channel activation in accordance with the sensitivity of PM anion currents to an increase in [Ca 2+ ] cyt [24,26] but have no effect on rapid ROS generation. Taken together, these data suggest that Ca 2+ influx and ROS generation probably act upstream of anion channel regulation. However, the anion channel inhibitors gli and 9-AC were also shown to decrease O 3 -induced O 2˙2 generation and [Ca 2+ ] cyt variation indicating a more complex interplay between ROS, Ca 2+ and ion channel activation in signal transduction processes. A putative explanation might be that a first non biological ROS generation, from O 3 reacting with the ascorbate pool [6], induced the Ca 2+ influx required for the activation of an anion channel. In turn, the Ca 2+activated anion channel and the ensuing PM depolarization possibly amplify the O 3 signal by activating a PM NADPHoxidase as recently described in animal cells [66,67]. The H 2 O 2 derived from the NADPH-oxidase activity then participates in the increase of [Ca 2+ ] cyt , through the activation of PM Ca 2+ channels [9,25,49] acting as a feedback loop on anion channel activation. This scenario is reminiscent of the activation of the oxidative burst in aequorin-transformed Nicotiana tabacum cells which was shown to be mediated by an anion channel-dependent increase in [Ca 2+ ] cyt [68,69]. However it is to be note that this ''wheel'' of interplay (Figure 7) might be fuelled by several other entries such salycilic acid which could also fuel the generation of H 2 O 2 and Ca 2+ influx involved in O 3 -induced cell death. SA-induced generation of ROS by cell wall peroxidases [70,71] and by NADPH-oxidase [58] is effectively known to lead to the Ca 2+ influx, and this could explain the delayed increase in anion current observed in response to the application of exogenous SA. Active SA could be released from the inactive SAG pool in apoplast through SAGase action [58,72]. This could explain why the extent of cell death could be minimized after an O 3 challenge in SA-related signalling cell lines (NahG, cpr5 and npr1), while that cell death level in the sid2 mutant impaired in SA synthesis could not be. Our data clearly confirm that SA is involved in the amplification of O 3 -induced cell death [4,16]. In the same way, ABA synthesis could also fuel this mechanism since this hormone is known to induce ROS generation, cytosolic Ca 2+ increases and anion channel increases in Arabidopsis cells [25]. The impact of ABA on ion fluxes in response to O 3 is reminiscent of the ABA effect on stomata allowing sustained ion efflux leading to a decrease in cell volume. This is in accordance with the recent observation that SLAC1 is essential for stomatal closure in response to ozone, ABA, Ca 2+ ions, and H 2 O 2 in Arabidopsis [31]. Thus, in our model SA release and ABA synthesis induced by O 3 could participate to sustain PM depolarization and anion effluxes leading to cell shrinkage.
The mitochondrial pathway was shown to be implicated in O 3triggered PCD [5]. However, it did not appear to be linked to anion channel activation since glibenclamide which decreased the activation of O 3 -induced anion channels, thus leading to less cell death failed to stop the O 3 -induced Dym ( Figure S1). Nevertheless, DABCO, a 1 O 2 scavenger which decreased the 1 O 2 level and cell death in our model, inhibited the effect of O 3 on Dym ( Figure  S1). The reaction of O 3 with ascorbate is known to lead to high yields of singlet oxygen [6] and therefore the detection of a rapid 1 O 2 production in response to O 3 was not a surprise. However, the involvement of 1 O 2 in O 3 -induced cell death might not involve the same pathway as that induced by anion channels, thus reinforcing the idea that several different pathways leading to cell death are triggered in response to O 3 . Further studies are needed to understand the role of 1 O 2 in O 3 -induced cell death.
To further assess the mechanisms by which the anion current increase contributes to cell death, its participation in the transcriptional activation of VPEs identified as key players in plant PCD [73] was explored. VPE is a family of cysteine proteases which exhibit enzymatic properties similar to that of caspase-1, a cysteine protease involved in the PCD pathway in animals. Our study was based on the finding that O 3 induces protease activities [4,5] and that VPE transcript levels are dependent on anion channel mediated NO 3 2 effluxes during cryptogein-induced cell death in tobacco [33]. Our results show that O 3 treatment upregulates the expression of VPEc and not the other Arabidopsis VPE-encoding genes. In the presence of 9-AC, glibenclamide or DPI, the accumulation of VPEc transcripts was reduced. The activation of O 3 -induced NADPH-oxidase and anion channels could thus be early prerequisites for VPEc synthesis. Our data are in accordance with the fact that VPE activities can be increased by SA treatment and that expression of VPEc is also transiently upregulated during the early phase of HR activation [35,74], at judged by increased transcript levels. Although the VPE target(s) that mediates HR cell death is unknown, Hatsugai et al. [34] nevertheless observed a dramatic inhibition of vacuole collapse in VPE-suppressed plants. Thus a VPE-dependent disruption of vacuole integrity might be a crucial step in O 3 -induced vacuolar shrinkage and cell death, as it is for some forms of HR cell death [33,34,73]. Therefore, anion channel activation might not be a passive secondary aspect of O 3 -induced cell death, but an event that drives the whole process.
Plant hormones as well as NO are involved in determining the duration and extent of O 3 -induced cell death and its propagation [3,11,19,52]. The extent of O 3 -induced cell death with the JAresistant mutant jar1-1 and cells treated with pharmacological agents that block ethylene or NO synthesis was similar to that It should be noted that methyl-jasmonate, jasmonic acid, diethylamineNONOate, used as an NO donor [56], or the ET generator aminocyclopropane carboxylate failed to increase anion channel activity under our experimental conditions (data not shown). However, our data are not in opposition with the involvement of JA and ET in determining the duration and extent of O 3 -induced cell death and its propagation [3]. In the same way, NO might be needed to generate the intracellular signals required for the cell-to-cell spreading of an O 3 -induced HR, but not necessary to induce a PCD [19]. In our experiments all of the cells were challenged by O 3 , thus systemic messages could not impact on the degree of cell death. Our data are thus in accordance with a direct role for SA and ABA in O 3 -induced cell death while JA, ET and NO probably act only in auxiliary signalling pathways, which stabilize and amplify the primary signal without excluding a role for SA and ABA in these auxiliary signalling pathways [3,11,75].
In summary, this work shows that anion channel activation is central to the signalling cascade leading to O 3 induced-cell death and provides evidence that anion movements are tightly correlated to the cellular and molecular events involved in this process. Indeed, anion channels are now recognized as important players in signaling pathways associated with plant cell responses to abiotic and biotic environmental stresses [21,22] and our findings highlight the notion that plants, like animals, use anion channels as components of cell death pathways.

Cell culture conditions
For this study, Arabidopsis thaliana L. cell line T87 [76] was used. Axelos et al. (1992) [76] have previously established a cell line (named T87) from the ecotype Columbia plant. Suspension cells have been shown to be a convenient model for identifying early physiological events induced by different biotic [32,37,42] and abiotic stress [10,77]. They show physiological responses to various stimuli, in a similar manner to autonomous cellular responses in intact tissues [78], especially the morphological features of dying cells during PCD [79], and thus allow the observation of events in each single cell or the real time behavioral monitoring of large populations of cells. A. thaliana suspension cells were grown in Gamborg or Murashige-Skoog medium (pH 5.8). They were maintained at 2262uC, under continuous white light (40 mE m 22 s 21 ) and continuous shaking (gyratory shaker) at 120 rpm. Cell suspensions were sub-cultured weekly using a 1:10 dilution. All experiments were performed at 2262uC using logphase cells (4 days after sub-culture).

Preparation of Arabidopsis mutant or transgenic cell lines
For the cell suspension cultures derived from Arabidopsis mutants (jar1-1, sid2-1, cpr5 and npr1) and transgenic lines (NahG and apoaequorin), the seeds of mutants and transgenic lines were sterilized in 1% (w/v) sodium hypochlorite and allowed to germinate on sterilized MS agar plates containing vitamin B5, but lacking 2,4-dichlorophenoxy acetic acid (2,4-D). The seedlings were grown on the agar plates under a 12/12 h light/dark regime at 2361uC for three weeks. Excised tissues from harvested seedlings were transferred onto agar medium containing 0.2 mg/ ml 2,4-D to promote callus formation. Suspension cultures of cells were initiated by addition of cut pieces of the resulting calli to the MS or Gamborg liquid medium (pH 5.8) containing 0.2 mg/ml 2,4-D. The cell suspension cultures (30 ml each in 100 ml conical flasks) were kept on gyratory shakers (120 rpm) at 2262uC under continuous light, and sub-cultured using 30 ml of 7-day cultures as inocula. Cells were harvested for the O 3 experiments 4-days after sub-culturing.

Ozone exposure
Ozone exposure of the cell suspension was performed as previously described [10]. Ozonized air (0.1 L/min; 10 mg O 3 /h) was passed on the surface of the cell suspensions (250 mL in 4 mL tubes). By this way the cells could be exposed to the pulse of O 3 for 3 or 10 min. Ozone was generated by a ceramic ozonizer (NAVI Super Ceramics Ozonizer EO-mini, Kenis Kagaku Kyoeisha Ltd., Tokyo, Japan), equipped with an air pump.

Cell viability assays
Cell viability was assayed using the vital dye, Evans blue, after air or ozone treatment with or without the appropriate pharmacological effectors (pretreatment of 15 min prior O 3 exposure). Cells (50 ml) were incubated for 5 min in 1 ml phosphate buffer pH 7 supplemented with Evans blue to a final concentration of 0.005%. Cells that accumulated Evans blue were considered dead. At least 1000 cells were counted for each independent treatment and repeated at least 4 times for each condition.
Cell viability was also checked using fluorescein diacetate (FDA) as previously described [43]. Briefly, after the appropriate treatment, 1 mL of cell suspension was gently stirred with a magnetic stirrer before FDA was added to a final concentration of 12 mM. The fluorescence increase was monitored over a 120 s period using a F-2000 spectrofluorimeter (Hitachi, Japan). Results are presented as the percentage of cell death = (slope of treated cells/slope of non treated cells) * 1006SE. The experiment was repeated at least 4 times for each condition.

DNA extraction and analysis
Frozen cells were ground in liquid nitrogen and genomic DNA was extracted according to the CTAB method [80]. DNA electrophoresis was performed to assess DNA fragmentation. DNA samples (5 mg/lane) were loaded onto a 1.8% agarose gel including 0.2 mg/ml ethidium bromide.

Electrophysiology
Individual cells were impaled and voltage-clamped in the culture medium using an Axoclamp 2B amplifier (Axon Instruments, Foster City, CA, USA) for discontinuous single electrode voltage clamp experiments as previously described [26,37,43]. Voltage and current were digitized using a computer fitted with a Digidata 1320A acquisition board (Axon Instruments). The electrometer was driven by pClamp software (pCLAMP8, Axon Instruments). Experiments were conducted on 4-day-old cultures at 2262uC (main ions in the medium after 4 d of culture: 9 mM K + , 11 mM NO 3 2 , [41]).

Aequorin luminescence measurements
Cytoplasmic Ca 2+ variations were recorded using freshly generated A. thaliana cell suspensions (T87) expressing apoaequorin [81]. For calcium measurements, aequorin was reconstituted by the overnight incubation of the cell suspension in Gamborg medium containing 2.5 mM native coelenterazine. Cell culture aliquots (500 mL) were transferred carefully to a luminometer glass tube, and the luminescence counts were recorded continuously at 0.2 s intervals with a FB12-Berthold luminometer. Treatments with air or ozone were performed directly in the luminometer. At the end of each experiment, the residual aequorin was discharged by addition of 500 mL of a 1 M CaCl 2 solution dissolved in 100% methanol. The resulting luminescence was used to estimate the total amount of aequorin for each condition. Calibration of calcium levels was performed using the equation: pCa = 0.332588(-logk)+5.5593, where k is a rate constant equal to luminescence counts per second divided by total remaining counts. Data are expressed as mM and are means 6 SE.

Monitoring of ROS Production
The production of singlet oxygen ( 1 O 2 ) and O 2˙2 was monitored by the chemiluminescence of the Cypridina luciferin analog (CLA) as previously described [13]. CLA-chemiluminescence specifically indicates the presence of O 2˙2 , and of 1 O 2 to a lesser extent, but not other ROS [82]. Chemiluminescence from CLA was monitored using a FB12-Berthold luminometer (with a signal integrating time of 0.2 s). For the statistical analysis of the data, the luminescence ratio (L/Lbasal) was calculated by dividing the luminescence intensities of CLA-luminescence (L) with the luminescence intensity before air or ozone treatment (Lbasal).

RT-PCR analysis of gene expression
Four-day-old cells were treated with O 3 , harvested and frozen in liquid nitrogen. Total RNA was prepared using the Genelute TM Mammalian Total RNA Kit (Sigma). RNA was treated by the Deoxyribonuclease I Kit (Sigma). Total RNA was quantified by spectrophotometry and their integrity checked on denaturing agarose gels. Total RNA (2 mg) was converted into first-strand cDNA with the Superscript TM II Rnase H 2 Reverse Transcriptase Kit (Invitrogen, Carlsbad, CA, USA) and oligo(dT) primers. One ml of cDNA was amplified in a 25 ml PCR mixture. VPEa, VPEb, and VPEc primers were designed (VPEa forw: GCGAAGAAC-GAGGAGAATCCAA, VPEa rev: TGCTCGTGCAAAGTCT-CTGTTT; VPEb forw: ACAATGACCACCGTCGTTTCCT, VPEb rev: TAGGGCGGAGACGAAGATCAAG; VPEc forw: GACCATGGTGGTCCTGGAGTTC, VPEc rev: ATTCG-CATCCGCAAAGGTAAAA). Control PCR was performed using the housekeeping gene EF1a4 [83]. Thermal cycling conditions were as follows: an initial denaturation step at 94uC 2 min, followed by 34 cycles (or by 26 cycles for EF1a4), of 94uC 30 s, 55uC 30 s, 72uC 1 min 30 s, and ending with a single step at 72uC 10 min. PCR products were seperated by gel electrophoresis and visualized by ethidium bromide fluorescence. Representative results from three independent experiments are shown.