Nitric oxide (NO) and hydrogen peroxide (H2O2) are key signalling molecules produced in response to various stimuli and involved in a diverse range of plant signal transduction processes. Nitric oxide and H2O2 have been identified as essential components of the complex signalling network inducing stomatal closure in response to the phytohormone abscisic acid (ABA). A close inter-relationship exists between ABA and the spatial and temporal production and action of both NO and H2O2 in guard cells. This study shows that, in Arabidopsis thaliana guard cells, ABA-mediated NO generation is in fact dependent on ABA-induced H2O2 production. Stomatal closure induced by H2O2 is inhibited by the removal of NO with NO scavenger, and both ABA and H2O2 stimulate guard cell NO synthesis. Conversely, NO-induced stomatal closure does not require H2O2 synthesis nor does NO treatment induce H2O2 production in guard cells. Tungstate inhibition of the NO-generating enzyme nitrate reductase (NR) attenuates NO production in response to nitrite in vitro and in response to H2O2 and ABA in vivo. Genetic data demonstrate that NR is the major source of NO in guard cells in response to ABA-mediated H2O2 synthesis. In the NR double mutant nia1, nia2 both ABA and H2O2 fail to induce NO production or stomatal closure, but in the nitric oxide synthase deficient Atnos1 mutant, responses to H2O2 are not impaired. Importantly, we show that in the NADPH oxidase deficient double mutant atrbohD/F, NO synthesis and stomatal closure to ABA are severely reduced, indicating that endogenous H2O2 production induced by ABA is required for NO synthesis. In summary, our physiological and genetic data demonstrate a strong inter-relationship between ABA, endogenous H2O2 and NO-induced stomatal closure.
Removal of the major known sources of either NO or H2O2 severely attenuates ABA-induced stomatal closure. Current reports show that there are at least two major enzymatic sources of NO in guard cells involved in ABA signalling: nitric oxide synthase (NOS) and nitrate reductase (NR). Nitrate reductase reduction of nitrite to NO has been demonstrated in vitro (Yamasaki and Sakihama, 2000) and in vivo in maize, sunflower, spinach, alfalfa and Arabidopsis (Desikan et al., 2002; Dordas et al., 2003; Rockel et al., 2002). In wild-type Arabidopsis, nitrite induces guard cell NO synthesis and stomatal closure whilst ABA fails to produce NO or induce stomatal closure in epidermal peels of the nia1, nia2 NR-deficient mutant (Desikan et al., 2002). A constitutive NOS enzyme, AtNOS1, is also important in ABA-induced NO-mediated stomatal closure (Guo et al., 2003). Atnos1 T-DNA insertional knock-out mutants have reduced stomatal responses to ABA and produce significantly less NO in response to ABA compared with wild-type plants (Guo et al., 2003). Enzymatic synthesis of H2O2 in guard cells in response to ABA occurs via two NADPH oxidase isoforms; AtrbohD and AtrbohF. The atrbohD/F double mutant has impaired ABA-induced H2O2 generation and stomatal closure (Kwak et al., 2003).
In this study, nia1, nia2, Atnos1 and atrbohD/F mutants were utilized to investigate the relationship between NO, H2O2 and ABA in guard cell signal transduction. By combining pharmacological and biochemical analysis with this genetic approach, we provide evidence to show that ABA-induced NO production via NR is required for ABA-induced H2O2-mediated stomatal closure.
Nitric oxide is required for hydrogen peroxide-induced stomatal closure
In order to investigate the possibility of an interaction between H2O2 and NO in ABA-induced stomatal closure, the effects of the NO scavenger 2-phenyl-4,4,5,5-tetremethylimidazolinone-1-oxyl 3-oxide (PTIO) and the H2O2 scavenger catalase (CAT) on ABA, NO and H2O2-induced stomatal closure were assessed. Abscisic acid and sodium nitroprusside (SNP)-induced stomatal closure were greatly reduced in the presence of PTIO, as shown in previous reports (Desikan et al., 2002; Neill et al., 2002b). Abscisic acid induced closure was also inhibited by the H2O2-scavenging enzyme CAT. Here, PTIO inhibition of H2O2-induced stomatal closure was also observed, similar to inhibition of ABA and SNP-induced stomatal closure (p<0.001), indicating a requirement for NO in H2O2-induced stomatal closure. Alone, PTIO has no effect on the stomatal aperture (data not shown). Moreover, although pre-treatment of leaves with CAT inhibited H2O2-induced closure (P < 0.001), it did not affect SNP-induced closure (P > 0.05), indicating that NO does not require H2O2 generation to initiate stomatal closure (Figure 1a). Incubation of leaves with either boiled CAT or bovine serum albumin (BSA) had no effect on ABA- or H2O2-induced stomatal closure (data not shown).
Hydrogen peroxide stimulates nitric oxide production in guard cells
The role of NO in H2O2-induced stomatal responses was further examined by monitoring NO synthesis in response to applied H2O2. Epidermal fragments were loaded with the NO-specific fluorescent dye diaminofluorescein diacetate (DAF2-DA) and confocal laser scanning microscopy (CLSM) used to monitor changes in NO-induced fluorescence. A significant increase in NO-induced guard cell fluorescence was observed in H2O2-treated epidermal fragments compared with control tissue (P < 0.001), demonstrating H2O2-mediated NO production in guard cells (Figure 1b,c). Nitric oxide production occurs at H2O2 concentrations as low as 5 μm and DAF2 fluorescence appears to reach saturation at 200 μm H2O2 (data not shown). Nitric oxide production was observed as little as 60 sec after the addition of 100 μm H2O2 and maximum fluorescence intensity was recorded after 20–25 min (data not shown). Importantly, H2O2-induced NO synthesis was abolished by co-incubation with PTIO, correlating these data with those from stomatal bioassays (P < 0.001; Figure 1a,b). Nitric oxide induced fluorescence in ABA- and H2O2-treated epidermal fragments was not significantly different (P > 0.05).
Nitric oxide does not induce hydrogen peroxide synthesis in guard cells
The data in Figure 1(a) indicate that H2O2 is not required for NO-induced stomatal closure. However, He et al. (2005) reported that in Vicia faba guard cells, exogenous NO, applied in the form of the NO donor SNP, did induce H2O2 production. Consequently, the effects of SNP on H2O2 generation in Arabidopsis guard cells were analysed using the fluorescent probe 2′,7′-dichlorofluorescein diacetate (H2DCFDA; Figure 1d). SNP did induce guard cell H2DCF fluorescence as did ABA treatment, previously shown to mediate H2O2 production (Desikan et al., 2004b; Pei et al., 2000). However, the H2DCFDA dye is not specific for H2O2 and also reacts with NO (Hempel et al., 1999; J. T. Hancock, unpublished data). To determine whether the effects on H2DCF fluorescence observed following SNP treatment were merely attributable to the reaction of NO with H2DCF, a number of experimental treatments were performed (Figure 1d). As anticipated, CAT was seen to reduce ABA-induced H2DCF fluorescence but had no effect on NO-induced fluorescence, whilst treatment with the antioxidant ascorbate (ASC) had the same effects. Importantly, treatment with PTIO greatly reduced H2DCF fluorescence in response to SNP but had no effect on ABA-induced fluorescence. The reaction of PTIO is specific to NO and not H2O2; therefore, these data indicate that the fluorescence observed in SNP-treated guard cells is due to exogenously applied NO and not H2O2. Together these data demonstrate unequivocally that NO does not induce H2O2 production in Arabidopsis guard cells.
Hydrogen peroxide induced nitric oxide synthesis and stomatal closure are sensitive to nitrate reductase or nitric oxide synthase inhibition
The potential sources of NO during H2O2-induced NO generation were investigated using NOS and NR inhibitors. Pre-treatment of epidermal fragments with the NOS inhibitor NG-nitro-l-arginine methyl ester (l-NAME) at 200 μm, a concentration expected to be effective (Guo et al., 2003), had only a slight significant effect on H2O2-induced NO synthesis (P < 0.001), whereas the potential NR inhibitor tungstate (Deng et al., 1989; Notton and Hewitt, 1971) had a more pronounced significant effect on H2O2-induced NO generation (P < 0.001; Figure 2a). Similarly, stomatal bioassay data demonstrated that H2O2-induced closure was also more sensitive to the effects of NR inhibition by tungstate than by l-NAME (P < 0.001; Figure 2b). In order to establish both the capacity of Arabidopsis NR to generate NO via nitrite reduction and the susceptibility of this reaction to tungstate, Arabidopsis NR was assayed for NO-generating activity. Addition of nitrite to purified Arabidopsis NR resulted in the dose-dependent production of NO, thereby demonstrating that Arabidopsis NR can indeed generate NO via nitrite reduction (Figure 3). Inclusion of tungstate in the reaction mixture before the addition of nitrite prevented NO production, whilst addition of tungstate 300 sec after reaction initiation caused marked inhibition of NR generation of NO providing evidence that tungstate does inhibit Arabidopsis NR–NO enzyme activity and that inhibition of H2O2-induced NO synthesis and stomatal closure by tungstate are likely related to an inhibition of NR activity. Hydrogen peroxide added directly to the in vitro incubation medium and inclusion of l-NAME had no effect on NO generation by NR. No NR activity was observed when NAD(P)H was omitted from the incubation medium (data not shown).
Hydrogen peroxide generation of nitric oxide occurs via nitrate reductase and not AtNOS activity
Atnos1 and nia1, nia2 mutants are impaired in ABA-induced NO synthesis in guard cells (Desikan et al., 2002; Guo et al., 2003). Therefore, to characterize further the source of NO during H2O2-mediated stomatal closure, Atnos1 and nia1, nia2 mutants were examined for their ability to produce NO in response to H2O2. Responses of these mutants to the NR substrate nitrite were also assayed in order to establish whether the Atnos1 mutation has any effect on NR–NO mediated synthesis. Abscisic acid responses were recorded as a positive control. As reported previously (Guo et al., 2003), Atnos1 guard cells were significantly impaired in NO generation in response to ABA compared with wild-type plants (P < 0.001, Figure 4), but to a lesser degree than those of nia1, nia2 mutants (P < 0.001, Figure 4). Atnos1 guard cells produced a significant level of NO in response to exogenous H2O2 in the same manner as did those of wild type (P < 0.001, Figure 4), indicating that AtNOS1 is not involved in NO production in response to exogenous H2O2. However, nia1, nia2 guard cells did not generate NO in response to H2O2 and were not significantly different to untreated controls (P > 0.05), demonstrating that NR activity is essential for NO synthesis in response to exogenous H2O2. Nitrite-mediated NO synthesis was not significantly impaired in Atnos1 mutants (P > 0.05) as it was in nia1, nia2 plants (P < 0.001), indicating that NR–NO activity is not affected by this mutation in the AtNOS1 enzyme (Figure 4).
This difference in the H2O2-stimulated NO-generating capacities of Atnos1 and nia1, nia2 guard cells prompted further evaluation of stomatal functioning in these mutants. Stomatal bioassay experiments revealed that neither mutant was impaired in stomatal responses to SNP compared with wild-type plants, indicating normal guard cell function in response to NO (Figure 5). Atnos1 stomata did close in response to H2O2 and this response did not differ significantly to wild-type responses (P > 0.05). The guard cells of the nia1, nia2 mutants were less sensitive to H2O2, and stomatal apertures in H2O2-treated leaves did not differ significantly from those in controls (P > 0.05, Figure 5). All these data indicate that both AtNOS1 and NR mediate ABA-induced NO production and that NR is responsible for the majority of NO production in response to H2O2. Therefore, both AtNOS1 and at least one of the two NR genes (Wilkinson and Crawford, 1993) should be expressed in guard cells. To confirm this, reverse transcription (RT) PCR on RNA prepared from purified guard cell enriched extracts from wild-type plants was used to determine such expression. AtNOS1 and both NR1 and NR2 are expressed in Arabidopsis guard cells (data not shown), with the expression of NR2 being at a higher level than that of NR1, correlating with the expression pattern described previously (Wilkinson and Crawford, 1993).
Abscisic acid-induced hydrogen peroxide production mediates nitric oxide synthesis and stomatal closure
Two NADPH oxidase isoforms, AtrbohD and AtrbohF, both expressed in guard cells, are responsible for H2O2 generation and stomatal closure in response to ABA in guard cells (Kwak et al., 2003). Here, the double mutant atrbohD/F was used to investigate the physiological significance of H2O2-mediated NO production in ABA guard cell signalling. Abscisic acid failed to induce significant NO production in the guard cells of the atrbohD/F double mutant (P > 0.05), although mutant guard cells responded significantly to applied H2O2 (P < 0.001) to approximately the same degree as wild type (Figure 6a,b). These data demonstrate that NO production occurs downstream of ABA induction of H2O2 synthesis. The response to H2O2 indicates that the lack of NO generation in response to ABA is not a result of a general impairment in NO synthesis in these mutants (Figure 6a). Stomatal closure responses to ABA are known to be reduced in atrbohD/F double mutants but responses to H2O2 are not affected (Kwak et al., 2003; Figure 6b). The data presented in Figure 6(b) shows that stomatal closure responses to NO are normal in atrbohD/F plants as they do not differ significantly to wild-type plants (P > 0.05), providing evidence that the atrbohD/F plants do not suffer impaired NO signalling. These data place H2O2 synthesis linearly before NO production in ABA signalling in Arabidopsis guard cells.
It is becoming increasingly evident that signalling mechanisms in plants often do not operate as isolated, linear pathways but that extensive cross-talk occurs between signal transduction pathways (Knight and Knight, 2001). This complexity is enhanced by the range of responses that are mediated by different stimuli but involve the production of the same second messenger molecules and, paradoxically, the fact that in some cases the same stimulus activates a multitude of signalling pathways that result in a single response. The intricate signalling web initiated by ABA in stomatal guard cells is an excellent example of how a vital response to an external stimulus is controlled by a complex array of signalling mechanisms (Assmann and Wang, 2001; Hetherington, 2001; Hetherington and Woodward, 2003). The complexity of signalling responses to certain stresses may have developed in such a way as to provide a ‘fail-safe’ mechanism: multiple, intertwined pathways are in place to ensure a contingency so that, in the event of dysfunction, a plant cell may still operate appropriately. Understanding how these signalling pathways interact is key to our understanding of fundamental plant physiological processes.
A requirement for both NO and H2O2 in ABA-mediated stomatal closure has previously been shown (Garcia-Mata and Lamattina, 2001; Neill et al., 2002b; Pei et al., 2000). However, NO and H2O2 synthesis were thought to occur in parallel, until Lum et al. (2002) demonstrated that exogenous H2O2 induces the rapid production of NO in the guard cells of P. aureus. A very recent report has provided pharmacological evidence indicating that endogenous H2O2-mediated NO generation has an important role in UV-B-induced stomatal closure in V. faba (He et al., 2005). In the present study, NO synthesis in response to H2O2 in Arabidopsis guard cells is demonstrated. Importantly, this process was correlated to the biological response of ABA-induced stomatal closure. Using a pharmacological approach, it has been shown that NO synthesis is required for H2O2-induced stomatal closure. In contrast to the data reported by He et al. (2005), the data in the present study demonstrate that SNP-induced stomatal closure is not affected by the H2O2 scavenging enzyme CAT, thereby indicating that NO does not induce H2O2 synthesis required for stomatal closure. Moreover, the data show unequivocally that NO does not induce H2O2 synthesis in Arabidopsis guard cells. Rather, the H2DCF fluorescence monitored in the presence of SNP was actually attributable to the reaction of the dye with NO, not with H2O2, and it is possible that He et al. (2005) observed a similar response. However, He et al. (2005) did report that ASC and CAT partially reversed SNP-induced H2DCF fluorescence and stomatal closure, so it is possible that in fact guard cell responses differ between Arabidopsis and V. faba.
Further evaluation of this response and the potential source of NO induced by exogenous H2O2 revealed that NO production and stomatal closure could be attributed to NR activity. The effects of the NR inhibitor tungstate on H2O2-induced NO synthesis and stomatal closure were significant and implicate NR activity in these responses. In vitro analysis of NR–NO activity provides unequivocal evidence that Arabidopsis NR has the capacity to generate NO from nitrite. Tungstate strongly inhibits this activity at the concentration used in bioassay and confocal microscopy experiments, indicating that the effects of tungstate observed in vivo are indeed related to attenuated NR activity. Hydrogen peroxide added directly to the in vitro incubation medium had no effect on NO generation by NR, indicating that H2O2 does not interact directly with NR. Thus, stimulation by H2O2 of NR in vivo must involve some intermediate signal transduction event. l-NAME had a lesser, yet significant, effect on the H2O2-induced response, implying that AtNOS1 or an NOS-like enzyme could also be involved (Guo et al., 2003).
Previous studies have reported that H2O2-induced NO generation in guard cells of P. aureus and V. faba is related to NOS activity, as l-NAME inhibits H2O2-mediated NO production in both species (He et al., 2005; Lum et al., 2002). In order to clarify the pharmacological observations regarding the source of NO during H2O2-induced responses in Arabidopsis, a genetic approach was adopted. Guard cells of the Atnos1 T-DNA insertion mutant produced NO in response to exogenous H2O2 in the same manner as did those of wild-type plants and the stomata closed in response to H2O2, indicating that AtNOS1 is not involved in H2O2-induced NO production. Furthermore, nitrite-mediated NO synthesis is not impaired in Atnos1 mutants, demonstrating that NR–NO activity is not affected in some way by the Atnos1 mutation. In contrast, guard cells of the nia1, nia2 double mutant do not generate NO in response to H2O2 and stomatal apertures in H2O2-treated nia1, nia2 leaves do not differ from those in controls. These data imply strongly that NR is responsible for the production of NO in guard cells in response to exogenous H2O2. Importantly, it has been demonstrated that nia1, nia2 plants do generate H2O2 in response to ABA (Desikan et al., 2002), providing further evidence that NR acts downstream of ABA-mediated H2O2 generation to produce NO and induce stomatal closure. Two NR genes, NR1 and NR2, are present in the Arabidopsis genome, with NR2 being by far the more abundant protein (Wilkinson and Crawford, 1991), so it is important to determine whether it is NR1 or NR2 that mediates ABA- and H2O2-activated NO generation in guard cells. Preliminary data indicate that the less abundant NR1 protein and not NR2 serves this function, as the mutant nia1 disocciation (nia1::Dsnia1::DS; Parinov et al., 1999; Wang et al., 2004) appears deficient in ABA and H2O2-induced stomatal closure and NO generation, whereas the NR2 deletion mutant, (Wilkinson and Crawford, 1991) responds normally to both ABA and H2O2 (J. Bright, unpublished data). As both NR1 and NR2 are expressed in guard cells, it is likely that subcellular localization, protein–protein interactions and differential activities of these proteins are important.
To determine the physiological significance of NO production induced by endogenous H2O2 in Arabidopsis guard cells, we investigated the effects of ABA on NO production and stomatal closure in atrbohD/F mutants. This key experiment provides evidence for a previously uncharacterized role for H2O2 in ABA-induced NO-mediated guard cell responses. The atrbohD/F mutants failed to generate NO in response to ABA, indicating the requirement of ABA-activated Rboh-mediated H2O2 generation for NO production. However, NO production in response to H2O2 still occurred in these mutants, indicating that atrbohD/F plants do not suffer a general impairment in NO synthesis.
The inter-relationship of NO and H2O2 during ABA-induced stomatal responses provides new insight into the complex signalling mechanisms controlling stomatal function (Hetherington and Woodward, 2003). Cross-talk between signalling pathways is a common phenomenon and, as a result of their coordinated production and apparent synergy in mediating a range of cellular processes, it has been speculated that H2O2 and NO signalling may converge. Here, it is revealed that in guard cells NO production is in fact dependent on H2O2 formation and that this mechanism governs stomatal responses to ABA. However, it is important to note that although linearity in ABA, H2O2 and NO signalling has been observed, it is likely that ABA–H2O2 guard cell signalling also remains divergent from NO signalling. Pharmacological evidence indicating that ABA-mediated NO production causes stomatal closure, at least partly, via the synthesis and action of cyclic guanosine monophosphate (cGMP; S. J. Neill, unpublished data) but that H2O2-induced stomatal closure is independent of cGMP has also been obtained, suggesting further complexities to ABA guard cell signalling, yet to be fully determined.
Wild-type and various mutants of the Columbia ecotype of Arabidopsis thaliana were sown in Levington's F2 compost (Avoncrop, Bristol, UK) and grown under a 16-h photo-period (60–100 μE m−2 s−1) in plant growth chambers (Sanyo Gallenkamp, Loughborough, UK). Fully expanded wild-type leaves were harvested for immediate use at 4–5 weeks. Atnos1 and nia1, nia2 leaves were harvested after they had reached an equivalent developmental stage (usually at 5–6 weeks). The nia1, nia2 double mutant seeds were obtained from the Nottingham Arabidopsis Stock Centre (NASC, Nottingham, UK). Atnos1, NR1 Ds and NR2 deletion mutant seeds were obtained from N. Crawford (University of California, San Diego, CA, USA). The atrbohD/F double mutant seeds were obtained from J. Jones (The Sainsbury Laboratory, Norwich, UK).
Stomatal bioassays were carried out essentially as described by Desikan et al. (2002). Whole leaves were incubated in MES-KCl buffer [10 mm 2-morpholino ethane sulfonic acid (MES), 5 mm KCl, 50 μm CaCl2, pH 6.15] for 2.5 h under light conditions (in 60–100 μE m−2 s−1) in a growth chamber before the addition of various compounds. Where indicated in figure legends, leaves were incubated in the presence of various inhibitory compounds for 30 min prior to treatment with ABA, SNP, H2O2 or nitrite. Compounds were dissolved in either water or ethanol and controls were appropriately treated with an equivalent amount of solvent. All chemicals were obtained from Sigma Aldrich, Poole, UK unless otherwise stated in the text. Stomatal apertures and areas were observed in epidermal fragments after a further 2 h incubation. Apertures were measured using a light microscope and imaging camera with Leica QWin image processing and analysis software (Leica Microsystems and Imaging Solutions, Cambridge, UK). Twenty to thirty apertures per treatment were recorded in 5–6 independent replicates.
Confocal laser scanning microscopy
Nitric oxide was visualized using the specific NO dye DAF2-DA (Calbiochem, Nottingham, UK), using the method essentially described by Desikan et al. (2002). Epidermal fragments in MES-KCl buffer (10 mm MES, 5 mm KCl, 50 μm CaCl2, pH 6.15) were loaded with 15 μm DAF2-DA for 15 min before washing with MES-KCl buffer for 20 min. Fragments were then incubated for a further 25 min in the presence of various compounds (as indicated in figure legends) before images were visualized. H2O2 production was monitored using the H2O2-sensitive fluorescent probe H2DCFDA (Molecular Probes, Leiden, the Netherlands) using the method essentially described by Desikan et al. (2004b). Epidermal fragments in MES-KCl buffer were loaded with 50 μm H2DCFDA for 15 min before washing in MES-KCl buffer for 20 min. Fragments were then incubated for a further 25 min in the presence of various compounds (as indicated in figure legend) before images were visualized. All images were visualized using CLSM (excitation 488 nm, emission 515 nm; Nikon PCM2000, Kingston upon Thames, UK). Images acquired were analysed using scion image software (Scion, Frederick, MD, USA). Data are presented as mean pixel intensities. Fifty to one hundred guard cells are observed per treatment for five independent replicates.
In vitro nitrate reductase assay
Purified Arabidopsis NR was obtained from The Nitrate Elimination Company, MI, USA. The fluorimetric detection of nitrite-dependent NR formation of NO was carried out using a fluorescence spectrophotometer and the NO indicator DAF2 (excitation and emission wavelengths 488 and 515 nm, respectively). Arabidopsis NR (30 mU) was added to MES-KCl buffer (10 mm MES, 5 mm KCl, 50 μm CaCl2, pH 6.15) containing 5 μm DAF2 and basal fluorescence was recorded for 200 sec. The reaction was initiated by the addition of 0.5, 1 or 2 mm nitrite in the absence or in the presence of 100 μm tungstate, added before or 300 sec after nitrite addition. Reactions were allowed to proceed for 1000 sec.
Isolation of guard cell RNA and reverse transcription PCR
Preparation of guard cell-enriched epidermal fragments (>95%) and RNA isolation were carried out as described previously (Desikan et al., 2005). PCR was performed on DNAsed RNA with PCR primers designed against sequences unique to NR1, (5′ TTGGTCAGACTGGTCAATCG 3′) and (5′ TCCAGATGAGTTTATCAGGC 3′), NR2, (5′ ATGGTGAAGAAGTCAAAGGG 3′) and (5′ ACGAACAGCAATCTCTTTGG 3′) and AtNOS1, (5′CGCTACGAACACTCTCAACG3′) and (5′ AGAATGTAGACATCTCGTCC 3′). Genomic DNA was used as a control to ensure PCR products originated from cDNA and PCR products were subsequently sequenced.
Data analysis and statistics
The P-values, estimated means and associated 95% confidence intervals (CIs) are derived from the minimal adequate models obtained using stepwise selection from the maximal model of the fractional factorial designs. All the final models have normal errors except Figure 6(b), which was more suited to a Gamma error structure. Model adequacy was assessed by a variety of residual diagnostic plots together with a Cook's distance plot to examine leverage. All modelling was carried out in the R statistics package (R Development Core Team, 2004).
We thank N. Crawford and colleagues (University of California) for the Atnos1, NR2 deletion and NR1 Ds mutant seeds, J. Jones (The Sainsbury Laboratory) for the atrbohD/F double mutant seeds and J. Gray and colleagues (University of Sheffield, Sheffield, UK) for advice on RNA isolation from Arabidopsis guard cells. We also thank J Tavaré, M Jepson and A Leard (Bristol University, Bristol, UK) for the use of the cell imaging facility. RD was supported by the BBSRC.