Oxidative burst and cognate redox signalling reported by luciferase imaging: identification of a signal network that functions independently of ethylene, SA and Me-JA but is dependent on MAPKK activity
John J. Grant,
Institute of Cell and Molecular Biology, University of Edinburgh, Daniel Rutherford Building, King's Buildings, Edinburgh EH9 3JH, UK
Recognition of avirulent microbial pathogens activates an oxidative burst leading to the accumulation of reactive oxygen intermediates (ROIs), which are thought to integrate a diverse set of defence mechanisms resulting in the establishment of plant disease resistance. A novel transgenic Arabidopsis line containing a gst1::luc transgene was developed and employed to report the temporal and spatial dynamics of ROI accumulation and cognate redox signalling in response to attempted infection by avirulent strains of Pseudomonas syringae pv. tomato (Pst). Strong engagement of the oxidative burst was dependent on the presence of functional Pst hrpS and hrpA gene products. Experiments employing pharmacological agents suggested that at least two distinct sources, including an NADPH oxidase and a peroxidase-type enzyme, contributed to the generation of redox cues. The analysis of gst1 and pal1 gene expression in nahG, coi1 and etr1 plants suggested that engagement of the oxidative burst and cognate redox signalling functioned independently of salicylic acid, methyl jasmonate and ethylene. In contrast, studies using a panel of protein kinase and phosphatase inhibitors and in-gel kinase assays in these mutant backgrounds suggested that a 48 kDa mitogen-activated protein kinase (MAPK) activity was required for the activation of gst1 and pal1 in response to redox cues. Thus the engagement of a bifurcating redox signalling pathway possessing a MAPK module may contribute both to the establishment of plant disease resistance, and to the development of cellular protectant mechanisms.
Following successful transgression of pre-formed physical and chemical barriers, attempted infection by an avirulent microbial pathogen elicits the deployment of a plethora of inducible defence mechanisms in both local and systemic tissues (Hammond-Kosack and Jones, 1996; Ryals et al., 1996). Pathogen recognition is thought to result from the direct or indirect interaction between a product of a microbial avirulence (avr) gene and the corresponding plant disease resistance (R) gene (Dangl, 1995; Staskawicz et al., 1995). Intriguingly, some bacterial avirulence proteins are thought to be recognized inside plant cells, their delivery mediated via a type III secretory system analogous to that employed by Gram-negative bacterial pathogens of animals (Fenselau et al., 1992). A key feature underlying successful pathogen recognition is the engagement of the so-called oxidative burst, a rapid biphasic production of reactive oxygen intermediates (ROIs), primarily superoxide (O2–) and hydrogen peroxide (H2O2), at the site of attempted infection (Apostol et al., 1989; Grant and Loake, 2000). Multiple cellular functions have been ascribed to ROIs, including the oxidative cross-linking of cell-wall structural proteins (Bradley et al., 1992) and direct antimicrobial activity (Peng and Kuc, 1992). Moreover, the accumulation of ROIs may also initiate the development of the hypersensitive response (HR) (Jabs et al., 1996; Levine et al., 1994) and engage the deployment of cellular protectant functions in distal cells to limit cell death expansion (Kliebenstein et al., 1999; Tenhaken et al., 1995). Nitric oxide (NO), a key signalling molecule in mammalian cells, has recently been proposed to function in concert with ROIs both to potentiate cell death and to induce the expression of specific defence genes (Delledonne et al., 1998; Durner et al., 1998).
While some of the cellular consequences of ROI accumulation have been uncovered, the identity of the molecular machinery underlying the oxidative burst remains to be rigorously established. Recent evidence has implicated a number of possible mechanisms, including: a plasma membrane-located NADPH-dependant oxidase (Groom et al., 1996; Keller et al., 1998); a cell-wall peroxidase (Bolwell and Wojtaszek, 1997; McLusky et al., 1999); an extracellular germin-like oxalate oxidase (Zhang et al., 1995); and apoplastic amine, diamine and polyamine oxidase-type enzymes (Allan and Fluhr, 1997; Tipping and McPherson, 1995). Of these proposed mechanisms the NADPH-dependant oxidase system, similar to that present in mammalian neutrophils, has received the most attention. Homologues of the mammalian gp91phox subunit of the NADPH oxidase complex have recently been identified in both rice and Arabidopsis (Groom et al., 1996; Keller et al., 1998). In other plant species, however, there is accumulating evidence for the involvement of apoplastic peroxidases in the oxidative burst: such enzymes have been shown to produce H2O2 at an alkaline pH, as found in the apoplast during an incompatible interaction (Bolwell and Wojtaszek, 1997). Moreover, they may be directly secreted to the sites of attempted microbial infection (McLusky et al., 1999).
In addition to their proposed role in local, R-gene mediated resistance, the oxidative burst and cognate redox signalling may also play a pivotal function in the establishment of acquired resistance in local and systemic tissues (LAR and SAR, respectively). In this context, the accumulation of H2O2 in catalase-deficient transgenic tobacco plants was shown to mediate the local and systemic induction of acidic pathogenesis-related (PR) proteins (Chamnongpol et al., 1998). Furthermore, a redox signalling network has been proposed to reiterate the oxidative burst at systemic microsites, which are thought to be a prerequisite for the establishment of SAR (Alvarez et al., 1998). Thus the oxidative burst and its cognate redox signalling network may function to integrate a number of diverse defence mechanisms, underpinning the establishment of both local R-gene mediated resistance and acquired disease immunity.
While the signalling networks orchestrating the responses to pathogen-derived products are now beginning to be uncovered (Hammond-Kosack and Jones, 1996), the defence pathways engaged following the release of secondary signals such as ROIs remain relatively undefined. Here we describe the construction of a novel transgenic Arabidopsis line containing a gst1:: luciferase transgene that faithfully reports the oxidative burst and cognate redox signalling. Employing this technology, we show there may be at least two distinct sources of ROIs generated in Arabidopsis and that their engagement is dependent on a functional bacterial type III secretory system. Furthermore, experiments employing nahG, coi1 and etr1 plants suggested that the induction of gst1 and pal1 gene expression in response to redox cues functioned independently of salicylic acid (SA), methyl jasmonate (Me-JA) and ethylene. In contrast, studies employing a panel of protein kinase and phosphatase inhibitors and in-gel kinase assays in these mutant backgrounds demonstrate that a 48 kDa mitogen-activated kinase (MAPK) activity may be required for the successful transmission of redox cues. This signalling network may contribute to both the establishment of plant disease resistance, and the development of cellular protectant mechanisms.
Identification of an ROI marker gene
After monitoring the expression of a number of Arabidopsis antioxidant defence genes, a gene encoding a glutathione S-transferase, designated gst1 (Greenberg et al., 1994), was selected as a robust molecular marker for ROI accumulation. Genes encoding other Arabidopsis gst genes have also recently been employed as molecular markers for the oxidative burst (Alvarez et al., 1998; Jabs et al., 1996). Northern blot analysis using a gene-specific probe demonstrated that gst1 transcripts were induced at 1 h, and showed maximum 18-fold accumulation at 2–3 h post-inoculation with Pseudomonas syringae pv. tomato (Pst) strain DC3000 expressing the avrB avirulence gene (Figure 1a,b). Leaf infiltration with glucose/glucose oxidase (G/GO), which produces a sustained, sublethal accumulation of H2O2– closely mimicking the kinetics of ROI production mediated via the oxidative burst (Levine et al., 1994) – strongly induced the accumulation of gst1 transcripts to similar levels after 2 h (Figure 1a). Staining with 3,3’-diaminobenzidine (DAB) was used to confirm that the temporal accumulation of endogenous H2O2 in response to Pst DC3000(avrB) was congruent with gst1 expression (Figure 3b).
A gst1 gene-specific probe was employed to isolate the corresponding genomic clone from a cosmid library (Schulz et al., 1994). A 909 bp promoter sequence upstream of the translation start site was uncovered: the translated region was 1092 bp long and contained two introns of 92 and 110 bp. During the course of this work a gst from a different Arabidopsis accession with 97% sequence identity to gst1 was reported (Yang et al., 1998), suggesting that these sequences may encode the same gene. The gst1 gene is composed of two introns and three exons; in accordance with evolutionary classification by Droog, this gene encodes a type I plant GST (Droog et al., 1993). This class of GST is thought to function as a key cellular protectant, and other members of this class are inducible by pathogen attack, wounding and lipid peroxidation (Marrs, 1996).
A gst1::luc transgene robustly reports the oxidative burst and cognate redox signalling
The gst1 promoter was transcriptionally fused to the firefly luciferase (luc) reporter gene (Figure 2a), and the resulting construct was transformed into Arabidopsis accession Col-0. The generated transgenic lines characteristically showed induction of LUC activity to similar levels following G/GO infiltration or inoculation with avirulent strains of Pst. The kinetics of gst1::luc transgene expression was determined using an ultra-low-light imaging camera system, following leaf inoculation with Pst DC3000 expressing either the avrB or avrRpt2 avirulence genes (Figure 2b). The temporal profile of LUC activity established following Pst DC3000(avrB) inoculation was congruent with the expression of the endogenous gst1 gene: LUC activity was first detected approximately 45 min post-inoculation, with a maximal induction of approximately 20-fold at 2–3 h, followed by a steady decay of LUC activity (Figure 2b). Similar results were obtained following inoculation of Pst DC3000(avrRpt2), although the maximum LUC activity measured was approximately 25% less than following Pst DC3000(avrB) inoculation. Infiltration with MgCl2 or virulent Pst DC3000 resulted in an approximately fivefold increase in LUC activity. In order to confirm that ROI accumulation is responsible for activation of the gst1::luc transgene, two enzymatic scavengers, catalase (CAT) and superoxide dismutase (SOD), were independently co-inoculated with Pst DC3000(avrB). While LUC activity was substantially reduced by 72% following co-inoculation of CAT (Figure 2f), SOD did not significantly diminish LUC activity (Figure 2e). This observation suggests that accumulation of H2O2 rather than O2– was the major cue responsible for gst1::luc induction.
The hypersensitive response and pathogenicity (hrp) genes of Gram-negative plant pathogenic bacteria are thought to encode a type III protein-secretion system that may deliver avirulence proteins inside host-plant cells (Fenselau et al., 1992). This prompted us to examine if engagement of the oxidative burst and cognate redox signalling was hrp gene-dependent. Strains of Pst DC3000 possessing a null mutation within either the hrpA or hrpS gene (Roine et al., 1997) were transformed with plasmid pVB01 containing the avrB gene (Innes et al., 1993), and assessed for their ability to engage the oxidative burst. Inoculation of Pst DC3000(avrB) hrpA or hrpS mutants failed to induce LUC activity above the levels recorded for PstDC3000 and MgCl2 control inoculations (Figure 3a). Moreover, no significant accumulation of H2O2 was detected by DAB staining following inoculation with Pst DC3000(avrB) hrpS (Figure 3b) nor hrpA strains (data not shown). Therefore functional hrpA and hrpS gene products are required for successful engagement of the oxidative burst during avrB/RPM1 mediated disease resistance.
Temporal and spatial gst1 expression programme established during HR formation
Using time-lapse image capture for 24 h we determined the temporal and spatial profile of LUC activity established during the dynamic events associated with the HR. Key images are shown in Figure 4. The presence of LUC activity within directly challenged cells at 8 h post-inoculation with Pst DC3000(avrB) suggested that gst1 transcript accumulation preceded their subsequent programmed execution during the HR (Figure 4b). This observation differs from that predicted from previous models in which gst gene expression was proposed to occur only in the distal unchallenged cells (Tenhaken et al., 1995). The onset of hypersensitive cell death usually occurred between 7 and 12 h post-inoculation with Pst DC3000(avrB), and resulted in a rapid reduction of LUC activity in directly challenged cells during the phase of lesion spread. At 16 h post-inoculation, viable unchallenged cells delimiting the developed HR lesion continued to express LUC activity, probably reflecting the impact of other defence signals temporally resolved from ROIs (Figure 4c). Observation of this cellular margin, using a light microscope attachment, revealed that the width of LUC activity was approximately 20 cells (data not shown).
At least two distinct sources of redox cues activate gst1 gene expression
A number of possible sources may contribute to ROI accumulation during the pathogen-activated oxidative burst, including an NADPH-dependent oxidase (Groom et al., 1996; Keller et al., 1998); cell wall-bound peroxidases (Bolwell and Wojtaszek, 1997); and apoplastic amine oxidase-type enzymes (Allan and Fluhr, 1997). The contribution of these enzymes to ROI accumulation in Arabidopsis has not previously been directly compared. Co-inoculation of diphenylene iodonium (DPI), an inhibitor of the NADPH-dependent oxidase complex, with Pst DC3000(avrB) reduced LUC activity by 36% compared to the value obtained with inoculation of Pst DC3000(avrB) alone (Figure 5). In a similar experiment, co-inoculation with sodium azide (NaN3), a peroxidase inhibitor, reduced LUC activity by 28% (Figure 5). The corresponding dose–response curves had previously revealed that these agents were employed at saturating concentrations. No reduction in LUC activity was observed when these phamacological agents were individually co-inoculated with H2O2, suggesting that they inhibit H2O2 production, not perception (data not shown). Neither DPI nor NaN3 blocked the induction of LUC activity in an Arabidopsis line containing a basic PR gene promoter fusion, suggesting that these agents specifically inhibited gst1::luc gene expression and, moreover, did not affect either Pst DC3000 or LUC activity in planta (data not shown). Amines can induce ROI production by acting as substrates for amine oxidases (Allan and Fluhr, 1997). We therefore tested the ability of putrescine and arginine to activate gst1::luc gene expression. Infiltration of these amines failed to induce LUC activity above the levels for control inoculations (Figure 5).
The emerging evidence suggests that NO may function as a key signal in disease resistance (Delledonne et al., 1998; Durner et al., 1998). This prompted us to examine if NO could impact gst1::luc gene expression. Infiltration of the NO donor sodium nitroprusside (SNP) failed to induce significant LUC activity, and co-inoculation of the NO synthase inhibitor Nω-nitro-l-arginine (L-NNA) with Pst DC3000(avrB) did not significantly blunt the activation of LUC activity (Figure 5). Moreover, co-inoculation with G/GO and SNP did not significantly potentiate the induction of LUC activity compared to G/GO alone. Similar results were obtained with Northern blot analysis (data not shown). In total, these observations suggest that NO did not directly or indirectly impact the redox signalling network that mediates gst1 gene expression. Our results suggest that at least two distinct enzymatic sources, including an NADPH oxidase and a peroxidase-type enzyme, generated the redox cues that engaged gst1 gene expression.
Engagement of the oxidative burst and cognate redox signalling is independent of ethylene, SA or Me-JA
Ethylene, Me-JA and SA are thought to play key roles in plant defence signalling (Gaffney et al., 1993; Xu et al., 1994). We therefore investigated if any of these signal molecules mediated either the activation of the oxidative burst, or the subsequent transduction of redox signals. Arabidopsis Col-0 plants containing the gst1::luc transgene were crossed with plants containing the ethylene insensitive allele etr1 (Bleecker et al., 1988), and with plants containing the nahG transgene which converts SA to catechol (an inactive metabolite with respect to defence signalling) (Gaffney et al., 1993). These plants were infiltrated with Pst(avrB) and LUC activity was recorded over time using an ultra-low-light imaging camera system. No consistent differences were observed in the profile of gst1::luc gene expression in either the etr1 or nahG genetic backgrounds compared to the wild-type Col-0 accession (Figure 6a,b). These observations suggest that neither ethylene nor SA affected the engagement of the oxidative burst or the subsequent transduction of redox signals. The accumulation of gst1 transcripts in nahG plants determined by Northern analysis closely paralleled that observed for wild-type Col-0 plants (Figure 6c), with maximum gst1 transcript accumulation at 2–3 h post-inoculation with Pst(avrB). Furthermore, infiltration of G/GO into nahG plants induced maximum gst1 transcript accumulation at approximately 2 h, similar to that obtained for wild-type Col-0 plants (Figure 6c). These experiments therefore suggest that local SA accretion was not necessary for the complete engagement of either the oxidative burst or cognate redox signalling. Northern analysis of gst1 gene expression in the Arabidopsis mutant coi1 (Feys et al., 1994), which is insensitive to Me-JA, was also indistinguishable from that observed with wild-type Col-0 plants (Figure 6d). Hence the lipid signalling molecule Me-JA also does not appear to play a role in the activation of the oxidative burst or in the subsequent transduction of redox cues.
ROI-induced gene expression is dependent on MAPKK activity
A key feature of redox signalling in animal cells is the pivotal role played by protein kinase cascades (Bauskin et al., 1991). We therefore tested a variety of protein kinase inhibitors for their ability to blunt H2O2-induced gst1 gene expression. Col-0 gst1::luc plants were inoculated with either G/GO or H2O2 in the presence or absence of a given kinase inhibitor, and the relative induction of gst1::luc gene expression was subsequently determined by measuring the level of LUC activity 2 h post-inoculation.
Neither staurosporine nor K252a (two broad-spectrum kinase inhibitors widely used in studies of plant biology) significantly inhibited the induction of LUC activity. These two pharmacological agents are well characterized inhibitors of protein kinase C, protein kinase A, calmodulin-dependent kinase and protein kinase G. In contrast, co-inoculation of G/GO with PD98059, an inhibitor of mitogen-activated protein kinase kinases (MAPKKs) (Cohen, 1997), strongly blunted the induction of LUC activity by 46% (Figure 7a). Therefore a MAP kinase cascade may be an integral component of the redox signalling network that engages gst1 gene expression in response to ROI accumulation.
To explore this possibility further, we undertook a complementary gain-of-function experiment, using the phosphatase inhibitor cantharidin, to examine if this pharmacological agent could induce gst1::luc gene expression in the absence of redox cues. To discriminate between activation of the oxidative burst (which is also regulated by a phosphorylation cascade; Levine et al., 1994) and engagement of redox signalling, cantharidin was co-inoculated with the H2O2-scavenging enzyme CAT, in order to blunt any H2O2 accumulation resulting from possible engagement of the oxidative burst machinery. Inoculation of 300 units ml−1 CAT had previously been demonstrated to strongly diminish the magnitude of gst1::luc gene induction in response to H2O2 generated by Pst DC3000(avrB) inoculation (Figure 2f). The results clearly demonstrate that cantharidin strongly induced LUC activity even in the presence of CAT (Figure 7a). The expression of the gst1 gene may therefore be regulated by a poise between phosphorylation/dephosphorylation. Similar results were obtained using Northern blot analysis (data not shown).
In addition, we also examined the response of a phenylalanine ammonia lyase (pal1) gene, which has been previously demonstrated to be strongly induced in response to ozone in an SA-independent manner (Sharma et al., 1996). Moreover, O2– has been shown to induce pal gene expression in parsley suspension cells (Jabs et al., 1997). Inoculation with Pst DC3000(avrB) resulted in strong accumulation of pal1 transcripts within 2 h (Figure 7b). Revealingly, co-inoculation of Pst DC3000(avrB) with CAT significantly blunted pal1 gene expression, while inoculation of G/GO strongly induced the expression of this gene. Thus ROIs may constitute a key signal for the activation of pal1 in response to Pst DC3000(avrB) inoculation.
In a similar fashion to gst1, engagement of pal1 gene expression in response to Pst DC3000(avrB) inoculation was not abrogated in either a nahG, coi1 or etr1 genetic background (Figure 7b). We therefore investigated if the engagement of pal1 gene expression by ROIs also depended on MAPKK activity. Co-inoculation of PD98059 with either Pst DC3000(avrB) or G/GO significantly reduced the induction of pal1 compared to that observed with either stimulus in the absence of PD98059 (Figure 7c,b). In the corresponding gain-of-function experiment, inoculation of cantharidin in the presence of CAT strongly induced pal1 gene expression in the absence of H2O2. Quantification of these results using phosphorImage analysis revealed that PD98059 decreased G/GO induction of pal1 by 99.5%. In contrast, cantharidin in the presence of CAT activated pal1 gene expression 17-fold (Figure 7d). The data therefore suggest that MAPKK activity, possibly as part of a MAPK module, was an integral component of the signal network that coupled the induction of both pal1 and gst1 gene expression to redox cues.
Sustained ROI accumulation induces two MAPK activities
To confirm the data derived from the deployment of pharmacological agents, we also determined MAPK activity directly using an in-gel kinase assay. Arabidopsis leaf tissue was infiltrated with G/GO, harvested over time, and protein extracts prepared. Protein kinase activity was determined via an in-gel kinase assay using myelin basic protein (MBP) as a substrate. Two MBP kinases of approximately 48 and 46 kDa were rapidly induced from a low basal level of activity (Figure 8a). Signals were not due to autophosphorylation, because no phosphorylation occurred after the same samples were separated on gels without substrate. Infiltration of G or GO alone did not significantly induce MBP kinase activity, and the equal loading of protein samples was demonstrated via a Coomassie stain of a parallel gel (data not shown). The molecular mass and substrate specificity of these kinases is indicative of a MAPK identity. Maximum relative induction of the 48 and 46 kDa MBP kinases, determined by phosphorImage analysis, was observed by 10 min post-infiltration with G/GO, attaining typical values of 14- and 10-fold, respectively (Figure 8b). After 15 min the 48 and 46 kDa MBP kinase activities had begun to decrease, and by 20 min their relative induction was five- and 4.8-fold, respectively. Maximum relative induction varied between fivefold in different experiments, although the kinetics remains unchanged.
To confirm the mode of action of the MAPKK inhibitor PD98059, which blunted the induction of gst1 and pal1 in response to redox cues, Arabidopsis leaves were infiltrated with G/GO in either the presence or absence of PD98059, and the corresponding protein extracts assayed for MBP kinase activity. PD98059 was observed to completely repress the activation of both 48 and 46 kDa MBP kinase activities (Figure 8c,e), suggesting that MAPKK activity is required for the activation of these two MBP kinases, underscoring their possible identity as MAPKs. Thus PD98059 may blunt the activation of gst1 and pal1 expression by inhibiting MAPKK activity, and consequently suppressing the activation of two MAPKs.
We had previously demonstrated that the production and transmission of the redox signals regulating gst1 and pal1 gene expression were independent of SA, Me-JA and ethylene signalling. Here we determined if the 48 and 46 kDa MBP kinases were activated in nahG, coi1 and etr1 plants in response to sustained ROI production using an in-gel kinase assay. The 48 kDa MBP kinase had a similar relative induction in nahG, coi1 and etr1 genetic backgrounds compared to the wild-type Col-0 accession (Figure 8c,d). In a similar fashion, the relative activation of the 46 kDa MBP kinase was not significantly different in coi1 and etr1 plants compared to Col-0 (Figure 8c,d). In nahG plants however, while the basal activity of the 46 kDa MBP kinase was similar to that determined for Col-0 plants, the relative induction of this kinase in response to sustained ROI accumulation was consistently reduced (Figure 8c,d). Hence the activation of these two MBP kinases can be resolved in nahG plants. The induction of gst1 and pal1 gene expression is not blunted in this genetic background, thus the 48 kDa rather than the 46 kDa MBP kinase may predominate in the redox signalling network regulating the expression of these genes.
To aid the functional dissection of the oxidative burst and cognate redox signalling, we developed a novel gst1::luc transgenic line to report, in real time, these dynamic cellular processes during the establishment of plant disease resistance. Leaf infiltration of MgCl2 or virulent Pst DC3000 resulted in a small but reproducible increase in LUC activity. This response probably reflected engagement of the so-called phase I oxidative burst, which is thought to be associated with wounding or inoculation with virulent microbial pathogens. In contrast, the large increase in LUC activity induced following Pst DC3000(avrB/avrRpt2) inoculation probably resulted from engagement of the phase II burst, which is thought to correlate with the establishment of disease resistance (Levine et al., 1994). Co-inoculation of the H2O2 scavenging enzyme CAT with Pst DC3000(avrB) significantly blunted the induction of LUC activity, confirming that ROIs functioned as the signal mediating gst1::luc gene expression. Due to size exclusion, infiltrated CAT was not expected to enter plant cells. The source of the ROI signal is therefore presumably the apoplast. Co-inoculation of the O2– scavenger SOD with Pst DC3000(avrB), in contrast, did not significantly decrease LUC activity. Hence H2O2, rather than O2–, is likely to be the ROI that cues the activation of gst1::luc expression. In bacteria and yeast, compelling evidence suggests that ROIs can engage distinct redox signalling pathways. For example, H2O2 signalling is mediated through the transcription factor OxyR in Escherichia coli (Storz et al., 1990), while the soxR and soxS gene products are required for the perception of O2– (Demple and Amabile-Cuevas, 1991). In a similar fashion, the redox signalling pathway mediating the activation of gst1, a prominent marker for the establishment of local and systemic disease resistance, may be engaged specifically by apoplastic H2O2.
Inoculation of Pst DC3000(avrB) containing mutations in either hrpA or hrpS failed to result in significant ROI accumulation or strong induction of LUC activity. The hrpS and hrpA gene products are thought to be required for the development and structural integrity, respectively, of the Hrp pilus, a filamentous surface appendage which may function as a conduit for the delivery of Pst avr gene products to the inside of plant cells (Roine et al., 1997). Hence Hrp pilus formation, and possibly the successful delivery of AVR proteins, may constitute a prerequisite for successful engagement of phase II of the oxidative burst.
While it is becoming increasingly apparent that ROIs integrate a diverse set of complementary defence mechanisms, the identity of the molecular machinery underlying the oxidative burst still remains to be rigorously established. In bean and cotton, the pathogen-activated oxidative burst is cyanide-sensitive, and apoplastic peroxidases are thought to be a direct source of ROIs in these cases (Bolwell and Wojtaszek, 1997; Martinez et al., 1998). However, a DPI-inhibited NADPH oxidase activity has been proposed to mediate the oxidative burst in soybean and tobacco cell-suspension cultures (Levine et al., 1994; Piedras et al., 1998). As many of the studies reported to date have been undertaken in different experimental systems, and have focused on only one possible enzymatic mechanism, we employed a variety of pharmacological agents to investigate the enzymatic source(s) of the redox cues responsible for engaging gst1 gene expression. Co-inoculation of either the NADPH oxidase inhibitor DPI or the peroxidase inhibitor NaN3 in combination with Pst DC3000(avrB/avrRpt2) significantly decreased the induction of gst1 gene expression. In contrast, no role could be found for the generation of ROIs via amine oxidase-type enzymes. Subject to the usual caveats associated with the deployment of these pharmacological agents (Barcelo, 1998), the data suggest that both an NADPH oxidase and a peroxidase-type enzyme contributed to the generation of redox signals that cued the engagement of gst1 gene expression in response to attempted Pst DC300(avrB) infection. Hence, mechanistically, the oxidative burst in Arabidopsis may resemble that of lettuce, which may also generate ROIs via both an NADPH oxidase system and apoplastic peroxidases (Bestwick et al., 1997). It will be useful to assess whether the contribution of these enzymes to ROI production varies in response to different pathogens.
Recently, NO has been proposed to potentiate the induction of hypersensitive cell death in soybean cells by ROIs, and to function independently of such intermediates to induce gene expression during the establishment of disease resistance (Delledonne et al., 1998; Durner et al., 1998). The deployment of both gain- and loss-of-function experiments, however, did not identify a direct function for NO in the engagement of gst1 gene expression. Moreover, NO was found not to potentiate the activation of gst1 via ROIs. No direct or indirect role for NO in the redox regulation of gst1 was therefore established. Thus the key redox responsive switches integral to the signalling network regulating gst1 gene expression are more likely to constitute critical regulatory thiols, the preferred targets of ROIs, rather than iron targets, the prototypic preference of NO (Stamler, 1994).
Ethylene, SA and Me-JA are thought to constitute key defence-signalling molecules that function to co-ordinate a diverse array of defence mechanisms underlying the establishment of disease resistance (Feys et al., 1994; Gaffney et al., 1993; Xu et al., 1994). We examined the potential role of these molecules on both the generation and perception of ROIs by monitoring gst1 gene expression in coi1 and etr1 plants, which are insensitive to ethylene and Me-JA, respectively. Neither mutant background significantly affected either the magnitude of induction or the temporal expression profile established by the gst1 gene in response to inoculation with Pst DC3000(avrB/avrRpt2). Moreover, similar observations were derived from Northern analysis of the redox-responsive pal1 gene. Thus engagement of the oxidative burst and cognate redox signalling may occur independently of Me-JA and ethylene. However, these conclusions do not preclude a potential role for these key defence-signalling molecules in the engagement of the oxidative burst and cognate redox signalling in tissues exhibiting either local or systemic acquired resistance, where their respective concentrations may exceed a critical threshold value.
Similar experiments were undertaken in nahG plants which show reduced accumulation of SA in response to attempted pathogen invasion (Gaffney et al., 1993). Neither expression of a gst1::luc reporter gene, nor the accumulation of endogenous gst1 transcripts, was affected in an nahG genetic background. Hence a suppression in the local accretion of SA was found not to affect either ROI production or the subsequent transmission of redox signals. These observations contrasted with previous studies that have suggested SA accumulation may potentiate the oxidative burst (reviewed by Van Camp et al., 1998). In cucumber hypocotyls, however, this phenomenon required an 18 h conditioning process that depended on de novo protein synthesis (Fauth et al., 1996). This mechanism would probably not be operational in our experimental system. In contrast, co-application of SA in conjunction with an avirulent pathogen has been reported to potentiate the oxidative burst without a prior conditioning step in soybean suspension cultures (Shirasu et al., 1997). However, the required SA concentration has only been measured routinely in tissues surrounding local HR lesions (Malamy et al., 1990). This mechanism also may not be operational in naive tissue, but deployed following the development of acquired resistance in local and possibly systemic tissues. Our observations of redox responsive gene expression in naive tissue of nahG transgenic plants would be consistent with this hypothesis.
A pivotal role for MAPKs in disease resistance has recently begun to emerge (Ligterink et al., 1997; Romeis et al., 1999), although these kinases are thought to function independently of the oxidative burst. It has recently been reported that the overexpression of constitutively active deletion mutants of a mitogen-activated protein kinase, kinase, kinase (MAPKKK), activated two of the six co-overexpressed MAPKs tested in Arabidopsis protoplasts exposed to acute H2O2 stress (Kovtun et al., 2000). Moreover, exposure of tobacco cell suspension cultures to necrosis-inducing concentrations of ozone has also recently been proposed to stimulate MAPK activity (Samuel et al., 2000). In the context of plant disease resistance, our observations have extended these studies by demonstrating that sustained production of sublethal concentrations of ROIs, mimicking the kinetics of the oxidative burst, resulted in the rapid activation of 48 and 46 kDa MBP kinase activities in Arabidopsis leaf tissue. The substrate specificity and molecular mass of these kinases predicted a MAPK identity. Moreover, PD98059, an inhibitor of MAPKKs, completely suppressed the activation of both kinases, suggesting that MAPKK activity is a prerequisite for their activation, further underscoring their proposed MAPK identity. The molecular mass and activation kinetics of the 48 and 46 kDa MBP kinases suggest they belong to the stress-activated class of MAPKs in Arabidopsis (Mizoguchi et al., 1997), which contain orthologues of the SA-inducible and wound-inducible MAPKs from tobacco, which may be important mediators of SA and Me-JA-dependent signalling, respectively (Seo et al., 1995; Zhang and Klessig, 1997).
Induction of the redox responsive genes gst1 and pal1 was found to occur independently of SA, Me-JA and ethylene. We therefore investigated the activation of the 48 and 46 kDa kinases in nahG, coi1 and etr1 genetic backgrounds in response to sustained ROI accumulation. Activation of the 48 kDa kinase in these genetic backgrounds paralleled that in wild-type plants; hence this kinase is activated independently of the action of SA, Me-JA and ethylene. In contrast, while the activation of the 46 kDa kinase was similar in coi1 and etr1 mutants compared to wild-type plants, in a nahG genetic background the activation of this kinase was consistently reduced in comparison with wild-type plants. Hence SA may be either necessary or sufficient for maximum activation of this kinase. The 48 and 46 kDa kinase activities can be resolved in the absence of SA. As the induction of gst1 and pal1 genes is not affected within a nahG genetic background, the 48 kDa rather than the 46 kDa kinase probably undertakes a predominant functional role in the transmission of redox cues. A MAPK module may therefore be a key feature of the redox signalling pathway engaged following activation of the oxidative burst during the establishment of disease resistance.
In animal cells there is compelling evidence for both direct and indirect mechanisms for the regulation of gene expression in response to changes in cellular redox status. In response to severe hyperoxic states, key cysteine residues of redox-modulated transcription factors may become oxidized, effecting changes in the expression profile of their target genes (Abate et al., 1990). In contrast, signal transmission in response to lower levels of ROIs may require the action of specific protein kinases. Recently, ROIs have been shown to engage the stress-activated class of MAPKs, including the c-Jun N-terminal kinase group, that contribute to a MAPK cascade activated in response to specific environmental stresses (Klotz et al., 1999). Hence our data, highlighting a pivotal role for a MAPK module in ROI-mediated signalling during the establishment of disease resistance, suggest that significant parallels may exist in the transduction of stress-induced redox signals in plants and animals.
Local ROI accumulation has recently been shown to lead to the establishment of acquired resistance in both local and systemic tissues (Alvarez et al., 1998; Chamnongpol et al., 1998). This mechanism is thought to operate via the SA-dependent induction of acidic PR proteins (Chamnongpol et al., 1998), and the production of SA precursors is known to be a major function of PAL during the development of disease resistance (Mauch-Mani and Slusarenko, 1996). We have demonstrated that expression of the Arabidopsis pal1 gene is mediated via redox cues following engagement of the oxidative burst. Moreover, this redox-signalling network functions independently of SA, Me-JA and ethylene, but is dependent on a 48 kDa MAPK activity. Thus local ROI production during the oxidative burst and reiterated systemic microbursts may activate the expression of pal1, thereby driving the biosynthesis of SA, and leading to the accumulation of PR proteins and the subsequent establishment of acquired resistance in both local and systemic tissues.
The oxidative burst may therefore engage a bifurcating redox signalling pathway that orchestrates the production of both key defence signals and pivotal anti-oxidant defences, leading to the development of both disease resistance and the limitation of HR lesion formation, respectively. We have utilized the gst1::luc transgenic line described here to undertake a saturating genetic screen to uncover mutations that affect redox signal transmission through this network (B.-W. Yun and G.J. Loake, unpublished results). The characterization of the corresponding mutants should provide significant insights into the mechanisms underlying redox signalling during the establishment of plant-disease resistance.
Unless otherwise stated, all reagents used were supplied by Sigma (Poole, Dorset, UK).
Nucleic acid analysis
For Northern analysis, 12 µg total RNA samples were separated by electrophoresis through formaldehyde–agarose gels and transferred to a nylon membrane (Amersham, Poole, Dorset). 32P-labelled DNA probes were prepared using a Prime-a-Gene labelling kit (Promega, Madison, WI, USA). Blot hybridizations were quantified with a phosporImager (Molecular Dynamics Inc., Sunnyvale, CA, USA) and normalized with reference to r18 hybridization.
Sequencing reactions were prepared and run on a HYBAID Omnigene Thermocycler using the Perkin-Elmer ABI PRISM Dye Terminator Cycle Sequencing Ready Reaction Kit, according to the manufacturer's instructions. Sequence data were transferred to the UNIX-based gcg package for further analysis.
Cloning of the gst1 promoter
A cosmid library of Arabidopsis (ecotype Ws) genomic DNA (Arabidopsis Stock Center, Ohio, USA) was screened at 65°C with a 32P-labelled gene-specific probe of the gst1 gene. The fragment was obtained by PCR amplification of the 3′ untranslated region of a template gst1 cDNA clone using primers previously designed by Sharma et al. (1996). DNA from a positive clone was isolated and subjected to restriction analysis followed by hybridization with the gst1 gene-specific probe. An 8 kb KpnI fragment containing gst1 was subcloned from a positive cosmid into pBluescript SK– (Stratagene, La Jolla, CA, USA) and further digested with EcoRI, and resulting fragments subcloned into pBluescript SK– and sequenced.
Generation of transgenic plants
PCR was used to amplify a 909 bp promoter fragment of gst1 with SacI and NcoI sites at 5′ and 3′ ends, respectively, using the following primers: 5′ primer-5′-TATAGAGCTCGGAAACAGCTATGACCATG-3′ and 3′ primer-5′-TTGATTCCTGCCATGGGTTAATACTGTGT-3′. This fragment was subsequently subcloned into pART7 (Gleave, 1992) containing a LUC reporter gene and OCS termination sequence, and transgenic plants generated by ‘floral dip’ (Clough and Bent, 1998).
Genetic crosses were undertaken by dissecting and emasculating unopened flower buds, then using the remaining pistils as recipients for pollen from four opened flowers. Homozygous transgenic plants were generated from their progeny. Pollen from nahG plants was used to fertilize Col-0 gst1:luc plants. Progeny containing nahG were identified by their brown deposits in root tissue when grown on MS media containing 1 mm SA.
Treatment of plants
Bacteria were maintained as previously described (Dangl et al., 1992). Mutant hrp Pseudomonas strains were transformed with the pVB01 plasmid by electroporation as described by Keen et al. (1990). Bacteria were inoculated into individual leaves by pressure infiltration using a 1 ml syringe, and 10 µl of bacteria were infiltrated into the abaxial leaf surface.
Aspergillus niger glucose oxidase 2.5 U ml−1 was added to 2.5 mm d-glucose in 20 mm sodium phosphate buffer pH 6.5, immediately prior to infiltration. Catalase (bovine liver), 300 U ml−1; superoxide dismutase, 25 U ml−1; DPI, 3 µm; sodium azide, 1 µm; K252a (Calbiochem, Beeston, Nottingham, UK), 1 µm; staurosporine (Calbiochem), 1 µm; and PD98059 (Calbiochem), 100 µm; sodium nitroprusside, 0.5 mm in 10 mm Tris–HCl pH 7.5; putrescine, 1 mm; l-arginine, 1 mm; L-NNA, 100 µm; and cantharidin, 5 µm were co-inoculated with 10 µl bacterial suspension or H2O2-generating system, as appropriate. Controls were also carried out using buffers alone. Bacterial growth curves were used to ensure inhibitors did not adversely affect bacteria.
Leaves of pathogen inoculated gst1::luc transgenic plants were painted with a solution containing 1 mm Luciferin (Promega) and 0.01% triton X-100 and 0.03% Silwet (Union Carbide) in a 1 mm sodium citrate buffer pH 5.8. All in planta LUC imaging was performed using an ultra-low-light imaging camera system (EG&G Berthold Luminograph 980). Images were routinely collected over a 1 sec period. Microscopy imaging was carried out using a Nikon Optiphot-2 microscope.
Preparation of protein extracts and in-gel kinase assays
Five-week-old plants were used for G/GO infiltration. Protein extraction and in-gel kinase assays were performed as described previously (Zhang and Klessig, 1997). Pre-stained molecular mass markers (Bio-Rad, Hercules, CA, USA) were employed to calculate the approximate mass of kinase activities. Quantification of relative kinase activity was determined via phophorImage analysis.
We are grateful to Scot Uknes, John Turner and the Arabidopsis stock centre, Nottingham for providing nahG, coi1 and etr1 Arabidopsis seed, respectively, and Tina Romeis for advice on in-gel kinase assays. The authors would also like to thank Roger Innes and Andrew Bent for the kind gift of Pseudomonas syringae strains expressing avrB and avrRpt2, respectively, Sheng He for Pseudomonas syringae mutants hrpS and hrpA, and Fred Ausubel for providing the gst1 cDNA clone. B-W.Y. was supported by grant F158BO from the Leverhulme Trust and J.G. was the recipient of a BBSRC research studentship.