Early events occurring during the hypersensitive resistance response (HR) were examined using the avrRpm1/RPM1 gene-for-gene interaction in Arabidopsis challenged by Pseudomonas syringae pv. tomato. Increases in cytosolic Ca2+ were measured in whole leaves using aequorin-mediated bioluminescence. During the HR a sustained increase in Ca2+ was observed which was dependent on the presence of both a functional RPM1 gene product and delivery of the cognate avirulence gene product AvrRpm1. The sequence-unrelated avirulence gene avrB, which also interacts with RPM1, generated a significantly later but similarly prolonged increase in cytosolic Ca2+. Accumulation of H2O2 at reaction sites, as revealed by electron microscopy, occurred within the same time frame as the changes in cytosolic Ca2+. The NADPH oxidase inhibitor diphenylene iodonium chloride did not affect the calcium signature, but did block H2O2 accumulation and the HR. By contrast, the calcium-channel blocker LaCl3 suppressed the increase in cytosolic Ca2+ as well as H2O2 accumulation and the HR, placing calcium elevation upstream of the oxidative burst.
Plants have developed an innate surveillance mechanism that enables them to respond rapidly to attempted invasion by pathogens or parasites. Pathogen recognition is a prerequisite for the induction of defence responses. In the widely studied gene-for-gene interaction the plant's resistance (R) gene product acts as a signalling adaptor for the pathogen avirulence (avr) gene product, leading to the elaboration of cell death recognized as the hypersensitive reaction, HR ( Flor, 1971; Yang et al. 1997 ). In the interactions occurring with plant pathogenic bacteria the proteins encoded by avr genes are, in most cases, delivered into plant cells by the bacterial hrp-dependent type III secretion system ( Alfano & Collmer, 1997). Plant cell collapse during the HR is associated with the generation of antimicrobial conditions and the restriction of colonization by potential pathogens ( Hammond-Kosack & Jones, 1996).
Experiments with cell suspensions, although implicating a role for Ca2+, do not replicate the modulating effects of unchallenged cells within responding tissues, or the continued presence and contribution of the pathogen to the interaction. It is clearly preferable to monitor Ca2+ fluxes within whole tissues, but the rigid cell wall and cell turgor make traditional methods of measuring cytosolic free Ca2+ difficult in whole plants ( Trewavas & Malho, 1998). There is only one report, involving ratio imaging of micro-injected epidermal strips from cowpea after challenge with basidiospores of Uromyces phaseoli, which provides compelling evidence for increased cytosolic calcium concentrations ([Ca2+]cyt) during incompatible interactions occurring in planta (Xu & Heath, 1998).
Transformed plants expressing the Ca2+-sensitive, luminescent protein aequeorin or yellow cameleon have been successfully used to quantify the intracellular fluxes associated with a range of diverse physical and pharmacological stimuli ( Allen et al. 1999 ; Chandra & Low, 1997; Knight et al. 1991 ; Trewavas & Malho, 1998). Although perhaps not as sensitive as fluorescent dyes, the use of aequorin allows quantification of responses in whole plants, thereby permitting measurement of the kinetics of a Ca2+ transient in a non-destructive manner, and avoiding potentially damaging loading methods associated with reporting dyes. We now describe the use of transgenic Arabidopsis plants expressing aequorin ( Knight & Knight, 1995; Knight et al. 1996 ) to investigate the role of [Ca2+]cyt as an important early signal during the gene- for-gene interaction between the bacterial pathogen Pseudomonas syringae pv. tomato DC3000 (DC3000) harbouring either the avrRpm1 or avrB avirulence genes, and Arabidopsis thaliana carrying the matching resistance gene RPM1 ( Dangl et al. 1992 ; Grant et al. 1995 ; Tamaki et al. 1988 ). RPM1 belongs to the nucleotide-binding site, leucine-rich repeat (NBS-LRR) class of R-gene proteins that are predicted to reside in the cytoplasm ( Hammond-Kosack & Jones, 1997; Jones & Jones, 1996). Sustained increases in [Ca2+]cyt were found during the incompatible interaction, prior to the onset of cell collapse characteristic of the HR. The sustained increase elicited by AvrRpm1 was temporally separated from the signature generated by AvrB, closely associated with the generation of ROS, that is, the oxidative burst, and was absolutely dependent upon both a functional type III secretion system in the bacterium and a functional R gene in the plant.
Generation of a gene-for-gene specific calcium signature
Infiltration of leaves with concentrated bacterial inoculum allows the simultaneous exposure of virtually all cells within the leaf to the challenging bacterium. The changes in pCa, resulting from Ca2+-dependent aequorin bioluminescence after the inoculation of leaves with DC3000 carrying different avr genes, are illustrated in Fig. 1. The traces shown are representative of at least five experiments carried out with each interaction. All strains, including DC3000 carrying the cloning vector alone (pVSP61, Innes et al. 1993 ), produced an early transient increase in [Ca2+]cyt after which luminescence returned to background levels. The first Ca2+ transient lasted for about 10 min, with a maximum 8–12 min after challenge. Leaves challenged with DC3000 carrying either avrRpm1 or avrB elicited a second sustained increase in [Ca2+]cyt of a similar magnitude (reaching between 0.3 and 0.4 pCa). Due to the inherent difficulty in establishing the time at which the second increase in [Ca2+]cyt began relative to background signals, we chose to use the time point of maximum calcium elevation as an unambiguous parameter for quantitative comparison of the timing of cytosolic Ca2+ influx elicited by different treatments. Bacteria carrying the avrRpm1 or avrB genes gave maximal increases in [Ca2+]cyt at 105 ± 10 min (SE, n = 21) and 137 ± 7 min (n = 12). The ability of avrRPM1 to elicit a significantly earlier influx of [Ca2+]cyt was clearly shown by using real-time imaging of Ca2+-dependent aequorin luminescence in plants challenged with both DC3000 derivatives, as shown in Fig. 2. Interestingly, no significant bioluminescence could be imaged in leaves challenged with DC3000(avrRpt2) which is recognized by the RPS2 resistance gene in Col-0 ( Bent et al. 1994 ; Mindrinos et al. 1994 ). This observation suggests either that the timing of Ca2+ influx may reflect the substantially delayed HR (at least 16 h, compared to 4–5 h for avrB/avrRpm1), or the aequorin reporter system may not be sensitive enough to detect the cytosolic Ca2+ increases generated during the avrRpt2/RPS2 interaction.
Requirement for signal delivery and RPM1
The time required before the onset of the second, avrRPM-specific transient (when the pCa rises above background levels), commencing within a window of 55–65 min after inoculation, may reflect a requirement for the expression of avrRpm1 by bacteria after inoculation, or for assembly of a functional secretion system. Figure 3 demonstrates that constitutive overexpression of the avirulence gene avrRpm1, under the control of a strong vector promoter, 3x lacUV on pDSK600 ( Murillo et al. 1994 ), did not affect the timing or magnitude of the Ca2+ signature. A similar lack of effect was observed when avrB was overexpressed in the same way ( Fig. 3). However, mutants of DC3000 compromised in assembly of the type III secretion apparatus (DC3000::hrpL or DC3000::hrpRS) did not generate the second increase (even if either avrRpm1 or avrB were overexpressed) and failed to elicit the HR. These results therefore demonstrate an absolute requirement for a functional type III secretion apparatus, and presumably delivery of the avr gene product into the plant cell before elevation of cytosolic Ca2+.
Functional RPM1 is required for the increase in cytosolic Ca2+
To confirm that functional RPM1 is required for the specific [Ca2+]cyt increase, we introduced pMAQ2 (for aequorin expression) into a loss of function rpm1 mutant (rps3-1) in which a frameshift results in the loss of the last three leucine-rich repeats from the encoded protein ( Bisgrove et al. 1994 ; Grant et al. 1995 ). Transgenic rps3-1 expressing reconstituted aequorin challenged with either DC3000(avrRpm1) or DC3000(avrB) failed to generate the [Ca2+]cyt increase ( Fig. 4) or to undergo the HR characteristic of the incompatible interaction. The rpm1 mutants were not, however, compromised in their ability to elicit aequorin-mediated bioluminescence in response to touch or cold shock, compared to RPM1 plants (data not shown).
Timing of signal delivery and the Ca2+ transients.
Figures 1and 2 show that the avrB/RPM1 Ca2+ signature peaks significantly later than that elicited by the avrRpm1/RPM1 interaction. As the difference was not abolished by constitutive overexpression of avr genes, an alternative explanation for the delayed response is that AvrB may be produced and delivered into the plant cells more slowly than AvrRpm1. In order to estimate the time required for DC3000 to deliver the Avr proteins, the induction time for the HR was analysed by injecting challenged sites with streptomycin (400 μg ml−1) in order to block bacterial protein synthesis and consequently prevent the activation of programmed hypersensitive cell death ( Klement & Goodman, 1967). In response to DC3000 expressing the avrB, avrRpm1 or avrRpt2 genes, the induction time was found to be identical, at 60–75 min ( Table 1). After this time, infiltration with the antibiotic did not prevent confluent tissue collapse in any of the gene-for-gene interactions tested. The induction time estimates the minimum time required for delivery of avr gene products into the plant cell. Our results indicate a rapid recognition of both AvrRpm1 and AvrB; however AvrRpm1 can activate Ca2+ influx more rapidly. The speed of response was supported by experiments in which the effects of antibiotic infiltration on [Ca2+]cyt were examined. We found that both the avrB/RPM1- and avrRpml/RPM1-specific calcium signatures continued to be detected after infiltration with streptomycin at 75 min but not 60 min after inoculation (data not shown). Interestingly, our data also suggest that AvrRpt2 is delivered within a similar time frame to AvrRpm1 and AvrB, even although development of a visible HR takes nearly four times longer.
Table 1. Effect of infiltration with streptomycin to block bacterial protein synthesis on development of the HR in ecotype Col-0
Occurrence of hypersensitive collapse caused by DC3000 expressing avr genes b
a Streptomycin concentration, 400 μg ml −1.
Data summarized from two experiments; the confluence of the HR was rated as –, none; +, patchy; ++, >50% collapse; +++, 100% of infiltrated zone collapsed 2 days after inoculation.
Elevation of [Ca2+]cyt and the oxidative burst
The relationship between timing of the major [Ca2+]cyt increase and activation of the oxidative burst in challenged cells was investigated. Accumulation of H2O2 was detected at the EM level using the CeCl3 staining technique ( Bestwick et al. 1997 ). Electron-dense deposits of cerium perhydroxides, indicating sites of H2O2 generation, were rarely observed during the compatible interaction with DC3000 alone ( Fig. 5a) or in the rps3-1 mutant genotype (data not shown), but were detected adjacent to bacteria in Arabidopsis tissues undergoing the HR ( Fig. 5b). The site of the major accumulation of cerium perhydroxide deposits was in the plant cell wall. Time-course studies, based on the calculation of a staining index, showed that no H2O2 accumulation was detected until 80 min after infiltration. The accumulation of H2O2 appeared to follow initiation of the major Ca2+ transient which was recorded from leaves harvested during one of these experiments ( Fig. 5c). However, our ability to determine the precise timing of processes occurring in responding cells was limited by the tissue-sampling techniques employed. It was therefore not clear if the generation of ROS was indeed downstream of the specific [Ca2+]cyt increase, or was activated at the same time in a parallel pathway.
To address the relationship between the calcium increase, ROS generation and hypersensitive cell death, we used a pharmacological approach on whole plants. The effects were examined of catalase to degrade H2O2, DPI to inhibit NADPH oxidase, and LaCl3 to block Ca2+ channels. The results obtained are summarized in Fig. 6. Co-infiltration of DC3000(avrRpm1) with DPI (7 μm) or catalase (25 μg ml−1) did not affect the sustained increase in [Ca2+]cyt ( Fig. 6a), but greatly reduced staining with CeCl3, that is, H2O2 accumulation ( Fig. 6b). Both DPI and catalase also only partially reduced the degree of cell collapse observed in the HR ( Fig. 6c). By contrast, LaCl3 abolished the Ca2+ signature, and blocked both H2O2 accumulation and the HR ( Fig. 6a–c). Overall our results therefore placed generation of ROS by the oxidative burst downstream of the increase in [Ca2+]cyt.
Molecular characterization of the RPM1 disease-resistance gene in Arabidopsis, and its matching but sequence-unrelated bacterial avirulence genes avrRpm1 and avrB, have allowed the analysis of early biochemical events in defined gene-for-gene interactions leading to the HR. We have used transgenic Arabidopsis Col-0 plants expressing aequorin to demonstrate the generation, during the HR, of a cytoplasmic Ca2+ signature that is unique in terms of its temporal characteristics. The increase in [Ca2+]cyt observed within intact leaves was biphasic. The initial calcium transient was elicited in an R gene-independent manner, and could also be generated by challenge with Escherichia coli (data not shown). Specificity was encoded in a second sustained increase which requires a functional RPM1 gene product and delivery of the cognate avirulence products, either AvrB or AvrRpm1. Our results suggest an intimately co-ordinated response mediated through the interacting gene products causing extremely rapid Ca2+ gating.
Significantly, we have been able to examine [Ca2+]cyt in tissues undergoing hypersensitive cell death, a response that is often not reproduced in experiments with cell-suspension cultures, either using non-specific elicitation or through gene-for-gene interactions ( Jabs et al. 1997 ; Piedras et al. 1998 ). The results obtained strongly implicate an important role for sustained cytoplasmic Ca2+ elevation as a post-recognition molecular switch, capable of communicating primary recognition responses to multiple downstream effectors, including activation of the oxidative burst.
Although the signalling events that directly follow the sustained rise in intracellular Ca2+ have yet to be determined, a role for post-transcriptional activation of Ca2+-dependent phosphatases and kinases is likely, as discussed by McAinsh & Hetherington 1998) and Trewavas & Malho, (1998). Recently, calmodulin and calmodulin-dependent NAD kinases have been suggested as the potential targets of elevated Ca2+ during the HR ( Harding & Roberts, 1998; Harding et al. 1997 ). Indeed, specific calmodulin isoforms have been demonstrated to be activated in a gene-for-gene specific manner and to participate in the Ca2+-mediated induction of defence responses in tobacco ( Heo et al. 1999 ). In tobacco suspension cultures, pharmacological and 45Ca2+ uptake studies implicate an unquantified Ca2+ influx as necessary for transient activation of MAP kinases, but the link between kinases, the oxidative burst and the HR has not been fully resolved ( Grant & Mansfield, 1999; Romeis et al. 1999 ). The possibility that increased cytosolic Ca2+ may play a role in the control of early transcription occurring during the incompatible interaction also merits detailed examination.
The temporal separation of Ca2+ signatures using two avr genes, avrB and avrRpm1, both of which genetically interact with RPM1, is intriguing. Despite their apparently simultaneous delivery, as indicated by the experiment on induction times, AvrRpm1 elicited maximal [Ca2+]cyt increases at least 20–30 min prior to AvrB-elicited responses. As the overexpression of the avr genes did not alter the timing of the Ca2+ transients, differing rates of expression of avrRpm1 and avrB could not have caused the delay observed with avrB. There are several other possible explanations for the differential responses observed: for example (i) an in planta modification may be necessary to make AvrB functionally active in this interaction; (ii) AvrRpm1 and AvrB may have different affinities for RPM1, leading to modulation of downstream regulation of [Ca2+]cyt; or (iii) the Avr proteins may activate RPM1 via different upstream intermediates. Although we cannot rule out a selective hierarchy of delivery by the bacteria, our induction experiments indicate that sufficient AvrB is delivered to the plant cell 1 h before maximal [Ca2+]cyt is attained.
The increases in [Ca2+]cyt elicited by the avrRpm1/RPM1 interaction may be generated from internal or external sources via modification of Ca2+ channels on the cytoplasmic face of membranes, through the action of the intracellular RPM1 protein. Such activation might be expected to differ mechanistically from [Ca2+]cyt increases generated through the interaction of extracellularly localized R-gene products such as Cf9 ( Piedras et al. 1998 ), or non-specific elicitors which probably bind to receptors in the plasma membrane ( Jabs et al. 1997 ). The potential involvement of RPM1 in Ca2+-channel activation is supported by its characterization as a peripheral membrane-associated protein ( Boyes et al. 1998 ), and also by recent studies in our laboratory that show the interaction of RPM1 with a predicted membrane protein in the two-hybrid system (M. deTorres and M. Grant, unpublished results). The proposed targeting of AvrRpm1 and AvrB to the plant plasma membrane for activity ( Nimchuk et al. 2000 ) suggests that Ca2+ elevation may be mediated via the action of a membrane-localized resistance interaction complex, somewhat analogous to the mammalian apoptosome ( Zou et al. 1999 ). Such an association would enable RPM1-mediated gating of Ca2+ channels in the immediate vicinity of the activated complex. This might result in the generation of ROS only in those regions of the plant cell in close proximity to bacteria that are in direct contact with the cell wall, as indicated by the localized deposition of cerium perhydroxides illustrated in Fig. 5(a).
The site of H2O2 accumulation found in Arabidopsis was very similar to that reported in lettuce by Bestwick et al. (1997) . However, DPI was a much more efficient inhibitor of ROS generation in Arabidopsis, indicating a more essential role for NADPH oxidase in the reaction described here. It was difficult to separate unequivocally the timing of the activation of sustained Ca2+ transient and the oxidative burst. However, time-course experiments suggested that generation of ROS followed the increase in [Ca2+]cyt ( Fig. 5). This order of response was strongly supported by experiments with the NADPH oxidase inhibitor which did not affect the [Ca2+]cyt, and LaCl3 which abolished both the calcium signature and ROS generation. There is accumulating evidence that ROS production in Arabidopsis closely resembles the respiratory burst of animal phagocytes, which has a requirement for continuous Ca2+ influx (reviewed by Bolwell, 1999). Reports from cell-culture systems have also indicated a requirement for Ca2+ to generate ROS. For example, reduced formation of ROS was found using Ca2+-depleted media by Schwacke & Hager, (1992). The R-gene-mediated increase in [Ca2+]cyt may facilitate Ca2+-dependent phosphorylation of the Arabidopsis homologues of the human neutrophil p67phox and p47phox during NADPH oxidase assembly ( Keller et al. 1998 ; Xing et al. 1997a ; Xing et al. 1997b ). Alternatively, Ca2+ may participate in the assembly or direct activation of NADPH oxidase through binding to the unique EF hand motifs located in the N-terminal region of the Arabidopsis gp91phox homologue ( Keller et al. 1998 ).
The sustained duration of the avrRpm1/RPM1 Ca2+ signature is reminiscent of intracellular calcium increases reported in studies of apoptosis in mammals ( Kong et al. 1997 ; Lipton & Nicotera, 1998). Slow increases in [Ca2+]cyt are important in tumour necrosis factor (TNF)-mediated cell death, even in the absence of extracellular calcium ( Kong et al. 1997 ). TNF-mediated cell death is accomplished via activation of the mammalian homologues of the Caenorhabditis elegans cell-death genes ced3/4/9. Intriguingly, the mammalian homologues of CED-4, Apaf-1 ( Zou et al. 1997 ) and FLASH ( Imai et al. 1999 ) have significant amino-acid homology to RPM1 and other NBS-LRR plant R genes within a region termed the apoptotic ATPase domain ( Aravind et al. 1999 ). As determined for CED-4 and Apaf1, this domain appears to be involved in protein oligomerization, a necessary requirement for activation of the associated caspase ( Zou et al. 1999 ).
One major restriction of the aequorin system is in determining the subcellular localization of calcium. Although the sustained Ca2+ increases reported here are relatively low, important Ca2+ signals are often extremely localized and are probably influenced by cell-wall architecture and distribution of organelles. A number of sources might contribute to the increase in [Ca2+]cyt, and this diversity highlights the need to use whole-plant systems to examine calcium signalling in plant–pathogen interactions. Visualization of free Ca2+ dynamics in cells challenged with pathogens carrying the avrRpm1 gene will require the use of one of the new generation of intracellular calcium indicators, such as the chameleon system. In chameleons, engineered variants of green fluorescent protein are used to probe their ionic environment using intramolecular fluorescence–resonance-energy transfer ( Miyawaki et al. 1997 ). Recently a pH-independent, green fluorescent protein-based calcium indicator, yellow cameleon, has been used to report cytoplasmic calcium dynamics in Arabidopsis guard cells ( Allen et al. 1999 ). If it is of sufficient sensitivity, this system may provide an informative way of determining the local address of the Ca2+ response reported here, and offer opportunities for further dissection of early events during gene-for-gene interactions.
Coelenterazine was obtained from Molecular Probes (OR, USA). All other chemicals were from Sigma (Gillingham, Dorset, UK).
Maintenance of bacteria
Pseudomonas syringae pv. tomato DC3000 containing avirulence genes cloned into the broad host-range vectors pVSP61 or pDSK600 were maintained under selection and cultured as described ( Grant et al. 1995 ; Murillo et al. 1994 ).
Growth and treatment of plants
Homozygous Col-0 plants expressing aequorin under control of the 35S promoter ( Knight & Knight, 1995; Knight et al. 1996 ) were sown in Levington's F2 compost and vernalized for 2 days at 4°C. Seven days after sowing, seedlings were transferred to pots (5 × 4 × 4 cm) containing Levington's F2 compost and grown under short-day conditions in controlled-environment chambers (12 h light, 23°C day, 20°C night) for 4–5 weeks before use.
Aequorin was reconstituted by pressure infiltration of a whole leaf on the abaxial surface with a 2 μm solution of coelenterazine using a 1 ml disposable syringe without a needle. Treated plants were left in the dark overnight. A DC3000 culture was grown overnight in 10 ml King's B solution ( King et al. 1954 ) supplemented with appropriate antibiotics. Overnight cultures were washed once in 10 m m MgCl2 and resuspended to 5 × 108 colony-forming units ml−1 (cfu ml−1) in 10 m m MgCl2.
Measurement of [Ca2+]cyt by aequorin luminometry
The bacterial suspension was pressure infiltrated into the abaxial surface of recently expanded leaves until the entire lamina appeared water-soaked. The leaf containing reconstituted aequorin was immediately detached and placed in a transparent plastic cuvette, and the luminescence read using a digital chemiluminometer with a discriminator, as previously described ( Knight & Knight, 1995). Luminescence counts were integrated every 5 sec over a 3 h period prior to discharging the remaining aequorin by injection with 0.75 m CaCl2 in 25% ethanol. Data were converted to pCa using a calibration curve empirically determined for the specific isoform of aequorin expressed in the transgenic line of Arabidopsis ( Knight et al. 1996 ). Each Ca2+ signature was derived from an individual leaf, and is representative of a typical response to that pathogen genotype derived from a minimum of five independent experiments. During all treatments a control inoculation using DC3000(avrRpm1) was run in parallel in order to confirm that the plants had the capacity to respond with the HR.
To address whether the signature was concentration- dependent, we performed a dose–response curve. The avrRpm1/ RPM1 response was saturated within a bacterial inoculum range of 3 × 108 to 8 × 108 cfu ml−1. For all experiments described here we used 5 × 108 cfu ml−1.
Construction of transgenic plants
Plants containing the rps3-1 frameshift allele of RPM1 were grown and transformed by vacuum infiltration with Agrobacterium C58 containing pMAQ2 ( Knight et al. 1991 ), as described by Clough & Bent, (1998). Primary transformants were isolated under kanamycin selection (40 μg ml−1) and homozygous transgenic plants were generated from the progeny.
Calcium-dependent aequorin luminescence was imaged ( Campbell et al. 1996 ) using an intensified CCD camera (model EDC-02), with camera control unit (HRPCS-2) and image acquisition and processing software (IFS216), all sourced from Photek (St Leonards-on-Sea, UK). In these experiments a 10 μm solution of coelenterazine was used in the overnight reconstitution of aequorin.
Accumulation of H2O2 was examined by electron microscopy using the CeCl3 staining technique. Deposits of electron-dense cerium perhydroxides, which formed as a result of the interaction between CeCl3 and H2O2, were scored by transmission electron microscopy in order to calculate an H2O2 index which had a maximum score of 2.0 ( Bestwick et al. 1997 ). Potential inhibitors of H2O2 accumulation were co-infiltrated with bacteria and added to staining solutions, also as described by Bestwick et al. (1997) .
We thank Jeff Dangl for the DC3000::hrpRS and ::hrpL strains, and Noel Keen for avrRpm1 and avrB cloned into pDSK600. Our research was supported by The Royal Society and BBSRC, UK; Hort-Research, New Zealand; and EU Framework IV project BIO-CT97-2244.