PopA is released by type III secretion from the bacterial plant pathogen Ralstonia solanacearum and triggers the hypersensitive response (HR) in tobacco. The function of PopA remains obscure, mainly because mutants lacking this protein are not altered in their ability to interact with plants. In an attempt to identify the site of PopA activity in plant cells, we generated transgenic tobacco plants expressing the popA gene under the control of an inducible promoter. Immunocytologic analysis revealed that the HR phenotype of these plants correlated with the presence of PopA at the plant plasma membrane. Membrane localization was observed irrespective of whether the protein was designed to accumulate in the cytoplasm or to be secreted by the plant cell, suggesting a general lipid-binding ability. We found that the protein had a high affinity for sterols and sphingolipids in vitro and that it required Ca2+ for both lipid binding and oligomerization. In addition, the protein was integrated into liposomes and membranes from Xenopus laevis oocytes where it formed ion-conducting pores. These characteristics suggest that PopA is part of a system that aims to attach the host cell plasma membrane and to allow molecules cross this barrier.
Gram-negative bacterial pathogens of mammals possess a pathogenicity mechanism, termed the type III secretion system (TTSS), that enables them to secrete proteins and to inject them into eukaryotic host cells (Anderson and Schneewind, 1999). During infection, the proteins released by this system are involved in mechanical functions, like the adhesion of the bacterium to the target cell surface and the assembly of a translocation apparatus across the eukaryotic cell membrane (Cornelis, 2002). These structures allow the delivery of effector proteins into the cytosol of the host cell where they corrupt the cell machinery (Cornelis, 2002; Waterman and Holden, 2003).
Type III secretion systems are also conserved in at least the four major genera of plant pathogenic bacteria, Erwinia, Pseudomonas, Xanthomonas and Ralstonia (Alfano and Collmer, 2004). Although little is yet known about the virulence function of TTSS substrates from plant pathogens, there is emerging evidence that some of them might be functional analogues of type III effectors from animal pathogens (Shao et al., 2002). An example are the cysteine proteases produced by some of the plant pathogens (Axtell et al., 2003; Hotson et al., 2003; Shao et al., 2003). Several Pseudomonas syringae effectors have recently been shown to suppress plant defence responses (Espinosa and Alfano, 2004) and have thus a role similar to effectors from animal pathogens suppressing immunity (Orth, 2002; Sauvonnet et al., 2002). One of the defence mechanisms suppressed by type III effectors from plant pathogens is the hypersensitive response (HR; Abramovitch et al., 2003). When pathogen-derived molecules (elicitors) are recognized by plant cells, a complex signalling cascade is triggered in the host, which results in gene activation, de novo protein synthesis, the production of antimicrobial compounds, and cell death at the infection sites. The HR resembles the programmed cell death of animal cells (Heath, 2000) and aims at preventing the spread of disease into healthy tissues (Lam et al., 2001). TTSS substrates from bacterial plant pathogens can also induce the HR (Nürnberger et al., 2004), suggesting that a fine tuning between the elicitation and suppression of the HR determines the outcome of plant–bacteria interactions (Alfano and Collmer, 2004). With respect to host plant specificity of elicitation, type III effectors are either specific or general elicitors of the HR (Nürnberger et al., 2004). Avirulence (Avr) proteins govern cultivar specificity of the plant–bacteria interactions through a direct or indirect contact with the corresponding resistance (R) gene-encoded protein inside the plant cell (Bogdanove and Martin, 2000; Shao et al., 2003). General elicitors, in contrast, like the harpins from Erwinia species and P. syringae pathovars (Wei et al., 1992; He et al., 1993), trigger the HR in a cultivar-independent manner when applied to the apoplastic spaces of plant leaves. The role of harpins for the plant–bacteria interaction is still unclear. They are secreted through a pilus and accumulate along the length of the growing structure (Li et al., 2002), indicating that they may assist pilus penetration through the cell wall matrix. The HrpZ harpin from P. syringae pv. phaseolicola can also form stable associates with lipid bilayers in vitro, suggesting a role of HrpZ in molecule translocation through a plant plasma membrane pore (Lee et al., 2001a).
Ralstonia (formerly Pseudomonas) solanacearum is a major bacterial phytopathogen, causing lethal wilting in more than 200 plant species worldwide (Denny, 2000). Its TTSS externalizes the Pseudomonas out proteins PopA (Arlat et al., 1994), PopB and C, PopF1 and F2, and PopP1 and PopP2. The proteins are secreted into the medium by cultured bacteria, but studies of PopP2 (Deslandes et al., 2003), PopB and PopC (Guéneron et al., 2000) suggest that these effectors are active inside the host cell during the plant–pathogen interaction. PopA induces the HR upon infiltration into tobacco leaves, but it is not known in which compartment of the plant cell its activity is localized. This general elicitor shares no sequence similarities with proteins and peptides for which the sequence has been identified. However, PopA is frequently considered as a harpin-like protein, because it induces the HR when applied to the apoplast, because it is abundantly secreted, and because it is glycine-rich and devoid of cysteine (Alfano and Collmer, 2004). PopA is composed of 344 amino acids, and N-terminal processing of this ‘full-length’ peptide, PopA1, leads to the generation of PopA2 (335 aa), and PopA3 (251 aa) (Arlat et al., 1994). Studies aiming at determining the site of PopA activity in planta, either during interaction with the bacterium or after leaf infiltration of the pure protein, have so far been unsuccessful (De Rycke and Van Gijsegem, personal communication).
In a previous work, we described the generation of transgenic tobacco plants expressing the popA gene under the control of the pathogen-inducible hsr203J promoter (Belbahri et al., 2001). The PopA proteins produced in the tobacco cells were designed either to be confined to the cytosol or to be secreted. The transgenic plants therefore provide excellent tools for analysing the localization of HR induction by PopA in tobacco cells. Here, we show that the induction of the HR phenotype in these plants correlated with the presence of PopA at the plasma membrane. We describe the biophysical characteristics that allow the protein to interact with the membranes and discuss the implications for its role in the plant–bacteria interaction.
Phenotypes of transgenic tobacco plants expressing the popA gene
To determine the cellular localization of the HR-inducing activity of PopA, we designed two principal constructs for the transformation of tobacco. In the pPopIn construct, the region coding for the native PopA protein was directly fused to the inducible hsr203j promoter, which can be activated by inoculation with an oomycete plant pathogen that does not elicit the HR (Keller et al., 1999; Belbahri et al., 2001). Because of the absence of a signal sequence, the transgene was supposed to direct the accumulation of an intracellular, cytoplasmic protein. In the second construct, pPopEx, the PopA coding region was fused at the N-terminus to the sequence coding for the signal peptide of the extracellular tobacco PR-1a protein. As a control, the promoterless construct pPop5 was included in the transformation experiments (Fig. 1A). Transgenic tobacco plants carrying these constructs were screened for their ability to form lesions (the visible part of the HR) after induction of the hsr203J promoter by leaf-infiltration with zoospores of the oomycete pathogen, Phytophthora parasitica, as described previously (Belbahri et al., 2001).
Spore infiltration led to the development of mycelium and to typical wilting symptoms in untransformed tobacco leaves. These symptoms were also observed in transgenic plants that had been transformed with the promoterless control construct, pPop5 (Fig. 1A). In contrast, transformants expressing the popA gene under the control of the inducible promoter displayed localized necrosis after zoospore infiltration. This necrosis restricted further growth of the oomycete pathogen (Fig. 1A). About 50% of the lines screened displayed HR symptoms following inoculation with the oomycete. This phenotype occurred regardless of whether the plants were transformed with the pPopIn or the pPopEx construct (Fig. 1B).
PopA localizes to the plasma membrane
To determine the subcellular location of PopA in the transgenic plants, we analysed extracellular fluid samples and soluble intracellular protein extracts collected from leaf tissues after P. parasitica-infection. In addition, we prepared microsomal preparations from these tissues, which were then enriched for plasma membranes by aqueous two-phase partitioning. Immunoblot analyses of these preparations revealed that the elicitor was present in the plasma membrane fraction only. PopA was not detected in the extracellular fluids or in the soluble protein cytoplasmic extracts (Fig. 2A). This localization to the plasma membrane was detected in preparations from plants transformed with the construct designed to target the PopA protein to the intracellular compartment (line PopIn.1C1) and in those from plants expressing the protein that was designed to be secreted (line PopEx.20C1).
To further analyse the subcellular localization, ultra-thin sections of transgenic tobacco leaf tissues were immunogold labelled using the PopA-specific antibody and then analysed by electron microscopy. Immediately after oomycete inoculation (time point 0), sections from leaves of transgenic plants immunolabelled with the PopA-specific antibody did not show any localization of gold particles over the still intact cell structures (Fig. 2B). Gold particles were present in the plasma membrane and tonoplast in sections prepared from material collected 24 h after Phytophthora-infection (Fig. 2C). At later time points, PopA was detected on other endomembranes, in particular the thylakoid membranes in the grana of disintegrating chloroplasts (Fig. 2D and E). The localization patterns found in tissues prepared from leaves of the tobacco line PopIn.14A2 (transformed with pPopIn; Fig. 2D) and those from the line PopEx.20C1 (transformed with pPopEx; Fig. 2E) were similar. Gold particles were barely detectable on thin sections prepared from tissues collected at time points earlier than 24 h post inoculation.
PopA binds to membrane lipids
The localization of PopA to the plasma membrane in transgenic tobacco and the association of the elicitor with the endomembranes suggest that the protein has a high affinity for lipid bilayers and that this affinity is independent on membrane type or orientation. Thus, we hypothesized that the PopA binding was dependent on the lipid composition of the membranes, rather than the protein components. We used a monolayer enzyme-linked immunosorbent assay to determine whether PopA has an affinity for lipids (Ghosh et al., 1996).
We found that PopA had a strong, saturable affinity for asolectin, a lipid preparation from soybean that is enriched for phosphatidylcholine (Fig. 3A). We determined the apparent binding constant (KD) of PopA for asolectin from Scatchard-plots and obtained a value of 43 nM. We then investigated the binding affinity of PopA for phospholipids (Fig. 3B), sterols (Fig. 3C), and sphingolipids (Fig. 3D). PopA was able to bind all the lipids tested, but the affinity was dependent on lipid type and was higher for sterols and sphingolipids, than for phospholipids. The KD values for cholesterol (Fig. 3C) and stigmasterol were estimated as 1 nM and 24 nM respectively. The KD values for the sphingolipids, d-erythro-dihydrosphingosine (Fig. 3D) and sphingomyelin, were 15 nM and 19 nM respectively. For the phospholipids, PopA affinity was highest for phosphatidylcholine (Fig. 3B; KD 48 nM), moderate for phosphatidylinositol (KD 135 nM), and relatively low for phosphatidylserine (KD 238 nM) and phosphatidylethanolamine (KD 322 nM). Because cleaved PopA3 is not recognized by our antibody under non-denaturing conditions (data not shown), the determined affinities have to be considered as accounting for the full-length protein.
In vitro association of PopA with lipid bilayers
Hydropathy plots according to the Eisenberg algorithm (Eisenberg et al., 1984) revealed the amphipathic nature of PopA (Fig. 4A). The protein harbours a central hydrophobic region flanked by two hydrophilic domains. At the N-terminus, PopA has a small, but highly hydrophobic region. During processing of the full-length protein, this domain is cleaved off, thus generating PopA3 with hydrophilic terminal regions (Fig. 4A).
To investigate whether PopA was able to interact with lipid bilayers, liposomes prepared from soybean asolectin were incubated with the elicitor. Substantial amounts of PopA were found associated with precipitated large unilamellar vesicles (LUVs). In contrast, control proteins, like cryptogein, the HR elicitor from the oomycete Phytophthora cryptogea (Keller et al., 1996), and cytochrome c, a peripheral protein from mitochondrial membranes were found exclusively in the supernatant (data not shown).
Porous silica beads coated with single phospholipid bilayers were used to further investigate this interaction. These beads (TRANSILs) are excellent tools for analysing protein binding to membranes (Schmitz et al., 1999). The lipid molecules on these beads are separated from the support by an ultrathin layer of water molecules, rather than being linked by covalent bonds. This arrangement confers characteristics on the beads that are similar to those of biological membranes. In addition, the beads are stable and the lipid and aqueous phases can be easily separated by short centrifugation steps. TRANSILs coated with the following phospholipids were used in this study: (i) PC (1,2-diacyl-sn-glycero-3-phosphocholine), a neutral bilayer made up of phosphocholine esterified with various fatty acids; (ii) a mixture of POPC (palmitoyl-oleoyl-phosphatidylcholine 80%) and POPE (palmitoyl-oleoyl-phosphatidylethanolamine 20%), an essential plant membrane phospholipid; and (iii) a mixture of DEPC (dielaidoyl-phosphatidylcholine 98%) and of DMPG (dimyristoyl-phosphatidylglycerol 2%), which introduces a negatively charged lipid into the bilayer.
Binding capacity was assessed following incubation of PopA and the control proteins with phospholipid-coated TRANSILs and phase separation. Cytochrome c, cryptogein, and BSA were consistently found in the supernatant after incubation with the POPC/POPE-coated beads (Fig. 4B), demonstrating that no stable interaction occurred between these proteins and the membrane bilayers. In contrast, when PopA was incubated with the beads, the majority of the protein coprecipitated with the TRANSILs (Fig. 4B). This coprecipitation was observed with both the full length and PopA3 cleavage product and was detected for silica beads coated with bilayers consisting of PC, POPC/POPE and DEPC/DMPG (data not shown). Thus, the interaction between PopA and the membrane bilayers was independent of lipid composition and net charge.
When increasing concentrations of PopA were incubated with the TRANSILs, the beads were found to be completely saturated at PopA concentrations of 3 µM (Fig. 4C). Interestingly, PopA1 appeared to have a stronger affinity for the beads than PopA3, for which a higher proportion was found in the soluble fractions. The addition of Ca2+ to the assay did not increase the binding capacities of the beads, but shifted the partitioning between the aqueous and lipid phases and favoured binding of PopA3 to the membrane bilayer (Fig. 4C).
From these experiments the questions arose whether PopA3 is not rather binding to PopA1 than to the membrane, and whether PopA associates as a peripheral, or an integral membrane protein. To answer both questions, full-length PopA1 and the 251 aa PopA3 protein were produced in Pichia pastoris, and we performed experiments incubating PopA1 and PopA3 individually or together with the TRANSILs. The beads were then washed in 1.5 M NaCl, 6 M Urea at pH 11 to disrupt possible peripheral binding. After centrifugation, the beads and the NaCl/Urea supernatant were analysed by immunoblotting. The results from this experiment showed that PopA3 is able to associate to membranes by its own and does not require PopA1 (Fig. 4D). Furthermore, PopA1 as well as PopA3 are almost exclusively found in the insoluble fraction, indicating that both are integral membrane proteins (Fig. 4D).
Membrane association requires Ca2+
Lipid binding proteins frequently require Ca2+ (Creutz et al., 1998; Gerke and Moss, 2002). To determine whether PopA behaves similar, we subjected the elicitor and other proteins to a 45Ca2+ overlay assay. In this assay, both PopA and the Ca2+-binding control protein calmodulin (CaM) were labelled by the isotope. None of the other standard proteins analysed bound 45Ca2+ (Fig. 5A). Autoradiography revealed furthermore that the 45Ca2+-binding capacities of PopA1 and PopA3 were similar, indicating that the N-terminal part of PopA1 that is cleaved to generate PopA3 is not involved in Ca2+-binding.
The amount of PopA bound to the lipids was higher for assays conducted in the presence of 2 mM Ca2+ ions than for those conducted in buffer only. The binding activity of PopA was completely abolished when the bivalent ion was chelated by the addition of EGTA to the assay mixture (Fig. 5B). Thus, calcium influences both lipid binding and the integration of PopA into the membrane.
To investigate this further, we analysed PopA1 and PopA3 oligomerization in buffer only and after the addition of Ca2+. When PopA1 was left at room temperature and then submitted to SDS-PAGE, stable protein dimers and trimers were detected by Western blotting. PopA3 dimers were barely detectable under these conditions and only the PopA1 oligomers were observed in a mixture containing equimolar concentrations of PopA1 and PopA3 (Fig. 5C, left). When Ca2+ was added to the assay a different band distribution was obtained. The banding pattern indicated that the presence of calcium promoted PopA1 oligomerization and led to the formation of SDS-stable PopA3 homodimers. Furthermore, Ca2+ favoured dimerization between PopA1 and PopA3 and its addition led to the appearance of immunoreactive bands representing high molecular weight oligomers (Fig. 5C, right).
PopA forms ion-conducting pores in artificial and biological membranes
Several type III effector proteins have been shown to form ion-conducting pores in lipid bilayers in vitro (Tardy et al., 1999; Dacheux et al., 2001; Marenne et al., 2003; Schoehn et al., 2003). To analyse whether membrane integration of PopA leads to pore formation, we prepared small unilamellar vesicles (SUVs) containing a 20 mM solution of carboxyfluorescein. At this concentration, fluorescence of the self-quenching dye is low. When the SUVs were incubated with either BSA, cytochrome c, or the cryptogein elicitor, no increase of fluorescence above the level for untreated vesicles was observed, indicating that these proteins do not affect the integrity of the liposomes (Fig. 6A). In contrast, incubation with PopA led to the release and dequenching of carboxyfluorescein resulting in an increase in fluorescence over a period of 20 min. The efflux then reached a steady state level with a slope of fluorescence increase similar to the one observed in untreated SUVs (Fig. 6A). Therefore, these results indicated that PopA was able to form pores in liposome bilayers, that pore opening was reversible, and that the lifetime of pore opening (in the order of minutes) was rather long.
To confirm this result, we analysed the pore formation in commercially available liposomes that sealed the Na+-sensitive fluorochrome Sodium GreenTM. When a pore-forming protein is added to these liposomes in the presence of NaCl, the influx of Na+ leads to an increase of fluorescence. Control experiments were performed with the HrpZ protein from P. syringae pv phaseolicola, for which a channel-forming activity has previously been shown (Lee et al., 2001a), and with BSA, which has no such activity. When PopA was added to the liposomes in the presence of NaCl, the increase in fluorescence showed that PopA was able to generate Na+-conducting pores in the pure lipid bilayers, similar to HrpZ (Fig. 6B).
To analyse whether PopA was also able to form ion-conducting pores in biological membranes, we performed two-electrode voltage-clamp experiments with oocytes of Xenopus laevis in the presence or absence of the elicitor. This system has been used previously for heterologous expression of membrane transporters (Soreq and Seidman, 1992) and to analyse the changes in membrane permeability induced by pore-forming molecules (Goudet et al., 1998; Charnet et al., 2003). We found that the steady state current recorded between −110 and + 30 mV was low in the absence of the elicitor. When PopA was added to the bathing solution, changes in membrane potential generated exogenous macroscopic currents (Fig. 6C). PopA1 and PopA3 induced similar conductivities, but the currents were not detected in oocytes that had been exposed to cryptogein (data not shown). The steady-state current-voltage plots revealed a strong outward-rectification with a threshold potential of about −80 mV. The currents had a reversal potential of about −10 mV. Under the experimental conditions used, it was not possible to determine the ionic selectivity of PopA. The macroscopic currents detected did not resemble the endogenous currents of Xenopus oocytes described by Weber (1999), strongly suggesting that the pores were induced by PopA. After washout of the added protein, the steady-state currents recovered to initial values, suggesting that PopA forms an ion channel with well-defined biophysical properties. Taken together, our data show that PopA is able to form ion-conducting pores not only in artificial lipid bilayers, but also in biological membranes.
Inactivation of the popA gene does not influence the virulence of R. solanacearum or its capacity to induce an HR (Arlat et al., 1994). Similar observations have been made for the hrpZ gene from P. syringae pv. tomato (Charkowski et al., 1998). However, in the absence of a mutant phenotype, approaches other than those relying on genetic tools have to be employed to determine the role of PopA in the plant–bacteria interactions.
Sequence analysis of PopA using different computer programs (Cuff et al., 1998; Combet et al., 2000; Pollastri et al., 2002) suggested that 108 amino acids of the full-length protein are organized in 5 α-helices. The rest of the protein appears to be randomly coiled. Tertiary structure predictions could not be made, because the degree of similarity between PopA and proteins of known 3D structure was too low to allow modelling by the SWISS algorithm (Schwede et al., 2003), and similarity searches did not reveal any significant homology between PopA and known proteins. However, computer programs predicted the presence of a transmembrane helix of 24 amino acids between positions 169 and 192 (Von Heijne, 1992; Hofmann and Stoffel, 1993). This prediction gave the first indication that PopA might associate with membranes. However, it was surprising to find that transgenic plants expressing PopA with and without the extracellular targeting signal displayed similar HR phenotypes and that the appearance of the HR correlated with the localization of PopA to the plasma membrane. Even if membrane localization of PopA in the pPopEx transformants might be explained by stacked transit to the ER lumen, no such transit should occur in the absence of the PR-1a leader peptide in the plants harbouring the pPopIn construct. The localization of the protein is thus rather independent of the protein translation machinery. The high affinity of PopA for lipids and Ca2+ more likely indicates that the biophysical properties of the protein allow it to spontaneously associate with the membrane.
The annexins and the copines are known Ca2+-dependent lipid-binding proteins, which are found in both animals and plants. In animals, annexins are involved in essential cellular processes such as linking membranes to the cytoskeleton, membrane trafficking, intracellular signal transduction and DNA replication (Gerke and Moss, 2002). In Medicago truncatula, these proteins are involved in the early stages of nodulation signalling in response to interaction with Rhizobium meliloti (De Carvalho-Niebel et al. 2002). Copines are thought to mediate resistance to bacterial and oomycete pathogens by regulating the cell death response (Jambunathan and McNellis, 2003). PopA contains neither the Ca2+-binding site typical of the annexins (the annexin-fold) nor the C2 domains present in the copines and other Ca2+-binding proteins. The lack of known Ca2+-binding motif suggests that PopA binds the bivalent ion through an unknown mechanism.
The lipid-binding activity of PopA was highest for cholesterol and the sphingolipid, d-erythro-dihydrosphingosine. These findings suggest that PopA preferentially binds to membrane regions that are enriched in these lipids. Cholesterol- and sphingolipid-rich microdomains in membranes (so-called lipid rafts) are major regulators of diverse cellular processes in animals (Simons and Ikonen, 1997; Brown and London, 2000). These domains act as concentration and partitioning platforms for receptors and signal transduction molecules that link the plasma membrane, the cytoskeleton and intracellular signalling pathways (Oliferenko et al., 1999; Simons and Toomre, 2000). The lipid rafts of animal cells are also exploited by pathogens to infect the host. The toxin VacA from Helicobacter pylori binds as a monomer to lipid rafts in HeLa cell membranes. The microdomains enable VacA to concentrate locally and to oligomerize efficiently (Schraw et al., 2002). In plants, microdomains have so far only be isolated from tobacco cells (Peskan et al., 2000; Mongrand et al., 2004) and little is known about their function.
We found that both full-length PopA1 and the processed PopA3 product were able to associate with lipid bilayers in a Ca2+-dependent manner. In addition, Ca2+ promoted PopA self association and the interaction between PopA1 and PopA3. PopA contains 25 partially overlapping GxxxG motifs, among which seven are in the putative transmembrane domain of the protein. These motifs provide a framework for the helix–helix interactions of membrane proteins (Russ and Engelman, 2000; Li et al., 2004). The 93 amino acids cleaved from PopA1 during processing contain only one GxxxG motif, suggesting that this N-terminal region plays only a minor role in protein association. The N-terminus of PopA1 does not seem to contribute to any of the biophysical characteristics of PopA described in our study. In addition, PopA1 and PopA3 induced the transcription of the same set of defence-related genes, indicating that the N-terminal region is not required for the HR-inducing activity of PopA in tobacco plants (Supplementary material, Fig. S1). In this context it is noteworthy that the 9-LOX-dependent peroxidative pathway is not activated by PopA, although it is one of the features of cryptogein-induced hypersensitive cell death (Rusterucci et al., 1999). This means that cryptogein and PopA trigger different, but complementary signalling pathways leading to the HR.
N-terminal cleavage of PopA occurs in cultures of R. solanacearum as well as in plants, leading to the suggestion that processing of the protein is an autoproteolytic event (Belbahri et al., 2001). AvrPphB, a type III effector from P. syringae pv. phaseolicola, has been reported to require autoproteolytic cleavage before it can induce the HR (Shao et al., 2002). However, in vitro experiments showed that PopA1 processing is not mediated by autoproteolysis but that cleavage required a proteinase inhibitor-sensitive protease activity that was present in plant plasma membranes (Supplementary material, Fig. S2). Similarly, the P. syringae pv. tomato type III effector, AvrRpt2, requires a plant factor to cleave the N-terminus of the full-length 32 kDa protein (Jin et al., 2003). But in contrast to PopA, cleavage of AvrRpt2 is protease-insensitive and reportedly occurs in the plant cell cytosol, where the protein has an avirulence function (Mudgett and Staskawicz, 1999; Coaker et al., 2005).
One of the major findings of our work was that PopA was able to form pores in artificial and biological membranes. This activity was reversible, allowed the conduction of sodium ions, and the pore openings created were wide enough to allow the 370 Da fluorophor carboxyfluorescein to cross the membrane. The ability to reversibly form ion-conducting pores in membranes has been reported for bacterial toxins, like the above mentioned VacA protein from H. pylori. VacA associates to the plasma membrane from human cells, forms a pore, and then becomes internalized by the cell where it binds to endomembranes (Cover and Blanke, 2005). Pore formation in artificial membranes has also been reported for HrpF, a protein from Xanthomonas campestris pv. vesicatoria (Büttner et al., 2002), and HrpZ, a harpin from P. syringae pv phaseolicola (Lee et al., 2001a). Both of these proteins are secreted by plant pathogen TTSSs. HrpF has no functional or structural similarities to PopA. In contrast, HrpZ shares several characteristics with PopA; it can be isolated from bacterial supernatants and it can induce an HR when applied exogenously to tobacco leaves. However, HrpZ and PopA are not similar in sequence or predicted structure and they probably induce the HR by different mechanisms. HrpZ is recognized by receptors on the surface of tobacco cells (Lee et al., 2001b). Thus, when the protein is expressed in planta, it has to be targeted to the apoplast to trigger the HR (Tampakaki and Panopoulos, 2000). In contrast to HrpZ, PopA does not require extracellular targeting to induce the HR in transgenic tobacco. We showed that the HR induction in PopA transgenics occurred when the protein was expressed either intra-, or extracellularly, and that the appearance of the HR correlated with PopA localization to the plasma membrane. It is thus most likely that PopA can attach the plasma membrane from the inside, as well as from the outside of the cell, indicating that PopA exerts its elicitor activity when anchored to the lipid bilayer. The presence of PopA in the plasma membrane may induce the plant cell signalling cascade(s) that activate defence responses and therefore trigger the HR. One early event in HR signalling is an influx of Ca2+ across the plasma membrane leading to an increase in cytosolic Ca2+ levels (Nürnberger and Scheel, 2001). Membrane-anchored PopA might interfere with Ca2+ signalling in the plant cell by binding the bivalent ion. Alternatively, the pores generated by PopA might provoke a Ca2+ influx that mimics the signalling event. A similar mechanism has already been proposed for certain annexins (Rojas et al., 1990). Another possibility is that PopA interacts directly with proteins involved in the cell signalling events leading to the HR. Such proteins may be the heterotrimeric G-protein subunits or the NADPH oxidase, which are associated with microdomains in tobacco cell plasma membranes (Peskan et al., 2000; Mongrand et al., 2004).
PopA secretion by R. solanacearum requires hrp-dependent pili. These pili reportedly pierce the cell wall during the interaction of the bacterium with the plant (Van Gijsegem et al., 2000). The results obtained in our study combined with those from previous investigations lead us to conclude that the protein is delivered directly to the host cell plasma membrane by the bacterium, where it binds lipids and calcium, oligomerizes and integrates to form a pore. We observed that PopA was localized to the endomembranes of transgenic tobacco cells in tissues sampled at the later time points of the HR. This localization was most likely a consequence of cytoplasm collapse rather than a cause of cell death. The membrane staples of the grana appeared to be the most stable structures in cells undergoing an HR. These structures retained their integrity even after other subcellular structures had been degraded. Thus, PopA may be liberated from disintegrating plasma membranes, or be translocated to the endomembranes by membrane vesicles. Invaginations and the formation of cytoplasmic vesicles from the plasma membrane have already been reported for tobacco cells undergoing an HR (Mittler et al., 1997).
The fact that popA can be deleted without affecting the pathogenicity of the bacterium indicates that the protein is not a component of the apparatus that translocates virulence factors. However, it is also possible that the translocon is composed of several proteins. These proteins may form homomeric structures like PopA that are sufficient to form the pore, but may also form heteromeric complexes to generate the translocation system. Similar mechanisms of membrane channel formation have been reported for YopB and YopD from Yersinia enterocolitica (Tardy et al., 1999) and PopB and PopD from Pseudomonas aeruginosa (Schoehn et al., 2003). This complementation system may also explain why deletion of PopA does not affect the pathogenicity of the bacterium. However, the suggestion that the pore formed by PopA serves simply to release nutrients from the host cells to feed the bacterium during infection cannot be excluded. Further studies are necessary to identify the proteins interacting with PopA. The interactants of PopA might be bacterial proteins able to form heteromers with the harpin-like elicitor and generate pores, or plant proteins capable of interfering with the HR signalling cascade. Recently, a protein binding site has been identified in HrpZ, which is supposed to allow the interaction between the harpin and a host plant protein (Li et al., 2005).
Generation and growth of transgenic plants
Growth of tobacco, cultivation of P. parasitica, and production of fungal zoospores were performed as described previously (Keller et al., 1996). The pPopIn and pPopEx constructs used for the transformation of Nicotiana tabacum cv Bottom special were the plasmids pPop12.1.2–101 and pPop126.96.36.199.1–101 respectively (Belbahri et al., 2001). To generate the construct pPop5, the insert from HindIII-SacI digested pPop6.1KS (Belbahri et al., 2001) was used to replace the uidA reporter gene between the HindIII and SacI restriction sites of the binary vector pBI101.2 thus generating pPop12.1–101.7.3. Transformation of tobacco, the selection of transformants, and the screening of the plants for a necrosis phenotype were performed as described (Keller et al., 1999).
Leaf tissue fragments were fixed, embedded, thin-sectioned and immunogold-labelled as described (Delannoy et al., 2003), except that the nickel grids were incubated for 3 h with a 1:50 dilution of Protein A-Sepharose CL-4B-enriched anti-PopA antibody. Incubation with a 1:20 dilution of the Gold Conjugate goat anti-rabbit IgG, labelled with 15 nm colloidal particles (EM GAT-15; BioCell Research Laboratory, Cardiff, UK), was performed for 1 h after washing. Grids were rinsed, stained for 20 min with a saturated solution of uranyl acetate, and incubated for 1 min with 0.3 M lead citrate, 20 mM NaOH. Controls used for immunolabelling included samples from wild-type lines at different time points post inoculation, samples from inoculated transgenic lines at time point 0, and samples from transgenic plants incubated without the primary antibody. Under those conditions, no labelling or only occasional non-specific staining to the cell wall has been observed.
Preparation of fluids, extracts, and of the plant plasma membrane
Intercellular washing fluids were prepared from P. parasitica-inoculated tobacco leaf areas, 72 h after infiltration with zoospores (1000 spores ml−1), according to Hugot et al., (2002). The solid residues were homogenized as described (Vauthrin et al., 1999), and the homogenate was submitted to a centrifugation at 100 000 g for 35 min in order to pellet the microsomal membranes. The supernatant containing the intracellular, soluble proteins was stored for analyses. Plasma membranes were prepared from the microsomal fraction by two-phase partitioning, as described (Vauthrin et al., 1999). For comparison, total proteins were extracted by grinding leaf tissues in the presence of 50 mM Na-phosphate at pH 7.0, containing 0.5%[w/v] SDS. The homogenate was subjected to centrifugation at 26 000 g, and the supernatant was stored for analyses. Samples containing 50 µg of protein were submitted to SDS-PAGE and transferred to nitrocellulose membranes. Labelling with antisera was revealed by chemiluminescence (ECL Western blotting protocols, Amersham, Buckinghamshire, UK).
Synthesis and purification of PopA
Native PopA, representing a mixture of the full-length protein and its processing products, was purified from concentrated culture supernatants of R. solanacearum strain GMI1000 (Arlat et al., 1994). PopA1 and PopA3 were synthesized using the P. pastoris expression system (Invitrogen). The popA1 coding region was generated by polymerase chain reaction (PCR) with the primers popA1.5 (5′-GTATCTCTC GAGAAAAGATCAGTCGGAAACATCCA-3′) and popA1.3 (5′-ATAATACCTAGGTTACATCGGCTGCGTCGAG-3′), using plasmid pGMI1911 (Arlat et al., 1994) as the template. The same template was used to prepare popA3 with the primers popA3.5 (5′-GTATCTCTCGAGAAAAGACCGCAGTCGGC CAACAAGAC-3′) and popA1.3, generating a fragment beginning 280 bp downstream of the popA1 start codon. The PCR products were ligated into XhoI/AvrII-digested pPIC9, to create the plasmids pPIC9-PopA1, and pPIC9-PopA3. Both plasmids were multiplied in Escherichia coli strain DH10B, linearized by SacI-digestion, and transformed into P. pastoris strain GS115, according to the supplier's protocol (Invitrogen). Positives were selected by Western blotting with the anti-PopA antibody, and transferred to a synthetic growth medium, which was supplemented every 24 h with 0.5%[v/v] methanol and incubated for 3 days at 28°C. After pelleting the yeast, the supernatants were concentrated 5 times using a vacuum rotary evaporator, dialysed against 25 mM K-phosphate, pH 6.5, containing 5%[v/v] acetonitrile, and applied to an anion exchange column, packed with Macro-Prep Sephadex HighQ (Bio-Rad, Marnes-la-Coquette, France), and equilibrated with the same buffer. Proteins were eluted by applying a discontinuous gradient of increasing KCl concentrations, eluting PopA1 and PopA3 at 30 mM and 60 mM KCl respectively. After dialysis against water, the protein fractions were equilibrated to 35%[v/v] acetonitrile in water with 0.1%[v/v] trifluoracetic acid, and applied to a column packed with a VYDAC 214TP Series C4 polymeric reversed-phase (Vydac, Mojave, CA). Proteins were eluted by applying a discontinuous gradient of increasing acetonitrile concentrations, eluting PopA1 and PopA3 at 65%[v/v] and 50%[v/v] acetonitrile respectively. Acetonitrile was evaporated, the solutions were dialysed against distilled water, and the proteins, which were pure at this stage, were lyophilized.
Lipid binding assays
All lipids were purchased from Sigma (St. Quentin Fallavier, France), suspended in chloroform and diluted in a solution of 30%[v/v] chloroform and 70%[v/v] methanol. Loading of 100 µl of 3 µM lipid solutions to 96-well titre plates (MaxiSorp; NUNC, Wiesbaden, Germany), binding, and blocking were performed as described (Ghosh et al., 1996). PopA proteins were added in dilution buffer (0.3%[w/v] BSA in PBS) to the wells and incubated for 1 h. Bound PopA was stained with the anti-PopA antibody and with peroxydase-conjuguated protein G (Pierce, Rockford, IL). Peroxydase activity was revealed by adding 2.2 mM of o-phenylenediamine dihychloride (Sigma) to the wells, and the reaction was stopped after 20 min by adding 50 µl of 1 M HCl. The absorbance was measured at 492 nm on a Multiskan RC ELISA plate reader (Labsystems, Farnborough, UK), and bound PopA was quantified by reference to a calibration curve established by direct coating of PopA to ELISA plates. The lipid concentration used for coating was determined to result in 50% of maximum binding, when blotting the binding of PopA at 200 nM to increasing concentrations of asolectin. Specifically bound PopA was defined as total bound elicitor minus elicitor that bound in the absence of lipids. For quantitative estimates of association, the binding data were fit to Scatchard plots (Motulsky and Christopoulos, 2004). To determine the Ca2+ requirement of lipid binding, the buffer was supplemented with either 2 mM CaCl2, 10 mM EGTA, or both 2 mM CaCl2 and 10 mM EGTA.
Liposomes were prepared at 4°C by dissolving 50 mg of asolectin in 1 ml of chloroform in a hemolysis tube. After evaporation of the chloroform under a continuous nitrogen stream, the samples were vacuum-dried for additional 45 min. To the dried lipids, 1 ml of 5 mM HEPES buffer, pH 7.4, containing 20 mM of carboxyfluorescein (Molecular Probes, Leiden, the Netherlands) was added, and vesicles were formed by sonication (Huang, 1969). After a first centrifugation at 100 000 g for 30 min, the supernatant was submitted to centrifugation at 160 000 g for 3 h, dividing SUVs and LUVs to the supernatant and pellet respectively (Barenholz et al., 1977). SUVs were portioned and stored at −20°C under nitrogen atmosphere. LUVs were resuspended in the initial buffer to maintain osmolarity, portioned, and stored like the SUVs. Liposomes composed of POPC (90%) and DPPG (dipalmitoyl-glycero-phosphosphoglycerol 10%) harbouring the sodium-sensitive fluorochrome Sodium GreenTM tetra(tetramethylammonium) salt (cell impermeant) at 20 µM were purchased from Novosom AG (Halle/Saale, Germany). The analysis of the interaction of proteins with TRANSILs (Nimbus GmbH, Leipzig, Germany) was performed using beads of 10- to 12 µm in size with 400 nm-wide pores. TRANSILs were coated with PC, POPC/POPE, and DEPC/DMPC, and had lipid concentrations of 7.4 (± 0,4) mM, 8.3 (± 0,4) mM, and 5.4 (± 0,3) mM respectively. PopA, cryptogein, BSA, and cytochrome c in varying concentrations were incubated at room temperature for 2 h under continuous shaking with 5 µl of TRANSILs in 20 mM Tris-HCl buffer at pH 7.4. After centrifugation at 1500 g for 5 min, unbound protein in the supernatant was collected and the beads were washed twice with 20 mM Tris-HCl buffer at pH 7.4. Proteins partitioning to the lipid- or aqueous phase were analysed by immunoblot (PopA and cryptogein), or by Coomassie blue staining (cytochrome c and BSA). To determine the Ca2+ requirement of membrane association, the 20 mM Tris-HCl buffer was supplemented with 2 mM CaCl2, before PopA was added at varying concentrations. To estimate whether PopA associates as a peripheral, or an integral protein, precipitated TRANSILs previously incubated with 0.5 µM PopA were resuspended in a solution containing 1.5 M NaCl and 6 M urea at pH 11. After centrifugation at 1500 g for 5 min, the beads and the supernatant were analysed by immunoblotting, as described above.
Ca2+ overlay and oligomerization
Calmodulin (Sigma; 0,25 nmol), PopA proteins (0.15 nmol each), and standard proteins were submitted to SDS-PAGE and transferred to PDVF membranes. 45Ca2+ overlay assays were performed as described (Maruyama et al., 1984). For analysis of oligomerization, 0.015 nmol of either PopA1 and PopA3, or of both were incubated for 30 min at room temperature in 20 mM Tris buffer, pH 7.4 alone, or in buffer supplemented with 2 mM CaCl2.
Small unilamellar vesicles charged with carboxyfluorescein were passed through Sephadex G25 (PD-10 columns; Amersham, Buckinghamshire, UK) and recovered in 5 mM HEPES, pH 7.4, containing 100 mM NaCl. Samples of 100 µl were spotted onto 96-well microtitre plates and dequenching of the dye was determined in a Fluoroskan II fluorometer (Labsystems, Farnborough, UK) at excitation and emission wavelengths of 493 nm and 520 nm respectively. After 10 min incubation at room temperature, BSA, cytochrome c, cryptogein, or PopA were added at 2.5 µM, 10 µM, 12.5 µM, and 4 µM, respectively, and fluorescence was monitored over an additional 1-h period. Assays with liposomes harbouring the sodium-sensitive fluorochrome were performed in 96-well plates and in a total volume of 100 µl containing 10 mM HEPES buffer (pH 7.0), 5 µl of the liposomes, and HrpZ, PopA, or BSA at 2 µM. After the addition of 50 µl of NaCl (50 mM), fluorescence was determined with a Cytofluor II fluorescence plate reader at 485/530 nm (Biosearch, Bedford, MA). The fluorescence from control experiments without protein were substracted from the obtained values.
Oocytes from X. laevis were obtained and maintained as described (Véry et al., 1995). Whole-cell currents were recorded using the two-electrode voltage-clamp technique (Stühmer, 1992) on oocytes from 3 different Xenopus. The tested protein was applied directly to the bath at a final concentration of 12 µM. Washing out of the protein was done by gravity driven perfusion of the chamber with the same solution without protein. All experiments were performed at room temperature. Current-passing and voltage-recording electrodes were filled with 3 M KCl and had tip resistances of 0.8–1.2 mΩ. The voltage-clamp amplifier was an Axoclamp 2 A (Axon Instruments, Foster City, CA) interfaced with a Digidata 1322 interface (Axon Instruments). Voltage-clamp protocol application, data acquisition and data analysis were performed using pCLAMP software (Axon Instruments).
We would like to thank Dr M. Ponchet for many fruitful discussions and Mrs C. Etienne for maintaining the plants. We are grateful to Dr L. Suty for the 9-LOX cDNA, to Dr E. Galiana for the anti-RNAse-NE antibody, to Dr S. S. Miller for the anti-cMDH antibody, and to Dr R. Serrano for the anti-H+-ATPase antibody. We would like to thank Dr F. Van Gijsegem for the anti-PopA antibody and for communicating unpublished results. This work was supported by grants from INRA, the French Ministry for Education, Research, and Technology (PRFMMIP) and the Association pour la Recherche sur les Nicotianées.