Metal Hyperaccumulation Armors Plants against Disease

Metal hyperaccumulation, in which plants store exceptional concentrations of metals in their shoots, is an unusual trait whose evolutionary and ecological significance has prompted extensive debate. Hyperaccumulator plants are usually found on metalliferous soils, and it has been proposed that hyperaccumulation provides a defense against herbivores and pathogens, an idea termed the ‘elemental defense’ hypothesis. We have investigated this hypothesis using the crucifer Thlaspi caerulescens, a hyperaccumulator of zinc, nickel, and cadmium, and the bacterial pathogen Pseudomonas syringae pv. maculicola (Psm). Using leaf inoculation assays, we have shown that hyperaccumulation of any of the three metals inhibits growth of Psm in planta. Metal concentrations in the bulk leaf and in the apoplast, through which the pathogen invades the leaf, were shown to be sufficient to account for the defensive effect by comparison with in vitro dose–response curves. Further, mutants of Psm with increased and decreased zinc tolerance created by transposon insertion had either enhanced or reduced ability, respectively, to grow in high-zinc plants, indicating that the metal affects the pathogen directly. Finally, we have shown that bacteria naturally colonizing T. caerulescens leaves at the site of a former lead–zinc mine have high zinc tolerance compared with bacteria isolated from non-accumulating plants, suggesting local adaptation to high metal. These results demonstrate that the disease resistance observed in metal-exposed T. caerulescens can be attributed to a direct effect of metal hyperaccumulation, which may thus be functionally analogous to the resistance conferred by antimicrobial metabolites in non-accumulating plants.


Introduction
Metal hyperaccumulation is described as the accumulation of exceptionally high concentrations of metallic elements in the aerial parts of a plant [1,2]. The phenomenon has evolved in around 450 plant species, distributed across several families [2,3,4], and most hyperaccumulator species are endemic to metalliferous soils, either natural or anthropogenic in origin [5]. This unusual characteristic has attracted considerable interest, and a number of hypotheses have been proposed to explain the evolution of the hyperaccumulation phenotype [6,7]. The possibility that accumulated metal provides a defense against herbivores or pathogens, termed the 'elemental defense hypothesis', has received most attention and support [7,8,9].
A number of studies have reported findings consistent with a defensive effect of hyperaccumulated metals against herbivores [10,11,12], but the interpretation of these results remains controversial, as other studies have failed to support the defense hypothesis. For instance, Noret et al. [13,14] were unable to find a defensive effect attributable to metal hyperaccumulation in the field, despite that reported in laboratory trials. There is also some evidence that the importance of metal-based defense is dependent upon the mode of herbivory [15]. In the case of defense against pathogens, considerably fewer tests have been carried out [e.g. [16]] and only one study [17] has explicitly tested the defense hypothesis with regard to bacterial pathogens. So far, although some evidence has been provided that plants exposed to high metal concentrations have reduced susceptibility to various pathogens, no study has demonstrated that the metal itself was directly responsible for this effect. It is therefore not possible, at present, to state that metal hyperaccumulation trait evolved as an antimicrobial defense.
Here, we test the possibility of a direct elemental defense against bacterial pathogens in the crucifer Thlaspi caerulescens, a hyperaccumulator plant characteristically associated with metalliferous soils [18]. In its natural habitats, this species can hyperaccumulate three different metals: zinc, nickel, and cadmium [19,20], and it has been widely used in studies of the hyperaccumulation trait [21,22,23,24,25].
We have tested the elemental defense hypothesis via a threepronged approach by means of studies in vitro, in planta, and in the field. First, we have monitored the growth of Pseudomonas syringae pv. maculicola M4, a model pathogen of Arabidopsis thaliana [26], in T. caerulescens plants treated with different metal concentrations. This pathogen grows in the apoplastic spaces between plant cells, so we have investigated metal concentrations in the apoplastic phase specifically to determine whether they are likely to be inhibitory to pathogen growth. A number of studies attempting to determine the location of hyperaccumulated metals within the leaf have found the majority to be accumulated in the epidermal vacuoles [27,28,29], but apoplastic metal concentrations have not been examined in previous studies. Further, we have assessed the growth of Psm mutants with altered zinc sensitivity in T. caerulescens plants grown on different zinc regimes. This has allowed us to test whether zinc tolerance is important for bacterial growth in planta, as expected if metals are directly involved in plant defense. Finally, we have tested the zinc tolerance of endophytic bacteria found in the leaves of T. caerulescens growing on a metal-rich soil at the site of a former lead-zinc mine to determine whether there is evidence that this form of defense may be effective under natural field conditions.

Pseudomonas syringae pv. maculicola M4 is a pathogen of Thlaspi caerulescens
When Thlaspi caerulescens plants were grown in the glasshouse, it was observed that occasional outbreaks of powdery mildew affected only those plants growing on low metal treatments ( Figure 1A and B). This observation prompted us to consider the importance of metals in defending T. caerulescens against pathogens. As powdery mildews are obligate biotrophs and therefore unculturable, we sought a more tractable model system for further study of this phenomenon. Bacterial plant pathogens are suitable for such investigations because they are easily cultured in vitro and are relatively straightforward subjects for genetic manipulations such as mutagenesis.
Of sixteen plant pathogenic bacteria tested, eight were found to cause necrotic symptoms in T. caerulescens within 2 to 4 days after inoculation [see Supporting Information Table S1]. Pseudomonas syringae pv. maculicola M4 (Psm), a rifampicin-resistant derivative of Psm 4326 [26], caused necrosis more rapidly and to a greater extent than any other tested strain. To confirm that Psm is pathogenic on T. caerulescens, we assessed the ability of Psm, Psm ES4326 (a streptomycin resistant derivative of Psm 4326) [30] and two type III secretion system (T3SS) mutants of ES4326 (gift of K. Schreiber and D. Desveaux) to grow in T. caerulescens. Over 5 days post-inoculation, Psm M4 and Psm ES4326 both multiplied two to three logs in the leaves of T. caerulescens plants grown on minimal (0.04 mM) zinc. However, neither T3SS mutant was able to grow in T. caerulescens at any of the zinc concentrations tested ( Figure 2). Both T3SS mutants grew similarly to wild-type bacteria in vitro ( Figure S1). This confirms that Psm is pathogenic towards T. caerulescens and shows that the ability of Psm to grow in T. caerulescens is T3SS-dependent.
Hyperaccumulation of zinc, nickel, or cadmium inhibits the growth of Pseudomonas syringae pv. maculicola M4 in Thlaspi caerulescens On inoculation of Psm into T. caerulescens plants treated with progressively higher concentrations of zinc, both symptom development and Psm growth were significantly reduced as the concentration of the zinc treatment increased ( Figure 1C and D; Figure 2).
T. caerulescens is able to accumulate three metals, zinc, nickel and cadmium, in its shoots. To determine whether nickel and cadmium also conferred increased resistance to infection, bacterial growth was quantified in leaves of T. caerulescens plants grown on nutrient solution supplemented with different concentrations of zinc, nickel, or cadmium. For all three metals, increasing metal concentration resulted in reduced bacterial growth. Zinc treatments at concentrations $30 mM, and nickel and cadmium treatments at concentrations $10 mM, caused significant inhibition of bacterial growth at 2 and 5 days post-inoculation ( Figure 3). Accumulation of any of these metals therefore inhibits bacterial growth in planta and defends T. caerulescens against disease.
Studies of plants that are not normally exposed to high concentrations of metals have shown that metal treatment can induce a range of stress responses, including up-regulation of genes associated with local plant defense responses and systemic acquired resistance (SAR) such as PR-1 [31]. In order to determine whether

Author Summary
Soils rich in heavy metals support communities of distinctive metal-tolerant plants, a number of which exhibit a remarkable trait known as metal hyperaccumulation. These plants accumulate exceptionally high concentrations of metallic elements in their leaves, but whether this trait confers any adaptive advantage is controversial. In this study, we test the hypothesis that metal hyperaccumulation provides protection against disease. We demonstrate that Thlaspi caerulescens becomes resistant to bacterial leaf spot caused by Pseudomonas syringae pv. maculicola (Psm) when it accumulates zinc, nickel, or cadmium, and show that the metal concentrations in these plants are sufficient to account for their observed disease resistance. We also show that there is a close correlation between the zinc tolerance of different strains of Psm and their ability to colonize T. caerulescens leaves with high zinc content. In a field study, bacteria isolated from the leaves of T. caerulescens plants growing on the site of a former lead-zinc mine were found to possess a higher degree of metal tolerance than bacteria isolated from crop plants. Our findings show that metal hyperaccumulation in plants can act as a mechanism to prevent attack by pathogenic microorganisms, but that metal tolerant pathogens can overcome this defense.
some of the protection conferred by high zinc was a consequence of the effect of high zinc concentrations on defense-associated gene expression, we compared the expression of PR-1 in leaves of T. caerulescens plants grown on 0.04 and 300 mM zinc by quantitative real-time PCR. Normalized PR-1 expression in the two metal treatments was similar to PR-1 expression in healthy A. thaliana leaves (relative expression levels in 0.04 mM, 1.42360.299; 300 mM Zn, 1.94260.520; untreated A. thaliana, 2.10261.435), and there was no significant increase in expression in response to increased zinc ( Figure S2). This indicates that the inhibitory effect of increasing metal treatments on Psm growth is not due to upregulation of SAR.
Apoplast extracts from Thlaspi caerulescens plants grown on high metal concentrations show decreased ability to support bacterial growth Psm infects the apoplastic phase between plant cells. To test whether metal accumulation reduces the suitability of this environment for bacterial growth, Psm was cultivated in vitro in apoplastic fluid extracted from plants treated with supplementary metals. Apoplast extracts from zinc-and nickel-treated plants supported significantly less bacterial growth than those from plants grown without supplementary metal ( Figure 4A and B). Apoplast extracts from plants treated with higher cadmium concentrations also inhibited bacterial growth for up to 18 hours, although this inhibition was eventually released ( Figure 4C). As such, it is clear that metal hyperaccumulation by T. caerulescens makes the apoplast a more hostile environment for the growth of Psm.
Metal concentrations in Thlaspi caerulescens leaves and extracted apoplast are sufficient to explain the high disease resistance of metal-treated plants To determine whether hyperaccumulated metal could be responsible for the observed reductions in bacterial growth in leaves, the ability of Psm to tolerate each metal was tested in a range of synthetic media and in extracted apoplast ( Figure S3) and compared with the metal concentrations of apoplast extracts and whole-leaf tissue samples ( Figure 5; Table 1). For all metals tested, concentrations in whole-leaf tissue of T. caerulescens plants grown at the two highest treatments were sufficient (i.e. higher than the IC 50 Figure 2. T3SS mutants of Pseudomonas syringae pv. maculicola are unable to colonise Thlaspi caerulescens. P. syringae pv. maculicola M4 (Psm M4), P. syringae pv. maculicola ES4326 (Psm 4326) and two T3SS mutants of Psm (hrpS 2 and hrcN 2 ) were inoculated into fully expanded leaves of T. caerulescens at 10 6 cfu/ml. Three samples were taken for each strain, zinc treatment and time point, each consisting of three leaf discs pooled together. Values are means6SE (n = 3). The mean growth of the four strains at each zinc treatment over 5 days post-inoculation was compared in ANOVAs; growth of the four strains was found to differ significantly in all treatments except in the 300 mM zinc treatment, where no strain was able to grow (P,0.0005 for 0.04 and 10 mM Zn; P = 0.001 for 30 mM Zn; P = 0.169 for 300 mM Zn). Within each zinc treatment, Bonferroni simultaneous comparisons were used to determine which means differed significantly at 5 days post-inoculation, and these are marked with different letters. The inset shows symptoms observed in leaves from T. caerulescens plants grown on 0.04 mM zinc 72 hours after inoculation with wild-type Psm ES4326 and the Psm hrcN 2 mutant at 10 6 cfu/ml compared with uninoculated control leaves. doi:10.1371/journal.ppat.1001093.g002 values for the respective metals) to explain the observed reduction in bacterial growth in planta. Apoplastic zinc and nickel concentrations from the highest treatments were also sufficient to explain bacterial growth reduction. Apoplastic cadmium concentrations were considerably lower, reaching around onethird of the IC 50 concentration for Psm in vitro in higher treatments, perhaps explaining the ability of Psm eventually to overcome inhibition by cadmium.
Pseudomonas syringae pv. maculicola zinc tolerance mutants show differential growth in Thlaspi caerulescens The importance of metal tolerance for bacteria growing in metal-hyperaccumulating T. caerulescens was assessed by screening a transposon mutant library of Psm to identify mutants with   Figure 3 were used as a growth medium for P. syringae pv. maculicola M4. Six samples of 100 ml of apoplast were used for each treatment. Values are means 6 SE (n = 6). The experiment was performed three times with similar results, and the results shown are from one experiment representative of the three. Concentrations of all three metals were significant predictors of growth at 18 and 24 hours (ANOVAs: P,0.0005 in all cases except P = 0.015 for cadmium at 24 h). Bonferroni simultaneous comparisons (a = 1%) show that all zinc and nickel treatments resulted in significantly less growth than the control at 18 h and 24 h, as did all cadmium treatments up to 18 h. doi:10.1371/journal.ppat.1001093.g004 increased or decreased zinc tolerance relative to wild-type Psm. The performance of four representative mutants in planta was then compared to that of the wild-type strain under four zinc regimes. Two mutants with increased zinc tolerance (9A6 and 9A3) and two with reduced zinc tolerance (10C1 and 7C11) were used ( Figure 6). These four mutants grew similarly to wild-type Psm in A. thaliana ( Figure S4) and were able to cause similar symptoms to wild-type Psm in A. thaliana and pak choi, Brassica rapa ssp. chinensis (data not shown). Mutant 7C11 showed slightly reduced growth relative to wild-type Psm during late log phase and stationary phase in in vitro growth assays in KB broth ( Figure S5), but the in vitro growth kinetics of the other three mutants were not significantly different from wild-type Psm.
The disrupted genes were sequenced using an inverse PCR method and their predicted functions are described in Table 2. Interestingly, mutant 7C11 was found to contain an insertion in a TonB-dependent siderophore receptor (PSPTO_2152), which suggests that the slight growth defect observed in KB broth for this strain may be linked to impaired iron acquisition. The other mutation giving rise to reduced zinc tolerance, in mutant 10C1, was located in an NAD-dependent DNA ligase gene (PSPTO_0382); this does not have an obvious role in metal transport or metal tolerance, but is located close to two operons predicted to encode a heavy-metal-sensing two component regulatory system and components of a cobalt-zinc-cadmium (Czc) cation efflux system (PSPTO_0375 -PSPTO_0379) in the genome of P. syringae pv. tomato DC3000. The two mutations giving rise to increased zinc tolerance were located in pslF (PSPTO_3533; 9A3), a gene within the psl operon, involved in exopolysaccharide synthesis and biofilm formation in Pseudomonas aeruginosa [32], and in a proline iminopeptidase (pip) gene (PSPTO_5164; 9A6). The pip gene of Xanthomonas campestris pv. campestris has been shown to be induced during plant colonisation and to be essential for pathogenesis on cabbage [33]. However, our results indicate that the pip gene of Psm is not required for pathogenesis in A. thaliana. Pip belongs to a family of metalloproteases and its enzymatic activity may be dependent on a metal cofactor such as zinc or cobalt. In the genome of P. syringae pv. tomato DC3000, pip is located upstream of, and in a putative operon with, a predicted D-Tyr-tRNAtyr deacylase, which may have a role in protecting cells against D-tyrosine toxicity. In E. coli, zinc and D-tyrosine have been shown to have opposite effects on the phosphatase activity of the aromatic amino acid biosynthesis regulator TyrR, which is stimulated by zinc and suppressed by Ltyrosine and D-tyrosine [34].
The ability of the four Psm mutants to grow in T. caerulescens plants cultivated on 0.04, 10, 30, or 300 mM zinc was found to vary according to the zinc tolerance of the inoculated strain ( Figure 7). Thus, while the wild-type showed significant growth reduction with plant zinc treatments of 30 mM or higher, the two mutants with increased zinc tolerance, 9A6 and 9A3, were capable of growth in plants treated with 300 mM zinc. The two mutants with decreased zinc tolerance, 7C11 and 10C1, were unable to grow even in plants treated with 10 mM zinc, at which concentration the wild-type grew well.  Table 1. Summary of bacterial growth across treatments in relation to metal concentration and metal-dependent growth inhibition of P. syringae pv. maculicola in vitro.  Because increased zinc tolerance could be correlated with reduced zinc uptake, we tested the ability of all four mutants to grow at low zinc concentrations. Wild-type Psm grew equally well in M9 minimal medium and in M9 supplemented with 1 to 10 mM Zn. Of the four mutants, only 9A6 had an increased zinc requirement relative to wild-type bacteria, showing optimal growth at concentrations ranging from 1 to 5 mM zinc.

Naturally occurring endophytes of Thlaspi caerulescens show high zinc tolerance
To test further the hypothesis that high metal tolerance is a prerequisite for bacterial growth in T. caerulescens, endophytic bacteria were collected from leaves of a natural population of T. caerulescens plants at Hafna Mine, a former lead-zinc mine in North Wales, UK [35], and their zinc tolerance assessed. The mean zinc content of the leaves of these plants was 16.261.39 g Zn per kg leaf dry mass (6 s.e.m., n = 23 leaves), slightly greater (p,0.05) than that of plants treated with the highest concentration of zinc (300 mM) used in the laboratory experiment (13.760.71 g Zn per kg dry mass (6 s.e.m., n = 18 leaves)). The average IC 50 for zinc of these naturally occurring endophytes was 9.4 mM (n = 86) in KB medium. In contrast, a set of plant pathogenic bacteria isolated from non-hyperaccumulating plants were found to have a significantly (p,0.001) lower mean IC 50 for zinc of 5.4 mM in KB (n = 7; Figure 8). When the zinc IC 50 values for individual strains isolated from the Hafna mine plants were compared to the zinc IC 50 of Psm, the most zinc tolerant of the plant pathogens used in this study, 65% were found to have a significantly higher zinc tolerance (p#0.05), while only a single strain out of 85 had a significantly lower zinc tolerance.  Therefore, in the field, as in the laboratory, only metal-tolerant bacteria can colonize T. caerulescens.

Discussion
In this work, we have developed a model system to study the elemental defense hypothesis for plant metal hyperaccumulation using the bacterial pathogen Pseudomonas syringae pv. maculicola M4. We have shown that Psm displays T3SS-dependent pathogenesis in Thlaspi caerulescens plants grown in low metal concentrations and that metal hyperaccumulation by T. caerulescens at higher metal concentrations provides an effective defense against Psm. Further, we have demonstrated that zinc tolerance is essential for bacterial colonization of zinc-hyperaccumulating plants. Thus, our work is consistent with the hypothesis that hyperaccumulation benefits plants by increasing their resistance to pathogens.
Although studies have found that zinc, nickel and cadmium are mainly stored in the leaf cell vacuoles of Thlaspi species, particularly in epidermal cells [27,28,36,37,38], metals are transported into the leaf through the extracellular spaces of the apoplast by means of the transpiration stream. By measuring metal concentrations in the apoplastic fluid, we were able to provide an approximation of the conditions experienced by invading pathogens in the leaves of T. caerulescens. Comparison of the metal concentrations in planta with the IC 50 values measured for Psm in vitro in extracted apoplastic fluid allowed us to demonstrate that direct inhibition of bacterial growth by hyperaccumulated zinc or nickel is a realistic possibility. T. caerulescens, when grown on at least 10 mM Zn or 30 mM Ni, accumulated sufficient metal in apoplastic fluid to provide an elemental defense against Psm. This result is particularly striking considering that, when grown at 10 mM Zn, T. caerulescens accumulated only an average of 0.3 g Zn per kg dry mass, while the threshold for designation as a zinc hyperaccumulating plant is 10 000 mg per kg [1]. This indicates that T. caerulescens could be protected against pathogens by zinc accumulation even when growing on relatively low-zinc soils. However, although T. caerulescens accumulates sufficiently high metal concentrations to account for its metal-dependent resistance to Psm, we cannot rule out the possibility that additional metaldependent factors act in conjunction with metals to limit the growth of Psm in T. caerulescens. Accumulation of metals requires mechanisms by which the plant may tolerate elevated intracellular metal concentrations, which have not been fully elucidated, but which have been shown to involve redox-related compounds such as glutathione [39], enzymes such as superoxide dismutase [40], and metal-binding ligands such as organic acids and amino acids [27,41], as well as proteins (e.g. metallothioneins [42]). Such changes may affect the quality of the plant environment for growth of Psm or zinc tolerance in Psm, without having any expressly defensive function.
One case in which additional factors seem likely to contribute to metal-dependent defenses is in plants treated with cadmium. Cadmium was able to defend T. caerulescens against Psm in planta when plants were treated with $10 mM cadmium. Bacterial growth was also inhibited for around 18 hours in apoplastic fluid extracted from these plants. After this time, inhibition was released, possibly as a result of changes in bacterial gene expression or in the composition of the apoplast during this time in vitro, which may include alterations in cadmium bioavailability. However, the cadmium concentrations detected in apoplast extracts from cadmium-treated plants were lower than those required for inhibition of bacterial growth in apoplast extracts from plants grown in the absence of cadmium, so it is possible that cadmium concentrations are correlated with another defensive factor which may be unstable (such as ROS) or volatile.
To investigate further the possibility of a true elemental defense, we tested the importance of zinc tolerance for bacterial growth in planta in zinc-treated T. caerulescens. The results obtained using mutants of Psm provide the clearest evidence to date of a direct role for zinc as an elemental defense against pathogens in T. caerulescens. Mutants with reduced zinc tolerance were unable to multiply in plants grown at 10 mM zinc, in leaves of which their wild-type counterpart was successful; conversely, mutants with increased zinc tolerance grew well in leaves of plants grown at 30 mM or even 300 mM zinc, in which the wild-type could not survive. This clear link between bacterial zinc tolerance and ability to colonize plants hyperaccumulating zinc provides strong support for the concept of an elemental defense by zinc in T. caerulescens.
Finally, we have compared the zinc tolerance of strains used in this work with that of bacteria isolated from the leaves of T. caerulescens plants growing under natural conditions in a zincpolluted field site. We have shown that bacteria naturally colonizing these plants in the field had a range of zinc tolerances much higher than those of plant pathogenic strains isolated from non-accumulating crop plants, providing evidence for local adaptation of these endophytes to their environment [43,44,45]. This is in agreement with previous studies demonstrating that endophytic bacteria isolated from nickel hyperaccumulators exhibit high nickel tolerance [46,47]. Plant pathogens that have not been subject to this selection may find it difficult to grow and cause disease in T. caerulescens.
Metal-dependent resistance in T. caerulescens may be particularly effective against airborne, foliar pathogens such as P. syringae and powdery mildew, which may be deposited onto the surface of T. caerulescens leaves by rain and wind having previously colonized non-accumulating or metal-excluding plants, with no prior selection for metal tolerance. Thus, hyperaccumulated metals may be functionally equivalent to the diverse array of antimicrobial secondary metabolites used by plants to provide protection against infection. Only pathogens that are able to tolerate or inhibit the chemical defenses present in a specific plant species or genotype can grow in plant tissues. Similarly, in the case of metal-hyperaccumulation, only a small number of organismsthose possessing high metal tolerance -are able to grow in these plants. When the plants are deprived of this form of defense by cultivation on a low-metal growth medium, they may become vulnerable to a wider range of pathogens, explaining the spontaneous outbreaks of mildew infection observed on such plants (Figure 1).
When considering the role of metals in protecting plants against infection in a natural setting it is important to note that metal concentrations in the leaves of hyperaccumulating plants are typically orders of magnitude higher than metal concentrations in the environment, as illustrated in Figure 5. Therefore, even though the environment surrounding these plants may favour the growth of moderately metal tolerant microorganisms, many of these organisms may have insufficient metal tolerance to be able to grow in the tissues of hyperaccumulating plants. In addition, the observation that T3SS mutants of Psm were unable to infect T. caerulescens plants grown on low metal concentrations indicates that the plants possess additional defense mechanisms that act in conjunction with metal hyperaccumulation to protect plants, but which can be suppressed by the action of T3SS effectors. Thus, successful pathogens of T. caerulescens must not only be adapted for growth in a metal-rich environment, but must also possess at least some of the pathogenicity mechanisms known to be required for infection of non-accumulating plants such as Arabidopsis thaliana.
We have demonstrated that hyperaccumulation of any of three metals, zinc, nickel, or cadmium, by T. caerulescens provides the plant with an elemental defense against the hemibiotrophic pathogen Psm. The validity of the defense hypothesis has been challenged by some studies in which no evidence was found that metal hyperaccumulation defended T. caerulescens from herbivory in the field [13,14]. Our work, however, suggests that the defensive effect of metal hyperaccumulation against pathogens remains relevant in field conditions, with only metal-tolerant bacteria found growing naturally in the leaves of T. caerulescens. Further, we have shown that metal concentrations in the leaves are sufficient to account for this defensive effect without invoking any other factors. For all of the metals hyperaccumulated by T. caerulescens, we have shown that growth is also inhibited in apoplastic fluid, and that both zinc and nickel are found in this specific compartment at concentrations sufficient to account for the defensive effect. Moreover, we have demonstrated that the zinc tolerance of Psm mutants is correlated with their ability to colonize zinchyperaccumulating T. caerulescens plants. This result is mirrored by our findings concerning the zinc tolerance of natural endophytes of T. caerulescens from a zinc-rich field site. We therefore believe that metal hyperaccumulation by T. caerulescens can provide an effective form of defense against a wide range of pathogens.

Plant material
Seeds of Thlaspi caerulescens J. & C. Presl from Prayon, Belgium (provided by A.J.M. Baker and C. Lefèbvre) were cultured hydroponically on modified 0.1-strength Hoagland solution [20] in a glasshouse. The Prayon population of T. caerulescens was chosen as it is a widely studied, well characterized population showing typical hyperaccumulation behavior and producing relatively large amounts of biomass under the growth conditions described [20,22,24,25]. Natural radiation was supplemented by sodium-vapor lamps for 14 hours per day. Night temperature was maintained at 14uC and day temperature at a minimum of 24uC. Two-week-old plants were transferred to modified 0.1strength Hoagland solution containing 0.04, 10, 30, or 300 mM ZnSO 4 , or 0, 3, 10, or 30 mM NiSO 4 or CdSO 4 . The lowest zinc concentration used was 0.04 mM, rather than 0 mM for Ni and Cd. This is because zinc is an essential micronutrient without which the plants do not survive. The Hoagland solution used for all Ni and Cd assays contained 10 mM Zn for this reason; at 0.04 mM Zn, signs of zinc deficiency become apparent. T. caerulescens plants were grown on these metal treatments for a further 8 weeks, the nutrient solution being exchanged fortnightly for the first 6 weeks and weekly for the final 2 weeks. Arabidopsis thaliana (L.) Heynh. (Col-0) were sown on peat-based compost and grown in a glasshouse under the same conditions for 6 weeks.

Bacterial growth conditions, media and antibiotics
Bacterial strains used in pathogenicity assays are listed in Table  S1. Psm ES4326, Psm hrpN 2 and Psm hrpS 2 were provided by K. Schreiber and D. Desveaux. Psm hrpN 2 and Psm hrpS 2 were isolated from a transposon library of Psm ES4326 constructed using a kanamycin-resistant derivative of mini-Tn5 ( [48]; D. Desveaux, personal communication). All bacterial strains were streaked onto Luria-Bertani (LB) agar [49] from stocks kept in glycerol at 280uC and incubated at 28uC (37uC for Escherichia coli) for 24 to 48 hours prior to use. Single colonies were transferred to LB broth and incubated at 28uC with shaking for most applications. King's B (KB) agar [50], supplemented with CFC (Cetrimide-Fucidin-Cephalosporin; Oxoid) at half the manufacturer's recommended concentration, was used to selectively culture bacteria isolated from plant tissues for in planta growth studies. LB and KB broths were used in metal-tolerance experiments. For zinc-requirement assays, the minimal medium M9 [49] was used.

Bacterial growth in vitro
To compare the in vitro growth of wild-type and mutant strains of Psm, bacteria were grown overnight on LB agar and resuspended in KB broth to give an OD 600 of 0.1. One hundred microliters of this suspension were added to a further 100 ml of media in each of 16 wells of a 96-well microplate. Sixteen wells were inoculated with media alone as a media control. OD was measured every 20 min for the next 48 hours using an Infinite M200 plate reader (Tecan Group Ltd., Mä nnedorf, Switzerland).

Pathogenicity assays and bacterial growth in planta
For assays to examine the ability of bacteria to cause disease symptoms in T. caerulescens, bacteria were grown overnight on LB agar and re-suspended in sterile 10 mM MgCl 2 at an optical density at 600 nm (OD 600 ) of 0.35. Bacterial suspensions were infiltrated into at least three fully expanded leaves from 6-week-old T. caerulescens plants grown on nutrient solution (10 mM zinc) or 6week-old A. thaliana using a blunt 1 ml syringe. Symptoms resulting from bacterial inoculation were monitored over 7 days. For growth assays, bacteria were resuspended in sterile 10 mM MgCl 2 at an OD 600 of 0.2. This suspension was diluted 100-fold to give a suspension of approximately 10 6 cfu/ml and infiltrated into fully expanded T. caerulescens or A. thaliana leaves through the abaxial surface using a blunt 1 ml syringe. Nine leaves on each of six plants were inoculated within each metal treatment. Leaf discs of 10-mm diameter were taken from three of the inoculated leaves immediately, and from three further leaves at 2 and 5 days after inoculation. Leaf discs were homogenized in 10 mM MgCl 2 and the resulting suspension spread onto agar plates with a minimum of three technical replicates used for each sample. After incubation at 28uC for 48 hours, the number of bacterial colonies was counted and used to estimate the number of bacterial cells per unit area of leaf.

Apoplast extracts
Apoplastic fluid was extracted from T. caerulescens and A. thaliana leaves by a modification of the method described by Rico and Preston [51], using vacuum infiltration of the intercellular spaces with distilled water followed by centrifugation to extract apoplastic fluid. This fluid was centrifuged for a further 10 minutes at 4uC, filter-sterilized, and stored at 280uC. The degree of apoplast dilution was estimated as described by Rico and Preston [51]. For bacterial growth experiments, apoplast extract was freeze-dried and resuspended in an appropriate volume of distilled water to return it to its estimated concentration in planta.

Bacterial growth in apoplast extracts
Apoplast extracts from T. caerulescens plants were used as a substrate for the growth of Psm in vitro. Six 100 ml samples of apoplast extract from each metal treatment were aliquoted into a 96-well microwell plate and inoculated with 5 ml of Psm suspended in 10 mM MgCl 2 to give a final OD 600 of 0.05. The plate was incubated at 28uC with shaking and the OD 600 measured at 20minute intervals for 24 hours using the plate reader.

Metal content of whole-leaf and apoplast extracts of Thlaspi caerulescens
For determination of whole-leaf metal concentrations, fresh leaf material was oven-dried at 80uC for 48 hours. Subsamples of 50 mg of dried leaf material were digested in 3 ml of concentrated (69%, v/v) nitric acid for 16 hours in glass vials. Samples were then diluted 10-fold with ultrapure water and filtered using Whatman grade 3 filter paper. Metal contents of samples were measured in an air-acetylene flame by atomic absorption spectrophotometry using a double-beam optical system with deuterium arc background correction (AAnalyst 100; Perkin-Elmer, UK). Samples were further diluted as necessary to fall within the linear range of calibration curves prepared using appropriate standard solutions and reagent blanks. The accuracy of the calibration curves was validated using Certified Reference Material LGC7162 (strawberry leaves; LGC Standards, Teddington, UK). Three technical replicates were analysed for each measurement, and two samples from each of three plants were analysed for each metal treatment. Measurements of the fresh and dry biomass of 24 individual T. caerulescens plants were used to provide an average ratio between fresh and dry biomass, which allowed the metal content to be expressed as an approximate molar concentration in the fresh tissue. Apoplast samples were diluted appropriately with ultrapure water and measured without further treatment.

Metal-tolerance assays
The metal tolerance of Psm in vitro was tested in LB. A 5 ml aliquot of an overnight liquid culture was added to 200 ml of media supplemented with zinc, nickel, or cadmium at a range of concentrations. OD 600 was read using the plate reader. Metal tolerance was also tested in apoplast extracted from T. caerulescens plants grown in 0.1-strength Hoagland solution (10 mM zinc). For these experiments, 2 ml of an overnight culture of P. syringae pv. maculicola M4 was added to 75 ml of apoplast extract or apoplast extract supplemented with metal.
Transposon mutagenesis of Pseudomonas syringae pv. maculicola M4 Transposon mutagenesis was carried out using the transposon miniTn5::gfp::lux cloned into pGP704 (generous gift of Phil Hill). The transposon was introduced into Psm by triparental mating using the helper plasmid pRK2013 [52]. The resultant bacterial mixture was spread onto LB agar containing 50 mg/ml kanamycin and 50 mg/ml rifampicin. Resulting colonies were inoculated into 150 ml of KB broth in 96-well plates and incubated at 28uC overnight. Thirty microliters of 50% (v/v) glycerol was then added to each well and plates were stored at 280uC as a mutant library.
Screening the Pseudomonas syringae pv. maculicola mutant library for zinc tolerance mutants Mutants were screened in a four-stage process. In round one, mutants were grown in KB in 96-well plates in which a wild-type was also included. A 5 ml aliquot of overnight culture was transferred to 200 ml of KB in a black, clear-bottomed 96-well plate and incubated at 28uC in the plate reader. Plates were maintained at 28uC with continuous shaking and the OD 600 and luminescence from each well was read at hourly intervals. After 3.5 hours, cultures were supplemented with 5 ml of 0.5 mM ZnSO 4 . OD 600 and luminescence were read immediately and then at intervals of 3 to 9 hours for the next 48 hours. Changes in luminescence after the addition of zinc were recorded and growth was compared to wild-type.
A total of 866 strains were selected with markedly increased or decreased tolerance to zinc shock. Each of these mutants was transferred to 200 ml KB in two separate wells of a 96-well plate and growth was monitored with and without 0.5 mM zinc over 48 hours. Of these, 134 mutants whose tolerance differed notably from the wild-type were selected and further screened for growth in 200 ml KB in 96-well plates with zinc concentrations from 0 to 3.5 mM for 48 hours. IC 50 values were then calculated and compared to wild-type. These results were validated in a second experiment for ten mutants that showed the largest consistent changes in zinc tolerance compared to wild-type. These ten mutants were then tested for their ability to cause symptoms in T. caerulescens, A. thaliana and pak choi (Brassica rapa spp. chinensis).

Inverse-PCR based sequencing of transposon mutants
Genomic DNA was extracted using a DNeasy Blood and Tissue kit (Qiagen) according to the manufacturer's protocol for Gram negative bacteria. Eight microliters of the subsequent DNA suspension were digested with 1 ml of either the restriction endonuclease SphI or NarI (New England BioLabs) and 1 ml of the appropriate 106 buffer according to the manufacturer's specifications. NarI cleaves the transposon DNA near the 39 end, while SphI cleaves it towards the 59 end; both also cleave the Psm genome frequently. Digested DNA was ethanol-precipitated, resuspended in 8 ml of ultrapure water, and self-ligated overnight at 14uC using 1 ml of T4 ligase and 1 ml of 106buffer (New England BioLabs). One ml of circularized DNA was then used as a template for inverse PCR. Primers designed to the ends of the short fragments of transposon resulting from NarI (AACAATCTAGCGAGGGCTTGGTAA-GGTGATCC and CTTGCAGTGGGCTTACATGACGATAG-CTAGAC) or SphI (GGAACGCCGCAGGAATG and CAGCA-GCTGTTACAAACTCAAGAAG) digestion, and facing outwards, were used to amplify the flanking DNA. This was then sequenced using a primer to the end of the transposon (CGGTTTACAAGC-TAAAGCTTGC for NarI-digested DNA and either CTTC-TTTAAAATCAATACC or TTCCAGTAGTGCAAATAA for SphI-digested DNA).
Assessment of zinc tolerance of naturally occurring endophytes of Thlaspi caerulescens from a zinc-rich site Leaves of T. caerulescens were collected in the field at Hafna mine (Snowdonia, North Wales, UK: 53u079N, 3u499W). They were then returned to the laboratory and immediately surface-sterilized by immersion in 10% (w/v) sodium hypochlorite for 3 minutes, followed by immersion in 100% ethanol for 3 minutes. Sterile leaves were rinsed in ultrapure water and macerated in 1 ml of 10 mM MgCl 2 solution. The resulting suspension was plated onto KB-CFC agar. Plates were incubated at 28uC for 48 hours. Resultant colonies were transferred to 150 ml of KB broth in 96well plates. After 48 hours of growth at 28uC, 30 ml of 50% (v/v) glycerol was added to each well and the plates stored at 280uC. For zinc tolerance experiments, bacteria were grown in 200 ml KB for 48 hours and replicated into 200 ml KB supplemented with 0, 1, 2.5, 5, 7.5, 10, 12.5, 15, or 20 mM ZnSO 4 . Bacterial growth at each of these zinc concentrations was determined by measuring OD 600 at 0 and 48 hours using the plate reader. Bacterial growth at 48 hours was then plotted against zinc concentration, from which the concentrations giving half-maximal inhibition (IC 50 ) were estimated.
Quantitative RT-PCR of PR-1 gene expression RNA for qRT-PCR analysis was isolated from leaves of 10week-old T. caerulescens grown on either 0.04 or 300 mM zinc and from leaves of 6-week-old, non flowering A. thaliana. A. thaliana leaves inoculated with Psm at 10 6 cfu/ml and incubated for 24 hours under standard plant growth conditions were used as a positive control for PR-1 expression. Leaves were snap frozen in liquid nitrogen and RNA was extracted using the RNeasy kit (Qiagen) according to the manufacturer's instructions. After elution, RNA was precipitated in 100 ml of 8 M LiCl overnight. After two washes in 70% (v/v) ethanol, pellets were resuspended in 30 ml ultrapure water. RNA concentration was measured using a Nanodrop-1000 spectrophotometer (Thermo Scientific) and integrity was checked by electrophoresis on an ethidium bromide gel. cDNA was prepared from 1 mg RNA using the Bioline cDNA synthesis kit with oligo dT primer according to the supplier's instructions. qRT-PCR was performed using SYBR green PCR master mix (Applied Biosystems) in a 7300 Realtime PCR machine (Applied Biosystems) and analysed by calibration to a standard curve of gene expression created from pooled cDNA from all samples under test, using the 7300 SDS system software v1.3.1 (Applied Biosystems). Four control genes were analysed in the same way, and PR-1 gene expression normalized to the geometric mean of the expression of these genes [53]. Genespecific primers used are listed in Table 3.

Accession numbers
The GenBank (http://www.ncbi.nlm.nih.gov) accession numbers for the genes and gene products discussed in this paper are: PSPTO_0382 (NP_790231.1); PSPTO_2153 (NP_791973); PSPTO_3533 (NP_793313.1); PSPTO_5164 (NP_794895). Figure S1 In vitro growth of wild-type Pseudomonas syringae pv. maculicola ES4326 and hrcNand hrpSmutants. Strains were inoculated into KB broth to give an OD at 600 nm of 0.1, and 100 ml of this suspension was added to a further 100 ml of KB in a 96 well microplate. OD 600 was monitored over 24 h. Values are means 6 SE (n = 16). Found at: doi:10.1371/journal.ppat.1001093.s001 (0.71 MB EPS) Figure S2 Expression of PR-1 in Thlaspi caerulescens in response to zinc. Quantitative real-time PCR was used to determine the expression of PR-1 in T. caerulescens grown on 0.04 or 300 mM zinc, and in A. thaliana, either after no treatment or 24 hours after inoculation with Psm at 10 6 cfu/ml. Expression was normalised to the geometric mean of the expression of four control genes: ubiquitin, EF-1a, b-tubulin and ornithine transcarbamylase. Within each species, t-tests were used to compare the two results obtained. These showed that there was no significant difference between PR-1 expression in T. caerulescens in the two zinc treatments (P = 0.2), while inoculation of A. thaliana with Psm resulted in a significant increase in PR-1 expression (P = 0.005). Values are means 6 SE (n = 12).  Figure S4 Growth of zinc tolerance mutants of Pseudomonas syringae pv. maculicola M4 in Arabidopsis thaliana. Six leaves of three plants were infiltrated with either Psm or one of four zinc tolerance mutants of Psm (9A6, 9A3, 7C11 and 10C1) suspended in 10 mM MgCl 2 at 10 6 cfu/ml. Leaves were sampled at 0 and 5 days after inoculation. Each replicate consisted of three leaves from one plant, giving a total of three replicates per time point and treatment. Values are means 6 SE (n = 5). The experiment was carried out twice with similar results. An ANOVA was used to test for an effect of strain upon growth. This was significant (P = 0.033), but Bonferroni simultaneous comparisons showed this significance to be due to a difference between the growth of mutants 9A6 and 9A3; no mutant grew to a level significantly different from that of Psm (P = 1.000, 1.000, 0.612 and 1.000 for 9A6, 9A3, 7C11 and 10C1, respectively).