Local adaptation is associated with zinc tolerance in Pseudomonas endophytes of the metal-hyperaccumulator plant Noccaea caerulescens

Metal-hyperaccumulating plants, which are hypothesized to use metals for defence against pests and pathogens, provide a unique context in which to study plant–pathogen coevolution. Previously, we demonstrated that the high concentrations of zinc found in leaves of the hyperaccumulator Noccaea caerulescens provide protection against bacterial pathogens, with a potential trade-off between metal-based and pathogen-induced defences. We speculated that an evolutionary arms race between zinc-based defences in N. caerulescens and zinc tolerance in pathogens might have driven the development of the hyperaccumulation phenotype. Here, we investigate the possibility of local adaptation by bacteria to the zinc-rich environment of N. caerulescens leaves and show that leaves sampled from the contaminated surroundings of a former mine site harboured endophytes with greater zinc tolerance than those within plants of an artificially created hyperaccumulating population. Experimental manipulation of zinc concentrations in plants of this artificial population influenced the zinc tolerance of recovered endophytes. In laboratory experiments, only endophytic bacteria isolated from plants of the natural population were able to grow to high population densities in any N. caerulescens plants. These findings suggest that long-term coexistence with zinc-hyperaccumulating plants leads to local adaptation by endophytic bacteria to the environment within their leaves.


Introduction
Metal-hyperaccumulating plants maintain exceptionally high concentrations of metals in their aerial tissues [1][2][3][4], often exceeding 1% of tissue dry biomass for elements such as zinc, cadmium, nickel and manganese, a phenotype that is thought to protect them from herbivory [5][6][7][8][9][10][11][12] and disease [13][14][15]. There is evidence that the use of metals in defence has led to an evolutionary trade-off between metal hyperaccumulation and some pathogen-induced defences [16][17][18]. At present, however, little evidence is available concerning the effect of metal hyperaccumulation and metal-based defences upon the endophytic bacteria of hyperaccumulator plants. Two studies have surveyed the endophytes associated with nickel hyperaccumulator plants on natural serpentine soils [19,20]. Both reported that these bacteria showed high nickel tolerance, supporting the notion that the plants provide a high-metal environment for bacteria. Bacterial nickel tolerance was found to vary in proportion to nickel concentrations in different tissues, while bacterial population densities decreased as nickel concentrations increased [19]. Similarly, high cadmium tolerance was found for the bacterial endophytes of a cadmium hyperaccumulator growing on mine tailings [21]. Thus, high metal tolerance is associated with bacterial growth within metal-hyperaccumulating plants. In previous work, we demonstrated that zinc concentrations in the aerial tissues of plants of a natural Noccaea caerulescens (synonym Thlaspi caerulescens) population growing on an abandoned lead-zinc mine at Hafna, Snowdonia, UK, [22] are higher than those found to prevent the growth of pathogenic Pseudomonas syringae bacteria under laboratory conditions [15], suggesting that zinc tolerance is important for naturally occurring endophytes of these plants. Here, we investigate the possibility that the endophytic pseudomonads of metal hyperaccumulators have developed increased metal tolerance as a result of local adaptation to the metalrich environment of the hyperaccumulator leaves, a scenario that might drive the evolution of further metal hyperaccumulation in plants that use metals as a defence against pathogenic microorganisms.
In this study, endophytic Pseudomonas associated with a natural population of the zinc hyperaccumulator N. caerulescens were characterized and compared with those associated with an artificial field population. Bacterial zinc tolerance and virulence on N. caerulescens were determined and the distribution of these traits across a phylogeny of the isolated bacteria was examined. The results provide evidence that bacteria isolated from the natural population of N. caerulescens are better able to grow and cause disease in this metal hyperaccumulator. These findings support the idea of local adaptation by endophytic bacteria communities in an established, natural population of N. caerulescens, and strengthen the concept of plant pathogens as a selective force in the evolution of metal hyperaccumulation.

Material and methods (a) Field sites
Plants of N. caerulescens (J.Presl & C.Presl) F.K.Mey. (¼Thlaspi caerulescens J.Presl & C.Presl) were studied at two field locations: an artificial population cultivated at Wytham, Oxfordshire, UK (51847 0 N, 1839 0 W) and a natural population growing on wasteland at Hafna mine, an abandoned lead -zinc mine in Snowdonia National Park, Conwy, UK (53807 0 N, 3849 0 W [22]). The artificial population was generated from seed collected from N. caerulescens at Prayon, Belgium (50835 0 N, 5840 0 E [23]). Plants were precultured in a glasshouse for eight weeks on soil supplemented with zinc oxide to contain either 0, 1, 2 or 5 g kg 21 Zn (w/w). They were then transferred to two closely adjacent field sites at Wytham. At each site, plants grown on different zinc treatments were split equally between three randomized blocks.

(b) Measurement of leaf zinc concentrations
Leaf material was sampled from plants of N. caerulescens at all field sites and pools of 10 -30 plants per treatment created. Zinc content in pooled samples of leaf biomass was measured following oven drying and extraction in concentrated nitric acid by atomic absorption spectrophotometry as described [24].
(c) Endophyte collection, identification and in planta growth assays Noccaea caerulescens leaves were collected at both field sites, sealed into plastic bags and transferred to the laboratory for further analysis. From the artificial population at Wytham, 10 leaves per plant per zinc treatment were sampled. From the natural population at Hafna mine, 10 leaves per plant were sampled. Incidence of disease symptoms in sampled plants was recorded. Leaves were surface sterilized by successive immersion in 10% (v/v) sodium hypochlorite solution and 100% ethanol for 5 min each. The sterilized leaves were rinsed in sterile, distilled water and dried. Each leaf was then macerated and incubated at room temperature for 5 min in 1 ml of potassium phosphate buffer at pH 6.8. One hundred microlitre aliquots of this buffer were spread onto King's B (KB) agar, which is particularly conducive to the growth of Pseudomonas species [25]. Plates were incubated at 288C for 48 h. Colonies were picked into KB broth and incubated overnight. Cultures were supplemented with 50% (v/v) glycerol to a final concentration of 20% (v/v) glycerol and stored at -808C. All endophyte strains were subjected to LOPAT and GATTa assays [26]. In addition, 16S, rpoD and gyrB genes were sequenced (primers are given in electronic supplementary material, table S1) and used as queries for NCBI BLAST. In planta growth and pathogenicity assays were carried out as described [15].

(d) Measurements of bacterial zinc tolerance
All bacterial strains isolated from N. caerulescens plants were grown in KB broth supplemented with a range of zinc concentrations from 0 to 20 mM, and the increase in OD 600 of the medium over 24 h recorded. Bacterial growth response curves for zinc were then created and used to determine the concentration of zinc causing a 50% reduction in bacterial growth (IC 50 ; [27]). Growth response curves were also clustered by shape, using k-means clustering in 'R' [28], into groups or classes representing different patterns of response to zinc. k-means clustering maps data to the nearest mean value, where 'k' is the number of means; here, k ¼ 4 was selected as this gave clear, discrete clusters. Strains in group 4 were the most tolerant, able to grow at 15 mM Zn, whereas those in group 1 showed strongly reduced growth even at 5 mM Zn. Both measures of zinc tolerance were used to describe the strains.

(e) Phylogeny reconstruction
Sequences of endophyte rpoD genes were used to build a phylogenetic tree. Alignment was carried out using MUSCLE (EBI [29,30]). Reference strain sequences were chosen based on BLAST results for the endophytic bacteria and obtained from NCBI (www.ncbi.nlm.nih.gov). GenBank IDs are given in electronic supplementary material, table S2. Following the initial alignment, strains with poor alignment to the rest or with poor sequence quality were removed from the analysis. Remaining sequences were trimmed to the region which consistently aligned well across all sequences, and the alignment repeated. Resulting good-quality alignments were used to produce a single gene tree. rpoD gene sequences were obtained from 60% (144/244) of isolated strains and 60% (86/144) of these sequences were of sufficient quality and sufficiently well aligned to use in phylogeny reconstruction. Phylogeny reconstruction was performed in MEGA [31] by maximum parsimony [32], using the Close-Neighbour-Interchange algorithm [33] with search level 3 in which the initial trees were obtained with the random addition of sequences with 10 replicates. An estimate of the trees' robustness was obtained by conducting a bootstrap analysis [34] with 500 replicates. To investigate whether the leaves of zinc-hyperaccumulating N. caerulescens provide a local environment that imposes a selective pressure in favour of zinc tolerance, we measured and compared the zinc tolerance of bacteria isolated from plants of the artificial and natural populations. Zinc tolerance curves showing bacterial growth versus zinc concentration were produced for each endophyte strain, which were used to determine IC 50 values for zinc for each strain. The endophytes isolated from leaves from the artificial population had a greater range of zinc tolerances by this measure, but the median for the strains isolated from the natural population was significantly higher than that for the artificial population (IC 50 of 9.5 versus 7.0 mM, respectively; Mann-Whitney test: p , 0.0005). Zinc tolerance curves for individual strains were clustered by shape using k-means clustering. These clusters then served to define four tolerance groups, with group 1 being the least, and group 4 the most, tolerant strains (electronic supplementary material, figure S1). The median tolerance group of the endophytes increased as the plant zinc concentration increased, either with zinc treatment in the artificial population or between the two patches of plants in Snowdonia (figure 1f ). Kruskal-Wallis tests on these data produce significant H-statistics (Wytham zinc treatments, p ¼ 0.014; Snowdonia sub-populations, p , 0.0001). In conjunction with the results shown in figure 1a -c, this indicates that greater zinc concentrations within the plants were correlated with higher zinc tolerance among their endophytes.

Results
(c) Artificial and natural populations of Noccaea caerulescens supported different bacterial endophytes The majority of isolated strains from both artificial and natural populations belonged to the genus Pseudomonas, with some representatives of Xanthomonas, Erwinia, Enterobacter, Pantoea, Escherichia and Aeromonas. Pseudomonads were of particular interest, as previous work has shown that some Pseudomonas species can be pathogenic on N. caerulescens [15]. By focusing on Pseudomonas, direct comparison of the results obtained here to those in our previous study was facilitated [15]. KB agar was therefore selected as an isolation medium [25]. Since KB is biased in favour of pseudomonads, we limit further consideration to this genus. The composition of the two populations was widely different, with Wytham plants supporting mainly strains whose top BLAST hit was to P. aeruginosa, whereas Snowdonia plants supported mainly strains whose top BLAST hit was to P. syringae, P. graminis and P. fluorescens (electronic supplementary material, figure S2).
(d) Phylogenies of Pseudomonas endophyte strains from natural and artificial field populations of Noccaea caerulescens suggest multiple origins of zinc tolerance To investigate whether zinc-tolerant strains of endophytes were related, an rpoD gene tree was created for all endophytic pseudomonads for which high-quality rpoD gene sequences could be obtained. Figure 2 shows the resulting tree with both measures of zinc tolerance (IC 50 and tolerance group) indicated. Highly zinc-tolerant strains are distributed across the tree and are not confined to strains isolated from either population of N. caerulescens. Also included are reference strains from the genus Pseudomonas, and non-Pseudomonas reference strains Hahella chejuensis, Azotobacter vinelandii and Acinetobacter ADP1. These show this tree to be in agreement with phylogenies produced by other researchers for Pseudomonas [36], with strains of Cluster I separated from Cluster II, and, within the clusters, the aeruginosa, syringae and fluorescens complexes differentiated. These deep branches in the tree also have high bootstrap support (99%, 99% and 97%, respectively; figure 2). The average IC 50 for zinc in these three species complexes were similar (9.8 mM, 9.5 mM and 9.0 mM, respectively), and species complex was not a significant predictor of IC 50 (ANOVA, p ¼ 0.52). On the other hand, tolerance groups were not uniformly distributed between the three species complexes, with more group 4 strains in the syringae and fluorescens complexes, and more group 2 strains in the aeruginosa complex (x 2 -goodness-of-fit test to uniform distribution of groups; p , 0.001). Smaller monophyletic clades with high bootstrap support can be seen within the syringae and fluorescens clusters, some of which contain strains with similar zinc tolerances, while others include a range of zinc IC 50 values and tolerance groups.
(e) Artificial and natural populations of Noccaea caerulescens supported different bacterial endophytes To identify the strains isolated, we sequenced 16S, rpoD and gyrB genes and used the results to query the NCBI database using BLAST [37] (electronic supplementary material, table S3).
Notably, a higher incidence of strains closely related to P. syringae was found in Snowdonia (electronic supplementary material, figure S2), which is consistent with disease incidence at this site, since at least some strains of P. syringae (g) Pathogenicity of endophytes on Noccaea caerulescens depends on both zinc tolerance and association with the natural population of Noccaea caerulescens To determine the effect of zinc tolerance on the virulence of candidate pathogens, six endophyte strains belonging to the species P. syringae (determined using sequence data and LOPAT and GATTa assays) and the previously studied laboratory strain P. syringae pv. maculicola M4 were inoculated into N. caerulescens plants (from Prayon, Belgium) grown with 0.04, 10, 30 or 300 mM zinc. Four strains were isolated from the natural N. caerulescens population, and displayed a range of zinc tolerances; the remaining two were from the artificial N. caerulescens population and displayed high-zinc  (table 1). As previously reported [15], Psm was able to multiply within plants grown on 0.04 and 10 mM Zn, but showed reduced or no growth in plants grown on 30 mM or 300 mM Zn. Bacterial growth in planta is shown in figure 3. The four strains isolated from the natural population of N. caerulescens were all able to grow in planta when the plants were grown on 0.04 mM Zn. The ability of these strains to multiply in plants grown on higher concentrations of zinc was correlated with their zinc tolerance, with the most tolerant, strain SnC10 (IC 50 for Zn ¼ 11.8 mM) able to multiply in all plants, although with reduced growth at higher zinc; strains SnB11 and SnD3 (IC 50 ¼ 7.1 and 7.6mM, respectively) able to multiply only in plants grown on 30 mM Zn or lower, again with IC 50 by Zn (mM) natural population artificial, 0 g kg -1 artificial, 1 g kg -1 artificial, 2 g kg -1 artificial, 5 g kg -1 Hahella  Figure 2. rpoD gene tree of Wytham and Snowdonia strains. The bootstrap consensus tree from 120 most parsimonious trees is shown. The percentage of replicate trees in which the associated taxa clustered together in the bootstrap test (500 replicates) is shown next to those branches where the bootstrap value exceeds 50% [27]. Branches are coloured red for endophytes of plants from the natural population in Snowdonia and blue for endophytes of plants from the artificial population at Wytham. Within the natural population, ** indicates that the plant from which the endophyte was isolated was symptomatic, whereas * indicates that the plant was from the same spatial sub-population as the symptomatic plants; no asterisk indicates that the isolate was from the asymptomatic sub-population. Within the artificial population, blue branches are further colour-coded to indicate the zinc treatment applied to the plant from which the endophyte was extracted (light blue: dashed: no added zinc, solid: 1 g zinc added per kilogram of soil; dark blue: dashed 2 g kg 21 , solid: 5 g kg 21 ). Zinc tolerance groups based on groupings for all strains on KB are shown in black and shades of blue, with group 4 (highest) in black, through to group 1 (lowest) in pale blue. Zinc concentrations giving IC 50 for strains are indicated with a heat map where red is highest (13-16 mM Zn) and yellow lowest (0 -4 mM).
rspb.royalsocietypublishing.org Proc. R. Soc. B 283: 20160648 less success at higher zinc; and the least tolerant strain, SnD4 (IC 50 ¼ 6.7 mM), only able to multiply in the plants grown on 0.04 mM Zn. Psm has an IC 50 of 7.25 mM, placing it between SnB11 and SnD3, whose in planta behaviour it closely replicated. By contrast, neither of the strains isolated from the artificial population of N. caerulescens was able to multiply in planta, regardless of plant zinc treatment, despite the high in vitro zinc tolerances of these two bacterial strains (IC 50 ¼ 10 and 11.3 mM).

Discussion
In this work, we investigated the possibility that populations of bacterial endophytes of metal-hyperaccumulating plants become locally adapted to the metal-rich environment they encounter within these plants. Previously, we have demonstrated that endophytic bacteria of the zinc hyperaccumulator N. caerulescens show higher zinc tolerance when compared with bacteria pathogenic on related non-accumulator plants [15], an idea supported by other studies of bacteria associated with metal-hyperaccumulating plants [18,19,[38][39][40]. Considered alongside evidence for a role of hyperaccumulated metals in defence of plants such as N. caerulescens against disease [13][14][15]41,42], this suggests that local adaptation of pathogens to high-metal concentrations in the rhizosphere and phyllosphere of metal-accumulating plants might drive the evolution of further defensive hyperaccumulation in a form of coevolutionary arms race [15,43].
Here, we have shown that bacteria isolated from the zinchyperaccumulating plant N. caerulescens display a degree of zinc tolerance correlated with the zinc concentrations found in the leaves of the plants from which they were isolated. This result supports the idea that the leaf environment provided by this zinc hyperaccumulator exerts a selection pressure for zinc tolerance. While in theory it is possible that disease leads to reduced metal accumulation, any reduction in metal uptake capacity would occur as a result of disease symptoms; zinc accumulation, however, occurs throughout the life of the plant, so that any measureable differences in the zinc content of aerial tissues are likely to predate the infection. Additionally, this is demonstrated in laboratory experiments (e.g. [15]) in which plants were first cultivated, in a healthy state, on a range of Zn concentrations and then artificially inoculated with P. syringae, resulting in symptoms with severity inversely proportional to measured plant Zn content. On the other hand, there is now an extensive body of evidence indicating that cross-talk occurs between metal and biotic stress signalling at the level of ROS (e.g. [15]), hormones (e.g. [44]), transcriptomic and proteomic changes (including enzymes of the secondary metabolism, PR proteins and more, see [45] for review). This cross-talk may confer cross-protection against disease when plants are exposed to and accumulate heavy metals such as zinc. Additionally, there is mounting evidence (e.g. [6,7,15,16]) that metal hyperaccumulation does indeed protect these plants from disease, as would be expected under the hypothesis that an evolutionary arms race taking place in the zinc-rich environment of hyperaccumulators and their surroundings is responsible for driving the evolution of the metal accumulation trait.
We have also compared the zinc tolerance of bacteria isolated from a natural population (Snowdonia) of N. caerulescens plants to those isolated from an artificial population (Wytham) created for this study. The preponderance of more zinc-tolerant bacteria originating from the Snowdonia site, and of less zinc-tolerant bacteria from the Wytham site, suggests that the natural population of N. caerulescens at Snowdonia provided a selection pressure acting in favour of zinc-tolerant bacteria. The scattered distribution of zinc tolerance across the rpoD gene tree implies multiple origins of zinc tolerance, within and among groups of related bacteria, suggesting that this trait might be acquired relatively easily, as seen in a library of transposon mutants of P. syringae pv. maculicola [15]. This supports the hypothesis that, if metal accumulation evolved as a defence, rapidly evolving pathogens are likely to have driven a subsequent arms race leading to the progressive enhancement of the metal-accumulation phenotype [16].
It is possible that differences in plant genotype between the N. caerulescens plants in the natural and artificial environments may have influenced the establishment of endophytic bacteria. However, differences in endophyte zinc tolerance that covary with plant zinc content are seen at the withingenotype level. Thus, the findings shown in figure 1f, that the median zinc tolerance group into which isolated endophytes fall is positively correlated with plant zinc content in the natural population, and with the zinc treatment on which the plants were grown in the artificial population, indicate that the plant zinc content affects the zinc tolerance of its bacterial endophytes in a manner independent of plant genotype.
Interestingly, while IC 50 values for zinc are not predicted by the species complexes into which a strain falls on the rpoD tree, zinc-tolerance group percentages do vary between these phylogenetic groups. Strains in the aeruginosa complex are exclusively found in tolerance groups 2 and 3, both of which show fairly linear zinc dose-response curves (electronic supplementary material, figure S1). Strains in the syringae and fluorescens complexes are, respectively, equally or more likely to fall into zinc tolerance group 4 than groups 2 and 3. Group 4 strains show little or no growth reduction until zinc concentrations are fairly high (5-10 mM zinc; electronic supplementary material, figure S1). The ability to withstand up to 10 mM zinc without growth reduction might give these bacteria an advantage when growing endophytically in N. caerulescens plants exposed to high zinc.
Notably, there are only three monophyletic groups with high bootstrap support (99%) containing exclusively highly zinc-tolerant strains (IC 50 . 9 mM Zn and tolerance group 3 or 4) isolated from plants of the natural population. These Table 1. Zinc tolerance of strains selected for in planta growth assays, showing zinc tolerance groups as defined by k-means clustering (electronic supplementary material, figure S1) and IC 50 values for zinc. This might indicate that the established, high-zinc environment of the natural population tends to select for the zinc tolerance trait wherever it arises, while the zinctolerant bacteria from the artificial site represent an earlier evolution of tolerance unrelated to the sudden, artificial appearance of Noccaea plants. The scattered distribution of zinc-tolerant bacteria across the tree could also reflect horizontal gene transfer among bacteria in Snowdonia, in addition to mutation towards greater zinc tolerance.
While pathogens isolated from the natural population were able to infect N. caerulescens under laboratory conditions, the reverse was not true of even highly zinctolerant bacteria from the artificial population. We have previously shown that the type 3 secretion system (T3SS), used to deliver bacterial disease effectors to the host, is essential for the growth of P. syringae pv. maculicola (Psm) in planta in N. caerulescens [15]. This indicates that, despite the importance of zinc as a defence [15] and the loss of certain other defences by the metal hyperaccumulators [16,17], the plants retain defences against which the T3SS is necessary, probably including some elements of PAMP-triggered immunity (PTI: [46,47]). It is also possible that some elements of effectortriggered immunity (ETI: [46,48]) are retained. Thus, to infect N. caerulescens, bacteria require both zinc tolerance and the ability to overcome pathogen-induced plant defences. As a result, N. caerulescens may only be vulnerable to pathogenic bacteria such as P. syringae, which possess the T3SS and effectors, but which have also been exposed to a zinc-rich environment and acquired zinc tolerance. We hypothesize that at the former mine site in Snowdonia, the natural population of N. caerulescens has selected for zinctolerant, N. caerulescens-specific pathogens. These are able to infect N. caerulescens, while endophytic bacteria from the artificial population, with no long-term exposure to N. caerulescens, proved non-pathogenic.
In conclusion, this study provides evidence for local adaptation of endophytic bacteria to the zinc-rich environment of leaves of N. caerulescens plants growing on a zincrich substrate. Pathogenicity, however, requires more than simply zinc tolerance, and was only found to be prevalent in the natural population. It will be of interest to determine which elements of PTI and ETI are retained by N. caerulescens, and whether these occur in all N. caerulescens, or vary with the degree of zinc availability, tolerance and hyperaccumulation in different populations. Investigation of endophytes and pathogens of other populations, and their ability to infect plants of other populations, would contribute to a more complete picture of the emergence and ongoing evolution of hyperaccumulation and associated endophytes.