Exploiting natural variation to uncover candidate genes that control element accumulation in Arabidopsis thaliana

Authors

  • Simon J. Conn,

    1. European Molecular Biology Laboratory, Grenoble Outstation, 38042 Grenoble, France
    2. School of Agriculture, Food, and Wine & The Waite Research Institute, University of Adelaide, Waite Campus, PMB1, Glen Osmond, South Australia 5064, Australia
    Search for more papers by this author
  • Philipp Berninger,

    1. European Molecular Biology Laboratory, Grenoble Outstation, 38042 Grenoble, France
    Search for more papers by this author
  • Martin R. Broadley,

    1. Plant and Crop Sciences Division, University of Nottingham, Sutton Bonington, Loughborough LE12 5RD, UK
    Search for more papers by this author
  • Matthew Gilliham

    1. School of Agriculture, Food, and Wine & The Waite Research Institute, University of Adelaide, Waite Campus, PMB1, Glen Osmond, South Australia 5064, Australia
    2. Australian Research Council Centre of Excellence in Plant Energy Biology, University of Western Australia, Crawley, Western Australia 6009, Australia
    Search for more papers by this author

  • Simon Conn was a finalist for the 2011 New Phytologist Tansley Medal for excellence in plant science, which recognises an outstanding contribution to research in plant science by an individual in the early stages of their career; see the Editorial by Dolan, 193: 821–822.

Author for correspondence:
Simon Conn
Tel: +33 476 20 77 25
Email: sconn@embl.fr

Summary

The plant ionome varies both inter- and intraspecifically despite the highly conserved roles for particular elements across the plant kingdom. Element storage requires transport across the plasma membrane and commonly deposition within the central vacuole. Therefore, tonoplast transport characteristics can be highly influential in controlling the plant ionome. As a result, individual cell types of the same plant, each with unique transcriptomes and vacuolar proteomes, can display very different elemental profiles. Here we address the use of natural variation in Arabidopsis thaliana for identifying genes involved in elemental accumulation. We present a conceptual framework, exploiting publicly available leaf ionomic and transcriptomic data across 31 Arabidopsis accessions, that promises to accelerate conventional forward genetics approaches for candidate gene discovery. Utilizing this framework, we identify numerous genes with documented roles in accumulation of calcium, magnesium and zinc and implicate additional candidate genes. Where appropriate, we discuss their role in cell-specific elemental accumulation. Currently, this framework could represent an alternate approach for identifying genes suitable for element biofortification of plants. Integration of additional cell-specific and whole-plant ‘omics’ datasets across Arabidopsis accessions under diverse environmental conditions should enable this concept to be developed into a scalable and robust tool for linking genotype and phenotype.

Elemental distribution within leaves of Arabidopsis thaliana

Plant storage vacuoles are a major site for accumulation of elements; with contents of the central vacuole commonly exceeding 0.5 M in total solute concentration (Fricke et al., 1994). Vacuolar partitioning achieves cellular detoxification and storage of nutrients for remobilization in times of deficiency, in addition to generating bulk cellular osmotic pressure (Marty, 1999). However, not all central vacuoles are equivalent. For example, vacuolar concentrations for a particular element can differ by over four orders of magnitude between plant species (Buescher et al., 2010; Conn & Gilliham, 2010). A large component of genetic variation for plant mineral composition exists at the species level. For example, Watanabe et al. (2007) showed that only 25% of the variation for 21 elements occurred at the family level and above, with most genetic variation occurring between or within species. Intraspecific variation in elemental accumulation is also evident among A. thaliana natural accessions, with leaf concentrations of calcium [Ca], magnesium [Mg] and zinc [Zn] varying by 3.3-fold, 3.7-fold and 7.2-fold, respectively (based on leaf weight normalized values (mg kg−1 DW) across 413 natural A. thaliana accessions obtained from http://www.ionomicshub.org). For both inter- and intraspecific variations, any differences in the properties of the central vacuole, as it is a dominant storage site, will impart a large effect on whole plant elemental accumulation.

Intriguingly, different cell types of the same plant can also vary in vacuolar concentration of specific elements by over four orders of magnitude, despite equivalent availability of ion(s) within the adjacent apoplast and similar unidirectional plasma membrane influx for some ions (Karley et al., 2000; Conn & Gilliham, 2010). As a result, some elements are rarely found to colocalize within the same cellular vacuoles at high concentrations, such as, Ca and phosphorus (P), possibly because of their predisposition to forming insoluble precipitates (Conn & Gilliham, 2010). The cell types that preferentially store particular elements are conserved within a species, but differ between families/orders of flowering plants (Conn & Gilliham, 2010). For example, Ca and Mg are preferentially accumulated within dicot leaf mesophyll tissue, which comprises over 60% of the Arabidopsis leaf volume, whereas epidermal cells that make up c. 10–15% of the total leaf accumulate significantly less. However, both Ca and Mg generally accumulate to significantly lower concentrations in leaves of cereal monocots, which accumulate these elements preferentially in the epidermis and bundle sheath, respectively (White & Broadley, 2003; Conn & Gilliham, 2010). Therefore, the abundance of the cell type accumulating the bulk of a specific element can also affect its concentration within the whole plant.

Many additional factors are also known to influence the Arabidopsis ionome. For example, [Ca] is strongly correlated with the transpiration rate of an organ and is low in purely phloem-fed organs, given its relative immobility in the phloem (White & Broadley, 2003). Furthermore, shoot Ca accumulation is also highly influenced by variations in root architecture, root plasma membrane uptake kinetics, root cation exchange capacity, growth rate, leaf cell wall Ca-binding properties and the extent of root suberization (White & Broadley, 2003; Dayod et al., 2010; Gilliham et al., 2011a). In many cases, though, it is the transport characteristics of the tonoplast that determine elemental storage capacity (Conn & Gilliham, 2010).

Elemental storage in particular leaf and root cell types is regulated by a specific subset of genes and is critical for control of whole plant concentrations for numerous elements, including sodium (Baxter et al., 2010), Ca (Conn et al., 2011a), Mg (Conn et al., 2011b), iron (Kim et al., 2006) and others (Puig & Peñarrubia, 2009). However, identifying the genes responsible for these distributions is a challenging and lengthy procedure, commonly involving large natural breeding populations or screens of laboratory-induced mutants. Therefore, alternate tools for expediting the prioritization of gene candidates that control elemental accumulation in plants are desirable.

Approaches for identifying genes involved in elemental accumulation

Classic forward genetics approaches

The genetic and phenotypic variation across > 1,500 Arabidopsis accessions has proved an excellent resource for identifying genes involved in various biological processes and traits, including element accumulation (Atwell et al., 2010; Koornneef et al., 2011). Quantitative trait locus (QTL) mapping, using populations of recombinant inbred lines (RILs) resulting from a cross between pairs of accessions divergent for a particular trait, is a common approach for finding loci that control or influence a phenotype. These are sensitive approaches with significant QTLs identified for Ca and Mg content between Arabidopsis accessions differing by < 1.6-fold in concentration, despite the broader available intraspecific variability mentioned above (Supporting Information, Table S1 and references therein). Depending on the heritability of the trait, QTLs comprising up to thousands of open reading frames (ORFs) can often be identified to explain various proportions of the phenotypic variation between parents. Subsequent high-resolution (fine) mapping is required to significantly reduce the number of candidate ORFs within a given QTL (Koornneef et al., 2011). Despite the multigenic process of ion uptake and storage, approaches that use Arabidopsis natural variation have been successfully utilized to map or identify transcripts contributing to elemental accumulation in > 30 reports (Table S1).

However, situations arise where highly complex traits may go uncharacterized by the use of natural variation across small sample sizes with limited genetic diversity. Thus, the advent of techniques which integrate wider sets of accessions and RILs can be employed such as multiparent advanced generation inter-cross (MAGIC) lines (Kover et al., 2009) and Arabidopsis multiparent RIL (AMPRIL) (Huang et al., 2011). Interestingly, the broader genetic spectrum of these crosses resulted in masking of a number of QTLs for flowering time compared with biparental crosses, presumably resulting from the genetic complexity introduced with the crosses. As detection of QTLs requires segregation in the progeny, the possibility of different founder alleles having similar effects or epistatic interactions can reduce the number of QTLs detected even in these studies with larger populations. Such approaches exploiting variation (including QTL and genome-wide association (GWA) analyses) are time-consuming, costly and thus may restrict many laboratories from undertaking them (Buescher et al., 2010). Thus, the framework we present here incorporates natural variation across multiple Arabidopsis accessions to implicate single genes, without introducing artefacts inherent in these breeding programmes.

An integrative framework for implicating genes involved in leaf elemental accumulation

Given that gene expression is a quantitative trait, a framework is presented here for identifying transcripts contributing to elemental accumulation by integrating published leaf transcriptomics and ionomics data across Arabidopsis accessions. Our study compares element content of plant leaves quantified by inductively coupled plasma-mass spectroscopy (ICP-MS) (http://www.ionomicshub.org) with genome-wide transcript profiles (microarray) of 31 Arabidopsis accessions (Lempe et al., 2005). While ionomics data were obtained from plants grown under consistent conditions at the same facility, the data were restricted maximum likelihood (REML)-normalized according to Broadley et al. (2010) in order to minimize any environmental variation between experimental trays. The overall result of this framework provides a filter, independent of gene annotation, to implicate genes involved in, or responsive to, leaf elemental accumulation. Furthermore, for elements that are preferentially accumulated in a cell-specific manner, the pool found in those cell types will dominate the total element accumulation of the plant (Conn & Gilliham, 2010). By cross-referencing candidates with cell-specific transcriptomes, one can predict whether gene candidates are also involved in cell-specific accumulation or exclusion.

Within the population studied here, there was a normal distribution of leaf [Ca], [Mg] and [Zn], despite the minimal variation across the 31 accessions: 1.3-fold (Ca), 1.4-fold (Mg) and 4.5-fold (Zn) (Fig. S1). By comparing the abundance of individual transcripts with leaf [Ca], [Mg] or [Zn] for each accession, over 1,500 transcripts were significantly correlated (< 0.05). By implementing more stringent conditions, including a minimum expression cutoff and a threshold Pearson correlation coefficient (r < −0.3, r > 0.3) deduced from the normal distributions of r-values (Fig. S2), c. 400 remaining transcripts with a strong linear correlation were identified and further investigated. Encouragingly, while various ontological classifications were represented in the ‘element-correlated’ subpopulation, there was a five- to sixfold enrichment of transporters for each element relative to the whole genome annotation (Fig. 1). Furthermore, for each element there was enrichment (> 2-fold) for unannotated transcripts without a predicted gene function. As a result, we can speculate that there remain uncharacterized roles for these transcripts in elemental accumulation. Within this review, we have concentrated on the transporter-related genes involved in elemental accumulation as this allows us to illustrate the power of the approach more easily.

Figure 1.

Enrichment for ion transporters and unannotated transcripts utilizing the population correlation framework; correlating transcript abundance with elemental accumulation across 31 Arabidopsis accessions. Enrichment of significantly correlated transcripts categorized into gene ontologies (GO) for Ca (black columns), Mg (white columns) and Zn (grey columns) compared with whole-genome annotation (fold-change) in the ‘element-enriched’ subpopulation. GO classification for transcripts significantly correlated with accumulation of Ca, Mg and Zn using the TAIR GO annotation tool (http://www.arabidopsis.org/tools/bulk/go/index.jsp).

Of the transporters found to be correlated with [Ca], [Mg] or [Zn] across these accessions, we identified two with known roles in cell-specific elemental accumulation for Ca (AtCAX1; Fig. 2a) (Conn et al., 2011a) and Mg (AtMRS2-1; Fig. 2b) (Conn et al., 2011b). To confirm the suitability of the microarray as a reliable measure of transcript abundance, AtCAX1 and AtMRS2-1 were measured using quantitative reverse transcriptase-polymerase chain reaction (qPCR). Abundance of both transcripts were found to be strongly correlated between the qPCR and array data across 10 accessions from within the microarray panel (Fig. 2c,d). However, as multiple transporters were specifically correlated with each element (Table 1), a series of verification steps were performed to reduce the likelihood of false positives, given the population size of 31 accessions. These include quantifying the ionomes of T-DNA insertion lines and probing GWA datasets for various alleles in close proximity to these genes.

Figure 2.

Correlation plots using the population correlation framework of ion transporters with a known role in leaf elemental accumulation. Expression intensity (log2) from Lempe et al. (2005) plotted against restricted maximum likelihood (REML)-normalized leaf element concentration (mg kg−1 DW) for AtCAX1 expression with leaf [Ca] (a), AtMRS2-1 expression with leaf [Mg] (b). Correlation of AtCAX1 (c) and AtMRS2-1 expression (d) by quantitative PCR (qPCR) on leaves of 5-wk-old Arabidopsis plants grown in hydroponics performed in our laboratory with raw array intensities (Lempe et al., 2005) for the following accessions: Columbia-0, C24, Landsberg erecta, Cape Verde Islands-0, Niederzenz-1, Kindalville-0, Vancouver-0, Estland, Bayreuth-0 and Shahdara. Refer to Fig. S4 for average linkage clustering of microarray data from each accession, resulting in the exclusion of the Frankfurt-2 accession. Pearson correlation coefficient (r) shown with P-value. Mean value for each accession is shown where replication was performed. Plant growth, RNA isolation, qPCR methodology and primers as per Conn et al. (2011a,b).

Table 1.   Transporters identified by the population correlation framework as specifically correlated with accumulation of Ca, Mg or Zn in Arabidopsis leaves
AGIGene annotationrLine analysis (percentage change in element concentration)Leaf cell-specific expression
Line nameCaMgZn
  1. NSD, no significant difference (from wildtype background; P > 0.05); –, no ionomics data available; GC, guard cell.

  2. Candidate transporters uniquely correlated with leaf concentration of Ca, Mg or Zn across 31 Arabidopsis accessions. Statistics performed using Pearson’s correlation coefficient (r), with significance of all correlations at P < 0.05. Change in element concentration for each T-DNA insertion line determined from http://www.ionomicshub.org or published literature and presented against wildtype levels. Gene annotation obtained from the Aramemnon website (http://aramemnon.botanik.uni-koeln.de/) and leaf cell-specific expression retrieved from the ePlant website (http://bar.utoronto.ca/eplant/).

  3. Scatterplots for each transporter referred to in the text against [Ca], [Mg] and [Zn] can be found in Fig. S3.

Calcium
At2g38170AtCAX1: high-affinity Ca2+/H+ cation exchanger0.510cax1-1NSDNSDNSDMesophyll
cax1/cax3−42% mesophyll Ca (Conn et al., 2011a)
At2g43950AtOEP37: chloroplast outer membrane ion channel0.427Mesophyll = GC
At5g47560AttDT: tonoplast dicarboxylate transporter0.395Mesophyll
At1g23090AtSultr3.3: putative sulphate transporter0.382SALK_000822 (5% lower [S])+11%NSDNSDEpidermis
At4g03560AtTPC1 (TWO-PORE CHANNEL 1); calcium channel/voltage-gated calcium channel0.378tpc1-2Unaltered Ca; but tpc1-2 had higher epidermal Ca (Gilliham et al., 2011b).NSDNSDEpidermis
At1g79230AtMST1/AtRDH1:mercaptopyruvate sulphurtransferase0.349Epidermis = mesophyll
At4g35920AtMCA1: putative Ca2+-permeable mechanosensitive channel−0.392mca1Unaltered Ca; but overexpressing line had higher root Ca (Nakagawa et al., 2007).GC (and roots)
Magnesium
At3g12100Putative cation diffusion facilitator (AtMTP5/AtMTPc2)0.563mtpc2-1NSD9–17% decreaseNSDEquivalent in each cell type
At3g01280AtVDAC1: voltage-dependent anion channel0.512SALK_011520NSDNSDNSDEpidermis
At2g23150AtNRAMP3: multispecific vacuolar metal cation transporter0.488nramp3, nramp3/nramp4NSDNSDNSDEpidermis = mesophyll
At1g15960AtNRAMP6: putative metal cation transporter0.390Epidermis = mesophyll
At1g54115AtCAX10/AtCCX4: putative cation/H+-antiporter0.384SALK_108983NSD+10%NSDGC = mesophyll
At1g55910AtZIP11: putative zinc(II)/iron(II) cation transporter0.353SALK_0856923–18% increase−12%NSDEpidermis
At1g16010AtMRS2-1/MGT2-1: putative magnesium cation transporter0.306mgt2-130% reduction in Mg compared with Col-0 under serpentine conditions (Conn et al., 2011b).Mesophyll
At4g23710AtVHA-G2: subunit G of vacuolar ATP synthase−0.330Epidermis = mesophyll
At1g60960AtIRT3: putative zinc(II)/iron(II) cation transporter−0.344Multiple linesNSD5–13% increaseNSDXylem
At2g01980AtSOS1/AtNHX7: putative sodium ion/proton antiporter−0.349Multiple linesNSD8–14% decreaseNSDEpidermis = mesophyll
At3g02050AtKUP3/AtKT4: putative potassium cation transporter−0.367SALK_002622NSD+5%NSDMesophyll
At4g33530AtKUP5/AtKT5: putative potassium cation transporter−0.397SALK_120707−6%−6%NSDEpidermis
Zinc
At3g26570AtPHT2;1: inorganic phosphate transporter0.342SALK_139411NSDNSD−13%Mesophyll
SALK_133096NSD+9%−17%
At2g46800AtMTP1/AtZAT1: vacuolar zinc/H+ anitporter0.321Overexpression lines show 25–33% decrease in leaf ZnMesophyll = GC
At4g39080AtVHA-A3: vacuolar V-ATPase subunit−0.368SALK_029786NSD−13%+ 30%Epidermis
At1g10130AtECA3/AtACA6: putative calcium cation-transporting P2A-type ATPase−0.380SALK_061433NSDNSD+ 16%Epidermis

Approaches and recommendations for validating the framework

The unprecedented number of Arabidopsis T-DNA insertion mutants and constitutive overexpression transgenics in public repositories is a great resource for plant researchers. The leaf ionomes of over 12,000 unique lines incorporating natural accessions, laboratory-induced mutants and transgenics for c. 2,500 unique genes (c. 10% of the Arabidopsis genome) have been catalogued at the iHUB website (http://www.ionomicshub.org). Of those transporter genes in the ‘element-correlated’ subpopulation for which mutants were catalogued, we found over 75% (14/18) were perturbed for the correlated element of interest (Table 1). While the lack of detectable ionomic differences in the remaining T-DNA insertion lines of the candidate transporters may indicate compensatory pleiotropic mechanisms following knockout, it may also be indicative of an indirect or chance correlation with the accumulation trait. Thus, we recommend additional validations before investigating or excluding candidates, including cross-referencing with GWA databases cataloguing single nucleotide polymorphisms (SNPs) within the genes across these accessions, or profiling the expression of other known family members in the tissue(s) of interest.

Candidate genes correlated with accumulation of specific elements

Calcium accumulation

The preferential accumulation of Ca in Arabidopsis leaf mesophyll vacuole ([Ca]vacc. 65 mM) compared with the epidermis and bundle sheath ([Ca]vacc. 5 mM) is controlled by a tonoplast-localized, mesophyll-specific member of the CAX (calcium/proton exchanger) transporter family, AtCAX1 (Conn et al., 2011a). As presented in Conn et al. (2011a), AtCAX1 was strongly positively correlated with Ca accumulation in Arabidopsis leaves (r = 0.510), and this new global framework identified it in the 35 strongest-associated transcripts (Table S2). Furthermore, AtCAX1 was found not to be significantly associated with either leaf [Mg] or [Zn] (Fig. S3). The lack of a measurable Ca accumulation phenotype in whole-leaf ICP-MS for the single T-DNA insertion line of AtCAX1 (cax1-1) was attributed to complementation by the homologue AtCAX3, which, when abolished in a double T-DNA insertion line (cax1/cax3), resulted in a 17% decrease in total leaf [Ca] compared with wildtype (Table 1). Furthermore, spatial Ca distribution was affected in this line having reduced mesophyll [Ca]vac (− 42%), unperturbed epidermal Ca, yet a threefold higher concentration of apoplastic Ca ([Ca]apo) (Conn et al., 2011a). Consistent with an increase in free [Ca]apo is a reduced growth rate, reduced leaf gas exchange, induction of transcripts for additional Ca2+-transporters targeted to various endomembranes, increased cell wall thickness and altered cell wall polymer profile. As growth in a medium with low [Ca] alleviated these phenotypes to wildtype levels, Conn et al. (2011a) proposed that cax1/cax3 plants were unable to appropriately partition Ca, with a concomitant increase in cell wall rigidity (including Ca-linked pectins) and perturbed cytosolic Ca-signalling. We presume the specific role of this transporter may extend to monocots, with the barley orthologue (HvCAX1) having greater abundance in the leaf epidermis where Ca is enriched (Fricke et al., 1994; Richardson et al., 2007). While further testing must be done to confirm the dependency on these transporters in other species, this infers a similar role for orthologous proteins in cell-specific elemental accumulation between plants.

AtTPC1, which is responsible for tonoplast SV channel activity, was also positively correlated with leaf [Ca] (not Mg, or Zn) (Table 1). Interestingly, the fou2 mutant, which has a point mutation within AtTPC1 that makes the SV channel less sensitive to luminal Ca, also accumulates more Ca in the mesophyll (Beyhl et al., 2009). It is proposed that AtTPC1 is part of a potassium (K) circuit that functions to efflux K+ from the vacuole to balance the positive charges (H+) entering the vacuole through function of the vacuolar H+-ATPase (V-ATPase). This prevents hyperpolarization of the tonoplast potential whilst providing a H+-gradient to drive vacuolar Ca accumulation via AtCAX1 (Hedrich & Marten, 2011). Therefore, an increase in [Ca] concurrent with increasing AtTPC1 expression would make sense. However, this theory is complicated by the tpc1-2 T-DNA insertion line displaying no whole leaf Ca phenotype (http://www.ionomicshub.org). This may result from AtTPC1 expression being restricted to the epidermis and vasculature, the leaf cell types accumulating the lowest [Ca]vac (Conn et al., 2011a). The observation that epidermal [Ca]vac is significantly increased in tpc1-2 compared with its parental background, Columbia-0, is further evidence against this theory (Gilliham et al., 2011b). Therefore, AtTPC1 may function to prevent vacuolar Ca accumulation in cells that do not accumulate Ca by a distinct mechanism, possibly by acting as a tonoplast Ca ion (Ca2+)-binding sensor, as demonstrated recently by Dadacz-Narloch et al. (2011), or in a vacuolar Ca2+ efflux pathway (Gilliham et al., 2011b). Clearly these mechanisms need to be further interrogated.

AtMCA1 encodes for a plasma membrane-localized mechanosensing Ca2+-permeable channel proposed to play a role in root growth (Nakagawa et al., 2007), yet its abundance is negatively correlated with Ca accumulation (Table 1). AtMCA1 is also highly abundant in the leaf guard cell, and thus AtMCA1 may also have a role in stomatal dynamics and regulation of transpiration (which is critical for leaf Ca accumulation) (Table 1). Interestingly, shoot growth of plants overexpressing AtMCA1 exhibited reduced growth and necrosis on high Ca media, similar to the phenotype of the cax1/cax3 mutant line which also had altered stomatal dynamics and reduced transpiration (Nakagawa et al., 2007; Conn et al., 2011a). Therefore, further analysis of the role of AtMCA1 in stomatal physiology or root-to-shoot signalling is warranted.

Dicarboxylic acids are known to be enriched in the Arabidopsis mesophyll, along with Ca (Hurth et al., 2005). Thus, these acids may provide a flexible counterion for Ca given the positive correlation of AttDT, a mesophyll-specific malate transporter critical for maintenance of vacuolar pH, with leaf Ca accumulation (Table 1) (Hurth et al., 2005). Using GWA across 300 Arabidopsis accessions, three SNPs were found to flank AttDT with the minor allele present in the Cape Verde Islands accession, which accumulates lower leaf Ca than Landsberg erecta (Ler) (I. Baxter & D. Salt, unpublished; White, 2005). Furthermore, a QTL has also been found in this region using RILs resulting from a cross between these accessions (Buescher et al., 2010). However, ionome profiling of this line is required to establish a conclusive role for AttDT in Ca accumulation.

Magnesium accumulation

As with Ca, Mg is preferentially accumulated within the Arabidopsis leaf mesophyll, which is in part controlled by AtMRS2-1 and AtMRS2-5, homologues of the bacterial CorA family of Mg2+ transporters (Conn et al., 2011b). Loss of AtMRS2-1 or AtMRS2-5 results in plants with lower mesophyll [Mg]vac and reduced growth, but only under serpentine (high Mg : Ca) conditions. No effect on leaf osmolality was found under these growth conditions as a result of compensatory increases in vacuolar Ca and potassium (K) in mesophyll cells, which correlated with increased transcript abundance of vacuolar transporters, including AtCAX1 (Conn et al., 2011b). Accordingly, two members of the KUP family of potassium (K+) transporters were found to be negatively correlated with Mg accumulation in this study (Table 1). This reciprocity seen for Ca, Mg and K substantiates the importance of maintaining the cell-specific ionome, with a rapid but sustained response of ‘compensating’ transporters measured at the transcript level (Conn et al., 2011b). Interestingly, AtCCX4 (a distinct, but closely related gene to those in the CAX family) is significantly correlated [Mg] and [Zn], but not Ca accumulation (Fig. S2), supporting the theory that the CCX family is able to transport divalent (and monovalent) cations, but not Ca2+ (Shigaki et al., 2006).

The strongest correlated transporter transcript with Mg accumulation was AtMTP5/AtMTPc2 (r = 0.563, P = 0.001), the homologue of which has higher abundance in the more serpentine-tolerant Thlaspi caerulescens compared with T. arvense (Hammond et al., 2006). Interestingly, no significant correlation was seen with Zn accumulation (r = 0.066, P = 0.724), despite T. caerulescens being a Zn hyperaccumulator, an observation supported by the ionome of the T-DNA insertion line (mtpc2-1), perturbed for only Mg (Table 1).

Zinc accumulation

AtMTP1, which is known to localize to the vacuole in roots and leaves, and to drive Zn accumulation in Arabidopsis leaf cells (Desbrosses-Fonrouge et al., 2005), was one of the few transporters positively correlated with Zn accumulation. Furthermore, with Zn predominantly accumulated in the Arabidopsis mesophyll, it is unsurprising that AtMTP1 is predominantly expressed in this cell type (Table 1) (Conn & Gilliham, 2010). Studies have shown that root cellular distribution of Zn in an AtMTP1 T-DNA insertion line differed from that of wildtype, and thus it would be interesting to probe Zn distribution in Arabidopsis leaves (Kawachi et al., 2009). The large amount of negatively correlated transcripts found for Zn accumulation could be explained by the regulatory role that the shoots play in acquiring Zn2+ from the roots. For instance, AtVHA-a3 (a subunit of the V-ATPase proton pump) is involved in establishing proton gradients across the tonoplast and its role in Zn tolerance and accumulation has been studied and shown to be antagonistic (Krebs et al., 2010), matching its negative correlation with Zn accumulation using this framework (Table 1). Therefore, with a number of candidate genes identified, in addition to the enrichment for (as yet) unannotated transcripts, there is great potential for identifying novel factors involved in element accumulation at the leaf level and, in certain cases, the cell-specific level.

Conclusions and future directions

The tool described in this study provides a framework for identifying transcripts involved in elemental accumulation in Arabidopsis leaves. As is commonplace with GWA/QTL analyses for elemental variation, the output of this framework was a number of transcripts, each contributing a small amount to the intraspecific variation in leaf element concentration. While we concede important root-driven transport pathways are overlooked with this approach, and false positives may arise, the identification of a number of element-specific transcripts which have been validated in previous reports, or whose T-DNA insertions display anticipated perturbations in their ionome, awards a greater degree of confidence in the output. Integration of additional accessions with broader leaf element concentration ranges under varying environmental conditions, at both the whole-leaf and cell-specific levels, into this tool would minimize false positives and widen its applicability. In addition, this tool can be further evolved in light of the cataloguing of genetic variation in Arabidopsis by sequencing genomes of 1001 accessions (Cao et al., 2011), public depositories for genome-wide transcript profiles (Gan et al., 2011) and the cell-specific transcriptomics databases for various cell types of Arabidopsis. We also suggest the scope of this framework could be expanded to integrate various other ‘omics’ datasets (proteomics, metabolomics) or physiological data (growth, nutrient use efficiency) to implicate gene candidates in diverse biological processes.

One immediate application of the output from the current version of this framework is in the field of plant biofortification. This may enable researchers to engineer a rational approach to spatiotemporally control the transgene and enrich for target elements. This is needed, as previous studies that have constitutively overexpressed genes have led to both biomass and harvest yield penalties. For instance, while heterologous expression of AtCAX1 in its constitutively active, truncated form (sCAX1) in tomatoes produced fruits with higher [Ca], this was associated with a yield penalty from perturbed Ca accumulation patterns with co-accumulation of some potentially hazardous ions and increased susceptibility to blossom end-rot (Park et al., 2005). However, cotransformation of a Ca2+-binding protein retained greater bioavailable Ca and recovered plant biomass (Wu et al., 2009). Accordingly, two calcium-binding proteins (At2g32450 and At3g24110) were positively correlated with leaf [Ca] and none were found to be negatively correlated using this tool (Table S2). Thus, before applying candidate genes in biofortification strategies, consideration must be given to the consequences of perturbing cell-specific accumulation and bioavailability of elements (Dayod et al., 2010). This has been used with great success by using grain endosperm cell-specific promoters to fortify white rice with bioavailable iron by targeted coexpression of nicotianamine synthase and ferritin (Wirth et al., 2009). Therefore, we provide all element/gene matrices (Table S2) to enable researchers to investigate, evaluate and build upon this tool and explore roles of nontransporter genes (including transcription factors) and unannotated transcripts in cell-specific elemental accumulation. Furthermore, lodging of additional ionomics/transcriptomics from other tissues, particularly the roots, are essential to validate the framework and broaden its application, enabling the fortification of other cereals and horticultural plants with wider consumption than the Brassicaceae.

Acknowledgements

We thank David Salt, Ivan Baxter and Mourad Ouzzani for their assistance in accessing and compiling ionomics data from the PiiMS dataset and GWA analyses; Dr Vanessa Conn for critical analysis of the manuscript, and Dr Ramesh Pillai for his support of this work. Funding for S.C. and P.B. was provided by the EMBL with additional funding for P.B. from the SNF Fellowship grant for prospective researchers (reference PBBSP3-133782); funding for M.R.B. was provided by BBSRC grant (reference BB/G013969/1) and funding for M.G. was provided by the University of Adelaide. This work was initiated by S.C. and M.G. supported by an ARC Discovery Project (reference DP0774063) awarded to Prof. Roger Leigh, Prof. Steve Tyerman and Dr Brent Kaiser, whose input and contributions were invaluable for this work.

Ancillary