Magnesium transporters, MGT2/MRS2-1 and MGT3/MRS2-5, are important for magnesium partitioning within Arabidopsis thaliana mesophyll vacuoles


  • Simon J. Conn,

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

    1. School of Agriculture, Food, & Wine and The Waite Research Institute, University of Adelaide, Waite Campus, PMB1, Glen Osmond, South Australia 5064, Australia
    2. Australian Centre for Plant Functional Genomics, Waite Campus, PMB1, Glen Osmond, South Australia 5064, Australia
    Search for more papers by this author
  • Stephen D. Tyerman,

    1. School of Agriculture, Food, & Wine and The Waite Research Institute, University of Adelaide, Waite Campus, PMB1, Glen Osmond, South Australia 5064, Australia
    Search for more papers by this author
  • Brent N. Kaiser,

    1. School of Agriculture, Food, & Wine and The Waite Research Institute, University of Adelaide, Waite Campus, PMB1, Glen Osmond, South Australia 5064, Australia
    Search for more papers by this author
  • Roger A. Leigh,

    1. School of Agriculture, Food, & Wine and The Waite Research Institute, University of Adelaide, Waite Campus, PMB1, Glen Osmond, South Australia 5064, Australia
    Search for more papers by this author
  • Matthew Gilliham

    1. School of Agriculture, Food, & Wine and The Waite Research Institute, University of Adelaide, Waite Campus, PMB1, Glen Osmond, South Australia 5064, Australia
    Search for more papers by this author

Authors for correspondence:
Matthew Gilliham
Tel: +61 8 8303 8145

Simon Conn


  • Magnesium accumulates at high concentrations in dicotyledonous leaves but it is not known in which leaf cell types it accumulates, by what mechanism this occurs and the role it plays when stored in the vacuoles of these cell types.
  • Cell-specific vacuolar elemental profiles from Arabidopsis thaliana (Arabidopsis) leaves were analysed by X-ray microanalysis under standard and serpentine hydroponic growth conditions and correlated with the cell-specific complement of magnesium transporters identified through microarray analysis and quantitative polymerase chain reaction (qPCR).
  • Mesophyll cells accumulate the highest vacuolar concentration of magnesium in Arabidopsis leaves and are enriched for members of the MGT/MRS2 family of magnesium transporters. Specifically, AtMGT2/AtMRS2-1 and AtMGT3/AtMRS2-5 were shown to be targeted to the tonoplast and corresponding T-DNA insertion lines had perturbed mesophyll-specific vacuolar magnesium accumulation under serpentine conditions. Furthermore, transcript abundance of these genes was correlated with the accumulation of magnesium under serpentine conditions, in a low calcium-accumulating mutant and across 23 Arabidopsis ecotypes varying in their leaf magnesium concentrations.
  • We implicate magnesium as a key osmoticum required to maintain growth in low calcium concentrations in Arabidopsis. Furthermore, two tonoplast-targeted members of the MGT/MRS2 family are shown to contribute to this mechanism under serpentine conditions.


The divalent magnesium cation (Mg2+) is fundamental to many processes in the biosphere, and appropriate concentrations of the ion need to be maintained in all living cells and cellular compartments (Maguire & Cowan, 2002). Magnesium ions contribute towards membrane stability, act as a cofactor in enzymatic reactions, and form the central atom of chlorophyll (Knoop et al., 2005). Magnesium (Mg) is unique in that it is the most charge-dense of all cationic nutrients; yet once hydrated its atomic radius increases 400-fold to be the largest of all divalent cations (Maguire & Cowan, 2002). For these reasons the phytoavailability of Mg from the soil solution, in contrast to calcium (Ca), is affected more by competing cations (particularly Al3+, Mn2+and Ca2+) than by binding to soil particles (Karley & White, 2009; White & Broadley, 2009). Given that Mg concentrations ([Mg]) in soil solutions extracted at field capacity range between 125 μM and 8.5 mM, Mg transport mechanisms must be able to adapt to fluctuations in supply in order to maintain the high Mg demand of plant aerial tissues (Broadley et al., 2008). The delivery of Mg to shoot tissues is determined by both transpirational water flows and phloem loading with [Mg] in the shoot apoplast approximately equal to that of the xylem sap (Wegner et al., 2007).

Best studied among the membrane transport systems for Mg2+ in Arabidopsis are the homologues of the bacterial CorA proteins, which were named after the cobalt resistance phenotype observed in bacterial mutants (Kehres et al., 1998; Moncrief & Maguire, 1999). CorA proteins comprise two closely spaced C-terminal transmembrane (TM) domains, the first of which ends with a Gly-Met-Asn (GMN) tripeptide motif (Knoop et al., 2005). Crystallization of the Thermotoga maritima CorA protein revealed that each subunit oligomerizes into a pentameric structure with their TM domains lining the membrane pore, while the five N-termini form a large cone-shaped funnel within the cytosol (Eshaghi et al., 2006; Lunin et al., 2006; Payandeh & Pai, 2006). Within Arabidopsis, homologues of the bacterial CorA Mg2+ transporters have been annotated as MRS2 (Schock et al., 2000) or MGT genes (Li et al., 2001) and comprise 10 members and one pseudogene (AtMRS2-9) (Gardner, 2003; Karley & White, 2009). All but AtMRS2-9 have been shown to rescue a Mg2+ transport-deficient Saccharomyces cerevisiae mutant (CM66) and, when targeted to mitochondria in transgenic yeast lines, enable Mg uptake into isolated mitochondria (Schock et al., 2000; Li et al., 2001; Drummond et al., 2006; Gebert et al., 2009). Members of this family vary in their affinities for Mg2+ transport, including high-affinity (AtMRS2-10, AtMRS2-11), low-affinity (AtMRS2-2, AtMRS2-7) and dual-affinity (AtMRS2-6) members (Li et al., 2001, 2008; Drummond et al., 2006; Mao et al., 2008; Chen et al., 2009). In planta, members of the MRS2 family have been localized to the plasma membrane (PM), mitochondria and chloroplast but not to the tonoplast, with the vacuole presenting the greatest potential sink for Mg storage (Li et al., 2001, 2008; Alexandersson et al., 2004; Drummond et al., 2006; Chen et al., 2009). However, the role of MRS2s in plant function is largely unknown.

Entry of Mg2+ into root cells is believed to be mediated by entry through Mg2+ permeable channels, with AtMRS2-10 thought to contribute most to this (Li et al., 2001; Karley & White, 2009). The Arabidopsis Mg2+/H+ antiporter (AtMHX1) encoded by a single gene has been suggested to contribute to the majority of vacuolar Mg uptake, despite no alteration in [Mg] in transgenic Arabidopsis plants carrying 35S:AtMHX1 constructs (Shaul et al., 1999). Both AtMRS2-2 and AtMRS2-3 (Waters & Grusak, 2008) and AtMHX1 (Vreugdenhil et al., 2004) co-localize with major quantitative trait loci (QTL) underscoring seed Mg concentration in Arabidopsis. Phenotypes of single and multiple Arabidopsis T-DNA loss-of-function lines of specific MRS2s include perturbations in pollen development (Li et al., 2008; Chen et al., 2009) and reduced growth under Mg-limiting conditions (Gebert et al., 2009). However, as yet none have demonstrated significantly altered tissue [Mg], a possible result of functional complementation by other family members. Furthermore, while transgenic Arabidopsis lines overexpressing AtMRS2-11 (Drummond et al., 2006) or AtMRS2-7 (Gebert et al., 2009) showed unperturbed ionomes, overexpression of the plasma membrane-targeted AtMRS2-10 in Nicotiana tabacum resulted in an increase in leaf [Mg] (Deng et al., 2006), supporting their annotation as Mg2+ transporters in planta.

Serpentine soils are defined by high Mg : Ca ratios and can also incorporate deficiencies of essential macronutrients (nitrogen, potassium, phosphorous), low water availability and/or high concentrations of heavy metals (Brooks, 1987). Polymorphisms within AtMRS2-2, AtMRS2-6 and AtMRS2-7 have been correlated with adaptation of Arabidopsis ecotypes to serpentine soils (Turner et al., 2010). According to the four criteria for serpentine adaptation proposed by Tyndall & Hull (1999), plants tend to show: greater tolerance of low Ca2+ and high Mg2+ concentrations; a higher Mg2+ requirement for maximum growth; lower Mg2+ absorption; and Mg2+ exclusion from leaves. Taken together, this would implicate the mitochondrial (AtMRS2-2, AtMRS2-6) and endomembrane-targeted (AtMRS2-7) MRS2s in catalysing Mg2+ uptake and partitioning within the plant under conditions of Ca2+ deficiency.

The vacuole is the major storage site for many nutrients in plant cells and also provides the majority of cell osmoticum. As Ca is an osmoticum to some plant cells (Conn & Gilliham, 2010), it is critical to balance for this under conditions of Ca deficiency. While it has been well documented that potassium can act as an osmoticum for perturbations in vacuolar cations (e.g. Ca2+, Na+) across numerous species (reviewed in Conn & Gilliham, 2010) and given the vacuole is the major storage site for intracellular Mg2+ (Marschner, 1995; Shaul, 2002), it is possible that under serpentine conditions Mg2+ fulfils this function.

In this study we sought to identify the key transporters involved in vacuolar Mg accumulation within the leaf and how these may be involved in growth under serpentine conditions. To achieve this, a cell-specific microarray was performed utilizing single cell sampling and analysis (SiCSA), and SiCSA in conjunction with X-ray microanalysis (XRMA) to compare epidermal and mesophyll transcriptomes and elemental accumulation, respectively. We identified two differentially expressed MRS2 candidates that were localized to the tonoplast, AtMRS2-1 and AtMRS2-5, and their respective T-DNA insertion lines displayed perturbed vacuolar Mg, leaf osmolality, chlorophyll concentrations and growth under serpentine conditions.

Materials and Methods

Plant materials and growth conditions

All chemicals were obtained from Sigma-Aldrich unless stated otherwise. Arabidopsis thaliana GAL4-VP16:UAS-GFP enhancer trap line KC464 with epidermal pavement cell-specific GAL4/GFP expression (Haseloff, 1999; Gardner et al., 2009) and wild-type, all of ecotype background Columbia-0 (Col-0) were germinated and grown in hydroponics as described later. Seeds for the T-DNA insertion lines of AtMRS2-1, AtMRS2-5 and AtMRS2-10 genes were purchased from the Arabidopsis Biological Resource Center (Columbus, OH, USA) and identified using primers specific for the T-DNA left and right borders and AtMRS2-specific primers (Supporting Information, Table S1). The T-DNA insertion site in the allele is described in the Salk Institute website ( and confirmed here along with the absence of coding sequence confirmed by PCR (Fig. S1).

All plants were constantly aerated in hydroponic tanks, under a 9 : 15 h, light : dark cycle, with 55% atmospheric humidity, at 22°C and an irradiance of 150 μmol photons m−2 s−1. Growth solutions used were basal nutrient solution (BNS) (¼ Hoagland’s solution; with Ca and Mg activities at 1 mM (aMg = aCa = 1.0 mM)) or the ‘serpentine’ solutions, aMg = 1.0 mM, aCa = 25 μM (low calcium solution; LCS) or aCa = 1 mM and aMg = 6.6 mM (high magnesium solution; HMS) (Table S2). Calcium was supplied as CaCl2 and Ca(NO3)2, while Mg was supplied as MgSO4 and MgCl2.

Magnesium leaf feeding assays

Leaf eight of 6-wk-old plants was excised under deionized water and the petiole inserted into degassed artificial sap (AS) solution within a 1.5 ml microcentrifuge tube and the petiole re-cut using small dissecting scissors to avoid cavitation within the xylem. AS contained (in mM) 1 K2HPO4, 1 KH2PO4, 1 CaCl2, 0.1 MgSO4, 1 KNO3, 0.1 MnSO4. Twenty millimolar MgCl2 was added to AS to generate Mg-supplemented AS solution (AS + Mg). Parafilm (Pechiney Plastic Packing Company, Chicago, IL, USA) was used to seal around the petiole and prevent evaporation from the tube, and XRMA and or SICSA/XRMA was performed on these leaves after 18 h of transpirational feeding as per (Møller et al., 2009; Conn et al., 2011).

Cell-specific elemental profiling: single cell sampling and cryoscanning electron microscopy (cryoSEM)/ X-ray microanalysis (XRMA)

Single-cell sampling and analysis (SiCSA) and cryoSEM/XRMA were performed essentially as described in Tomos & Sharrock (2001) and Storey & Leigh (2004), respectively, with the following modifications. All samples were obtained from cells situated in the middle of leaf eight (as described by Tsukaya, 2002) of 5- to 8-wk-old Arabidopsis plants. Elemental composition of samples taken from individual leaf cells were measured using XRMA with spectra analysed using EDAX eDXi software (EDAX Inc., Mahwah, NJ, USA) and converted from peak over background ratios to concentration values using calibration standards (Fig. S2). All XRMA data were processed using an automated algorithm, XRMAplot (Australian Centre for Plant Functional Genomics, Adelaide, SA, Australia).

RNA isolation and amplification for microarray and quantitative polymerase chain reaction (qPCR)

RNA was isolated from specific cell types of Arabidopsis enhancer trap line KC464 by SiCSA as per Roy et al. (2008), amplified and hybridized to custom 4 × 44 k Agilent microarrays or used for qPCR (Conn et al., 2011). Details of the RNA amplification/validation and microarray analysis are presented in Supplementary Methods. Real-time qPCR was performed using Bio-Rad iQ SYBR Supermix (Bio-Rad) on both whole-leaf- and SiCSA-amplified cDNA. Normalization was carried out using control genes AtEF1α, Atβ-Tubulin5, AtActin2 and AtGAPDH-A (Table S3) and the final concentrations of mRNAs of the genes of interest are expressed as arbitrary units that represent the numbers of copies per 30 ng of cDNA from total RNA (or amplified equivalent assuming 5% mRNA component), normalized against the geometric means of the three best control genes using the geNORM software ver. 3.5 (Vandesompele et al., 2002). Primers for qPCR analysis along with their amplification efficiencies (qGene software) are given in Table S3.

Subcellular localization

The full coding sequences of AtMRS2-1 (At1g16010, NM_101469), AtMRS2-5 (At2g03620, NM_126412) and AtMRS2-10 (At1g80900, NM_106738) with and without the stop codon were amplified from Arabidopsis leaf cDNA (primers listed in Table S4) using Phusion Hot Start High Fidelity DNA Polymerase (Finnzymes, Espoo, Finland). Products were A-tailed, and cloned into pCR8/GW/TOPO according to manufacturer’s instructions (Life Technologies, Carlsbad, CA, USA). Correct orientation of the insert was confirmed by restriction digest analysis and sequencing with the coding sequence (with or without stop codon, respectively) recombined into the pBS 35S::YFP-attR (Genbank accession: AY995137) or the pBS 35S::attR-YFP (Genbank accession: AY995141) vectors using LR clonase II (Life Technologies) to generate N- and C-terminal yellow fluorescent protein (YFP) fusions under the control of 35S CaMV promoter (Subramanian et al., 2006). Following LR recombination, there is a physiologically uncharged 13-amino-acid spacer in both N-terminal (DSLYK+K+AGSEFAL) and C-terminal (K+GEFDPAFLYK+VV) fusions between the MRS2 and YFP proteins. Plasmid was purified by Zyppy Maxiprep Kit (Zymo Research, Irvine, CA, USA) with 10 μg used to transform Col-0 Arabidopsis mesophyll protoplasts according to the methods of Yoo et al. (2007). Following 18 h of dark incubation, cells were imaged using a Leica SP5 spectral scanning confocal microscope with fluorescence images captured sequentially and merged using LEICA SPS software (Leica Microsystems, Wetzlar, Germany). Excitation/emission conditions for YFP (514 nm/527–552 nm), cyan fluorescent protein CFP, (458 nm/470–500 nm) and chlorophyll autofluorescence (488 nm/664–696 nm) were used throughout.

Vector construction

For stable plant transformation, modifications were made to the pMDC100 vector (Curtis & Grossniklaus, 2003) to introduce a dual CaMV 35S promoter, nos terminator sequence and bar herbicide resistance gene (pVC4-bar; Genbank accession number: HQ175992). The 35S:bar gene was obtained from the pTOOL2 binary vector (Møller et al., 2009) and directionally inserted into the pVC4 construct using flanking AloI (Thermo Fisher Scientific, Waltham, MA, USA) and BsaXI (New England Biolabs, Ipswich, MA, USA) restriction sites common to both vectors to replace the entire 35S:nptII region of pVC4. Full-coding sequences for AtMRS2-1, AtMRS2-5 or AtMRS2-10 in pCR8/GW/TOPO were recombined into pVC4-bar by LR clonase II, as per manufacturer’s instructions.

Transformation and selection of T-DNA insertion lines

Col-0, mrs2-1a and mrs2-5 lines were transformed by the floral dip method using the Agrobacterium tumefaciens GV3101 line. Transformants were selected on 0.5× MS plates in the presence of both kanamycin sulphate (50 μg ml−1) and glufosinate-ammonium (50 μM) (Bayer Crop Science, Monheim, Germany) as per Harrison et al. (2006), and confirmed by PCR screening for the full-length gene and qPCR forward primers (Tables S2, S3) and a vector-specific reverse primer (pVC4-R: GTACAAGAAAGCTGGGTCG). Six to eight T1 transformants were screened per line/construct combination, and at least two lines were selected for further analysis using leaf-feeding assays.

Bulk leaf analyses: osmolality, chlorophyll concentration and soluble sugar analysis

Leaf osmolality was calculated from whole leaf samples by snap freezing leaf number eight within a Parafilm-sealed 1 ml syringe on liquid N2, while the remainder of the shoot was also frozen, then lyophilized for 18 h for soluble sugar analysis. The single leaf was allowed to defrost at room temperature, then ground to a slurry using a fine metal rod, with the sap ejected onto a clean glass microscope slide (20–30 μl). Ten microlitres of this leaf sap was analysed using a vapour pressure osmometer (Model 5500; Wescor, Logan, UT, USA).

Freeze-dried shoot samples were ground into a fine powder using the LM1 laboratory ring mill for 2 min (Labtech Essa, Belmont, WA, Australia). Five milligrams was used to extract soluble sugars using an anthrone colorimetric assay, with a fructose standard curve as described previously (Leyva et al., 2008). Chlorophyll (Chla,b and total chlorophyll) was extracted and quantified from 15 mg of freeze-dried material with 1 ml methanol as per Warren (2008).


Magnesium is preferentially accumulated in the Arabidopsis mesophyll vacuole

Vacuolar magnesium ([Mg]vac) was quantified in Arabidopsis leaf cells by cryoSEM/XRMA. Mg predominantly accumulated in mesophyll cells, especially under high magnesium supply, over both epidermal and bundle sheath cells (Fig. 1a). SiCSA, a technique able to compare between epidermal and mesophyll cells, is a more sensitive method for Mg quantification than cryoSEM, with detection limits of 15 and 25 mM, respectively (Fig. S2), and was used to examine the relationship between [Mg]vac, calcium ([Ca]vac) and potassium ([K]vac) (Fig. 1b). Palisade mesophyll cells were found to accumulate 57 mM higher [Ca]vac and 15 mM higher [Mg]vac than neighbouring adaxial epidermal cells, while adaxial epidermal cells were enriched for [K]vac by 55 mM. The quantification of [Mg]vac was shown to be independent of the SiCSA detection limit by the serial concentration of cell samples before XRMA (three to 10 cell samples per analysis), generating equivalent concentrations as with single cell inputs (data not shown).

Figure 1.

Magnesium accumulates preferentially within the mesophyll vacuole of Arabidopsis leaves and is correlated with abundance of magnesium transporters. (a) Magnesium accumulation within Arabidopsis leaf cell vacuoles measured using cryoscanning electron microscopy (cryoSEM)/X-ray microanalysis (XRMA). Data are presented as the ratio of Mg-supplemented artificial sap (AS) solution to nonsupplemented AS solution showing enrichment in mesophyll cells only. (b) Concentration of magnesium (Mg), calcium (Ca) and potassium (K) in adaxial epidermal (closed bars) and palisade mesophyll (open bars) cell vacuoles in Arabidopsis leaves as determined by single cell sampling and analysis (SiCSA)/XRMA. (c–d) Expression of known magnesium transporters in 5-wk-old, basal nutrient solution (BNS)-grown Arabidopsis KC464, an epidermal pavement cell enhancer trap plant (Col-0 background) with epidermis-specific green fluorescent protein (GFP), within the leaf adaxial epidermis (closed bars) and palisade mesophyll (open bars) by microarray (c) and qPCR (d) analyses on amplified RNA, normalized against AtEF1α, Atβ-Tubulin5 and AtActin2 as per described methods (mean ± SEM; n = 3 amplifications from adaxial epidermis and palisade mesophyll from independent plants). Asterisks indicates a significant difference in transcript abundance between epidermis and mesophyll (Student’s t-test, < 0.05).

Transcript abundance of specific AtMRS2 genes are correlated with preferential magnesium accumulation in leaf mesophyll cells

A microarray screen was performed on epidermal and mesophyll tissue from Arabidopsis line KC464 using SiCSA in combination with RNA amplification to correlate gene expression of candidate Mg2+- transporters with [Mg]vac. Expression of green fluorescent protein (GFP) and other known cell-specific genes were used to exclude contamination of samples with other cell types (Fig. S3). While six Mg2+-transporters were found to be highly expressed in the mesophyll (log2 intensity > 8.0), four of these were correlated with the higher accumulation of Mg in the mesophyll as a result of their low abundance in the epidermis (Fig. 1c). AtMRS2-10, AtMRS2-1, AtMRS2-4 and AtMRS2-5 were 18-fold, 24-fold, 82-fold and 98-fold higher in abundance within the mesophyll than the epidermis, respectively. qPCR on RNA from these cell types isolated by SiCSA found all these candidates except AtMRS2-4 to be undetectable within the epidermis and an equal abundance between the epidermis and mesophyll for AtMRS2-11 and AtMHX (Fig. 1d). Furthermore, a meta-analysis using 23 Arabidopsis ecotypes uncovered a strong positive correlation between leaf [Mg] as measured by inductively coupled plasma mass spectroscopy (ICP-MS) and transcript abundance by microarray of AtMRS2-1 and AtMHX but not AtMRS2-4, AtMRS2-10 or AtMRS2-11 (Fig. S4). Conversely, there was an inverse correlation between leaf [Ca] and expression of both AtMRS2-1 and AtMHX, suggesting some interplay between these nutrients.

Subcellular localization of AtMRS2s

While AtMRS2-10 has been localized to the plasma membrane (Li et al., 2001), and AtMRS2-4 has been putatively localized to the mitochondria (S. Conn, unpublished), subcellular localization has yet to be established for AtMRS2-1 and AtMRS2-5. N- and C-terminal YFP fusions of AtMRS2-1, AtMRS2-5 and AtMRS2-10 were developed for transfection of Arabidopsis mesophyll protoplasts (Fig. 2). All N-terminal YFP fusions resulted in cytosolic YFP fluorescence (Fig. S5), whereas C-terminal YFP fusions were targeted to the tonoplast for both AtMRS2-1 and AtMRS2-5 (Fig. 2) and, as previously shown, AtMRS2-10 localized to the plasma membrane (Li et al., 2001; Fig. S5). Given these localizations, any of the three proteins could contribute to Mg supply, or storage within the mesophyll vacuoles, and thus further analyses focused only on these genes.

Figure 2.

Subcellular localization of AtMRS2-1, and AtMRS2-5 in Arabidopsis mesophyll protoplasts. Confocal images of Arabidopsis protoplasts which were osmotically disrupted to release intact vacuoles following transfection with 35S-AtMRS2-1-YFP (a) and 35S-AtMRS2-5-YFP (b). Images captured on a Leica SP5 spectral scanning confocal microscope as described in the Materials and Methods section. Bars, 20 μm. DIC, differential interference contrast; YFP, yellow fluorescent protein.

Loss-of-function lines for tonoplast-localized AtMRS2 genes have lower [Mg]vac under serpentine conditions

The preferential accumulation of [Mg]vac and expression of two tonoplast-localized (AtMRS2-1, AtMRS2-5) and one PM-targeted (AtMRS2-10) AtMRS2 genes within the mesophyll led us to investigate whether their loss could perturb Mg accumulation in leaves. Loss-of-function T-DNA insertion lines of Arabidopsis for AtMRS2-1 (two lines designated 2-1a, 2-1b), AtMRS2-5 (mrs2-5) and AtMRS2-10 (mrs2-10) were identified (Table S1) and analysed for their [Mg]vac accumulation phenotype within leaves. Despite loss of detectable full-length coding sequence (Fig. S1), none of these lines were found to have altered cell-specific Mg accumulation under standard hydroponic conditions by SiCSA/XRMA (Fig. S6a), as has been shown previously for most lines by ICP on whole leaves (Gebert et al., 2009). When excised leaves from these lines were fed AS solution with supplemental Mg (aMg = 12.5 mM) for 18 h, [Mg]vac was greater within the palisade mesophyll of all lines when compared with the same lines fed nonsupplemented AS solution (Fig. 3a). At the same time there was a reduction in both mesophyll [Ca]vac and [K]vac. However, under these conditions mrs2-1 and mrs2-5 lines had reduced palisade mesophyll [Mg]vac when compared with Col-0 and mrs2-10 (Fig. 3a). When whole plants were grown in HMS (aCa = 1 mM, aMg = 6.6 mM) for 4 d, palisade mesophyll [Mg]vac was similarly reduced (22–25%, Fig. S6b) in only mrs2-1 (2-1a and 2-1b) and mrs2-5 lines, but all lines showed phytotoxic phenotypes, including elevated chlorophyll concentration, flaccid leaves and also irreversible reduced growth (data not shown), as seen previously (Visscher et al., 2010).

Figure 3.

T-DNA insertion lines for AtMRS2-1 and AtMRS2-5 have reduced mesophyll vacuolar Mg accumulation under serpentine growth conditions. Difference in mesophyll vacuolar [Mg] (black bars), [K] (grey bars) and [Ca] (white bars) in a detached leaf assay between leaves fed artificial sap (AS) solution and Mg-supplemented AS solution for 18 h (a) or in whole plants grown for 7 d in low-calcium solution (LCS) (b), presented as mean difference from the equivalent basal nutrient solution (BNS)-grown plants (n = 32 samples per condition). (c) Quantitative polymerase chain reaction (qPCR) abundance of Mg2+-transporters in leaves correlates with vacuolar Mg accumulation. qPCR performed on RNA isolated from 5-wk-old Arabidopsis Col-0, mrs2-1, mrs2-5 and cax1/cax3 plants grown in BNS and a 7 d treatment in LCS. (mean ± SEM; n = 3 biological replicates; qPCR performed in triplicate). (d) Complementation of T-DNA insertion lines mrs2-1a and mrs2-5 with 35S CaMV promoter-driven AtMRS2-1, AtMRS2-5, AtMRS2-10, or empty vector controls. Single cell sampling and analysis (SiCSA)/X-ray microanalysis (XRMA) performed on adaxial epidermis (closed bars) and palisade mesophyll (open bars) cells using Mg-supplemented AS solution on excised leaves (n = 25 cells of each type, from five independent plants from a single representative T1 transformant). Lettering (a, b, c) indicates significance categories, relative to Col-0 values (Student’s t-test, P < 0.05).

Alternative serpentine conditions were imposed by maintaining Mg activity (aMg = 1.0 mM) and reducing aCa to 25 μM (LCS). Growth of Col-0 in LCS resulted in a reduction in mesophyll [Ca]vac from day 3 to day 7, and a concomitant enrichment of [Mg]vac and [K]vac (Fig. S7). While all lines showed a similar decrease in mesophyll [Ca]vac, the mrs2-1 and mrs2-5 lines accumulated 33–39% less [Mg]vac than Col-0 after 7 d of treatment. This was associated with a 75–82% increase in mesophyll [K]vac (Fig. 3b).

To investigate the inverse relationship between mesophyll [Mg]vac and [Ca]vac, another mutant which has been shown to require serpentine conditions for maximal growth was analysed (Cheng et al., 2005). The Arabidopsis cax1-1/cax3-1 (cax1/cax3) mutant line abolishes expression of AtCAX1 and AtCAX3 (known to partially complement for loss of AtCAX1) (Catala et al., 2003; Bradshaw Jr., 2005; Cheng et al., 2005; Visscher et al., 2010). When grown in BNS and fed Mg-supplemented AS solution, cax1/cax3 showed a similar increase in [Mg]vac as Col-0 and mrs2-10, yet showed a greater loss of [K]vac and higher retention of [Ca]vac than all other lines tested (Fig. 3a). However, when grown in LCS, cax1/cax3 accumulates greater [Mg]vac within the mesophyll, significantly above all other lines, while there is a significantly smaller increase in [K]vac (Fig. 3b).

MRS2 expression correlates spatiotemporally with Mg accumulation

Transcript profiling by qPCR was utilized to correlate expression of specific MRS2s with alterations in vacuolar [Mg] imposed by the various treatments. In both Col-0 and cax1/cax3, four MRS2s –AtMRS2-1, AtMRS2-5, AtMRS2-10 and AtMRS2-11– were induced following 7 d of LCS treatment, while AtCAX1 is repressed and AtMRS2-4 is unaffected by the LCS treatment (Fig. 3c). While all other MRS2s and AtMHX1 were induced to a similar extent as Col-0, both AtMRS2-1 and AtMRS2-5 were up-regulated 7.8-fold and 8.7-fold in cax1/cax3, as opposed to 3.8-fold and 3.5-fold in Col-0, respectively (Fig. 3c). Within the mrs2 T-DNA insertion lines, a degree of transcriptional interplay between family members was also evident, with AtMRS2-1 and AtMRS2-5 being 1.6- to 1.8-fold higher in the mrs2-5 and mrs2-1 lines under all treatments compared with Col-0 (Fig. 3c).

Complementation of mrs2-1 (line 2-1a) and mrs2-5 by CaMV 35S promoter-driven MRS2s elevated epidermal Mg in all transgenic lines (Fig. 3d). However, recovery of mesophyll [Mg]vac with the Mg-supplemented AS solution assays in excised leaves was only observed with the self-complementation (mrs2-1 + 35S::AtMRS2-1 and mrs2-5 + 35S::AtMRS2-5) (Fig. 3d).

Correlation of leaf osmolality and chlorophyll concentration with growth

Cellular osmolality and the rate of solute uptake into cells are major drivers for cell expansion and leaf growth (Fricke & Peters, 2002; Cosgrove, 2005). Bulk leaf osmolality was measured for each line under BNS and LCS conditions and osmolality increased with LCS treatment for all lines, except cax1/cax3, where it decreased (Fig. 4a). Within T-DNA insertion lines of AtMRS2-1 and AtMRS2-5, osmolality was further increased by 25–30 mosmol kg−1 (0.061–0.073 MPa osmotic pressure) over Col-0 (Fig. 4a); however, leaf water content of all lines under all conditions did not differ (Table S5). Soluble sugar concentration of bulk leaf extracts of each line was quantified to ascertain its contribution to leaf osmolality. LCS treatment of Col-0, or any mrs2 T-DNA insertion line, was found not to effect the soluble sugar concentration compared with BNS-grown plants (< 1 mg g−1) (P > 0.05). However, a 12% reduction in sugar concentration was seen in the cax1/cax3 line, which had higher sugar concentration than Col-0 when grown in the BNS media (P < 0.001) (Table S5). Total chlorophyll concentration was found not to be correlated with mesophyll vacuolar [Mg] but was found to be positively correlated with plant growth (Fig. 4b) (r2 = 0.8035; Fig. S8a). Bulk leaf osmolality was found to be inversely correlated with growth rate for all mrs2 lines and conditions, including the slow-growing cax1/cax3 line, which increased its growth rate and chlorophyll concentration in LCS, with a concomitant reduction in leaf osmolality to that of LCS-grown Col-0 (Fig. 4a,c) (r2 = 0.8854; Fig. S8b).

Figure 4.

Changes in bulk leaf osmolality and chlorophyll concentration correlated with growth in Arabidopsis under serpentine conditions. Osmolality (a) and chlorophyll concentration (b) in bulk leaf extracts and shoot growth rate (c) in basal nutrient solution (BNS)-grown (closed bars) and low-calcium solution (LCS)-grown (open bars) Arabidopsis lines. Plants were 5 wk old and grown for an additional 7 d in BNS or LCS. Osmolality of Arabidopsis bulk leaf extracts performed on leaf number eight of plants, with the remainder of the shoot material freeze-dried and used for analysis of chlorophyll concentration (mean ± SEM.; n = 6 independent plants per treatment, across duplicate experiments). Changes in shoot biomass are shown as mean growth rate (mean FW gain d–1 (g FW d−1) ± SEM; n = 6 plants per treatment). Lettering indicates significance categories (< 0.05) compared with Col-0 (annotated as ‘a’) within BNS and LCS treatments.


Cell-specific correlation of transporter expression with elemental compartmentation is a promising tool in contemporary forward genetics (Conn & Gilliham, 2010; Conn et al., 2011). Here we identified a suite of Mg2+ transporters enriched in the mesophyll that enable the accumulation of greater vacuolar Mg ([Mg]vac) than in the epidermis. Two of these transporters, AtMRS2-1 and AtMRS2-5, were localized to the tonoplast. T-DNA insertion lines accumulated less vacuolar Mg and displayed reduced growth and perturbed transcriptomes under serpentine conditions.

As with Ca, Mg is highly accumulated within transpiring organs of dicotyledonous plants (Broadley et al., 2008) and sorts preferentially to the mesophyll as measured by cryoSEM and SiCSA (Figs. 1a,b; Conn et al., 2011). SiCSA/XRMA was chosen preferentially over cryoSEM/XRMA to determine concentrations in vacuoles as a result of its > twofold higher sensitivity for Mg than on fully hydrated Arabidopsis leaf cross-sections (Fig. S2). Magnesium was detected at higher concentrations than Ca in the epidermis of Arabidopsis, a possible result of the presence of greater Mg uptake capacity of the epidermis, specifically the high expression of AtMHX1 (Fig. 1c). Elbaz et al. (2006) showed that the transcript for the AtMHX homologue in the nickel hyperaccumulator A. halleri (AhMHX) was constitutively higher than AtMHX in A. thaliana, yet was unresponsive to the metal status of the plant, or altered Mg growth conditions, implicating a post-transcriptional level of control. We confirmed that AtMHX1 was unresponsive in Col-0, mrs2 and cax1/cax3 T-DNA lines in standard and serpentine conditions, despite differences in mesophyll [Mg]vac. While it is widely held that the single protein encoded by AtMHX1 is responsible for the majority of Mg loading into the vacuole (Shaul et al., 1999), we show that AtMRS2-1 and AtMRS2-5 may contribute significantly to mesophyll [Mg]vac based on both their mesophyll-specific expression profile under standard growth conditions (Fig. 1a) and the 33% reduction in [Mg]vac in T-DNA insertion lines under serpentine conditions (Fig. 3b). Furthermore, the increase in [Mg]vac under serpentine conditions was positively correlated with expression of both AtMRS2-1 and AtMRS2-5, being enriched in LCS-grown Col-0 and cax1/cax3 lines, the latter accumulating more mesophyll [Mg]vac than Col-0 (Fig. 3b,c). Our results contrast with the overall unresponsiveness of these genes in the reports of Visscher et al. (2010) and Hermans et al. (2010a, 2010b). However, these studies utilized microarray analysis of roots and/or shoots to determine transcriptional reprogramming in response to Mg excess and depletion, respectively, from the early response (45 min) to the 7 d response (as measured in this study). This is the first report to detect alterations in abundance of specific MRS2s by qPCR in plant leaves in a Mg concentration-dependent manner.

Our interest in the role of these membrane-bound Mg2+- transporters in cell-specific vacuolar accumulation led us to exclude further investigation of AtMRS2-4 on the basis of its subcellular localization within the mitochondrial membrane (S. Conn, unpublished). Of the three remaining transporters enriched in the mesophyll, AtMRS2-10 has been previously localized to the plasma membrane (Li et al., 2001). The localization of AtMRS2-5 had yet to be demonstrated by fluorescent protein fusions but has been found within tonoplast (Whiteman et al., 2008) and plasma membrane proteomes (Alexandersson et al., 2004) of Arabidopsis leaves. Furthermore, Schock et al. (2000) were unable to detect fluorescence using a GFP fusion with full-length AtMRS2-1 or AtMRS2-2, while C-terminal truncations of both were found to be diffuse within the cytosol. We observed cytosolic fluorescence when YFP was fused to the N-terminus of AtMRS2-1, AtMRS2-5 or AtMRS2-10. We considered this a probable result of proteolytic cleavage of an MRS2 inherent signal peptide, given a predicted cleavage site between 16 and 19aa by SignalP version 3.0 (Bendtsen et al., 2004) (Fig. S4). Use of the more stable YFP protein and introduction of a short, uncharged spacer between the MRS2 and the YFP at the C-terminus in this report saw the localization of both AtMRS2-1 and AtMRS2-5 to the tonoplast and AtMRS2-10 to the plasma membrane, as previously seen (Li et al., 2001), validating our method of subcellular localization.

These are the first members of the MRS2 family to be localized to the tonoplast in any species and to show a significant perturbation in elemental concentration of the plant. While knockouts of an endoplasmic reticulum (ER)-targeted member (AtMRS2-7) showed perturbed growth of seedlings on low Mg concentrations (50–100 μM), the mrs2-1 and mrs2-5 lines used in this study were unaffected on this media (Gebert et al., 2009). This suggests that the ability of these lines to sequester Mg into the ER is unaffected. In light of the reciprocal transcriptional complementation of AtMRS2-1 in mrs2-5 and AtMRS2-5 in mrs2-1 (Fig. 3c), it would appear both proteins are capable of maintaining [Mg]vac.

While T-DNA insertion lines of AtMRS2-1, AtMRS2-5 and AtMRS2-10 did not show significant perturbation in any element under standard growth conditions (Fig. S6; PiiMS database), knockout plants (mrs2-1 (2-1a and 2-1b) and mrs2-5) fed higher [Mg] in either detached leaves (Fig. 3a), or whole plants (Fig. S6) accumulated significantly less Mg in the mesophyll than either Col-0 or mrs2-10. In addition, both Ca and K were found to adjust for the large increase in [Mg]vac (c. 40 mM), by decreasing in concentration by c. 20 and c. 35 mM, respectively (Fig. 3a). Taking univalent anions to balance charge of the cations, the total osmotic adjustment caused by reduced Ca2+ and K+ would be about equal to the gain from increased Mg2+. This adjustment, presumably to maintain osmotic pressure, was also observed in a growth solution with very low Ca2+ concentrations (LCS; 25 μM aCa) (Fig. 3b). Under these conditions, [Mg]vac increased along with [K]vac, showing the versatility of the K response and the large contribution of Mg to this adjustment. However, this adjustment appears to overcompensate, since leaf osmolarity increased under LCS conditions (Fig. 4a). In the mrs2-1 lines and mrs2-5, increased [K]vac compensated for the reduced accumulation of [Mg]vac; however, the degree of compensation based on an osmotic equivalent (using univalent charge balance) for the reduced [Ca]vac was very similar to that of Col-0 and mrs2-10 (i.e. c. 140 mosmol kg−1). Overcompensation of osmolarity (bulk leaf sap) was greater in these lines and correlated with much higher [K]vac. While K is known as a major compensating osmoticum for nutrient excess and deficiencies (Leigh et al., 1986; Conn & Gilliham, 2010), this role can also now be attributed to Mg for loss of [Ca]vac. Furthermore, the positive correlation between leaf [Mg] and both AtMRS2-1 and AtMHX expression (but not AtMRS2-10) combined with their inverse correlation with leaf [Ca] across Arabidopsis ecotypes further strengthens this balancing between the two divalent cations (Fig. S4). It must be noted that probes for AtMRS2-5 transcript are absent from Affymetrix Genechip® probe arrays and thus this correlation was unable to be made using the data from Lempe et al. (2005).

Growth in LCS was retarded for all plants, except for the cax1/cax3 line. This growth retardation may be related to the decrease in leaf chlorophyll concentration, perhaps related to more Mg being sequestered to the vacuole from the cytoplasm to compensate for reduced [Ca]vac, with less available for chlorophyll biosynthesis within chloroplasts. The increased leaf sap osmolarity which is also associated with lower growth, may be a consequence of reduced cell expansion while solute accumulation may not be down-regulated sufficiently to completely compensate. This phenomenon has been observed in barley plants grown under saline conditions where increased leaf osmolality was correlated with lower growth rates (Fricke & Peters, 2002; Fricke, 2004). As soluble sugars are not significantly affected by growth in LCS, it is likely that the osmolality increase in plants in LCS results primarily from changes in vacuolar K+, Mg2+ and Ca2+. Similarly, the observed increase in osmolality in mrs2-1 and mrs2-5 T-DNA lines correlates with an increase in K+ concentration over wild-type leaves (Figs 3b, 4a).

Complementation of the lower mesophyll [Mg]vac phenotype under serpentine conditions for mrs2-1 and mrs2-5 was only seen with transformation of the ‘knocked out’ gene, not its tonoplast-localized homologue, that is, AtMRS2-1 and AtMRS2-5, respectively (Fig. 3d). There is 63% protein similarity between AtMRS2-1 and AtMRS2-5, suggesting that each may play a unique role in vacuolar Mg sequestration. We also report an increase in epidermal [Mg]vac when either AtMRS2-1 and AtMRS2-5 (shown to be absent from the epidermis) were transformed into the T-DNA lines under the control of the ubiquitous 35S promoter, showing the importance of cell-specific expression of these genes in controlling elemental accumulation patterns (Fig. 3d). As mentioned earlier, T-DNA insertion lines for AtMRS2-1 and AtMRS2-5 show reciprocal transcriptional complementation (Fig. 4c), indicating functional compensation under standard conditions, but incomplete compensation under serpentine conditions. While this may reflect a difference in affinity for Mg2+ between the two transporters, we cannot discount that post-transcriptional or post-translational levels of control contribute to this situation.

Absence of AtMRS2-10 transcript from leaf epidermal cells suggests that this protein is not the only route for Mg2+ influx into leaf cells. AtCNGC10, one of the 20 cyclic nucleotide-gated ion channels in Arabidopsis, has recently been shown to transport Ca2+and Mg2+ (Guo et al., 2010) and has been localized to the plasma membrane of epidermal and mesophyll cells in Arabidopsis leaves (Christopher et al., 2007). AtCNGC10 transcript was present in both cell types in our microarray analysis (data not shown) and thus may contribute to mesophyll Mg2+ influx along with AtMRS2-10. However, vacuolar sequestration of Mg seems independent of AtMRS2-10-mediated Mg2+ influx across the plasma membrane, given no alteration in mesophyll [Mg]vac even under serpentine conditions (Fig. 3a,b). This points to the importance of the tonoplast in control of plasma membrane fluxes of Mg2+, as noted previously for chloride (Cl) (Laties et al., 1964), K+ (MacRobbie, 1971), Ca2+ (Conn & Gilliham, 2011) and other ions (Conn & Gilliham, 2010).

Together we show that control of vacuolar magnesium storage is an important process for maintaining plant growth and that differential, cell-specific mechanisms play important roles in this homeostasis. Understanding this sequestration and the plants’ responses to enrichment of Mg will be important for biofortification efforts in transpiring leafy vegetables and other phloem-fed tissues (White & Broadley, 2009).


The authors would like to acknowledge the assistance of Lyn Waterhouse and Ken Neubauer (Adelaide Microscopy) in the confocal fluorescence microscopy and XRMA; Kendal Hirschi for supplying cax1/cax3 lines, Ute Baumann and Andreas Schreiber for assistance with microarray analysis, and Wendy Sullivan for soluble sugar analysis. This work was supported by an ARC grant awarded to R.A.L., B.N.K. and S.D.T., an Australian Professorial Fellowship awarded to S.D.T. and University of Adelaide, Faculty of Science grants awarded to R.A.L. and M.G.