Natural variation among Arabidopsis thaliana accessions in tolerance to high magnesium supply

High magnesium (Mg2+) in some extreme serpentine soils or semi-arid regions is an important factor affecting crop growth and development. Specific loci that form the genetic framework underlying high Mg2+ homeostasis, however, are not well understood. By using GWA mapping on 388 accessions of Arabidopsis thaliana selected from a worldwide collection and genotyped at approximately 250,00 SNPs, we successfully identified 109 and 74 putative genetic regions associated in nutrient traits under normal (1,000 µM) and high Mg2+ (10,000 µM), respectively. Above 90% SNPs associated with nutrients including Mg2+ and only two SNPs shared between normal and high Mg2+. A single strong peak of SNPs associated with Ca concentration corresponding to candidate gene At1g60420 ARABIDOPSIS NUCLEOREDOXIN (AtNRX1) under high Mg2+ was further determined. Compared with wildtype, mutants of Atnrx1-1 and Atnrx1-2 supplied with high Mg2+ had higher Ca concentrations in the plant, and higher cytosolic Ca2+ concentrations during root elongation, as well as higher fresh weight and lateral-root number. This suggests that AtNRX1 was a critical gene negatively regulating Ca uptake under high Mg2+ conditions. The discovery could help to breed/select crops that can adapt to high-Mg2+ soils such as serpentine soils (high ratio of Mg2+: Ca2+) or Mars soil with high levels of magnesium sulfate.

Magnesium (Mg) is the 8 th most abundant mineral element on earth and the fourth abundant mineral element in plants following nitrogen (N), potassium (K) and calcium (Ca) 1 . Magnesium in soils originates from source rock material containing various types of silicates and carbonates. In extreme serpentine (high magnesium: calcium) soils 2 and in semi-arid regions, where water stress conditions 3 can lead to Mg accumulation. In addition, Mars regolith (unconsolidated surface material) is a potential medium for plant growth in bioregenerative life support systems. However, analyses by the Mars Exploration Rover Landers at Meridiani Planum and Gusev Crater have also suggested there are particularly high levels of magnesium sulfate minerals in outcrops and soils [4][5][6] .
Although plants rely on a sufficient supply of Mg 2+ for normal growth and development, excessive Mg 2+ accumulation often causes toxicity to plant cells. Plants grown in high-Mg 2+ soils like serpentine or Mars soils may alleviate Mg toxicity by limiting internal Mg accumulation, Mg excretion from leaves and/or increasing plant tolerance to high Mg 2+ concentrations via altering leaf size, sclerophylls, stature and root systems 7,8 . Previous reciprocal transplant experiments between serpentine populations (or species) and their spatially adjacent populations (or sister species) growing on serpentine soils demonstrated unequivocally that adaptation to serpentine soils had a genetic basis [9][10][11][12] . Several quantitative trait loci that control contrasting adaptive traits in serpentine-tolerant Microseris douglasii and serpentine-intolerant M. bigelovii have also been identified 13 . A high degree of polymorphism has been reported among plant species in terms of growth tolerance to serpentine soils [14][15][16] .
This present study employed genome-wide association studies (GWAS), an approach that has been applied successfully to identify many different traits ranging from genetically simple traits to more complex features 17 , to identify candidate genes likely to be involved in adaptation to high Mg 2+ , based on 388 accessions of Arabidopsis Arabidopsis cultivation. The 388 accessions of Arabidopsis thaliana were grown under 2 Mg 2+ treatments, which were achieved by altering the concentration of MgSO 4 in the basal medium. To minimise maternal effects prior to phenotyping, natural accessions were grown for one generation during 2015 under controlled greenhouse conditions at Zhejiang University (N30°18′25, E120°04′54). For surface sterilization, seeds of various accessions were placed for 1 h in open 200-μL PCR tubes in a sealed box producing chlorine gas generated from 13 mL of 10% sodium hypochlorite and 350 μL of 37% hydrochloric acid. Sterile seeds were then put on the surface of 30 mL agar media, containing 1.2% (w/v) agar and 0.6% (w/v) sucrose (A-1296; Sigma-Aldrich; http://www. sigmaaldrich.com) in 10 × 10-cm 2 plates with grid schematic engraved below the plate. Plates were positioned in racks and oriented in a vertical position, and were kept at 4 °C for 48 h in the dark for seed stratification. Thereafter, the racks containing the plates were transferred to a growth chamber under a 10-h light/14-h dark photoperiod at constant temperature of 22 °C, 60% relative humidity and light intensity of 120 µmol photons m −2 s −1 . Twelve seedlings were grown in each plate and each treatment received at least four independent replicates. The racks were removed to the image acquisition room once per day and then immediately returned to the growth chamber. Throughout the experiments, the plate position within the box and box position in the growth chamber were re-randomized daily. For assays on agar plates, studies were performed on 8-d-old plants; that is, at an early stage of their stability and homogeneity growth phase. Moreover, many previous publications adopted 6-to 8-d-old Arabidopsis seedlings under stresses for determining Arabidopsis morphology, physiology, and development as well as responses to nutrient stresses 29-31 . Mg 2+ Treatments. The basal medium (Normal Mg 2+ , N), which was used as control, contained (µM) 1500 KNO 3 , 500 NaH 2 PO 4 , 1000 CaCl 2 , 250 (NH 4 ) 2 SO 4 , 1000 MgSO 4 , 25 Fe-EDTA, 10 H 3 BO 3 , 0.5 MnSO 4 , 0.5 ZnSO 4 , 0.1 CuSO 4 and 0.1 (NH 4 ) 6 Mo 7 O 24 . The normal and high Mg 2+ treatments contained 1,000, and 10,000 µM MgSO 4 , respectively, which was in accordance with Mg 2+ concentrations in soil solutions 27 . 10,000 μM Mg 2+ was selected in the present study for the quantitative analysis of the natural variation of Arabidopsis in response to high Mg 2+ stress because this concentration had an adverse effect on growth without toxic symptoms of most natural accessions 27,32 . Besides, the concentrations of Mg 2+ in the control were 1000 µM, that has been adopted for Arabidopsis growth by many plant biologists [33][34][35] . Though the concentration of SO 4 2− increased 10 times comparing with that in the normal medium, SO 4 2− had hardly affected morphogenesis of Arabidopsis 27,36 . The pH of the growing media was adjusted to pH 5.8 with MES (N-morpholino) ethane-sulphonic acid)-KOH buffer before autoclaving. Phenotype analysis. Each line had four biological replicates for each treatment. Seeds that had not germinated at day 6 were discarded from further analysis, resulting in approximately 3-4 seedlings analyzed per genotype per treatment. The 21 traits under each treatment were classified into 2 categories: 7 morphological traits and 14 nutrient traits. Seedlings were photographed with a high-resolution digital camera (Sony RX100, Japan) daily for determination of root and shoot growth. Root or shoot biomass, and the number of days from seeding to emergence (>50% seedlings having the first radicle or cotyledon), were recorded. Meanwhile, photographs after 8 days of treatment were analyzed and quantified for phenotype using the public domain image analysis program Image J version 1.43 (http://rsb.info.nih.gov/ij/) 27 . The length of primary roots, rosette diameter and epicotyl were determined across the median seedling using Image J. Lateral-root number was determined by counting the number of true roots (>1 mm long lateral root primordia) per primary root. For each condition, an independent sample accession consisted of four plants at the same growth stage, and the time of sampling was the end of the light period of day 8.
Genome-Wide Association (GWA) mapping, SNP selection, and statistics. The GWA mapping was performed on the values of epicotyl length, rosette width, primary root length, and lateral root number of the median seedling of each accession. For traits of and nutrients, means values of 4 biological replicates were used. We linked 21 traits under normal or high Mg 2+ (Supporting Information Table S1) to published genomic data on accessions from a 250 K SNP chip. The GWA analysis was performed in the GWAPP web (http://gwas. gmi.oeaw.ac.at/) interface using the mixed model algorithm method 37 . A kinship matrix was generated using the identity in state of SNPs between each pair of accessions 38 . The P-value above bias due to population stratification was evaluated with Q-Q plots. Candidate loci were selected based on the minor allele frequency, LOD score and trait-associated SNPs and their position in the genome to pinpoint candidate genes within 5 kb up-and down-stream of the identified SNP (Supporting Information Table S2). The putative candidate genes corresponding to each SNP are listed in Supporting Information Table S3. For each of the candidate genes, the annotations were retrieved from TAIR10 (Arabidopsis.org).
Extraction of total RNA and quantitative PCR. Samples were collected from 8-day-old Arabidopsis after grown at normal and high Mg 2+ treatments. For each condition, four samples were taken on independent seedling materials from eight different plants at the same growth stage and the time of sampling was the end of the light period of day 8. Total DNA was extracted by Plant DNA Isolation Reagent (Takara) from about 80 mg of fresh seedling tissues of four independent biological replicates. Two homozygous T-DNA insertion lines (SALK_100357C, Atnrx1-1) and (SALK_113401C, Atnrx1-2) was conducted by PCR with gene-specific primers SALK_100357C LP: 5′-AACCCAACCATCTTTGGACTC-3′; SALK_100357C RP: 5′-TCAAGACTTCAAGACCAAGCC-3′ and T-DNA border primer LBb1.3:5′-ATTTTGCCGATTTCGGAAC-3′; SALK_113401C, LP: 5′-ATCCAC TTTTGTCGTTGAACG-3′; SALK_113401C RP: 5′-CGATCGCAACTTCTTCTGATC-3′ and T-DNA border primer LBb1.3:5′-ATTTTGCCGATTTCGGAAC-3′; Total RNA was extracted by RNAisoPlus (Takara, Otsu, Shiga, Japan) from about 70 mg of fresh seedling tissues of four independent biological replicates. All RNA samples were checked for DNA contamination before cDNA synthesis. cDNA was synthesized, and possible residual genomic DNA contamination was verified as described in our previous study 39 . The mRNA levels of AT1G60420 in two lines of SALK_100357C and SALK_113401C were detected by the SYBR Green RT-PCR kit (Takara) with following pairs of gene-specific primers: AT1G60420, fw: 5′-TTCTCGGCCCTGATGGAAAAACC-3′, rv: 5′-GCTTCATTCCGCTCCTTTATCTGCT-3′ A pair of housekeeping gene of UBQ10 (fw: 5′-GGTTCGTAC CTTTGTCCAAGCA-3′, rv: 5′-CCTTCGTTAAACCAAGCTCAGTATC-3′) was used for a control PCR. Melting curve analysis and gel electrophoresis of the PCR products were used to confirm the absence of non-specific amplification products. Relative expression levels were calculated by subtracting the threshold cycle (Ct) values for UBQ10 from those of the target gene (to give ∆Ct) and then calculating 2 −∆Ct , where C t refers to the cycle number at which the fluorescence rises above the set threshold qPCR. UBQ10 was chosen as the housekeeping reference 40 . Phylogenetic and Statistical Analyses. DNA sequences were obtained from the 1001 Genomes project (www.1001genomes.org). All statistical analyses were conducted with DPS software (Stirling Technologies Inc., China). Means were compared by the t test or the Fisher's least significant difference test at P = 0.05.

Results
Distribution and correlation of different 21 traits under normal and high Mg 2+ . Distribution of 388 Arabidopsis accessions collected from the wild and available in the stock center or soon-to-be-released collections is shown in Fig. 1A. In general, high Mg 2+ supply (10,000 μM) had an adverse effect on the growth of most natural accessions. This high Mg 2+ concentration induced a large degree of phenotypic variation and thus was selected for the quantitative analysis of the natural variation of Arabidopsis in response to high Mg 2+ stress (Fig. 1B). A subset of 388 accessions (250 K SNPs) was characterized phenotypically for multiple traits, including root emergence (RG), shoot emergence (SG), primary root length (PR), lateral root number (LR), epicotyl length (EL), rosette width (WD), fresh weight (FW) of the plants and fourteen nutrient traits, including absorption per plant and concentration of K, Ca, Mg, S, Fe, Mn and Na under normal and high Mg 2+ supplies (see Table 1 and 'Phenotype analysis' in Materials and Methods).
Before performing the quantitative genetic analyses, we assessed the distributions of all traits based on all raw data (Supporting Information Table S1). The phenotypic analysis of the 388 accessions showed that the frequency distribution of all traits was close to normality ( Fig. 1C; Table 1). Further detailed analysis showed that Mg 2+ ) showed a great right-skewed probability density with the right-skewed being greater for Mg-C-H than that for Mg-A-H, suggesting that high Mg 2+ supply increased Mg concentration of most Arabidopsis accessions at least in the early phase of development. However, only slight skewness and kurtosis was observed for S absorption and concentration (Fig. 1C). Distribution of RW, EL, PR, FW, Ca-A (Ca absorption), Ca-C (Ca concentration), Na-A (Na absorption) and Na-C (Na concentration) presented a standard normal distribution but with a sharp peak upon high Mg 2+ as compared with normal Mg 2+ (Fig. 1C), indicating that the high Mg 2+ treatment resulted in a more focused response in those traits. Notably, the right side of the curve of absorption and concentration of Ca, Fe and Mn under both normal and high Mg 2+ fitted a normal distribution, but the left-hand side deviated significantly from the normal distribution, displaying a decapitation tail (Fig. 1C).
Data of the traits are close to normal distribution reflecting the homogeneity of the investigated population. However, what was the response of each individual domain to high Mg 2+ ? According to the date (Supporting  Information Table S1), the effects of high Mg 2+ on all phenotypes can be divided into two types. First, high Mg 2+ had a similar promoting effect on LR, Mg-A, Mg-C, S-A and S-C of most accessions. Second, high Mg 2+ enhanced traits of PR, EL, RW, FW and absorption and concentrations of K, Ca, Fe and Mn in the accessions with low value while reduced them in the accessions with high value as compared with normal Mg 2+ supply. Furthermore, the initial comparison of 21 plant traits between normal and high Mg 2+ was performed using statistical analysis by a paired two-tailed Student t test, with a P-value < 0.01 (Table 2). Accordingly, the above first type traits were consistently changed by high Mg 2+ as compared with normal Mg 2+ . Apart from this, the above second traits were inconsistently altered by high Mg 2+ ( Table 2, Fig. 2A,B).
Pearson's correlation coefficients (r 2 ) were used to examine the natural variation within the population of quantified traits between normal and high Mg 2+ conditions ( Table 2). The RG and SG showed positive correlations between normal and high Mg 2+ treatments (r 2 = 0.53) ( Table 2). A weak correlation was observed for nutrient traits, especially Ca, Fe and Mg concentrations between normal and high Mg 2+ treatments (r 2 < 0.2), suggesting different physiological functions of nutrients under different Mg 2+ supplies.
Cluster of Pearson's correlation coefficient between phenotypes (Heat map) showed a negative correlation between Ca-C-H, Fe-C-H and Na-C-H and Mg-C-H (Fig. 2C), suggesting that Mg 2+ would compete for uptake with Ca 2+ , Fe 2+ and Na 2+ in most Arabidopsis accessions when supplied with high Mg 2+ . Increased Mg 2+ absorption would decrease the uptake of these ions and hence plant growth. However, Mg-A, Mg-C, S-A, S-C, K-A and K-C were positively correlated only under high Mg 2+ (Fig. 2C), indicating a common mechanism or genetic pathway between Mg 2+ , S and K under high Mg 2+ (Fig. 2C). In contrast, there were highly significant positive correlations between PR, RW, LR, FW as well as S-A, Mg-A and K-A of the phenotypes under normal but not high Mg 2+ supply (Fig. 2C).
GWAS analysis for nutrient metabolism in response to high Mg 2+ . The phenotypic data collected from normal and high-Mg 2+ conditions (Supporting Information Table S1) were used as an input for genome-wide association study (GWAS). All SNPs were converted to homozygote or heterozygote according to the genotype, and tri-allelic or tetra-allelic SNPs were removed. Candidate loci were selected based on the minor allele frequency, LOD score and number of significantly associated SNPs within one locus. Using −log 10 (p) > 6.15 as the nominal genome-wide significance threshold, the GWASs had 109 hits in total, 32, 10, 25, 18 and 24 located  (Fig. 3A,B). There were 2 shared hits between normal and high Mg 2+ supplies (Fig. 3C). Under normal Mg 2+ supply, 41 and 22 loci were associated with Mn absorption (Mn-A-N) and concentration (Mn-C-N), and accounted for 7.47% and 2.67% of the total phenotypic variance, respectively. 17 and 21 loci were associated with Fe absorption (Fe-A-N) and concentration (Fe-C-N), and accounted for 3.60% and 5.04% of total phenotypic variance individually (Fig. 3B  accounted for 2.93% and 12.04% of total phenotypic variance individually. Thirty three and 11 loci were associated with Fe absorption (Fe-A-H) and Fe concentration (Fe-C-H), which accounted for only 5.07% and 2.60% of total phenotypic variance individually (Fig. 3B, Supporting Information Table S2). The GWAS analysis confirmed that only 2 common loci were detected differed between normal Mg 2+ and high Mg 2+ (Fig. 3C). Single nucleotide polymorphisms (SNP) 1_3884526 (Chromosome 1, 3884526 bp) was found to associate with SG-H, RG-H, RG-N and Na-C-N. It had a Minor allele frequency (MAF) of 0.0633, and is within the OST3A, a gene related to oligosaccharyl transferase complex/magnesium transporter family (Supporting Information  Table S2). SNP 4_2080234 was associated with Mn-C-H and Fe-C-N. It had a MAF of 0.050 and is within AT4G042960 and AT4G04293, both transposable element genes.
No significant associations were observed for the Mg 2+ (Mg-A) and Ca absorption (Ca-A) under normal Mg 2+ , but 2 loci were identified to associate with these traits under high Mg 2+ . Furthermore, no significant associations were found for Mg 2+ (Mg-C) and S concentration (S-C) under high Mg 2+ , whereas 7 and 6 loci were identified to associate with these traits under normal Mg 2+ (Fig. 3B). No significant associations for K absorption (K-A) were identified under normal and high Mg 2+ .
We speculated that the absorption and concentration of nutrients should share many candidate loci associated with the same element under the same Mg 2+ level. However, this was not the case in our data set, as no common candidate loci were observed to associate with absorption and concentrations of K, Ca, Mg, S and Na under either normal or high Mg 2+ (Supporting Information Table S2). Though common loci were found between the absorption and concentration of Fe or Mn in plants grown under both normal and high Mg 2+ , these only accounted for a small part of the variance in attitudes. While six common loci were detected between the traits of Fe-A-H and Fe-C-H, only 2 common SNPs were for Fe-A-N and Fe-C-N. Likewise, 10 common loci were associated with Mn-A-H and Mn-C-H and 6 common loci associated with Mn-A-N and Mn-C-N (Fig. 4A,B). Furthermore, manhattan plot showed the detailed distribution of the GWAS hits for Fe and Mn (Fig. 4C,D).

Characterization of Atntrl mutants in response to high Mg 2+ treatment. GWAS is conducted to
identify causative/predictive factors for a given trait, or to determine aspects of the genetic architecture of the trait (i.e. the number of loci that contribute to the phenotype) 41 . A single strong peak of SNPs 22262848 between position of 22262147 and 22266184 bp on chromosome, which is associated with Ca concentration, was identified under high but not normal Mg 2+ (Fig. 5A,B). This SNP accounted for 0.326% of the total variance of an estimated 2.20% of the total variance in Ca concentration under high Mg 2+ . It contained three neighboring significant SNP hits (−log 10 (p) > 6.9) corresponding to only a candidate gene At1g60420 (AtNRX1) (Fig. 5C), and was calcium locus CD-domain containing protein.
To validate the GWAS results in the present study, the above putative significant candidate gene was determined through a collection of well-characterized mutants (Atnrx1-1 and Atnrx1-2) and physiological and morphological comparisons in mutants and wild-types. Before this, we further verified that the mutant stocks were homozygous T-DNA insertion lines, by PCR with gene-specific primers LP, RP and LB. As expected, the results confirmed that both Atnrx1-1 and Atnrx1-2 were only mutation in At1g60420 (Fig. 6A). Moreover, the expression of AT1G60420 were greatly reduced in two mutants under both normal and high Mg 2+ conditions as compared to wildtype (Fig. 6B). Compared with normal Mg 2+ , high Mg 2+ increased the fresh weight of Atnrx1-1 and Atnrx1-2 but did not affect that of wildtype. It is worth mentioning that compared with wildtype, fresh weight of two mutants was not changed under high Mg 2+ but was reduced under normal Mg 2+ (Fig. 6C). It is suggested that high Mg 2+ reduced wildtype growth but had little effect on the growth of Atnrx1 mutant.
As shown in Fig. 6D, a tip-focused [Ca 2+ ] c gradient was formed during the initiation and elongation of roots of both wildtype and mutants, and such a tip-focused gradient of [Ca 2+ ] c was weak under normal Mg 2+ . In contrast, under high Mg 2+ , Atnrx1-1 and Atnrx1-2 but not wildtype formed a strong tip-focused [Ca 2+ ]c gradient in the initiation and elongation of root of the mutant (Fig. 6D). Furthermore, high Mg 2+ greatly enhanced concentration of Ca in Atnrx1-1 and Atnrx1-2 but decreased it in wildtype as compared with normal Mg 2+ (Fig. 6E). Meanwhile, when compared with wildtype, Ca concentration of high Mg-fed Atnrx1-1 and Atnrx1-2 mutants was greater while that of normal Mg-fed Atnrx1-1 and Atnrx1-2 mutants was similar (Fig. 6E). As expected, high Mg 2+ enhanced Mg concentrations in both mutants and wildtype (Fig. 6F).
Compared with the wildtype, both Atnrx1-1 and Atnrx1-2 had shorter primary root and less number of lateral roots under normal Mg 2+ but had greater length of primary roots and a similar number of lateral roots under high Mg 2+ (Fig. 6G,H). Under high Mg 2+ , there is no significant difference in K, Fe, Mn and S concentrations in plants between the two mutants and wildtype, but the mutant had a higher Na content (Data not shown). These results indicate that AtNRX1 is a special and critical negative gene regulating Ca uptake and probably the plant growth under high Mg 2+ stress.

Discussion
To adapt to the constantly high availability of Mg 2+ in the environment, plants have evolved mechanisms of reduced ion uptake and/or sequestering excessive ions in the vacuole or developing higher Mg 2+ requirement for the maximum growth, which finally improves the tolerance to the stress of certain elements. By linking the mapping with reverse genetic experiments, we identified the causal gene and established that polymorphisms at AtNRX1 were the major genetic determinant for the variation in Ca concentration in this global A. thaliana population (Fig. 6). Atnrx1 mutants have most of the phenotypes that would associate with the tolerance to serpentine soils, including survival in solutions with a low Ca 2+ : Mg 2+ ratio; requirement for a high concentration of Mg 2+ for maximum growth; greater biomass and more lateral roots, reduced leaf tissue concentration of Mg; and poor growth performance on 'normal' levels of Mg 2+ .
SCienTifiC REPORTS | (2018) 8:13640 | DOI:10.1038/s41598-018-31950-0 Shoot and root morphology variation in response to high Mg 2+ . In order to understand molecular mechanisms of plant adaptation to high Mg 2+ stress, we grew 388 wild Arabidopsis accessions at normal and high Mg 2+ concentrations, and observed a substantial quantitative variation in plant growth between these accessions ( Fig. 1; Table 1). We also showed a diversity in shoot and root morphology across two levels of Mg 2+ supply and among these natural accessions. High Mg 2+ supply resulted in an overall reduction in primary root length, lateral root number, epicotyl length and fresh weight but its effect on the rosette width varied among the accessions ( Fig. 2; Table 2). The phenotypic data indicate that Arabidopsis thaliana has evolved some capacities to adapt to high Mg 2+ stress and that the number of lateral roots of seedlings is a suitable index of high-Mg 2+ responses.  Information Table S1) and GWAS, this study showed that a great extent of variation in the length of primary and lateral roots, rosette growth and hypocotyl length among the accessions tested was strongly Mg-dependent. As we known, the growth of primary and lateral roots, hypocotyl and rosette is determined by a number of extrinsic and intrinsic factors in addition to Mg 2+ supply. The significant associations with the leaf or root growth-related phenotypes according to −log 10 (p) > 6.15 reflects the high complexity of the underlying genetic architecture or that these traits were not influenced by any of the SNPs alone. This notion is supported by the presence of obvious sub-threshold peaks associated with shoot and root morphology under both normal and high Mg 2+ supply. Meanwhile, the low heritability of investigated traits in the present study suggested a weak genetic component determining the observed phenotypic variability. Hopefully, it is known that many traits of Arabidopsis are polygenic with small effect size, and that increasing the sample size would improve the power to recover meaningful associations 42 . Therefore, increasing sampling size was likely to underestimate the range natural variation in photomorphogenic responses or growth-related pathways and processes that strongly interact with the high Mg 2+ stress.

Arabidopsis showed remarkable natural genotypic variation for ionomics under high Mg 2+ .
The present study showed that most of the SNPs were derived from the nutrient concentration and uptake, SNPs differed between normal and high Mg 2+ , the effective number of SNPs (−log 10 (p) > 6.15) was also different ( Fig. 3; Supporting Information Table S2). For example, positive SNPs associated with Ca-A and Mg-A were located only under high Mg 2+ (Fig. 3B). Likewise, no significant associations were observed for either Mg-C and S-C under high Mg 2+ but seven and six significant SNPs were associated with them under normal Mg 2+ (Fig. 3), indicating that different genetic systems were responsible for Mg 2+ and S uptake under normal and high Mg 2+ . Meanwhile, the SNP significance tests showed that different traits were generally associated with different genes or gene regions, but many of the traits also shared common genes or gene regions. The significant score >10 association was for the SNP on Mn under high Mg 2+ in a narrow region of chr3, where multiple SNPs were associated with different traits in Chr1, 3, 4 and 5 under normal Mg 2+ (Fig. 3).
Though not all cases, the chromosomal regions harbouring the significantly associated SNPs are coincided with the location of the gene(s) co-expressed with known-function genes that control these traits. Of the 105 loci under normal Mg 2+ , the majority of the loci were associated with Mn-A and Mn-C then with Fe-A and Fe-C. Among these, SNP 1_3953597, which is within the SS3, a gene encoding a starch synthase, is co-expressed with SPLICING ENDONUCLEASE 1 (SEN1) 43 . Likewise, SNP 3_19424421 of p-value 12.55, associated with Mn-C under the normal Mg 2+ condition, had a MAF of 0.058 and was within gene AT3G52380 CHLOROPLAST RNA-BINDING PROTEIN 33 (CP33) that is co-expressed with SEN1. SEN1 gene was reported encoding an integral membrane protein homologous to Glycine max nodulin-21, and also to CCC1, a vacuolar Fe/Mn transporter of Saccharomyces cerevisiae of Arabidopsis 41 . Besides, SNP 1_210424, associated with Fe-C, had a MAF of 0.055 and was within gene FRO2 which encodes the low-Fe-inducible ferric chelate reductase. It is suggested to be the major Fe(III) chelate reductase in Arabidopsis 44 . The most SNP hits under high Mg 2+ were identified associations with Fe-A and Fe-C and then Mn-A and Mn-N. SNP 1_26930918 had a MAF of 0.074 and was located within AT1G71480 and AT2G33410 genes which are involved in nuclear transport factor 2 family gene, and are co-expressed with Fe transporter gene FERRIC REDUCTION OXIDASE 6 (FRO6) and 7 (FRO7) 45 and 1 (FRO1) 46 . Furthermore, SNP 2890675, associated with Fe-C under the high Mg 2+ on Chromosome 5, had a MAF of 0.074 and was located within AT5G09310 gene and co-expressed with Fe transporter gene FRO1 46 . Likewise, SNP 3_6977829 of P-value 8.40, associated with Fe-A, within gene AT3G20015 (ASPG2), and was co-expressed with H( + )-ATPASE 2 (AHA2), a main actor in Fe uptake and signaling 47 .
In addition, SNP 4_772594 associated with Mg + C under the normal Mg 2+ , had a MAF of 0.079 and was located within AT4G01800 ALBINO OR GLASSY YELLOW 1 (AGY1) gene which was co-expressed with MAGNESIUM CHELATASE I2 (CHLI2) while CHLI2 regulates the function of Mg 2+ chelatase 48 . It should be mentioned that SNP 1_21762851 of P-value 9.05, associated with S-C under the normal Mg 2+ , had a MAF of 0.090 and contained only one gene, AT1G51980, which encodes an insulinase and is co-expressed with SULFATE TRANSPORTER 1;2 (SULTR1;2) that is involved in sulfur metabolism 49 . As discussed above, changing the Mg 2+ availability in the soil would alter absorption and utilization of other elements, and transporters, channels, or/ and the genes that encode and regulate them in Arabidopsis. However, a few of the genes in proximity of the most significant associations were among the major known regulators of ion uptake or transport, highlighting the potential of GWAS to identify previously unknown regulators 50 .
Changes in nutrient availability influence plant vegetative growth and phenotypic traits of A. thaliana, but the effect may vary among populations or genotypes [51][52][53] . Decreases in nutrient availability result in phenotypic responses (such as smaller leaf areas, shorter primary root and reduced lateral root formation) that vary across A. thaliana genotypes.
A novel candidate gene AtNRX1 negatively regulates Ca concentration at high Mg 2+ supply.
Calcium is an essential plant macronutrient with key structural and signaling roles. Calcium ions (Ca 2+ ) act as an osmoticum within vacuoles, a stabilizing element in the membranes, a strengthening agent in cell walls, and a secondary messenger for a multitude of signals [54][55][56] . By using GWA mapping on the 388 A. thaliana accessions, we successfully identified a single strong peak of SNP 1_22263976/22264331/22262848 (Chromosome 1, 22263976 bp) associated with Ca concentration under high Mg 2+ (Fig. 5). Mutants of Atnrx1-1 and Atnrx1-2 supplied with high Mg 2+ had higher cytosolic Ca 2+ concentrations ([Ca 2+ ] c ) during root elongation, meanwhile produced higher biomass and more lateral roots as compared with wildtype (Fig. 6). This suggested that the increase in Ca concentration in mutants of Atnrx1-1 and Atnrx1-2 by high Mg 2+ cannot be attributed to increased biomass or photosynthesis alone, but also to the improved Ca nutrition of the plants through morphological, physiological and molecular responses to high Mg 2+ .
It has been known that AtNRX1 is involved in DC1 domain-containing protein, which is co-expressed with calcineurin B-like Ca sensor 1-protein kinase gene CBL-INTERACTING PROTEIN KINASE 23 (CIPK23) 57,58 . Study of Tang et al. 59 showed that tonoplast CBL-CIPK calcium signaling network regulated ion homeostasis and vacuolar sequestration of Mg 2+ , thereby protecting plants from Mg 2+ toxicity 59,60 . Besides, loss-of-function cax1 (gene encoding vacuolar H + /Ca 2+ ) mutations could produce phenotypes characteristic of plants adapted to serpentine 8 .
This locus has a MAF of 0.0712 and is within the AT1G60420 (AtNRX1), a gene involved in catalytic activity of NADP-NADPH and disulfide-reductase activity 61 . In special cases, toxic ROS molecules are largely accumulated under excess stresses and diverse enzymes played as ROS scavengers. Arabidopsis contains 3 NADPH-dependent thioredoxin reductases (NTRs) for scavenging ROS and Nucleoredoxins (NRX) are potential nuclear TRX found in most eukaryotic organisms 61 . Recently, excellent research presents evidence that activity of the plant stresses-inducible TRX superfamily member, NRX1 is necessary for the integrity of antioxidant systems 62 . Meanwhile, it is the fact that many publications indicated ROS produced by NADPH oxidase activates Ca 2+ channels at the plasma membrane, leading to an increase in Ca 2+ concentration and a tip-focused Ca 2+ concentration Data are means ± SD (n = 5). Means followed by a common letter within a root segment are not significantly different at P < 0.05. FW, fresh weight. gradient and subsequent signaling inherent to plant growth 32,[63][64][65][66] . Withal, our previous study also showed that high Mg 2+ could inactivate NADPH and the production of ROS 27 probably through regulating ABA-DELLA signaling 28,67 .
To sum up, it is likely that under high Mg, the wild-type plant is associated with cellular accumulation of ROS and redox imbalance, which triggers class of antioxidant enzymes activation and subsequently results in a decrease of intracellular ROS. ROS restrains ion channels and Ca 2+ concentration; While in the nrx1 mutants, ROS could not be effectively removed without protection of NRX1 due to NRX1 mutation, which leads to ROS accumulation. Then ROS activate Ca 2+ -channels at the plasma membrane and promote cytosolic Ca 2+ concentration. When both wild-type and nrx1 mutants under normal Mg conditions, the cell is normally controlled with H 2 O 2 and redox balance, thus the protective effect of NRX1 gene is not prominent, or maybe even little dedication.
Notably, Mg 2+ concentrations in Atnrx1-1 and Atnrx1-2 mutants was increased compared to that in the wildtype plants under high Mg 2+ supply. This implied that nrx1 mutants required for a high concentration of Mg 2+ for the maximum growth associating with tolerance to high Mg 2+ soils. The AtNRX1 gene is co-expressed with MAGNESIUM TRANSPORTER 4 (MRS2-3), a transport gene related to Mg 2+ nutrition 68 . These indicate that Mg 2+ transporter genes have some interactions with AtNRX1. However, there is no significant difference in K, Fe, Mn and S concentration in plants between the two mutants and wildtype, but the mutant had a higher Na content (Data not shown). These was consistent with other study showed that the cbl2 cbl3 double mutant was hypersensitive to a handful cations but not to Na + 59 . Further work will be directed to identify other key components in this novel of NRX-signaling pathway in regulation of Ca uptake and how the positive effects of AtNRX1 can be maximized through plant adaptive strategies to improve the tolerance to high Mg 2+ .
To sum up, our study suggests that AtNRX1 was a critical gene negatively regulating Ca uptake and probably improving the plant tolerance to high Mg 2+ soils. The discovery of the functions of AtNRX1 gene helps to breed/ select crops that can adapt to high-Mg 2+ soils such as serpentine soils (high Mg 2+ : Ca) and those in semi-arid regions or even serpentine or Mars soils with high levels of magnesium sulfate. Also, new genes associated with natural variation in A. thaliana can be used for future bioengineering.