Arabidopsis thaliana Columbia (Col-0) ecotype was used in this study. The T-DNA insertion mutant atmgt6 (SALK_203866) was obtained from the Arabidopsis Biological Resource Center. Homozygous individuals of atmgt6 were screened by PCR using primers listed in Supplementary Table 1. For on-plate growth assays, seeds were sterilized with 75% ethanol for 2 min, washed three times, and sown on half-strength Murashige and Skoog (MS) medium containing 0.75 mM Mg²⁺, 1.5 mM Ca²⁺, 1% sucrose (Sigma) and solidified with 0.8% phytoblend (Caisson Labs). The plates were kept at 4°C for 2 days and then were placed vertically in growth chamber under 90 μmol⋅m^-2⋅s^-1 light intensity with a 16 h light/8 h dark photoperiod. Three-day-old seedlings were transferred onto media containing various ions as indicated in the figure legends. For hydroponic cultures, 7-day-old seedlings germinated in half-strength MS (1/2 MS) were transferred to one-sixth-strength (1/6 MS) hydroponic medium containing 0.25 mM Mg²⁺ and 0.5 mM Ca²⁺ without sucrose for another 7 days. Plants were then transferred to hydroponic 1/6 MS media containing various contents of Mg²⁺. Plant materials were harvested for further analyses 2 days after treatment.

The S. typhimurium mutant MM281, which lacks the Mg²⁺ transporter-CorA, MgtA, and MgtB, is used as a system for functional complementation analysis of candidate Mg²⁺-transporter genes. MM281 competent cells were transformed with empty pTrc99A vector, AtMGT10-pTrc99A or OsHKT2;4-pTrc99A plasmid by electroporation. Cells were plated onto LB medium containing 10 mM Mg²⁺ and indicated antibiotics (34 μg⋅mL^-1 chloramphenicol and 100 μg⋅mL^-1 ampicillin), and incubated at 37°C overnight. The transformants were confirmed by PCR amplification and individual positive ones were grown in liquid LB medium containing 10 mM Mg²⁺ and antibiotics as indicated above. Fifty micrometer IPTG was applied for the induction of protein expression. The liquid cultures were adjusted to OD₆₀₀ = 1.0, diluted in a 10-fold series, and spotted 3 μL onto N-minimal medium supplemented with different concentrations of MgSO₄ and the antibiotics. Growth of different strains was pictured after incubation at 37°C for 2 days. The growth rate of the three strains in liquid medium was also monitored as previously described (Mao et al., 2014). After growing in liquid LB medium to OD₆₀₀ of 0.6–0.8, cells were harvested by centrifugation at 5000 × g for 10 min, washed twice with distilled water to remove excess Mg²⁺, and resuspended in distilled water. N-minimal medium was prepared with various concentrations of MgSO₄ (0.1, 0.5, 1, and 10 mM). Cells were then adjusted to a final OD₆₀₀ of 0.001–0.002. The growth of the cultures was monitored and was plotted as a function of growth time.

For the constructs used in functional complementation assay in MM281 strain, the OsHKT2;4 cDNA fragment was amplified using the primers OsHKT2;4-FC and OsHKT2;4-RC and ligated to the pTrc99A vector. For overexpressing OsHKT2;4 in wild type and the atmgt6 mutant, the genomic fragment (containing a 1.92 kb promoter region upstream of the ATG starting codon and 1731 bp coding region of OsHKT2;4) was amplified using primer pair OsHKT2;4-OE-F and OsHKT2;4-OE-R and cloned into pCAMBIA1300 vector. This construct was introduced into Agrobacterium tumefaciens strain GV3101 by electroporation and was selected on 1/2 MS medium containing kanamycin. The selected positive transformant was used to transform developing floral tissues of 4-week-old atmgt6 plants using the flora dip method (Clough and Bent, 1998). For expression in X. laevis oocytes, OsHKT2;4 cDNA was cloned into the pGEMHE vector downstream from the T7 promoter using primers OsHKT2;4-FP and OsHKT2;4-RP. All primer pairs were listed in Supplementary Table 1.

Total RNA was extracted from rosette leaves using the TRizol Reagent (Invitrogen), and the first-strand cDNA was synthesized by M-MLV Reverse Transcriptase (Promega) following the manufacturer’s instructions. The semi-quantitative RT-PCR analysis of gene expression using cDNA of Col-0, atmgt6, OE29, and OE24 followed by a 26 cycles of PCR amplification. AtActin2 (AT3G18780) was used as the internal reference. Primers used are listed in the Supplementary Table 1.

cRNA was synthesized from 1 μg linearized DNA template using a mMessage mMachine in vitro transcription kit (Ambion) according to the manufacturer’s recommendations and stored at -80°C. Stage V to VI X. laevis oocytes were harvested, defolliculated, and cultured in ND96 solution containing 96 mM NaCl, 2 mM KCl, 1.8 mM CaCl₂, 1 mM MgCl₂, 25 μg⋅mL^-1 gentamicin, pH 7.4 adjusted with 5 mM HEPES/NaOH. Approximately 50 ng of cRNA, in a total volume of 23 nL, was injected into each X. laevis oocyte. Oocytes of 2 days after injection were used for two-electrode voltage-clamp analysis. The perfusion solution was used as described previously (Lan et al., 2010) with some modifications. The perfusion solution contained (in mM) 1 K-gluconate, 1 Na-gluconate, 185 mannitol, and 10 Mes-Tris (pH 7.4). The recording pipette contained 3 M KCl. The currents were recorded by hyperpolarized pulses of a 0.2 s prepulse at -40 mV followed by voltage steps of 60 to -150 mV (in 15 mV decrements, 1.8 s duration) followed by a 1.5 s deactivation at 0 mV. The current-voltage (I-V) curves plot current values at the end of each voltage-clamp episode (t = 2 s, n = 6 for each group).

Two-week-old hydroponically grown plants were exposed to solution containing different concentrations of Mg²⁺. After 2-day exposure, both the roots and shoots were harvested and sampled for analysis. The dry weight (DW) of the samples was measured after drying for 48 h at 60°C. Subsequently, the samples were digested in 0.5 ml of 70% HNO₃ at 100°C for 30 min on a digester (DigiBlock ED16, LabTech). Ion concentration was measured by Inductively Coupled Plasma-Mass Spectrometry (ICP-MS) (PerkinElmer NexION 300).

To determine whether OsHKT2;4 functions in Mg²⁺ transport, a cDNA fragment containing the complete open reading frame with 1530 bases encoded a protein of 509 residues was cloned and expressed in S. typhimurium mutant strain MM281. MM281 is incapable of loading Mg²⁺ into cellular compartment, as it lacks three functional Mg²⁺ transporters CorA, MgtA, and MgtB, and its growth is retarded or arrested when the culture medium contains less than 10 mM Mg²⁺ (Townsend et al., 1995; Li et al., 2001). Therefore, complementation of this strain has proved useful in identifying and developing information about potential Mg²⁺ transporters, including AtMGTs (Li et al., 2001, 2008; Hicks et al., 2003; Mao et al., 2008; Chen et al., 2009; Mao et al., 2014).

We used AtMGT10 (AT5G22830), also named as AtMRS2-11, as the positive control in the complementation assay of MM281 due to its high affinity in Mg²⁺ transport in MM281 system (Li et al., 2001). As shown in Figure 1A, the MM281 mutant strains exogenously expressing empty vector pTrc99A, AtMGT10-pTrc99A, or OsHKT2;4-pTrc99A grew normally in the medium with 10 mM Mg²⁺. The strains expressing OsHKT2;4-pTrc99A exhibited faster growth than those expressing the empty pTrc99A vector in the media containing low concentrations of Mg²⁺ (1 and 2 mM), and still grew but to a less extent in the media containing 500 and 100 μM Mg²⁺, while the control did not grow at these conditions (Figure 1A), suggesting that OsHKT2;4 renders the mutant strains more tolerant to Mg²⁺ deficiency by enhancing the Mg²⁺ transport activity. However, OsHKT2;4 was less effective to restore the growth of mutant strains compared with AtMGT10 in the media containing insufficient Mg²⁺. For example, AtMGT10 rescued MM281 growth in medium containing 10 μM Mg²⁺ (Figure 1A), as shown before (Li et al., 2001), while OsHKT2;4 did not (Figure 1A). These results indicated that although OsHKT2;4 had the Mg²⁺ transport activity similar to AtMGT10, it might have the kinetic property with lower affinity to Mg²⁺ in heterologous MM281 system.

To further verify the complementation of OsHKT2;4 for the growth of MM281 tested in the agar plates (Figure 1A), the bacteria were cultured in the liquid media containing 0.1, 0.5, 1, or 10 mM Mg²⁺, and their growth curves were established within 24 h after cultured. As shown in Figure 1B, the strain expressing AtMGT10 grew the most rapidly at these Mg²⁺ concentrations, supporting that AtMGT10 is a high-affinity Mg²⁺ transporter. In addition, MM281 expressing OsHKT2;4 displayed faster growth than those with empty vector pTrc99A under the conditions in which the Mg²⁺ concentration was 1 or 10 mM. By contrast, in the presence of 0.1 or 0.5 mM Mg²⁺, the strains expressing OsHKT2;4 displayed the similar rate of growth to the strains expressing empty vector pTrc99A. These results were consistent with the ones observed on the agar plates, and demonstrated that OsHKT2;4 might mediate low-affinity Mg²⁺ uptake in vivo.

To further assess the transporting properties of OsHKT2;4 under different Mg²⁺ concentrations, two-electrode voltage-clamp experiment using X. laevis oocytes was performed. OsHKT2;4-dependent currents were recorded from the oocytes injected with OsHKT2;4 cRNA or the oocytes injected with water perfused with different Mg²⁺ concentrations. The oocytes injected with water produced small endogenous currents in perfusion medium with 6 mM Mg²⁺ (Figures 2A-a). In contrast, OsHKT2;4-expressing oocytes generated the larger currents in the solutions containing 1.2, 6, and 20 mM Mg²⁺ (Figures 2A-b–d). The current-voltage relationship displayed the currents from OsHKT2;4-expressing oocytes perfused with 6 or 20 mM Mg²⁺ were significantly larger than those from OsHKT2;4-expressing oocytes perfused with 0.3 or 1.2 mM Mg²⁺. It was noteworthy that the currents from OsHKT2;4-expressing oocytes perfused with 0.3 mM Mg²⁺ were similar to those with 1.2 mM Mg²⁺, implying they might not be Mg²⁺ sensitive under low-Mg²⁺ conditions (Figure 2B).

To test this possibility, we compared the amplitude and reversal potential of the currents generated from the oocytes perfused with 0, 0.3, 1.2, 6, or 20 mM Mg²⁺. The currents generated from the oocytes expressing OsHKT2;4 perfused with 0.3 or 1.2 mM Mg²⁺ displayed the similar levels of the amplitudes and reversal potentials, even were similar to those without Mg²⁺ (Figures 2C,D). These Mg²⁺ insensitive currents were larger than the currents from the oocytes injected with water (Figure 2C), and thus they may result from other ions, such as Na⁺ or K⁺ currents generated by OsHKT2;4, as suggested by the previous studies (Lan et al., 2010; Horie et al., 2011). By contrast, in the presence of 6 or 20 mM Mg²⁺, the oocytes expressing OsHKT2;4 produced the currents with larger amplitudes and less negative reversal potentials compared with the others (Figures 2C,D). Thus, these results further supported the hypothesis that OsHKT2;4 exhibits permeability for Mg²⁺ only under the conditions containing high-Mg²⁺ concentrations.

We have shown previously that AtMGT6, a Mg²⁺ deficiency-induced Mg²⁺ transporter, mediates directly Mg²⁺ uptake in roots and is required for plant adaptation to low-Mg²⁺ environment (Mao et al., 2014). An independent study reported ethyl methanesulfonate (EMS)-mutagenized AtMGT6, or named as AtMRS2-4, caused plant growth defects under both low and high-Mg²⁺ conditions (Oda et al., 2016). Considering the critical role of AtMGT6 in Mg²⁺ acquisition, we suggested the activity of Mg²⁺ transport conducted by OsHKT2;4 might be covered by this transporter, and thus generated transgenic OsHKT2;4 Arabidopsis lines with the disruption of AtMGT6 to examine the potential relevance of OsHKT2;4 to Mg²⁺ responses in planta. We used an Arabidopsis T-DNA insertion line (SALK_203866), in which T-DNA was inserted into the third exon of AtMGT6 gene (Figure 3A). Transcript of AtMGT6 in the line SALK_203866 was not detected by RT-PCR (Figure 3C), indicating that the T-DNA insertion line was a knockout allele, and referred to as atmgt6 line hereafter. We then expressed the coding region of OsHKT2;4 into atmgt6 line driven by its native promoter (Figure 3B). The transformants were screened by hygromycin, and were further analyzed for the expression levels of OsHKT2;4 by RT-PCR. We selected two of them as the representative transgenic lines due to their relatively high OsHKT2;4 expression levels, and referred to as OE29 and OE34, respectively, to perform subsequent experiments (Figure 3C).

We examined the growth of the Col-0, atmgt6, and transgenic lines OE29 and OE34 in Mg²⁺-depleted medium supplemented with various contents of Mg²⁺ as indicated in Figure 3D. The mutant atmgt6 exhibited growth defects in the medium containing 0, 0.01, 0.1, or 0.25 mM Mg²⁺, and had lower fresh weight and shorter roots than those of the Col-0 plants, while the growth retardation could be rescued in the Mg²⁺-sufficient medium (2 mM Mg²⁺) (Figure 3D), consistent with the idea that AtMGT6 confers low-Mg²⁺ tolerance for Arabidopsis (Mao et al., 2014; Oda et al., 2016). However, OsHKT2;4 overexpression could not rescue the growth deficiency of atmgt6 in low-Mg²⁺ conditions as expected, as transgenic lines OE29 and OE34 displayed the similar growth phenotype to atmgt6 with no significant differences on fresh weight and root length under these tested conditions (Figures 3D–F). These results suggested that OsHKT2;4 might not function in low-Mg²⁺ conditions in planta.

As shown in the experiments in vitro, OsHKT2;4 exhibiting Mg²⁺ transport activity in both heterologous MM281 system and X. laevis oocytes happened only at high external Mg²⁺ concentrations (Figures 1, 2). Therefore, we presumed that OsHKT2;4 might mediate Mg²⁺ transport when plants were cultivated under the Mg²⁺ abundant conditions, though it was unable to function in low-Mg²⁺ conditions in planta (Figures 3D–F). To conduct assessment of the sensitivity to high-Mg²⁺ condition of OsHKT2;4, we used 1/6 MS medium supplemented with several concentrations of Mg²⁺ (2, 4, 6, 8, and 10 mM) for growth assays. After growing on 1/6 MS for 2 weeks, both the root length and fresh weight of transgenic lines OE29 and OE34 were comparable to those of atmgt6 under the normal condition. However, addition of 2 mM Mg²⁺ resulted in growth arrest of OE29 and OE34 compared with atmgt6. Further increases of extra Mg²⁺ (up to 10 mM) demonstrated a consistent dosage-dependent inhibitory manner (Figure 4A). Quantitative analysis of root length (Figure 4B) and fresh weight (Figure 4C) indicated that, compared with Col-0 and atmgt6, the aerial parts of OE29 and OE34 exhibited a more severe growth retardation in Mg²⁺-abundant conditions, while their root length were not altered. Taken together, these results demonstrated that OsHKT2;4 results in Mg toxicity on aerial tissues in high-Mg²⁺ conditions in Arabidopsis, supporting the idea of low-affinity Mg²⁺ uptake of OsHKT2;4.

To examine whether the high-Mg²⁺ toxic phenotype is a consequence of the ectopic expression of OsHKT2;4 in OE lines, we conducted an RT-PCR analysis to verify the expression of OsHKT2;4 under a high-Mg²⁺ condition supplemented with 6 mM Mg²⁺. As demonstrated in Supplementary Figure 5, OsHKT2;4 was mainly expressed in shoot tissues under normal growth conditions, consistent with the previous report in rice (Lan et al., 2010). Moreover, the expression of OsHKT2;4 in the shoots was not significantly induced by 6 mM Mg²⁺, and even decreased after 24 h′ treatment with 6 mM Mg²⁺. However, expression of OsHKT2;4 in the roots was dramatically induced after 4 h′ treatment with 6 mM Mg²⁺, and became even stronger after 24 h of this treatment (Supplementary Figure 5). The oppose effect on OsHKT2;4 expression roots and shoots upon high Mg²⁺ suggested a disturbance on Mg²⁺ balance between roots and shoots under the high-Mg²⁺ conditions.

To probe the reason responsible for the increased sensitivity to high external Mg²⁺ in the OE plants, Mg²⁺ concentration of the atmgt6 and OsHKT2;4 overexpression lines was determined using ICP-MS. Plants were grown hydroponically for 2 weeks and then transferred to a fresh hydroponic medium containing 0, 0.25 (referred to as “Control” in Figure 5), and 6 mM Mg²⁺ for another 2 days before the roots and shoots were harvested, respectively, for analysis. As shown in Figure 5A, Mg²⁺ content was consistently higher in the OE lines than atmgt6 in shoot tissues when plants grown in all Mg²⁺ regimes tested (0, 0.25, and 6 mM). In analysis of Mg²⁺ content of root tissues among different plants, although they exhibited similar and incremental Mg²⁺ content in the Mg²⁺-deficient and normal medium, the OE lines contained ∼30% less Mg²⁺ compared with the atmgt6 plants in 6 mM Mg²⁺ condition (Figure 5B). These results indicated an altered Mg²⁺ distribution ratio in shoot and root. We thus analyzed the Mg²⁺ partitioning between shoots and roots in atmgt6 and OE lines, and noticed that as the Mg²⁺ level elevated, more Mg²⁺ sequestered in root tissues than in low-Mg²⁺ condition. In medium containing 6 mM Mg²⁺, the shoots of atmgt6 accumulated ∼37%, while the OE lines accumulated over ∼50% of the total Mg²⁺ enclosed in plants (Figure 5C), further confirming the critical role of OsHKT2;4 in the Mg²⁺ allocation between shoots and roots.

Due to the similar physical properties, Ca²⁺ and Mg²⁺ compete for the same sites of substrates (Yermiyahu et al., 1994), and the balance of Ca²⁺ and Mg²⁺ is an important factor for plant growth. Previously, evidence was presented that OsHKT2;4 acts as a channel for the transport of both Ca²⁺ and Mg²⁺ (Lan et al., 2010; Horie et al., 2011). To examine whether Ca²⁺ affects the high Mg²⁺-sensitive phenotype in OsHKT2;4 overexpressing lines, we assessed the growth of Col-0, atmgt6, atmgt6 overexpressing OsHKT2;4 lines (OE29 and OE34) on Mg²⁺-abundant 1/6 MS medium supplemented with different concentrations of excess Ca²⁺.

In normal 1/6 MS medium, OE29 and OE34 exhibited the similar growth as atmgt6. As the concentration of external Mg²⁺ increased, OE29 and OE34 started to show a more severe growth arrest than atmgt6 (Figure 4A). However, increasing Ca²⁺ improved the plant growth (Figure 6A) of all genotypes, and the improvement was more obvious in the OE29 and OE34 lines (Supplementary Figure 2). For example, fresh weight of OE lines was ∼50% of that in atmgt6 in 1/6 MS medium with extra 8 mM Mg²⁺ (“basal medium” in Figure 6), however, when 1 mM Ca²⁺ was added to this basal medium, the fresh weight of OE lines was restored and reached to ∼80% to that in atmgt6. Moreover, when 3 mM Ca²⁺ were added to the basal medium, both root length and fresh weight of OE29 and OE34 were recovered to almost an identical level with that of atmgt6 (Figures 6B,C and Supplementary Figure 2), supporting the notion of Ca²⁺-Mg²⁺ antagonism.

It has been reported that low Ca²⁺ in the medium triggered the increase of Mg²⁺ concentration, mimicking high-Mg²⁺ conditions (Rios et al., 2012). To analyze further whether Ca²⁺ deficiency is also responsible for the phenotype induced by high Mg²⁺, we thus tested the sensitivity among different plants to low-Ca²⁺ conditions. In Ca²⁺-depleted 1/6 MS medium (“basal medium” in Figure 7), similar phenotype was observed in atmgt6 and OE lines, suggesting that OsHKT2;4 is not responsible for Ca²⁺ deficiency. On the contrary, as the increasing content of extra Mg²⁺ (2, 4, 6, and 8 mM) was added to the basal medium, differences of fresh weight between atmgt6 and OE lines started to occur (Figure 7 and Supplementary Figure 3). These results demonstrated that high Mg²⁺, rather than Ca²⁺ deficiency, is the primary factor that caused growth defects in the OE lines than atmgt6.