Constitutive Overexpression of the OsNAS Gene Family Reveals Single-Gene Strategies for Effective Iron- and Zinc-Biofortification of Rice Endosperm

Background Rice is the primary source of food for billions of people in developing countries, yet the commonly consumed polished grain contains insufficient levels of the key micronutrients iron (Fe), zinc (Zn) and Vitamin A to meet daily dietary requirements. Experts estimate that a rice-based diet should contain 14.5 µg g−1 Fe in endosperm, the main constituent of polished grain, but breeding programs have failed to achieve even half of that value. Transgenic efforts to increase the Fe concentration of rice endosperm include expression of ferritin genes, nicotianamine synthase genes (NAS) or ferritin in conjunction with NAS genes, with results ranging from two-fold increases via single-gene approaches to six-fold increases via multi-gene approaches, yet no approach has reported 14.5 µg g−1 Fe in endosperm. Methodology/Principal Findings Three populations of rice were generated to constitutively overexpress OsNAS1, OsNAS2 or OsNAS3, respectively. Nicotianamine, Fe and Zn concentrations were significantly increased in unpolished grain of all three of the overexpression populations, relative to controls, with the highest concentrations in the OsNAS2 and OsNAS3 overexpression populations. Selected lines from each population had at least 10 µg g−1 Fe in polished grain and two OsNAS2 overexpression lines had 14 and 19 µg g−1 Fe in polished grain, representing up to four-fold increases in Fe concentration. Two-fold increases of Zn concentration were also observed in the OsNAS2 population. Synchrotron X-ray fluorescence spectroscopy demonstrated that OsNAS2 overexpression leads to significant enrichment of Fe and Zn in phosphorus-free regions of rice endosperm. Conclusions The OsNAS genes, particularly OsNAS2, show enormous potential for Fe and Zn biofortification of rice endosperm. The results demonstrate that rice cultivars overexpressing single rice OsNAS genes could provide a sustainable and genetically simple solution to Fe and Zn deficiency disorders affecting billions of people throughout the world.


Introduction
Rice is the primary source of food for roughly half of the world's population yet the polished grain, also known as white rice, contains nutritionally insufficient concentrations of iron (Fe), zinc (Zn) and pro-vitamin A to meet daily requirements in diets based on this staple [1,2]. Other widely consumed cereals, such as wheat and maize, are also poor sources of several key micronutrients. As a result, micronutrient deficiencies afflict billions of people throughout that world and are particularly prevalent in developing countries where cereals are widely consumed. Fe deficiency affects more than two billion people worldwide, with symptoms ranging from poor mental development and depressed immune function to anaemia, and is the most widespread nutritional deficiency in the world [3]. The development of new cereal varieties containing increased concentrations of Fe and other essential micronutrients, an approach known as biofortification, offers an inexpensive and sustainable solution to the chronic micronutrient malnutrition problems that currently plague people in developing countries.
Rice has the lowest Fe concentration of the cultivated cereal crops and a striking lack of genetic variation for this trait has hindered conventional breeding efforts from increasing its Fe concentration beyond 6 mg g 21 in polished grain [4,5]. To produce polished grain with 14.5 mg g 21 Fe, the target concentration that nutritionists have recommended to meet Fe requirements in a rice-based diet, novel sources of genetic diversity for grain Fe concentration are required [6,7]. Numerous biotechnological strategies have been employed to produce rice with increased concentrations of Fe in endosperm tissues; the principle constituent of polished grain. Grain-specific overexpression of genes encoding ferritin, a Fe storage protein found in plants, animals and bacteria, has been utilized to increase the ''sink'' for Fe in endosperm [8]. While this approach has resulted in a two-fold increase in endosperm Fe concentration, the relatively modest increases in Fe concentration do not match the 13-fold increase in endosperm ferritin protein levels that often occur via this strategy [9], suggesting that transport of Fe to the endosperm sink is also limiting.
Nicotianamine (NA) is a chelator of transition metals that plays important roles in long-and short-distance transport of metal cations, including Fe 2+ and Fe 3+ , in higher plants [10,11]. NA is biosynthesized by trimerization of S-adenosylmethionine, a reaction catalyzed by the NA synthase (NAS) enzymes. Genes encoding NAS are known to be differentially regulated by iron status in a variety of plant species including maize, Arabidopsis, barley and rice [12][13][14][15], and show strong induction by Fe deficiency. Two of the three rice NAS genes (OsNAS1 and OsNAS2), for instance, show stelar-specific root expression under Fe sufficiency that becomes more ubiquitous throughout root and shoot tissues under Fe deficiency, while a third NAS gene (OsNAS3) changes from a primarily shoot-specific expression pattern under Fe sufficiency to more root-specific expression under deficiency [13]. Aside from its role in metal transport in plants, NA is an antihypertensive substance in humans and rice lines with enhanced NA concentration have been developed as potential candidates for the functional food industry [16].
Transgenic approaches to increase NA concentration have often focused on overexpression of exogenous NAS genes in plants. Constitutive overexpression of a barley NAS gene, HvNAS1, in Arabidopsis and tobacco led to a several-fold increase in seed Fe, Zn and Cu concentration of both species [17]. Similar overexpression of HvNAS1 in rice led to greatly enhanced NA concentration (15-fold increase over wild type) and 2.3-and 1.5fold increases in Fe and Zn concentrations of polished grain, respectively [18]. Constitutive expression of an Arabidopsis NAS gene, AtNAS1, in conjunction with endosperm-specific expression of genes encoding ferritin and phytase, led to a 6.3-fold increase in Fe concentration of rice endosperm [19].
Relatively few studies have been done to overexpress the endogenous rice NAS genes (OsNAS) in rice. Endosperm-specific overexpression of OsNAS1 resulted in polished grain with significantly increased concentrations of NA and Zn and, although Fe concentration was not increased by this strategy, the bioavailability of Fe was double that of controls as measured by ferritin synthesis in Caco-2 cells [20]. Recently, activation tagged lines of rice with increased expression of OsNAS2 and OsNAS3 were identified and characterized [21,22]. An OsNAS2 activation tagged line had 20fold more NA and 2.7-fold more Zn in polished grain, while two OsNAS3 activation tagged lines had up to nine-fold more NA, 2.6fold more Fe and 2.2-fold more Zn in polished grain. Most significantly, polished grain from an OsNAS3 activation tagged line reversed signs of Fe-deficiency when fed to anemic mice [21].
The overall aim of this study was to constitutively overexpress all three members of the OsNAS gene family, individually, to assess their utility for Fe biofortification of polished rice grain via a single-transgene approach. Characterization of more than 90 independent transgenic lines overexpressing these genes revealed that all three OsNAS genes increase not only Fe, but also Zn concentrations in unpolished and polished grain when expressed constitutively and those increases are positively correlated with NA concentration. One member of the OsNAS gene family, OsNAS2, was particularly effective at increasing Fe and Zn concentrations in rice endosperm and this increase was mapped in unprecedented detail using synchrotron X-ray fluorescence spectroscopy (m-XRF).

Results
Construction of three rice populations overexpressing OsNAS1, OsNAS2 and OsNAS3 The 0.7 kb dual CaMV 35S promoter contained in the pMDC vector system [23] was used to drive constitutive expression of the OsNAS1 (LOC_Os03g19427), OsNAS2 (LOC_Os03g19420) and OsNAS3 (LOC_Os07g48980) coding sequences in rice. The T-DNA region of the binary vectors used for transformations also contained the selectable marker gene neomycin phosphotransferase II that detoxifies aminoglycoside antibiotics such as geneticin (G418) and kanamycin ( Figure 1). Embryogenic callus of japonica rice cultivar Nipponbare was used for Agrobacterium-mediated transformation of the binary vectors containing the three different OsNAS coding sequences. The production of 30 independent transgenic lines carrying the OsNAS1 overexpression vector, designated the OE-OsNAS1 population, 39 independent transgenic lines carrying the OsNAS2 overexpression vector, designated the OE-OsNAS2 population, and 24 independent transgenic lines carrying the OsNAS3 overexpression vector, designated the OE-OsNAS3 population, was confirmed by resistance to geneticin in T 0 and T 1 plants, PCR and Southern blot analysis (data not shown).

Constitutive overexpression of the OsNAS genes leads to increased Fe and Zn concentrations in unpolished and polished grain
Because nicotianamine is known to chelate and mobilize a variety of metal cations including Fe, Zn, Mn, Cu and Ni in plants [24,11] we employed inductively coupled plasma optical emission spectrometry (ICP-OES) to characterize the elemental composition of unpolished T 1 grain harvested from a single season of growth of the three transgenic rice populations in the glasshouse. Two metals in particular, Fe and Zn, were several-fold higher in unpolished grain of the OE-OsNAS population, relative to WT grain ( Figure 2), while Mn, Cu and Ni did not show significant differences from WT grain (Table S1).
Unpolished grain Fe concentrations ranged from 25 to 56 mg g 21 dry weight (DW) in the OE-OsNAS1 population, 19 to 81 mg g 21 DW in the OE-OsNAS2 population and 21 to 63 mg g 21 DW in the OE-OsNAS3 population, representing up to 2.4-fold, 3.5-fold and 2.7-fold increases, respectively, over wild type (WT) in each population ( Figure 2A). Unpolished grain Zn concentrations ranged from 40 to 59 mg g 21 dry weight (DW) in the OE-OsNAS1 population, 30 to 95 mg g 21 DW in the OE-OsNAS2 population and 30 to 79 mg g 21 DW in the OE-OsNAS3 population, representing up to 1. -fold, 2.5-fold and 2.1-fold increases, respectively, over wild type (WT) in each population ( Figure 2B). Furthermore, Fe and Zn concentrations of unpolished grain were highly correlated in transgenic lines (r = 0.83, 0.94 and 0.97 for the OE-OsNAS1, OE-OsNAS2 and OE-OsNAS3 populations, respectively) which accounts for the similar Fe and Zn profiles in Figure 2A and 2B (genotype order is the same in both panels).
Polished grain was produced from selected transgenic lines using a modified non-contaminating Kett Mill. One transgenic line overexpressing OsNAS1 (OE-OsNAS1S), two independent transgenic lines overexpressing OsNAS2 (OE-OsNAS2B and OE-OsNAS2J) and one transgenic line overexpressing OsNAS3 (OE-OsNAS3B) were selected for this analysis based on large numbers of available grain to mill and Fe concentrations that fell within the upper 20% of each overexpression population. WT had a Fe concentration of 4.5 mg g 21 DW in the polished grain, representing approximately 20% of Fe concentration in the unpolished grain ( Table 1) Constitutive overexpression of the OsNAS genes leads to increased NA concentrations in unpolished grain that are positively correlated with Fe and Zn concentration Liquid chromatography-mass spectrometry (LC-MS) was employed to determine if OsNAS overexpression leads to significantly increased NA concentration of the grain. Single-insert transgenic lines were selected for nicotianamine quantification experiments to ensure that null segregant lines (lines that have the lost the overexpression vector due to meiotic segregation) were produced as additional controls to WT. Three sibling T 1 lines, comprising two transgenic lines and one null segregant line, were derived from a single T 0 parental line within each of the three OE-OsNAS populations and grown to maturity in a growth room to yield T 2 grain. The T 0 parents of the OE-OsNAS1, OE-OsNAS2 and OE-OsNAS3 T 1 siblings had unpolished grain Fe concentrations of 56, 64 and 51 mg g 21 DW, respectively, and unpolished grain Zn concentrations of 59, 80 and 65 mg g 21 DW, respectively. In addition to nicotianamine quantification by LC-MS, the Fe and Zn concentrations of T 2 grain were determined by ICP-OES.
The unpolished grain NA concentration was 18 mg g 21 DW for WT and did not differ significantly from unpolished grain NA concentrations of null segregant (NS) lines ( Figure 3A). By contrast, unpolished grain NA concentrations ranged from 96 to 115 mg g 21 DW in the OE-OsNAS1 sibling lines, 152 to 168 mg g 21 DW in the OE-OsNAS2 sibling lines and 174 to 210 mg g 21 DW in the OE-OsNAS3 sibling lines, representing up to 6.4-fold, 9.3fold and 11.7-fold increases, respectively, over wild type (WT) concentrations of NA. Figure 3B demonstrates the statistically significant, positive correlation that was found between unpolished grain NA concentration and Fe and Zn concentration for the ten genotypes utilized in this experiment. While the OE-OsNAS3 siblings produced T 2 grain with the same Fe and Zn concentrations as the T 0 parent, the Fe concentration of T 2 grain from the OE-OsNAS1 siblings and Fe and Zn concentrations of T 2 grain from the OE-OsNAS2 siblings were lower than that of the T 0 parents (approximately 10-15 mg g 21 DW lower). These results indicate that grain Fe and Zn concentrations of certain transgenic events are more consistent than others across varying environments (in this case glasshouse vs. growth room) and that all events should be evaluated over several sexual cycles and under differing conditions including the field. m-XRF elemental maps reveal significant increases in Fe and Zn accumulation in specific tissues of OE-OsNAS2 grain Synchrotron X-ray fluorescence spectroscopy (m-XRF) was used to generate elemental distribution maps of several NA-related cations (Fe, Zn, Mn and Cu) in two longitudinal sections each of WT and OE-OsNAS2A grain (four sets of elemental distribution maps total). Elemental distribution in the two longitudinal sections of each grain type was very similar and therefore only one set of images for each grain type is presented in Figure 4.
The maps of Fe distribution in WT and OE-OsNAS2A grain (Figures 4B and C, respectively) show a striking lack of detectable Fe signal in large portions of the endosperm. In WT grain, the highest signal occurs in scutellum and outer regions of the embryo while a very low Fe signal is detected in the outermost layers of the endosperm and the single-layered aleurone that surrounds the endosperm. No signal is detected from the inner layers of the endosperm. The OE-OsNAS2A grain, by contrast, has a high Fe signal in the outer endosperm and aleurone layers in addition to high aleurone and embryo signals. The inner layers of endosperm, as with WT grain, have no signal.
The maps of Zn distribution in WT and OE-OsNAS2A grain ( Figures 4D and E, respectively) depict a radically different distribution of this element compared to Fe. In WT grain, the highest signal is observed inside the embryo (likely corresponding to the plumule) while the scutellum and outer embryo has intermediate signals. A low signal is detected in a thick band comprising many outer layers of endosperm and the single-layered aleurone that surrounds the endosperm. Unlike Fe, the Zn signal extends (albeit at very low levels) throughout the endosperm of WT grain. In OE-OsNAS2A grain a similar pattern of Zn distribution is observed, however, the signal in the embryo and throughout endosperm tissues is considerably higher. In the outermost layers of the endosperm and the single-layered aleurone, the signal borders on intermediate signal intensity.
The maps of Mn and Cu in WT and OE-OsNAS2A grain (Figures 4F and G, 4H and I, respectively) show that these metals, like Fe, have no signal in much of the endosperm. The Mn signal is higher in the outermost layers of the endosperm and the singlelayered aleurone of OE-OsNAS2A compared to WT, and this may be explained by the slightly higher Mn concentration detected by ICP-OES for the OE-OsNAS2A grain relative to WT (14 vs. 11 mg g 21 DW Mn, respectively). The Cu signal is higher in the embryo of OE-OsNAS2A compared to WT, and this may be explained by the slightly higher Cu concentration detected by ICP-OES for the OE-OsNAS2A grain relative to WT (9 vs. 7 mg g 21 DW Cu, respectively).
XRF line scans reveal significantly more Fe and Zn, and larger Fe:Zn ratios, in aleurone, subaleurone and endosperm tissues of OE-OsNAS2 grain Two 135 mm line scans across the grain, away from the embryo region, are reported for WT and OE-OsNAS2A in Figure 5. These line scans were obtained by laterally averaging a box (represented by a square in Figure 4A) with a width of 23 pixels so that each count represents the average of 23 line scans of the grain. The rice grain contains a single aleurone layer that is rich in phosphorus (P, primarily in the form of phytic acid) while starchy endosperm tissues contain only trace P [25,26]. The P distribution, which had highly similar counts and profile in both WT and OE-OsNAS2A grain, was thereby used to assign regions of the line scan to aleurone, subaleurone and endosperm layers of grain. As the rectangular-shaped aleurone cells of rice endosperm are approximately 25-30 mm in length [27], we conservatively assigned the first 50 mm of the line scan to the aleurone layer (the additional 20 mm accounting for the pericarp, seed coat and nucellus that precede the aleurone layer). Consistent with the phytic acidenriched aleurone layer, the average P counts for both WT and OE-OsNAS2A grain were by far the highest in this 50 mm section (70 and 81 counts, respectively). P counts began to rapidly drop off after 50 mm and we assigned 51-90 mm to the subaleurone layer and 91-135 mm to the endosperm (assuming average cell lengths of 40 mm in these two layers). Average P counts for WT and OE-OsNAS2A grain were roughly halved in the subaleurone layer (44 and 30 counts, respectively) and nearly background levels in the endosperm layer (20 and 18 counts, respectively). The line scan of WT grain ( Figure 5A) shows that Fe and Zn counts increase rapidly through the aleurone layer and are nearly equal towards the start of the subaleurone layer (,1550 counts for both Fe and Zn at 50 mM). The Fe count begins to drop off beyond this point while Zn continues to climb for most of the subaleurone and endosperm layers before leveling off in the last 10 mm of the scan. The results demonstrate that while the aleurone layer has significant amounts of Fe and Zn, the subaleurone layer and endosperm layers, combined, have higher amounts of both of these metals. The results also show that there is more Zn relative to Fe for nearly all of the scanned region, leading to Fe:Zn signal ratios of ,1 for the aleurone, subaleurone and endosperm layers ( Table 2).
The line scan of OE-OsNAS2A grain ( Figure 5B) shows that Fe and Zn counts increase rapidly through the aleurone layer but at a much steeper slope for Fe, so that Fe counts surpass Zn counts early in the aleurone layer. The Fe count shows a first peak in the aleurone layer (5482 counts at 41 mm), similar to WT. Unlike the WT grain, however, the OE-OsNAS2A grain has two, successively higher, Fe peak regions in the subaleurone (6498 counts at 81 mm) and endosperm (6958 counts at 101 mm) layers. The Fe count begins to drop off beyond this point, most rapidly in the final 20 mm of the scan. The Fe counts in WT and OE-OsNAS2A grain are plotted on linear scale in Figure 6 to clearly visualize the differences in Fe quantity and distribution between the two genotypes.
The Fe count is approximately 4.4-fold higher in the aleurone, subaleurone and endosperm layers of OE-OsNAS2A grain relative to WT ( Table 2). The Zn count in OE-OsNAS2A grain rises progressively through the scan, but does not surpass the Fe count until the final 20 mm of the scan. The Zn count is 1.4-fold higher in the aleurone layer and two-fold higher in the subaleurone and endosperm layers of OE-OsNAS2A grain relative to WT. These fold increases for Fe and Zn count are remarkably consistent with the 4.2-fold and 2.2-fold increases of Fe and Zn concentration, respectively, that were reported for polished OE-OsNAS2B grain, relative to WT, as determined by ICP-OES (Table 1). Because Fe counts are higher than Zn counts for most of the line scan, Fe:Zn signal ratios are much larger for OE-OsNAS2A and .1 for both the aleurone and subaleurone layers ( Table 2).

Discussion
The results of this study demonstrate significantly increased Fe and Zn concentrations in rice endosperm as a result of constitutive overexpression of single OsNAS genes. Within each of the three transgenic rice populations -OE-OsNAS1, OE-OsNAS2 and OE-OsNAS3 -lines were identified with at least 2-and 1.5fold increases in Fe and Zn concentrations, respectively, of unpolished grain. Large differences, however, were observed regarding the upper limits of Fe and Zn enrichment that were found in each population (Figure 1). The OE-OsNAS1 and OE-OsNAS2 populations differed most significantly from each other, with the highest Fe-containing OE-OsNAS1 line (56 mg g 21 DW) representing only 70% of the unpolished grain Fe concentration of the highest OE-OsNAS2 line (81 mg g 21 DW). In fact, the five highest Fe-containing OE-OsNAS2 lines had more than 60 mg g 21 DW Fe in unpolished grain. The observed differences between the OE-OsNAS1 and OE-OsNAS2 populations are surprising considering that the coding sequences of OsNAS1 and OsNAS2 share 87% identity and the first 233 amino acids of the two, roughly 330 aa long enzymes, are identical. The OsNAS1 and OsNAS2 enzymes may show differential activity in the synthesis of NA, or overexpression of the OsNAS1 and OsNAS2 coding sequences may cause significant, as yet unknown, pleiotropic effects on nutrient transport processes in rice. Studies regarding both of these possibilities are now underway.
While only a subset of lines were milled to produce polished grain, results obtained with the two OE-OsNAS2 lines revealed some of the highest Fe concentrations that have been reported for rice endosperm. OE-OsNAS2B and OE-OsNAS2J had Fe concentrations of 19 and 14 mg g 21 DW, respectively, in rice endosperm (Table 1). These concentrations are 4.2-and 3-fold higher, respectively, than the Fe concentration observed for WT polished grain (4.5 mg g 21 DW) and represent the first time that rice lines have been reported with Fe concentrations at or above the 14.5 mg g 21 DW threshold recommended for a Fe-biofortified rice diet [6]. Zn concentrations of the two OE-OsNAS2 lines were also 1.5-to 2.2-fold higher than WT polished grain. The increased Fe concentrations of polished OE-OsNAS2 grain appear due to not only higher metal concentration in unpolished grain, but also reduced losses of Fe during milling of the grain. While Fe concentrations of WT, OE-OsNAS1 and OE-OsNAS3 polished grain represented 20% of unpolished grain Fe concentrations, Fe concentrations of the two OE-OsNAS2 polished grain samples represented 26-30% of unpolished grain Fe concentration (Table 1). These results suggest that Fe penetrates further into endosperm tissues of the OE-OSNAS2 grain relative to WT, however, more overexpression lines within each of the three populations require similar characterization to determine whether this trait is specific to only the OE-OsNAS2 population. The OE-OsNAS overexpression lines should also be evaluated under field conditions to determine the stability of the high-Fe trait under different environments where Fe may be more limiting. Field trials will also enable agronomic performance of the lines to be accurately assessed. Counts are plotted on logarithmic scale in the y-axis to account for the low P counts (blue) relative to Fe (red) and Zn (green). As the rice grain contains a single aleurone layer that is rich in phosphorus (P, primarily in the form of phytic acid) while endosperm contains little P, 1-50 mm was assigned to the aleurone layer, 51-90 mm to the subaleurone layer and 91-135 mm to the endosperm. doi:10.1371/journal.pone.0024476.g005 The positive correlations between Fe and Zn concentrations in both unpolished and polished grain of the three OE-OsNAS populations provided strong evidence that a common mechanismmost likely NA -was responsible for transporting these micronutrient metals into the grain. The LC-MS experiments with segregating T 1 lines confirmed this hypothesis and showed that NA concentrations in unpolished grain were 6.4-to 11.7-fold higher in the OsNAS overexpression progeny relative to WT and null segregant lines. The NA concentration that we calculated for WT unpolished grain using the LC-MS method -18 mg g 21 DW -is very close to published values of 21.2 mg g 21 DW for unpolished rice grain [28] and gave us confidence that our analytical technique was accurate and representative of actual NA concentrations. The fact that null segregant lines did not differ significantly from WT with respect to unpolished grain NA concentration, while all of the transgenic progeny had significantly higher concentrations, provided conclusive evidence that the OE-OsNAS constructs were responsible for the increases in NA concentration ( Figure 3A). Additionally, these experiments demonstrated that the OE-OsNAS constructs of singleinsert lines were transmitted to progeny lines in typical Mendelian fashion and stably expressed in progeny. The statistically significant, positive correlations between NA concentration and Fe and Zn concentrations (r = 0.9769 and 0.9288, respectively) demonstrated that increased NA concentrations of unpolished grain were not only higher in the OsNAS transgenic progeny, but also indicative of Fe and Zn concentration ( Figure 3B). NA concentration can therefore be considered a major regulator of Fe and Zn concentrations in rice grain, and NA may very well be a limiting factor in the accumulation of Fe and Zn in WT rice grain.
To further explore the distributions of Fe, Zn and other metal cations (that are known to chelate with NA) in rice grain, we employed synchrotron X-ray fluorescence spectroscopy (m-XRF) to generate detailed elemental maps of WT and OE-OsNAS2 longitudinal grain sections. A single-insert transgenic line from the OE-OsNAS2 population, OE-OsNAS2A, with high levels of Fe and Zn (64 and 80 mg g 21 DW, respectively) was selected as a comparison to WT grain. The elemental maps of Fe and Zn in WT grain ( Figure 4B and D) are in agreement with what we know about the distribution of these two metals in cereal grain -namely that Zn has a higher concentration and more broad distribution profile in the grain compared to Fe and, away from the embryo region, is not limited to outer layers of endosperm and the aleurone. The abundance of Zn in the central portion of the embryo, most likely in the plumule, has been observed in similar m-XRF studies of barley grain [29]. The complete absence of Fe signal from much of the endosperm, as opposed to Zn which extends (faintly) throughout the endosperm, demonstrates why polishing of rice grain causes much greater losses of Fe compared to Zn and is in line with previous findings [30]. When maps of WT grain were compared to those of OE-OsNAS2A grain, one of the most striking differences was in Fe distribution ( Figure 4B and C). Whereas WT had very low signals of Fe in the outer endosperm and aleurone layer of the grain, OE-OsNAS2A grain had intermediate to high signals in the same position. Another major difference between WT and OE-OsNAS2A grain concerned the overall higher signals for Zn throughout the transgenic grain. Although the Zn distribution pattern did not appear altered in OE-OsNAS2A grain, the signal intensity was considerably higher. The small differences observed in Mn and Cu intensity between the two grain types were likely due to small (2-3 mg g 21 DW) increases of those elements in the embryo and/or aleurone layer of OE-OsNAS2A grain.
Line scans allowed us to focus in on the outer region of the grain, away from the embryo region (the area highlighted, as an example, in green in Figure 4A), where the large differences for Fe signal intensity had been observed between WT and OE-OsNAS2A grain. Of crucial importance to this experiment was the ability to detect phosphorus (P) as a direct indicator of the aleurone layer. The aleurone cells of cereals accumulate high levels of phosphorus-containing phytic acid (PA), which normally occurs as a mixed salt of potassium (K), magnesium (Mg), calcium (Ca), Fe and Zn in the cells [31]. The primary function of PA is to provide storage of phosphorus and minerals for germinating seeds. The consequences of PA binding to minerals and micronutrients such as Fe and Zn, however, are undesirable from a nutritional point of view. PA is a strong inhibitor of mineral and micronutrient absorption and is reported to inhibit Fe, Zn, Ca and manganese (Mn) absorption in humans [32]. It is thought that mineral binding to PA forms an insoluble complex that precipitates, thereby rendering the mineral unavailable to human intestinal absorption. By detecting P in our line scans, we were able to not only distinguish aleurone cells from starchy endosperm (which has only trace levels of P), but also accurately determine if Fe and Zn were localized to a region where they were likely to be bound by PA (and therefore unlikely to be bioavailable).
The line scans of WT grain showed the highest counts of Fe towards the aleurone/subaleurone junction, after which Fe counts began to slowly decline through the subaleurone and endosperm layers ( Figure 5A). Zn counts, on the other hand, steadily increased through the aleurone, subaleurone and endosperm layers, so that the highest Zn counts were detected in the endosperm. Because Zn count was higher than Fe for most of the scan, the Fe:Zn signal ratio in all three layers was ,1 ( Table 2). The line scan of WT grain provides novel insights into the distribution of Fe in the outer layers of rice grain. While significant quantities of Fe and Zn are localized in the aleurone cells, and therefore likely bound to PA, the subaleurone and endosperm contain substantial quantities of Fe and Zn that are likely to be bioavailable; particularly in the endosperm region where only trace P was detected. What molecule(s) the micronutrient metals are chelated to in this region remains unknown.
The line scans of OE-OsNAS2A grain revealed a radically different distribution and quantity of Fe and Zn in all three layers ( Figure 5B). A major difference was apparent in the relative amounts of Fe and Zn, with more Fe than Zn counts detected in most of the scanned region. This is essentially the reverse of what was seen in WT grain, and is reflected in Fe:Zn signal ratios of .1 for aleurone and subaleurone cells, and close to 1 in the endosperm, for OE-OsNAS2A grain (Table 2). Furthermore, the Fe count did not tail off after a peak in the subaleurone layer, rather, it continued to peak at progressively higher levels in the subaleurone and endosperm layers. In fact, the highest count of Fe in OE-OsNAS2A grain (6958 at 101 mm) occurred in a region of endosperm where only trace counts of P (21) were detected, thereby indicating that most of the Fe could not be complexed with PA and may be readily bioavailable. To better visualize the distribution of Fe in OE-OsNAS2A grain, and how it differs from that of WT, the Fe counts for both grain types were plotted on linear scale in Figure 6. Increased Fe concentrations in the aleurone, subaleurone and endosperm layers of OE-OsNAS2A grain, relative to WT, are readily apparent in this chart and, furthermore, the enrichment of Fe in the endosperm layer of transgenic grain is clear. The Zn counts in OE-OsNAS2A grain were also significantly higher than those of WT grain, but followed a similar trend to that of WT by gradually increasing towards the endosperm.
It is tempting to speculate that the increased concentrations of Fe and Zn in aleurone, subaleurone and endosperm layers of OE-OsNAS2A grain are present as complexes with NA. NA is known to have high binding affinities for Fe and Zn at alkaline pH, while the Fe 2+ NA complex in particular demonstrates unusually high kinetic stability that does not show autoxidation at physiological pH ranges [11]. Preferential binding of NA to Fe 2+ as it is transported to the grain through phloem tissues, and high stability of the Fe 2+ NA complex within the grain, could explain why the trend of Zn.Fe counts in WT grain is reversed in OE-OsNAS2A grain. The hypothesis that Fe and Zn are bound to NA is bolstered by recent analyses of grain from previously mentioned OsNAS3 activation tagged line of rice which contains nine-fold more NA in the grain [21]. Whereas WT and the OsNAS3 activation tagged grain did not differ with regards to the amount of PA-bound Fe, the OsNAS3 activation tagged grain had seven-fold more Fe bound to a low molecular weight mass compound that is likely to be NA. A similar result was found in grain of the OsNAS2 activation tagged line with regards to Zn [22]. We are currently using X-ray Absorption Near Edge Structure (XANES) to identify compounds that bind to Fe and Zn in the endosperm of WT and OE-OsNAS2A grain. XANES should also yield speciation information for Fe (Fe 2+ vs. Fe 3+ ), and we expect the OE-OsNAS2A grain to have more Fe 2+ relative to Fe 3+ due to preferential chelation of Fe 2+ by NA under aerobic conditions [11].
More than 2 billion people are currently afflicted by iron deficiency, a serious nutritional problem that has been exacerbated by high dependences on nutrient poor cereal crops in many developing countries of the world. Billions also suffer from equally devastating micronutrient disorders such as Zn and Vitamin A deficiency. Worryingly, micronutrient malnutrition problems may become even more prevalent as the Earth's atmospheric concentration of carbon dioxide (CO 2 ) continues to rise. Many studies have shown that carbon enrichment, while increasing productivity of many crops, also causes significant decreases in the concentration of key micronutrients such as Fe and Zn [33]. In light of these results it is imperative that conventional breeding and biotechnology are exploited to the fullest extent to increase nutritional composition of the world's major food staples. Using constitutive overexpression of single members of the OsNAS gene family, we have produced biofortified rice lines that contain significantly enhanced Fe and Zn concentrations in polished grain. The use of rice genes to increase the micronutrient concentration of rice shows that cisgenic plants could be developed using similar technology [34]. Most importantly, the Fe concentrations detected in particular OsNAS2 overexpressing lines meet or surpass the target concentration for Fe biofortification of rice endosperm. The enhanced Fe concentrations are preferentially located in areas of the rice grain where they are unlikely to be bound by phytic acid and therefore likely to be bioavailable in human diets.

Plant growth conditions
Oryza sativa ssp. japonica cv. Nipponbare was used for all experiments. Seeds were germinated on moist filter paper for one week before transfer to 15 cm (1 L capacity) containers of University of California (UC) potting mix in a glasshouse or growth room maintained at 28uC day, 24uC night, 12 h light/ dark. The UC potting mix was prepared by mixing 1,200 litres of sterlised sand with 750 litres of peatmoss with the addition of calcium hydroxide (hydrated lime, 1 kg), calcium carbonate (agricultural lime, 1.8 kg) NPK fertilizer and 4 kg Osmocote per 1000 kg soil. Transgenic plantlets were grown under the same conditions. Grain harvested from plants was dried for 3 d at 37uC and then used for elemental, NA and m-XRF studies.
Vector construction and rice transformation RNA was extracted from 2-week old seedlings of japonica rice cultivar Nipponbare and used for cDNA synthesis. The OsNAS1 coding sequence was PCR amplified from cDNA with forward primer 59 -ATGGAGGCTCAGAACCAAGAGGTCG -39and reverse primer 59 -GTTAGACGGACAGCTCCTTGTTGGC -39 to yield a 1000 bp fragment containing the OsNAS1 cDNA; the OsNAS2 coding sequence was PCR amplified with forward primer 59-ATGGAGGCTCAGAACCAAGAGGTCG -39 and reverse primer 59 -ATGCACGCACTCAGACGGATAGCCT -39to yield a 991 bp fragment containing the OsNAS2 cDNA; and the OsNAS3 coding sequence was PCR amplified with forward primer 59 -ATGACGGTGGAAGTGGAGGCGGTGA -39 and reverse primer 59 -GGTGAGGTAGCAAGCGATGGAAGCA -39 to yield a 1072 bp fragment containing the OsNAS3 cDNA. The three PCR fragments were cloned, separately, into the Invitrogen GatewayH Entry vector pCR8H/GW/TOPOH. Error free sequences were then recombined into a modified pMDC100 vector [23] that placed the OsNAS coding sequences under the control of a dual CaMV 35S promoter (Figure 1). Embryogenic nodular units arising from scutellum-derived callus were inoculated with supervirulent Agrobacterium tumefaciens strain AGL1 (carrying the OsNAS overexpression vectors) and 200 l 21 geneticin-resistant shoots were regenerated after nine weeks using established protocols [35]. Rooted T 0 plantlets were transferred to the growth room in Jiffy peat pots, and moved to soil after 15 days.

Elemental analyses of rice grain
Samples consisting of approximately 25 unpolished grain, or 35 polished grain, were analyzed by ICP-OES to determine metal concentrations. Polished grain was produced using a modified non-contaminating Kett Mill with a milling time of 2 min 30 sec, as preliminary studies with KOH staining had shown this time period to be sufficient for removal of the bran layer.

Liquid chromatography-mass spectrometry (LC-MS)
The LC-MS method is based on published methods [36]. Four individual rice grains were obtained from each line, 20 mg of ground rice from each grain was weighed into separate Eppendorf tubes and 200 mL of EDTA solution (5 mM), containing the internal standard 2-aminobutyric acid (25 mM), was added. The EDTA was used to release any metal complexed with NA. The derivatization of nicotianamine involved mixing of 10 mL supernatant with 70 mL borate buffer (0.2 M; pH 8.8), followed by the addition of 10 mL of AQC solution (10 mM) in dry acetonitrile. The reaction mixture was then heated at 55uC for 10 min and analyzed by LC-MS. Chromatograms and mass spectra were evaluated using the MassHunter Quantitative analysis program (Agilent). Quantification was based on the external calibration curve method using the internal standard for error correction. A 1 mg/mL stock nicotianamine standard was prepared and subsequently diluted with the EDTA solution (5 mM) to prepare calibration standards in the concentration range between 2.75 -100 mM. Synchrotron X-ray fluorescence spectroscopy (m-XRF) Thin longitudinal sections of rice grain were obtained using published methods [30]. Briefly, grains were glued to a plastic support and then sliced using a vibrating blade microtome in order to obtain a flat surface (Leica VT1000 S). A piece of Kapton polyimide film was then pressed on the surface of the sample with the blade of the microtome cutting underneath. In this way, longitudinal sections, 70 mm thick, were directly placed on Kapton tape without the need for embedding. Two longitudinal sections each of WT and OE-OsNAS2A grain were analyzed. m-XRF elemental maps were collected at the X-ray Fluorescence Microscopy (XFM) beamline at the Australian Synchrotron. Whole grain elemental maps were collected at 7.5 keV using a 96-element prototype Maia detector [37]. The detector was placed perpendicular to the beam path at a distance of 20 mm and was used to collect the full spectra fluorescence signal from the sample. The samples were analysed continuously in the horizontal direction with steps of 1.25 mm in the vertical direction. The sample stage was set to a speed of 2 mm s 21 , resulting in a pixel transit time of roughly 0.6 ms. The full XRF spectra were analysed using GeoPIXE [38,39].
In order to map P distribution together with the distribution of the micronutrients of interest, a Vortex detector was employed as the Maia detector is unable to analyse elements lighter than K. Rectangular areas of 406135 mm were mapped at the margin of grains with a dwell time for pixel of 1 sec. The line scans were obtained by laterally averaging the rectangular areas scanned (represented by a green box in Figure 4A). Table S1 Average concentrations