**3. Discussion**

The frequent lack in intake of essential mineral elements in human diets can be significantly improved by shaping agronomic practice and/or designing staple crops to generate mineral-dense produce [7]. Agronomic approaches have been efficient when (i) the soil contains insufficient amounts

of certain element(s), which can be added to the agricultural system as fertilizers or (ii) when changes to phytoavailability of elements in the rhizosphere are required and pH-related intervention can offer solutions. On the other hand, variability in the elemental composition of the edible produce can be exploited to (i) introduce cultivars with superior mineral-use e fficiency, provided there is no penalty to agronomically important traits or (ii) to identify candidate genes for future genetic optimisation [9]. After these interventions are implemented and a produce with the largest possible inherent concentration is available, the Fe status of the individual and other food components (e.g., dietary fibre, organic acids) will still play crucial roles in the availability of a certain element, with Fe being particularly problematic [34], further complicating the e fforts to ensure optimal nutrition in humans.

### *3.1. Awned Cultivars Accumulate More Ca and Mn But Less Fe in Grain than Awnletted Wheat Cultivars*

We investigated the diversity in grain mineral element accumulation in 20 wheat cultivars and found that there is a link between the awn length and the Ca, Mn and Fe concentrations (Figure 1). In wheat, awns have been shown to have transpiration and photosynthetic activity [20], thus their presence could contribute to the translocation of elements taken up by roots on the one side and/or to the phloem-driven (re)allocation of assimilates on the other side, thereby a ffecting mineral element density in the grain. This connection has, however, not been investigated so far. Awnletted wheats have been shown to produce significantly more grains per unit area and per spike, resulting in a significant reductions in grain size and an increased frequency of small, shrivelled grains [35]. Our observations did not fully support this report, since there was no significant di fference in 1000-grain weight between the awned and awnletted cultivars, both for the established agronomic values and for those from our experiment (Table S1). The observed di fferences in elemental concentrations are conceivably not a consequence of dilution by grain weight, but rather arise from genetic di fferences in uptake, allocation and/or mobilisation or Ca, Mn and Fe in these cultivars. It may however be, that by accumulating larger Fe concentration in the embryo and particularly in the nucellar projection in awned cultivars compared to awnletted cultivars as observed in our study, less Fe is being translocated to other grain tissues, resulting in larger total Fe concentration in the whole grain.

On average, total Fe concentrations in grain from our experiment showed 2.03-fold variability (Figure 2), which is in a similar range as the value 1.76 reported for 150 bread wheat cultivars [14]. The whole grain Fe concentrations in all studied cultivars exceed the reported maximum value (50.8 mg Fe·kg−1) for the bread wheat cultivars [14], but were within the range of observations in spelt (*Triticum spelta* L.) grain, for which up to 99 mg Fe·kg−<sup>1</sup> was found, but with significant variation due to the year and the location [36]. Iron concentration in barley grain exceeding 100 mg·kg−<sup>1</sup> has also been reported [37].

The positive correlation between P and Fe concentrations was observed (Figures S2 and S3) in agreemen<sup>t</sup> with findings in bread wheat [14] and in spelt [36], suggesting that the increased Fe concentration may be accompanied by a decreased bioavailability (i.e., due to phytate), presenting a further challenge for biofortification. However, the positive relationship has not been observed in all instances [38,39] and a strong genotype, environment and/or genotype × environment interaction has been shown to a ffect the relationship [40,41]. The positive correlation between grain Fe and Zn observed also in our experiment (Figures S2 and S3), seems to be quite stable as it has been consistently reported, over di fferent seasons and locations, for example in wheat [14,38], durum wheat [39], spelt [36], and barley [15,37]. It could be attributed to the limited specificity of transporters and metal ligands for either Zn or Fe [42], suggesting that the increased density could be achieved simultaneously for a larger number of trace elements. Still, the issues with the bioavailability of these trace elements will have to be addressed before any such observations are implemented into breeding strategies.

### *3.2. Iron Speciation and Iron Ligands in Wheat Grain Are Stable across Cultivars Di*ff*ering in Total Iron Concentration*

To complement the current knowledge on Fe speciation and Fe ligands in whole wheat grain [25] and some of its tissues [28] we studied these traits in five grain tissues of four wheat cultivars. Initial X-ray fluorescence mapping in frozen hydrated cross-sections of the grain revealed Fe hotspots and provided information on P and S distribution as well. Based on the co-localisation of Fe, P, and S (Figures 4 and 5), selected regions of interest (Figure 3) were easily distinguished and high-Fe pixels were investigated by micro-XANES. There was no apparent difference in the Fe speciation or Fe ligands in the grain of awned and awnletted cultivars, regardless of the differences in the Fe concentrations (Figure 6a,c) indicating that the total Fe concentration in the grain does not influence Fe species or Fe ligands. Similar results were found by Singh et al. [25], who included in the analysis a wild relative of common wheat, *Aegilops kotschyi* Boiss., which contains up to three times larger Fe concentrations than grain of wheat landraces. The relative amounts of ferrous (36%) species in wheat grain—up to one third of total Fe as assessed by micro-XANES in our experiment (Figures 4–6)—was somewhat larger than in previous findings in whole wheat with values between 14% and 24% [25]. Furthermore, on average 53% of Fe was bound to phytate. The proportion of non-phytate ligands is in agreemen<sup>t</sup> with another study [28] on wheat and Tartary buckwheat grain, where 22% of total Fe was bound to non-phytate ligands [33]. Perplexingly, no direct association of the Fe ligand profile could be found with Fe bioavailability (assessed using Caco-2 cell system) in Tartary buckwheat sprouts containing a much larger proportion (55%) of Fe3<sup>+</sup> citrate [33]. In legume seeds, which store large amounts of Fe in ferritin, progressive accumulation of phytate with seed maturity limits Fe bioavailability, as demonstrated by comparing immature and mature pea (*Pisum sativum* L.) seeds [43]. Apparently, more studies are required to reach consensus on the connection between Fe ligands and Fe bioavailability from different plant-based sources.

### *3.3. Distinct Tissue Specificity in Iron Speciation, Iron Ligands and Iron Concentration in Wheat Grain*

At the tissue level a large variability in the Fe speciation and Fe ligands was found (Figure 6) in line with a previous report [28], suggesting differences in bioavailability of Fe from different grain tissues. All tissues contained >60% of ferric species in line with analyses in whole grain [25]. The only exception was nucellar projection, where equal amounts of Fe3<sup>+</sup> and Fe2<sup>+</sup> were found (Figure 6b). Regarding the Fe ligands, the presence of non-phytate ligands in the pericarp and nucellar projection was particularly striking (Figure 6d). The pericarp is a maternal tissue and is typically accessed by xylem, which may be a reason for the Fe-citrate pool. There is some evidence that Fe from the Tartary buckwheat pericarp [33] and from the wheat bran [44] is relatively bioavailable. While Tartary buckwheat pericarp is not edible, the inclusion of wheat bran in the meal would, despite markedly increased phytate concentration, outweigh the negative effect of phytate-induced losses in bioavailability.

The major part of cereal grain loading with micro and macronutrients is presumed to take place through vasculature, which in mature grain is shrunk and borders the pigment strand, which in turn borders the nucellar projection (Figure 3). During grain maturity the tissues within the crease undergo a series of transformations, and nutrients passing into the endosperm must cross the pigment strand, the nucellar projection, and the endosperm transfer cells [45]. The nucellar projection is a part of the nucellar tissue that faces the vascular tissue, has a morphology characteristic of transfer cells [46] and contains large concentration of Fe, while the pigment strand was rich in Mn, making these two tissues clearly distinguishable [12,25,32]. As already follows from the apparent lack of co-localisation of Fe and P in the nucellar projection (Figures 4 and 5; [12,25]) only non-phytate ligands were found there. Our analysis also confirms the presence of nicotianamine in the nucellar projection reported previously in [28]. For the first time, ferric and ferrous nicotianamine compounds are demonstrated to exist in a grain tissue (together 45% of Fe ligands). In addition, some Fe was bound to sulphate (22%), in agreemen<sup>t</sup> with previous reports [28] and predicted from S localisation in this tissue (Figures 4–6, Figures S4 and S5) and exclusion of P [25].

In embryo, scutellum and aleurone, the only non-phytate ligand found was citrate, which is in agreemen<sup>t</sup> with results for cotyledons in Tartary buckwheat grain [33]. Most of the Fe present in the liquid endosperm in pea was found as a mixture of Fe3<sup>+</sup> citrate and malate [47]. The mixture has been shown to undergo an ascorbate-driven reduction [47] which makes Fe more mobile within the developing seedling. Ferric citrate-malate complexes have been demonstrated as the main form of Fe circulating in pea (*Pisum sativum* L.) plants [19]. In pea, the mother plant transports Fe3<sup>+</sup> malate/citrate complexes via the seed coat to the embryo, which in turn secretes ascorbate to reduce Fe3<sup>+</sup> to Fe2<sup>+</sup> for uptake during germination [47]. These observations sugges<sup>t</sup> active participation of all grain tissues and not only crease tissues in Fe loading of the grain and seed.
