Using natural variation to understand plant responses to iron availability

Abstract

Iron bioavailability varies dramatically between soil types across the globe. This has given rise to high levels of natural variation in plant iron responses, allowing members of even a single species to thrive across a wide range of soil types. In recent years we have seen the use of genome-wide association analysis to identify natural variants underlying plant responses to changes in iron availability in both Arabidopsis and important crop species. These studies have provided insights into which genes have been important in shaping local adaptation to iron availability in different plant species and have allowed the discovery of novel regulators and mechanisms, not previously identified using mutagenesis approaches. Furthermore, these studies have allowed the identification of markers that can be used to accelerate breeding of future elite varieties with increased resilience to iron stress and improved nutritional quality. The studies highlighted here show that, in addition to studying plant responses to iron alone, it is important to consider these responses within the context of plant nutrition more broadly and to also consider iron regulation in relation to additional traits of agronomic importance such as yield and disease resistance.

Exploring natural variation is a powerful approach to understanding plant responses to iron availability and has allowed the discovery of novel genetic regulators and mechanisms that have shaped plant adaptation.

Introduction

More than 400 million years have passed since plants colonized land, transforming the terrestrial biosphere and creating a world within which future species would come to thrive. As sessile organisms, plants must make effective use of the resources available within their very local environment. This includes vital nutrients such as iron which, although highly abundant in the Earth’s core, is often largely unavailable to plants (Morrissey and Guerinot, 2009). Plants depend on iron for a wide range of biological processes including electron transport—a central component of both photosynthesis and respiration (Zhang et al., 2019). However, when present in excess, iron can prove toxic and thus while plants must work hard to scavenge this often-scarce resource, they must do so with caution (Connolly and Guerinot, 2002; Morrissey and Guerinot, 2009; Zhang et al., 2018). Improving our understanding of how plants cope with iron stress would allow us to develop future, more stress-resilient crops. Furthermore, finding novel ways through which we can maximize iron content in crop plants could contribute significantly to human health, where low dietary iron continues to be a key concern (Abbaspour et al., 2014).

Much of the progress made in understanding plant responses to iron has been the result of mutant studies in Arabidopsis thaliana (Arabidopsis), which have enabled the discovery of key regulators of iron uptake (Robinson et al., 1999; Vert et al., 2002), transport (Schuler et al., 2012), and storage (Kim et al., 2006). While this is a valuable approach, there are limitations. For example, the majority of Arabidopsis research is focused on a single accession, Col-0, which may risk drawing general conclusions not widely applicable at the whole-species level. Furthermore, while mutant screens facilitate the identification of genes underlying mutant phenotypes, these loci do not necessarily relate to phenotypic diversity observed between individuals and may not be relevant for selection and adaptation.

In nature, iron content varies drastically between soil types, and the availability of iron to plants is influenced by many factors including soil pH, soil redox state, and soil aeration (Johnson et al., 2012). Iron exists naturally in the soil in two ionic states, Fe³⁺, also known as ferric iron, and Fe²⁺, or ferrous iron. The majority of plants, with the exception of graminaceous species, make use of iron in its reduced, ferrous state. The balance of Fe²⁺ and Fe³⁺ in the soil varies with oxygen availability. In highly aerobic soils, there is high potential for iron oxidation, which reduces iron bioavailability for most plants. In contrast, in oxygen-depleted soils, such as those that experience flooding, there is risk that very high levels of Fe²⁺ will accumulate (Sahrawat, 1979; Becker and Asch, 2005; Pezeshki and DeLaune, 2012). Soil pH has a particularly dramatic effect on iron bioavailability. Under alkaline conditions, Fe–hydroxyl complexes form, resulting in reduced iron solubility and, consequently, reduced uptake by plants. Alkaline soils, such as calcareous and saline soils, are widespread across the globe, particularly in semi-arid and arid climates, and it is estimated that as much as 30% of the Earth’s soils can be considered iron deficient (Chen and Barak, 1982; Lindsay and Schwab, 1982). In acidic soils, often the result of poor drainage, high sulfide content, or high rates of organic matter decomposition, iron solubility is increased, again resulting in potential iron toxicity to local plants (Johnson et al., 2012). Of course, iron-deficient and iron-toxic soils represent extreme conditions forming just part of a wide spectrum of soil iron profiles. This natural variation in iron availability has resulted in a high level of natural variation in plant iron responses. While members of a given species are, by definition, largely indistinguishable at the genetic level, subtle differences in DNA sequence arise as a result of random mutation and local selection pressures which, over time, give rise to distinct populations comprising unique genetic strains. While chance events (genetic drift) play a key role in determining the prevalence with which specific alleles persist within a population, allele frequencies are also shaped by the level of fitness they confer (Chen et al., 2019). If we consider members of a given species growing in soils varying in iron availability, while certain alleles may be favored within the context of one soil type, they may prove benign or even detrimental within the context of another. These differences in selection pressure will, over time, become reflected in the DNA sequence, and exploring the natural variation that results can allow us, using quantitative genetics methods, to assign allelic variation with variation in adaptive traits of interest such as tolerance to iron stress. One widely used approach has been to explore variation segregating between progeny resulting from a biparental cross. However, exploring variation between just two parental genotypes is unlikely to reflect the level of variation available across that species and, due to the low level of recombination events that would have taken place since the establishment of the mapping population, mapping resolution is limited (Xu et al., 2017). An alternative, or indeed complementary, approach is to make use of multiparent populations which increases the level of genetic variation available (Mackay and Powell, 2007; Xu et al., 2017). Another powerful approach is genome-wide association (GWA) analysis, which, making use of large collections of diverse accessions, provides a more comprehensive overview of the natural variation available within a given species and, given the many historic recombination events that would have taken place throughout the evolutionary history of these accessions, provides increased mapping resolution (Mackay and Powell, 2007; Xu et al., 2017). Several studies have made use of GWA analysis to understand plant responses to iron. This review aims to celebrate this progress and, taking examples from both Arabidopsis and important crop species, highlight the different ways through which natural variation can be explored to understand the complexities of iron regulation.

Quantifying plant iron responses: possible approaches and available resources

What we learn from methods such as GWA analysis will ultimately depend on the phenotype used. An ideal phenotype should provide an accurate readout of the biological process of interest and should be easily quantifiable. Plants respond rapidly to shifts in local iron availability by adjusting root growth (Ward et al., 2008; Long et al., 2010; Li et al., 2016). Tracking changes in root growth response (RGR) may thus provide valuable information. Later in development, changes in iron availability may result in altered loading of iron into developing seeds and leaves, the latter of which, depending on severity, can result in observable changes in leaf color. The differences in timing of these iron-related phenotypes may allow us to uncover genetic regulators underlying different stages of the plant iron response pathways. While it may not be feasible to quantify RGRs at the required throughput in natural soil conditions, recent efforts have resulted in the development of high-throughput phenotyping platforms for the assessment of these traits within a laboratory setting (Slovak et al., 2014; Gioia et al., 2016; Shahzad et al., 2018). Using flatbed scanners, Slovak et al. (2014) developed a method that allows high-resolution image capture across thousands of seedlings grown on agarose plates, per day. To facilitate image analysis, a semi-automated image quantification pipeline, BRAT, was also engineered (Slovak et al., 2014). An alternative option is the GroScreen-PaGe phenotyping system described by Gioia et al. (2016), which involves growing seedlings on germination paper suspended in liquid nutrient solution. This method enables screening of larger root systems and has already proven successful in studying natural variation in root traits for wheat, barley, and rapeseed (Gioia et al., 2016). Shifts in leaf pigmentation such as that resulting from iron deficiency-induced leaf chlorosis can be scored visually or can be quantified using colorimetric methods. Iron levels in seeds or other organs can be assessed visually using histochemical methods, including Perl staining (Perls, 1867; Meguro et al., 2007). An alternative would be to use X-ray fluorescence (Sosa et al., 2018) or inductively coupled plasma optical emission spectrometry (Bettinelli et al., 1998), which, although they require access to expensive, highly specialized equipment, promise increased accuracy.

Finally, vast transcriptional shifts have been reported in response to changes in iron availability (Yang et al., 2010; Rodríguez-Celma et al., 2013). Using gene expression data as trait input for GWA analyses, or expression quantitative trait locus (eQTL) analyses, can provide a means through which both cis and trans transcriptional regulation of key iron-response genes can be identified (Rodríguez-Celma et al., 2013). In Arabidopsis, leaf transcriptome data are now available, summarizing natural variation in expression across a large number of accessions (Schmitz et al., 2013; Dubin et al., 2015; Kawakatsu et al., 2016). Similar resources have been developed in crop species including Brassica napus (Harper et al., 2012). A number of studies have shown that the use of the leaf transcriptome can serve as a proxy for gene expression patterns in distant tissues, including roots and seed (Harper et al., 2012; Li et al., 2019; Miller et al., 2019). These resources may therefore prove powerful for studying various aspects of plant responses to iron. Indeed, at least in Arabidopsis, very high levels of natural variation in the expression of genes related to multiple levels of plant iron responses can be seen across accessions (Kawakatsu et al., 2016). Examples of this are illustrated in Fig. 1 for genes involved in iron uptake (FRO2 and AHA2), iron transport (YSL1 and NAS4), and iron storage (VTL1 and NRAMP4). Making use of these data in eQTL analysis may prove powerful in identifying novel regulators upstream of these genes.

Fig. 1.

Profound natural variation of iron homeostasis gene expression indicates a high potential for using eQTL analysis to uncover regulators of iron responses in Arabidopsis. Histogram plots of publicly available transcriptome data (Kawakatsu et al., 2016) showing the frequency distribution of expression values (RPKM) for key iron uptake (AHA2 and FRO2), iron transport (OPT3 and NAS4), and iron storage (NRAMP4 and VTL1) genes across 728 accessions. These data, revealing high levels of natural variation across accessions, highlight the potential for using eQTL analysis to further our understanding of the transcriptional regulation of iron-response genes in Arabidopsis.

Arabidopsis: a model for exploring natural variation in plant iron responses

In recent years we have seen that the value of Arabidopsis extends beyond its use in classical genetics and that it also serves as a good model system within which to explore natural variation underlying complex traits. As a species, Arabidopsis grows naturally across Europe, Asia, and East Africa, and has been introduced to parts of America and Australia (Shindo et al., 2007; 1001 Genomes Consortium, 2016). This broad geographic range, at least in part, reflects the capacity of this species to adapt to different soil types varying in micronutrient availability. One of the earliest studies highlighting the potential for using natural variation to understand plant responses to iron availability specifically came from a transcriptome study, comparing the expression of genes in response to iron deficiency, across a small group of divergent Arabidopsis accessions (Stein and Waters, 2012). This study revealed extensive genotypic differences, with only 10 genes commonly differentially expressed in response to iron deficiency. Such a narrow gene set is striking and highlights the presence of significant natural variation in iron responses in Arabidopsis (Stein and Waters, 2012). Since this initial insight, we have seen the first examples of GWA analyses in Arabidopsis aimed at understanding plant responses to iron availability. These studies, each making use of natural variation in RGR to iron, provide examples of how exploring this natural variation can not only further our understanding of already known iron regulatory genes but can also fuel the discovery of novel regulators that underlie these responses.

Exploring natural variation in root growth responses to iron in Arabidopsis

Exploring variation across 134 Swedish Arabidopsis accessions, Satbhai et al. (2017) showed that non-coding variation at the FRO2 locus underlies natural variation in RGRs to iron deficiency. This study revealed that natural variants at the FRO2 locus are able to differentially complement the reduced root growth phenotype of fro2 mutants under iron deficiency and that this is coupled with both differential FRO2 expression and ferric chelate reductase activity. The pattern of intronic and exonic single nucleotide polymorphisms (SNPs) in contrasting accessions at the FRO2 locus supports the possibility that FRO2 RNA or protein stability might play a role in the observed allele-dependent variation (Satbhai et al., 2017). Iron deficiency-dependent FRO2 RNA stability has been indicated by the induction of FRO2 mRNA levels in response to iron deficiency even when it was driven by the ubiquitous 35S promoter, which should not respond to Fe deficiency (Connolly et al., 2003). FRO2 regulation at the protein level has been indicated by Johnson and Vert (2016) where, under iron-replete conditions, FRO2 is targeted for K63 ubiquitylation, a protein modification known to regulate the stability of proteins at the plasma membrane. One intriguing observation made in this study is that FRO2 was identified as a regulator of RGRs to iron deficiency in a GWA analysis carried out on plants grown on ferrozine and thus expected to be completely deprived of iron. This seems surprising given that FRO2 is known to facilitate iron uptake (Yi and Guerinot, 1996) and thus in the absence of iron we may not expect natural variation at this locus to influence growth. While this may suggest that FRO2 has other, unknown functions that serve to regulate RGRs to iron deficiency, there are two alternative explanations. One is that the phenotypes observed under iron deficiency are the result of altered iron content in the maternal seed which provides an important source of iron during early seedling development. The other is that the ferrozine might not have entirely removed all traces of iron in the medium. Regardless of the mechanism employed, the results of this study indicate that selection for allelic variation at the FRO2 locus has been important in shaping local adaptation to iron deficiency in Arabidopsis.

While it is clear that there is high value in exploring natural variation in RGRs specific to iron deficiency, a study by Bouain et al. (2019) highlights that the regulation of these responses can vary in the presence of additional nutrient stresses. For example, it has been noted that, when coupled with phosphorus depletion (–P), iron depletion (–Fe)-induced root growth phenotypes can be rescued (Ward et al., 2008). Bouain et al. (2019) have shown that this phenomenon is common across 227 natural Arabidopsis accessions and, by combining GWA analysis with gene network and classical reverse genetics approaches, they identified novel alleles, genes and pathways that may coordinate this response. Using GWA analysis, natural variation in RGRs to –Fe and –P individually and in combination (–Fe–P) were explored. To more specifically investigate the mitigation of –Fe root growth phenotypes under –P, they also assessed variation in the difference in root growth under –Fe and –Fe–P, expressed as [(ΔRGR (–P–Fe, –Fe)]. No common associations were observed for –P and –Fe. Furthermore, very little relationship was seen between those markers associated with RGRs to individual nutrient deficiencies and the nutrient stresses combined, suggesting that distinct regulatory mechanisms underlie single and combinatorial nutrient stress responses. For [ΔRGR (–P–Fe, –Fe)], four associating loci were uncovered. Three of these loci were significantly associated with RGRs under –Fe–P. The fourth locus, however, was identified for natural variation underlying the difference in RGRs observed under –Fe and –Fe–P specifically. In close proximity to associating markers at this locus, FRO4 was identified. This gene, as described for FRO2, is involved in iron reduction (Wu et al., 2005). Exploring associations for [ΔRGR (–P–Fe, –Fe)] with a less stringent significance threshold, revealed 949 associating SNPs and, based on proximity to these associating markers, 92 candidate genes (Bouain et al., 2019). Assessment of this gene list using AraNet (Lee and Lee, 2017) showed over-representation of genes involved in chromatin modification, the cell cycle and DNA replication. For the majority of candidate genes, T-DNA mutants were identified. Mutants defective in VARIANT IN METHYLATION 1 (involved in DNA methylation) (Woo et al., 2007), FORMIN-LIKE PROTEIN 6 (involved in cytoskeleton organization) (Favery et al., 2004), and VOLTAGE-DEPENDENT ANION-SELECTIVE CHANNEL PROTEIN 3 [involved in regulation of reactive oxygen species (ROS)] (Zhang et al., 2015) showed altered RGRs specific to –Fe–P (Bouain et al., 2019). These findings provide insight into the mechanisms that may underlie the complex but fascinating interaction observed between iron and phosphorus nutrition in plants.

In addition to this progress made in understanding Arabidopsis RGRs to iron deficiency, a recent study by Li et al. (2019) provided the first insights into natural genetic variants underlying RGRs to iron toxicity. Iron toxicity can be detrimental to plant fitness and can be a key yield-limiting factor in waterlogged soils and specific soil types in crop species. This issue is particularly pronounced in rice (Fageria et al., 2008; Stein et al., 2014). Li et al. (2019) reveal that variation at the GSNOR (S-nitrosoglutathione reductase) locus in Arabidopsis, that results in its differential expression across accessions, contributes to natural variation in RGRs to iron toxicity. Making use of eQTL analysis, natural variation in GSNOR expression was explored across 665 Arabidopsis accessions, which indicates that variation at the GSNOR locus itself (indicative of cis regulation) underlies natural variation in the expression of this gene (Li et al., 2019). Knockout mutants for this gene display increased sensitivity to iron toxicity, and allelic variants segregating at the GSNOR locus can differentially complement this mutant phenotype. Further exploration of this variation may allow for the discovery of some novel transcriptional regulator of GSNOR and may thus further contribute to our understanding of plant responses to iron toxicity. GSNOR is involved in S-nitroglutathione degradation, which is the product of the reaction of nitric oxide (NO) and glutathione (GSH). Li et al. (2019) show that treating seedlings with an NO donor could significantly affect root growth, suggesting that the phenotypes observed in the gsnor mutants may be the results of NO accumulation. Consistent with this, they show that in wild-type (WT) plants NO accumulates in the root meristem (where high GSNOR expression is observed) in response to iron toxicity and that this is observed even under control conditions in mutant plants. In addition to a clear role in nitrosative stress, they show that GSNOR is required to avoid H₂O₂-induced oxidative stress, with the gsnor mutants displaying increased sensitivity to H₂O₂. Closer inspection of root meristem cells revealed that under iron toxic conditions, the root meristem is significantly smaller in knockout mutants relative to the WT and that there is a significant increase in cell death. Li et al. (2019) propose that under iron toxic conditions, GSNOR serves to protect meristem cells from the damage caused by iron-dependent NO-induced nitrosative and H₂O₂-induced oxidative toxicity in Arabidopsis. This study was the first to identify causal variation underlying iron toxicity responses in plants and to describe a mechanism whereby the effects of iron toxicity are not solely explained by H₂O₂ production. Furthermore, using CRISPR (clustered regularly interspaced palindromic repeats), Li et al. (2019) show that rice plants carrying mutations at the GSNOR locus also display increased sensitivity to iron toxicity. Iron toxicity continues to limit rice breeding worldwide (Sikirou et al., 2016), and thus the results of this study could have widespread implications for the development of future elite rice varieties expressing increased tolerance to iron toxicity.

Identifying natural variants underlying resilience to iron stress and seed iron content for crop improvement

In addition to the progress made in making use of natural variation to understand iron responses in Arabidopsis, we have seen examples of studies aimed at understanding iron regulation in different crop species. In addition to its value in uncovering gene function, identifying markers associated with a trait of interest provides breeders with the tools necessary to select for important genetic variation by marker-assisted selection (MAS). As previously mentioned, low iron intake continues to be a widespread issue in human nutrition (Abbaspour et al., 2014). Uncovering natural variants that confer increased iron content in edible portions of crop plants would help to alleviate this problem. Furthermore, both iron deficiency and toxicity continue to be key yield-limiting factors for a wide range of important crop species. Identifying alleles that confer resilience to iron stress could thus contribute considerably to future yield improvements. Unlike Arabidopsis, crop species have been subjected to significant artificial selection. While this of course limits the natural variation available to explore using quantitative methods such as GWA analysis, as we will see, even within these relatively limited gene pools, natural variation in key iron-related traits persists, suggesting that there is potential for the improvement of these traits in future elite varieties.

Exploring natural variation in iron-related leaf chlorosis and leaf bronzing: key yield-limiting factors in crops

Iron deficiency chlorosis (IDC), characterized primarily by leaf yellowing, is a common indicator of iron deficiency stress in plants. In crop species, such as soybean, IDC is highly correlated with yield losses (Bai et al., 2018). Exploring variation across 270 advanced soybean accessions, Mamidi et al. (2014) identified seven loci associated with variation in IDC. A subset of these loci had been reported previously in QTL analyses, including an association on Gm03, which represents a historic IDC QTL (Lin et al., 1997, 2000). Based on proximity to the associating markers, potential candidate genes that may underlie the natural variation in IDC observed were identified, including homologs of Arabidopsis AHA11, NAS3, and FRO2. AHA11, encodes an ATPase membrane protein involved in the extrusion of protons from cells to increase acidification of the surrounding environment (Haruta and Sussman, 2012). Proton extrusion by AHA11 would provide a mechanism through which iron solubility could be improved in the calcareous soils in which much of our soybean crops are grown (Mamidi et al., 2014). NAS3 is involved in the synthesis of nicotianamine which is vital for the phloem-based transport of iron to sink organs (Schuler et al., 2012). In Arabidopsis, plants defective in nicotianamine biosynthesis display severe leaf chlorosis (Klatte et al., 2009). Finally, FRO2 is, as we have seen, important in the reduction of iron in the rhizosphere, which serves to increase iron bioavailability and consequently uptake (Robinson et al., 1999). While further work would be required to prove causality of these genes in regulating IDC, identifying genes such as FRO2 and NAS3 suggests that natural variation in responses to iron deficiency, that facilitate iron uptake and transport, may be key in contributing to variation in IDC across these soybean accessions. In addition, numerous disease resistance genes were also identified in close proximity to associating markers (Mamidi et al., 2014). Previous studies have shown that pathogens compete with plants for an iron source (Aznar et al., 2015). These disease resistance genes may therefore also represent plausible candidates. However, it is also possible that these genes are just by chance in close proximity to the true causal variation, which, depending on the alleles co-segregating, may make it complicated to select for increased tolerance to IDC without compromising immunity.

A more recent GWA analysis study, reported by Aseffa et al. (2020), explored natural variation in IDC across a more diverse panel of soybean accessions collected across 27 countries. This study again revealed important natural variation at the historic Gm03 locus which, using a SNP clustering analysis, they deconstruct into four distinct linkage blocks. This suggests that what has long been thought to be a single QTL may in fact represent a genomic hotspot comprising multiple IDC-related genes. Based on proximity to associating markers, differential expression in response to iron and previous known functions in regulating iron deficiency responses, candidate genes within these discrete blocks were identified. One of the four intervals contained two homologs of the Arabidopsis bHLH38—a transcription factor known to regulate the expression of key iron uptake genes, FRO2 and IRT1 (Wang et al., 2007). In a second linkage disequilibrium (LD) block, they identified Glyma.03g128300 (AtGLU) (Assefa et al., 2020). This locus is predicted to encode a ferredoxin-glutamate synthase and, in Arabidopsis, knockout mutants for this gene display increased chlorosis (Coschigano et al., 1998). Further adding to the complexity of this Gm03 region, Assefa et al. (2020) identified interactions between the Gm03-associating SNPs and SNPs located in other genomic regions. This included an interaction between Glyma.03g128300 and distant SNPs located on Gm05 and Gm06. Using StringDB, the authors identified interactions between Glyma.03g128300 and two genes in close proximity to the Gm05- and Gm06-based SNPs. These genes were Glyma.05 g127900 (At1G72550), encoding a tRNA synthase, and Glyma.06 g206600 (At5G08110/AtHRQ1), encoding a helicase protein required for genome stability and repair (Assefa et al., 2020). This study provides the first evidence that epistasis is important in regulating IDC in soybean, and exploring the potential roles of these proteins in regulating IDC and, more specifically, their interaction with Glyma.03g128300, represents an exciting area for future research.

We have already seen that iron toxicity is a key issue in rice breeding. This is due to the often very high levels of ferrous iron in flooded paddy fields which can, when severe, lead to complete yield loss (Becker and Asch, 2005; Sahrawat, 2005). In rice, iron toxicity leads to what has been named ‘leaf bronzing’, the severity of which is closely correlated with yield losses (Matthus et al., 2015). Matthus et al (2015) explored variation in iron toxicity-induced leaf bronzing across 329 accessions of Asian rice (Oryza sativa L.). Correlation of leaf bronzing severity and shoot iron content across a subset of accessions showed that tolerance to leaf bronzing could be seen coupled with both high and low shoot iron content, suggesting that while some accessions achieve tolerance by preventing iron uptake, others employ mechanisms that promote tolerance despite high iron uptake. Using GWA analysis, this study revealed three highly associating SNPs for variation in leaf bronzing on chromosome 1. In close proximity to these associating markers, they identified two genes encoding glutathione S-transferases (GSTs), which have been implicated in plant oxidative stress responses (Sappl et al., 2009), making them promising candidate genes for phenotypes related to iron toxicity. Closer assessment of associating marker alleles segregating on chromosome 1 revealed haplotype groups representing the most tolerant and susceptible allelic combinations of the associating markers. In a subset of accessions representing these different haplotype groups, they explored genes in LD with the associating markers for sequence variation that displayed clear segregation with variation in leaf bronzing. Clear segregation patterns were observed in the two genes encoding GSTs. Furthermore, both genes displayed increased expression in response to iron toxicity. Finally, Matthus et al. (2015) show that the different haplotype groups varied significantly in dehydroascorbate reductase activity, which has previously been associated with GST protein function. While further work would be required to prove causality, this work indicates that natural variation in genes encoding GSTs may underlie variation in leaf bronzing in Asian rice.

Understanding iron regulation to drive biofortification efforts in crop species

A number of studies have been carried out with the aim of increasing seed nutrient content in important crop species. A study by Upadhyaya et al. (2016) made use of natural variation segregating between 92 chickpea accessions, with the aim to uncover loci regulating iron and zinc content of seeds. A high positive correlation was observed between the natural variation in zinc and iron content across accessions, suggesting that it should be possible to improve the seed content of both micronutrients simultaneously and that there may be common loci underlying the regulation of these traits. The GWA analysis combined global SNP data, comprising 16591 SNPs, mined by genome-wide GBS (genotyping by sequencing) with SNPs identified by more thorough sequencing of known candidate genes (Upadhyaya et al., 2016). The candidate gene screening approach revealed a significant association between iron content and natural variation at the FERRITIN1 locus. In addition, YSL1-, ZIP1-, and HMA2/4-localized SNPs were significantly associated with variation in both iron and zinc content. These genes are known to be involved in the regulation of iron and zinc storage and transport in Arabidopsis (Grotz et al., 1998; Le Jean et al., 2005; Sinclair et al., 2018). The global SNP analysis identified associating SNPs located in 12 further genes, including an auxin-related gene and a gene encoding an late embryogenesis abundant (LEA) protein. LEA proteins have been implicated in phloem iron transport (Conte and Walker, 2011), and interactions between auxin and various aspects of iron regulation in plants have been identified (Chen et al., 2010; Sun et al., 2017). This study also showed, across a subset of accessions, that the expression of all 16 associating SNP-containing genes is positively correlated with micronutrient content (Upadhyaya et al., 2016). The majority of the associating SNPs were located in genes not previously implicated in regulating seed iron and zinc content and may thus represent completely novel regulators.

A study by Hindu et al. (2018) explored variation across 923 maize inbred lines to study the genetic regulation of kernel iron and zinc content. This study identified 26 marker associations for kernel iron content. Despite iron and zinc content being positively correlated, only one common associating locus, located on chromosome 9, was identified for these traits. In close proximity to the associating markers, they identified gene model GRMZM2G311974, which encodes a NAC domain transcription factor family protein (Hindu et al., 2018). This family of transcription factors has been implicated in regulating iron and zinc content in wheat grain (Uauy et al., 2006). As a final example, on chromosome 1, they identified two SNPs located within gene model GRMZM2G302373 (Hindu et al., 2018). This gene is predicted to encode a protein with GST activity which has been previously implicated in stress responses to heavy metal toxicity (Hossain et al., 2012; Gao et al., 2020).

While in many cases high mineral content may be seen as a positive quality in breeding programs, some minerals, such as cadmium, are known to be toxic to humans and animals. Exploring natural variation in 10 key grain minerals, iron, calcium, cadmium, cobalt, copper, lithium, magnesium, manganese, nickel, and zinc across 123 synthetic wheat genotypes, Bhatta et al. (2018) show that wheat iron content is highly positively correlated with all other nine nutrients including cadmium. Using this synthetic wheat panel, the authors revealed two loci associated with grain iron content. One marker association, located on chromosome 3, was also associated with variation in zinc and manganese, suggesting that this locus may be used to co-select for increased iron, zinc, and manganese in wheat grain. Importantly, this region was not associated with variation in cadmium (Bhatta et al., 2018). The second association related to variation in seed iron content, located on chromosome 1, was, however, seen to co-segregate with variation in cadmium levels. In close proximity to associating markers at this locus, Bhatta et al. (2018) identified a sodium-translocating NADH-quinone reductase subunit A (TraesCS1A01G432900). NADH-quinone reductases are known to bind iron–sulfur clusters (Nakamaru-Ogiso et al., 2002) and are expected to have roles in protecting plants against oxidative stress (Wang and Maier, 2004). Taken together, and given the known relationship between both Cd and iron in the generation of ROS (Maksymiec and Krupa, 2006; Reyt et al., 2015), this may represent a good candidate gene underlying the natural variation observed for these related traits.

Conclusions and future perspectives

In recent years we have seen the first steps taken in exploring natural variation underlying plant responses to iron. This has fueled the discovery of important natural variation in known iron-response genes as well as completely novel regulators not previously identified using classical genetics approaches. These findings not only help to build our knowledge of regulators and mechanisms underlying plant responses to iron, but also provide information regarding genes that have been important in shaping natural populations facing different environmental challenges. In addition to the progress made using our model species, Arabidopsis, several studies have been successful in describing natural variation underlying iron-related phenotypes in important crop species. Assessment of genes in close proximity to markers associating with this variation revealed several genes known, in Arabidopsis, to regulate iron responses, including FRO2, AHA2, and NAS4. This might suggest that there is a high level of conservation of iron response regulatory mechanisms between model and crop species. With information regarding the function of any causal genes and of any genes in LD with causal variants, it becomes possible to identify potential pleiotropic effects that may result from the selection of specific alleles. This valuable information can then be used to make more informed decisions that align with specific breeding aims. Iron is an essential micronutrient and thus interactions with a wide range of biological processes that ultimately affect plant fitness and agronomic performance are expected. Given this, while studying plant iron responses alone is of interest, future efforts should also consider these responses within the context of organismal or environmental parameters.

Using both classical and quantitative genetics approaches, we have uncovered a whole suite of genes and variants that contribute to the regulation of plant responses to iron. There is, however still much to learn. For example, very little is known about how plants sense iron, allowing them to trigger the appropriate downstream response. The E3 ligase protein and known negative regulator of the iron deficiency response, BRUTUS (BTS), has recently been identified as a potential iron sensor in plants (Long et al., 2010; Hindt et al., 2017). BTS is able to directly bind iron (Selote et al., 2015), which offers a way through which this protein may be able to track shifts in cellular iron status. Given the complexity of iron homeostasis and plant responses to iron, it does however seem likely that there would be multiple iron sensors, maybe relaying information describing the iron status of different cellular compartments and tissue types. Furthermore, while it seems that BTS is involved in intracellular iron sensing, we are yet to identify possible extracellular sensing mechanisms in plants. To identify regulators upstream of the currently known iron-response pathways, including iron sensors, it would be advantageous to focus on any phenotypic effects that can be observed very early in response to shifts in iron availability. We know that vast shifts in gene expression can be observed after between 12 h and 72 h exposure to iron deficiency for example (Long et al., 2010). One option would be to make use of data describing natural variation in gene expression for eQTL analysis. Using these data as trait input for GWA analysis might allow for upstream regulators to be identified, which may bring us closer to discovering sensor mechanisms. We have seen several examples here where exploring natural variation in root growth responses has been valuable for the discovery of iron-response genes. Developing high-throughput microscopy methods that provide the resolution necessary to quantify the very earliest shifts in root growth may be another powerful way to capture natural variation related to the onset of plant iron responses.

It is clear that the genetic regulation of plant responses to iron is complex and involves the orchestrated effort of many genes. One factor that can influence the likelihood of detecting trait-controlling loci is the effect size of a given locus. Plant iron responses are highly quantitative and many of the genes contributing to the regulation, despite having important roles, may individually contribute very little to overall phenotype. When considered alone, these loci may go undetected. To alleviate this potential issue, it may be of value to instead focus on natural variation within gene sets, that is, groups of genes known to share some common factor related to biological function. Subramanian et al. (2005) describe a method called Gene Set Enrichment Analysis (GSEA). After grouping genes based on prior knowledge of function, they perform correlation analyses, exploring the relationship between gene expression and phenotype. The strongest and most highly significantly correlated genes are then queried to identify if there is enrichment of genes assigned to any specific gene sets. Identification of enrichment might suggest that the biological process or pathway represented by that gene set is somehow important in regulating variation in the phenotype of interest. In addition to making use of transcriptome data in this way, it would also be possible to take a similar approach using SNP information for GWA analysis. In the same way, genes would be grouped into gene sets. Markers would then, based on proximity, be assigned to genes, thus serving as a proxy for the linked genes and therefore the gene set to which they belong. One could then perform GWA analysis and rank the resulting marker P-values in order of significance. Querying this list would allow for any enrichment of markers assigned to genes belonging to any specific gene sets to be identified. Using such an approach, it may be possible to uncover modules of novel regulators of plant iron responses not detectable using classical GWA analysis methods.

The work celebrated here represents just the initial efforts in exploring natural variation to understand plant responses to iron availability. However, it is clear, even from this limited number of studies, that there is great promise in using this approach. Based on work carried out in Arabidopsis, we have seen that the greatest power comes from combining information across multiple omics levels. In recent years we have seen the development of methods such as associative transcriptomics, a novel GWA study approach which combines the use of natural variation at both the DNA sequence and gene expression level. This method has already proven powerful in dissecting the genetic regulation of complex traits in both wheat (Miller et al., 2016; Wang et al., 2017) and rapeseed (Harper et al., 2012; Miller et al., 2019). These continued efforts and collaborations will not only allow us to continue in our journey to understand gene function, but will also reveal unparalleled power to identify causal alleles that can be used to drive the development of elite varieties with increased tolerance to iron stress and enhanced nutritional quality.

Author contributions

This manuscript was written by CNM. Edits and revisions were made by CNM and WB.