High‐affinity iron uptake is required for optimal Epichloë festucae colonization of Lolium perenne and seed transmission

Abstract Epichloë festucae uses a siderophore‐mediated system to acquire iron, which is important to maintain endophyte–grass symbioses. Here we investigate the roles of the alternative iron acquisition system, reductive iron assimilation (RIA), via disruption of the fetC gene, which encodes a multicopper ferroxidase, either alone (i.e., ΔfetC) or in combination with disruption of the gene sidA, which encodes a siderophore biosynthesis enzyme (i.e., ΔfetC/ΔsidA). The phenotypic characteristics of these mutants were compared to ΔsidA and wild‐type (WT) strains during growth under axenic culture conditions (in culture) and in symbiosis with the host grass, perennial ryegrass (in planta). Under iron deficiency, the colony growth rate of ΔfetC was slightly slower than that of WT, while the growth of ΔsidA and ΔfetC/ΔsidA mutants was severely suppressed. Siderophore analyses indicated that ΔfetC mutants hyperaccumulate ferriepichloënin A (FEA) at low iron concentrations and ferricrocin and FEA at higher iron concentrations. When compared to WT, all mutant strains displayed hyperbranching hyphal structures and a reduced ratio of Epichloë DNA to total DNA in planta. Furthermore, host colonization and vertical transmission through infection of the host seed were significantly reduced in the ΔfetC/ΔsidA mutants, confirming that high‐affinity iron uptake is a critical process for Epichloë transmission. Thus, RIA and siderophore iron uptake are complementary systems required for the maintenance of iron metabolism, fungal growth, and symbiosis between E. festucae and perennial ryegrass.


| INTRODUC TI ON
Iron is a micronutrient that is required as a cofactor involved in a range of key biological activities, including DNA synthesis, respiration in almost all species, and photosynthesis in plants, due to its capacity to acquire and lose electrons (Guerinot & Yi, 1994;Kermeur et al., 2023;Zhang et al., 2019). Although iron is plentiful in nature, it has poor bioavailability because it is often present in the form of insoluble oxides. There are two major oxidation states for iron: ferrous (Fe 2+ ) and ferric (Fe 3+ ). Iron excess may harm tissues and cause diseases by acting as a redox catalyst in cell-damaging reactive oxygengenerating processes (Sutton & Winterbourn, 1989). To maintain | 1431 ZHANG et al.
iron homeostasis, fungi have evolved a variety of iron acquisition and storage systems as they lack iron excretion systems (Sørensen et al., 2014).
Iron acquisition methods for fungi fall into two broad categories: high-affinity and low-affinity, which serve as the primary iron assimilation systems in iron-limited and iron-abundant environments, respectively (Hause et al., 2002). When iron is scarce, most fungi rely on one or two high-affinity iron acquisition processes to gather sufficient iron from the plant, the animal, or the environment.
One approach is based on the secretion, by microorganisms, of small, high-affinity Fe 3+ -chelating compounds called siderophores.
These siderophores act as carriers of iron across cell membranes (Haas, 2003;Liu et al., 2021). Extracellular siderophores aid in iron absorption, while intracellular siderophores are involved in iron storage, sequestration, and iron transportation Johnson et al., 2013). Extracellular siderophores bind extracellular Fe 3+ , and ferrisiderophores are subsequently transported into the cell through specialized transport mechanisms and Fe 3+ is released; excess Fe 3+ may be retained by intracellular siderophores (Stintzi et al., 2000). The alternative route for iron uptake is via reductive iron assimilation (RIA), for example, in Saccharomyces cerevisiae, which uses a multicopper ferroxidase enzyme, Fet3p (orthologous to FetC in this study), which receives Fe 2+ from cell surface iron reductases such as Fre1p and Fre2p and then transfers Fe 3+ to the iron permease Ftr1p. FetC is a type I membrane protein with a single transmembrane domain at its C-terminus and contains three distinct copper sites previously identified in copper proteins (Solomon et al., 1996). RIA requires the complex of Fet3p and Ftr1p assembled in the endoplasmic reticulum to achieve proper cell surface targeting (Kosman, 2003).
Epichloë species (family Clavicipitaceae) are filamentous fungal endophytes that form persistent, mutualistic, and host-specific relationships with Pooideae cool season grasses (family Poaceae). Epichloë festucae dwells in the intercellular spaces (apoplast) of plant aerial tissues, where it acquires nutrients from the host plant (Christensen et al., 2002;Sattelmacher, 2001). In E. festucae, two hydroxamate siderophores have been characterized in culture: secreted epichloënin A (EA, ferrated form is ferriepichloënin A [FEA]) and intracellular ferricrocin (FC) Johnson et al., 2013). A key early biosynthesis enzyme, N 5 -hydroxyl-ornithine monooxygenase, encoded by the sidA gene, is required to produce both siderophores. Subsequently, two nonribosomal peptide synthetases, SidN (siderophore synthetase) and SidC (siderophore synthetase for FC), assemble EA and FC, respectively, from ornithine-derived intermediates. Hence, in the E. festucae mutant strains ΔsidN and ∆sidC, the production of EA and FC, respectively, is abolished, while in ΔsidA, siderophore production is abolished . Aberrant hyphal growth is a phenotype of these mutants that is observed under iron-deficient conditions . However, the cultures are still viable, presumably due to the presence of the com- ftrA (GenBank accession JN132405.1), in the genome of E. festucae strain Fl1; the expression of these genes was controlled by a shared bidirectional promoter. The current evidence suggests that both genes are required for a functioning RIA system (Albarouki & Deising, 2013;Condon et al., 2014;Kosman, 2003); hence, fetC was chosen for further study. This study aims to elucidate (1) the functions of the RIA system in E. festucae in culture and in planta and (2) the contribution of each iron uptake system to vertical endophyte transmission, a critical component for the persistence of this symbiosis.

| Inactivation of high-affinity iron uptake systems
This study characterized the role of RIA by disrupting the multicopper ferroxidase gene (∆fetC) and by abolishing both RIA and siderophore-mediated iron uptake (∆fetC/∆sidA). The mutants were characterized in culture and in planta and were compared to the previously characterized siderophore mutant ΔsidA . Table S1 provides plasmids and biological material generated in this study or from other sources.
The previously identified fetC and ftrA genes in E. festucae Fl1 form a gene cluster with an approximately 1.3-kb intergenic region containing gene promoters for both genes (Johnson et al., 2013).
The fetC locus is a 2.7-kb stretch of genomic DNA with a coding sequence of 1.9 kb ( Figure S1). BLAST analyses using the computed protein sequence of FetC from E. festucae Fl1 as a query against S. cerevisiae proteins or using the S. cerevisiae Fet3p protein (accession AJS70674.1) as a query against the proteins and genome of E. festucae Fl1 showed high sequence identity between them (49%). These results and the colocation with the ftrA gene suggest that fetC is the only gene encoding the multicopper ferroxidase involved in RIA in E. festucae Fl1.

| Colony growth in response to iron supply
For all iron concentrations, the colony features of ΔfetC and the gene-complemented strains, ΔfetC/fetC and ΔsidA/sidA, showed a similar morphology to that of the wild type (WT) over the study period (Figures 1 and 2). Analysis of colony growth indicated that ΔfetC colonies grew at a slightly slower rate than WT and genecomplemented strains over the 10-day measurement period.
Beginning from Day 6, ΔfetC growth was significantly reduced (88%-89% of WT) until Day 10 on defined medium with iron chelation and from Day 8 (86%-90% of WT) on defined medium with 0 and 25 μM FeCl 3 , as determined using a least significant difference (LSD) test (p < 0.05) (Figure 2a). Consistent with a study , ΔsidA growth was most delayed under iron starvation, with radial and aerial hyphal growth becoming gradually restored with increasing iron concentration (>400 μM) (Figures 1 and 2).
The growth of ΔfetC/ΔsidA colonies was retarded in all growth conditions, even on potato dextrose agar (PDA). Iron starvation severely limited the growth of ΔfetC/ΔsidA, which barely grew after 30 days (Figure 1). A small increase in colony diameter was recorded with increasing iron concentrations at Day 10 (17%-18%, 19%-20%, and 22%-26% of WT under defined medium with iron chelation, 0 μM FeCl 3 , or 25 μM FeCl 3 , respectively; Figure 2), with all measurements being significantly different from WT, ΔsidA, ΔfetC, and their complemented strains. From Day 2 to Day 10, all strains showed significantly higher growth rates than ΔfetC/ΔsidA when the FeCl 3 concentration was increased to 50 μM. However, there were no significant differences in the growth rates of WT, ΔfetC, ΔsidA, and the complemented strains under these conditions ( Figure S2). Despite ΔsidA having a growth rate similar to that of WT, it exhibited a deficiency in aerial hyphae ( Figure S3).

| Iron-dependent siderophore production in E. festucae
As expected, no siderophore production was detected in any sample derived from ΔsidA or ΔfetC/ΔsidA strains ( Figures S5 and S6).
Two-way analysis of variance (ANOVA) showed that there were significant interactive effects of Epichloë strain and growth medium on the production of intracellular FC (p < 0.05) and FEA (p < 0.05) siderophores. The greatest significant difference in the accumulation of FC between ΔfetC and WT as well as gene-complemented strains was observed when iron supply was increased to 100 μM  (Table 1).

| Hyphal growth characteristics in the host plant
Neither infection with E. festucae WT nor the disruption of fetC, sidA or fetC/sidA in E. festucae had a significant effect on perennial ryegrass growth parameters including tiller height (p = 0.107 by ANOVA), tiller number (p = 0.081 by ANOVA), and plant fresh weight at Week 8 (p = 0.263 by ANOVA) (Figure 3). However, sporadically occurring structures comprising hyperbranched hyphae were observed in leaf sheath tissues that were colonized by ΔfetC, ΔsidA, and ΔfetC/ΔsidA mutants under both light and confocal microscopy (Figure 4a,b). In comparison, the hyphae of the WT and gene-complemented strains grew parallel to the leaf axis and were rarely branched, which is consistent with normal growth (Christensen et al., 2002). The ratio of E. festucae DNA to total DNA was significantly reduced in pseudostem tissues infected with ΔfetC (1.3 pg/ng), ΔsidA (1.1 pg/ng), and ΔfetC/ ΔsidA (1.2 pg/ng) compared to WT (1.7 pg/ng) (p < 0.05 using Dunn's multiple comparison test) (Figure 4c).
To understand the effect of iron supply on the growth of  tissues (5.5 pg/ng total DNA) was lower than in WT-infected pseudostem tissues (8.9 pg/ng total DNA), but there was no statistical difference between the groups (p = 0.09) (Figure 5b).

F I G U R E 3
Boxplots of (a) tiller height, (b) tiller number, and (c) plant fresh weight of ryegrass plants 8 weeks after planting. Plants were mock-inoculated (E−) or infected with Epichloë festucae wild type (WT), ΔfetC, ΔfetC/fetC, ΔsidA, ΔsidA/sidA, and ΔfetC/ΔsidA. Each blue dot represents one biological replicate. Different letters on the bars indicate significant differences between strains as determined using the LSD test (p < 0.05).

F I G U R E 4
Hyphal growth in the leaf sheath and endophyte biomass in the pseudostem tissues. Leaf sheaths were taken from ryegrass plants infected with Epichloë festucae wild type (WT), ∆fetC, ∆fetC/fetC, ∆sidA, ∆sidA/sidA, and ∆fetC/∆sidA and analysed by (a) light microscopy and (b) confocal laser scanning microscopy. Confocal microscopy images of aniline blue-stained hyphae (purple pseudocolour) and septa labelled with WGA-488 (green pseudocolour). Bars = 20 μm. (c) DNA concentrations (pg endophyte DNA per ng total [plant + endophyte] DNA) in the basal pseudostem tissues of ryegrass plants. Different letters beside the bars indicate significant differences between treatments as determined using the LSD test (p < 0.05).
However, the disruption of genes involved in either siderophore or RIA-mediated iron uptake had no effect on the spike length  seeds harvested from mock-inoculated ryegrass plants continued to be free of Epichloë infection (0%), as confirmed using the tissue-print immunoblotting and seed squash methods, respectively (data not shown). For WT, the seed transmission rate (25%) was much lower than the vegetative infection rate (99%), but it was significantly higher than that of ΔfetC/ΔsidA-infected seeds (AB41: 3%; AB44: 8%) (p < 0.05) as determined using Dunn's multiple comparison test (Figure 6a,b). The results indicate that maintenance of endophyte iron metabolism is important for the reproductive tiller percentage and vertical endophyte transmission.

| DISCUSS ION
Iron is a micronutrient that is required for the mutualistic symbiosis between perennial ryegrass and E. festucae. Previous research has established that the E. festucae siderophore system, which consists of the cellular siderophore FC and the dual localized siderophore EA, is essential for E. festucae survival in culture and also for the mutualistic interaction Johnson et al., 2013).  (Hassett et al., 2000), whereas the rice false smut pathogen Ustilaginoidea virens uses Uvt3277 as the low-affinity iron transporter (Zheng et al., 2017), and an as yet to be characterized low-affinity iron uptake system has also been proposed for A. fumigatus, which are also viable in the absence of high-affinity uptake systems (Schrettl et al., 2004). Therefore, we conclude that siderophore-mediated iron uptake is the dominant strategy in E. festucae, at least in culture, and that low-affinity uptake systems are likely to exist.
E. festucae produces two siderophores, FC and EA (Johnson et al., 2013). Our previous research showed that FC is the primary intracellular siderophore under iron-sufficient conditions, while EA is the dominant intracellular and secreted (extracellular) siderophore when iron access is limited Johnson et al., 2013).
Concomitant with those studies, our results (Table 1) suggest that siderophore-assisted iron assimilation is greater in ΔfetC mutants than in WT. Under low-iron conditions, the significantly higher EA concentrations suggest that ΔfetC is more iron-starved than WT, but the higher concentrations of FC suggest less restricted access to iron by FC. Conversely, under iron-sufficient conditions, the sustained high concentrations of intracellular EA and accumulation of FC suggest an increased requirement for iron sequestration by FC, as the iron internalized as FEA can be transferred to FC, thus allowing EA to be recycled. Therefore, it could be that the mechanisms that coordinate siderophore production and transport are partly deregulated by the lack of RIA activity.
In the absence of extracellular siderophores, EA enhances the formation of H 2 O 2 in the iron-depleted ΔsidA and ΔsidN mutants Johnson et al., 2013). Further, this study has found little effect of deletion of fetC alone, but the ΔfetC/ΔsidA mutants appeared to form less H 2 O 2 precipitates compared to ΔsidA, which implies that the RIA system adds to oxidative stress under low-iron conditions (Figure 2b). Release of free iron by RIA into the cell without interception by siderophores may be the reason, with the ΔfetC mutants ameliorating some iron stress through the absence of the RIA system in combinatorial mutants and also via enhanced production of siderophores, particularly EA in the single ΔfetC mutants (Table 1). This further supports the protective role of EA against reactive oxygen stress in E. festucae. Similar to our results, in studies of C. heterostrophus, the combinatorial deletion of the RIA component FTR1 and the extracellular siderophore nonribosomal peptide synthetase enzyme NPS6 (but not the intracellular siderophore biosynthetic enzyme NPS2) led to hypersensitivity to oxidative stress, but at the same levels as the nps6 mutant alone (Condon et al., 2014). To minimize free excess iron stimulation of hydroxyl radical formation (Fenton reaction) (Halliwell et al., 1992;Schützendübel & Polle, 2002), the balanced interplay between the siderophore and RIA systems in E. festucae is probably important to maintain a metabolic balance between iron uptake and iron sequestration.
All mutant strains lacking one or two high-affinity iron uptake systems displayed aberrant hyphal morphologies in leaf sheath tissues of infected host plants and reduced biomass (Figures 4 and 5).
The reduced Epichloë biomass of all mutant-infected ryegrass plants was associated with swollen, branched hyphae and epiphyllous hyphal structures in the pseudostem tissues. It appeared from the microscopic images that the biomass of E. festucae mutant strains may be higher than that of the WT, which conflicted with the DNA quantification results. This phenomenon could suggest a patchy growth pattern of the mutant strains that lack one type of high-affinity iron uptake system. These strains could colonize more in iron-enriched regions (e.g., xylem vessels) (Ariga et al., 2014) as areas with low iron concentrations probably do not support mutant hyphal growth.
We also found that for both WT and ΔfetC, endophyte biomass was higher under low-iron than under high-iron conditions, which might be due to a dilution effect, presuming an increase of grass growth and no change in fungal growth under high-iron conditions. Similar results were found in ryegrass plants under low nitrogen and phosphate supply (Liu et al., 2011;Rasmussen et al., 2007).
Hyphal biomass has been shown to be correlated with seed transmission (Albarouki et al., 2014;Gagic et al., 2018), and we hypothesized that, because ΔfetC, ΔsidA, and ΔfetC/ΔsidA had reduced biomass, they would not colonize new tillers or transmit to the seed as well as the WT strain. However, this was true only for ΔfetC/ ΔsidA, in which vegetative and floral transmission was significantly reduced relative to WT (Figure 6), suggesting that possessing functional RIA (in ΔsidA) or siderophore (in ΔfetC) systems can support Epichloë transmission. The failure to form hyphal fusions in ΔfetC/ ΔsidA ( Figure S4) might be indicative of a restrictive growth pattern of E. festucae in symbiosis with ryegrass plants, suggesting that host colonization could be defective (Becker et al., 2014). This further demonstrates the critical role of iron for balancing host-endophyte compatibility and that both high-affinity iron uptake systems are required for normal in planta growth. Although we saw a decrease in  (Haas, 2003;Heymann et al., 2002;Philpott & Protchenko, 2008).
In contrast to some pathogens that appear to prioritize the use of one type of iron uptake mechanism over another (Burbank et al., 2015;Eichhorn et al., 2006), this study confirms that E. festucae uses both high-affinity iron uptake systems concurrently in iron metabolism, with some level of redundancy to maintain iron homeostasis, similar to C. graminicola and C. heterostrophus (Albarouki et al., 2014;Albarouki & Deising, 2013;Condon et al., 2014). This redundancy confers E. festucae more flexibility, with increased capacity to obtain and use iron from its environment. Similar findings have been reported for C. heterostrophus, a maize pathogen, where the role of RIA in pathogen virulence is overshadowed by the involvement of extracellular siderophores as a high-affinity iron uptake mechanism, with RIA serving as a critical backup for the fungus (Condon et al., 2014). We found in this study that disruption of either one or two high-affinity iron uptake systems reduced the percentage of reproductive tillers in flowering plants. This may reflect reduced floral induction and/or termination of inflorescence development through somehow impaired interactions with these mutants. Moreover, the complemented strains for ∆fetC and ∆sidA did not fully recover the phenotype as observed in the WT-infected plants ( Figure S7). The likely reason for this was because the complementation method was performed through ectopic integration of the gene into a non-native genomic location, which may cause the overall function of the complemented gene to differ from its native context, particularly for sidA and fetC, which both form part of a divergently coexpressed gene pair; the physical decoupling of the genes may prevent full complementation. Therefore, the differences in reproductive tiller percentage between plants infected with WT or treated with mock inoculation and the complemented strains are reasonable. Additionally, the design of p∆fetC involved the deletion of an approximately 500-bp intergenic region including the fetC promoter sequences ( Figure S1), which may have also impacted the gene expression of ftrA in ∆fetC and thereafter ∆fetC/fetC under iron-depleted conditions. The gene expression of ftrA ( Figure S8) appeared reduced, but the increased FC concentration in these strains under iron deficiency (Table 1) might also suggest that higher iron availability could lead to reduced expression of the iron deprivation-responsive ftrA gene (Schrettl et al., 2008). The disruption of either RIA or siderophore systems not only perturbs endophyte hyphal phenotype and biomass levels in the host interaction, but also impedes Epichloë transmission between tillers and to the subsequent seed generation, which impacts fitness. Seed trans- In this study, we have examined the role of the high-affinity iron uptake system RIA as a complementary iron uptake system in E. festucae Fl1. Although some redundancy with the siderophore iron uptake system is apparent, the RIA system can also contribute independently to endophyte iron uptake under both in culture and in planta growth conditions. However, the loss of both high-affinity uptake systems critically affects the Epichloë ryegrass interaction phenotype and the transmission of Epichloë during the plant life cycle.

| Plant and Epichloë growth conditions and endophyte inoculations
E. festucae Fl1 (wild-type [WT] strain) was originally isolated from perennial ryegrass cultivar SR3000 (Leuchtmann et al., 1994(Leuchtmann et al., , 2014. All E. festucae strains were maintained on 3.9% (wt/vol) PDA (Oxoid Ltd) at 22°C in an 8 h light/16 h dark cycle. Defined medium (Johnson et al., 2013) was used as a nutritional base for fungal growth assays and chemical analyses. Endophyte-free seeds for Epichloë inoculation were sourced from the Margot Forde Germplasm Centre (Palmerston North, New Zealand). Fungal saprophytes and other seedborne endophytes were removed by heat treatment using the method of Bouton et al. (1993) with extended incubation (47°C for 3 weeks at 45% relative humidity). Seed packets were placed in a desiccator jar on a gauze mat above 100 mL of a 75% glycerol and 25% water mixture (vol/vol) to maintain relative humidity. Epichloë inoculation into perennial ryegrass seedlings was previously described (Becker et al., 2018;Latch & Christensen, 1985). Tissue-print immunoblotting was used to confirm Epichloë infection (Simpson et al., 2012). Epichloë-infected perennial ryegrass plants were grown in 1-L pots filled with potting medium, kept in a glasshouse, and wa-

| Construction, validation, and selection of deletion mutants
Details can be found in File S1.

| In culture growth assay
All strains were evaluated for their ability to grow under a range of iron concentrations and culture conditions, in addition to complete medium (i.e., PDA). Defined medium (<1 μM iron with 2% agar, +0 Fe) was supplemented with BPS (Sigma-Aldrich) to generate iron-

| Measurements of siderophore concentrations
Starter cultures were grown in 100-mL flasks containing 50 mL liquid defined medium for 10 days at 22°C, with shaking at 100 rpm. A 2-mL aliquot of each culture was transferred to a 2-mL microcentrifuge tube and centrifuged at 3000 g for 10 min. The supernatant was

| Microscopy of hyphae in culture and in planta
Using cultures grown on PDA for 7 days, we examined the morphology of hyphal tips and hyphal fusion. Hyphae were stained following the method described by Shoji et al. (2015). Images were taken with an FV1000-D confocal laser scanning microscope (Olympus) and FV10i-ASW 3.1 Viewer software (Olympus).

| Quantification of Epichloë biomass in perennial ryegrass
The Epichloë biomass in pseudostem tissues was quantified by using a quantitative PCR (qPCR) method (Cook et al., 2009 A list of primers used in this study can be found in Table S2.

| Gene expression quantification
Details can be found in File S1.

| Flowering spike morphology and Epichloë seed transmission
At the seed ripening stage, the flowering spike length (cm), total and fertile spikelet numbers, seed weight per spike (g), and grain number per spikelet were determined. Spike length was determined by measuring the distance between the base of the most basal floret and the tip of the most apical floret. Fertile spikelets produce at least one grain, whereas infertile spikelets do not set any grain (Guo et al., 2017). Fertility is defined as the ratio of the fertile to total spikelet counts. For trait measurements, at least four plants infected with each transformant were randomly selected.

| Measurement of vegetative and vertical transmission of Epichloë
To determine the efficiency of Epichloë transmission in vegetative tillers, 1-year-old plants infected with each strain were analysed for endophyte presence in the tillers by tissue-print immunoassay.
Approximately 40 randomly selected tillers per plant and at least eight plants infected with each construct were tested.
Seeds harvested from mock-inoculated endophyte-free ryegrass plants were analysed for the presence of Epichloë using a common seed squash method (Clark et al., 1983). To determine the rate of viable Epichloë transmission from maternal plants to seedlings, approximately 108 seeds from each maternal plant were sown and grown for 5 weeks until one to three tillers were present. Seedlings were then tissue-print immunoblotted to assess infection status. At least five maternal parent plants infected with each construct were used, and the percentage of progeny seeds with viable endophyte was determined.

| Growth assays of hydroponically grown plants
A hydroponic system was used to control nutrient availability and evaluate the impact of iron deficiency on plants infected with ∆fetC more precisely. Iron-free hydroponic solution was prepared . The iron source was a chelated EDTA ferric sodium salt solution, which was used to create two distinct treatments (high iron: 50 μM Fe 3+ , low iron: 500 nM Fe 3+ ) ( Figure S9).
Root-trimmed mature tillers infected with WT, ∆fetC, and ∆fetC/fetC were transferred from soil to hydroponic conditions. Two independent 24-site hydroponic propagation systems were set up with separate containers containing approximately 3.5 L hydroponic solution with two iron concentrations (Bothe et al., 2018). Three tillers of a single genotype were inserted into a neoprene plug at each site. For each inoculated treatment, four biological replicates and two technical replicates were included, which were randomly arranged in the propagators. To regenerate new roots, the tillers were grown for 10 days in hydroponic solution containing 20 μM iron, followed by 4 weeks in high-or low-iron conditions. Hydroponic solutions were refreshed every 5 days.
Two-way or one-way ANOVA was conducted using the Anova function in the "car" package (Fox et al., 2007). An LSD method was adopted for further post hoc analysis with the "agricolae" package (de Mendiburu & de Mendiburu, 2019). Pairwise treatment comparisons were carried out with Fisher's protected LSD test (α = 0.05).
Data that did not satisfy the assumptions of normality and homogeneity of variance were analysed by one-way ANOVA with the nonparametric Kruskal-Wallis method using the "stats" package and Dunn's multiple comparison test with the "dunn.test" package.
The data were plotted using either R functions or with the "ggplot2" package (Wickham, 2011). Librarians.

DATA AVA I L A B I L I T Y S TAT E M E N T
Raw data that support the findings of this study are available in Table S3.