Maize seeds were sterilized for 30 min in a 10% solution of H₂O₂, washed with distilled water, soaked in saturated CaSO₄ for 10 h, and then germinated on moist filter paper in the dark at room temperature. Two days later, the germinated seeds were wrapped in moist filter paper roll and grown. At the stage of two visible leaves, the seedlings were selected and transferred into a 40-L black container. Solution pH was set at 5.5–6.0. The adjusted Hoagland nutrient solution contained the following parameters (mmol L⁻¹): 0.5 NH₄NO₃, 0.5 CaCl₂, 1.5 Ca(NO₃)₂, 0.75 K₂SO₄, 0.65 MgSO₄, 0.1 KCl, 0.25 KH₂PO₄, 1.0 × 10⁻³ H₃BO₃, 0.01 Zn-EDTA, 8.0 × 10⁻³ CuSO₄, 1.2 × 10⁻² MnSO₄, 4.0 × 10⁻⁵ (NH₄)Mo₇O₂₄, and 4.0 × 10⁻³ NiCl. All the experiments were conducted in randomized blocks with three replications. All the experiments were conducted in the growth chamber under strictly controlled environments. The condition of the growth chamber was set as a 14-h light period from 8:00 to 22:00 with 28 °C and a 10-h dark period with 22 °C. The average light intensity measured at the canopy was 350 μmol m⁻² s⁻¹.

Experiments were terminated 14 days after transplanting. During the period from 9:00 to 11:00 A.M., SPAD values of the youngest fully expanded leaf were measured on the 1/3 part from the leaf base for three times using the SPAD-502 leaf chlorophyll meter. In addition, the mean of three observed values was recorded for each plant. The plant height and root length for each plant were measured. All samples were heat-treated at 105°C for 30 min and dried at 75°C till constant weight. Concentrations of Fe, copper (Cu), and zinc (Zn) in shoots and roots were analyzed by ICP-AES (inductively coupled plasma-atomic emission spectroscopy). Relative ratios of root to shoot (R/S) efficiency and nutrient contents were estimated using the following Equations (1) and (2), respectively. (1) Relative ratios of R/S = R/S (-Fe) / R/S (CK)(2) Nutrient content = Concentration × Dry weight For GWAS, Fe score for each plant has been visually recorded for three times since the 12th day after transplanting. Five scales (0–4) were designed to assess the ability to tolerate Fe deficiency of each line in the GWAS panel under Fe-deficient condition [-Fe: 0.6 μmol L⁻¹ Fe(II)-2,2′-bipyridyl] (Supplementary Figure S1). Under Fe-deficient condition, Fe-deficiency tolerance scoring mainly depends on the degree of inhibition in growth and development for the whole plant, as well as the degree of chlorosis of young leaves. Score 0: plants developed about three leaves with one sprout and showed the severest stunted growth and reduced plant height, distorted leaf growth, and yellowish-white necrotic lesions distributed on the whole youngest leaves. Score 1: plants developed four leaves and one sprout and showed reduced plant height and wrinkled leaf margins. Fe-deficient chlorosis with necrotic patches was distributed on the youngest leaves. Score 2: plants developed four leaves and one sprout and exhibited a better growth compared with Score-0 and Score-1 plants, but still showed chlorosis on the youngest leaves. Score 3: plants developed four leaves and one sprout and obtained an improved plant height compared with Score-2 plants. Young leaves showed Fe-deficient chlorosis without wrinkled leaf margins. Score 4: plants grew well without reductions in plant height but still display banded chlorosis in the middle of young leaves.

In Experiment 4, total RNA was extracted from shoots and roots of plants using the Total RNA Extraction Kit (TIANGEN, China). The cDNA was synthesized in accordance with Fast Quant RT Super Mix Reverse Transcription Kit instructions (Transgene, Beijing, China). Quantitative real-time PCR was performed using SYBR Green Real-time RT-PCR (Takara) and an ABI7500 Fast Real-Time PCR System (Applied Biosystems). The relative gene expression level was calculated using the 2^−ΔΔCt method. Each real-time PCR experiment contained three technical replicates.

The means for each trait were compared using one-way ANOVA at a 0.05 level of probability followed by the least significant difference (LSD) test using SPSS 20.0. The linear mixed effect function lmer in the lme4 package of R version 3.1.1 was fitted to each RIL to obtain the BLUP (best linear unbiased prediction) value for each trait.

A genome-wide association analysis on the traits associated with Fe-deficiency tolerance was performed using a mixed linear model (MLM) (Li et al., 2013) which considered the population structure and relative kinship matrix (Yu et al., 2006) using TASSEL 3.0. Using “no compression” and “population parameters previously determined” (P3D) algorithms, the MLM program in TASSEL 3.0 was conducted to detect the association between the phenotype and genotype. The threshold of P = 1 × 10⁻⁴ was used to determine a significant SNP, which was determined based on the quantile–quantile plots and distribution of P-values for all the traits (Supplementary Figure S2).

According to the previous study, the LD distance of this association panel was 50 kb (Li et al., 2013). Therefore, candidate genes within a 100-kb region (50 kb upstream and downstream of the lead SNP) flanking each leading SNP were identified. Their functional descriptions were identified using the maize B73 reference genome assembly version 2 available on the MaizeGDB Database (http://www.maizeGDB.org/) and Gramene Database (https://www.gramene.org/).

In Experiment 1, Fe-efficient inbred line Ye478 and Fe-inefficient inbred line Wu312 were hydroponically grown during the same period. Compared with Wu312, Ye478 exhibited significantly higher leaf SPAD, shoot and root dry weight (Figure 2). On the Fe-deficient supply of 0.6 μM Fe(II)-2,2′-bipyridyl, the leaf SPAD, shoot, and root dry weight, and shoot Fe content of Ye478 were 1.4-fold, 2.2-fold, 2.5-fold, and 1.2-fold higher than those of Wu312, respectively (Figures 2A–C,F). Especially, the largest difference in shoot dry weight between two inbred lines was found at the ferrous concentration of 0.6 μM among ten different treatments (Figure 2B).

We hypothesized that compared to Wu312, the roots of Ye478 had the genetic advantage when seedlings were transplanted. When the two inbred lines were hydroponic cultured together, Ye478 displayed an absolute advantage in competition under the condition of a limited total amount of Fe supply. To test the above hypothesis, we set up this experiment at different developmental stages.

In Experiment 2, compared with Wu312, Fe-efficient inbred line Ye478 was transplanted and grown in hydroponics 7 days later. Although Wu312 has been grown for 14 days after transplanting, the 7-day Ye478 still displayed advantages in plant growth under Fe deficiency (Figure 3). On the supply of 0.6 μM ferrous concentration, the leaf SPAD, shoot, and root dry weight, and shoot Fe content of 7-day Ye478 were 2.8-fold, 1.8-fold, 2.4-fold, and 2.2-fold higher than those of 14-day Wu312, respectively (Figures 3A–C,F).

Compared to the phenotypes at the same developmental stage in Experiment 1, plant growth of Wu312 should be significantly improved when grown 7 days earlier in Experiment 2. However, the results of this experiment refuted our original hypothesis. Because even when the roots of Wu312 had a significant advantage over Ye478, Wu312 eventually developed to the fourth leaf and then stopped growing in Experiment 2, and the phenotypes of Wu312 in Experiment 2 were similar to the phenotypes at the same development stage in Experiment 1.

The difference in the ability to tolerate Fe deficiency between the two different genotypes, which have been observed in Experiments 1 and 2, cannot be explained clearly by physiological traits. Therefore, the GWAS panel which has extensive genetic variation was utilized to perform genome-wide association analysis and further identify candidate genes responding to low Fe stress, preliminarily revealing the genetic basis conferring Fe-deficiency tolerance in maize.

The phenotypic means of each trait in the GWAS panel are shown in Supplementary Table S1. On average, Fe deficiency decreased the means of plant height, root length, shoot dry weight, and root dry weight by 38.6, 30.6, 51.7, and 46.2%, respectively, and the mean of R/S ratio was increased by 75.9% under the Fe-deficient condition. To identify the loci associated with Fe efficiency, we used all traits in the -Fe treatment, shoot and root dry weights, and R/S ratio in the -Fe/CK treatment to perform GWAS. All the traits of the GWAS panel varied widely in the -Fe or the -Fe/CK treatments (Figure 4). The coefficients of variance for each trait under Fe deficiency ranged from 22.8 to 71.6% (Supplementary Table S1). Furthermore, significant correlations were observed among plant height, root length, shoot, and root dry weights under Fe deficiency (Supplementary Table S2).

We performed a genome-wide association study by fitting a mixed linear model (MLM) with population structure and familial relatedness. The quantile–quantile plots (Q-Q plots) for GWAS based on BLUP value of each trait implied that the associations were well controlled for population structure (Supplementary Figure S2). At a significance level of P < 1 × 10⁻⁴, a total of 353 significant SNPs were mapped on 10 chromosomes (Figure 5, Supplementary Table S3): 16 SNPs for Fe score, 21 SNPs for plant height, 22 SNPs for root length, 63 SNPs for shoot dry weight, 85 SNPs for root dry weight, and 146 SNPs for R/S ratio. For all traits, the phenotypic variance explained by each SNP varied from 5.1 to 13.3%.

Based on the physical position (B73_v2 reference), 472 candidate genes were identified within 100-kb flanking 353 significant SNPs which were associated with Fe score, plant height, root length, shoot and root dry weight, and R/S ratio (Supplementary Table S3). According to the functional descriptions of total 472 candidate genes in Arabidopsis and rice on MaizeGDB Database (http://www.maizeGDB.org) and Gramene Database (https://www.gramene.org), seven genes were proposed to be associated with Fe-deficiency tolerance in maize, including the genes participating in Fe homeostasis of strategy I plants and strategy II plants (Table 1).

Among them, GRMZM2G060952, also known as ZmDMAS1 which encodes deoxymugineic acid synthase, was located in the 1.1 kb downstream of the SNP chr1.S_31743584 which was significantly associated with root dry weight. GRMZM2G096958, also known as ZmNAAT1 which encodes nicotianamine aminotransferase, was containing the SNP chr4.S_219582768 that is significantly correlated with R/S ratio under Fe deficiency. GRMZM5G812538 also known as ZmYSL11 encoding a yellow stripe-like transporter, was identified within the overlapped regions co-localized by chr10.S_131843508 and qFe(II)-SPAD10-2 which is a reported QTL controlling leaf SPAD detected in the Ye478 × Wu312 RIL population under Fe deficiency (Xu et al., 2022).

Apart from strategy II genes, two candidate genes probably associated with Fe homeostasis in strategy I plants were identified by two significant SNPs. GRMZM2G118821, encoding an iron transporter IRT in maize, was detected in the 89.4 kb upstream of the significant SNP chr1.S_258444638 which was associated with root length under Fe deficiency. GRMZM2G111300, which encodes another member of ZIP (ZRT/IRT-like protein) family, was localized in the 25.9 kb upstream of chr4.S_26684019 on chromosome 4. GRMZM2G171616, encoding an EIN3-binding F-Box protein that may be involved in the regulation of ethylene signaling and Fe homeostasis in strategy I plants, was identified within the co-localization of chr9.S_90366645 and qFe(II)-SPAD9-1 which is another locus mapped on chromosome 9 in the Ye478 × Wu312 RIL population reported by Xu et al. (2022).

To evaluate the effect of Fe deficiency on the expression of candidates participating in Fe homeostasis of strategy II and strategy I plants, the relative expression levels of these genes were analyzed in the shoots and roots of Ye478 under Fe-deficient [-Fe: 0.6 μmol L⁻¹ Fe(II)-2,2′-bipyridyl] and Fe-sufficient [CK: 350 μmol L⁻¹ Fe(II)-EDTA] conditions.

Overall, strategy II genes ZmNAAT1 (GRMZM2G096958), ZmDMAS1 (GRMZM2G060952), and ZmYSL11 (GRMZM5G812538) were all significantly upregulated in both shoots and roots under low Fe stress (Figure 6). In the shoots, 9.4-fold and 13.0-fold upregulation were observed in the expression of ZmNAAT1 and ZmDMAS1, respectively (Figures 6A,C). By contrast, the relative expression levels of ZmNAAT1 and ZmDMAS1 in the roots were 5.1-fold and 6.4-fold upregulated under Fe deficiency, respectively. These results indicated that the expression of ZmNAAT1 and ZmDMAS1 responded to Fe deficiency more in shoots than in roots. For ZmYSL11, 2.0-fold and 2.4-fold upregulations were found in the shoots and roots under Fe-deficient conditions, respectively (Figures 6E,F).

Interestingly, three strategy I genes displayed Fe-deficiency-inducible upregulation in maize which is considered as the strategy II plant (Figure 7). ZmIRT1 (GRMZM2G118821) was found to be significantly upregulated in the roots under Fe deficiency (Figure 7A). Furthermore, the expression of ZmZIP4 (GRMZM2G111300) displayed a 1.5-fold upregulation in the shoots in response to low Fe stress (Figure 7B). ZmEBF1 (GRMZM2G171616), which encodes EIN3-binding F-Box protein 1 and plays a key role in the ethylene signaling pathway and Fe homeostasis of strategy I plants, was found to be significantly upregulated in both shoots and roots under Fe deficiency (Figures 7C,D).

To date, numerous studies concentrate on the concentrations of Fe and other minerals in different tissues of maize (Qin et al., 2012; Šimić et al., 2012; Zdunić et al., 2014; Gu et al., 2015; Zhang et al., 2017; Hindu et al., 2018; Ma et al., 2021). However, there were some reports of mapping the QTLs associated with shoot and root biomass accumulation, yield, plant height, and root traits (Azevedo et al., 2015; Burton et al., 2015; Li P. et al., 2016; Wang et al., 2016; Luo et al., 2017). Therefore, the loci identified in these previous studies were compared with the current study based on the physical position of each significant SNP.

In total, 113 of 353 significant SNPs identified by GWAS have been found to be co-localized with the reported QTLs or SNPs in previous studies (Supplementary Table S4). Among these loci, five significant SNPs which were associated with root dry weight and R/S ratio were co-localized with two significant SNPs associated with leaf necrosis under Fe-deficient condition detected by GWAS (Benke et al., 2015). In addition, eight QTLs related to root length, shoot and root dry weight, and R/S ratio detected in the current study were overlapped by previously reported QTLs or SNPs associated with iron (Fe), zinc (Zn), and magnesium (Mn) concentration in grains detected by linkage analysis or GWAS (Gu et al., 2015; Hindu et al., 2018), suggesting that these co-localized genomic regions may have pleiotropic effects on Fe and other mineral concentration of grains in maize.

In addition, a total of 53 significant SNPs associated with Fe score, plant height, root length, shoot and root dry weight, and R/S ratio detected in the present study was co-localized with the loci controlling plant height under salt stress or normal conditions reported by Luo et al. (2017). Besides, 64 significant SNPs controlling six different traits in this work overlapped with ten loci controlling root morphology under low-phosphorus (P) stress detected by Azevedo et al. (2015) and nitrogen (N) stress identified by Li P. et al. (2016). These findings indicate that these genomic regions may harbor several genes with pleiotropic effect on plant height or root morphology under abiotic stress at the maize seedling stage.

Moreover, the Fe-deficiency-inducible candidate gene ZmNAAT1, containing the R/S-associated SNP, was co-localized with the QTL controlling ear height, leaf angle, and yield in maize, and the candidate gene ZmDMAS1, which was markedly upregulated by Fe deficiency in shoots and roots (Figure 7B), was identified within the co-localization of the significant chr1.S_31743584 associated with root dry weight identified in this work and the QTL controlling root diameter under low-phosphorus stress reported by Azevedo et al. (2015). These results further confirm that these genomic regions may be highly valuable for identifying the genes associated with Fe-deficiency tolerance.

Strategy II plants acquire Fe through mugineic acid family phytosiderophores (MAs) (Bashir and Nishizawa, 2006). In the MA biosynthetic pathway, nicotianamine aminotransferase (NAAT) and deoxymugineic acid synthase (DMAS) enzymes catalyze the formation of 2'-deoxymugineic acid (DMA) from nicotianamine (NA) (Takahashi et al., 2003), and then, DMAs are secreted into the rhizosphere (Li et al., 2019). Subsequently, Fe(III)-DMA and Fe(II)-NA complexes are taken up from the rhizosphere by transporters, such as Yellow Strip 1 and Yellow Stripe1-Like proteins (Curie et al., 2001; Inoue et al., 2009). In this study, strategy II genes were identified, including ZmNAAT1, ZmDMAS1, and ZmYSL11.

OsNAAT1, a homolog of ZmNAAT1, was upregulated in all cells of Fe-deficient roots and shoots, and its transcripts were more abundant in yellow young leaves than in green old leaves under Fe deficiency (Cheng et al., 2007; Inoue et al., 2009). Herein, we further confirm that ZmNAAT1 expression was not only induced by Fe-deficiency stress in roots, but also in shoots (Figures 6A,B). DMAs have been detected in xylem and phloem, and the amount of DMA is reported to increase in Fe-deficient shoots and roots (Nishiyama et al., 2012; Ariga et al., 2014; Nozoye et al., 2014). ZmDMAS1, OsDMAS1, HvDMAS1, and TaDMAS1 are highly homologous to each other and are all strongly upregulated under Fe-deficient condition (Bashir and Nishizawa, 2006; Benke et al., 2014; Beasley et al., 2017). Furthermore, OsDMAS1 expression is restricted to the cells participating in long-distance transport under Fe-sufficient condition and is detected in vascular bundles specifically under Fe-deficient condition (Bashir et al., 2006). More importantly, our results further confirm that ZmDMAS1 responded to Fe deficiency more in shoots than in roots (Figures 6C,D), which was inconsistent with the results of gene expression analysis performed by Nozoye et al. (2013).

YSL genes, such as AtYSL1 and OsYSL2, are reported to transport metals via NA complexes, including Zn(II)-NA, Cu(II)-NA and Fe(II)-NA, Mn(II)-NA (DiDonato et al., 2004; Koike et al., 2004; Le Jean et al., 2005; Ishimaru et al., 2010). OsYSL15 is found to encode an Fe-specific transporter which is not only responsible for Fe uptake from the rhizosphere but also for phloem transport of Fe by Fe(III)-DMA (Inoue et al., 2009; Lee et al., 2009). In this study, ZmYSL11 was furthermore confirmed to be significantly upregulated by Fe deficiency in both shoots and roots (Figures 6E,F).

In this study, genes involved in strategy I, including ZmIRT1, ZmZIP4, and ZmEBF1, were identified within the genomic regions flanking the significant SNPs detected by GWAS (Table 1). There is basically no research on these genes in maize. However, the function summary of the homologous genes of these genes in other plants is helpful to understand their functions in maize. Beyond that, we also analyzed their expression pattern in response to Fe deficiency in maize.

IRT1 is the bifunctional transporter-receptor at the heart of metal sensing and signaling, and IRT1 transcription in response to low iron is dually controlled by local and long-distance signals (Cointry and Vert, 2019). In this research, we further confirm that ZmIRT1 was Fe-deficiency-inducible in the roots of maize inbred line (Figure 7A).

The exact role of most ZIP members in planta has not been understood, including ZmZIP4. OsZIP4 and, its close homolog, OsZIP3, are both highly expressed in the nodes. OsZIP3 is mainly involved in unloading Zn from the xylem of enlarged vascular bundles (Sasaki et al., 2015). OsZIP4 is involved in transporting Zn to the phloem of diffuse vascular bundles in the nodes for subsequent distribution to the developing tissues, including tiller buds, new leaf, and panicles (Mu et al., 2021). It also functions in transporting Zn to meristem cells in the tiller buds. Our expression results showed that ZmZIP4 was significantly upregulated in the Fe-deficient roots of maize (Figure 7B).

So far, the role of hormones and signaling substances on the regulation of Fe-deficiency responses in strategy II species has been rarely studied (Lucena et al., 2015). Under Fe deficiency, ethylene is implicated in the activation of some Fe-related genes (Kobayashi et al., 2014). The responses of strategy I plants to ethylene are mediated by regulation of EBF1/2-dependent degradation of the ETHYLENE INSENSITIVE3 (EIN3) transcription factor, which is critical for growth in plants and has been analyzed by many studies (Gagne et al., 2004; An et al., 2010). Moreover, in the current study, we found that ZmEBF1 was not only significantly upregulated in the shoots, but also in the roots (Figures 7C,D). These findings suggest that strategy I genes, namely ZmIRT1, ZmZIP4, and ZmEBF1, may be involved in the mechanisms underlying Fe-deficiency tolerance in maize.