A genome‐wide association study identifies a transporter for zinc uploading to maize kernels

Abstract

The Zn content in cereal seeds is an important trait for crop production as well as for human health. However, little is known about how Zn is loaded to plant seeds. Here, through a genome‐wide association study (GWAS), we identify the Zn‐NA (nicotianamine) transporter gene ZmYSL2 that is responsible for loading Zn to maize kernels. High promoter sequence variation in ZmYSL2 most likely drives the natural variation in Zn concentrations in maize kernels. ZmYSL2 is specifically localized on the plasma membrane facing the maternal tissue of the basal endosperm transfer cell layer (BETL) and functions in loading Zn‐NA into the BETL. Overexpression of ZmYSL2 increases the Zn concentration in the kernels by 31.6%, which achieves the goal of Zn biofortification of maize. These findings resolve the mystery underlying the loading of Zn into plant seeds, providing an efficient strategy for breeding or engineering maize varieties with enriched Zn nutrition.

The metal‐NA (Nicotianamine) transporter ZmYSL2 loads Zn from the maternal tissue into maize kernels. Overexpression of ZmYSL2 increases the amount of Zn in maize kernels, providing a potential approach for breeding Zn‐rich crop.

Introduction

Zn biofortification of cereal‐based food is a fundamental method to address the problem of Zn deficiency, which affects 17% of people worldwide, and genetic improvement of the Zn content in cereal grains is the most economical and effective measure (White & Broadley, 2009; Maqbool & Beshir, 2019). Maize is one of the most important food crops in the world and one of the leading edible crops in regions with Zn malnutrition (Maqbool & Beshir, 2019; Gashu et al, 2021). However, maize typically has the lowest Zn concentration of all cereal crops, people relying on maize‐based diets are therefore most likely to suffer from Zn deficiency (Gashu et al, 2021). The large variation in kernel Zn content has been found (Long et al, 2004) to be largely due to genetic contributions (Simic et al, 2012; Baxter et al, 2013), suggesting that genetic improvement of Zn biofortification in maize is feasible. However, no genetic locus controlling natural variation in the Zn content of maize kernels has been successfully cloned, limiting genetic strategies for Zn biofortification.

Based on studies conducted mainly in Arabidopsis thaliana, a series of genes controlling the uptake and long‐distance transport of Zn have been identified and characterized. Zn uptake from the rhizosphere is controlled by metal‐ion transporters of the ZRT‐ and IRT‐like protein (ZIP) family (Robinson et al, 1999; Vert et al, 2002), while heavy metal‐transporting P1B‐ATPases (HMAs) 2 and 4 are responsible for the long‐distance transport of Zn from roots to shoots (Hussain et al, 2004; Hanikenne et al, 2008). Yellow stripe 1 (YS1) and YS1‐like (YSL) proteins function in transporting metals complexed with phytosiderophores or nicotianamine (NA) into the cell and have been extensively studied for their roles in root uptake and phloem loading of Fe (Curie et al, 2001; Le Jean et al, 2005). However, there is conflicting evidence on the role of YSL transporters in Zn homeostasis (Waters & Grusak, 2008), questioning whether and how YSL transporters control Zn homeostasis.

The transport of Zn from the mother plant to seeds is critical for biofortification. This process involves at least two steps: unloading of Zn from the mother plant to the maternal–filial interface and uploading of Zn from the maternal–filial interface to the seed domain. In A. thaliana, HMA2 and HMA4 were found to mediate the export of Zn from maternal tissues to the apoplastic space between the seed and the mother plant (Olsen et al, 2016). Their ortholog in barley, HvHMA4, was also found to be expressed in the cells connecting the maternal domain and the filial tissue (Tauris et al, 2009), suggesting that a similar mechanism of uploading of Zn from the mother plant exists in barley. However, it is unknown how Zn is loaded to filial tissues from the maternal–filial interface.

ZmYSL2 was reported to function as metal–NA transporter to control endosperm development by affecting Fe distribution between the embryo and the endosperm (Zang et al, 2020; He et al, 2021). However, there were conflicting results about whether ZmYSL2 affects the total Fe content of the kernels (Zang et al, 2020; He et al, 2021). Besides, the reported expression pattern of ZmYSL2 in aleurone and sub‐aleurone layers in endosperm seemed inconsistent with the ion phenotype of the mutants (He et al, 2021), questioning the functional mechanisms of ZmYSL2. In addition, Fe³⁺‐DMA was found to be the major form of Fe in the phloem of cereal plants (Nishiyama et al, 2012), and DMA but not NA affects the delivery of Fe into the seeds of gramineous plants (Diaz‐Benito et al, 2018; Chao & Chao, 2022). Instead, Zn‐NA is the major form of Zn in cereal phloem and NA but not DMA controls the translocation of Zn into the gramineous seeds (Nishiyama et al, 2012; Diaz‐Benito et al, 2018). These data suggested that the role of ZmYSL2 in metal homeostasis needs to be addressed.

In this study, we performed a genome‐wide association study (GWAS) to identify a QTL controlling natural variation in the kernel Zn content. We established that the YSL family transporter ZmYSL2 is the causal gene in the kernel Zn QTL, which is specifically expressed in the maize basal endosperm transfer layer (BETL) and functions in the uploading Zn‐NA into the filial domain from the maternal–filial interface. Overexpression of ZmYSL2 was able to significantly increase the Zn concentration by 30%. This study not only is the first to clone a QTL controlling the kernel Zn content of maize and identified a transporter for zinc uploading to maize kernels but also offers a new method for breeding Zn‐biofortified crops.

Results

GWAS of Zn content in maize kernels

We analyzed the kernel Zn concentrations of 1,404 inbred lines from a synthetic, complete‐diallel design plus an unbalanced breeding‐like inter‐cross (CUBIC) population (Liu et al, 2020b) and observed kernel Zn concentrations ranging from 9.2 to 48.3 μg g⁻¹ dry weight (DW) across the population (Fig 1A). We performed a GWAS by applying a mixed‐model approach to reveal a 300 kb interval on chromosome 2 containing 85 SNPs that were closely associated with variation in the kernel Zn content (P‐values < 10⁻⁷) (Fig 1B and C). The most closely associated SNP (−log P‐value = 9.03) is Chr5_190655347 (B73 v3). The lines with two guanines (GG) at Chr5_190655347 had average kernel Zn concentrations 22.8% higher than those of the lines with two thymines (TT), and the minor T allele was present in 6.7% of the lines studied (Fig 1D).

Figure 1. GWAS for the kernel Zn content in maize.

• A
  Frequency distribution of the seed Zn concentration in 1311 progenies of the CUBIC population.
• B
  Manhattan plot for the GWAS. The dashed horizontal line depicts the Bonferroni‐adjusted significance threshold (P = 1.0 × 10⁻⁷). The SNPs located within the candidate genes are labeled as red dots. The red arrow indicates the most significant SNP chr5.s_190655347.
• C
  Regional Manhattan plot of the candidate genomic region on chromosome 5 and pairwise LD analysis. The 2‐Mb genomic region on either side of the most significant SNP is shown. The SNPs located within the candidate genes are labeled as red dots. The red arrow indicates the most significant SNP.
• D
  Zn content of the genotypes of the CUBIC population separated by the peak SNP. n denotes the number of genotypes. The central band in the boxplot represents the median of the sample values and the boxes depict the interquartile range. The whiskers indicate variability outside the upper and lower quartiles.
• E, F
  Haplotypes (Hap) of the ZmYSL2 gene body (E) and promoter (5 kb before ATG) (F) region among maize natural variations. SNPs in each haplotype are indicated in different colors, where red indicates T, black indicates G, blue indicates C, and green indicates A. Gray indicates the same sequences among each haplotype.
• G
  Zn content of four haplotypes. n denotes the number of genotypes belonging to each haplotype group. Statistical significance was determined by ANOVA. The different lowercase letters above the bars indicate significant differences by one‐way ANOVA (P < 0.01). The central band in the boxplot represents the median of the sample values and the boxes depict the interquartile range. The whiskers indicate variability outside the upper and lower quartiles.
• H
  Substrate transport activity test of four haplotypes. The zrt1zrt2 mutant yeast strain was transformed with different vectors. Yeast cells were spotted on SD‐Ura medium containing ZnSO₄ (10 μM) or Zn‐NA (10 μM). Plates were incubated for 3 days at 30°C.
• I
  Correlation analysis of the Zn content in mature maize seeds and ZmYSL2 expression level at 20 DAP. Each haplotype was labeled with different colors as indicated.

Source data are available online for this figure.

Of the 10 genes localized in the 300 kb LD region of the top SNP (Table 1), GRMZM2G135291 and GRMZM2G156599 encoded two proteins belonging to the YSL family, namely ZmYSL2 and ZmYS1, which could be the best candidates for controlling the natural variation in the kernel Zn concentrations of maize. However, ZmYS1 has been reported to control iron (Fe) uptake from the soil, and loss of function of ZmYS1 results in a severe shoot deficiency phenotype (Curie et al, 2001), but we did not identify any SNPs around this locus that were significantly associated with the leaf or kernel Fe concentrations (Fig EV1A–F), suggesting that there was no functional variation present in ZmYS1 and that ZmYSL2 may thus be responsible for the natural variation in the kernel Zn concentrations of maize.

Table 1.

Genes within 300 kb of the SNP most closely associated with the kernel content Zn in the CUBIC population.

Gene_ID	Start	Stop	Annotation	Ortholog in A. thaliana	Distance to peak SNP
GRMZM2G117164	190526952	190528628	Homeobox leucine zipper protein ATHB‐6	AT3G61890	126,719 bp
GRMZM2G117238	190532564	190536335	Origin recognition complex subunit 2	AT2G37560	119,012 bp
GRMZM2G072865	190605603	190612676	ATP‐dependent zinc metalloprotease FTSH7	AT3G47060	42,671 bp
GRMZM2G135385	190614417	190615314	Cytochrome b5	AT2G46650	40,033 bp
GRMZM2G135291	190656253	190659274	Iron‐nicotianamine transporter yellow stripe like 2	AT5G53550	906 bp
GRMZM2G156599	190726051	190729182	Iron–phytosiderophore transporter yellow stripe 1	AT5G53550	70,703 bp
GRMZM2G156585	190735441	190737796	Threonine synthase 1	AT4G29840	80,094 bp
GRMZM2G156575	190738088	190739754	Unknown protein	AT5G51170	82,741 bp
GRMZM6G037947	190741052	190741387	Unknown protein	–	85,705 bp
GRMZM2G031952	190786926	190791823	ELMO domain‐containing protein 2	AT3G60260	131,579 bp

Figure EV1. GWAS for the kernel Fe, leaf Zn and Fe levels in maize, and correlation analysis of the Fe content in maize seeds and ZmYSL2 expression level.

1. Frequency distribution of the leaf Fe content in 1311 progenies of the CUBIC population.
2. Genome‐wide association analysis of leaf Fe content.
3. The frequency distribution of kernel Fe content.
4. Genome‐wide association analysis of the kernel Fe content.
5. Frequency distribution of the leaf Zn content.
6. Genome‐wide association analysis of the leaf Zn content.
7. Correlation analysis of the Fe content in mature maize seeds and ZmYSL2 expression level at 20 DAP.

Data information: The dashed horizontal line in (B), (D), and (F) depicts the Bonferroni‐adjusted significance threshold (P = 1.0 × 10⁻⁷).

Source data are available online for this figure.

We analyzed the ZmYSL2 sequences of the 24 parents used to generate the CUBIC population and found numerous polymorphic sites present in the ZmYSL2 genomic region (Fig 1E), which can be divided into four major haplotypes (Table 2), with haplotype I accumulating 18.6% less Zn in the kernels than the other three haplotypes (Fig 1G). The four ZmYSL2 haplotypes encoded four different ZmYSL2 protein variants, but all the amino acid variations in four haplotypes were non‐conserved sites among YSL2 orthologs (Appendix Fig S1) and all these ZmYSL2 variants were able to complement the Zn‐uptake defective yeast mutant zrt1zrt2 when the growth medium was supplemented with 10 μM Zn‐NA but not when the growth medium was supplemented with 10 μM Zn (Fig 1H). These data supported that ZmYSL2 functions in uploading Zn‐NA to the kernel, but the amino acid changes among these variants are not responsible for natural variation in the kernel Zn concentrations.

Table 2.

Nucleotide and amino acid variations in exons of ZmYSL2 among four haplotypes.

Nucleotide acid position (reference to B73)		196	209	705	1201	1291	1495	1540
Haplotype	Maize cultivar
Hap1	HuangC	G	AGC	A	A	T	C	T
	E28	G	AGC	A	A	T	C	T
	B73	G	AGC	A	A	T	C	T
Hap2	HZS	C	AGC	G	T	T	C	G
	HYS	C	AGC	G	T	T	C	G
	NX110	C	AGC	G	T	T	C	G
	Chang7‐2	C	AGC	G	T	T	C	G
	Lx9801	C	AGC	G	T	T	C	G
	Ji853	C	AGC	G	T	T	C	G
	W22	C	AGC	G	T	T	C	G
	W64A	C	AGC	G	T	T	C	G
Hap3	Lv28	C	–	G	T	T	G	T
	Ji53	C	–	G	T	T	G	T
	K12	C	–	G	A	T	G	T
	Shuang741	C	–	G	T	T	G	T
	Zong31	C	AGCAGC	G	A	T	G	T
	F349	C	AGCAGC	G	A	T	C	G
	Zi330	C	AGCAGC	G	A	T	G	T
Hap4	YFH	C	AGCAGC	G	A	A	C	G
	Xi502	C	AGCAGC	G	A	A	C	G
	Q1261	C	AGCAGC	G	A	A	C	G
	TY4	C	AGCAGC	G	A	A	C	G
	Dan340	C	AGCAGC	G	A	A	C	G
	H21	C	AGCAGC	G	A	A	C	G
	Zong3	C	–	G	A	A	C	G
	81,515	C	AGCAGC	G	A	A	C	G
	5,237	C	AGCAGC	G	A	A	C	G
	KN5585	C	AGCAGC	G	A	A	C	G
Amino acid change and location (reference to B73)		V66L	Q	N179D	I261F	W291R	L324V	Y339D

Nucleotide acid position (reference to B73 genome)		1814	1816	1943	2016	2522	2725	2729
Haplotype	Maize cultivar
Hap1	HuangC	G	A	A	C	C	G	G
	E28	G	A	A	C	C	G	G
	B73	G	A	A	C	C	G	G
Hap2	HZS	C	A	A	C	T	T	A
	HYS	C	A	A	C	T	T	A
	NX110	C	A	A	C	T	T	A
	Chang7‐2	C	A	A	C	T	T	A
	Lx9801	C	A	A	C	T	T	A
	Ji853	C	A	A	C	T	T	A
	W22	C	A	A	C	T	T	A
	W64A	C	A	A	C	T	T	A
Hap3	Lv28	C	G	A	G	C	G	G
	Ji53	C	G	A	G	T	T	A
	K12	C	G	A	G	T	G	G
	Shuang741	C	G	A	G	C	G	G
	Zong31	C	G	A	G	T	G	G
	F349	C	A	A	C	T	T	A
	Zi330	C	G	A	G	T	T	A
Hap4	YFH	C	G	G	C	C	G	G
	Xi502	C	G	G	C	C	G	G
	Q1261	C	G	G	C	C	G	G
	TY4	C	G	G	C	C	G	G
	Dan340	C	G	G	C	C	G	G
	H21	C	G	G	C	C	G	G
	Zong3	C	G	A	G	T	G	G
	81,515	C	G	G	C	C	G	G
	5,237	C	G	G	C	C	G	G
	KN5585	C	G	G	C	C	G	G
Amino acid change and position (reference to B73)		S399T	S400G	I411V	A435G	P604S	L671F	V673M

Reference B73, three wild‐type background W22, W64A, KN5585, and 24 parents of the CUBIC population were shown and all the materials were classified into four haplotypes according to sequence similarity. Numbers in the first row show the positions of nucleotide acid variation which causes sense amino acid variations on the gene body of ZmYSL2 in reference B73 starting from the start codon ATG. The last row shows the amino acid changes and positions on the protein of ZmYSL2 in reference B73 starting from the first amino acid Met.

In contrast to the gene body region, the promoter region of ZmYSL2 exhibited much greater variability. Except for the 300‐bp region upstream of the start codon of ZmYSL2, the 5‐kb promoter regions shared little similarity (Fig 1F). The expression of ZmYSL2 in the kernels was linearly positively correlated with the kernel Zn concentration (Fig 1I) but not with the kernel Fe concentrations (Fig EV1G). In addition, the lines with haplotype I showed significantly lower ZmYSL2 expression in kernels than the lines with the other three haplotypes (Fig 1I). In contrast, there is no polymorphism in the intron region significantly associated with the variation in the kernel Zn concentration (Appendix Fig S2). These data indicated that polymorphisms in the promoter of ZmYSL2 most likely drive natural variations in the expression of ZmYSL2 and the kernel Zn concentrations.

ZmYSL2 controls total Zn but not Fe content in the kernels

To further study the role of ZmYSL2, we isolated four genetic mutants for this gene (Figs 2A–G and EV2). Among them, criysl2‐1 and criysl2‐2 were knockout mutants generated by CRISPR‐Cas9 technology in the KN5585 background (Figs 2B and C and EV2B). The mutant zmysl2‐1 was a transposon‐insertion mutant in the W22 background (Figs 2D and EV2A). The mutant zmysl2‐2 was an EMS‐mutagenized mutant in the W64A background (Figs 2E–G and EV2C), in which the conserved alanine at the 133^rd residue was substituted with a valine (Fig EV2C).

Figure 2. ZmYSL2 is responsible for loading Zn into maize kernels.

• A
  Information for the ZmYSL2 mutants.
• B–G
  Representative ear images of heterozygous criysl2‐1 (B), criysl2‐2 (C), zmysl2‐1 (D), and zmysl2‐2 (E) plants and homozygous zmysl2‐2 (G) and the corresponding wild‐type W64A (F) plants. The white arrow indicates genotyped homozygous criysl2‐1, criysl2‐2, zmysl2‐1, and zmysl2‐2 seeds and the red arrow indicates corresponding genotyped wild type. Scale bars, 1 cm.
• H–J
  μXRF analysis of the Zn content and distribution in kernels of homozygous criysl2‐1 (H), criysl2‐2 (H), zmysl2‐1 (I), and zmysl2‐2 (J) mutants and the corresponding wild‐type genotypes. The upper panel shows the mature kernels cut in half, and the lower panel shows the Zn content and distribution. The dashed line depicts different parts of the kernel, where the red line indicates the whole kernel, the yellow line indicates the embryo and endosperm, and the green line indicates the embryo.
• K–S
  Zn content in whole kernels (K, N, Q), embryos (L, O, R), and endosperm (M, P, S) of criysl2‐1/2 (K–M), zmysl2‐1 (N–P), and zmysl2‐2 (Q–S) detected by ICP‐MS. Samples of mature kernels were collected. The data represent the means ± SDs. n = 6–12 biological replicates. Statistical significance was determined by ANOVA in (K–M) and Student's t‐test in (N–S). The different lowercase letters above the bars indicate significant differences by one‐way ANOVA (P < 0.01) (K–M). **P < 0.001 (Student's t‐test).

Figure EV2. CRISPR/Cas9‐mediated mutagenesis in ZmYSL2 .

1. Schematic representation of ZmYSL2 and the positions of the CRISPR/Cas9 target, Mu insertion site, and EMS mutation site on the genes. Red letters show the sequence of the target site, and blue letters show the mutant nucleotide type between W64A and the EMS mutant zmysl2‐2.
2. Genotypes of ZmYSL2 in the criysl2‐1 and criysl2‐2 mutants. Red lines show the sequences mutagenized in the mutants.
3. EMS mutation site on ZmYSL2.

By using a micro‐X‐ray fluorescence spectrometer (μXRF), we observed that Zn was enriched in the embryos of the kernels, and the Zn content in the kernels carrying a homozygous knockout or impaired allele of ZmYSL2 was notably lower than that in the corresponding wild‐type kernels from the same self‐pollinated heterozygous ear (Fig 2H–J; i.e., kernels carrying homozygous criysl2‐1 accumulated much less Zn than wild‐type kernels from the same ear). This observation was further confirmed by elemental analysis by inductively coupled plasma‐mass spectrometry (ICP‐MS), which showed that the kernels carrying homozygous mutant alleles of ZmYSL2 accumulated much less Zn in either the embryos or endosperm than the kernels with the corresponding wild‐type genotypes from the same ear (Fig 2K–S). These data showed that loading Zn to the kernels requires ZmYSL2; specifically, it requires ZmYSL2 expressed in the filial tissues of the kernels.

Compared with the Zn content, the total Fe content in the kernels was not affected by mutation of ZmYSL2 (Fig EV3A, D, and G). However, the Fe distribution was changed in the kernels of crisysl2‐1, crisysl2‐2, and zmysl2‐1, in which Fe was accumulated in the endosperm rather than embryo (Fig EV3B, C, E, F, and H–L), indicating that the mother‐to‐offspring delivery mechanism of Fe is different from that of Zn and ZmYSL2 may control Fe transport from endosperm to embryo in an unknown manner. As ZmYSL2 can transport Fe(II)‐NA (Fig EV3M) (Zang et al, 2020; He et al, 2021), we wondered whether Fe is transported into the filial tissue of maize in another form that is not a substrate of ZmYSL2. Previous studies reported that Fe(III)‐DMA is the major Fe form in the phloem of gramineous plants (Nishiyama et al, 2012) and DMA rather than NA is essential for Fe accumulation in cereal seeds (Diaz‐Benito et al, 2018), indicating that Fe(III)‐DMA rather than Fe(II)‐NA is the major form for Fe transport into the maize kernel. We then confirmed that Fe(III)‐DMA is not a substrate of ZmYSL2 by using yeast complementation experiment (Fig EV3M), explaining why mutation of ZmYSL2 does not affect the total Fe content in the kernels. Phytic acid is another Fe chelator in seeds and its content was not changed in the kernels of zmysl2‐1 (Appendix Fig S3).

Figure EV3. Fe phenotypes of the criysl2‐1, criysl2‐2, zmysl2‐1, and zmysl2‐2 mutants and the corresponding wild type.

• A–I
  Fe content in the whole kernels (A, D, G), embryos (B, E, H), and endosperm (C, F, I) of criysl2‐1/2 (A–C), zmysl2‐1 (D‐F), and zmysl2‐2 (G‐I) determined by ICP‐MS. Samples of mature seeds were collected. The data represent the means ± SDs. n = 6–12 biological replicates. Statistical significance was determined by ANOVA in (A–C) and Student's t‐test in (D–I). The different lowercase letters above the bars indicate significant differences by one‐way ANOVA (P < 0.01) (A–C). **P < 0.001 (Student's t‐test).
• J–L
  μXRF analysis of the Fe content and distribution in kernels of homozygous criysl2‐1 (J), criysl2‐2 (J), zmysl2‐1 (K), and zmysl2‐2 (L) mutants and the corresponding wild‐type genotypes. The upper panel shows the mature seeds cut in half, and the lower panel shows the Zn content and distribution. The dashed line depicts different parts of the kernel, where the red line indicates the whole kernel, the yellow line indicates the embryo and endosperm, and the green line indicates the embryo.
• M
  Transport activities of ZmYSL2 for Fe. The fet3fet4 mutant yeast strain was transformed with expression vectors as indicated. Yeast cells were spotted on SD‐Ura medium containing FeSO₄ (0, 10, and 100 μM), Fe²⁺‐NA (10 μM), or Fe³⁺‐DMA (10 μM).

Source data are available online for this figure.

The opaque phenotype of ZmYSL2 mutants is due to deficiency of Zn in the kernels

In addition to the ion phenotype, we found that the kernels in homozygous criysl2‐1, criysl2‐2, or zmysl2‐1 genotypes derived from the self‐pollinated heterozygous knockout mutants showed an opaque endosperm and were not able to germinate (Fig 2A–G), which is consistent with previous reports (Zang et al, 2020; He et al, 2021). In contrast, although the kernels in homozygous zmysl2‐2 also showed an opaque phenotype (Fig 2A and G), they were germinable and viable, suggesting that zmysl2‐2 was a weak allele. Moreover, the Zn concentrations in the kernels of crisysl2‐1, crisysl2‐2, and zmysl2‐1 were decreased 73.0%–71.3% compared with those in the corresponding wild types, while zmysl2‐2 accumulated Zn in the kernels 62.7% less than wild‐type W64A (Fig 2K, N, and Q). This result further indicated that zmysl2‐2 was a weak allele rather than a knockout allele. ZmYSL2 was recently reported to control endosperm development by affecting the distribution of Fe between the embryo and the endosperm (Zang et al, 2020). Considering that the Fe distribution was not affected in zmysl2‐2 and kernels of zmysl2‐2 were opaque and viable, these data suggested that the opaque phenotype was caused by Zn deficiency, while the embryo lethal phenotype was caused by Fe deficiency, which is consistent with our previous observation in A. thaliana (Chao et al, 2021). Kernels carrying homozygous mutant alleles of ZmYSL2 also accumulated much less cobalt (Co) than the kernels with the corresponding wild‐type genotypes from the same ear (Tables 3, 4), indicating that loading Co to the kernels also requires ZmYSL2. These data are also consistent with our finding in A. thaliana, which revealed that translocation of Co to seed is dependent on NA but lack of Co does affect seed development (Chao et al, 2021).

Table 3.

Kernel element profile of crizmysl2 mutants and the corresponding wild‐type.

Element	Content (μg/g DW)			Percentage difference from KN5585 (%)		P‐values
	KN5585	criysl2‐1	criysl2‐2	criysl2‐1	criysl2‐2
B	103.55 ± 7.98	101.26 ± 15.96	106.38 ± 16.47	−2.22	2.73	ns
Na	0.15 ± 0.04	0.14 ± 0.04	0.16 ± 0.04	−6.95	9.57	ns
Mg	473.74 ± 43.98	500.19 ± 76.00	455.10 ± 86.59	5.58	−3.93	ns
P	2800.65 ± 412.75	2625.68 ± 316.05	3097.46 ± 360.83	−6.25	10.60	ns
S	577.48 ± 61.40	531.61 ± 70.60	566.38 ± 49.11	−7.94	−1.92	ns
K	4630.48 ± 513.13	4684.84 ± 706.40	4549.11 ± 580.89	1.17	−1.76	ns
Ca	328.06 ± 60.76	343.51 ± 67.09	276.79 ± 79.57	4.71	−15.63	ns
Ti	7.42 ± 0.73	8.60 ± 1.86	7.86 ± 1.91	16.01	5.94	ns
Cr	0.07 ± 0.02	0.07 ± 0.01	0.07 ± 0.03	5.41	14.86	ns
Mn	9.80 ± 1.89	9.97 ± 1.83	10.31 ± 1.61	1.80	5.18	ns
Fe	17.32 ± 1.71	18.58 ± 2.85	16.92 ± 2.08	7.25	−2.31	ns
Co	0.0044 ± 0.0007	0.0021 ± 0.0006	0.0020 ± 0.0007	−53.07	−53.96	< 0.0006
Ni	0.12 ± 0.04	0.11 ± 0.55	0.12 ± 0.03	−7.89	−2.91	ns
Cu	1.10 ± 0.09	1.10 ± 0.16	1.20 ± 0.26	0.03	8.43	ns
Zn	18.70 ± 1.67	6.99 ± 1.24	5.46 ± 1.41	−62.61	−70.82	< 0.0006
As	0.04 ± 0.01	0.05 ± 0.02	0.04 ± 0.01	22.01	−4.76	ns
Se	1.24 ± 0.43	1.58 ± 0.40	1.32 ± 0.33	27.33	6.28	ns
Rb	1.58 ± 0.19	1.47 ± 0.40	1.42 ± 0.50	−6.95	−10.13	ns
Sr	0.64 ± 0.13	0.68 ± 0.19	0.51 ± 0.18	6.22	−19.31	ns
Mo	0.03 ± 0.003	0.02 ± 0.004	0.02 ± 0.005	−10.12	−5.31	ns
Cd	0.009 ± 0.003	0.009 ± 0.002	0.009 ± 0.003	−0.71	−9.52	ns
Pb	0.64 ± 0.14	0.61 ± 0.13	0.58 ± 0.15	−4.55	−8.98	ns

Data represent the mean ± sd (n = 9–12). Data in bold represent elements with a significant difference (P ≤ 0.01) between the crizmysl2 mutants and KN5585 in both mutants. ns, not significant. Statistical significance was determined by Student's t‐test.

Source data are available online for this Table.

Table 4.

Kernel element profile of zmysl2‐1, zmysl2‐2, and the corresponding wild type.

Element	Content (μg g DW−1)				Percentage difference from WT (%)		P‐values
	W22	zmysl2‐1	W64A	zmysl2‐2	zmysl2‐1	zmysl2‐2
B	75.86 ± 16.87	81.12 ± 21.79	90.15 ± 17.58	97.97 ± 12.99	6.94	8.68	ns
Na	0.10 ± 0.03	0.12 ± 0.03	0.12 ± 0.01	0.12 ± 0.03	13.59	0.07	ns
Mg	415.61 ± 46.61	401.31 ± 74.60	586.57 ± 49.68	562.73 ± 52.29	−3.44	−4.06	ns
P	2879.30 ± 410.01	2565.63 ± 677.66	3618.49 ± 573.40	3501.93 ± 361.83	−10.89	−3.22	ns
S	392.11 ± 56.58	418.49 ± 93.96	568.28 ± 92.46	601.32 ± 52.79	6.73	5.81	ns
K	2702.49 ± 377.53	2516.47 ± 222.70	3527.67 ± 577.60	3743.58 ± 604.17	−6.88	6.12	ns
Ca	129.24 ± 44.38	144.26 ± 34.32	156.26 ± 28.42	162.59 ± 38.01	11.62	4.03	ns
Ti	7.40 ± 1.17	7.97 ± 1.93	8.76 ± 1.25	8.46 ± 1.10	7.65	−3.35	ns
Cr	0.05 ± 0.01	0.05 ± 0.01	0.03 ± 0.01	0.04 ± 0.01	−4.83	16.54	ns
Mn	5.79 ± 0.77	5.28 ± 1.14	6.18 ± 0.70	6.66 ± 0.096	−8.80	7.89	ns
Fe	17.77 ± 2.63	17.95 ± 3.19	20.61 ± 2.42	20.01 ± 2.28	1.03	−2.91	ns
Co	0.0053 ± 0.0005	0.0029 ± 0.0007	0.0138 ± 0.0015	0.0067 ± 0.0009	−45.63	−51.57	< 0.0006
Ni	0.34 ± 0.11	0.35 ± 0.08	0.80 ± 0.84	0.61 ± 0.66	4.14	−23.01	ns
Cu	2.04 ± 0.31	1.50 ± 0.21	0.81 ± 0.14	0.85 ± 0.16	−26.71	4.89	ns
Zn	18.96 ± 2.49	6.98 ± 1.50	15.71 ± 1.99	7.08 ± 1.25	−63.17	−54.91	< 0.0006
As	0.05 ± 0.01	0.05 ± 0.01	0.06 ± 0.01	0.06 ± 0.01	5.91	1.02	ns
Se	1.39 ± 0.51	1.47 ± 0.59	0.86 ± 0.11	0.97 ± 0.20	5.85	13.78	ns
Rb	1.55 ± 0.21	1.59 ± 0.27	1.19 ± 0.19	1.17 ± 0.24	2.63	−2.04	ns
Sr	0.22 ± 0.06	0.24 ± 0.05	0.15 ± 0.05	0.15 ± 0.05	10.71	2.26	ns
Mo	0.05 ± 0.006	0.04 ± 0.009	0.02 ± 0.003	0.02 ± 0.002	−10.76	3.51	ns
Cd	0.008 ± 0.003	0.008 ± 0.002	0.007 ± 0.001	0.007 ± 0.001	−0.25	−6.40	ns
Pb	0.51 ± 0.12	0.51 ± 0.13	0.53 ± 0.10	0.47 ± 0.08	−1.34	−11.39	ns

Data represent the mean ± sd (n = 12). Data in bold represent elements with a significant difference (P ≤ 0.01) between the mutants and wild type in both mutants. ns, not significant. Statistical significance was determined by Student's t‐test.

Source data are available online for this Table.

Allelism tests for different ZmYSL2 mutants

To confirm that the phenotypes above were caused by mutation of ZmYSL2, we performed pairwise crosses among criysl2‐1/+, zmysl2‐1/+, and homozygous zmysl2‐2 (Fig 3). We found that the low Zn kernels and high Zn kernels on the ears from the crosses between zmysl2‐1/+ and criysl2‐1/+ segregated in a 1:3 segregation ratio (Fig 3A and B and Table 5) as expected and the low Zn phenotype and the opaque phenotype were co‐segregated. In addition, all the kernels with low Zn content and opaque phenotype bear a criysl2‐1/zmysl2‐1 genotype, indicating that those phenotypes of criysl2‐1 and zmysl2‐1 are caused by the same gene (Fig 3B and C and Table 5).

Figure 3. Allele test of the criysl2, zmysl2‐1, and zmysl2‐2 alleles.

• A
  Images of reciprocally pollinated ears among heterozygous criysl2‐1, zmysl2‐1, and homozygous zmysl2‐2 plants. The white arrow indicates genotyped homozygous ysl2 mutant kernels and the red arrow indicates corresponding genotyped wild‐type kernels. Scale bars, 1 cm.
• B
  Images of genotyped mature kernels of reciprocally pollinated homozygous ysl2 mutants and corresponding heterozygous ysl2 and wild‐type seeds.
• C–D
  μXRF analysis of the Zn and Fe content and distribution in genotyped mature kernels of reciprocally pollinated homozygous ysl2 mutants and corresponding heterozygous ysl2 and wild‐type seeds. The dashed line depicts different parts of the kernel, where the red line indicates the whole kernel, the yellow line indicates the embryo and endosperm, and the green line indicates the embryo.

Table 5.

Allele test of the ysl2 mutants.

Ear	+/+ and +/− (high Zn and normal endosperm)	ysl2/− (low Zn and opaque endosperm)	Observed ratio	Expected ratio	P‐value for Chi‐square test
zmysl2‐1 /+ x criysl2‐1 /+	171	54	3.17:1	3:1	0.80
criysl2‐1 /+ x zmysl2‐1 /+	52	16	3.25:1	3:1	0.95
zmysl2‐2 x criysl2‐1 /+	99	105	0.94:1	1:1	0.77
criysl2‐1 /+ x zmysl2‐2	23	21	1.09:1	1:1	0.83
zmysl2‐1 /+ x zmysl2‐2	104	109	0.95:1	1:1	0.81
zmysl2‐2 x zmysl2‐1 /+	64	58	1.1:1	1:1	0.70

Consistent with the above observation, the segregation ratio of the low Zn kernels and high Zn kernels on the ears from the crosses between zmysl2‐1/+ and zmysl2‐2 was 1:1 (Fig 3A and B and Table 5) and the opaque phenotype was also co‐segregated with the low Zn phenotype. Moreover, all the kernels with a zmysl2‐1/zmysl2‐2 genotype showed opaque endosperm and low Zn phenotype (Fig 3B and C and Table 5). These data confirmed that zmysl2‐1 and zmysl2‐2 are two loss‐of‐function/reduced alleles of the same locus. The crosses between criysl2‐1/+ and zmysl2‐2 confirmed that criysl2‐1 and zmysl2‐2 are two alleles of the same gene. These data together established that ZmYSL2 is the causal gene responsible for the low Zn and opaque phenotypes of the three mutants.

Furthermore, the pairwise crossed ears with criysl2‐1, zmysl2‐1, and zmysl2‐2 genotypes all exhibited low‐Zn phenotypes while abnormal Fe distribution phenotype was only observed in kernels with zmysl2‐1/criysl2‐1 and criysl2‐1/zmysl2‐1 genotype (Fig 3D). These data further supported that the opaque phenotype was caused by Zn deficiency.

ZmYSL2 is responsible for loading Zn from the placenta–chalaza domain to the BETL

According to the qRT–PCR results, ZmYSL2 was predominantly expressed in kernels after 12 days after pollination (DAP) (Fig 4A). Further investigation showed that ZmYSL2 was dramatically upregulated in the endosperm beginning at 12 DAP but not in the embryo or the endosperm without BETL tissue (Fig 4B). These data suggested that ZmYSL2 is predominantly expressed in the BETL during the filling stage, which was then confirmed by in situ hybridization of ZmYSL2 (Fig 4C). Expression of ZmYSL2‐GFP in protoplasts isolated from leaf of A. thaliana showed that ZmYSL2‐GFP is localized on plasma membrane (Fig 4D). An immunofluorescence assay using a specific ZmYSL2 antibody further indicated that ZmYSL2 was specifically localized on the plasma membrane of BETL cells facing the apoplastic space between the maternal tissue and the filial domain (Fig 4E). These data established that ZmYSL2 functions in the uptake of Zn‐NA from the maternal–filial interface to the BETL, a zone for transferring nutrients from the maternal tissue to the filial domain.

Figure 4. ZmYSL2 is specifically localized on the plasma membrane facing the maternal tissue of the BETL.

1. qRT–PCR analysis of ZmYSL2 during seed development. Error bars represent the standard deviations from three biological replicates. DAP, day after pollination. n = 3 biological replicates.
2. qRT–PCR analysis of ZmYSL2 in embryos, endosperm, and endosperm without BETL during seed development. BETL, basal endosperm transfer layer. Error bars represent the standard deviations from three biological replicates. DAP, day after pollination. n = 3 biological replicates.
3. RNA in situ hybridization of ZmYSL2 in seeds at 16 DAP in B73. Positive signals (indicated by the red arrow) are clearly restricted to BETL cells. No signals were observed in the negative control material. Scale bar, 500 μm.
4. Subcellular localization of ZmYSL2 revealed by transiently expressing ZmYSL2‐GFP in Arabidopsis mesophyll protoplasts using the pA7 vector. pA7‐GFP was used as a vector control. Scale bar, 20 μm.
5. Immunofluorescence assays of ZmYSL2 in seeds at 20 DAP in KN5585 and criysl2‐1. Scale bar, 500 μm.

Source data are available online for this figure.

Perfectly matching the expression pattern of ZmYSL2, the Zn concentrations in the wild‐type kernels gradually increased from 12 DAP (Fig 5A–C), while the Zn concentrations in the mutant kernels did not change until the late filling stage (Fig 5A–C). These results were further confirmed by μXRF observation (Fig 5D–I). In addition, we found that Zn was deposited in the maternal tissues next to the BETL of the kernels carrying homozygous zmysl2‐1 (Fig 5D–H). ICP‐MS data also showed that the Zn concentrations in the adjacent maternal tissues of the mutant kernels were notably higher than those in the wild‐type kernels (Fig 5J–L). These data further confirmed that ZmYSL2 functions in transporting Zn‐NA from maternal tissues to filial BETLs. However, Zn deposition in the adjacent maternal tissues of the mutant kernels disappeared in the mature stage, suggesting that the deposited Zn might be recirculated to the mother plant (Figs 4I and 2H–J). Moreover, we did not observe Fe deposition in the adjacent maternal tissues of the mutant kernels, either in ICP‐MS data (Fig EV4A–I) or in μXRF results (Fig EV4J–L), further indicating that Fe is not transported to the kernel via the same pathway as Zn.

Figure 5. ZmYSL2 is responsible for loading Zn from the maternal tissue into the BETL.

• A–C
  Zn concentration in the criysl2‐1(A), criysl2‐2(A), zmysl2‐1 (B), and zmysl2‐2 (C) mutants during seed development. Fresh kernels of each genotype on different days after pollination were weighed and examined by ICP‐MS after genotyping. Error bars represent the standard deviations from six biological replicates. DAP, day after pollination.
• D–I
  μXRF analysis of the Zn distribution in kernels of W22 and homozygous zmysl2‐1 mutants during seed development. The dashed line depicts different parts of the kernel, where the red line indicates the whole kernel, the yellow line indicates the embryo and endosperm, and the green line indicates the embryo. DAP, day after pollination.
• J–L
  Zn content in different parts of homozygous criysl2‐1 (J), criysl2‐2 (J), zmysl2‐1 (K), and zmysl2‐2 (L) seeds and the corresponding wild‐type genotypes at 20 DAP. Genotyped fresh kernels of each genotype at 20 DAP were first dissected into different parts, including the whole kernel, embryo, endosperm, and maternal tissue, and then weighed, followed by ICP‐MS analysis after drying. The data represent the means ± SDs. n = 6 biological replicates. **P < 0.001 (Student's t‐test).

Source data are available online for this figure.

Figure EV4. Fe content and distribution phenotypes of W22 and the homozygous zmysl2‐1 mutant during seed development.

• A–F
  μXRF analysis of the Fe distribution and content in kernels of W22 and the homozygous zmysl2‐1 mutant during seed development. The dashed line depicts different parts of the kernel, where the red line indicates the whole kernel, the yellow line indicates the embryo and endosperm, and the green line indicates the embryo.
• G–I
  Fe content in different parts of homozygous criysl2‐1 (G), criysl2‐2 (G), zmysl2‐1 (H), and zmysl2‐2 (I) seeds and the corresponding wild‐type genotypes at 20 DAP. Genotyped fresh kernels of each genotype at 20 DAP were first dissected into different parts, including the whole kernel, embryo, endosperm, and maternal tissue, and then weighed, followed by ICP‐MS analysis after drying. The data represent the means ± SDs. n = 6 biological replicates. **P < 0.001 (Student's t‐test).

Source data are available online for this figure.

Overexpression of ZmYSL2 increases the Zn content in maize kernels

To further explore the application potential of ZmYSL2, we generated three independent transgenic maize lines with a construct of pBETL9::ZmYSL2 that specifically overexpressed ZmYSL2 in the BETL using ZmBETL9 promoter (Zhan et al, 2015; Fig EV5A and B) and three lines with a construct of pUbi::ZmYSL2 that constitutively overexpressed ZmYSL2 (Fig EV5C and D). Elemental analysis and μXRF observation showed that the Zn contents in the kernels of the transgenic lines either constitutively overexpressing ZmYSL2 or overexpressing ZmYSL2 specifically in the BETL were significantly increased (Fig 6A, C, E, and F and Table 6). Among the overexpression lines, pUbi::ZmYSL2‐1 accumulated Zn at 39.9 mg kg⁻¹ (Fig 6C and Table 7), which was 31.6% higher than that in the wild type and 20% higher than the breeding goal for Zn biofortification (33 mg kg⁻¹) (Bouis & Welch, 2010).

Figure EV5. Relative expression of ZmYSL2, leaf Zn and Fe contents, and storage reserve accumulation in ZmYSL2 overexpression lines.

• A–D
  Relative expression of ZmYSL2 in KN5585, and three independent pBETL9::ZmYSL2 ^B73 transgenic lines (A‐B) and three independent pUbi::ZmYSL2 ^B73 transgenic lines (C–D) in 16 DAP kernel (A, C) and root (B, D). The values are shown as the means ± SDs (n = 3 independent samples).
• E–H
  Zn (E, G) and Fe (F, H) levels in leaves of KN5585 and three independent pBETL9::ZmYSL2 ^B73 transgenic lines (E, F) and three pUbi::ZmYSL2 ^B73 transgenic lines (G, H). Error bars represent the means ± SDs. n = 6 biological replicates.
• I
  Accumulation of zein and non‐zein storage proteins in mature seeds of KN5585 and three ZmYSL2 overexpression lines.
• J, K
  Total storage protein (J) and starch (K) content in mature seeds of KN5585 and three ZmYSL2 overexpression lines. The data represent the means ± SDs. n = 3.

Source data are available online for this figure.

Figure 6. Overexpression of ZmYSL2 increases the Zn content in maize kernels.

• A–D
  Zn (A, D) and Fe (B, D) concentrations in the kernels of KN5585, three independent pBETL9::ZmYSL2 ^B73 transgenic lines (A, B), and three pUbi:ZmYSL2 ^B73 transgenic lines (C, D). Error bars represent the means ± SDs. n = 6–12 biological replicates. The different lowercase letters above the bars indicate significant differences by one‐way ANOVA (P < 0.01) (K–M). **P < 0.001 (Student's t‐test). n = 6 biological replicates.
• E–F
  μXRF analysis of the Zn and Fe content and distribution in mature kernels of KN5585, three pBETL9::ZmYSL2 ^B73 transgenic lines (E), and three independent pUbi:ZmYSL2 ^B73 transgenic lines (F). The dashed line depicts different parts of the kernel, where the red line indicates the whole kernel, the yellow line indicates the embryo and endosperm, and the green line indicates the embryo. The photograph and SAXF images shown in (E) and (F) represent composite images generated by the TORNADO system.
• G
  Proposed model of ZmYSL2 function in maize kernel. ZmYSL2 is specifically localized on the plasma membrane facing the maternal tissue of the basal endosperm transfer cell layer (BETL) and functions in loading Zn‐NA into the BETL. Fe‐DMA complexes were transported into BETL by an unknown transporter. An HMA transporter may function in the export of free Zn from maternal tissue to the maternal–filial interface.

Source data are available online for this figure.

Table 6.

Kernel element profile of KN5585 and ZmYSL2 overexpression lines driven by pBETL9 promoter.

Element	Content (μg g DW−1)				Percentage difference from KN5585 (%)			P‐values
	KN5585	pBETL9::ZmYSL2‐1	pBETL9::ZmYSL2‐2	pBETL9::ZmYSL2‐3	pBETL9:: ZmYSL2‐1	pBETL9:: ZmYSL2‐2	pBETL9:: ZmYSL2‐3
B	13.16 ± 1.63	13.14 ± 1.29	13.15 ± 2.63	12.73 ± 2.15	−0.21	−0.13	−3.34	ns
Na	11.66 ± 2.28	11.21 ± 1.63	10.87 ± 1.76	10.60 ± 1.81	−3.84	−6.74	−9.08	ns
Mg	1045.55 ± 155.24	941.01 ± 140.40	940.25 ± 170.97	1023.88 ± 119.05	−10.00	−10.07	−2.07	ns
P	4175.04 ± 569.72	3813.68 ± 400.45	3934.58 ± 326.75	4283.34 ± 416.46	−8.66	−5.76	2.59	ns
S	568.28 ± 41.60	581.12 ± 57.89	539.43 ± 47.28	565.46 ± 76.81	2.26	−5.08	−0.50	ns
K	5367.44 ± 478.15	5146.50 ± 649.68	5158.04 ± 468.23	5407.55 ± 314.52	−4.12	−3.90	0.75	ns
Ca	198.44 ± 50.06	201.67 ± 38.78	199.49 ± 31.58	206.26 ± 51.62	1.63	0.53	3.94	ns
Ti	203.33 ± 52.34	200.94 ± 46.45	190.46 ± 61.81	211.44 ± 39.89	−1.18	−6.33	3.99	ns
Cr	0.32 ± 0.07	0.36 ± 0.06	0.34 ± 0.07	0.34 ± 0.04	12.08	6.06	7.11	ns
Mn	9.35 ± 1.36	9.24 ± 1.24	9.29 ± 1.38	9.77 ± 1.24	−1.16	−0.64	4.43	ns
Fe	23.2 ± 1.38	23.39 ± 1.40	22.94 ± 1.39	22.52 ± 1.31	0.83	−1.10	−2.94	ns
Co	0.0031 ± 0.0005	0.0048 ± 0.0008	0.0065 ± 0.0006	0.0038 ± 0.0009	53.31	104.87	21.95	< 0.03
Ni	0.11 ± 0.03	0.12 ± 0.03	0.11 ± 0.02	0.10 ± 0.03	5.54	−0.62	−7.07	ns
Cu	2.78 ± 0.50	2.83 ± 0.34	2.68 ± 0.28	2.82 ± 0.44	1.77	−3.42	1.36	ns
Zn	29.18 ± 1.59	34.45 ± 2.00	36.85 ± 1.29	31.18 ± 2.32	18.05	26.27	6.83	< 0.02
As	0.15 ± 0.02	0.15 ± 0.03	0.15 ± 0.02	0.15 ± 0.02	0.29	0.14	−1.10	ns
Se	2.53 ± 0.39	2.38 ± 0.49	2.69 ± 0.53	2.48 ± 0.42	−5.87	6.35	−2.13	ns
Rb	1.59 ± 0.13	1.53 ± 0.18	1.63 ± 0.20	1.50 ± 0.21	−3.90	2.81	−5.35	ns
Sr	0.62 ± 0.08	0.56 ± 0.08	0.58 ± 0.08	0.59 ± 0.06	−9.47	−5.76	−4.47	ns
Mo	0.17 ± 0.02	0.18 ± 0.03	0.17 ± 0.02	0.17 ± 0.02	2.63	−0.88	−0.13	ns
Cd	0.05 ± 0.008	0.06 ± 0.008	0.06 ± 0.007	0.06 ± 0.007	0.77	−3.67	−2.16	ns
Pb	3.83 ± 1.60	4.57 ± 0.78	3.76 ± 1.70	3.49 ± 1.02	19.18	−2.03	−8.82	ns

Data represent the mean ± sd (n = 12). Data in bold represent elements with a significant difference (P ≤ 0.05) between the overexpression lines and KN5585 in three independent lines. ns, not significant. Statistical significance was determined by Student's t‐test.

Source data are available online for this Table.

Table 7.

Kernel element profile of KN5585 and ZmYSL2 overexpression lines driven by Ubi promoter.

Element	Content (μg g DW−1)				Percentage difference from KN5585 (%)			P‐values
	KN5585	pUbi::ZmYSL2‐1	pUbi::ZmYSL2‐2	pUbi::ZmYSL2‐3	pUbi:: ZmYSL2‐1	pUbi:: ZmYSL2‐2	pUbi:: ZmYSL2‐3
B	9.10 ± 1.25	8.94 ± 1.03	8.95 ± 1.16	9.27 ± 0.63	−1.81	−1.65	1.88	ns
Na	9.94 ± 0.80	10.13 ± 0.74	10.29 ± 0.45	9.24 ± 0.93	1.92	3.53	−7.05	ns
Mg	1133.87 ± 64.48	1104.17 ± 45.57	1129.21 ± 29.75	1128.82 ± 33.50	−2.62	−0.41	−0.45	ns
P	4077.23 ± 179.30	4275.37 ± 188.26	4321.80 ± 230.65	4112.21 ± 328.26	4.86	6.00	0.86	ns
S	656.61 ± 28.81	640.87 ± 24.43	664.11 ± 21.81	627.49 ± 27.11	−2.40	1.14	−4.43	ns
K	4325.04 ± 220.48	4395.61 ± 186.76	4414.65 ± 161.91	3999.06 ± 238.04	1.63	2.07	−7.54	ns
Ca	167.19 ± 54.13	197.13 ± 19.25	176.81 ± 43.33	176.88 ± 49.17	17.91	5.76	5.79	ns
Ti	85.87 ± 9.24	77.80 ± 12.57	84.86 ± 14.69	80.97 ± 17.86	−9.39	−1.18	−5.70	ns
Cr	0.31 ± 0.04	0.31 ± 0.02	0.31 ± 0.01	0.30 ± 0.01	1.08	1.46	−4.78	ns
Mn	8.91 ± 1.00	11.07 ± 1.17	10.19 ± 0.88	10.00 ± 1.38	24.36	14.44	12.27	ns
Fe	25.90 ± 0.60	30.59 ± 0.68	26.86 ± 0.67	28.53 ± 0.75	18.13	3.72	10.15	< 0.02
Co	0.0032 ± 0.0006	0.0094 ± 0.0011	0.0055 ± 0.0003	0.0079 ± 0.0008	188.28	69.53	140.60	< 0.0006
Ni	0.11 ± 0.01	0.11 ± 0.01	0.10 ± 0.01	0.11 ± 0.01	0.36	−2.08	−0.76	ns
Cu	1.62 ± 0.17	2.41 ± 0.26	1.83 ± 0.15	2.14 ± 0.25	48.52	12.83	32.03	< 0.04
Zn	30.33 ± 0.65	39.92 ± 1.03	33.66 ± 0.55	36.75 ± 1.37	31.61	10.97	21.16	< 0.0006
As	0.11 ± 0.01	0.12 ± 0.01	0.11 ± 0.01	0.11 ± 0.01	6.86	1.67	−0.03	ns
Se	1.70 ± 0.29	1.68 ± 0.24	1.75 ± 0.21	1.58 ± 0.21	−1.25	−3.22	−7.70	ns
Rb	1.94 ± 0.23	2.03 ± 0.17	1.81 ± 0.10	1.89 ± 0.15	4.35	−6.82	−2.37	ns
Sr	0.61 ± 0.05	0.64 ± 0.06	0.59 ± 0.04	0.60 ± 0.07	6.20	−3.12	−0.53	ns
Mo	0.13 ± 0.01	0.13 ± 0.01	0.12 ± 0.01	0.14 ± 0.01	4.75	−3.23	5.58	ns
Cd	0.06 ± 0.005	0.06 ± 00006	0.05 ± 0.004	0.06 ± 0.004	−5.43	−8.86	1.02	ns
Pb	2.29 ± 0.31	2.13 ± 0.24	2.50 ± 0.34	2.09 ± 0.44	−6.94	9.12	−8.67	ns

Data represent the mean ± sd (n = 6). Data in bold represent elements with a significant difference (P ≤ 0.05) between the overexpression lines and KN5585 in three independent lines. ns, not significant. Statistical significance was determined by Student's t‐test.

Source data are available online for this Table.

Constitutive overexpression of ZmYSL2 also led to an increase in the Fe level in the kernels but to a lesser extent (Fig 6D and Table 7) while specific overexpression of ZmYSL2 in the BETL did not affect kernel Fe concentration (Fig 6B), suggesting that the increased Fe in the kernels of the constitutive overexpression transgenic lines was most likely due to ectopic effect of ZmYSL2. Consistent with this, the leaf Fe and Zn concentrations were also increased in the pUbi::ZmYSL2 transgenic lines but not in the pBETL9::ZmYSL2 transgenic lines (Fig EV5E–H). Interestingly, we did not observe any other phenotypic changes in the ZmYSL2 overexpression lines in addition to the changes in the Zn and Fe levels (Fig EV5I–K). These data strongly imply that ZmYSL2 has great potential to resolve the Zn‐deficiency problem associated with cereal‐based food.

Discussion

The transport of nutrients into seeds is an essential process for seed development and plant reproduction, as well as for human nutrition. It has been reported that AtHMA2 and AtHMA4 are responsible for the unloading of Zn from the maternal tissue in A. thaliana (Olsen et al, 2016), but how Zn is uploaded to the filial tissue in plants remains unclear. In this study, we revealed that ZmYSL2 is the major transporter responsible for loading Zn to filial tissues in maize, filling a major gap in the knowledge regarding Zn homeostasis in plants and broadening our understanding of the molecular mechanisms underlying nutrient delivery from the mother plant to the offspring (Fig 6G). In addition, ZmYSL2 is also the first characterized mineral nutrient transporter specifically expressed in BETL, which not only supports the essential function of BETL in nutrient transport but also provides a marker for BETL.

Our data and previous studies showed that ZmYSL2 functions as a Zn‐NA transporter (Zang et al, 2020; He et al, 2021), and the finding that ZmYSL2 dominates the mother‐to‐offspring transport of Zn thus suggests that Zn is transported into the seeds in the form of Zn‐NA complexes. However, AtHMA2 and AtHMA4 export the ionic form of Zn out of the maternal tissue in A. thaliana (Olsen et al, 2016). Therefore, if maize employs the same Zn unloading strategy as A. thaliana, there must exist a chelation process for forming Zn‐NA complexes at the maternal–filial interface. In this case, it would be interesting to study which one of HMAs and ZmYSL2 is the bottleneck for mother‐to‐offspring transport of Zn, as it could be used for further improving Zn biofortification. Otherwise, the mother‐to‐offspring delivery mechanism of Zn should be different between maize and A. thaliana. Consistent with the latter possibility, we recently found that impaired NA secretion in A. thaliana does not affect seed Zn levels (Chao et al, 2021), while other studies found that Zn‐NA is enriched in the phloem of cereal plants and is essential for Zn accumulation in the seeds of the cereal plants (Nishiyama et al, 2012). However, much more research should be conducted to prove or falsify these hypotheses.

ZmYSL2 also transports Fe‐NA in vivo (Zang et al, 2020; He et al, 2021). Surprisingly, the loss of function of ZmYSL2 did not affect the total Fe content in the kernel, and the weak allele zmysl2‐2 did not affect Fe distribution in the kernel. These data suggested that there is mechanistic divergence in the transport of Fe and Zn from the mother plant to the seeds. Consistent with this hypothesis, Fe was found to be mainly present in rice phloem sap in the form of Fe(III)‐DMA, while Zn was present mainly in the form of Zn‐NA (Nishiyama et al, 2012). In addition, DMA concentration rather than NA concentration is essential for Fe accumulation in cereal seeds (Diaz‐Benito et al, 2018; Chao & Chao, 2022). Furthermore, we found that ZmYSL2 is not able to transport Fe(III)‐DMA. These data together suggest that Fe might be loaded into the BETL in the form of Fe(III)‐DMA through some other YSLs. However, it is possible that Fe(II)‐NA is important for the distribution of Fe between embryo and endosperm, given that the null allele of ZmYSL2 may affect the total NA content in kernels. This hypothesis is worthy of further study to address the complexity of Fe homeostasis in seeds.

Previous studies claimed that the opaque phenotype of the kernels of ZmYSL2 mutants could be attributed to the disorder of Fe distribution in the kernels (Zang et al, 2020; He et al, 2021). This hypothesis was falsified by our identification of the weak allele zmysl2‐2, which produced opaque kernels but exhibited a normal Fe distribution. Instead, the opaque phenotype of the ZmYSL2 mutant kernels most likely resulted from a lack of Zn, given that Zn is required for starch and protein synthesis (Jyung & Ehmann, 1975; Broadley et al, 2007) and that all kernels with impaired ZmYSL2 were low in Zn. Fe is more likely to be required for embryo development, given that only the knockout mutants of ZmYSL2 were not viable. Consistent with this, a lack of Fe in the seeds of A. thaliana also caused severe defects in embryo development and seed germination (Klatte et al, 2009; Chu et al, 2010; Chao et al, 2021). This study thus also contributes to our understanding of the role of trace elements in seed development.

Breeding high‐Zn crops, which relies on the genetic diversity of the crops, is one of the most important strategies to achieve Zn biofortification. Although previous studies have uncovered the rich genetic diversity controlling the natural variation in the Zn content of maize kernels (Baxter et al, 2013; Maqbool & Beshir, 2019), none of the QTLs had been cloned. The power of GWAS allowed us to identify ZmYSL2 as the first successfully cloned QTL controlling natural variation in mineral nutrients in maize kernels, and these findings will aid the identification of additional ionome‐related QTLs in maize by using GWAS.

The polymorphisms of ZmYSL2 associated with the maize kernel's Zn content represent important and reliable molecular markers to assist in the breeding of high‐Zn maize varieties. These findings will also aid the engineering of high‐Zn maize via enhancement of the expression of ZmYSL2. Overexpression of ZmYSL2 driven by the Ubi promoter confirmed that this idea is feasible, as we have shown that this strategy was able to increase the Zn content in kernels to 31.6% higher than that in wild‐type kernels and 20% higher than the breeding goal of Zn biofortification for maize. In addition, overexpression of ZmYSL2 driven by the BETL‐specific promoter pBETL9 was also able to significantly enhance Zn accumulation in kernels, but the increase was slightly lower than that of the constitutive overexpression lines. These data suggest that overexpression of ZmYSL2 in the roots might also contribute to the increase of Zn in the kernels. However, it is hard to estimate how constitutive overexpression of ZmYSL2 contributes to the final increase in the kernel Zn as ectopic expression of ZmYSL2 might also lead to more distribution of Zn to other tissues such as leaves. Thus, it will be good to evaluate the effectiveness of more promoters in enhancing the expression of ZmYSL2 and Zn accumulation in kernels.

Material and Methods

Plant material and growth conditions

The two knockout mutants criysl2‐1 and criysl2‐2 were generated by CRISPR/Cas9 technology in the KN5585 background. The allele zmysl2‐1 was obtained from the UniformMu Transposon Resource. The allele zmysl2‐2 was obtained from EMS mutagenesis in the W64A background. The ZmYSL2 overexpression lines were obtained from transgenic events in the KN5585 background separately using pBETL9::ZmYSL2 ^B73 and pUbi::ZmYSL2 ^B73 platforms. The complete‐diallel design plus unbalanced breeding‐like inter‐cross (CUBIC) population (Liu et al, 2020b) derived from extensive intercrosses of 24 elite Chinese maize inbred lines was planted in a nursery field at Sanya, Hainan (2016, winter). All maize plants were planted in one‐row plots (2.5 m per row with 0.25 m spacing between plants and 0.75 m between rows) in each location. Field management followed standard procedures. The mutants and T₃‐generation overexpression lines were planted in a nursery field in Songjiang, Shanghai (2020 and 2021, summer).

Vector construction

To generate mutants of criysl2‐1 and criysl2‐2, two sgRNAs targeting ZmYSL2 (AAGGTGCGGGCAGATCATCC and AGCTGGTCATGCTGCTCTC) were inserted into the binary vector pCPB‐ZmUbi::hSpCas9 vector (Li et al, 2017). To construct the ZmYSL2 overexpression vectors, the coding region of ZmYSL2 was amplified from endosperm cDNA in B73 and inserted into the proZmUbi::eYFP vector to generate the pUbi::ZmYSL2 ^B73 constructs. To construct the expression vectors pBETL9::ZmYSL2 ^B73 , CDS fragment of ZmYSL2 was amplified from B73 cDNA using the primers pBETL9‐ZmYSL2‐ov‐F/pTF102‐ZmYSL2‐R. The 2 kb fragment of ZmBETL9 promoter was amplified from B73 genomic DNA using the primers pTF102‐pBETL9‐F/pBETL9‐ZmYSL2‐ov‐R. Thereafter, two fragments were fused by overlapping PCR with the primers pTF102‐pBETL9‐F and pTF102‐ZmYSL2‐R. The fused fragment was inserted into the cloning site between Spe I and Pst I of the binary expression vector pTF102‐p27 by using the Hieff Clone one‐step PCR Cloning kit (Yisheng Co. Ltd, Shanghai, China). Both the gene‐editing and overexpression constructs were transferred to the Agrobacterium strain EHA105 and transformed into the immature embryo of the maize inbred line KN5585 (Liu et al, 2020a) through Agrobacterium‐mediated transformation by WIMI Biotechnology Co., Ltd.

For the construction of the yeast expression vector pDR196‐ZmYSL2 for different genotypes, the coding region of ZmYSL2 was amplified from endosperm cDNA in HuangC, HZS, Lv28, YFH, W64A, or zmysl2‐2 by using the primers pDR196‐ZmYSL2‐F and pDR196‐ZmYSL2‐R. For the construction of the yeast expression vector pDR196‐OsYSL15, the coding region of OsYSL15 was amplified from cDNA in Oryza sativa cv. Nipponbare by using the primers pDR196‐OsYSL15‐F and pDR196‐OsYSL15‐R. For the construction of the yeast expression vector pDR196‐AtIRT1, the coding region of AtIRT1 was amplified from the cDNA of Col‐0 by using the primers pDR196‐AtIRT1‐F and pDR196‐AtIRT1‐R. For the construction of the yeast expression vector pDR196‐ZmYS1, the coding region of ZmYS1 was amplified from the cDNA of B73 by using the primers pDR196‐ZmYS1‐F and pDR196‐ZmYS1‐R. Each amplified fragment was cloned into the cloning site between Spe I and Xho I of the yeast expression vector pDR196.

To construct the transient expression vectors of pA7‐ZmYSL2‐GFP, the coding regions of ZmYSL2 without the stop codon from B73 were amplified and inserted into the pA7‐GFP vector. The primers used are listed in Table 8.

Table 8.

Primers used in this study.

Use	Primer name	Primer sequence (5′‐3′)
qRT for ZmYSL2	qZmYSL2‐F	GCCAGCTTCGCCATCGACAT
	qZmYSL2‐R	CTAGCTTCCAGGTGTAAATTTC
qRT for Zm00001d044172	W130‐F	CGTCCACCCATAGTGCCATGAG
	W130‐R	ACGGCAAGTAGCACTCAGACAC
Haplotype analysis	Hap1‐ProZmYSL2‐F	TAAATCGTCTCCGACCCACAG
	Hap2‐ProZmYSL2‐F	TAACGGCTCATCCACTTCCG
	Hap3‐ProZmYSL2‐F	CTCCCAAACCGGCGATACTT
	Hap4‐ProZmYSL2‐F	ACCCCTTTCGGCCTCTAGAA
	Hap‐ProZmYSL2‐R	GAGCAAGAACTACTATCAATCAG
ZmYSL2 RNA probes construct	ZmYSL2‐anti‐F	ATGCCGAGGACCACAACTCCAC
	ZmYSL2‐anti‐R	ATACGACTCACTATAGGGCGAACGGCGACGCCCAGGCGGGACA
pA7‐ZmYSL2 construct	pA7‐ZmYSL2‐F	TTACGAACGATACTCGAGATGCCGAGGACCACAAC
	pA7‐ZmYSL2‐R	CCATCACTAGTACGTCGACGAAATTTACACCTGGAAGC
pDR196‐ZmYSL2 construct	pDR196‐ZmYSL2‐F	ATCCCCCGGGCTGCAGGAATTCATGCCGAGGACCACAAC
	pDR196‐ZmYSL2‐R	GGTACCGGGCCCCCCCTCGAGCTAGCTTCCAGGTGTAAATTTC
pDR196‐OsYSL15 construct	pDR196‐OsYSL15‐F	ATCCCCCGGGCTGCAGGAATTCATGGAGCACGCCGACGCGGAC
	pDR196‐OsYSL15‐R	GGTACCGGGCCCCCCCTCGAGTTAGCTTCCAGGCGTAAACTTCA
pDR196‐AtIRT1 construct	pDR196‐AtIRT1‐F	ATCCCCCGGGCTGCAGGAATTCATGGCTTCAAATTCAGCACT
	pDR196‐AtIRT1‐R	GGTACCGGGCCCCCCCTCGAGTTAAGCCCATTTGGCGATAATC
pDR196‐ZmYS1 construct	pDR196‐ZmYS1‐F	ATCCCCCGGGCTGCAGGAATTCATGGACCTTGCACGGAGCG
	pDR196‐ZmYS1‐R	GGTACCGGGCCCCCCCTCGAGCTAGCTTCCAGGAGTGAAC
pTF102‐pBETL9::ZmYSL2 construct	pTF102‐pBETL9‐F	AGCAATGCACGGGCATATAACTAGTACACGCCCATGAAACCATGA
	pBETL9‐ZmYSL2‐ov‐F	CGTCAACAGTTGAAGTTATACCCATGCCGAGGACCACAAC
	pBETL9‐ZmYSL2‐ov‐R	GTTGTGGTCCTCGGCATGGGTATAACTTCAACTGTTGACG
	pTF102‐ZmYSL2‐R	GGTGATTTTTGCGCCGCTGCACTAGCTTCCAGGTGTAAATTTC

Elemental profiling

The elemental analysis was conducted by inductively coupled plasma mass spectrometry (ICP‐MS) as previously described (Danku et al, 2013). For mature kernels of CUBIC individuals, 10 dry mature kernels of each maize line were ground into powder by a customized grinder with a grinding tank (material: polytetrafluoroethylene) and beads (material: Jargonia) manufactured by Jingxin Industry Co., Ltd., Shanghai. For mature kernels of mutants and overexpression lines, after harvest, kernels were detached from the same or different ears and grouped based on seed phenotype and genotype, and at least six kernels were ground into powder using the method described above. Then, 2–5 mg of each sample was used for ICP‐MS analysis. To investigate the element changes that occurred during kernel development, each fresh kernel (20 days after pollination [DAP]) was sampled for weighing and then dried for ICP‐MS analysis. To detect the ionome in different parts of 20 DAP seeds, genotyped fresh kernels of each genotype at 20 DAP were first dissected into different parts, including the whole kernel, embryo, endosperm, and maternal tissue, and then weighed, followed by ICP‐MS analysis after drying. The maternal tissue indicates an area in the kernel excluding the embryo and endosperm that is close to BETL.

The powdered samples were placed into Pyrex test tubes (16 × 100 mm) to dry in an oven at 65°C for 24 h. After weighing 12 samples (these weights were used to calculate the weights of the rest of the samples), Primar Plus nitric acid for trance metal analysis (Fisher Chemicals) spiked with an indium internal standard was added to the tubes (1 ml per tube). The samples were then digested in a block heater (DigiPREP MS, SCP Science; QMX Laboratories, Essex, UK) at 115°C for 5 h. The digested samples were diluted to 10 ml with 18.2 MΩ Milli‐Q Direct water (Merck Millipore). Elemental analysis was performed using ICP‐MS (NexION 350D; PerkinElmer, USA) coupled with an Apex desolvation system and an SC‐4 DX autosampler (Elemental Scientific Inc., USA). All the solid samples were normalized with a heuristic algorithm using the best‐measured elements as previously described (Lahner et al, 2003).

For each set of ICP‐MS measurements of CUBIC individuals, the median value from each run was used to correct run‐to‐run variations. The corrected values were then log2 transformed for genome‐wide association analysis.

Genome‐wide association study

A genome‐wide association study (GWAS) was performed using 10.8 million SNPs (MAF ≥ 0.05) in the CUBIC population. The same relatedness matrix and the top 10 PCs used in the previous study (Liu et al, 2020b) accounting for the familial relationship (K) and population structure (Q) were routinely fitted into the mixed linear model (MLM) of TASSEL 5.0 (Bradbury et al, 2007). The significance threshold for the association was determined to be 9.3 e‐08 according to the Bonferroni correction (P = 1/n; n = total markers used for GWAS).

Functional complementation assay in yeast

The pDR196‐ZmYSL2 vector for different genotypes, including HuangC, HZS, Lv28, and YFH, and pDR196‐OsYSL15 as well as the pDR196 empty vector were introduced into the zrt1zrt2 (MATa, his3, leu2, met1, lys2, ura3, zrt1, zrt2) yeast strain. The pDR196‐ZmYSL2 vector for different genotypes, including W64A and zmysl2‐2 and pDR196‐AtIRT1, pDR196‐ZmYS1, as well as the pDR196 empty vector were introduced into the fet3fet4 (DEY1453, MATα, ade2, can1, his3, leu2, trp1, ura3, fet3::HIS3, fet4::LEU2) yeast strain. Transformed yeast cells were diluted from an optical density of 1 to 0.001 at 550 nm and spotted onto synthetic defined (SD)‐Ura medium, in which YDB with low Zn and Fe was gotten from USBiological (L18022001). Chemically synthesized NA was purchased from Cayman Chemical, USA, and the DMA was kindly provided by Luqing Zheng. For the zrt1zrt2 yeast strain, the medium was supplemented with 10 μM ZnSO₄ or 10 μM NA‐ZnSO₄ at pH 7. For the fet3fet4 yeast strain, the medium was supplemented with 0 μM FeSO₄, 10 μM FeSO₄, 10 μM NA‐FeSO₄, 10 μM DMA‐FeCl₃, and 100 μM FeSO₄ at pH 7. NA‐ZnSO₄ was prepared according to the following protocol: 11 μl of 10 mM NA or DMA and 10 μl of 10 mM ZnSO₄ or FeSO₄ were mixed and incubated for 10 min at 60°C, and then, 10 ml of SD‐Ura medium was added to the plates (von Wiren et al, 1999). NA‐FeSO₄ was prepared by mixing 10 mM FeSO₄ in 20 mM sodium ascorbate with 5 mM NA solution and 200 mM MES/Tris buffer (pH 7.0) to produce 0.5 mM Fe, 2.5 mM NA, and 50 mM MES/Tris buffer, followed by 10 min incubation at 65°C. DMA‐FeCl₃ was prepared by mixing 10 mM FeCl₃ in 10 mM HCl with 5 mM DMA and 200 mM MES/Tris buffer (pH 7.0) to produce 0.5 mM Fe, 2.5 mM DMA, and 50 mM MES/Tris buffer, followed by 3 h incubation at room temperature.

Micro‐X‐ray fluorescence visualization

The Zn and Fe distribution in the kernels of different maize genotypes was obtained using a microenergy‐dispersive X‐ray fluorescence (mEDXRF) system (M4 Tornado; Bruker, Germany). For mature kernels, seeds were cut in half by using a ceramic knife; half of the kernels were used for genotyping, and the other half were used for SXAF analysis. For immature kernels, fresh seeds were cut in half immediately by using a ceramic knife; half of the kernels were used for genotyping, and the other half were used for SXAF analysis within 10 min. Zn and Fe fluorescence was observed under Rh X‐ray tube illumination. The X‐ray generator was operated at 50 kV and 600 mA. For better visualization, a set of filters between the X‐ray tube and the sample were used, which was composed of three foils of Al/Ti/Cu with thicknesses of 100/50/25 mm. The elemental mappings were performed with a pixel spacing of 30 μm to completely cover the kernel, with a measurement time of 2 ms per pixel.

For SXAF image larger than 14 mm × 11 mm, a mosaic image was recorded by M4 TORMADO. Like a microscope, the M4 TORNADO applies a simple imaging system, in which the position of the X‐ray tube and camera is stationary, and the X–Y–Z stage can move to enable the exact focusing and positioning of the sample required by the small spot of the excitation radiation. A 10× magnification camera with a field view of 14 mm × 11 mm was used for the overview of the sample. When the sample's size exceeds the field of view of the 10× camera, for ease of navigation and overview, the M4 TORNADO allows to create a so‐called mosaic image, which can cover an area as large as the stage area.

RT–qPCR analysis

The roots, leaves, stems, silks, tassels, seeds, endosperms, and embryos from B73 plants were collected and flash‐frozen in liquid nitrogen. The roots, leaves, stems, and nodes were collected at the L10 stage. Silks and tassels were collected at the plant flowering stage. The seeds, endosperm, and embryos were collected on different days after pollination. Total RNA was extracted by TRIzol reagent (Invitrogen, catalog number 15596018) and then purified with an RNeasy Mini Kit (Qiagen, catalog number 74106) after DNaseI digestion (Qiagen, catalog number 79254) following the manufacturer's protocol. For RT–qPCR analysis of ZmYSL2 expression, digested RNA was used for reverse transcription with a Super Script III First Strand Kit (Invitrogen). The resulting cDNA was diluted to 50 ng μl⁻¹ for RT–qPCR with SYBR Green (TOYOBO) on a CFX Connect Real‐Time System (Bio‐Rad). The maize gene (Zm00001d044172) was used as the internal control (Wang et al, 2018). The relative expression of the gene was calculated by the $2^{-\mathrm{\Delta }{C}_{\mathrm{t}}}$ method. All the experiments were performed with three independent RNA samples. Primers are listed in Table 8.

Subcellular localization of ZmYSL2

For transient expression of ZmYSL2 in Arabidopsis protoplasts, vectors were transferred into protoplasts of Col‐0 following previously described methods (Wu et al, 2009). At 36 h after incubation, fluorescence signals were recorded under a Leica TCS SP8 confocal laser scanning microscope. To observe the GFP signals, an excitation wavelength of 488 nm from an argon laser was used, and the emission signal was detected at 500–550 nm.

RNA in situ hybridization

RNA in situ hybridization for ZmYSL2 was performed following the method described previously (Zhang et al, 2015) with modifications. At 16 DAP, seeds from the middle region of the KN5585 ears were harvested and fixed in 4% paraformaldehyde solution with 0.1% Triton X‐100 and 0.1% Tween 20 in phosphate‐buffered saline (PBS) (Takara, Cat# T900) overnight. The seeds were then dehydrated using a gradient concentration of ethanol and embedded in paraffin. A 400‐bp cDNA fragment of ZmYSL2 was amplified using the primer pair ZmYSL2‐anti‐F and ZmYSL2‐anti‐R. The antisense and sense RNA probes were synthesized according to the instructions for the DIG RNA Labeling Mixture (Roche Cat# 11175025910). Ten‐micrometer seed sections were cut and then hybridized with the RNA probes at 50°C overnight. After blotting with anti‐digoxigenin AP‐conjugated antibody (Roche Cat# 11093274910) and incubation with NBT solution (Roche Cat# 11383213001), the sections were observed and photographed with an ECLIPSE 80i microscope.

Measurement of the proteins, starch, and phytic acid

Approximately 10 kernels were selected from the middle of each ear that had their embryo, seed coat, and aleurone removed. The resulting endosperm was dried in an oven at 60°C and then ground into fine powder by a customized grinder. The extraction and analysis of zein and non‐zein proteins were conducted as previously described (Liu et al, 2016). For assaying the total protein content, 60 mg of endosperm powder was used for measurement by an ELEMENTAR Rapid N exceed instrument. The endosperm powder was filtered through an 80‐mesh sieve to prepare samples for starch measurement. The starch content was measured with a Total Starch Assay kit (K‐TSTA; Megazyme) following the standard protocol (Deng et al, 2020). Three biological replicates were performed. The phytic acid contents were measured by using a Total Phytic Acid Assay kit (K‐PHYT; Megazyme) following the protocol provided by the manufacturer.

Immunostaining of ZmYSL2

ZmYSL2 antibodies were produced by ABclonal. The cDNA fragment encoding the partial ZmYSL2 protein from amino acids 1–90 was cloned into the prokaryotic expression vector pET‐28a‐ZmYSL2. Purified protein was injected into rabbits to produce ZmYSL2 antibodies. After affinity purification, the antibodies were used for immunofluorescence histochemistry.

Immunostaining of ZmYSL2 was performed following the method described previously (Huang et al, 2019) with modifications. In brief, hand‐cut sections of 16 DAP seeds in KN5585 and criysl2‐1 were fixed for 1 h in freshly prepared 4% (v/v) formaldehyde in PHEM buffer (60 mM PIPES, 25 mM HEPES, 10 mM EGTA, and 2 mM MgCl₂; pH 6.9) with 0.02% (v/v) Triton X‐100 and 5% (v/v) DMSO. After being rinsed several times in PHEM buffer, the samples were digested in an enzyme solution containing 1% (w/v) β‐glucuronidase (LabTop, catalog number S10046), 0.1% (w/v) cellulose (Yakult, catalog number 9012‐54‐8), 0.1% (w/v) pectinase (LabTop, catalog number S10007), 0.1% (w/v) lyticase (Yeasen, catalog number 10403ES92), and 1% (w/v) glucose at 37°C for 30 min to remove cell walls, followed by washing with PHEM buffer. Samples were put into PHEM buffer with 1% Triton X‐100, 5% DMSO, and 3% BSA (w/v) at 37°C for 30 min. After washing again with PHEM buffer, the samples were placed in PBS (10 mM Na₂HPO₄, 1.8 mM KH₂PO₄, 2.7 mM KCl, and 137 mM NaCl; pH 7.4) with anti‐ZmYSL2 antibody (1:200) and incubated overnight in the dark at 4°C. Then, the samples were washed five times with PBS and incubated with Alexa Fluor 555‐conjugated goat anti‐mouse IgG secondary antibody at a dilution of 1:300 (Abmart, catalog number M213411M) in PBS for 1 h at 37°C. After washing five times in PBS, the samples were then observed using a confocal laser scanning microscope (TSC SP8 STED 3X, Leica). To observe the secondary antibody signals, an excitation wavelength of 555 nm from an argon laser was used, and the emission signal was detected at 565–590 nm.

Author contributions

Zhen‐Fei Chao: Conceptualization; resources; data curation; software; formal analysis; validation; investigation; visualization; methodology; writing—original draft. Yuan‐Yuan Chen: Conceptualization; resources; data curation; software; formal analysis; validation; investigation; visualization; methodology. Chen Ji: Conceptualization; data curation; formal analysis; validation; investigation; methodology. Ya‐Ling Wang: Data curation; software; validation; methodology. Xing Huang: Data curation; formal analysis; investigation; methodology. Chu‐Ying Zhang: Investigation; methodology. jun Yang: Investigation; methodology. Tao Song: Investigation; methodology. Jia‐Chen Wu: Investigation; methodology. Liang‐XIng Guo: Investigation; methodology. Chu‐Bin Liu: Software; investigation. Mei‐Ling Han: Data curation; investigation. Yong‐Rui Wu: Conceptualization; supervision; writing—review and editing. Jianbing Yan: Conceptualization; supervision; writing—review and editing. Dai‐Yin Chao: Conceptualization; data curation; formal analysis; supervision; funding acquisition; investigation; writing—review and editing.

Disclosure and competing interests statement

The authors declare that they have no conflict of interest.

Supporting information

Appendix

Expanded View Figures PDF

Source Data for Tables

Source Data for Expanded View and Appendix

PDF+

Source Data for Figure 1

Source Data for Figure 4

Source Data for Figure 5

Source Data for Figure 6

EMBO Reports (2023) 24: e55542.

Data availability

All data are present either in the main study or the supplementary materials. This study includes no data deposited in external repositories.