Engineering a Gibberellin‐Switchable Dual‐Use Line Based on Ent‐Kaurene Oxidase Gene ZmKO1 Enables Two‐Line Hybrid Seed Production in Maize

ABSTRACT

Male‐sterile systems are pivotal for maize hybrid seed production, yet existing systems are constrained by significant limitations. Here, we identified a novel single‐gene‐controlled recessive genic male sterility (GMS) mutant, defective in filament elongation 1 (def1), which exhibits non‐elongated filaments, unopened glumes, and non‐exserted anthers, yet undergoes normal microsporogenesis and produces viable pollen. Map‐based cloning revealed that def1 is caused by a loss‐of‐function mutation in the ZmKO1 gene. Previous studies have identified ZmKO1 and ZmKO2 as the two putative KO‐encoding genes in maize. Although heterologous expression evidence suggested that only ZmKO1 possesses KO activity, we demonstrate that both ZmKO1 and ZmKO2 are functional KOs for GA biosynthesis. The def1 mutant or ZmKO1 knockout lines exhibited reduced plant height during early seedling development but recovered rapidly to wild‐type levels. In contrast, ZmKO2 knockout lines displayed severe dwarfism and pollen‐less male sterility. These contrasting phenotypes demonstrate a clear functional divergence between the two genes. Utilising marker‐assisted backcrossing and CRISPR/Cas9 editing, we generated ZmKO1‐based GMS lines across diverse genetic backgrounds. These lines maintained stable male sterility and could be efficiently propagated via exogenous GA₃ application under different genetic backgrounds and environments, demonstrating robust “dual‐use” properties. Data indicated that this dual‐use trait has no significant effect on other agronomic traits of the male sterile line and its derived hybrid combinations. Based on this, we established a novel two‐line hybrid seed production system in maize, termed the Gibberellin‐switchable Dual‐use Line (GDL), and successfully applied it in the hybrid seed production of Chuandan 99.

1. Introduction

Heterosis breeding, as an effective means of enhancing crop productivity, has been widely adopted in agriculture. Its commercial success depends on scalable and efficient hybrid seed production. Maize, being monoecious with separate male and female flowers, became a pioneer in hybrid crop commercialization in the 1920s, largely due to the relative ease of its hybrid seed production system (Duvick 2001). To enforce cross‐pollination during maize hybrid seed production, several methods are employed to prevent self‐pollination in female parents. These include detasseling, the application of male‐specific gametocides, and the utilization of male sterility (Perez‐Prat and van Lookeren Campagne 2002; Wan et al. 2019). Among these, detasseling remains the industry standard for maize hybrid seed production. However, it is not only labor‐intensive and costly but also poses safety risks for workers required to operate in challenging field conditions. Furthermore, it impairs both seed yield, by inflicting physical damage on female plants, and seed purity, by permitting self‐pollination due to incomplete emasculation (Feng et al. 2014). The practical application of chemical gametocides remains limited owing to inconsistent pollen sterilization, negative effects on female fertility, and environmental safety concerns (McRae 1985). By eliminating the need for detasselling, the use of male‐sterile lines as female parents reduces labour and production costs and ensures high seed purity, thus representing an optimal strategy for large‐scale hybrid maize production (Wan et al. 2019).

Plant male sterility is broadly categorised as cytoplasmic male sterility (CMS) or genic male sterility (GMS) based on their inheritance patterns. Although CMS lines have been utilised in hybrid maize breeding, their application remains constrained by limitations such as narrow genetic diversity, pathogen susceptibility, and unstable sterility/fertility restoration (Williams 1995; Weider et al. 2009). In contrast, most GMS lines, controlled by recessive nuclear genes, offer stable sterility and theoretically universal restorers, making them theoretically ideal for hybrid seed production. However, this same nuclear inheritance poses a major obstacle to the efficient, large‐scale propagation of pure male‐sterile lines (Wan et al. 2019).

Advances in biotechnology have led to innovative systems to overcome this limitation. The Roundup Hybridization System (RHS), developed by Monsanto (Feng et al. 2014), utilises glyphosate‐induced male sterility for hybrid seed production. The system employs two transgenic constructs, RHS and RR, which confer distinct spatiotemporal expression patterns of the glyphosate tolerant CP4 EPSPS gene. The RHS construct in the female inbred line drives weak CP4 EPSPS expression in the tapetum and microspores but strong expression in other tissues. In contrast, the RR construct in the male inbred line provides high constitutive expression throughout the plant. In hybrid seed production fields, interplanting rows of RHS females with RR males enables over‐the‐top spray of glyphosate to induce male sterility in RHS females, allowing subsequent pollination by RR males. Without glyphosate treatment, the RHS females maintain full fertility and can be propagated on a large scale through self‐pollination. However, the system's necessity of glyphosate application across vast areas, combined with the transgenic nature of its hybrid seeds, has prevented its broad commercial use. Based on the need for non‐transgenic hybrid seed production, Wu et al. (2016) developed the Seed Production Technology (SPT), which utilises a transgenic maintainer line to propagate non‐transgenic male‐sterile lines. However, the scalable application of SPT is constrained by pollen‐mediated transgene transmission rates ranging from 0.002% to 0.518%, depending on the specific construct, transformation event, and genetic background (Wu et al. 2016). To mitigate this risk, Zhang et al. (2018) engineered a Multi‐Control Sterility (MCS) system that reduces the transgene escape frequency to approximately 1/8–1/7 of the SPT level. Nevertheless, complete prevention of transgene flow remains an unresolved challenge.

Another effective strategy for large‐scale propagation of GMS is the use of dual‐use lines. A prominent example is the two‐line hybrid rice breeding programs based on the application of environment‐sensitive genic male sterility (EGMS) dual‐use lines. These lines remain sterile under specific conditions for hybrid seed production but regain fertility under different conditions to enable self‐pollination and propagation (Xie et al. 2019). The scarcity of elite dual‐use germplasm has limited the adoption of two‐line hybrid breeding in maize. In an effort to overcome this barrier, Li et al. (2017) engineered a maize temperature‐sensitive genic male sterile (TGMS) line by targeting the ZmTMS5 gene with CRISPR/Cas9. The tms5 mutants exhibit sterility at 32°C and fertility at 24°C, with a transitional phase near 28°C. However, the commercial application of this system remains limited, primarily owing to the risks associated with unpredictable temperature fluctuations under field conditions. Therefore, developing novel dual‐use male‐sterile germplasms is pivotal for innovating maize hybrid seed production systems.

In this study, we identified a novel recessive GMS mutant, defective in filament elongation 1 (def1), from space‐induced mutation in maize. The def1 mutant exhibits male sterility due to suppressed filament elongation, failure of spikelet opening, and non‐exserted anthers, yet produces morphologically normal and viable pollen. Map‐based cloning revealed that the def1 phenotype is caused by a mutation in ZmKO1 (Zm00001eb385100). The ent‐kaurene oxidase (KO) catalyses the oxidation of ent‐kaurene to ent‐kaurenoic acid, representing a critical step in the gibberellin (GA) biosynthesis pathway (Salazar‐Cerezo et al. 2018). Mutations affecting GA biosynthesis (e.g., ga1‐3) can lead to male sterility due to the arrest of stamen filament cell elongation in Arabidopsis (Cheng et al. 2004). Through homology‐based sequence analysis, Mao et al. (2017) identified two putative KO‐encoding genes in maize, ZmKO1 and ZmKO2 (Zm00001eb385070). Heterologous expression in E. coli suggested that ZmKO1 exhibits KO activity, whereas ZmKO2 appears to lack this function. Our study establishes that both ZmKO1 and ZmKO2 are functional KOs in the maize GA pathway. A key finding is that the male sterility of def1 can be completely reversed by exogenous bioactive GAs, permitting the large‐scale propagation of this GMS line. Capitalising on this dual‐use trait (male sterility under natural conditions and fertility recovery upon GA treatment), we established a transgene‐free, cost‐effective, environmentally friendly, and readily applicable two‐line hybrid seed production system in maize, designated as Gibberellin‐switchable Dual‐use Line (GDL), and successfully applied it to produce the hybrid cultivar Chuandan 99.

2. Materials and Methods

2.1. Plant Materials

The def1 mutant originated from a space‐mutagenized population of the maize inbred line RP125 and has since been maintained by sib‐mating. Homozygous fertile siblings from the same ear were used as wild‐type (WT) controls. The EMS‐316919 mutant (EMS‐mutagenized B73 background) was obtained from the Maize EMS Database (Lu et al., 2017). The other inbred lines were derived from the germplasm resources collected by the Maize Research Institute of Sichuan Agricultural University.

2.2. Paraffin Sections

Tassel spikelets from WT and def1 plants at different developmental stages were fixed in FAA solution (3.7% formaldehyde, 5% acetic acid, 50% ethanol) at 4°C for 5 days. The fixed samples were dehydrated through a graded ethanol series. Ethanol concentrations and treatment durations were as follows: 30% (1 h), 50% (1 h), 75% (2 h), 85% (1.5 h), 90% (1.5 h), 95% (1.5 h), 100% (1 h), 100% (1 h), followed by clearing in three changes of xylene (8, 20, and 20 min, respectively). Subsequently, the samples were infiltrated with three changes of paraffin (0.5, 1, and 2 h) and embedded in paraffin blocks. Sections of 5 μm thickness were cut using a microtome (RM2016, Leica), dewaxed in xylene, stained with toluidine blue, and examined under a light microscope (DM2000 LED, Leica).

2.3. Evaluation of Pollen Viability

For pollen staining assessment, 200 μL of either 1% I₂‐KI or Alexander staining solution was pipetted onto a glass slide. For in vitro germination assay, 1.5 mL of pollen germination medium (containing 10% sucrose, 0.01% boric acid, 0.1% yeast extract, 10 mM CaCl₂, 50 μM KH₂PO₄, and 15% PEG4000) was dispensed into a 3‐cm Petri dish. At anthesis stage, anthers of unflowered male spikelets were placed into the staining solution or medium and gently crushed with forceps to release pollen. After GA₃ treated, freshly shed pollen grains were directly transferred into the solution or medium. After staining for 10 min, a coverslip was applied, and the samples were observed and photographed under a stereo microscope (XTL‐165‐MT, Phenix). For germination assays, the germination medium containing pollen was incubated in darkness at room temperature (~22°C) for 4 h, after which germination was assessed and photographed under a stereo microscope.

2.4. Genetic Analysis and Molecular Marker Mapping

The def1 mutant was crossed as the female parent with the inbred lines B73, Mo17, and Huangzaosi. The resulting F₁ hybrids were self‐pollinated to generate three F₂ populations, which were grown in Chengdu (Sichuan) for genetic analysis based on field fertility records. The (def1 × Mo17) F₂ population was used for the molecular marker mapping. Genomic DNA (gDNA) was extracted from leaves using the CTAB method. The gDNA from def1, Mo17, and their F₁ hybrid was used to screen for co‐dominant Simple Sequence Repeat (SSR) markers. Bulked segregant analysis (BSA) was conducted by pooling equal amounts of leaf tissue from 10 sterile and 10 fertile (def1 × Mo17) F₂ plants to construct sterile and fertile DNA bulks, respectively. The co‐dominant SSR markers were used to amplify gDNA from the two bulks, Mo17, and def1 to identify markers linked to the male sterility. Subsequently, InDel and SNP markers were developed by sequencing ~900 bp genomic PCR products for fine mapping. All polymorphic markers are listed in Table S1.

2.5. RNA Extraction and qRT‐PCR Analysis

Tissues were snap‐frozen in liquid nitrogen and stored at −80°C. Total RNA was extracted with the RNAprep Pure Plant Kit (TIANGEN, China). RNA integrity and concentration were verified by 1.5% agarose gel electrophoresis and NanoDrop 2000 (Thermo Fisher Scientific, USA). First‐strand cDNA was synthesized using the HiScript III 1st Strand cDNA Synthesis Kit (+gDNA wiper) (Vazyme, China). gDNA contamination was assessed by PCR with primers (qPCR‐Act1) flanking a 107‐bp intron of the maize Actin1 (Zm00001d010159), which yield amplicons of 298 bp from gDNA and 191 bp from cDNA. qRT‐PCR was carried out on a CFX96 Real‐Time System (Bio‐Rad, USA) with Taq Pro Universal SYBR qPCR Master Mix (Vazyme, China). Actin1 was used as the internal control. Three biological replicates, each with three technical replicates, were analyzed for each sample. Relative expression levels were calculated by the 2^(‐ΔΔCt) method. Primers used are listed in Table S1.

2.6. Comparative Transcriptome Analysis

When WT plants reached the anthesis stage, a subset of def1 individuals were treated with exogenous GA₃ spraying. On the following day, freshly opened male spikelets from the main axis were simultaneously collected from WT, def1 + GA₃ (GA₃‐treated def1 plants), and untreated def1 plants for transcriptome sequencing. RNA‐Seq libraries were constructed using the TruSeq Standard RNA LT Sample Preparation Kit (Illumina, USA) and sequenced on an Illumina NovaSeq 6000 platform. Raw reads were quality‐checked with FastQC (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/) and processed with Trimmomatic for adapter trimming and quality filtering (Bolger et al. 2014). Clean reads were aligned to the maize reference genome (Zm‐B73‐REFERENCE‐GRAMENE‐4.0) using HISAT2 (Kim et al. 2019). Gene expression was quantified as TPM (transcripts per kilobase million), and differentially expressed genes (DEGs) (p < 0.05, |fold change| ≥ 2) were identified with edgeR in default parameters (Chen et al. 2016). Gene IDs of DEGs in Zm‐B73‐REFERENCE‐GRAMENE‐4.0 model were translated to the IDs in Zm‐B73‐REFERENCE‐NAM‐5.0 for gene ontology (GO) analysis and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses using OmicShare Tools (Mu et al. 2024).

2.7. Gene Functional Validation

For functional complementation, the Ubi‐1 promoter was amplified from B73 gDNA using primers ProUbi and cloned into a SacI‐linearized pCAMBIA3301 vector using the ClonExpress II One Step Cloning Kit (Vazyme, China), yielding the intermediate construct 3301‐ProUbi. The coding sequence (CDS) of ZmKO1 was amplified from the leaf cDNA of WT using primers KO1OE and inserted into a SmaI‐digested 3301‐ProUbi vector, producing the final expression construct ProUbi::ZmKO1CDS (Figure S1a). The construct was introduced into maize inbred line B104 via A. tumefaciens mediated transformation by Edgene Bio (Wuhan, China). Transgenic T₀ plants were identified by genomic PCR with primers OEcheck. PCR‐positive T₀ plants were crossed as male parents with the def1 mutant. PCR‐positive F₁ plants were self‐pollinated to generate the F₂ populations, which were subsequently genotyped and phenotyped. Primers used are listed in Table S1.

CRISPR‐P was used for single‐guide RNA (sgRNA) design of CRISPR‐system in plants (Lei et al. 2014). For ZmKO1 knockout, primers KO1GE were designed with the forward primer containing the sgRNA (KO1sgRNA) targeting the exon 2 of ZmKO1. A fusion fragment comprising KO1sgRNA and the gRNA scaffold was amplified by PCR using the modified CRISPR/Cas9 vector pCrisprV1 as template. This fragment was cloned into a BamHI/HindIII‐linearized pCrisprV1, generating the final gene editing construct designated pCrisprV1‐KO1 (Figure S1b). For ZmKO2 knockout, primers KO2GE1 were designed with an sgRNA (KO2sgRNA1) targeting exon 4, and the intermediate vector pCrisprV1‐KO2m1 was constructed. The U6 promoter was amplified from pCrisprV1 with primers ProU6 and assembled into a HindIII‐linearized pCrisprV1‐KO2m1 to generate pCrisprV1‐KO2m2. Primers KO2GE2 were designed with another sgRNA (KO2sgRNA2) targeting exon 5, resulting in the dual‐target gene editing construct pCrisprV1‐KO2 (Figure S1c). The pCrisprV1‐KO1 and pCrisprV1‐KO2 constructs were transformed into B104. Transgenic T₀ plants were screened by genomic PCR with Cas9check primers. Positive plants were backcrossed with B104 to generate BC₁F₁ populations. Non‐transgenic edited plants were selected from BC₁F₁ using Cas9check primers, and mutations in ZmKO1 and ZmKO2 were verified by sequencing with primers KO1site and KO2site, respectively. Confirmed edited plants were self‐pollinated to generate BC₁F₂ populations for subsequent genotyping and phenotypic analysis.

Heterozygous mutants (ZmKO1 ^G/A ) from the EMS mutant line EMS5‐316919 were identified by genomic PCR and sequencing using primers EMS919. Self‐pollination of heterozygous mutants segregated wild‐type (ZmKO1 ^G/G ), heterozygous mutants, and homozygous mutants (ZmKO1 ^A/A ) progeny. Primers used were listed in Table S1.

2.8. Subcellular Localization

The CDS of ZmKO1 (stop codon excluded) was amplified with primers KO1Sub and cloned into a KpnI‐linearized pCAMBIA2300 vector, generating the 35S::ZmKO1‐eGFP construct for subcellular localization. The empty 35S::eGFP construct was used as a negative control. 35S::ZmKO1‐eGFP and 35S::eGFP constructs were introduced into A. tumefaciens strain GV3101 and transiently expressed in N. benthamiana leaf epidermal cells via agroinfiltration. After 48 h of incubation at 25°C, fluorescence signals were visualised with a LSM800 confocal microscope (Zeiss, Germany). Primers used are listed in Table S1.

2.9. Detection of GUS Activity

A 1978‐bp promoter region upstream of the start codon of ZmKO1 was amplified with primers KO1P‐gus and cloned into an EcoRI/NcoI‐linearized pCAMBIA3301 vector, yielding the ZmKO1Pro::GUS reporter construct. The construct was introduced into the inbred line B73 (performed by Edgene Bio). Transgenic T₂ plants were identified by genomic PCR using primers KO1P‐gus‐TD. At anthesis, spikelets from transgenic and B73 plants were harvested and subjected to histochemical GUS staining with a GUS Stain Kit (Solarbio, China). Following decolorization, tissues were imaged under a stereo microscope. Primers used are listed in Table S1.

2.10. HPLC‐MS/MS Analysis of GAs and Ent‐Kaurenoic Acid

Following the same transcriptome sampling protocol, spikelets from the main axis of WT, def1, and def1 + GA₃ tassels were sampled. The shoots of ZmKO2 knockout lines and B104 were harvested at 10 days after sowing (DAS), since the phenotypic differences at this stage were pronounced and the aboveground biomass was adequate for reliable HPLC‐MS/MS analysis. All samples were immediately snap‐frozen in liquid nitrogen and stored at −80°C. Extraction and quantification of ent‐kaurenoic acid and GAs were conducted by WeBiolotech Co. Ltd. (Wuhan, China).

3. Results

3.1. def1 Is a Single‐Gene‐Controlled Recessive GMS Mutant

While WT plants showed normal anthesis, the def1 mutant exhibited complete male sterility characterised by persistently closed glumes, non‐elongated stamen filaments, and non‐exserted anthers, but maintaining normal plant architecture and female fertility (Figure 1a–1e). Histological observations of anthers at sporogenous cell, dyad, and uninucleate stages showed that microsporogenesis in def1 was indistinguishable from WT (Figure 1f). Although the spikelets of def1 failed to open and shed pollen, results from I₂‐KI staining, Alexander staining, and in vitro germination assays demonstrated that the def1 produced morphologically normal pollen grains, with viability showing no significant difference compared to the WT (Figure 1g; Figure S2a). F₁ hybrids derived from crosses between def1 and the inbred lines B73, Mo17, and Huangzaosi exhibited fully fertility. The corresponding F₂ populations consistently segregated in a 3:1 fertile‐to‐sterile ratio (Table S2), indicating that def1 is a single‐gene‐controlled recessive GMS mutant.

FIGURE 1.

Phenotypic comparison of the WT and def1. (a) Morphology of WT and def1 at pollen shedding (PS) stage. (b) Root morphology of WT and def1 at PS stage. (c) Close‐up view of the tassels from WT and def1 at PS stage. (d) Ear seed set in WT and def1. (e) Close‐up view of the spikelets of WT (e1) and def1 (e2), and a manually opened spikelet of def1 (e3) at PS stage. (f) Paraffin sections of male spikelets at various stages of microspore development. (g) Pollen viability in WT and def1 plants.

3.2. def1 Results From a Loss‐Of‐Function Mutation in ZmKO1

Initial BSA located the def1 locus to chromosome 9. Subsequent mapping with 275 male‐sterile individuals confined the locus to the region between the markers idp4 and umc1271 (Figure 2a). To refine the mapping, the population was expanded to 4213 male‐sterile individuals, ultimately delimiting the locus to a 1.24‐Mb region between markers idp8 and SNP2 (Figure 2a). This interval, based on the Zm‐B73‐REFERENCE‐NAM‐5.0 genome, contains 11 annotated protein‐coding genes, including two putative KO genes ZmKO1 and ZmKO2 (Figure 2a; Table S3). Genomic PCR and sequencing revealed no sequence variation in ZmKO2 between def1 and WT. In contrast, a 5202‐bp insertion was identified at the 3′‐terminal region of exon 1 in ZmKO1 of def1. This insertion resulted in the replacement of a native 19‐bp segment in exon 1 by an 82‐bp fragment at the transcript level, causing a frameshift and introducing a premature termination codon predicted to truncate the encoded protein from 508 to 46 amino acids (Figure 2b).

FIGURE 2.

Map‐based cloning of the def1 gene. (a) Fine mapping of the def1 gene. N, number of male‐sterile F₂ plants used for mapping. (b) Sequence comparison of candidate gene ZmKO1 (Zm00001eb385100) between WT and def1 mutant. A 5202‐bp insertion was identified in the first exon of the def1 allele. At the transcriptional level, an 18‐bp segment (underlined) in the WT was replaced by an 82‐bp sequence (underlined) in the def1 mutant. (c) From thirteen independent regeneration events (TG‐1 ~ TG‐13), seven ProUbi::ZmKO1CDS‐positive transgenic T₀ lines (T0‐1 ~ T0‐7) were identified through genomic PCR screening using primers OEcheck. +, the ProUbi::ZmKO1CDS plasmid. (d) Allele‐specific markers Mar‐W and Mar‐m can distinguish the ZmKO1 ⁺ and ZmKO1 ^def1 , respectively. (e) Genotyping of the ZmKO1 locus in the (def1 × T0‐1)F₂ population using Mar‐W and Mar‐m markers. (f) All Zmko1 ^def1/def1 homozygous plants in the (def1 × T0‐1)F₂ population were assayed for the presence of the ProUbi::ZmKO1CDS transgene by genomic PCR with primers OEcheck. All transgenic plants (e.g., plants 3 and 4) were male fertile, whereas all non‐transgenic plants (e.g., plants 7 and 8) displayed male sterility. The numbers above the agarose gel lanes in (e) and (f) correspond to individual plant IDs in the population. (g) Knockout of ZmKO1 via CRISPR/Cas9. KO#1, KO#2, and KO#3 carry a 1‐bp insertion, a 1‐bp deletion, and a 16‐bp deletion, respectively, in the second exon of ZmKO1, all resulting in frameshift mutations (sequence highlighted in red underline) and premature termination (red star). (h) Wild‐type B104 exhibits normal fertility, whereas ZmKO1 knockout lines exhibit male sterility. (i) Mutation types and male fertility phenotypes of allelic mutants derived from EMS5‐316919 line at the ZmKO1 locus.

To validate the causal relationship between the loss‐of‐function mutation in ZmKO1 and the def1 mutant, a complementation construct ProUbi::ZmKO1CDS (Figure S1a) was transformed into inbred line B104, yielding seven independent transgenic T₀ lines (Figure 2c). Three transgenic T₀ lines (T₀‐1, T₀‐3, and T₀‐7) were crossed as male parents with def1 to generate three F₁ populations. One transgenic plant from each F₁ population was self‐pollinated to produce the corresponding F₂ population. As shown in Table S4, the non‐transgenic (def1 × B104)F₂ population segregated at a 3.64:1 fertile‐to‐sterile ratio, consistent with the monogenic recessive inheritance of the male sterility. In contrast, fertile‐to‐sterile ratios in the three transgenic F₂ populations strongly deviated from the 3:1 expectation, with values of 152:7, 211:12, and 219:4, respectively, demonstrating that the ProUbi::ZmKO1CDS transgene dramatically reduced the frequency of male‐sterile plants. We designated the alleles of ZmKO1 locus in WT and def1 as ZmKO1 ⁺ and Zmko1 ^def1 , respectively, and developed allele‐specific markers Mar‐W and Mar‐m for their discrimination (Figure 2d). Genotyping of the (def1 × T₀‐1)F₂ population at the ZmKO1 locus identified 41 ZmKO1 ^+/+ , 78 ZmKO1 ^+/def1 , and 40 ZmKO1 ^def1/def1 plants (Figure 2e), fitting a 1:2:1 segregation ratio (χ² = 0.0286, p = 0.986). Among the 40 ZmKO1 ^def1/def1 plants, all 33 transgenic individuals were fertile, while all seven non‐transgenic individuals exhibited male sterility (Figure 2f). This result demonstrates that the ProUbi::ZmKO1CDS transgene fully restores male fertility in ZmKO1 ^def1/def1 homozygotes.

CRISPR/Cas9‐mediated editing of exon 2 in ZmKO1 generated three non‐transgenic homozygous knockout lines: KO#1 (1‐bp insertion), KO#2 (1‐bp deletion), and KO#3 (16‐bp deletion). All three mutations resulted in frameshifts and premature termination codons (Figure 2g). Each knockout line displayed a male sterility phenotype identical to that of def1 (Figure 2h). The EMS5‐316919 mutant carries a G‐to‐A nonsense mutation in exon 6 of ZmKO1, introducing a premature termination codon and truncating the C‐terminal 199 amino acids of the encoded protein (Figure 2i). Genotyping and phenotyping of the EMS5‐316919 individuals showed that wild‐type (ZmKO1 ^G/G ) plants and heterozygous (ZmKO1 ^G/A ) mutants exhibited normal anthesis, whereas homozygous (ZmKO1 ^A/A ) mutants showed strongly suppressed anthesis, with only partial spikelets undergoing anthesis (Figure 2i). Collectively, these results conclusively demonstrate that the male sterility of def1 is caused by the loss‐of‐function mutation in ZmKO1.

3.3. ZmKO1 Links GA Biosynthesis to Male Fertility Regulation

Subcellular localization confirmed plastid targeting of ZmKO1‐eGFP through co‐localization with chlorophyll autofluorescence (Figure 3a), consistent with prior reports of KO proteins in Arabidopsis and rice (Helliwell et al. 2001; Zhang et al. 2020). ZmKO1 transcripts were detected in all examined developmental stages and tissues. Expression of the mutant allele ZmKO1 ^def1 was significantly lower than that of the wild‐type allele ZmKO1 ⁺ (Figure 3b). To determine its spatial expression pattern in male spikelets, a ZmKO1pro::GUS reporter construct was transformed into the inbred line B73. Histochemical staining revealed specific GUS activity localized to the lodicules of male spikelets in transgenic plants (Figure 3c).

FIGURE 3.

Functional characterisation of ZmKO1. (a) Subcellular localization of ZmKO1 protein. (b) Spatiotemporal expression patterns of ZmKO1 in WT and def1. Data are presented as mean ± SD (n = 3 biological replicates). DAS, day after sowing. (c) Histochemical GUS staining of male spikelets. (d) Analysis of ent‐kaurenoic acid and GA levels in tassel spikelets of WT and def1. Data are presented as mean ± SD (n = 3 biological replicates). Significance test was determined by Student's t‐test (*, p < 0.05; **, p < 0.01; ***, p < 0.001). n.d., not detected. Black numbers denote values for the WT; red numbers denote values for def1. The accompanying percentages reflect the increase (+) or decrease (−) in def1 relative to WT. (e) Restoration of def1 male fertility by GA₁, GA₃, GA₄ and GA₇. Treatment concentration: 200 ppm. (f) Close‐up view of a male spikelet of def1 following GA₃ treatment. (g) Close‐up view of the anthers in (f). (h) Filament length in spikelets of WT and GA₃‐treated def1 at anthesis Data are presented as box plot (n = 25). Significance test was determined by Student's t‐test (ns, not significant). (i) Pollen viability in WT and def1 + GA₃ plants. (j) Relative levels of GAs in def1 + GA₃ treatment, normalised to the corresponding GA levels in untreated def1 plants (set as “1”). Data are presented as mean ± SD (n = 3 biological replicates). Significance test was determined by Student's t‐test (***, p < 0.001; ns, not significant). (k) Top 15 enriched GO terms for the 1826 common DEGs. (l) Top 15 enriched KEGG pathways for the 1826 common DEGs. (m) The differential expression patterns of the seven SAUR genes that enriched in the GO:0009725 term.

ZmKO1 encodes a putative KO catalysing the conversion of ent‐kaurene to ent‐kaurenoic acid in GA biosynthesis (Mao et al. 2017). Quantification of ent‐kaurenoic acid and 12 GAs (including bioactive GA₁, GA₃, GA₄, and GA₇) in male spikelets at anthesis revealed that def1 had significantly lower levels of ent‐kaurenoic acid and 11 GAs compared to WT, while GA₃ content was uniquely increased by 61.73% (Figure 3d). Exogenous application of GA₁, GA₃, GA₄, or GA₇ individually restored filament elongation, glume opening, and pollen shedding in def1 spikelets (Figure 3e–3h). Following GA₃ treatment, the pollen released from def1 exhibited no significant difference in viability compared to that of the WT (Figure 3i; Figure S2b). On the next day after GA₃ treatment, the concentrations of GA₃ and GA₁ in def1 + GA₃ plants increased by 610.97‐fold and 38.09‐fold, respectively, compared to untreated def1 plants. In contrast, neither GA₄ nor GA₇ levels differed significantly between the two groups (Figure 3j). These results confirm that ZmKO1 functions as a KO enzyme, and its disruption impairs GA biosynthesis, leading to GA‐responsive male sterility in def1.

Comparative transcriptome analyses of male spikelets from WT, def1, and def1 + GA₃ plants were performed. A total of 3679 and 3628 DEGs were identified in the def1 vs. WT and def1 + GA₃ vs. def1 comparisons, respectively, with 1826 DEGs shared between them (Data S1 and S2). Notably, all common DEGs displayed fully reversed expression patterns across the two comparisons (Data S2), indicating that their transcription is directly responsive to GA‐signalling changes. GO and KEGG enrichment analyses of the common DEGs revealed significant enrichment for the GO term response to hormone (GO:0009725) and the pathway plant hormone signal transduction (ko04075), respectively (Figure 3k,l; Data S3 and S4). All seven SMALL AUXIN UP RNA (SAUR) genes enriched in GO:0009725 were significantly down‐regulated in def1 relative to WT, while exogenous GA₃ treatment up‐regulated their expressions (Figure 3m).

3.4. ZmKO2 Also Functions as an Ent‐Kaurene Oxidase for GA Biosynthesis

While mutations in KO genes typically caused dwarf phenotypes (Helliwell et al. 1998; Chen et al. 2019), the ZmKO1 mutations in def1 and knockout lines only led to a transient reduction in plant height during early seedling development, with subsequent recovery to wild‐type levels (Figure 4a). Ultimately, these mutations had no significant effect on final plant height, ear height, or internode length (Figure S3). Homology analysis indicated that maize possesses an additional putative KO gene, ZmKO2 (Mao et al. 2017). In all examined tissues (except roots at 66 DAS), the expression levels of ZmKO2 were markedly higher at both the V1 and V8 stages than at anthesis (66 DAS) (Figure 4b). To investigate whether loss of ZmKO2 function induces dwarfism, a dual‐target CRISPR/Cas9 construct was designed to target exons 4 and 5 of ZmKO2 (Figure 4c) and transformed into B104. Non‐transgenic edited plants were selected from BC₁F₁ progeny after backcrossing T₀ positive plants with B104, and BC₁F₂ populations were generated by self‐pollination. From three independent BC₁F₂ populations, we identified two homozygous knockout lines, KO2#1 and KO2#2, carrying distinct editing profiles (Figure 4c). No off‐target edits were detected in ZmKO1. In addition, CRISPR‐P predicted the top 20 potential off‐target sites genome‐wide. Sites with an off‐target score > 0.5 or located within gene CDS regions were selected for off‐target editing validation by genomic PCR amplification and sequencing, and no off‐target editing was detected at any examined site (Table S5).

FIGURE 4.

Effects of ZmKO1 and ZmKO2 on maize plant height. (a) Plant height performance of def1 and ZmKO1 knockout lines at 8 DAS and 16 DAS. (b) Spatiotemporal expression patterns of ZmKO2 in B104. Data are presented as mean ± SD (n = 3 biological replicates). (c) Knockout of ZmKO2 via CRISPR/Cas9. The short red line indicates a 1‐bp deletion, and the inverted black triangle shows the site of base insertion. The red amino acid sequence signifies a frameshift mutation; the red star indicates the stop codon. (d) The dwarf phenotype and pollen‐less male sterility of ZmKO2 knockout lines. (d1) Plant height performance of B104 and the KO2#1 and KO2#2 lines at 10 DAS. (d2) Plant height performance of B104 and KO2#2 line at 66 DAS. (d3, d4) Close‐up view of a tassels from B104 (d3) and KO2#2 (d4) plants. (d5, d6) Comparison of male spikelets (d5) and anthers (d6) from the B104 and KO2#2 line at 66 DAS. (d7, d8) I₂‐IK staining of pollen derived from B104 (d7) and KO2#2 line (d8). (e) Analysis of ent‐kaurenoic acid and bioactive GAs concentrations in shoots of ZmKO2#2 and B104 at 10 DAS. Data are presented as mean ± SD (n = 3 biological replicates). Significance test was determined by Student's t‐test (*, p < 0.05; **, p < 0.01; ***, p < 0.001). (f) Rescue of the dwarf phenotype in KO2#2 seedlings by exogenous GA₃ application. Treatment concentration: 200 ppm. DAS, day after sowing. DAT, day after treatment.

Both knockout lines exhibited severe dwarfism from the early seedling stage (Figure 4d). As exemplified by KO2#2 line, loss of ZmKO2 resulted in extreme dwarfism throughout all growth stages and a pollen‐less male‐sterile phenotype (Figure 4d). Quantification of ent‐kaurenoic acid and bioactive GAs in 10 DAS shoots showed that KO2#2 had significantly reduced levels of ent‐kaurenoic acid, GA₁, GA₃, GA₄, and GA₇, with reductions of 64.81%, 50.86%, 98.39%, 55.16%, and 42.49%, respectively, compared to B104 (Figure 4e). The dwarf phenotype of KO2#2 seedling was fully rescued by exogenous GA₃ application (Figure 4f). These results demonstrate that ZmKO2, like ZmKO1, functions as an authentic KO in GA biosynthesis.

3.5. Creating ZmKO1‐Based Dual‐Use GMS Lines in Maize

Exogenous GA₃ application effectively restored fertility in the def1 mutant, confirming its utility as a dual‐use GMS line. To assess the genetic background independence of this trait, we generated ZmKO1‐based dual‐use GMS lines in diverse inbred backgrounds through marker‐assisted backcross (MABC) and CRISPR/Cas9 editing (Figure S4a,b). The Mar‐m marker reliably tracked the ZmKO1 ^def1 allele across backcross generations (Figure S4c,d). We developed eight male‐sterile lines, Zheng58‐def1, DH351‐def1, Mo17‐def1, PH6WC‐def1, B73‐def1, Y9614‐def1, ZNC442‐def1, and Jing724‐def1, via MABC, and two non‐transgenic male‐sterile lines, B104‐def1 and SCML0849‐def1, via CRISPR/Cas9 (Figure S5). All lines remained male‐sterile under natural conditions but fully regained fertility upon exogenous GA₃ treatment (Figure 5a), enabling successful seed set through self‐pollination after GA₃ application (Figure 5b). Aside from male fertility, no significant differences were observed between the sterile lines and their corresponding inbred lines in key agronomic traits, including plant height, ear height, silking date, kernel number per ear, 100‐kernel weight, and field emergence rate (Table S6). Phenotypic evaluations of male sterility and fertility restoration were conducted across multiple locations: Chengdu (Sichuan), Qingtongxia (Ningxia), Yining (Xinjiang), Menghan (Yunnan), Linze (Gansu), and Ledong (Hainan) (Table S7). Stable male sterility and reliable fertility restoration were consistently observed at all sites, indicating that the fertility performance of this dual‐use system is robust across diverse genetic backgrounds and environments (Figure 5a,c,d).

FIGURE 5.

Phenotypic characterisation of ZmKO1‐based dual‐use GMS lines. (a) Fertility performance of ZmKO1‐based dual‐use GMS lines in Chengdu. (b) Comparison of self‐pollinated ears between ZmKO1‐based dual‐use GMS lines after exogenous GA₃ application and their corresponding inbred lines. (c) Stability of male sterility of ZmKO1‐based dual‐use GMS lines at different locations. (d) Fertility restoration performance of ZmKO1‐based dual‐use GMS lines after exogenous application of GA₃ at different locations.

3.6. ZmKO1‐Based Dual‐Use GMS Line Enables Two‐Line Hybrid Seed Production in Maize

Gibberellic acid, a widely used gibberellin‐based plant growth regulator in agriculture, was applied in this study using a commercial formulation containing 75% active ingredient. The optimal concentration for restoring male fertility in def1 was determined to be ~150 mg/L (Figure 6a). According to agricultural technology extension guidelines, the recommended ambient temperature for applying gibberellic acid is between 15°C–30°C. We established an efficient propagation protocol for ZmKO1‐based dual‐use GMS lines: when 60% of plants reach silk emergence, gibberellic acid (150 mg/L, ~600 L/ha) is sprayed around 18:00 using a backpack agricultural sprayer (Figure 6b). The following day, under suitable field conditions, the male‐sterile lines flower and shed pollen normally (Figure 6c). A second application is made on the third day after pollen shedding initiation. We observed that even if rain occurred the day after spraying, which temporarily suppressed flowering and pollen release, normal anthesis resumed once the weather cleared (Figure 6d). Gibberellic acid treatment did not alter plant architecture, plant height, ear height, or ear height coefficient of the male‐sterile lines (Figure 6e,f). Taking def1 as an example, the ZmKO1‐based dual‐use GMS lines can be efficiently propagated through exogenous gibberellic acid application under open‐pollination conditions (Figure 6g).

FIGURE 6.

Propagation of ZmKO1‐based dual‐use GMS lines. (a) Screening for optimal gibberellic acid concentration. (b) Spray treatment of the dual‐use GMS line DH351‐def1. (c) Fertility restoration performance of DH351‐def1 (c1), Mo17‐def1 (c2), def1 (c3), SCML0849‐def1 (c4) and B104‐def1 (c5) at one day after gibberellic acid treated in Chengdu. (d) Fertility restoration in dual‐use GMS line ZNC442‐def1 subsequent to exogenous gibberellic acid application, with rainfall on day 1 and clear conditions on day 2 in Chengdu. (e) Plant architecture of gibberellic acid‐treated and untreated ZNC442‐def1 line in Menghan. (f) Statistical analysis of Plant height, ear height and ear height coefficient in gibberellic acid‐treated and untreated ZmKO1‐based dual‐use GMS lines in Menghan. Data are presented as box plot (n = 28). (g) Seed set performance in WT and def1 under isolated and open‐pollination conditions in Menghan.

To evaluate the impact of the ZmKO1‐based dual‐use GMS on hybrid performance, we crossed eight elite inbred lines as male parents with male‐sterile lines (def1 and B104‐def1) and with corresponding wild‐type lines (WT and B104). The resulting F₁ hybrids were grown in Chengdu, and nine agronomic traits were measured. Pairwise comparisons between hybrids derived from male‐sterile lines and those from wild‐type lines showed no significant differences in pollen shedding, plant height, ear height, ear weight, ear diameter, ear length, ear row number, ear grain weight, and 100‐kernel weight (Figure 7a,b; Figure S6; Table S8). These results confirm the suitability of the ZmKO1‐based dual‐use GMS line for maize hybrid seed production. Finally, we established a Gibberellin‐switchable Dual‐use Line (GDL) system for maize hybrid seed production based on the ZmKO1‐based dual‐use GMS line and successfully applied it in the seed production of the hybrid cultivar Chuandan99 (Figure 7c,d).

FIGURE 7.

Hybrid seed production using ZmKO1‐based dual‐use GMS line. (a, b) The hybrids, including H1 (WT × DH351), H2 (def1 × DH351), H3 (WT × DH382), H4 (def1 × DH382), H5 (WT × PHBIM), H6 (def1 × PHBIM), H7 (B104 × PH6WC), H8 (B104‐def1 × PH6WC), H9 (B104 × DH382), H10 (B104‐def1 × DH382), H11 (B104 × SCML0849), and H12 (B104‐def1 × SCML0849), exhibited normal male fertility (a) and yield performance (b). (c, d) Schematic diagram of the Gibberellin‐switchable Dual‐use Line (GDL) system for hybrid seed production of Chuandan 99 including the propagation of dual‐use GMS line ZNC442‐def1 (c) and the hybrid seed production using dual‐use GMS line ZNC442‐def1 (d). SCML0849: Paternal line of Chuandan 99.

4. Discussion

4.1. ZmKO1 And ZmKO2 Encode Functional KOs in Maize

Plant GA biosynthesis begins with geranylgeranyl diphosphate, which is cyclized to ent‐kaurene by ent‐copalyl diphosphate synthase (CPS) and ent‐kaurene synthase (KS). Ent‐kaurene is subsequently oxidized to ent‐kaurenoic acid by KO, and then converted to GA₁₂ by ent‐kaurenoic acid oxidase (KAO) (Figure 3d). In maize, genes encoding CPS, KS, and KAO have been functionally characterised through studies of classical dwarf mutants an1, ks3‐1, and gad5‐1 (Bensen et al. 1995; Lu et al. 2018; Li et al. 2023), yet the KO remained molecularly uncharacterized. Although ZmKO1 and ZmKO2 were previously annotated as putative KO genes and heterologous expression suggested only ZmKO1 possessed KO activity (Mao et al. 2017), their in‐planta functions remained unexplored. Here, we provide forward genetic evidence that ZmKO1 functions as a KO in maize. Unlike loss‐of‐function mutants of OsKO1, OsKO2, OsKO5, and AtKO1, which exhibit semi‐dwarf or severe dwarf phenotypes in rice and Arabidopsis (Helliwell et al. 1998; Sakamoto et al. 2004; Swain et al. 2005; Chen et al. 2019), def1 and ZmKO1 knockout lines only showed a transient height reduction at the seedling stage (Figure 4a; Figure S3). Given the high expression of ZmKO2 during vegetative growth (Figure 4b), we examined its role in plant stature. ZmKO2 knockout lines exhibited severe dwarfism and pollen‐less male sterility, accompanied by significantly decreased levels of ent‐kaurenoic acid and bioactive GAs (Figure 4d,e). At the seedling stage, the dwarf phenotype was fully rescued by exogenous GA₃ application (Figure 4f). These results collectively demonstrate that, contrary to the previous report (Mao et al. 2017), ZmKO2 also acts as a functional KO in GA biosynthesis and is essential for regulating both plant height and microsporogenesis in maize.

4.2. Stamen Filament Elongation Is Potentially GA₁ ‐Dependent and May Involve SAUR Gene Induction in Maize

While hormonal regulation of stamen filament elongation by auxin, GAs, and jasmonic acid has been well characterised in Arabidopsis (Peng 2009; Cheng et al. 2009; Song et al. 2013; Gastaldi et al. 2020; He et al. 2023; Wang et al. 2023), it remains poorly understood in monocots. The def1 mutant exhibited impaired stamen filament elongation due to impaired GA biosynthesis (Figure 1e; Figure 3d). GA₁, GA₃, GA₄, and GA₇ are recognised bioactive GAs that regulate plant development, with GA₁ and GA₄ being predominant in most plants (Cowling et al. 1998; Ueguchi‐Tanaka et al. 2005; Yamaguchi 2008; Hedden and Thomas 2012; Magome et al. 2013; Nomura et al. 2013; Chaudhry et al. 2024). GA₃ and GA₁ levels in spikelets of GA₃ treated def1 plants were 610.79‐fold and 38.09‐fold, respectively, compared to untreated def1 plants, while GA₄ and GA₇ levels remained unchanged (Figure 3j). The sharp rise in GA₃ is likely attributed primarily to the direct uptake from the exogenous application, whereas the increase in GA₁ is probably due to GA₃ induced biosynthesis. Although exogenous application of GA₁, GA₃, GA₄ and GA₇ each fully restored the fertility of def1. endogenous GA₃ levels in its tassel spikelets were 61.73% higher than those in WT (Figure 3d,e). These results strongly suggest that GA₁, rather than GA₃, is likely the primary bioactive GA regulating filament elongation in maize. Meanwhile, other possibilities should also be considered. For instance, ZmKO1 expression is specifically localised to the lodicules of the male spikelet (Figure 3c), whereas our hormone quantification used whole spikelets and thus reflects GA₃ dynamics at the whole‐organ level. Tissue‐specific analysis of lodicules might reveal a different concentration profile. It is also possible that endogenous GA₃ bioactivity is lower in the def1 spikelets than in the WT spikelets, which may lead to feedback‐mediated accumulation of GA₃ in the mutant. These hypotheses warrant further experimental verification. SAUR genes were initially characterised as auxin‐responsive genes, but later studies revealed that certain SAUR63 subfamily members are GA‐inducible and mediate GA‐dependent filament elongation in Arabidopsis (Bai et al. 2012; Chae et al. 2012; Ren and Gray 2015; Gastaldi et al. 2020). The mechanism of GA‐regulated filament elongation remains unknown in maize. Our comparative transcriptomic analysis identified seven SAUR genes that were significantly down‐regulated in def1 and consistently up‐regulated following exogenous GA₃ treatment (Figure 3m). These GA‐responsive SAUR genes thus represent strong candidates for mediating GA‐dependent filament elongation, and their functional validation via overexpression, knockout, or spatial expression manipulation constitutes an important goal for ongoing and future research.

4.3. Key Factors Determining the Applicability of the ZmKO1‐Based GMS Line for Maize Hybrid Seed Production

Using MABC and CRISPR/Cas9 editing enables the rapid and straightforward development of ZmKO1‐based GMS lines across diverse genetic backgrounds (Figure 5a). Their fertility can be efficiently restored by exogenous application of GA₃ (Figure 5a), conferring a “dual‐use” capability that resolves the challenge of efficiently propagating GMS line. Fertility stability is one of the key factors affecting the application of male‐sterile lines in hybrid seed production. In China, major hybrid maize seed production bases are located in Gansu and Xinjiang, with additional hubs in Ningxia, Sichuan, and Yunnan. Additionally, Hainan serves as the national winter multiplication core zone. Despite distinct environmental profiles in temperature and humidity in these locations (Table S7), the ZmKO1‐based GMS lines maintained stable male sterility, and fertility was effectively restored upon GA₃ treatment (Figure 5a,c,d). These results indicate that the dua‐use trait of the ZmKO1‐based GMS line is robust across diverse environments and genetic backgrounds. Besides fertility stability, the effect of the ZmKO1‐based male sterility on agronomic traits is also crucial for its application in hybrid seed production. Table S6 shows that the male‐sterile lines and their corresponding inbred lines did not differ significantly in key agronomic traits. Furthermore, F₁ hybrids derived from ZmKO1‐based GMS female parents displayed normal male fertility and agronomic traits (Figure 7a and 7b; Table S8). These results suggest that the ZmKO1‐based male sterility has no adverse effects on the agronomic traits of the male‐sterile line or the F₁ hybrids.

The seed yield, spraying cost, and operability of the male‐sterile lines propagation are also critical factors affecting their practical application. GA are essential for many developmental processes in plants (Davière and Achard 2013). In maize production, lodging leads to substantial yield losses, complicates harvesting, and ultimately raises production costs. Plant height, ear height, and the ear height coefficient are the key agronomic traits directly related to lodging resistance. We found that the gibberellic acid spraying protocol did not cause visually detectable alterations in plant architecture, plant height, ear height, and the ear height coefficient (Figure 6e,f). Under open‐pollination conditions after gibberellic acid treatment, the seed set of the def1 population showed no significant difference from that of the WT population (Figure 6g). The above results indicate that gibberellic acid spraying is unlikely to affect the yield performance of male‐sterile lines.

Commercially available gibberellic acid is cost‐effective. The price for the formulation used in this study is only ~1.6 RMB per gram. Based on our spraying protocol, the estimated gibberellic acid cost for propagating male‐sterile lines is ~288 RMB per hectare. The spraying equipment is common in agricultural production, with no specific model required as long as it can atomize the solution effectively. For large‐scale seed production, we suggest that unmanned aerial vehicle spraying would be a superior option. The propagation coefficient of maize inbred lines typically ranges from 1:200 to 1:350. Conservatively estimated, the yield from one hectare of male‐sterile line propagation can meet the demand for 200 ha hybrid seed production. According to publicly available data from 2025 in major maize hybrid seed production regions like Zhangye and Xinjiang, the manual detasseling cost for 200 ha ranges approximately from 780 000 to 1 200 000 RMB, while mechanical detasseling costs about 300 000 to 450 000 RMB. Therefore, the overall cost of gibberellic acid spraying (including the expenses for the gibberellic acid, labor, and spraying equipment) is negligible. So, during the propagation of male‐sterile lines, the spraying protocol does not require extreme precision. The gibberellic acid dosage can be moderately increased and the application frequency raised to ensure the highest fertility restoration rate and seed yield. Additionally, 100% restored fertility is not needed due to partial fertility of maize being enough to accomplish self‐pollination. Collectively, the ZmKO1‐based GMS line is well‐suited for establishing a two‐line hybrid seed production system in maize. Ultimately, we established a GDLsystem and successfully applied it in the hybrid seed production of Chuandan 99 (Figure 7c,d).

4.4. Advantages of the GDL System

Currently, three main male‐sterility systems are used in hybrid maize production: the CMS system, the RHS system, and the transgenic‐mediated non‐transgenic seed production system (SPT or MCS) (Table S9). The CMS system has been constrained by limited genetic diversity, disease susceptibility, and unstable fertility restoration (Williams 1995). For the RHS system, both parental lines and their hybrids are transgenic, and implementing this system requires blanket glyphosate application across the entire seed‐production field, raising concerns about its ecological and health impacts (Singh et al. 2024). In addition, the RHS system induces male sterility by spraying glyphosate within a specific time window. The developmental asynchrony among field plants necessitates a more precise operational protocol to ensure fertility meets production requirements. Although the transgenic‐mediated non‐transgenic seed production system, such as SPT and MCS, employs transgenic strategies to enable the propagation of non‐transgenic GMS lines, it is still associated with a certain risk of transgene flow (Wu et al. 2016; Zhang et al. 2018). Moreover, the transgenic vector used in this system involves the concatenation of multiple expression cassettes, resulting in an extended transgenic fragment. This considerable length raises the likelihood of recombination events, potentially leading to systemic errors in seed production.

Here, the GDL system effectively overcomes the limitations associated with these existing systems. (i) Stable male sterility. The male sterility of the ZmKO1‐based GMS line remains stable across diverse genetic backgrounds and environmental conditions (Figure 5a,c). (ii) Broad restoration spectrum. ZmKO1‐based GMS is recessive; all plants used as male parents would, by virtue of being male fertile, contain the wild‐type allele of ZmKO1 and would be restorers (Figure 7a). (iii) User‐friendly. As discussed in Section 4.3, propagating the ZmKO1‐based GMS lines does not require highly precise spraying. The protocol is straightforward and easy to implement. (iv) Non‐transgenic and environmentally friendly. Maintenance of the ZmKO1‐based GMS line does not rely on transgenic technology and is therefore free from transgene‐associated risks. Gibberellic acid is an endogenous plant hormone and a widely adopted plant growth regulator in agricultural production, posing minimal environmental risk.

4.5. The Potential Transferability of the GDL System to Other Crops

Loss‐of‐function mutations in KO encoding genes in rice and Arabidopsis result in extreme dwarfism (Helliwell et al. 1998; Chen et al. 2019). The extreme dwarfism phenotype often severely compromises other agronomic traits, rendering it impractical for actual agricultural production. Although a weak allele of OsKO2 confers a semidwarf phenotype, it exhibits normal fertility. We conclude that the feasibility of engineering a practical GDL system via KO genes varies significantly across species. Coix, a close relative of maize, is monoecious. We performed knockout of the KO gene homologous to ZmKO1 in Coix and preliminarily observed that the male flowers exhibited a male‐sterile phenotype similar to that of ZmKO1‐based GMS. Although the stability of sterility and its effects on other agronomic traits require further investigation, we propose that the GDL system can be replicated in some other crops.

4.6. Other Issues Requiring Further Clarification

Certainly, this system warrants further discussion across additional dimensions. For instance, the impact of extreme weather events on fertility stability and the potential influence of the introduced ZmKO1‐based male‐sterile trait on overall plant resistance require thorough evaluation. Researches on the compatibility of the GDL system with different maize heterotic groups and its impact on hybrid yield stability across multiple locations should be expanded. We hope that the GDL system will be widely adopted in the hybrid seed production for maize in the near future.

Author Contributions

Chuan Li and Moju Cao secured funding, conceived the study, and provided overall project supervision. Chuan Li, Siyu Wang, Yong Xiang, Tao Yu, Rujun Wei, Wenqian Mei, Xiaowei Liu, Peng Yuan, Hongyang Yi, and Shufeng Zhou performed the experiments. Chuan Li wrote the original draft of the manuscript. Chao Xia, Moju Cao, and Hongyang Yi reviewed and edited the manuscript. All authors read and approved the final manuscript.

Funding

This work was supported by the National Key Research and Development Program of China, 2023YFD1201101, 2022YFD1200704‐3. National Natural Science Foundation of China, 32101673. Major Science and Technology Project of Sichuan Province, 2022ZDZX0013. Academician Expert Workstation of Yunnan Province, 202305AF150052.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1: Schematic map of the constructs. (a) Schematic map of the ZmKO1 expression construct ProUbi::ZmKO1CDS. (b, c) Schematic map of the ZmKO1 (b) and ZmKO2 (c) CRISPR/Cas9 gene editing constructs.

Figure S2: Pollen viability assessment at the anthesis stage. (a) Pollen viability assessment for the unflowered male spikelets. (b) Pollen viability assessment for the freshly shed pollen after GA₃ treatment. Data for each box in the box plot were derived from five independent individual plants, with more than 100 pollen grains counted per plant. Significance test was determined by Student's t‐test (ns, not significant).

Figure S3: The effect of ZmKO1 mutation on plant height, ear height, and internode length. (a, b) Comparison of plant height and ear height between def1 and WT (a) and between ZmKO1 knockout lines (KO1#1, KO1#2) and B104 (b) at 8 DAS and anthesis stage, respectively. Data are presented as box plot (n = 20 biological replicates) (c) Comparison of internode length between def1 and WT and between ZmKO1 knockout lines (KO1#1, KO1#2) and B104 at anthesis stage, respectively. Data are presented as mean ± SD (n = 10 biological replicates). Significance test was determined by Student's t‐test (***, p < 0.001; ns, not significant).

Figure S4: Development of ZmKO1‐based dual‐use GMS lines via marker‐assisted backcrossing (MABC) and CRISPR/Cas9 genome editing. (a) Schematic diagram of the MABC workflow. For MABC, Mar‐m as the ZmKO1 ^def1 ‐specific marker for foreground selection and def1 mutant as the donor parent for the male‐sterile allele ZmKO1 ^def1 . (b) Schematic diagram of the CRISPR/Cas9 genome editing ZmKO1 workflow. (c, d) A representative marker‐assisted selection for ZmKO1 ^+/def1 individuals from the BC₁F₁ generations (c) and the BC₆F₁ generations (d).

Figure S5: ZmKO1‐based GMS lines in diverse inbred backgrounds. Zheng58‐def1, DH351‐def1, Mo17‐def1, PH6WC‐def1, B73‐def1, Y9614‐def1, ZNC442‐def1, and Jing724‐def1 male‐sterile lines derived from MABC; non‐transgenic B104‐def1 and SCML0849‐def1 male‐sterile lines derived from CRISPR/Cas9 genome editing.

Figure S6: Plant architecture of hybrid progeny.

Table S1: Primers used in this study.

Table S2: Genetic analysis for the def1 mutation.

Table S3: Functional annotation of genes in the fine‐mapped region.

Table S4: Genetic analysis for the F2 populations used for functional complementation assay.

Table S5: Sequencing analysis of potential off‐target sites.

Table S6: Statistical analysis of agronomic traits in male‐sterile and inbred lines.

Table S7: Temperature and humidity across different locations.

Table S8: Statistical analysis of agronomic traits in F1 hybrids.

Table S9: Comparison of different male‐sterility systems in maize.

Data S1: The 3679 and 3628 DEGs identified from the two RNA‐seq comparisons.

Data S2: Expression patterns and functional annotations of the 1826 common DEGs.

Data S3: GO and KEGG analyses of the 1826 common DEGs.

Data S4: Expression patterns and functional annotations of genes enriched in GO:0009725 and ko04075.

Data S5: Supplementary Information.

Data Availability Statement

The data that supports the findings of this study are available in the Supporting Information of this article.