Haploid induction by editing the conserved GEX3 gene in monocots and dicots Authors: Lan Wang, Shu-Yan Chen, Jing Wang, Shu Chang, Peng-Fei Jia, Wei-Cai Yang, Hong-Ju Li Published: 2026 Dear Editor, The growing global population and escalating challenges to food security require the accelerated development of crops that combine high yield with stress resilience. Conventional hybrid breeding relies on laborious, resource-intensive procedures for generating and selecting inbred lines (Zhang et al., 2023). In contrast, doubled haploid (DH) technology can significantly accelerate this process: haploid induction followed by chromosome doubling enables the production of fully homozygous DH lines in only two generations (Jacquier et al., 2020). This approach substantially shortens breeding cycles and improves selection efficiency. Because spontaneous haploid formation occurs at very low frequencies, artificial induction methods have been developed. In vitro haploid induction—via gametic embryogenesis—involves the culture of post-meiotic germ cells, but its application is limited by technical complexity, high cost, and strong species and genotype specificity (Gilles et al., 2017). In contrast, in vivo haploid induction, also referred to as in situ gynogenesis, triggers embryogenesis directly from the egg cell and therefore offers broader applicability (Gilles et al., 2017; Kalinowska et al., 2019). In vivo induction operates mainly through three mechanisms: (1) parthenogenesis, in which ectopic expression of embryo-promoting factors in egg cells initiates embryo development (Quiroz et al., 2024). Although this strategy is promising, its optimization is hampered by the limited understanding of the key regulators and regulatory networks that control egg-cell embryogenesis across different species. (2) Chromosome elimination, in which either maternal or paternal chromosomes are lost after fertilization (Ishii et al., 2016; Xiong et al., 2023). (3) Single fertilization, in which the egg cell is activated by sperm entry without integration of the paternal genome (Zhong et al., 2020; Zhang et al., 2023; Chen et al., 2025). Despite recent advances, the limited availability of haploid-inducing genes and the incomplete understanding of their mechanisms continue to hinder the widespread implementation of DH technology. Therefore, identifying efficient and broadly applicable haploid inducers remains essential for accelerating crop improvement. Through transcriptomic analysis, we identified AtGEX3 as a sperm cell-specific gene that functions synergistically with DMP8/9 in gamete adhesion and fusion. In the gex3 dmp8 dmp9 triple mutant, seed set was lower than in either the gex3 or dmp8 dmp9 mutants alone, as detailed in a companion manuscript (Wang et al., 2026). Among the progeny of three independent gex3 knockout lines generated using CRISPR–Cas9 in the wild-type (WT) Col-0 background, a subset of mutant plants exhibited small stature, reduced floral organ size, shortened siliques, and complete sterility (Figures 1A–1C; Supplemental Figures 1A and 1B). Flow cytometry of leaf nuclei from these plants showed a single peak corresponding to the 1n DNA content, indicating a haploid genome (Figure 1D). To further validate ploidy, we performed chromosome spreads using male meiocytes. While WT diploids displayed 20 sister chromatids at anaphase II, the haploid plants showed five unpaired chromosomes at anaphase I that segregated unequally, consistent with the absence of homologous chromosome pairing (Supplemental Figure 1C). These cytological results confirmed that these plants were haploid. To evaluate the haploid induction rate (HIR), we systematically screened the self-pollinated progeny of the gex3 mutant. Among 1,916 selfed progeny, 11 were identified as haploids, yielding an HIR of 0.57% (11/1,916, haploids/total selfed gex3 progeny), compared with 0% (0/692, haploids/total WT progeny) in WT plants (Supplemental Figure 1D). DMP genes were the first reported to enable haploid induction in both monocots and dicots (Zhong et al., 2019, 2020). In our hands, Arabidopsis dmp8 dmp9 double mutants showed an HIR of 1.98% (14/708, haploids/total selfed dmp8 dmp9 progeny) (Supplemental Figure 1D). Notably, CRISPR–Cas9-mediated knockout of DMP8 and DMP9 in the gex3-2 mutant increased the HIR to 3.13% (14/447, haploids/total selfed gex3 dmp8 dmp9 progeny) (Supplemental Figures 1C and 1D). This result suggests that these genetic loci exert an additive or synergistic effect in promoting haploid induction. Loss of GEX3 function induces haploid formation across monocots and dicots. (A) Inflorescences of diploid and haploid gex3 plants. Scale bar, 1 cm. (B and C) Flower (B) and siliques (C) of diploid and haploid gex3 plants. Scale bars, 1 mm. (D) Flow cytometry analysis confirming the ploidy of diploid and putative haploid gex3 plants. The x-axis represents nuclear DNA content, and the y-axis shows the number of nuclei. (E) Panicle morphology of WT and Osgex3 mutant plants. Scale bar, 2 cm. (F) Representative images of 3-week-old diploid, haploid, and triploid Osgex3 plants. Scale bar, 2 cm. (G) Flow cytometry analysis confirming the ploidy of diploid, putative haploid, and putative triploid Osgex3 plants. (H) Representative tomato plants showing diploid WT and haploid Slgex3 individuals. Scale bar, 10 cm. (I and J) Leaves (I) and flowers (J) of diploid WT and haploid Slgex3 plants. Scale bars, 2 cm in (I) and 1 cm in (J). (K) Flow cytometry analysis confirming the ploidy of diploid and haploid tomato plants. (L) Representative plants showing diploid WT and haploid Gmgex3 soybean individuals. Scale bar, 10 cm. (M) Flow cytometry analysis confirming the ploidy of diploid and putative haploid soybean plants. In (G), (K), and (M), the x-axis represents nuclear DNA content, and the y-axis shows the number of nuclei. To determine whether the GEX3 mutation induces maternal haploids, we pollinated WT pistils with pollen from a gex3 mutant expressing OLEO1:OLEO1-mCherry, a reporter expressed in both the embryo and endosperm (Zhong et al., 2020). Among the resulting 1,858 F1 seeds, two with faint mCherry fluorescence were isolated as putative maternal haploids (Supplemental Figure 1E). Subsequent genotyping confirmed the absence of the paternal gex3 allele in these seedlings (Supplemental Figure 1F), further validating their maternal origin. To evaluate the functional conservation of GEX3 across species, we identified its ortholog in rice (Oryza sativa), a single-copy gene designated OsGEX3 (Os01G42060). Public transcriptome data indicated that OsGEX3 is specifically expressed in anthers. We generated two OsGEX3 knockout lines using CRISPR–Cas9: Osgex3-1, which harbors a single-base insertion in the N-terminal region that results in a premature stop codon, and Osgex3-2, which carries a large deletion causing the loss of the first four β-propeller repeats in the deduced protein (Supplemental Figure 2A). These mutants displayed similarly reduced seed set (Figure 1E; Supplemental Figure 2B). Among the 163 progeny plants screened, one plant exhibited dwarfism and complete sterility, both of which are characteristic of haploidy (Figure 1F). Flow cytometry combined with fluorescence in situ hybridization confirmed the haploid status of this individual (Figures 1F and 1G; Supplemental Figure 2C). By contrast, screening of a WT rice population (n = 463) yielded no haploids. These results collectively demonstrate that loss of GEX3 function can induce haploid formation in rice. Strikingly, 9.82% of selfed Osgex3 progeny were triploid (16/163, triploids/total selfed progeny; Figures 1F and 1G), whereas no triploids were detected among WT progeny (0/354, triploids/total WT progeny). This indicates that triploid induction occurs at a higher frequency than haploid induction in the Osgex3 background. Cytological analysis of triploid Osgex3 meiocytes revealed abnormal chromosome pairing and segregation during male meiosis (Supplemental Figure 3A). Compared with diploid and WT controls, triploid plants developed enlarged caryopses, pistils, and stamens but produced no viable pollen (Supplemental Figures 3B and 3C). Copy-number variation analysis confirmed balanced chromosomal dosage across all chromosomes in both haploids and triploids, thereby ruling out aneuploidy (Supplemental Figures 3D–3F). In contrast to Osgex3, Atgex3 mutants did not produce triploids, suggesting that GEX3-mediated fertilization is conserved but exhibits species-specific differences. These findings suggest that loss of OsGEX3 disrupts genome transmission during fertilization. One possible explanation is that the sperm fusion defect in Osgex3 is not fully penetrant; therefore, the presence of multiple sperm cells around the egg cell may occasionally allow two sperm cells to fuse with a single egg cell. In addition, the polyspermy block in rice may be less stringent than in the dicot species exmined here, in which no triploid progeny were detected. Together, these factors may contribute to the formation of both haploid and triploid progeny. In tomato (Solanum lycopersicum), we identified an apparent single-copy ortholog of AtGEX3, designated SlGEX3 (Solyc03g045080). Using CRISPR–Cas9, we generated a knockout line (Slgex3-1) carrying a 98-bp deletion that introduced a premature stop codon, which is predicted to truncate the protein from 920 to 331 amino acids (Supplemental Figure 4A). Self-fertilized Slgex3-1 plants exhibited increased seed abortion and a higher frequency of undeveloped ovules compared with WT plants, indicating impaired reproductive development (Supplemental Figures 4B–4D). No haploids were detected in the WT tomato population (n = 463). By contrast, among 220 Slgex3 progeny, we identified one haploid, which was confirmed by flow cytometry (Figure 1K). Cytological analysis of root tip cells further validated its haploidy, revealing 12 chromosomes compared with 24 chromosomes in diploid WT controls (2n = 24) (Supplemental Figure 4E). Phenotypically, the haploid plant displayed mild dwarfism, smaller leaves and flowers, and complete sterility (Figures 1H–1J). Together, these results indicate that loss of SlGEX3 function can induce maternal haploid formation in tomato, although at a relatively low frequency under the conditions tested. To investigate whether GEX3 is conserved in legumes, we identified two GEX3 homologs in soybean (Glycine max): GmGEX3A (Glyma.13G047800) and GmGEX3B (Glyma.14G157500). Using CRISPR–Cas9, we simultaneously targeted both loci and generated two independent double-knockout lines, Gmgex3a/3b-1 and Gmgex3a/3b-2. Each line carried frameshift insertions or large deletions that were predicted to disrupt both genes (Supplemental Figure 5). Selfed progeny of the double mutants included dwarf plants that exhibited typical haploid phenotypes, including complete sterility (Figure 1L). Flow cytometry confirmed the haploid genome status of these plants (Figure 1M). The HIR in Gmgex3a/3b mutants was 8.99% (8/89, haploids/total selfed mutant progeny), whereas no haploids were identified among WT progeny (n = 315). These findings demonstrate that GEX3 plays a conserved role in soybean and that its disruption induces haploid formation in this legume species. Collectively, our work identifies loss of GEX3 function as a conserved genetic route for haploid induction across both monocot and dicot lineages, including Arabidopsis, rice, tomato, and soybean. These findings establish GEX3 as a conserved, sperm-specific factor essential for gamete fusion and genomic transmission, supporting its potential as a broadly applicable target for advancing in vivo doubled haploid technologies in diverse crops. Funding This study was supported by the National Natural Science Foundation of China (Grant nos. 32425009, 32130032, 32388101, and 32500293), the National Key Research and Development Program of China (Grant nos. 2022YFF1003500 and 2023YFF1001504), the Strategic Priority Research Program of the Chinese Academy of Sciences (CAS; Grant no. XDB1090000), and the CAS Project for Young Scientists in Basic Research (Grant nos. YSBR-078 and YSBR-134). Acknowledgments The authors declare that they have no conflicts of interest. Author contributions L.W. and J.W. performed the experiments; S.-Y.C. analyzed the data and wrote the manuscript; S.C. and P.-F.J. provided technical assistance for haploid characterization; and W.-C.Y. and H.-J.L. supervised the project and edited the manuscript. Published: May 5, 2026 Footnotes Supplemental information is available at Plant Communications Online. Supplemental information Document S1. Supplemental Figures 1–5, Supplemental Table 1, and supplemental methods Document S2. Article plus supplemental information References Associated Data Supplementary Materials Document S1. Supplemental Figures 1–5, Supplemental Table 1, and supplemental methods Document S2. Article plus supplemental information