Four paternal factors (MTL/PLA1/NLD, DMP, PLD3, and POD65) have been identified from maize. RMZM2G47124, the first-identified maternal haploid-inducing gene cloned from qhir1, was named MATRILINEAL (MTL) [58], PHOSPHOLIPASE A1 (PLA1) [59], and NOT LIKE DAD (NLD) [60] by three different groups. MTL/PLA1/NLD (hereafter referred to as MTL) belongs to the phospholipase family and is an enzyme that hydrolyzes the phospholipids that function in membrane remodeling [61,62,63]. MTL is localized at the pollen endo-plasma membrane, a special plasma membrane that originates in the plasma membrane of vegetative cells and closely surrounds two sperm cells [64]. A 4 bp insertion leads to a frame-shift mutation in the MTL, which results in seed abortion and haploid induction. The second-identified maternal haploid inducer gene is ZmDMP, which encodes a DUF679 domain membrane protein and is specifically expressed in sperm cells. ZmDMP mutation induced an HIR of 0.1~0.3%, but it significantly increased HIR at a five-to-six-fold in combination of MTL mutation [65]. In addition to maize, recent studies have demonstrated that the role of DMP in haploid induction is conserved between monocots and eudicots, which has been achieved in A. thaliana, B. napus [66,67], B. oleracea [68], Medicago truncatula [69], Nicotiana tabacum [67,70], S. lycopersicum [71], and S. tuberosum [72]. In addition to MTL and DMP, vegetative cell-expressed ZmPLD3 and sperm cell-expressed ZmPOD65 were also shown to be able to induce haploids in maize [73,74]. ZmPLD3 belongs to phospholipase D (PLD) family, and it is localized in the ER, plastids, the Golgi apparatus, and the cytosol of vegetative cells [73], whereas ZmPOD65 encodes a peroxidase (POD) protein and is highly expressed in pollen at the tricellular stage [74].

In addition to the paternal factors in haploid embryogenesis, recent reports have demonstrated that egg cell-expressed maternal factors can also be used to induce haploid embryos. EGG CELL-SPECIFIC1/2 (ECS1/2) encodes egg cell-specifically expressed aspartic proteases, which are secreted to the synergid cell region upon fertilization to avoid polytubey by degrading the pollen tube attractant LURE1 [75]. Recently, two independent studies demonstrated that the mutation of ECS1 and ECS2 can also induce haploid embryogenesis [76,77]. Unfused sperm nuclei were observed in zygotes and early embryos, suggesting that karyogamy defects occurred in the sperm–egg fertilization, and the haploids from the ecs1 ecs2 mutant progenies may have resulted from the post-fertilization genomic elimination [77]. In summary, ECS-mediated haploid induction is likely caused by pseudogamy and potential post-fertilization genomic elimination.

Since a low efficiency of haploid embryogenesis is observed in most inducer lines, efficiency has become a major barrier for its application in crop breeding. To improve the efficiency of haploid embryogenesis, the combination of MTL and DMP [65] or MTL and PLD3 [73] was performed to test whether they could improve haploid production. The mutation of PLD3 or DMP in mtl mutant background could significantly increase its haploid induction rate, but the HIR was still lower than the expected. A dmp–mtl–pld3 triple mutant was also created to test whether it could increase the HIR. However, dmp–mtl–pld3 triple-homozygous plants could not be obtained, which was likely due to the pollen developmental defect or the fertilization defect in triple mutants. Hence, it is still worth testing other combinations of haploid inducer genes to improve the efficiency of haploid induction in further studies.

Two mechanisms have been proposed to explain the haploid formation in maize. The first is a single fertilization-induced haploid and the second is post-fertilization genome elimination. The former may produce haploid embryos through the parthenogenesis of the egg cell while the central cell fertilizes normally to form the endosperm. In the latter, double fertilization occurs normally, but the zygote undergoes uniparental genome elimination that results in haploid embryo formation.

The mechanisms for haploid embryogenesis are primarily focused on MTL-induced haploid embryogenesis, and whether it is conserved among different inducers, such as DMP, remains largely unknown. Several recent studies have demonstrated that MTL-induced haploids may form through post-fertilization genome elimination, rather than single fertilization-mediated parthenogenesis. The markers, including B chromosomes and CENH3-YFP derived from the paternal genome, were detected in haploid progenies [78], suggesting that the egg cells were fertilized successfully, and that uniparental genome elimination occurred during haploid induction. In addition, when WT pistils were pollinated by mtl mutant pollen grains (which were carrying the Cas9 and gRNA expression cassette) [79], genome-edited haploids without the Cas9 expression cassette were detected in the progenies. Since the CRISPR/Cas9 system only exists in the paternal genome, this result strongly suggests that the paternal genome is transmitted to the egg cell upon fertilization and is eventually eliminated after fertilization. A multi-omics analysis of mtl pollen grains revealed that ROS signals were involved in post-fertilization genome elimination [74]. Elevated ROS levels may cause DNA damage in pollen from the mtl mutant [80], which may lead to chromosome fragmentation in sperm cells and then induce genome elimination after fertilization. Spermatid chromosome fragmentation in the CAU5 haploid inducer line was detected through single nucleus sequencing [81]. In addition, the in vitro treatment of pollen grains with ROS-inducing reagents can also result in sperm DNA fragmentation and lead to the formation of haploids when pollinated to the WT plants. Studies on the sperm cell-expressed peroxidase gene ZmPOD65 further confirmed the role of ROS in haploid embryogenesis [74]. Peroxidases wildly exist in the plant kingdom [82], and they convert hydrogen peroxide (H₂O₂) into H₂O during the POD catalytic reaction [83]. Therefore, POD65 may be functional in the removal of H₂O₂ in sperm cells, and POD65 mutation can cause H₂O₂ to burst in sperm cells, which leads to sperm DNA fragmentation and, eventually, haploid production.

In summary, vegetative cell-expressed MTL and sperm cell-expressed POD65 are involved in the regulation of ROS levels, as well as the expression of ROS-related genes in pollen grains. Mutations in MTL and POD65 do not appear to impair fertilization, but the zygotes generated from the cross between mtl or pod65 mutants and WT plants will undergo genome elimination to form haploid embryos (Figure 2b). An MTL mutation may cause an imbalance in the hydrolyzed phospholipids (such as PC) between sperm cells and vegetative cells, resulting in the over-accumulation of PCs in sperm cells, which disrupts mitochondrial homeostasis and leads to increased ROS levels [84]. Elevated ROS levels in sperm cells will induce DNA damage and impair the expression of ROS-related genes, leading to the chromatin fragmentation. The fragmented chromatins of sperm cells are then eliminated after fertilization, leading to the formation of haploid embryos. This genome elimination mechanism is different from that in CENH3-mediated haploid embryogenesis. In the MTL- or POD65-induced haploid embryos, chromosome fragmentation occurs before fertilization and the fragmented paternal genome cannot function properly after fertilization and is eventually eliminated, whereas the CENH3-induced post-fertilization genome elimination is largely due to the incompatibility of the parental genome.