Haploid induction technology has been used in crop breeding since decades. The major advantage of generating haploid plants is that a genome can be doubled in a single step to generate doubled haploid plants whose genome is genetically fixed at each locus. This avoids the time and money consuming processes to generate homozygous plants required for plant breeding and seed production, which otherwise have to be generated after many cycles of inbreeding from diploid heterozygous material1-5. Compared with in vitro methods involving, for example, culture of explants containing haploid cells like isolated microspores, anthers or ovule, whose regeneration to plants is often limited to certain crops and genotypes, respectively, in vivo haploid induction techniques have the potential to be more widely applied and thus could significantly improve breeding efficiency and praxis4-5.
Currently, in vivo haploid induction techniques are based on three major strategies: (i) induction of parthenogenesis, (ii) application of paternal inducer lines and (iii) uniparental genome elimination. Transgenic approaches have demonstrated that ectopic expression of a zygotically expressed gene BABY BOOM LIKE1 (BBML1) encoding an AP2 transcription factor that is normally activated after fertilization from the paternal genome already in the egg cell is sufficient to trigger parthenogenesis in rice6. This approach has been transferred to other cereal crops like maize7 and thus indicates that a single gene can be used to generate haploid embryos in grass crops. Similarly, in the dicot dandelion, expression of PARTHENOGENESIS (PAR) encoding a zinc finger EAR-domain protein in egg cells can trigger embryo development without fertilization. Recently, it was reported that expressing PAR in egg cells of lettuce can also induce parthenogenesis8. A popular technique in maize to generate double haploids is based on paternal Stock6-derived in vivo haploid inducer lines. The underlying gene was recently identified by three groups and named as ZmPLA1/MTL/NLD (Zea mays PHOSPHOLIPASE A1/MATRILINEAL/NOT LIKE DAD) encoding a sperm cell-expressed phospholipase lacking the very C-terminal domain and thus is incapable to be located in the sperm plasma membrane. If this mutated version is expressed in sperm cells, it could be used to trigger haploidization in maize9-12. The system has been successfully transferred to other cereals like rice and wheat13,14, but the molecular mechanism of haploidization remained unclear. Similarly, mutants of ZmDMP8/9 encoding members of the conserved DUF679 domain membrane protein family that are highly expressed in pollen were used for haploid induction, but the molecular mechanism is unclear15,16. Uni-parental chromosome elimination was reported in CENH3-mediated haploid induction protocols. Mutations in either maternal or paternal CENH3 was reported to lead to a reasonable number of haploids, for example, in Arabidopsis, maize, wheat and cotton17-20. However, there are still problems associated especially with male inducer lines: e.g. some haploid inducer lines produce abnormal pollen that fail to germinate and thus affect the seed setting rate, and paternal inducer lines are still limited to a very few crops. Thus, due to the application limits of current haploid technology, plant breeders are highly interested in methodological improvements as well as in novel principles of haploidization21.
Here, we report the identification of a novel female haploid inducer system based on mutations in egg cell-specific genes. ECS1 and ECS2 encode aspartic endopeptidases that are exclusively expressed in the unfertilized egg cell of Arabidopsis22. Pollen development and pollen germination of the ecs1 ecs2 double mutant appears normal like those from wild type (WT) plants. However, selfed ecs1 ecs2 double mutant plants contained about 15% abnormal seeds. We distinguished two types of abnormal seeds: (i) undeveloped seeds slightly larger than unfertilized ovules and (ii) developed but small seeds (Fig. 1a, b). By using a whole-mount ovule/seed clearing technique, we found that undeveloped seeds were arrested at the zygote stage. After fertilization (fusion of egg and sperm cell) these zygotes became vacuolized and finally degenerated (Fig. 1c, d). To make sure that they were indeed zygotes, we labelled the egg cell membrane of ecs1 ecs2 double mutants using the DD45::GFP-LTI6b fusion construct (Supplementary Fig. 1) and carefully calculated the egg cell and sperm cell fusion rate. We found that sperm cells carrying the nuclear marker HTR10-RFP could perfectly fuse with all egg cells (n = 481; Fig. 1e, f). However, occasionally we found an unfused male nucleus inside the egg cell cytoplasm (Fig. 1e; Supplementary Fig 2). 3D image analysis of these zygotes clearly showed a sperm nucleus in the egg cell cytoplasm confirming successful gamete fusion (Supplementary Fig. 2, Supplementary movie. 1). We next analyzed small sized seeds and found aborted embryos showing abnormal cell division planes. When WT embryos developed to the eight-celled stage, embryos from the ecs1ecs2 double mutant divided less frequent and cell division planes appeared irregular. These abnormal embryos finally arrested at different early developmental stages (Fig. 1g, h).
Furthermore, among the offspring of selfed ecs1ecs2 double mutants, we found plants with small flowers and short siliques. We then geminated hundreds of seeds and checked the ploidy level of seedlings by flow cytometry. Among 149 randomly selected ecs1 ecs2 double mutant offspring plants we identified two haploid plants (Fig. 2a). At anthesis haploid plants contained small flowers and generated short siliques (Fig. 2b, c). To further confirm the haploid nature of these plants, chromosome spreads were generated from male meiocyte cells at meiosis I showing that the haploid plants contained five chromosomes compared to 10 in diploid WT plants (2n=10) (Fig. 2d).
Although we have never found a single fertilization phenotype in the double mutants as mentioned above, we tested the possibility that haploid plants originate from fertilization failure due to excessive sperm cell number. We used tes mutant pollen possessing more than two sperm cells23 to pollinate ecs1 ecs2 double mutants and also WT pistils as a control. Among 190 progeny plant derived from ecs1 ecs2 double mutants, we obtained two haploid plants, while haploids were not detected among 154 WT offspring plants. Haploidy was first confirmed by flow-cytometry (Fig. 2e). Haploid plants contained smaller flowers and generated shorter siliques compared with WT plants (Fig. 2f). Their haploid nature was further confirmed by the chromosome spread technique. Experiments revealed that haploid plants contained only five chromosomes in their somatic cells (Fig. 2g). In short, after double checking haploid plants were obtained at a rate of 1.2%.
To explore the origin of haploids, we pollinated ecs1 ecs2 plants expressing the egg cell membrane marker DD45::GFP-LTI6b24 with pollen carrying the sperm nucleus marker HTR10-RFP. In WT plants, karyogamy (fusion of gamete nuclei) usually occurs shortly after egg cell and central cell each fuse with a sperm cell. Thereafter, zygotes start to elongate showing a typical zygote morphology (Fig. 3a). In ecs1 ecs2 double mutants we observed semigamy (a type of fertilization in which the sperm cell fuses with the egg cell, but its nucleus does not fuse with the egg nucleus) at 24 hours after pollination (HAP) (Fig. 3b, Supplementary Fig. 3, Supplementary movie 2). Few semigamous zygotes (2/1500) divided at 24 HAP, but still kept egg cell morphology and did not show zygote elongation. The male nucleus could yet be seen after semigamous zygote division (Fig. 3c, Supplementary Fig. 4, Supplementary movie 3). These findings indicate that sperm entry is sufficient to trigger cell division. A similar phenomenon has been reported in fertilized central cells, in which sperm nuclei did not fuse with polar nuclei, but central cell division could be promoted25 Divided semigamous zygotes may eventually generate haploid plants. Furthermore, to understand the process of haploidization, we labelled the chromosomes of ecs1 ecs2 egg cells and zygotes by using the DD45::CENH3-GFP fusion construct17 (Supplementary Fig. 5). We then crossed ecs1 ecs2 plants expressing CENH3-GFP with pollen from WT plants. At 12 and 24 hap, we observed aneuploidy (6-8 chromosomes in mutant zygotes compared with 10 chromosomes in the control) (Fig. 3d-f, Supplemental movie 4-6). This indicates that paternal and/or maternal chromosome elimination occurs at the beginning of embryogenesis. The higher rate of obtained haploid plants (1.2%; see above) compared with divided semigamous zygotes (0.13%) suggests that haploidization also occurs in ecs1 acs2 zygotes/early embryos when karyogamy was initially successful.
In summary, we found that in addition to its reported role in degrading the pollen tube attractor LURE1 and thus to prevent polytubey22, ECS1 and ECS2 endopeptidases possess additional roles after gamete fusion. We found that haploid offspring of the ecs1 ecs2 double mutants could be generated after forming semigamous zygotes and/or by chromosome elimination during early embryogenesis. These phenotypes suggest that ECSs play critical roles in fertilization to ensure successful fusion of male and female nuclei and parental chromosome incorporation. Although the detailed molecular mechanism underlying the process of nuclear fusion is still not known and further ECS1 ECS2 target proteins need to be identified, we found that ecs1 ecs2 mutant plants can be used as female haploid inducer lines, which adds to the tools of generating double haploid plants. Notably, 3.5% haploid progeny plants were also observed in another report studying ecs1 ecs 2 double mutants indicating that the haploidization rate could be further increased. However, we did not observe single fertilization and parthenogenetic activation of unfertilized egg cells as suggested in that report (BioRxiv, 476184 (2022); doi: https://doi.org/10.1101/2022.01.20.476184) and think it is a misinterpretation. Currently, different in vivo paternal haploid inducer lines are widely used for haploidization, especially in maize, which provide powerful tool for crop breeding. Staging ecs1 ecs2 mutations with other inducers like dmp8 dmp9 and/or pla1/mtl/nld might significantly increase the haploid induction rate as well as manipulation of the post-fertilization process that is regulated by ECS1 ECS2 activity. In addition, among the haploid offspring derived from ecs1 ecs2 mutants, we found that it is easy to classify seeds or seedling into different groups according to their morphology to enrich potential haploid plants and therefore to greatly accelerate the selection and confirmation process. Thus, the female haploid induction and selection system reported here provides a novel strategy of haploidization for breeding and will now be tested in crop plants.