Main

Haploid cells and organisms contain only one set of chromosomes, making them extraordinarily useful for basic research and breeding. By whole-genome duplication of a haploid, a completely homogeneous line, Doubled Haploid, is produced in two generations, while conventional inbreeding often takes more than six or eight generations. Haploid breeding technology therefore dramatically reduces the length and costs of the breeding process. In most crops, inducing haploids efficiently is the primary obstacle. In vitro gametophyte tissue culture techniques have been hindered by laborious procedures, are effective for only a limited number of species and are often genotype dependent1. In recent years, in vivo haploid production by crossing with inducers has received the most attention1. Haploid inducers provide convenience, cost-effectiveness and, most importantly, the ability to function independently of genotypes, allowing haploid technology to be easily integrated into traditional and modern breeding procedures. According to the nuclear genome origin of haploids, haploid inducers are divided into two groups: maternal and paternal haploid inducers. Maternal haploid inducers are widely adopted in breeding because only the genome of the target plants is transmitted to the haploid embryos. In contrast, the cytoplasmic genome of the paternal haploid inducer is transmitted to haploids, which may be beneficial or detrimental depending on the applications of the haploids.

Most haploid inducers work in one direction, maternally or paternally. However, a haploid inducer that was serendipitously discovered during functional studies of CENH3 in Arabidopsis thaliana and GFP-tailswap has capacities to act as a haploid inducer in both directions2. CENH3 is the epigenetic determinant of the centromere and activates the formation of the kinetochore, a protein complex that facilitates faithful chromosome segregation3. Similar to canonical H3, CENH3 shares a conserved histone fold domain, but its amino-terminal tail is highly divergent in both length and amino acid sequence. Because of the critical function of CENH3, a null allele of cenh3, cenh3-1, results in chromosome segregation failure, which leads to embryo-lethal phenotypes4. In an attempt to rescue the cenh3-1 embryo by chimaeric CENH3 constructs, Ravi and Chan developed a novel method to construct haploid inducers in plants2. One of the complement constructs, GFP-tailswap (encoding a chimaeric CENH3 with the N-terminal tail replaced by a GFP-tagged tail of H3.3), restored the embryo-lethal phenotype of cenh3-1 (ref. 2). But when the resulting rescued line, also called GFP-tailswap, was fertilized with pollen from plants with wild-type CENH3, a substantial amount of haploid progeny (20–40%) was identified, which suggested that GFP-tailswap is a highly effective paternal haploid inducer line2. Since then, GFP-tailswap has been widely used as a paternal haploid inducer in Arabidopsis, such as for the construction of immortal mapping populations5, chromosome substitution6, the creation of novel plasmotype–nucleotype combinations7 and facilitating reverse breeding8. However, the cytoplasm of the paternal haploid is derived from the inducer line, which limits its applications in practice.

Besides being a paternal haploid inducer, GFP-tailswap can be used as a maternal haploid inducer, providing high degrees of flexibility. Maternal haploids are applicable to more scenarios because there is no interference of cytoplasmic genetic information. But when GFP-tailswap serves as a pollen donor for haploid induction, the haploid induction rate (HIR) is much lower than when it is used as a paternal haploid inducer, generally not exceeding 5% (ref. 2). Moreover, the pollen vigour of GFP-tailswap is extremely low, making the induction process more challenging2. These two key issues, especially the severe defects of the pollen vigour, lead to GFP-tailswap rarely being used as a maternal haploid inducer. Improving the male fertility of the inducer line would therefore be the first step.

Since many genic male sterilities are temperature dependent9, we wondered whether the male sterility of GFP-tailswap is also sensitive to ambient temperatures. To inspect the effects of ambient temperatures, we performed Alexander staining assays on anthers of wild-type Col-0 and GFP-tailswap plants grown under various temperatures. We found that pollen vigour was perfectly recovered at cool ambient temperatures (18 °C) but declined more severely at relatively warm ambient temperatures (25 °C), whereas wild-type Col-0 did not show any notable differences between these conditions (Fig. 1a). In accordance with pollen vigour, the fertility of GFP-tailswap was remarkably affected by changes in ambient temperature (Fig. 1a). GFP-tailswap exhibits low fertility under conventional conditions with temperatures at 22 °C, in which most siliques produced from self-pollination of GFP-tailswap are non-elongated (Extended Data Fig. 1a) and do not contain normal seeds (Fig. 1b and Extended Data Fig. 1e), while a few of those that emerge late are slightly elongated (Extended Data Fig. 1a) and contain only one or two normal seeds (Fig. 1b and Extended Data Fig. 1e). When we grew GFP-tailswap at different temperatures, we found that a slight reduction of ambient temperatures to 18 °C resulted in impressive fertility recovery in GFP-tailswap, and both the size of the siliques (Fig. 1b and Extended Data Fig. 1a,d) and the number of normally developing seeds recovered to approximately 80% of those of the wild type (Fig. 1b and Extended Data Fig. 1e). By contrast, when ambient temperatures were moderately raised to 25 °C, the fertility of the inducer line was almost completely lost, the siliques did not elongate and most siliques did not contain normal seeds (Fig. 1b and Extended Data Fig. 1d,e). To further confirm the sensitivity of GFP-tailswap to ambient temperatures, we transferred flowering inducer plants between cool and warm conditions and found that the siliques were clearly divided into two groups according to their emergence sequences and the corresponding ambient temperatures (Extended Data Fig. 1b,c). It should be mentioned that for the GFP-tailswap line, whether under cold or warm conditions, a few early emergent siliques are normally sterile (Extended Data Fig. 1a). We then wondered whether growing the inducer plants in cool temperatures could restore the pollen vigour from the first flower. Alexander staining results clearly demonstrated that the pollen vigour of the first flower had already been restored by the low ambient temperatures (Extended Data Fig. 2a), even though the filaments remained short, which was probably responsible for the sterility of the first silique (Extended Data Fig. 2b). We therefore concluded that the pollen vigour and the fertility of GFP-tailswap can be restored by cool temperatures.

Fig. 1: The pollen vigour and the fertility of GFP-tailswap is restored by cool temperatures.
figure 1

a, The viability of pollen grains of Col-0 and GFP-tailswap under different conditions was tested by coloration with Alexander staining; the purple staining indicates that the pollen is viable. Around the tenth flower on the stem was used for the Alexander staining experiments. Scale bar, 200 μm. The experiments were repeated three times independently, with similar results. b, The fertility of Col-0 and GFP-tailswap under different conditions as revealed by the split of green mature siliques. Scale bar, 1 mm. c, gl1 Ler (22 °C) plants were fertilized by recovered pollen from GFP-tailswap (18 °C). The gl1 Ler plants were then transferred to low-temperature conditions (18 °C), and the siliques were opened to reveal the developing seeds. Scale bar, 1 mm. d, The frequencies of haploids, aneuploids and diploids in offspring resulting from the outcrosses described in c. Sterile offspring with narrow leaves, small flowers and glabrous phenotypes were scored as haploid; fertile wild-type offspring were scored as diploid; and offspring with developmental defects were scored as aneuploid. n indicates the number of progeny scored. The experiments were repeated two times independently, with similar results.

Source data

Full size image

The first key issue concerning using GFP-tailswap as a maternal haploid inducer thus appears to have been resolved. We then used the recovered pollen to pollinate a trichomeless mutant, gl1 (Ler background), grown in normal conditions (22 °C) to determine whether the recovered pollen can be used in maternal haploid induction. The gl1 Ler mutant has a more compact inflorescence than the GFP-tailswap plants (Col background). After outcrossing, we found that most of the seeds survived (Fig. 1c) and around 1.6% (Fig. 1d) of the offspring showed trichomeless and compact inflorescence phenotypes (Extended Data Fig. 3a), which normally are haploids; some of these were confirmed by flow cytometry analysis (Extended Data Fig. 3b,c). The HIR at lower temperatures is much lower than that previously reported (5%) (ref. 2), which suggests that, similarly to the male sterility of GFP-tailswap, haploid induction might be sensitive to ambient temperatures.

GFP-tailswap is easily and commonly used as a paternal haploid inducer. To further explore the effect of ambient temperatures on haploid induction, we performed the haploid induction at different temperatures by using gl1 Ler pollen to pollinate inducer flowers. According to previous studies, the HIR and abortion rate of hybrid seeds of haploid induction are positively correlated10. We first checked the developmental status of hybrid seeds of haploid induction before calculating the HIRs. We found that hybrid seed development is highly sensitive to moderate changes in temperature (Extended Data Fig. 4a). On the basis of the shrink levels, the seeds were divided into four groups: dark inviable seeds, wrinkled seeds (size < 1/2), wrinkled seeds (size > 1/2) and plump viable seeds. We found that the abortion rate of hybrid seeds increases with a rise in temperature (Extended Data Fig. 4b). By combining trichomeless, compact inflorescence and haploid typical phenotypes, we further confirmed that the haploid induction ability of GFP-tailswap is highly sensitive to ambient temperatures as expected, with HIR rising from 17.5% at 18 °C to 22.1% at 22 °C and 77.0% at 25 °C (Fig. 2a).

Fig. 2: The induction ability of CENH3-based haploid inducers is temperature sensitive, with temperature effects occurring after fertilization.
figure 2

a, The proportions of diploid, aneuploid and haploid offspring from outcrosses with inducers pollinated by gl1 Ler. The HIRs of GFP-tailswap and GFP–CENH3 were tested at 18 °C and 25 °C using a Percival (AR-41L3) growth chamber, and the HIR of cenh3-8 was tested at 18 °C, 25 °C and 30 °C using a Ruihua (AR-1200) growth chamber; all tests at 22 °C were performed in a greenhouse. The experiments were repeated eight, three and two times independently for the GFP-tailswap, GFP–CENH3 and cenh3-8 assays, respectively. All experiments showed the same trends. b, The proportions of haploid, diploid and aneuploid offspring from outcrosses with GFP-tailswap fertilized by Ler (22 °C) with ambient temperature switch. The temperatures before and after the arrow indicate the growth conditions of GFP-tailswap before and after hybridization, respectively. The experiments (18 °C and 25 °C) were performed with Percival (AR-41L3) growth chambers, but the pollen was collected from gl1 Ler plants grown in a greenhouse at normal temperatures (22 °C). The experiments were repeated two times independently, with similar results.

Source data

Full size image

In addition to chimaeric GFP-tailswap, the direct fusion of GFP at the N terminus of CENH3 (GFP–CENH3) nearly fully rescued cenh3-1 mutants, indicating that GFP–CENH3 has little impact on the function of CENH3 (ref. 2). However, the resulting GFP–CENH3 line still showed a weak haploid induction ability when it was used as the female parent under normal conditions2. Similarly, we found that the paternal haploid induction ability of GFP–CENH3 responds to ambient temperatures; the HIR increased from 0% at 18 °C to 1.9% at 22 °C and 4.3% at 25 °C (Fig. 2a).

Besides complementation of the null allele of cenh3, numerous studies have demonstrated that CENH3 subjected to point mutations or fragment deletions could convert the plants into haploid inducers10,11,12. Here we show that the haploid induction abilities of GFP-tailswap and GFP–CENH3 are highly sensitive to ambient temperatures. We thus wondered whether other CENH3-based inducers are also susceptible to temperature changes. Using the CRISPR–Cas9 technique, we obtained a cenh3-8 allele with a 27-amino-acid deletion at the N-terminal tail of CENH3 (Extended Data Fig. 5a,b). Both vegetable tissues and pollen vigour showed no apparent differences from wild-type plants under normal conditions (Extended Data Fig. 5c,d). We further determined that changes in growing temperatures do not appear to affect the fertility of cenh3-8 plants (Extended Data Fig. 5d). A previous study showed that the expression of CENH3 depends on the intragenic regulatory elements within the second intron13. In accordance with the deletion of the first and second introns in cenh3-8 (Extended Data Fig. 6a), CENH3 expression levels decreased in cenh3-8 at all temperature conditions (Extended Data Fig. 6b). All these results indicate that cenh3-8 could be a weak mutant allele of CENH3. Similar to the previous study4, the self-fertilization of heterozygous cenh3-1 produced 75.32% (n = 158) normal seeds (Extended Data Fig. 7), indicating that either copy of parental CENH3 is enough to support seed development. Thus, if cenh3-8 is a weak mutant allele, the reciprocal crossing between cenh3-1/+ and cenh3-8 would rescue the seed phenotype in both directions. As expected, when cenh3-8 plants were used as pollen donors, 97.59% (n = 166) of the seeds developed normally (Extended Data Fig. 7). Surprisingly, when cenh3-8 plants were used as pollen receivers, only 78.79% (n = 165) of the seeds developed normally (Extended Data Fig. 7), which indicates that at least the maternal cenh3-8 displays severe function defects during seed development.

We therefore examined the cenh3-8 line to determine whether it has the capacity to induce haploids and whether it reacts to ambient temperatures. The cenh3-8 mutant plants were grown at different temperatures, and the plants were then pollinated with pollen from gl1 Ler and kept growing under the same conditions. Our results indicate that no haploid could be identified at ambient temperatures of 18 °C and 22 °C, but at 25 °C, we obtained an HIR of 13.6% (Fig. 2a), which further increased to 28.6% at 30 °C (Fig. 2a). These results indicate that a mutant allele such as cenh3-8 without haploid induction capacity at normal temperatures could gain haploid ability at warm ambient temperatures. According to these results, higher temperatures could be used as a common factor to improve the haploid induction ability of the CENH3-based inducer lines.

Here we demonstrate that the haploid induction ability of CENH3-based inducers is often ambient temperature dependent. This finding is further supported by a recent report that a weak allele, cenh3-4 (with a G-to-A transition in the splicing donor site of the third exon of the CENH3 gene, leading to a substantial decrease in CENH3 at centromeres), gains haploid induction ability at high temperatures14. These findings therefore have the potential to be directly utilized to enhance paternal haploid production by CENH3-based haploid inducers. However, as previously reported10, we showed that haploid induction capacities are positively correlated with seed abortion rates, possibly diminishing the final effectiveness of ambient temperatures in haploid induction. To validate the enhanced effects of warm temperatures on haploid induction of GFP-tailswap, we collected all the seeds from five siliques of each outcross between GFP-tailswap and pollen donors of Ler plants at cool and warm conditions and divided them into groups with different shrink levels as mentioned previously. Each group of seeds was germinated on 1/2 Murashige and Skoog plates and then transplanted in nutrient soil; the ploidy level was analysed separately. We found that seeds germinated even in the group that was considered dark inviable (Extended Data Table 1). Finally, we obtained 15 haploid plants in cool conditions, and most of them (10 of 15) originated from well-developed plump seeds (Extended Data Table 1). The increase in temperature to 25 °C led to more than double the production of haploid plants, of which the majority (23/33) were derived from severely wrinkled (size 1/2) seeds (Extended Data Table 1). However, when the temperatures were further increased to 30 °C, all the resulting seeds were classified as dark inviable, and hardly any of them from around 68 siliques (26 seeds per silique by counting 6 siliques) germinated. These results show that elevating the ambient temperatures within a certain range can effectively increase the haploid induction efficiency of GFP-tailswap. In addition, for inducer lines such as GFP-tailswap that are completely male sterile at high temperatures, this method provides extra advantages because it eliminates the process of emasculation, which reduces the difficulty of outcrossing (especially for crops such as rice, wheat and soybean, which are difficult to emasculate).

We have therefore identified methods that can enhance haploid induction by manipulating ambient temperatures. However, with respect to the two issues of maternal haploid induction, the ambient temperatures present a pair of conflicting (yin-yang) effects on GFP-tailswap. It is therefore essential to determine whether these two effects of ambient temperatures can be separated—in other words, whether ambient temperatures affect the ability to induce haploids in early embryos, directly or in some way through gametophyte inheritance. To address this concern, instead of performing the entire assays under the same conditions, we switched the growth conditions between cool and warm temperatures right after the GFP-tailswap plants had been pollinated. The GFP-tailswap plants grown at warm (25 °C) and cool (18 °C) temperatures were emasculated, and the next day they were pollinated with gl1 Ler pollen collected from plants grown at normal temperatures (22 °C). Following pollination, half of the plants remained under the original conditions, and the rest were immediately transferred to different temperatures to continue growing. We found that the haploid induction ability of GFP-tailswap is principally affected by the ambient temperatures of post-pollination rather than pre-pollination. For inducer plants originally grown at cool temperatures (18 °C), the HIR was approximately 8.0% when the pollinated inducer plants remained at 18 °C, whereas it increased to 79.7% when the plants were moved to 25 °C (Fig. 2b). As for inducer plants grown at warm temperatures (25 °C), the HIR was around 80.5% when the pollinated inducer plants remained at 25 °C, whereas it decreased to 13.1% when the plants were moved to 18 °C (Fig. 2b). The haploid induction capacity of GFP-tailswap is therefore probably influenced by ambient temperatures directly on the formation of the haploid embryos.

Compared with paternal haploids, maternal haploids without the interference of the cytoplasmic genome are highly demanded in breeding. GFP-tailswap has the ability to induce maternal haploids, but with a very low HIR compared with that of paternal haploid induction. The serious defect of pollen vigour of the inducer plants also makes it hard to utilize this line in practice. Our study demonstrates that ambient temperatures present a pair of yin-yang effects on GFP-tailswap inducer plants regarding maternal haploid induction. Cool temperatures increase pollen vigour while depressing HIR, and warm temperatures do the opposite. We further proved that these yin-yang effects are separated, which motivated us to design a strategy to address the two key issues (low pollen vigour and low HIR of maternal haploid induction of GFP-tailswap) by simply switching the ambient temperatures. The target gl1 Ler plants were grown at normal temperatures (22 °C) to flowering, and the inducer plants were grown at cool temperatures (18 °C) to harvest recovered pollen. After emasculation, the gl1 plants were pollinated by recovered inducer pollen, and the pollinated plants were then moved to higher temperatures (22 °C and 25 °C) for haploid induction (Fig. 3a,b). We found that compared with the HIR of 1.6% for plants always kept at cool temperatures (18 °C) (Fig. 1c), the HIR was increased to 7.0% for post-pollinated plants moved to normal temperatures (22 °C), and it was further increased to 24.8% for post-pollinated plants moved to warm temperatures (25 °C) (Fig. 3c). The finding of yin-yang effects of ambient temperatures on the haploid inducer therefore led us to develop a simple strategy to solve the two key limitations of GFP-tailswap for maternal haploid production.

Fig. 3: Utilizing the yin-yang effects of ambient temperatures to optimize maternal haploid induction.
figure 3

a, gl1 Ler (22 °C) plants were fertilized with recovered pollen from GFP-tailswap (18 °C) and then transferred to 22 °C or 25 °C. Siliques were split open to reveal the developing seeds. Scale bar, 1 mm. b, The proportions of haploid, diploid and aneuploid offspring resulting from the outcrosses described in a. The experiments were repeated two times independently, with similar results.

Source data

Full size image

In our study, we found that the enhancement of haploid induction by higher ambient temperatures could be a common feature of CENH3-based haploid inducers. Attempts to develop haploid inducers based on CENH3 in crops have not been as successful as expected, and the inducers that have been produced achieve relatively low levels of haploid induction capability15,16,17. Accordingly, in most scenarios, low HIR is the first issue to be addressed in terms of haploid induction. We believe that the enhancement of HIR by increasing ambient temperatures offers important potential for optimizing the CENH3-based haploid induction method, and it would be worth examining whether ambient temperatures have a similar effect on other haploid induction systems, including those based on MTL and DMP. Nevertheless, when the HIR is optimized to a high level, a balance must be struck between HIR and seed abortion rate. For instance, we observed almost no germination when utilizing GFP-tailswap for inducing paternal haploids at 30 °C.

In light of these findings, we provide methods to optimize the process of developing efficient haploid inducer lines in crops from at least two perspectives. First, defects in CENH3 may result in sterility issues, limiting the acquisition of lines with high induction capability. In this context, it would be beneficial to develop and gather potential lines at low ambient temperatures. Second, due to the high sensitivity to ambient temperatures, some potential lines could be overlooked when verifying induction ability. It is therefore worthwhile to test haploid induction ability under various conditions. In summary, the findings presented here will guide us in developing new effective CENH3-based haploid inducers for crops in the future.

Our findings not only are beneficial to haploid production in practice but also provide substantial insights into the mechanisms related to genome elimination. CENH3 is well known to play a crucial role in maintaining the faithful separation of chromosomes during both mitosis and meiosis. Functional defects of chimaeric CENH3 in the GFP-tailswap line allow plants to acquire haploid induction capacity along with sterility of male gametophytes. In spite of this, we know little about how these processes are regulated in plants. Zhu et al. demonstrated that slowing development could be a general mechanism for temperature-sensitive genic male sterility (TGMS) lines to undergo sterility–fertility conversion18. According to Zhu et al., res1, a weak allele of cdka;1, restores male fertility in many TGMS mutants18. Low temperatures were thought to slow the progression of male meiosis in a similar manner as res1. By crossing res1 with GFP-tailswap, we obtained GFP-tailswap inducers with homozygous res1 backgrounds. We found that res1 is unable to restore the male sterile phenotype of GFP-tailswap (Extended Data Fig. 8). This result indicates that ambient temperatures might act differently between GFP-tailswap and the pollen-wall-related TGMS mutants.

Here we show that the sterility and the haploid induction ability of GFP-tailswap are temperature dependent, and most importantly these effects are separated. Our results therefore suggest a different mechanism for haploid induction between CENH3-based haploid inducers in Arabidopsis thaliana and Stock-6-derived haploid inducers in maize. In maize, it is believed that the chromosome fragmentation observed in sperm cells of CAU5 (a maternal haploid inducer line derived from Stock 6) is responsible for both the sterility of male gametophytes and the subsequent formation of female haploids19,20,21,22,23. Similarly, Tan et al. found that the elimination of chromosomes inherited from the GFP-tailswap plants during embryonic cell division could result from DNA damage in these chromosomes24. But they showed that the truncated or shattered chromosomes observed in their induction assays were associated with mitotic errors in the early embryo instead of inherited from the sperm cell24, which coincides with our observation that haploid cell formation in the early embryo is directly influenced by ambient temperatures. In addition, the newly identified cenh3-8 presents new ways to investigate the function of the fast-evolving N-terminal tail of CENH3 (Extended Data Fig. 9), which has been reported to be involved in loading the protein during meiosis4,25,26. However, cenh3-8 (with a 27-amino-acid deletion at the highly diverse N-terminal tail) grows normally and fully fertile. Moreover, a complete deletion (Δ11) of the highly conserved αN helix (the junction of the N and C termini) was reported to fully rescue the loss of function of cenh3-1 during meiosis10. Therefore, even if the whole N-terminal tail of CENH3 is necessary for meiosis4,25,26, this process appears to be tolerant of fragment deletions at both ends of the N-terminal tail of CENH3 (Extended Data Fig. 9). Regarding haploid induction, Δ11 presents an impressive HIR (8–25.7%), while cenh3-8 shows effective HIRs only at high temperatures, which is probably due to mutations occurring at the CENH3 domains with different levels of conservation (Extended Data Fig. 9). The mechanism underlying all these phenomena remains a mystery. Further investigations are required to determine the function of each domain of CENH3 and how ambient temperatures coordinate with CENH3 to regulate chromosome segregation during meiosis and mitosis.

Methods

Plant materials and growth conditions

Arabidopsis thaliana seeds for GFP-tailswap2, GFP-CENH3 (ref. 2), res1 (ref. 18) and Ler gl1-1 (ref. 24) were described previously. The plants were grown under long-day conditions in a cultivation room (16 h light/8 h dark) at 22 °C. Cool (18 °C) or warm (25 °C) temperature treatments were performed in Percival (AR-41L3) and Ruihua (AR-1200) plant growth chambers.

CRISPR–Cas9-mediated generation of the cenh3-8 mutant

The CRISPR–Cas9 plasmid was modified on the basis of pHEE401E27. The guide RNA of the CENH3 was designed by CRISPR-GE28; the target sequence was target 1: 5′-CGTTACCAGGTCACAACCT-3′, target 2: 5′-CCGACAAGGAGAGGCGGTGA-3′. The final plasmid was transformed into Agrobacterium GV3101 and then transformed to Arabidopsis Col-0 by the inflorescence dip method. T1 plants were screened for the CENH3 locus by agarose gel electrophoresis and Sanger sequencing, and the Cas9-free T2 seeds were selected by GFP fluorescence. The primers for cenh3-8 identification are F: 5′-GAGTGTCGAGCGGGAAAGTA-3′ and R: 5′-GGCATAGCCTGTCTGGATCT-3′. The DNA of all samples was extracted by the CTAB method29 to identify the genotypes.

Real-time quantitative PCR

To analyse the difference in the expression of CENH3 in Col-0 and cenh3-8 at different temperatures, we collected the inflorescences of Col-0 and cenh3-8 grown at 18 °C, 22 °C and 25 °C for expression detection. We first used Total RNA Extractor (Sangon Biotech, B511311) extraction reagent to extract RNA from the sample, and we then used HiScript II RT SuperMix for the quantitative PCR (qPCR) (Vazyme, R223) kit to synthesize complementary DNA. Real-time qPCR was performed on an Agilent AriMx Real-Time PCR System using AceQ qPCR SYBR Green Master Mix (Vazyme, Q111). PUX7 was used as an endogenous control for normalization. The positions of the primers are shown in Extended Data Fig. 6a, and the primers used for qPCR are F: 5′-TTCTTATTCCGGCTGCCAGT-3′ and R: 5′-ATCTTCTGCCGCCTCTTGAA-3′.

Alexander staining

Flower buds were collected from the samples, the anthers were isolated and placed on slides, and 50 μl of Alexander stain solution (Solarbio, G3050) was added to the slides. The slides were incubated overnight at 4 °C in the dark and were then evaluated using a stereo microscope (Olympus, SZX16).

Crossing scheme and haploid screening

For the hybrid combination used in this study, the inducers were GFP-tailswap, GFP–CENH3 and cenh3-8, and the induced plants were the wild type of Ler or the trichomeless mutant gl1 in the Ler background. The flowers were emasculated at night and pollinated on the morning of the third day. Successful hybrids were harvested two to three weeks after pollination. The siliques were dried in a 28 °C incubator for one week, and then the seeds were collected. The proportion of aborted seeds was counted under a microscope. Diploid, haploid and aneuploid plants were confirmed by phenotype as described above2,12.

Flow cytometric ploidy measurements

The nuclei of rosette leaves were isolated and subjected to flow cytometry using a method described previously30. In short, we prepared MgSO4 solution (1.23 g MgSO4·7H2O, 1.85 g KCl and 0.6 g HEPES dissolved in 480 ml ddH2O, pH-adjusted to 8.0 using KOH and then brought to a final volume of 500 ml with ddH2O) before the experiment, then put a petri dish on ice and added 400 μl of Aru buffer (9.65 ml MgSO4 buffer, 100 μl 1 M DTT and 250 μl Triton X-100). We put about 25 mg of young rosette leaves washed with ddH2O into the Aru buffer, and the nuclei were released by three-minute razor cutting. All mixtures were transferred to a 30 μm cell strainer (BD Falcon 352235), and the nuclei in the culture dish were washed with 200 μl of Aru buffer and collected into the cell strainer. We then added 2.5 μl of 20 mg ml−1 RNase A and 5 μl of 10 mg ml−1 propidium iodide (Sangon Biotech, E607306) solution to a 5 ml collection tube, mixed it well and placed it in the dark at 4 °C for 30 minutes. The samples were run on a Beckman CytoFLEX LX flow cytometer. Propidium iodide emission was detected with the Y585-PE-A filter; around 10,000 events were counted for each sample.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.