Production of double haploid watermelon via maternal haploid induction

Watermelon (Citrullus lanatus, 2n = 2x = 22) is an economically important Cucurbitaceae species that is widely cultivated worldwide. Commercial watermelon cultivars are usually F1 hybrids due to harness heterosis, which used pure lines as parents. However, obtaining pure inbred lines requires at least eight generations of self‐pollinating, which is time‐consuming, resource‐intensive and cost‐prohibitive. By contrast, doubled haploid (DH) technology provides homozygous lines within two generations. Parthenogenesis is the best‐known method to obtain DHs in cucurbits, but this method in cucurbits presents many limiting factors which impede efficient production of haploids, especially in watermelon (Sari and Solmaz, 2021). New approach for producing DH efficiently in Cucurbitaceae is urgent for their application in breeding programs.

Significant advances have been made in DH production in dicots, with the application of a haploid‐inducing (HI) gene, Domain of unknown function 679 membrane protein (DMP) (Zhong et al., 2019), whose homologues have been identified in both monocots and dicots (Zhong et al., 2020). Importantly, mutations in DMP homologues trigger HI and produced DHs in Fabaceae (Wang et al., 2022), Solanaceae (Zhong et al., 2022a,b) and Brassicaceae (Li et al., 2022; Zhong et al., 2022b). However, this approach has not yet been applied in Cucurbitaceae.

Here, we identified six putative DMP‐like genes in the watermelon genome. In accordance with the pipeline proposed for selection of DMP candidate genes for HI (Zhong et al., 2020), we selected ClDMP4 (Cla97C06G121370; the most similar gene to ZmDMP) (Figure S1), which is specifically expressed in male flower buds and pollen (Figure S2). To introduce mutations in ClDMP4, we designed a construct for genome editing via clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR‐associated nuclease 9 (Cas9), targeting the first exon of ClDMP4; the binary construct also included a cassette (Figure 1a) driving the expression of enhanced fluorescence protein (Venus) under the control of the cauliflower mosaic virus (CaMV) 35S promoter for the later identification of haploid seeds. After Agrobacterium (Agrobacterium tumefaciens)‐mediated transformation (Tian et al., 2017), two homozygous cldmp4 mutants with insertion (1 bp) or deletion (2 bp) that resulted in translational frame shifts and premature stop codons were generated in the watermelon inbred line PI179878 background (Figure 1b). Compared with wild type, cldmp4 mutants reduced the number of filled seeds and increased the percentage of aborted seeds, as previously reported.

Figure 1.

Production of double haploid watermelon via maternal haploid induction. (a) The CRISPR/Cas9 mutagenesis vector targeting ClDMP4. (b) Schematic representation of ClDMP4 gene structures. Black blocks, the coding region. Green blocks, the regions encoding the four predicted transmembrane domains (TMs); Red line, the targeted region. (c) Peeled seeds under white light and fluorescent light. The red arrow indicates the non‐fluorescent seeds. (d) Flow cytometry analysis of the diploid and haploid plants. The diploid is labelled “2n”, and the haploid is labelled “n”. (e) Chromosome number in meiotic sample of diploid (n = 22) and haploid (n = 11). (f) Images of haploid and diploid seedlings. (g–k) Phenotypic differences between watermelon diploid and haploid plants (whole plant, leaf, flower bud, male and female flower). (l) Rind phenotype of the offspring in DHs and YJX1 × Cldmp4. (m) Haploid induction rate (HIR) determined by crossing. For crossing, F1 hybrid watermelon plants (YJX1, JM2K and YXF3) were used as the female parent and were pollinated with Cldmp4. Scale bars: 10 μm (e), 5 cm (f and g), 2 cm (h–k) and 10 cm (l)

To investigate whether Cldmp4 mutants can induce maternal haploids when used as the male parent, we used pollen from Cldmp4 mutants to pollinate the F1 hybrid ‘yellow JingXin No.1(YJX1)’. YJX1 is a yellow‐rind hybrid derived from a cross of a dark green female parent and a yellow‐rind male parent, whose offspring display a variety of rind colours. To identify haploid plants among the F1 offspring of the YJX1 × Cldmp4 cross, we looked for green fluorescence in peeled seeds as an indicator of diploid hybrid seeds (Figure 1c). We then grew seeds devoid of fluorescence and genotyped the resulting seedlings with a pair of kompetitive allele‐specific PCR (KASP) markers. KASP markers can distinguish the difference of one nucleotide between two parents, seedlings with the same genotyping as YJX1 are considered as possible haploids. Finally, we confirmed the haploids by flow cytometry (Figure 1e), chromosome counting (Figure 1f) and plant phenotyping (Figure 1g). To further test the maternal origin of these haploids, we analysed 42 single nucleotide polymorphisms (SNPs) that differ between Cldmp4 (generated in the watermelon inbred line GS24, whose rind is striped dark green) and YJX1 using the KASP platform (Yang et al., 2022). We established that none of these haploid seedlings carries SNPs from the paternal parent, as we only detected maternally‐derived SNPs.

Compared to diploid controls, haploid watermelon seedlings were smaller (Figure 1g,h) and produced smaller leaf and reproductive organs than the diploids (Figure 1i–l). In addition, the haploid watermelon plants were male sterile, similar to previously described haploids (Zhong et al., 2020). Haploid plants are typically sterile and their chromosome number needs to double to develop into fertile diploid homozygous plants. We thus applied 25 mg/L oryzalin directly to the shoot apex of haploid watermelon seedlings (Bae et al., 2020). We observed the successful conversion of three haploid seedlings into DH plants, with restoration of fertility. The offspring of these three DHs have stable rind colours, like the dark green or yellow rind parents of YJX1. The F2 offspring of the YJX1 × Cldmp4 cross display a variety of rind colours (Figure 1m).

We further crossed Cldmp4 mutants (as male parents) to two other F1 hybrid watermelon plants (JM2K and YXF3). The average haploid induction rate (HIR) ranged from 0.55% to 1.08% (Figure 1n). These results indicate that dmp mutants can be used for efficient and genotype‐independent maternal DH production in watermelon.

In summary, we developed a simple and cost‐effective DH method with no apparent genotype recalcitrance in watermelon. Homozygous DH plants may directly become new varieties or can be valuable as parents to produce F1 hybrid plants. This report presents the first example of successful DH production by in vivo (seed‐based) haploid induction system in cucurbit crops. Our success in watermelon offers a breakthrough approach that could lead to a new era for cucurbit breeding.

Conflicts of interest

The authors declare that there is no conflict of interest.

Author contributions

S.T., C.W. and Y.X. designed the research; S.T., J.Z., H.Z. and M.Z. performed the experiments; Other authors analysed and interpreted the data; S.T. and Y.X. drafted the manuscript; C.W. critically revised the manuscript. All authors have read and approved the final manuscript.

Supporting information

Figure S1 Phylogenetic analysis of ZmDMP and its homologues in watermelon.

Figure S2 The expression of ClDMP4 in the indicated tissues.