Tomato (Solanum lycopersicon) is the second largest vegetable crop and the largest fruit crop (Costa and Heuvelink, 2018). Progress in tomato breeding has been achieved by classical breeding, introgression of traits found in related wild Solanum species and exploiting heterosis in F₁ hybrid crosses (Lin et al., 2014). These approaches require the development of inbred lines to reduce or largely eliminate heterozygosity. Classically, multiple rounds of selfing or backcrossing are used to generate inbred lines (Gale, 1980), but homozygous lines can also be obtained in a single generation using doubled haploid (DH) technology (Jacquier et al., 2020). However, tomato is highly recalcitrant for haploid induction (HI) (Seguí-Simarro, 2010).

Significant breakthroughs in DH production were made after identification of the zmpla1/mtl/nld and zmdmp mutant genes that induce in vivo maternal haploid embryos in maize and extension of this technology to other monocot crops (Jacquier et al., 2020 and references therein). The utility of dmp mutants for HI in dicots was demonstrated in Arabidopsis (Zhong et al., 2020), but it is not known whether this approach can be applied in dicot crops. Here we show that dmp mutants can also be used for efficient and genotype-independent maternal HI in tomato.

The tomato genome contains a single DMP gene (SlDMP; Solyc05g007920) that is highly expressed in pollen and flower buds (Zhong et al., 2020). We generated sldmp mutants in cv. Ailsa Craig using a CRISPR-Cas9 mutagenesis construct that includes the FAST-Red marker for haploid identification (Zhong et al., 2020, Figure 1a). Homozygous or biallelic sldmp mutants with insertions and/or deletions that resulted in translational frame shifts and premature stop codons were generated (Figure 1b). Compared with wild type, sldmp mutants significantly reduced the number of filled seeds (Figure 1c,d) and increased the percentage of both aborted seeds and undeveloped ovules (Figure 1e), as previously reported (Zhong et al., 2020).

We determined whether sldmp mutants can induce maternal haploids after selfing. In the absence of segregating molecular markers, we first identified putative haploid plants based on their phenotype, that is, smaller organs and sterility (Zhong et al., 2019, 2020). Among 55 T₁ seedlings, one plant (Figure 1f) showed the typical haploid phenotype (Figure 1g–j). This plant was confirmed to be a true haploid by flow cytometry (Figure 1k). Our data suggest that sldmp mutation facilitates in vivo haploid embryo development in tomato.

Next, we crossed four wild-type F₁ female plants (listed in Figure 1o) from different genetic backgrounds with sldmp mutants to determine whether sldmp pollen can also induce maternal haploids upon outcrossing. Ten haploids derived from these crosses were first screened by molecular markers and confirmed by chromosome counting, flow cytometry and plant phenotype (Figure 1l–o). To confirm their maternal origin, three of these haploid seedlings were used for whole-genome resequencing. Single-nucleotide polymorphism (SNP) analysis showed that none of the seedlings carried paternally derived SNPs, suggesting that sldmp induces ‘clean’ maternal haploids (Figure 1p).

The Arabidopsis FAST-Red marker is expressed in the embryo and endosperm and can be used in crosses to distinguish between diploid seeds derived from double fertilization (marker expression in the embryo and the endosperm) and maternal haploid seeds (marker expression only in the endosperm) (Zhong et al., 2020). To determine whether FAST-Red can be used to identify tomato haploids, we analysed FAST-Red expression in seeds from a DF199 × dmp cross. Imbibed seeds were first classified into red and white seed groups based on their colour under white light. Under fluorescent light, the red seeds showed weak RFP expression in the endosperm and strong RFP expression in the embryo, while 70% of white seeds showed weak RFP expression in the endosperm and no RFP expression in the embryo (Figure 1q). Some white seeds showed RFP expression in the embryo and endosperm under florescent light and were recategorized as red/RFP-expressing seeds. These two groups were confirmed by checking root tip RFP expression (Figure 1q). The embryos in red seeds were scored as putative diploids, while the white seeds with weak RFP expression in the endosperm and no RFP expression in the embryo were scored as putative maternal haploids. Ploidy analysis of 218 putative haploid and 2303 putative diploid seedlings showed that FAST-Red can be used with 100% accuracy for selection of maternal haploids in tomato (Figure 1r).

Next, we used FAST-Red marker for haploid seed selection in crosses between diverse tomato genotypes and sldmp FAST-Red lines. The haploid induction rate (HIR) after crossing 36 different female genotypes with the sldmp inducer lines ranged from 0.5% to 3.7%, with an average HIR of 1.9% (Figure 1s). These data suggest that sldmp mutants can be used for genotype-independent HI.

To summarize, we demonstrate that sldmp mutants induce in vivo maternal haploids in a major dicot crop, tomato, and that identification of haploid embryos is facilitated by the FAST-Red marker. Given the presence of DMP-like genes in dicot species and the ability of both Arabidopsis and tomato dmp mutants to induce maternal haploids (Zhong et al., 2020), it is highly likely that dmp haploid inducers can be generated in other dicot crops. Extending this system would represent a major advance over in vitro haploid production, especially for members of the Solanaceae, Fabaceae and Cucurbitaceae that are recalcitrant for DH production (Hooghvorst and Nogués, 2020).