A previous study reported a Japanese radish (Raphanus sativus L.) landrace, ‘cv. Nishimachi-Riso’, which contains a GRS1 mutation (Ishida et al., 2015). As the genotype of GRS1 was not fixed within the ‘cv. Nishimachi-Riso’ population, the genotype of GRS1 was determined using DNA markers (see 2.2 Genotyping) and plants carrying heterozygouse GRS1 (GRS1/grs1) were designated as AKO lines. Kale (Brassica oleracea var. acephala) inbred line ‘KK45-2’ is a collard-type kale whose leaves contain 4MSOB used as the paternal parent. All the plants were grown in plastic pots (diameter of 210 mm) in a greenhouse. A Nippi-engei-baido (Nihon Hiryo Co.,Ltd, Gunma, Japan) based soil was used and 1/1000 HYPONeX (HYPONeX JAPAN CORP., LTD., Osaka, Japan) solution was applied once a week. To obtain hybrids, the anthers were removed from radish buds 1–3 d before flowering and were pollinated with kale pollen on the day of flowering. Raphanobrassica seeds were sown on September 5, 2020, and genomic DNA was purified from the true leaves two weeks later and used for genotyping. For genomic DNA purification, DNeasy 96 Plant Kit (QIAGEN, Venlo, Netherlands) was used. The plants were then planted in a field on August 31, 2020, with a gap of 35 cm between plants and 60 cm between rows, at the Institute of Vegetable and Floriculture Science (34°46′N, 136°25′E; Tsu, Mie, Japan).

Previous studies have shown that insertion into the first exon of GRS1 causes complete functional defects (Kakizaki et al., 2017). To detect the insertion into GRS1, primers were designed at the genomic positions shown in Figure 2A , and polymerase chain reaction (PCR) was used to amplify the region between Rs270 (5′-GCAGGAGAGGATGCTTGAAGG-3′) and Rs271 (5′-TGAAACCTTACCCCAAAACG-3′) for the functional type, Rs270 and Rs272 (5′-TCCAGGTTGGGATAGCTTGT-3′) for the defective type. PCR was performed under the following conditions: initial denaturation at 94°C for 1 min, cycling at 94°C for 15 s for heat denaturation, annealing at 60°C for 15 s, and extension at 72°C for 50 s for 32 cycles. The amplified PCR products were separated on a 2% agarose gel, and the functional and defective types were distinguished based on differences in fragment length ( Figure 2B ). To confirm that the kale genome was inherited by the hybrid, PCR was performed using primers Bo-Fw (5′-CTAGTATGAGGACTCGTTCAGTTACCTCCCTTAGCAGC-3′) and Bo-Rv (5′-GTTTCTTAGAATATGGTGATTGCTGGCTT-3′) to amplify UDP-sulfoquinovose synthase (LOC106306866), which is located on chromosome C1.

Leaf, root, stem, and flower bud are used for GSL analysis ( Figure 3 ). For leaf analysis, 10 cm from the tip of the 20 cm long leaf was collected, and the central vein was removed. Three leaves per plant were collected and analyzed as one bulk sample. For root analysis, a section was cut 5 cm below the stem/hypocotyl border in the form of a disk (thickness, 0.5–1.0 cm) and collected. For bud analysis, apical flower buds were collected. For stem analysis, a 10 cm section below the apical flower bud of the main stem, which was elongated after bolting, was collected.

After sampling, each part was frozen in liquid nitrogen and dried using a lyophilizer (Labconco, Kansas City, MO, USA). The dried samples were crushed using a multi-bead shocker (Yasui Machinery, Miyazaki, Japan); 0.1 g of ground samples were weighed, mixed with 4.8 mL of 80% methanol and 0.2 mL of 5 mM 2-propenyl GSL (Sigma-Aldrich, St. Louis, MO, USA) as an internal standard and shaken for 30 min at 25°C. After centrifugation at 3,000 rpm for 10 min, the supernatant was collected, GSLs were adsorbed onto DEAE-Sephadex A-25 (Sigma-Aldrich, St. Louis, MO, USA) and desulfonated using arylsulfatase (Type H-1, EC 3.1.6.1, Sigma-Aldrich, St. Louis, MO, USA) at 25°C for 18 h. Desulfo-GSL solutions eluted ion-exchange water were used as samples for high-performance liquid chromatography (HPLC) analysis (LC-20A; Shimadzu Corp., Kyoto, Japan). A reverse-phase column (COSMOSIL 5C18-II, 150 × 4.6 mm; Nacalai Tesque Inc., Kyoto, Japan) was used at 30°C and a flow rate of 1.5 mL/min. The mobile phase comprised 20% acetonitrile, and detection was performed using UV light with a wavelength of 229 nm. The GSL molecular species were estimated based on the retention time of the peaks according to our previous report (Ishida et al., 2015). The individual GSL contents were calculated by the ratios of the individual desulfo-GSL peak areas to the peak areas of an internal standard, 2-propenyl GSL (Sigma-Aldrich, St. Louis, MO, USA), and a response factor (The International Organization for Standardization, 1992). Table 1 shows the list of GSLs analyzed, and Figure 4 shows representative chromatograms of each desulfo-GSL molecular species detected using HPLC.

To generate intergeneric hybrids that harbored the grs1 allele, we used a radish AKO line that was heterozygous for the GRS1 gene (GRS1/grs1) as a seed parent. The kale KK45-2 line, containing 4MSOB in mature leaves, was used as the pollen parent. At the time of crossing, all stamens were removed from radish buds 1–2 d before flowering to avoid self-fertilization, and pollination was performed using kale pollen on the day of flowering. Twenty-one AKO plants were pollinated using KK45-2 pollen, and the pod formation rate varied from 0 to 0.24 among radish plants ( Table 2 ). Similarly, there was a difference in the number of seeds per pod among the AKO plants. Among the 21 combinations, the seeds of four radish plants (AKO103, AKO108, AKO110, and AKO118) that yielded a large number of seeds were sown in petri dishes, and their germination rates and genotypes were investigated ( Table 3 ). The germination percentage ranged from 71.4–97.1%, and 47–87 plants were obtained from each combination. To confirm that the obtained plants were hybrids, PCR was performed using primers specific for kale UDP-sulfoquinovose synthase (LOC106306866). This confirmed the inheritance of the kale genome. No amplification of the kale genome was detected in only four plants obtained from AKO110, and these plants were inbred radish plants. Genotyping using the GRS1 marker ( Figure 2 ) resulted in a 1:1 match for GRS1 segregation in all combinations. No plants heterozygous for the GRS1 marker were identified. Based on these results, we obtained hybrids of radish and kale, and GRS1 segregation followed theoretical values. These data allowed us to evaluate the relationship between the GRS1 genotype and 4MSOB quantity in the Raphanobrassica population.

The obtained hybrids were planted in a field on September 5, 2020, and the GSL composition in the true leaves (20 cm length) was analyzed 72 d after planting. Growth was vigorous, and leaf shape was similar to that of radish ( Figures 3A, B ). No plants with pollen fertility were observed for any of the cross combinations. The root shape was snarled and the main root branched into several branches ( Figure 3C ). The timing of bolting was earlier than that of the parents, and the shape of the flower buds was similar to that of kale ( Figure 3D ). The 4MSOB content in the true leaves of intergeneric hybrids and parent plants is shown in Figure 5 . Notably, in all cross combinations, the grs1-type had a 4MSOB content approximately twice as high as that of functional GRS1 hybrids ( Table 4 ). Furthermore, these grs1-type plants had higher 4MSOB concentrations than those in kale KK45-2 and radish AKO103 (grs1/grs1) plants. There were significant differences in the concentration of 4MSOB between GRS1 genotypes, but even within the same GRS1 genotype the concentration of 4MSOB varied widely ( Figure 5 ). The average of 4MSOB in four lines in GRS1-type and grs1-type were 15.5 ± 0.5 and 34.1 ± 1.0 µmol·g^-1 dry weight in leaves, respectively. The highest 4MSOB concentration was 69.4 µmol·g^-1 dry weight. In grs1-type Raphanobrassica, 4MSO3B was almost undetectable and the accumulation of 4MTB was detected ( Figure 6 ). The contents of indolic GSLs, such as Indol-3-ylmethyl GSL (glucobrassicin, I3M), were not affected by the GRS1 mutation.

The grs1 mutation increased 4MSOB concentration by approximately 2-fold in the leaves of the hybrids. Next, we analyzed the GSL profiles in various edible plant parts such as roots, flower buds, and stems ( Table 5 ). The 4MSOB content was highest in the flower buds, followed by stems and roots. The grs1 mutation increased the concentration of 4MSOB by more than 2-fold in flower buds and stems, similar to that in leaves. The major GSLs in the roots were 4MT3B and 4MTB, whereas those in the flower buds and stems were 4MSO3B and 4MSOB ( Figure 7 ). 2-Hydroxy-3-butenyl GSL (progoitrin, 2H3B) and I3M, which are rarely detected in the roots, were detected in flower buds and stems. Regarding the total GSL, the GRS1-type plants had a higher content than that in the grs1-type, similar to that in the leaves. The total GSL content was highest in flower buds for both GRS1 and grs1-types, followed by roots and stems.

Raphanobrassica, an intergeneric hybrid between the genera Raphanus and Brassica, contains large amounts of 4MSOB and 4MSO3B (Schutze et al., 1999). Lyophilized powder derived from Raphanobrassica inhibits Helicobacter pylori-induced gastritis in Mongolian gerbils (Yamada et al., 2014). Although Raphanobrassica has useful properties, the hybridization rate between Raphanus and Brassica is considerably low. In 1973, McNaughton reported that the number of seeds per pollinated flower between R. sativus and B. oleracea is between 0.38 and 0.4 (McNaughton, 1973). This value is comparable to that of the most efficient combination in the crossing experiment of the present study (AKO118 × KK45-2, 0.34). Further, the formation rates of hybrids differ between varieties (Kakizaki, 1925). This phenomenon is known as the “hybridization barrier” and is divided into two types: pre-zygotic and post-zygotic barriers. The pre-zygotic barrier is caused mainly by defects in fertilization, such as interspecific incompatibility and defects in pollen tube guidance (Dresselhaus and Marton, 2009). The post-zygotic barrier includes hybrid embryo breakdown and hybrid sterility. Candidate genes or quantitative trait loci responsible for these barriers have been cloned (Udagawa et al., 2010; Tonosaki et al., 2013). In the present study, the hybridization rates were markedly different between AKO plants originating from the same radish ‘cv. Nishimachi-Riso’ ( Table 2 ). It is not known whether the barriers observed in the present study are pre- or post-zygotic, but differences in their degree within the same species may provide good material for genetic analysis.

The hybridization between a radish and Chinese cabbage produces a cultivar called ‘Baemoochae’ (Lee et al., 2011). Baemoochae contains GSL molecular species present in both radish and Chinese cabbage. However, the total GSL content is similar to that of both parents (Nugroho et al., 2020). In contrast, the total GSL content of Raphanobrassica produced in this study was higher than that of either parent, suggesting that the entire GSL synthesis pathway was activated. It is reported that the GSL synthesis gene expression is elevated in resynthesized Brassica allotetraploids compared with that in their diploids (Zhang et al., 2015). Further studies are needed to determine whether the increased 4MSOB in Raphanobrassica developed in this study is solely due to mutations in GRS1 or is also influenced by the increased expression of other biosynthetic enzyme genes. GSLs are hydrolyzed by myrosinases to not only ITCs but other products (thiocyanates, nitriles, or epithionitriles) depending on pH and the presence of epithiospecifier protein (Bones and Rossiter, 1996). Distribution of degradated products vary widely between plant species. B. oleracea has a higher proportion of nitriles and epithionitriles than ITCs, whereas R. sativus has a higher proportion of ITCs (Cole, 1976). Therefore, analysis of the abundance ratios of degradation products in Raphanobrassica, which has both genomes, is of great importance for its use as a functional vegetable.

Hybrids containing grs1 showed almost no 4MT3B or 4MSO3B content in their leaves ( Table 4 and Figure 6 ). This result supports the hypothesis that the introduction of grs1 into Raphanobrassica enhances 4MTB utilization in 4MSOB synthesis and increases the 4MSOB content. However, in all cross combinations, the total content of each GSL in the population possessing grs1-type tended to be lower than that in GRS1-type plants ( Figure 6 ). Therefore, increased 4MSOB content in plants lacking GRS1 function may cause feedback inhibition of FMO. For example, YUCCA, an enzyme in the auxin synthesis pathway, also belongs to the monooxygenase family, similar to FMO. YUCCA transcription levels are negatively regulated by the synthetic product auxin (Suzuki et al., 2015). If FMO activity and transcription levels are negatively regulated by 4MSOB, understanding the underlying mechanisms and applying them in breeding could lead to higher 4MSOB content.

4MSOB is present in the genera Brassica, Eruca, and Raphanus (Ciska et al., 2000; Ishida et al., 2014). A study in 1992 reported that sulforaphane (ITC derived from 4MSOB) in broccoli functions as a major inducer of anti-carcinogenic defense enzymes. Consequently, the recognition of broccoli as a representative vegetable containing 4MSOB has increased, which has prompted research in various fields. To develop broccoli varieties with high 4MSOB content, MYB28 of the wild species B. villosa (BvMYB28) has been introduced into cultivated species, and the F₁ hybrid Beneforté^® has been cultivated (Traka et al., 2013). In Arabidopsis, AtMYB28 positively regulates aliphatic GSL biosynthesis (Sonderby et al., 2007). The expression of several aliphatic GSL genes is elevated in broccoli, which is homozygous for BvMYB28. The 4MSOB concentration in the homozygous BvMYB28 inbred broccoli line is 20 µmol·g^-1 dry weight in floret (Neequaye et al., 2022). In contrast, the average value of 4MSOB in the leaves of Raphanobrassica with grs1 grown in this study was 34.1 µmol·g^-1 dry weight ( Table 4 ). Although simple comparisons cannot be made because the analyzed sites contained the highest contents in adult plants and the content varies with the growing region and cultivation method, the variety with defective GRS1 in Raphanobrassica may be used in various cuisines as a new leafy vegetable that contains as much 4MSOB in broccoli Beneforté^®. Compared with kale, the taste of Raphanobrassica leaves is softer and less gruel-like, making it suitable for a variety of dishes such as stir-fry and salads. The low hybridization affinity between radish and kale and the inability to ensure a commercial level of seed production represent issues that need to be addressed to popularize the variety in the future. However, chromosome doubling may be used to breed seed-fertile allotetraploid Raphanobrassica (Niimi et al., 2015).