Enhancing health-promoting isothiocyanates in Chinese kale sprouts via manipulating BoESP

Abstract Glucosinolates (GSLs) are a group of sulfur-containing secondary metabolites, which are abundant in Brassica vegetables. GSL breakdown products (GBPs), especially isothiocyanates (ITCs) benefit human health. Chinese kale is a native Brassica vegetable in China, and its sprouts are rich in GSLs and nutritional substances. ITCs are the predominant GBPs while alternative products are formed in the presence of specifier proteins. However, fewer ITCs are formed in the sprouts. Epithiospecifier (ESP) promotes the formation of epithionitriles at the expense of ITCs in Arabidopsis, but a systematic study of different isoforms of ESPs in most vegetables is still missing. In this study, changes in the content of GBPs and the precursor GSLs, as well as thiols per plant were monitored during sprout development. The proportions of epithionitriles and ITCs in total GBPs were found to be increased and decreased, respectively. RNA-seq showed enhanced expression of numerous genes involved in GSLs biosynthesis and degradation, as well as sulfur assimilation in sprouts compared to seeds. Four copies of BoESPs were isolated and BoESP2 was the most abundant isoform. Generally, transcription of BoESPs showed a strong response to abscisic acid and gibberellin, and consequently epithionitriles increased under these treatments. Knockdown of BoESP2 expression through virus-induced gene silencing system could effectively increase total ITCs and decrease total epithionitriles. Overall, dynamic GSL metabolic flux exists in the sprouting period, and the expression of BoESPs determines the pattern of GBPs, suggesting that improving the health-promoting ITCs in Chinese kale sprouts through manipulating BoESPs by metabolic engineering is feasible.


Introduction
Sprout development starts from seed germination, which is an important stage in the plant life cycle. In this process, storage materials such as proteins, fats, and starches are degraded to small molecule compounds [1]. After this process, well developed sprouts are even more nutritious than the original seeds, which makes sprouts a popular vegetable [2]. Chinese kale (Brassica oleracea var. alboglabra) is a native Chinese Brassica vegetable, whose tender leaves and bolting stems are usually harvested for food. In recent years, Chinese kale sprouts are favored by producers and consumers because they are economical and rich in health-promoting compounds such as carotenoids, vitamin C, and glucosinolates (GSLs) which can be hydrolyzed by degradation enzymes to release breakdown products (GSL breakdown products, GBPs), conferring them additional health-promoting properties [3,4]. Lots of work has been carried out to investigate the concentration of GSLs/GBPs at different plant developmental stages [5][6][7][8][9][10]. However, because the water content in sprouts at different developmental stage is different and is much higher than that in seeds, it is worth exploring the change of GSLs and GBPs per plant during sprout development to accurately report the metabolism of GSL and better access the health-promoting value of sprouts. To date, only a few studies have paid attention to the change in GSL content per plant during sprout development [11][12][13], and the change in GBP content per plant has not been reported.
GSLs are a group of plant secondary metabolites containing sulfur and nitrogen, which are mainly found in Brassicaceae plants, especially ecologically important Brassica crops. According to their amino acid precursors, GSLs can be classified into aliphatic GSLs (derived from methionine, alanine, leucine, isoleucine, and valine), indole GSLs (derived from tryptophan), and benzenic GSLs (derived from phenylalanine and tyrosine). Upon tissue damage, a group of beta-glucosidases termed myrosinases (thioglucoside glucohydrolases, TGGs) interact with GSLs and hydrolyze them into glucose and unstable aglucone [14]. The unstable aglucone undergoes spontaneous rearrangement and forms isothiocyanates (ITCs). Besides ITCs, thiocyanates, nitriles, and epithionitriles can be formed in the presence of thiocyanateforming proteins (TFPs), nitrile specifier proteins (NSPs), and epithiospecifiers (ESPs). Among them, nitriles can be formed in the presence of NSPs, which is independent of the alkenylation of GSL structure, or in the absence of NSPs at low pH and high ferrous ion concentration [15][16][17]. ESP catalyzes alkenylated GSLs to epithionitriles and alkylated GSLs to nitriles [15,18]. The wellknown bioactivity of GSLs in plant defense and human health promotion is conferred by their hydrolysis products, particularly ITCs [19][20][21][22]. However, we found the GBPs in Chinese kale sprouts are mainly epithionitriles instead of much healthier ITCs in our previous study, indicating high ESP activity [4]. So far, the metabolic pathways of GSLs in model plant Arabidopsis have been fully studied [23]. However, in vegetables, research mainly focused on the biosynthesis of GSLs, such as clarifying the structural genes and regulatory factors (Miao, 2020), while the hydrolysis of GSLs has received limited attention. The function of ESP has been clarified in Arabidopsis since its first separation from Crambe abyssinica [24][25][26][27]. As Brassica species experienced a whole-genome triplication and very recent genome duplications after their divergence from Arabidopsis [28], the function of each ESP isoform might be more complex. In vegetables, so far partial ESP isoforms have been characterized in several B. olearacea cultivars, and their function has been analyzed through prokaryotic expression, or heterologous overexpression in Arabidopsis [18,[29][30][31]. Cloning and systematic study of all ESP isoforms in most vegetables are still missing.
In this study, we investigated the change of GBP pattern per plant from seed to 7-day-old seedling of Chinese kale to better evaluate the health-promoting value of Chinese kale sprouts. The underlying mechanisms were ascertained through studying GSL turnover and whole-genome transcriptome analyses. Also, the systematic characterization of all ESP isoforms in Chinese kale sprouts was carried out and in sprouts higher ITCs but lower epithionitriles were obtained via manipulating GBP pattern through virus-induced gene silencing (VIGS)-mediated gene silence of BoESP.

Epithionitriles increase during sprout development
A total of 15 kinds of GBPs were detected in seeds and seedlings of Chinese kale ( Fig. 1a; Table S2, see online supplementary material). There were seven kinds of ITCs, five kinds of nitriles, and three kinds of epithionitriles. Generally, the compositions of GBPs were consistent in seeds and different stages of seedlings, except for the lack of 4-(methylsulfinyl) butyl ITC in seeds. In seeds, 3-butenyl ITC was the predominant one (0.044 μmol/plant, accounting for 55.4% of total GBPs), followed by 2-propenyl ITC (0.013 μmol/plant, accounting for 16.0% of total GBPs). After germination, 1-cyano-3,4-epithiobutane became the richest GBP, ranging from 0.105 to 0.231 μmol/plant, accounting for 34.3% to 50.4%. The second abundant one was 3-butenyl ITC in soaked seeds (d0) and early seedlings (1-day-old to 3-day-old, d1 to d3), and 1-cyano-2, 3-epithiopropane in late seedlings (d5 to d7), respectively.
Overall, the content of individual and total GBPs varied during sprout development. The content of total GBP increased sharply by 247.9% (seed to d0) and 75.8% (d0 to d1) in very early seedling stage, and then stabilized at a high level until d7, except for a remarkable decrease of 27.4% showing up in d3 seedlings. The content of total epithionitriles and nitriles varied similarly to the total GBPs, while total ITCs went up first but then went down, with a peak at d1 (Fig. 1b). However, the proportion of ITCs in total GBPs dropped dramatically from seed (76.5%) to d0 (32.2%), and then showed a wavelike decrease during the rest days, accounting for 15.4% in d7 seedlings. In contrast, the proportion of epithionitriles increased along with seedling development, accounting for 71% in d7 seedlings (Fig. 1c).

Dynamic changes in the content of glucosinolates and thiols during seedling development period
To explore why the content of epithionitriles kept rising and ITCs showed a declining trend during seedling development, we measured the content of precursor GSLs and thiols which correlate closely to GSLs. As shown in Fig. 1d and Table S3 (see online supplementary material), 12 GSLs were detected in seeds and all different ages of seedlings, including eight aliphatic GSLs and four indole GSLs. No benzoic GSL was found. During seedling development, aliphatic GSLs were always the predominant GSL, accounting for 92.1%∼98.9% of total GSLs (Fig. 1f). Among them, the content of alkenyl GSL was 1.563 to 2.750 times higher than that of alkyl GSL (Fig. 1e), with gluconapin being the richest one (accounting for 39.0% to 51.0%), followed by sinigrin and glucoerucin (Fig. 1d).
From seed to d7, total GSL level in seeds plummeted to 0.134 μmol/plant after being soaked for one day, which was only 69.3% of that in seeds (Fig. 1d). It then maintained the upward trend overall, and finally reached 0.248 μmol/plant in d7 seedlings. The changes in the content of total aliphatic GSL and alkenyl GSL were similar to that of total GSLs, and alkyl GSL content changed little except for an increase at d1 (Fig. 1e). The content and proportion of indole GSL remained unchanged from seed to d1, and then kept rising ( Fig. 1e and f).
The backbone of GSLs harbor two to three S atoms [32]. Primary sulfur metabolites, cysteine and glutathione, offer reduced sulfur in the biosynthesis of GSL. The hydrolysis of GSL, in turn, provides sulfur for the biosynthesis of primary sulfur metabolites [33,34]. In this study, during seedling development, the content of cysteine went up first but then went down, with a peak at d3 (5.4 nmol/plant) (Fig. 1g). From seed to d7, the change of glutathione content was wavelike. The lowest level was observed in d0 seedlings (0.04 μmol/plant), while the highest one was in d3 seedlings (0.18 μmol/plant) (Fig. 1h). Taken together, these results indicated that dynamic sulfur metabolic f lux exists during seedling development.

Transcriptome analysis of GSL metabolism-related genes in Chinese kale seeds and seedlings
To further investigate the potential mechanism of GSL turnover, we conduct RNA-seq in seeds and d3 seedlings, as the accumulation of both cysteine and glutathione was highest in d3 seedlings, and there was a noticeable difference in GBP pattern between seeds and d3 seedlings (Fig. 1). A total of 20 667 differentially expressed genes (DEGs) were found, including 12 508 up-regulated and 8159 down-regulated genes (Fig. S2a, see online supplementary material). Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis revealed that many genes were enriched in signaling pathways closely to photosynthesis and energy metabolism, while 'cysteine and methionine metabolism' and 'sulfur metabolism' were also included in the top 20 of KEGG pathways (Fig. S2b, see online supplementary material).
The DEGs involved in GSL turnover were paid close attention, namely GSL biosynthesis and degradation, as well as the sulfur assimilation pathway. Overall, a total of 32 DEGs related to GSL biosynthesis were identified, and most of them showed higher expression level in d3 seedlings than in seeds (Fig. 2a). Among them, two DEGs (IPMDH1 and BCAT3) were involved in side-chain elongation of aliphatic GSL, and 19 DEGs involved in the core structure biosynthesis of GSL, including nine DEGs participated in  Table S1 (see online supplementary material).
both aliphatic and indole GSL biosynthesis, and three and seven DEGs were linked to aliphatic GSL and indole GSL, respectively. In addition, 11 DEGs were involved in the side-chain modification of GSL, including aliphatic GSL-related FMO GS-OX , indole GSL-related CYP81F and IGMT. For sulfur assimilation pathway, we identified a total of 14 DEGs that almost covered the whole pathway, including ATPS, APK, SIR, OASTL, CGS, CBL, GSH1, and GSH2. All of them had higher expression level in d3 seedlings than in seeds, except for GSH1-1 (Fig. 2a).
For classical GSL degradation pathway, a total of 13 DEGs were identified, including eight TGG1s, three ESPs, and two NSPs ( Fig. 2b). Almost all identified TGG1s showed higher expression levels in d3 seedlings than in seeds, which might facilitate more myrosinases accumulation in d3 seedlings than in seeds. ESP and NSP have been reported to promote the formation of epithionitriles and nitriles [35]. Here, only the expression of ESP was observed higher in d3 seedlings than in seeds, indicating that ESP might be the key to explaining higher content of epithionitriles and nitriles in d3 seedlings than that in seeds. Taken together, these results confirmed our findings of the dynamic metabolic f lux at the transcriptional level, and ESP might be the vital gene affecting the pattern of GBPs in Chinese kale sprouts.

Four copies of BoESP were isolated from Chinese kale
The putative homologous ESP protein sequences from cabbage, namely BoESP1 cabbage (Bol039072), BoESP2 cabbage (Bol006380), BoESP3 cabbage (Bol024137) and BoESP4 cabbage (Bol013374), were identified through the BLASTP program in BRAD by using Arabidopsis ESP amino acid sequence as a query sequence. Then, four copies of ESP, being named BoESP1, BoESP2, BoESP3, and BoESP4, were isolated from Chinese kale by referring to the four ESP homologous sequences of cabbage. The coding sequences of BoESP1, BoESP2, and BoESP3 were 1029 bp, and BoESP4 was 1044 bp (Fig. S3, see online supplementary material). The multiple sequence alignment analysis showed that the protein sequences of the four BoESPs were 100% identical with the corresponding ESPs from cabbage, and more than 75% identical with AtESP (Fig. S4, see online supplementary material). Even the identity between every two BoESPs was more than 80%. The phylogenetic tree was constructed based on ESPs from Arabidopsis, Chinese kale, and six species in U's triangle (Fig. S5, see online supplementary material). BoESP1, BoESP2, and BoESP3 were grouped into a subgroup, which was separated from BoESP4. Generally, BoESPs were grouped most closely with the ESPs from B. oleracea, Brassica napus, and Brassica carinata, rather than Brassica nigra, Brassica rapa, and Brassica juncea.

Subcellular localization and spatio-temporal expression pattern of BoESPs
To investigate the possible cellular location that individual BoESP function in, we carried out transient expression in tobacco leaf cells (Fig. 3a). Generally, when BoESPs were expressed in GFP fusions under the control of the Caulif lower mosaic virus (CaMV) 35S promoter, all proteins distributed throughout the cytoplasm and the nucleus. These results were confirmed by heterologous overexpression of BoESP-GFP in Arabidopsis (Fig. 3b). The expression of BoESP2 was the highest in almost all stages of seedling development, and it was much higher in d3 seedlings (approximately 8.75-fold) than in seeds, which was in accordance with the results of RNA-seq. The expression of BoESP1, BoESP3, and BoESP4 was quite low during sprout development (Fig. 3c). The expression of all four BoESPs in different developmental stages of young plant and different organs of mature plants showed that BoESP2 was the isoform with the highest expression, followed by BoESP1, except that BoESP3 was strongly expressed in the root of the mature plant. Besides, the overall expression of BoESP4 was extremely weak and was undetectable in inf lorescence and seed pod (Fig. 3d).

The expression of BoESPs and the pattern of GBPs in response to plant hormones
The analysis of cis-acting elements in the promoter region of ESPs was carried out to better understand the regulatory mechanisms involved in ESP expression. As the coding sequences of ESPs exhibit extreme conservation between Chinese kale and B. oleracea, the promoter sequences of B. oleracea from BRAD were employed to conduct the analysis. As shown in Fig. 4a and Table S4 (see online supplementary material), abundant hormone-responsive cis-elements were identified in the promoter region of all four ESPs from cabbage with a difference in quantity and variety. Specifically, there existed abscisic acid (ABA) and methyl jasmonate (MeJA)-responsive cis-elements in the promoter region of BoESP2 cabbage , BoESP3 cabbage and BoESP4 cabbage ; gibberellin (GA)-responsive cis-elements in BoESP1 cabbage , BoESP3 cabbage and BoESP4 cabbage ; auxin-responsive cis-elements in BoESP1 cabbage , BoESP2 cabbage , and BoESP4 cabbage ; salicylic acid (SA)-responsive ciselements in BoESP3 cabbage . It is noteworthy that the promoter region of all four genes contained many light-responsive ciselements.
We then experimented to verify the response of BoESPs to plant hormones by exogenous application of ABA, GA 3 , and SA on Chinese kale sprouts, as they represent the richest cis-element in the promoter of BoESP2 cabbage , the cis-element that exists in the promoter of other three ESPs from cabbage, and the one that only exists in the promoter of BoESP3 cabbage . As shown in Fig. 4b, the expression of four BoESPs homologs could be induced by ABA treatment, with BoESP1 being the strongest one. The expression of BoESP1 was increased significantly by 41.519-and 26.020-fold, respectively, at 12 h and 48 h after ABA treatment. The transcription level of BoESP2 was significantly increased after 48 h treatment of ABA, which was 2.211 times higher compared to the control. The expression of BoESP3 responded positively to ABA treatment at the early stage (before 12 h). Compared to mock, BoESP4 showed a higher expression level upon ABA treatment, but it is not statistical significance. For GA 3 treatment, the expression of BoESPs could be induced generally. The transcription levels of BoESP2, BoESP3, and BoESP4 were significantly enhanced by 14.17%, 19.22%, and 10.64%,  Next, we measured the changes in GBPs in response to ABA and GA 3 treatment. Both hormone treatments did not affect the GBPs profiles but changed the content of individual GBPs considerably in Chinese kale sprouts (Table S5, see online supplementary material). After ABA and GA 3 treatment, the predominant 1-cyano-3,4-epithiobutane was remarkably increased by 1.554-and 0.414-fold, respectively, while the accumulation of total epithionitriles was also notably accelerated by 1.271-and 0.513-fold; the content of total nitriles was significantly boosted by 1.346and 0.497-fold, respectively (Fig. 5a, see online supplementary material Table S5). However, the content of total ITCs was decreased markedly by 48.72% and 56.88%, respectively (Fig. 5a). In general, ABA treatment promoted the formation of total GBPs while GA 3 did not. We also calculated the proportion of each kind of GBPs in total GBPs. Upon ABA and GA 3 treatment, the proportion of epithionitriles was increased by 44.44% and 39.86%, respectively, while ITCs was decreased by 67.67% and 59.97%, respectively (Fig. 5b). The proportion of nitriles was increased significantly upon ABA treatment. Overall, our findings suggest ABA and GA 3 regulate the pattern of GBP by affecting the expression level of BoESPs.

VIGS-mediated gene silencing of BoESP enhances the accumulation of ITCs and reduces the content of epithionitriles
To ascertain the function of ESP in GBPs formation in Chinese kale sprouts, we employed the VIGS system to generate Chinese kale sprouts with decreased expression of BoESPs. A batch of material with decreased expression of BoESPs was obtained, and those with BoESP2 being decreased by approximately 50% were selected for further detection (Fig. 6a). GBPs detection showed that BoESP2-silenced sprouts had comparable content of total GBPs to the control (Fig. 6b). Specifically, the contents of total ITCs were significantly increased by 15.03%, and total epithionitriles were decreased by 34.37% in BoESP2-silenced sprouts, while no significant change was observed in total nitriles. From the perspective of proportion, total epithionitriles in BoESP2-silenced sprouts consumed a smaller proportion in total GBPs when compared to the control, whereas the proportion of both total nitriles and ITCs was higher (Fig. 6b). Thus, these results indicate that BoESP2 plays an important role in the formation of epithionitriles, but not nitriles.

Discussion
In this study, we measured the content and composition of GBPs throughout sprout development and calculated the level per plant. Generally, sprouts form more ITCs and total GBPs than seeds ( Fig. 1a and b). Considering the constantly rising biomass of sprouts, d1 sprouts are recommended to harvest for providing ITCs or total GBPs as fewer of them can generate more GBPs. To date, extremely few studies have been devoted to monitoring the changes of GBPs during the sprouting period and a unified conclusion has not been formed yet. In broccoli, Williams et al. (2008) reported a general decrease in sulforaphane and sulforaphane nitrile concentration (μmol/g FW) during the first 7 days of seedling development, while Gu et al. (2012) observed that the concentration of sulforaphane (mg/g DW) went down sharply during the first day of seed germination, then went up and reached the same level as that in seeds at 48 h before decreasing again [36,37]. In the current survey, the content of total ITCs per plant showed an inverted V-shaped trend throughout the whole Chinese kale sprouting period. It might be the different representation methods that resulted in inconsistent conclusions among different studies. To our surprise, more epithionitriles and nitriles were observed in sprouts than in seeds, and both the content and proportion of epithionitriles exhibited an increasing trend. This is not beneficial to Chinese kale sprout quality because ITCs have stronger potential health-promoting effects [19,20,22]. In contrast to GBPs, the content of precursor substances GSL dropped sharply from seeds to d0 seedlings ( Fig. 1d and e). This is in agreement with the observations in several Brassica crops [6]. Correspondingly, only a trace level of cysteine was detected and the content of glutathione dropped sharply. It might be because GSLs and glutathione were hydrolyzed to provide substrates for the synthesis of life's basic chemicals to ensure survival and growth because sprouts cannot harvest energy and chemicals from photosynthesis at this stage. As it has been reported, GSLs are an important source of sulfur, which can be degraded by β-glucosidase BGLU28 and BGLU30 to provide sulfur for cysteine synthesis [33]. Low level of cysteine might be due to active sulfur metabolic f lux that cysteine turns into other urgent chemicals quickly as it is the precursor of massive biomolecules and possesses low-accumulation but high-f lux property [38,39]. However, the reason why high GSLs generate low GBPs in seeds is probably because the hydrolysis product of GSLs may be further converted to other chemicals in the seed homogenate [40]. Besides, from d0 to d7, the content of total GSLs changed similarly to that of total GBPs, but the proportion of alkenyl GSLs and alkyl GSLs was kept almost unchanged while that of ITCs decreased and epithionitriles increased. Correspondingly, the transcription of ESPs rose to a higher level in the later stage of sprout development (Figs 2b and 3c). Thus, it might be the enhancement of ESP that led to more epithionitrile formation. Moreover, during seed germination and seedling growth, radicle protrudes through the seed coat and the interaction between plants and environments increases gradually. Therefore, more defense chemicals like GSLs are needed to resist biotic and abiotic stresses. Following this, we found the content of total GSLs kept rising from d0 to d7 (Fig. 1d and e). However, conf licting results have been observed in Arabidopsis, which were planted in the nutritional matrix [11], suggesting that circumstance conditions affect the GSL metabolism greatly. Interestingly, the content and proportion of indole GSL increased markedly from d3 to the end of the experiment (Fig. 1e and f). Consistently, RNA-seq results showed that more GSL-related up-regulated DEGs are indole GSL biosynthetic genes. As the young seedling has been well developed 3 days after sowing, more biotic stresses will be encountered, thus, more defense substances are needed, particularly indole GSL which is important in the resistance to herbivorous insects and pathogens. Hanschen et al. (2017) compared the content of GSLs between sprouts and fully developed heads of five kinds of Brassica crops, and found that the latter had higher indole GSL content [41]. What's more, Wiesner et al. (2013) demonstrated that mature leaves accumulated more indole GSL than sprouts in Pak Choi [42]. All these reports suggest that plants increase indole GSL storage along with growth and maturity.
We identified and characterized four BoESP isoforms in Chinese kale. High identity between BoESPs and AtESP was observed (Fig. S4, see online supplementary material), and the amino acid sequences of ESP from oilseed rape and broccoli were more than 75.0% identical with AtESP [18,29]. These results suggest that ESP is highly conserved in the Brassicaceae family. Both transient expression in tobacco leaf cells and heterologous overexpression in Arabidopsis showed clear cytoplasmic and nuclear localization of BoESP, which is in line with the localization of AtESP in Arabidopsis [24]. This cytoplasmic and nuclear localization facilitates the dual functions of ESP as an epithiospecifier involved in GSL degradation and regulator of WRKY53 involved in leaf senescence [43]. Spatio-temporal expression pattern of BoESPs in Chinese kale indicated that BoESP3 was rich only in root, while BoESP1 and BoESP2 were abundant in other tested tissues. A similar result has been found by Witzel et al. (2019) in white cabbage [31]. However, AtESP was detected in all organs of Arabidopsis (Ler) except for the roots [24]. It is known that the At-α whole-genome duplication occurred near the origin of the Brassicaceae family. Subsequently, retained gene duplicates of ancestral f lavin monooxygenase glucosinolate S-oxygenase (FMO-GSOX) enzymes undertook tandem duplication and subfunctionalization to catalyze side chain modifications of Met-derived GSLs [44,45]. There exists a monophyletic origin of ESPs from NSPs, and the split between them occurred as the consequence of the α-whole genome duplication followed by neofunctionalization [46]. Correspondingly, the generation of non-isothiocyanate GBPs is widespread in the Brassicaceae, but not common in plants of other families of the Brassicales [46]. Although isothiocyanates are more toxic than their corresponding nitriles, some specialist herbivores use isothiocyanates as cues to identify host, thus specifier proteins may confer additional resistance to specialist herbivores on plants [21,46]. As Brassica species experienced genome duplications after their divergence from Arabidopsis [28], different isoforms of ESPs in Brassica vegetables may evolve diverse functions in specific tissues to better adapt to the environment or selection pressures from specialist herbivores. Overall, the expression of BoESPs in various organs throughout the whole plant implies an indispensable role in the life cycle.
Exogenous ABA and GA 3 treatments caused notable BoESPs induction. Correspondingly, the content and proportion of epithionitriles were significantly increased by ABA and GA 3 treatments, while those of ITCs were decreased markedly. Besides, the content of total GBPs was boosted by ABA application but not GA 3 , which is consistent with previous reports that ABA treatment rather than GA 3 enhanced the accumulation of GSL [47,48]. Thus, increasing the content and proportion of ITCs in Chinese kale sprouts through inhibition of GA 3 signaling might have potential to improve the health-promoting quality of Chinese kale sprouts. Overall, these results imply that the function of ESP in promoting the formation of epithionitriles at the expense of ITCs might be conserved in Chinese kale [18,26]. This was testified in vivo as BoESP2-silenced sprouts showed a significant decline of epithionitriles, and an increase of ITCs in comparison to the control (Fig. 6). Besides, the content of nitriles did not change significantly when the expression of BoESP2 was knocked down, suggesting BoESP2 is mainly in charge of the formation of epithionitriles rather than nitriles in Chinese kale sprouts.
In summary, dynamic GSL metabolic f lux exists during sprout development to fine-tune the growth and resistance. The content and proportion of total epithionitriles increased along with sprout development, while ITCs showed a decreasing trend. However, this pattern could be altered by regulating BoESPs via VIGS-mediated gene silence, implying the possibility of artificially improving the health-promoting quality of Chinese kale sprouts.

Plant materials, growth conditions, and chemical treatments
The seeds of Chinese kale (B. oleracea var. alboglabra Bailey cv. Sijicutiao) were sterilized in 20% bleach for 10 min, and washed with distilled water 5-6 times. Then they were immersed in distilled water for 24 hours at 25 • C without light. Sprouts were grown in plastic trays with distilled water at 23 • C under a 16 h/8 h light/dark cycle with 80% relative humidity (Fig. S1, see online supplementary material). Distilled water in the plastic trays was replaced every third day, and no fertilizer was added. Sprouts were harvested 0 day (d0), 1 day (d1), 3 days (d3), 5 days (d5), and 7 days (d7) after sowing. The dry seeds and harvested sprouts were frozen in liquid nitrogen immediately. Part of them was lyophilized by a freeze-dryer and was crushed into powder for GSL analysis, while the rest were stored at −80 • C for the detection of GBP and thiols. Besides, the weights of 80-300 fresh or dried plants at each time point were recorded for the calculation of the weight per plant.
Chinese kale for spatio-temporal expression detection were grown in soil, and samples at different developmental stages and organs were harvested, frozen in liquid nitrogen immediately, and then stored at −80 • C for RNA extraction.
For chemical treatments, Chinese kale seeds were sterilized and soaked as above. Sprouts were grown in plastic tissue culture vessels (7 × 8 × 7 cm) with five pieces of wet absorbent gauze in the same chamber as Chinese kale sprouts. Four days after sowing, sprouts were watered with 4 mL 200 μmol/L abscisic acid or 4 mL 50 μmol/L or sterile-distilled water (control group). Sprouts were harvested at 0 h, 6 h, 12 h, 24 h, and 48 h after treatment for RNA extraction, and 72 h after treatment for GBP detection.

Gene cloning and sequence analysis
Primer pairs of BoESP1F-BoESP1R, BoESP2F-BoESP2R, BoESP3F-BoESP3R, and BoESP4F-BoESP4R (Table S6, see online supplementary material), respectively, were used for BoESPs amplication, which were designed according to the sequences of ESPs from cabbage obtained from the Brassicaceae Database (BRAD) (http:// brassicadb.cn). The cDNA of Chinese kale sprouts was used as the template, and PrimeSTAR ® HS DNA Polymerase (premix) (Takara, Japan) was used for DNA amplification. The amino acid sequences of ESP from other species were acquired from BRAD and The Arabidopsis Information Resource (TAIR). Multiple sequence alignment of BoESPs, ESPs from cabbage, and AtESPs was conducted by ClustalW (https://www.genome.jp/toolsbin/clustalw). The image showing the aligned sequences was generated by ENDscript/ESPript (https://espript.ibcp.fr/ESPript/ cgi-bin/ESPript.cgi). MEGA 7.0 was employed to construct the phylogenetic tree with the neighbor-joining (NJ) method and bootstrap analysis (500 replicates). The results were expressed as a figure drawn by using iTOL (https://itol.embl.de/itol.cgi). The regulatory elements in the promoters were analysed by PlantCARE (http://bioinformatics.psb.ugent.be/webtools/plantcare/html/), and TBtools was used to visualize the results [49].
In the VIGS experiment, 40 bp target oligonucleotides were selected in the highly conserved region of four BoESPs coding sequences. Two sets (pTY-S/BoESP-1 and pTY-S/BoESP-2) were synthesized to increase the efficiency of VIGS. We used the vector reported in a previous study [51]. The oligonucleotides were annealed and mixed with SnaBI-digested pTY-S plasmids to produce the working plasmids, namely pTY-BoESP-1 and pTY-BoESP-2.

Transient expression in Nicotiana benthamiana and transformation in Arabidopsis
N. benthamiana plants were grown in the same growth chamber as Chinese kale sprouts. The Agrobacterium strain GV3101 was used as the carrier for transient expression. Bacterial suspensions carrying corresponding constructs were infiltrated into tobacco leaves according to the previous report [52]. After infiltration, N. benthamiana plants were placed in the growth chamber for 48 hours before observation under a confocal microscope.
Arabidopsis plants (Col-0) were grown at 21 • C under a 16 h/8 h light/dark cycle with 70% relative humidity. The Agrobacterium tumefaciens strain GV3101 was employed for the transformation experiment. Selection of transformants was carried out according to the antibiotic markers associated with the corresponding vectors used. The root of transformants that heterologously overexpressing BoESPs was used for the observation under a confocal microscope.

Virus-induced gene silencing of BoESPs in Chinese kale sprouts
The inoculum was prepared by adding pTY-BoESP-1 and pTY-BoESP-2 or empty plasmid pTY-S into infection media containing 10 mM MgCl 2 and MES with a final concentration of 14 ng/μL. Germinated Chinese kale seeds with radicle protrusion through the seed coat were immersed in the inoculum and infiltrated via vacuuming for 30 seconds twice. The infiltrated seeds were then planted and grown under the same condition as samples for GSL detection. After 9 days, 10 sprouts were collected as one replicate to determine the expression level of BoESPs and the content of GBPs.

RNA extraction and expression analysis
Total RNA was isolated by Trizol reagent and mRNA was reverse transcribed into cDNA according to the manufacturer's instruc-tion (Takara, Japan). The qPCR was conducted with CFX96 Real-Time PCR Detection System (Bio-Rad, USA). β-ACTIN was used as a housekeeping gene. The relative expression level of target genes was computed with the 2 − CT method [53]. Gene-specific primer sequences for qPCR are listed in Table S6 (see online  supplementary material).

Determination of glucosinolates and glucosinolate breakdown products
Glucosinolates and glucosinolate breakdown products were extracted and determined as previously described with some modifications [4]. Glucosinolates were extracted from 20 mg samples by 2 mL of 90% methanol. Data were given as μmol/g or μmol/plant (the value of μmol/g times the weight of a single plant). For the detection of glucosinolate breakdown products, frozen sprouts (300 mg) were homogenized in liquid nitrogen. Data were given as μmol/g or μmol/plant (the value of μmol/g times the weight of a single plant).

Determination of thiols
The detection of cysteine and glutathione was carried out as previously described [54]. In this study, about 300 mg of frozen sprouts were homogenized in liquid nitrogen.

Statistical analysis
At least three biological replicates were taken for each experiment, and values were represented as the mean ± SD. The SPSS package program version 16.0 was used to do statistical analysis. Multiple comparisons were subjected to one-way ANOVA test and the least significant difference (LSD) test with a significant level of 5% (P < 0.05), while the values of BoESP expression in response to plant hormones were analysed by two-way ANOVA and Tukey's test (P < 0.05). Pairwise comparisons were subjected to Student's t-test (P < 0.05). Significance was indicated by asterisks * or different letters.

Data availability
The datasets that support the findings of this study are available from the corresponding author upon reasonable request.