Broccoli (B. oleracea L., Lvyu), leaf mustard (Brassica juncea L., Jinsha), and radish (Raphanus sativus L., Yanghua) seeds were purchased from Nanjing Jinshengda Seed Co., Ltd. (Nanjing, China). The broccoli heads (Lvyu) were bought from the Suguo Supermarket (Nanjing, China). The seeds were planted according to the method of Wang et al. (2020). After germination, about 150 sprouts of similar size were chosen and transferred to plastic chambers which contained quarter-strength Hoagland’s solution (Wang et al., 2020). Then mustard and radish sprouts were illuminated by LED with white light in a growth chamber (Ningbo Sai Fu Instrument Co., Ltd., Ningbo, China). Besides, the broccoli sprouts were illuminated by LED with white, red and blue light, respectively. The photoperiod was set at 16 h light/8 h dark and the light intensity was set at 200 μmol m^–2s^–1. The humidity was set at 75 ± 5%, while the temperature was set at 25 ± 1°C.

After growing in the incubator for 2 days, the broccoli sprouts were treated with SA (Pérez-Balibrea et al., 2011). Firstly, broccoli sprouts were treated with 0, 40, 45, 50, 55, and 60 μM SA to select the suitable dose which has an inhibitory effect on the I3C content of the edible part. Then, the sprouts were divided into four groups, and the following treatments were carried out: sprouts were treated with 10 mL distilled water (Con), only with 10 mL 40 μM SA (SA), only with 10 mL 100 μM Paclobutrazol (a SA biosynthesis inhibitor, Pac), and with 5 mL 40 μM SA and 5 mL 100 μM Pac (SA + Pac), respectively. After 3 days of treatment, the sprouts were collected for measurement.

The A. thaliana eds5-1 (SALK-091541), ics1-l1 (SALK-133146), and ics1-l2 (SALK-093400) mutants were bought from the AraShare.¹ The A. thaliana was cultivated by the method of Wu et al. (2021). Seven days later, the seedlings were transferred to the soil and cultivated for 2 weeks under blue light (200 μmol m^–2s^–1).

Growth parameters [fresh weight (FW), hypotocyl length, and root length] of broccoli, mustard and radish sprouts were determined on three independent experiments (n = 3). The growth parameters from five sprouts were determined by a ruler (0.1 cm) and an electronic scale (0.0001 g), respectively. The fresh samples were photographed by camera (Canon, EOS 6D Mark II, Tokyo, Japan).

Indole-3-carbinol concentration was determined by the method described previously (Lee et al., 2010) with some modification. Briefly, 0.5 g edible portion of fresh samples (broccoli, mustard, and radish sprouts) were ground with liquid nitrogen and then the samples were left to autolyze with 8 mL 10 mM phosphate-buffered saline (pH 7.4). At the same time, the sample was incubated for 5 h at 37°C. After that, the crude suspension was centrifuged at 12,500 rpm at 4°C for 10 min, then the supernatant was collected and dried in a Rotary Evaporator (Hualida, LNG-T1OO, Nanjing, China). The residue was extracted twice with 500 μL of ethyl acetate and then dried in a Rotary Evaporator. The residue was dissolved in 1 mL of ethanol. Before the High-performance Liquid Chromatography (HPLC) analysis, the crude extract was filtered with a 0.22-μm filter. Then the extraction liquid was separated by a HPLC system (DIONEX, UltiMate 3000, RS Pump, Milford, CT, United States) with a UV detector on a T3 (ACQUITY UPLC^® HSS, 1.8 μm, 2.1 × 100 mm, Milford, CT, United States) column. The parameters were set according to the method of Lee et al. (2010).

Salicylic acid was extracted from edible portion excised from broccoli, mustard, and radish sprouts, and quantified using HPLC as previously described (Lee et al., 2011). 0.5 g samples were used to extract SA. Furthermore, all the methanol was used in this experiment is chromatographic methanol. Besides, all the supernatant was dried in a Rotary Evaporator (Hualida, LNG-T1OO, Nanjing, China). Finally, the residue was dissolved with 600 μL 40% chromatographic methanol. Before the HPLC analysis, the crude extract was filtered with a 0.22-μm filter. Then the extraction liquid was separated by a HPLC system with a UV detector on a C18 (Rapid Resolution HD, 2.1 × 50 mm, 1.8-Micron) column. The parameters were set according to the method of Lee et al. (2011).

Besides, as a key enzyme for the synthesis of SA, the activity of the phenylalanin ammonia-lyase (PAL) in broccoli sprouts were measured by the method of Lister et al. (2015).

The total RNA was extracted from broccoli sprouts treated with white, red, and blue light (W, R, B), respectively, with three biological replicates using RNA extraction kit by the manufacturer’s instructions (Invitrogen, Gaithersburg, MD, United States). And each biological replicate composed of a pool of sprouts. Then the qualified RNA was sent to Genedenovo Biological Co., Ltd. (Guangzhou, China) for RNA sequencing (RNA-seq) on the illumina sequencing platform. To ensure data quality, FASTP is used for quality control of raw reads, filtering low quality data and obtaining clean reads (Chen et al., 2018).

Then the clean reads aligned to the B. oleracea (wild cabbage²) reference genome. DEGs were defined as genes with |log₂FC| ≥ 1 and P-adjust < 0.05. The DESeq2 software was used to analysis DEGs between pairwise comparison (Love et al., 2014). The fragments per kilobase of transcript per million mapped reads (FPKM) can be directly used to compare gene expression differences between different samples. The Gene Ontology (GO) and the Kyoto Encyclopedia of Genes and Genomes (KEGG) of the differentially expressed genes (DEGs) were analyzed with an online tool.³

Five-day-old broccoli sprouts with different treatments (0.5 g) were collected and the RNA was extracted by RNA extraction kit (Invitrogen, Gaithersburg, MD, United States). Single-stranded cDNA synthesis was carried out with total RNA using the reagent kit (TOYOBO CO., LTD., Osaka, Japan). Next, the Mastercycler^® ep realplex real-time PCR system (ABI7500, MD, United States) system was used to carry out the reaction in a system of 20 μL (10 μL 2 × ChamQ Universal SYBR qPCR Master Mix, 0.5 μL forward/reverse primer, 9 μL cDNA). The program was used by the method of Wu et al. (2021). The Actin was used for a housekeeping gene and the transcription level of these genes were calculated by the 2^–ΔΔCt method (Livak and Schmittgen, 2001). Primers were designed by National Center for Biotechnology Information (NCBI), and the primer sequences were listed in Supplementary Tables 1, 2. Each cDNA sample was run in triplicate.

Arabidopsis thaliana DNA was extracted by the manufacturer’s instructions (Genenode Solution for Life science, Beijing, China). Then the leaves of ics1-L1, ics1-L2, eds5, and wild type (WT) seedlings were collected and the homozygosity was detected through PCR according to the method of Wu et al. (2021). The primers were designed by online website.⁴ Besides, the primers sequences were listed in Supplementary Table 3.

All experimental data were analyzed by the SPSS (Chicago, IL, United States). The results were shown as the means ± standard deviation (SD) of three independent experiments. The difference between the treatments were tested by analysis of variance (ANOVA) combined with Duncan’s multiple range, P < 0.05 represents the significant difference between the treatments. R package (version 4.0.5)⁵ was used for Pearson correlation analysis (Core et al., 2015). The graph was made by GraphPad.⁶

First we compared growth parameters among different species. The fresh weight (FW) of broccoli and radish sprouts were higher than mustard sprouts (Supplementary Figure 1B). Compared with broccoli sprouts, the hypocotyl length of radish and mustard sprouts were higher (Supplementary Figure 1C). However, among the three species, broccoli sprouts have the longest root length (Supplementary Figure 1D).

Subsequently, we explored the changes in I3C and SA contents among the three species. As shown in Supplementary Figure 1E, the I3C content in broccoli sprouts was 28.5 and 29.8% higher than mustard and radish sprouts, respectively. By contrast, the content of SA in different species was just the opposite. In all species, broccoli sprouts presented the lowest SA concentration, while the content of mustard and radish sprouts were higher. There were no differences in SA and I3C content between mustard and radish sprouts (Supplementary Figure 1F). Since broccoli sprouts had the highest content of I3C, it would be used in subsequent experiments. In order to further verify whether there were differences in I3C content of broccoli sprouts during different periods, the I3C content was detected using HPLC (Supplementary Figure 2). The I3C content of sprout stage was 43% higher than mature stage (Supplementary Figure 2B). Therefore, in this study, broccoli sprouts would be used in the next experiments.

Light, one of the key environmental factors, regulates plant growth and metabolism. Next, we explored the effect of different light on the phenotype, I3C content, and SA content of broccoli sprouts (Figure 1). Effects of light on plant growth, biomass, and root length of broccoli sprouts were analyzed (Figure 1). The FW, hypocotyl length, and root length of broccoli sprouts were 17.4, 24.6, and 34.6% lower in the blue light treatment than white light treatment, respectively (Figures 1A–D). In addition, the hypocotyl of broccoli sprouts was the longest under red light. However, there were no significant differences in FW and root length under red and blue light treatments (Figures 1A–D).

What’s more, according to the Figure 1E, we found that blue light treatment significantly increased the content of I3C. In contrast, the SA content was the lowest under blue light (Figure 1F), which was consistent with our suppose. Interestingly, there were no difference in I3C and SA contents under white and red light treatments.

In order to further verify our conjecture, we explored the influence of exogenous SA on I3C content. In order to determine the appropriate concentration of SA, we conducted a screening experiment for the concentration of SA (Supplementary Figure 3). As shown in Supplementary Figure 3, exogenous addition of SA could significantly reduce the content of I3C as compared to the control (Supplementary Figure 3B). Compared with the control, the content of I3C decreased by 51.7, 41.1, 43.4, 37.7, and 40.6% under 40, 45, 50, 55, and 60 μM SA treatments, respectively. Considering that excessive SA treatment may affect the growth and metabolism of plants, we selected 40 μM SA for subsequent analysis.

Next, we analyzed the influence of exogenous SA on endogenous SA and I3C content. As shown in Figure 2A, under the same light treatment, there was no significant difference in the growth parameters of broccoli sprouts in SA experiment (Figure 2A). The content of SA significantly decreased by sole Pac treatment compared with that of the control. Besides, the content of SA was elevated with SA + Pac co-treatment, as compared with that of the sole Pac treatment (Figure 2B). In contrast, low I3C content could be observed in broccoli sprouts, after SA treatment. However, compared with control, Pac treatment alone significantly increased the content of I3C. Compared with sole Pac treatment, I3C content was significantly reduced in the co-treatment (Figure 2C).

In order to further reveal the response of broccoli to blue light, RNA-seq analysis of broccoli sprouts treated with different light for 5 days was performed. In our research, we designed three treatment groups including white-vs-red (W-vs-R), white-vs-blue (W-vs-B), and red-vs-blue (R-vs-B), then the DEGs [FDR < 0.05 and |log2(FC)| > 1] were analyzed (Figure 3A). In the W-vs-R group, 482 DEGs were identified with 239 upregulated DEGs and 182 downregulated DEGs. There were 1104 DEGs in the R-vs-B group, among which 549 DEGs were upregulated and 555 DEGs were downregulated. By contrast, blue light induced a large number of DEGs, and the number of DEGs found in the W-vs-B group rose to 1193 including 561 upregulated DEGs and 632 downregulated DEGs.

Gene Ontology analysis is an internationally standardized classification system, which is used to comprehensively describe the function of uncharacterized genes. The DEGs were more related to biological processes, cell components, and molecular functions under different comparison groups (Supplementary Figure 4). Specifically, we observed that the enriched categories between different treatment groups were same, but there were certain differences in the number of DEGs (Supplementary Figure 4). The results of GO analysis indicated that the significantly enriched categories in three comparable groups were “metabolic process,” “single-organism process,” “cellular process,” “binding,” “catalytic activity,” “cell,” “cell part,” and “organelle process.”

Next, we analyzed the top 20 significantly enriched metabolic pathways by KEGG. In the W-vs-R group, the enriched pathways were “metabolic pathways,” “biosynthesis of secondary metabolites,” “phenylpropanoid biosynthesis,” “starch and sucrose metabolism,” and “plant hormone signal transduction” (Supplementary Figure 5A and Supplementary Table 4). A total of 410 DEGs were identified in the R-vs-B group, and these DEGs were mainly enriched in “metabolic pathways,” “biosynthesis of secondary metabolites,” “carbon metabolism,” “starch and sucrose metabolism,” and “glyoxylate and dicarboxylate metabolism” (Supplementary Figure 5B and Supplementary Table 5). However, the DEGs enriched in the W-vs-B group were quite different from other groups. These 439 DEGs were significantly enriched in “metabolic pathways,” “biosynthesis of secondary metabolites,” “plant hormone signal transduction,” “phenylpropanoid biosynthesis,” and “plant-pathogen interaction” (Figure 3B and Supplementary Table 6). These results revealed that the DEGs related to “plant hormone signal transduction” were more enriched in the W-vs-B group.

In order to verify the accuracy of the transcriptome data, 12 DEGs were selected randomly for quantitative reverse transcription PCR (RT-qPCR analysis). The results were very similar to the trend of the RNA-seq data (Supplementary Figure 6), which indicated that the RNA-seq data have a high degree of credibility.

Some genes have been reported to be involved in the biosynthesis of SA, such as Mate efflux family protein 5 (EDS5), Isochorismate synthase 1 (ICS1), Auxin-responsive gh3 family protein3 (PBS3), and PAL (Rekhter et al., 2019; Torrens-Spence et al., 2019). To explore the role of blue light in SA biosynthesis, the transcript responses of SA metabolism genes to blue light illumination were investigated. According to the transcriptome, we found blue light treatment led to lower gene expression levels of SA biosynthesis pathway, namely, the ICS pathway and the PAL pathway, as compared to control (Figure 4A and Supplementary Table 7). In PAL pathway, the expression level of ncbi_106300639 encoding PAL was significantly reduced by blue light treatment, and the heat map showed that the levels of FPKM in white and blue light treatment were 2.792 and 1.953, respectively. Besides, four genes encoding AIM1 were also downregulated under blue light treatment. We further compared genes in ICS pathways according to transcriptome by heat map (Figure 4A and Supplementary Table 7), and founded four genes (encoding ICS1) and one gene (encoding EDS5) were remarkably downregulated under blue light treatment compared with other treatments, but there were no significantly differences in the seven genes (encoding PBS3) and three genes (encoding EPS1). Taken together, these results highlight that blue light remarkably affects the expression of genes implicated in SA biosynthesis processes in broccoli sprouts. In particular, in PAL pathway. However, the genes encoding benzoic acid 2-hydroxylase (BA2H) in plants have not yet been resolved (Leon et al., 1993). Compared with white light treatment, RT-qPCR analysis confirmed that the transcript levels of BoEDS5, BoICS1, BoPBS3, and BoPAL were obviously downregulated under blue light, which is consistent with the RNA-seq data. Under red light treatment, with the exception of BoEDS5, these genes are slightly downregulated in comparison with the white light treatment (Figures 4B–E). Interestingly, there were no significant difference in the level of transcription of these genes between white and red light treatments, as compared with that of the blue light treatment (Figure 4).

Previous studies have shown that PAL plays an important role in the biosynthesis of SA (Huang et al., 2010). Therefore, we further analyzed the activity of PAL under various light conditions. We found that the activity of PAL was significantly reduced under blue light compared with white light and red light (Supplementary Figure 7).

Indole-3-carbinol is produced by the hydrolysis of I3M (Esteve, 2020). Besides, there are many genes involved in the production of I3C, including CYP79B2 (Cytochrome p450, family 79, subfamily b, polypeptide 2), CYP79B3 (Cytochrome p450, family 79, subfamily b, polypeptide 3), CYP83B1 (Cytochrome p450, family 83, subfamily b, polypeptide 1), UGT74B1 (Udp-glucosyl transferase 74b1), SOT16 (cytosolic sulfotransferase 16), TGG1 (Thioglucoside glucohydrolase 1), and ESP (Epithiospecifier protein), etc (Zang et al., 2010). As a comparative analysis of DEGs in different treatments is very important to explore key genes in I3C biosynthesis regulation, we further compared genes in I3C pathways according to the FPKM by heat map (Figure 5A), and revealed vital genes took part in I3C metabolism, as listed in Supplementary Table 8. As shown in Figure 6A, the genes involved in side-chain elongation stage, namely, BoTSB1 (tryptophan synthase beta-subunit 1) and BoASA1 (anthranilate synthase alpha subunit 1) showed higher levels in blue light treatment than other treatments. The core-structure biosynthesis stage is the important pathway in the biosynthesis of I3C, and the key genes in this stage include CYP79B2/B3, CYP83B1, GSTF9/10 (glutathione S-transferase F9/10), SUR1 (S-alkyl-thiohydroximate lyase SUR1), and UGT74B1 (Grubb and Abel, 2006). Under blue light treatment, most of these genes were upregulated compared with other light treatments. Noticeably, the expression level of ncbi_106328484 (encoding GSTF10) was significantly downregulation. As for the side-chain modifications stage, we found two genes (encoding STO16) and one gene (encoding TGG1) were upregulated under blue light treatments. However, the two genes, namely, MSTRG.17449 and ncbi_106306884 (encoding ESP) were lower in blue light treatment than that in other light treatments (Figure 5A and Supplementary Table 8). These results indicate that the upregulation of genes may play an important role in I3C metabolism.

To determine how blue light is involved in the regulation of I3C metabolism in the cells, the RT-qPCR assay was performed to analyze the transcript levels of four GLs biosynthesis genes, namely, CYP79B2, CYP79B3, CYP83B, and UGT74B1, and the transcript levels of two major GLs-degrading genes, namely, ESP and TGG1, in the edible part of broccoli sprouts. The results confirmed that the relative expression of BoCYP79B2/B3, BoCYP83B1, and BoUGT74B1 were significantly upregulated under blue light (Figures 5B–E). Interestingly, these genes had similar expression levels under white light and red light except BoUGT74B1 and BoCYP83B1 (Figure 5). Meanwhile, compared with the control, the gene expression of BoTGG1 involved in the breakdown process had a higher level under blue light (Figure 5G). However, blue light treatment significantly decreased the expression level of BoESP as compared to other light treatments (Figure 5F). Interestingly, in contrast, the transcript levels of these genes did little respond to red light illumination.

Arabidopsis thaliana, the model plant of the cruciferae family, can also produce I3C through secondary metabolism processes. Therefore, in order to further verify the negative regulatory effect of SA in the process of I3C biosynthesis, we used A. thaliana wild-type (WT) and the SA biosynthesis defective mutants (eds5, ics1-l1, and ics1-l2) for subsequent analysis. Mutations in these genes cause a decrease in SA accumulation (Rekhter et al., 2019; Torrens-Spence et al., 2019). The homozygous verification of A. thaliana indicated that all the A. thaliana mutants in our experiment were homozygous, which was conducive to our further experimental analysis (Figures 6A–F). In this research, we found that eds5, ics1-l1, and ics1-l2 seedlings displayed shorter hypocotyl and root length compared with WT (Figure 6G). Next, we used RT-qPCR to further verify the expression of these mutant genes. As shown in Figure 6H, RT-qPCR analysis confirmed that the expression of AtEDS5 and AtICS1 was significantly downregulated in mutants, compared with WT, which supported the loss of function of these genes in the mutant.

Next, we used A. thaliana, which grown under blue light for 2 weeks, to detect SA and I3C content. The results showed that in contrast to the WT, the SA content in the mutant was significantly reduced, and the eds5 had the lowest SA content (Figure 6I). Compared with the WT, the content of SA in eds5, ics1-l1, and ics1-l2 was reduced by 96, 49, and 37%, respectively (Figure 6I). At the same time, the content of I3C was just the opposite of the SA content (Figure 6J). In contrast to ics1-l1, ics1-l2, and WT, the I3C content was highest in eds5 sprouts (Figure 6J). However, compared with WT, eds5, ics1-l1, and ics1-l2 effectively elevated the content of I3C by 97, 88, and 90%, respectively (Figure 6J). According to correlation analysis, we found that SA and I3C were negatively correlated, and the correlation coefficient was −0.9574 (Figure 6K).