Sterile Salvia miltiorrhiza plantlets were cultured on a Murashige and Skoog basal medium, as described previously (Yan and Wang, 2007). All chemicals were obtained from Sigma Chemical Co. (St. Louis, MO, United States). Solvents were of high-performance liquid chromatography (HPLC) grade. Standards of RA, Sal B, and JA were purchased from the National Institute for the Control of Pharmaceutical and Biological Products (Beijing, China). All were prepared as stock solutions in methanol and stored in the dark at -18°C. Primer pairs are listed in Supplementary Tables 1, 2.

To construct the SmbHLH37 overexpression vector, we amplified the full-length open reading frame (ORF) of SmbHLH37 (GenBank Accession Number KP257470.1) with primers GVSmbHLH37-F/R, which introduced attB sites, and subsequently re-combined it into the pDONR207 vector (BP reaction Gateway^®) according to the protocol from the Gateway manufacturer (Invitrogen, United States). The ENTRY vector pDONR207-SmbHLH37 was sequenced and inserted into the pEarleyGate 202 vector (Earley et al., 2006) by an LR reaction (Gateway^®) to generate the pEarleyGate 202-SmbHLH37 overexpression vector. Agrobacterium-mediated gene transfer was performed based on protocols established in our laboratory (Yan and Wang, 2007).

To evaluate whether the OE box had been integrated into the transgenic plant genome, we amplified the CaMV35S promoter from isolated genomic DNA, using previously published protocols (Yang et al., 2017).

Total RNA from the roots of S. miltiorrhiza transgenic lines was extracted and converted into cDNA. Gene expression was monitored via real-time quantitative PCR (RT-qPCR), with housekeeping gene SmUbiquitin serving as an internal reference. Quantitative reactions were performed on a LightCycler^® 96 real-time PCR detection system (Roche, Switzerland), using SYBR Premix Ex TaqTM (Takara, Beijing, China). The reaction mixture contained 10 μl of 2 × SYBR Premix Ex Taq II, 0.5 μM each of sense and antisense primers, 20 ng of first-strand cDNA, ddH₂O up to 20 μl. Initial thermal-cycling at 95°C for 30 s was followed by 45 cycles of 95°C for 10 s and 60°C for 30 s. All experiments were performed on three independent biological experiments with each including three technical replicates. Relative expression was analyzed according to the 2^-ΔΔCt method compared with the WT (Livak and Schmittgen, 2001). Statistical significance was calculated using the by two-tailed Student’s t-test.

Based on the transcript levels of SmbHLH37, we conducted RT-qPCR analysis to determine the expression levels of key enzyme genes for the biosynthetic pathways of Sal B, JA, and anthocyanin.

Extraction and quantification of anthocyanins was performed in accordance with the protocols of Mano et al. (2007), with minor modifications. 20 mg samples of powder from transgenic or wild type (WT) plants were extracted with 1 mL of acidic methanol [1% (v/v) HCl] for 1 h at 20°C, with moderate shaking (100 rpm). After centrifugation (12000 rpm, room temperature, 5 min), 0.7 mL of the supernatant was added to 0.7 mL of chloroform. Absorption of the extracts at wavelengths of 530 and 657 nm was determined photometrically (DU 640 Spectrophotometer, Beckman Instruments). Quantitation of anthocyanins was performed using the following equation: Q (anthocyanins) = (A530–0.25 × A657) × M^-1, where Q (anthocyanins) is the concentration of anthocyanins, A530 and A657 are the absorptions at the wavelengths indicated, and M is the dry weight (in grams) of the plant tissue used for extraction.

Roots were collected from 2-month-old transgenic plantlets and air-dried at 20 ± 2°C. The phenolic compounds were extracted and determined as described by Li et al. (2018).

To determine the concentration of JA, we extracted JA using a modified protocol as described (Yang et al., 2012). Approximately 0.1-g root samples were homogenized and added to 10 mL of cold extraction buffer (acetone: 50 mM citric acid, 7:3, v/v). After this mixture was vortexed and then left to stand 30 min at 4°C, 10 mL of ethyl acetate was added before vortexing again. Following centrifugation at 5000 g for 10 min at 4°C, the supernatants were transferred to new 50-mL tubes and evaporated to dryness in a freeze dryer. The residue of each sample was re-suspended in 1 mL of 80% methanol (v/v) and sonicated for 10 min, then passed through a 0.22-μm organic filter. The extracts were loaded onto an Agela Cleanert SPE-NH2 (500 mg/6 mL); sonication and filtration steps were repeated. The combined supernatants were used for JA detection.

We determined the concentrations of JA in the plant samples by LC-QQQ-MS. Briefly, analyses were conducted using an Agilent 1260 HPLC system coupled to an Agilent 6460 QQQ LC-MS system equipped with a dual electrospray ion source operated in the negative mode. The extracts were separated on a Welch Ultimate XB-C18 column (2.1 × 150 mm, 3 μm). The chromatographic separation was performed over an 8-min analysis time, using a linear gradient of 85% to 50% A (0–6 min), 50% to 0% A (6–7 min), and 0% to 0% A (7–8 min). The flow rate of the gradient mobile phase was 0.4 mL/min, and the column temperature was 30°. Conditions for mass spectrometry included a drying gas temperature of 300°C, drying gas flow of 10 L/min, nebulizer pressure of 45 psi, ion spray voltage of 3500 V, and sheath gas of 11 L/min, at a temperature of 350°C. Retention time was 6.9 min for JA. The precursor/product ion of JA was 209.1 > 59.1. The concentrations were quantified based on standard curves prepared with authentic reference standards.

The ORFs of SmJAZ1/3/8 and SmMYC2 without the termination codon were individually cloned into the pDONR207 vector through Gateway reactions and re-combined into the pEarleyGate202-YC (YC) vector to generate YC-SmJAZ1/3/8 and YC- SmMYC2. Likewise, the ORF of SmbHLH37 without the termination codon was inserted into the YC vector or pEarleyGate201-YN (YN) to construct YC-SmbHLH37 and YN-SmbHLH37. The YC and YN recombinant plasmids were mixed at equal densities before co-transformation.

The plasmids were transiently transformed into onion epidermis cells by particle bombardment (helium pressure, 1100 psi) with the PDS-1000/He system (Bio-Rad, CA, United States). After 24 h of incubation, those cells were stained with DAPI (Vector Labs, CA, United States) for 20 min and then observed using a Leica DM6000B microscope (Leica, Germany) with an excitation wavelength of 475 nm.

The full-length coding sequence of SmbHLH37 was cloned into the pGADT7 or pGBKT7 vector, while those of SmJAZ1/3/8 and SmMYC2 were cloned into the pGADT7 vector. To test potential auto-activation of the prey, vectors of pGBKT7-SmbHLH37 and pGADT7 were co-transformed. Empty vectors of pGADT7 and pGBKT7 were also co-transformed as a negative control. The two types of recombinant vectors were co-transformed into yeast strain AH109 by the PEG/LiAC method (Zhou and Memelink, 2016). Interaction assays were performed according to manufacturer’s protocol for the Matchmaker Gold Yeast Two-Hybrid System (Clontech, United States), and Y2H images were taken on Day 5 of incubation.

The ORFs of SmbHLH37 and SmMYC2 were individually amplified by PCR using primers containing BamHI and EcoRI restriction sites. They were fused to the GAL4 activation domain in vector pGADT7-Rec2 (Clontech) to create the fusion proteins pGADT7-SmbHLH37 and pGADT7-SmMYC2. The ∼798-bp, ∼1350-bp, and ∼1146-bp promoter regions of SmPAL1, SmTAT1, and SmCYP98A14, respectively, were amplified and cloned into pHIS2 (Clontech). These recombinant vectors were co-transformed into yeast strain Y187 according to the reported protocol (Huang et al., 2013). The transformed cells were cultured on an SD/-Leu/-Trp medium and then selected on an SD/-Leu/-Trp/-His medium supplemented with 60 mM 3-amino-1, 2, 4-triazole to examine any protein-DNA interactions.

For assaying transient transcriptional activity, we amplified and cloned the ∼798-bp, ∼1350-bp, and ∼1146-bp promoter regions of SmPAL1, SmTAT1, and SmCYP98A14, respectively, into the pGreenII 0800-LUC (luciferase) vector (Hellens et al., 2005) to generate our reporter construct. The full-length coding sequences of SmMYC2 and SmbHLH37 were inserted into the pGreenII62-SK vector as the effector. Transient expression was monitored in N. benthamiana leaves according to the protocols of Sparkes et al. (2006). After 3 d of infiltration, activities of firefly LUC and renillia luciferase (REN) were measured using a dual-luciferase reporter gene assay kit (Beyotime Biotechnology, China) and a GloMax 20/20 luminometer (Promega, United States). Relative LUC activity was calculated by normalizing it against REN activity.

SmbHLH37 can be dramatically induced by exogenous MeJA (Zhang et al., 2015). We previously showed that SmbHLH37 is most similar to AtJAM3 (Su et al., 2017), which interacts with JAZs in Arabidopsis (Fonseca et al., 2014; Sasaki-Sekimoto et al., 2014). To detect whether SmbHLH37 and SmJAZs could interact with each other in S. miltiorrhiza, we performed BiFC and Y2H assays. Because SmJAZ1 and SmJAZ8 quickly respond exogenous MeJA treatment and SmJAZ3 shows highest expression in roots (Ge et al., 2015), we selected SmJAZ1/3/8 for the experiments. Our results demonstrated that SmbHLH37 interacts with SmJAZ3/8 (Figures 1A,B). Among the JAZ proteins in S. miltiorrhiza, SmJAZ8 has been established as being involved in repressing the biosynthesis of salvianolic acids and tanshinones (Ge et al., 2015; Pei et al., 2018). The bHLH protein usually forms a homodimer or heterodimer to develop their function (Feller et al., 2006). Our findings indicated that SmbHLH37 does form a homodimer (Figures 1A,B). Moreover, SmbHLH37 does not interact with SmMYC2.

To check whether the expression box had been integrated into the genome of S. miltiorrhiza, we performed PCR to amplify the CaMV 35S promoter. Our result showed that the transgenic plants indeed contained an expected 721-bp fragment (Supplementary Figure 1A). Real-time quantitative PCR demonstrated that expression of SmbHLH37 was highest in Lines OE-4 and OE-7 when compared with the non-transformed WT (Supplementary Figure 1B). Therefore, we chose those two lines for further analysis.

JA is derived from a-linolenic acid and the biosynthesis pathway was shown in Figure 2A. The transcript levels of genes encoding LOX (lipoxygenase), AOS (allene oxide synthase), AOC (allene oxide cyclase), and OPR3 (12-oxophytodienoic acid reductase) were significantly down-regulated in OE lines (Figure 2B). We performed LC-MS to determine the concentrations of endogenous JA in fresh root samples from OE and WT lines. The MRM chromatograms of JA were shown in Supplementary Figure 2. The results of the LC-MS analysis revealed that those JA levels were significantly decreased in OE-4 and OE-7 when compared with the control (Figure 2C). These results implied that overexpression of SmbHLH37 lowed JA biosynthesis in S. miltiorrhiza.

We further examined the transcription changes of genes encoding JAZ proteins and MYC2, core factors in the JA signaling pathway. Our RT-qPCR results showed that overexpression of SmbHLH37 significantly decreased the transcript levels of SmJAZ1/3/8 and SmMYC2 (Figure 3).

Our lab previously proved that activation of JA signaling can improve the accumulation of anthocyanin in S. miltiorrhiza (Ge et al., 2015). We tested whether this regulation of a transcriptional cascade by SmbHLH37 alters anthocyanin levels and found that concentrations of this pigment were significantly lower in the roots of OE-4 and OE-7 than in the WT (Figures 4A,B). We also investigated the expression profiles of genes for anthocyanin biosynthesis, e.g., CHS (chalcone synthase), F3′H (flavonoid 3′-hydroxylase), F3′5′H (flavonoid 3′5′-hydroxylase), FLS (flavonol synthase), and DFR (dihydroflavonol 4-reductase). All were significantly down-regulated in OE lines, with DFR showing the largest fold-change (Figure 4C).

The biosynthetic pathway of RA and Sal B included both phenylpropanoid-derived and tyrosined-derived pathway (Figure 5A). Many of the genes encoding enzymes on the pathway have been identified in S. miltiorrhiza, including SmPALs, SmC4Hs, Sm4CLs, SmTATs, SmHPPRs, SmRASs, and SmCYP98A14 (Di et al., 2013; Hou et al., 2013; Wang et al., 2015). We predicted that the production of salvianolic acids would be decreased in OE lines because of the decline in JA levels. To test this, we performed LC-MS to determine the concentrations of RA and Sal B. The MRM chromatograms of RA and Sal B were shown in Supplementary Figure 3. The results of the LC-MS analysis were consistent with our expectations, i.e., the levels of RA and Sal B were significantly declined in OE lines (respective reductions of 2.0- and 1.8-fold for RA and Sal B in OE-4; 1.7- and 1.6-fold for RA and Sal B in OE-7) when compared with the WT (Figure 5). To evaluate how the expression of genes related to phenolic acid biosynthesis is influenced in transgenic lines, we monitored relative transcript levels for SmPAL1/3, SmC4H1, Sm4CL2/3, SmTAT1/3, SmHPPR1, SmRAS1/6, and SmCYP98A14 in the WT and OE lines (Figure 5). Expression of all tested genes was significantly decreased in OE plants (p < 0.01). In particular, transcript levels of RAS6 were decreased 6.0- and 5.1-fold in OE-4 and OE-7, respectively.

The bHLH TFs function by binding to the E/G-box of the target gene promoter. Although 29 enzyme genes have been predicted to participate in phenolic acid biosynthesis in S. miltiorrhiza (Wang et al., 2015), only a few have been verified as doing so, including SmPAL1 (Song and Wang, 2011), SmTAT1 (Xiao et al., 2011), and SmCYP98A14 (Di et al., 2013). Each of them carries E/G-box sequences in its promoter (Figure 6A). We speculated whether SmbHLH37 is directly involved in regulating the pathway of phenolic acid biosynthesis. We conducted a Y1H experiment with promoter regions of ∼798-bp, ∼1350-bp, and ∼1146-bp length from genes SmPAL1, SmTAT1, and SmCYP98A14, respectively. The results shown in Figure 6B indicate that SmbHLH37 directly binds to the promoters of SmTAT1 and SmPAL1 rather than SmCYP98A14 in yeast.

We then conducted an assay of transient transcriptional activity in N. benthamiana leaves. The promoter regions of SmTAT1, SmPAL1, and SmCYP98A14 were fused individually with LUC to generate the reporter, and SmbHLH37, driven by the 35S promoter, was used as an effector (Figure 6C). As showed in Figure 6D, the LUC activity under promoters of SmTAT1 and SmPAL1 were reduced, which is due to repression of expression resulting from SmbHLH37 binding to the promoters. The same was not true for SmCYP98A14. Therefore, these results demonstrated that SmbHLH37 directly binds to the promoter regions of SmTAT1 and SmPAL1 to repress their expression.

We have reported that overexpression of SmMYC2 strongly increases the production of RA and Sal B, and those transcript levels of SmTAT1 and SmPAL1 are dramatically improved in SmMYC2-OE lines (Yang et al., 2017). However, the molecular mechanism had not yet been characterized. Here, we performed Y1H assays and examined transient transcriptional activity to verify whether SmMYC2 directly binds to the promoter regions of these genes to activate their expression. Results from our Y1H assay showed that SmMYC2 did bind to the promoter regions of SmTAT1, SmPAL1, and SmCYP98A14 (Figure 7A).

To conduct transient transcriptional activity analysis in N. benthamiana, SmMYC2 was used as an effector (Figure 7B). SmMYC2 activated LUC expression under promoters of SmTAT1, SmPAL1, and SmCYP98A14, based on our data from the assay of transient transcriptional activity (Figure 7C). We also learned that SmbHLH37 can repress SmMYC2-activated LUC expression, as driven by the promoters of SmTAT1 and SmPAL1 (Figure 7C). Together, these findings suggested that SmbHLH37 antagonizes transcription activator SmMYC2 in the Sal B biosynthesis pathway.