The Protein Kinase SmSnRK2.6 Positively Regulates Phenolic Acid Biosynthesis in Salvia miltiorrhiza by Interacting with SmAREB1

Subclass III members of the sucrose non-fermenting-1-related protein kinase 2 (SnRK2) play essential roles in both the abscisic acid signaling and abiotic stress responses of plants by phosphorylating the downstream ABA-responsive element (ABRE)-binding proteins (AREB/ABFs). This comprehensive study investigated the function of new candidate genes, namely SmSnRK2.3, SmSnRK2.6, and SmAREB1, with a view to breeding novel varieties of Salvia miltiorrhiza with improved stress tolerance stresses and more content of bioactive ingredients. Exogenous ABA strongly induced the expression of these genes. PlantCARE predicted several hormones and stress response cis-elements in their promoters. SmSnRK2.6 and SmAREB1 showed the highest expression levels in the leaves of S. miltiorrhiza seedlings, while SmSnRK2.3 exhibited a steady expression in their roots, stems, and leaves. A subcellular localization assay revealed that both SmSnRK2.3 and SmSnRK2.6 were located in the cell membrane, cytoplasm, and nucleus, whereas SmAREB1 was exclusive to the nucleus. Overexpressing SmSnRK2.3 did not significantly promote the accumulation of rosmarinic acid (RA) and salvianolic acid B (Sal B) in the transgenic S. miltiorrhiza hairy roots. However, overexpressing SmSnRK2.6 and SmAREB1 increased the contents of RA and Sal B, and regulated the expression levels of structural genes participating in the phenolic acid-branched and side-branched pathways, including SmPAL1, SmC4H, Sm4CL1, SmTAT, SmHPPR, SmRAS, SmCHS, SmCCR, SmCOMT, and SmHPPD. Furthermore, SmSnRK2.3 and SmSnRK2.6 interacted physically with SmAREB1. In summary, our results indicate that SmSnRK2.6 is involved in stress responses and can regulate structural gene transcripts to promote greater metabolic flux to the phenolic acid-branched pathway, via its interaction with SmAREB1, a transcription factor. In this way, SmSnRK2.6 contributes to the positive regulation of phenolic acids in S. miltiorrhiza hairy roots.


INTRODUCTION
Salvia miltiorrhiza Bunge (family Labiatae) is one of the traditional bulk medicinal materials. Its dried roots have been prescribed for the clinical treatment of many human diseases, such as irregular menstruation, cardiovascular and cerebrovascular diseases, and inflammation (Kai et al., 2011;Ma et al., 2013). This plant contains two main bioactive ingredients: lipid-soluble tanshinones and water-soluble phenolic acids (Chen et al., 2001). Phenolic acids have attracted considerable attention, largely due to their important medicinal effects and their convenient extraction by decoction, the main mode of application in traditional Chinese medicine. At present, the biosynthetic pathway of phenolic acids and its competition bypass branches in S. miltiorrhiza have been proposed, and the key enzyme genes involved in the phenolic acid-branched and side-branched pathways have been cloned Ma et al., 2013;Zhang Y. et al., 2014).
Faced with an ever-increasing demand for S. miltiorrhiza, its supply from wild resources is no longer sufficient. In response, many studies have since focused on increasing this species phenolic acid contents by either overexpressing or suppressing its transcription factors or key enzyme genes within the relevant metabolic pathways. For example, the overexpression of SmTAT, SmC4H, SmHPPR, AtPAP1, AtEDT1, and SmPAP1 induced a substantial accumulation of phenolic acids in the transgenic S. miltiorrhiza (Zhang Y. et al., 2010;Xiao et al., 2011;Hao et al., 2016;Liu Y. et al., 2016), whereas the downregulation of SmMYB39, SmHPPD, SmCCR1, or SmCHS all increased the content of phenolic acids Wang et al., 2012;Zhang et al., 2013;Zhang et al., 2015).
However, the effective components of S. miltiorrhiza are in the form of secondary metabolites. Because their quality and quantity strongly depends on environmental stresses, such as drought, cold, and salinity, this results in a lower yield of S. miltiorrhiza (Akula, 2011;Liu et al., 2011;Zhao et al., 2014). Several studies have reported that by overexpressing SmLEA2, AtDREB1A, AtDREB1B, and AtDREB1C, the salt and drought tolerance of transgenic S. miltiorrhiza was improved (Wei et al., 2016a(Wei et al., ,b, 2017Wang et al., 2017). More of such genes should be investigated via genetic engineering methods, as it could assist in breeding new varieties of S. miltiorrhiza that have a higher content of active ingredients and a stronger tolerance to stresses.
Sucrose non-fermenting 1 (SNF1)-related protein kinase 2 (SnRK2) is a plant-specific Serine/Threonine (Ser/Thr) protein kinase family. Based on their amino acid sequence similarity, the members of the SnRK2 family have been grouped into three subclasses. Compared to the subclass I/II members, those of subclass III play central roles in the positive regulation of abscisic acid (ABA) signaling and in the responses to osmotic stress (Fujii and Zhu, 2009). Overexpression of subclass III SnRK2 subfamily members SAPK9, TaSnRK2.8, and ZmSAPK8 could enhance drought, salt, or cold tolerance in transgenic plants. The subclass III SnRK2 reportedly regulate the expression of the ABAresponsive gene, primarily through four ABA-responsive element (ABRE)-binding proteins (AREB/ABFs): ABF1, ABF2/AREB1, ABF3, and ABF4/AREB2, all of which are highly inducible by osmotic stress and ABA treatments applied to vegetative tissues (Yoshida et al., 2010(Yoshida et al., , 2015Fujita et al., 2013).
Although the SnRK2 family has received extensive and indepth research, their subclass III members in S. miltiorrhiza have not yet been explored. Moreover, previous studies in our laboratory showed that exogenous ABA, as an elicitor, is capable of promoting the accumulation of phenolic acids . Therefore, it is possible that the subclass III SnRK2 members from S. miltiorrhiza have key roles to play in the synthesis of phenolic acids in this plant. With this in mind, we aimed to identify the orthologous genes of SnRK2s in S. miltiorrhiza, and to then evaluate whether these proteins contribute to the regulation of phenolic acid accumulation and whether they are involved in the stress responses of this plant. In this study, we not only cloned and functionally characterized the two subclass III SnRK2 members (i.e., SmSnRK2.3 and SmSnRK2.6), but we also identified and functionally characterized SmAREB1, which has a close affinity with AtAREB1. Additionally, we analyzed the physical interaction between SmSnRK2.3/2.6 and SmAREB1. These results not only fill a gap in our knowledge of the subclass III SnRK2 genes in S. miltiorrhiza, but they also provide a theoretical foundation to improve the stress tolerance and bioactive ingredient contents of S. miltiorrhiza via genetic engineering. More broadly, this study further contributes to the preliminary exposition of the molecular mechanism underpinning the exogenous ABAinduced accumulation of phenolic acid content in plants.

Plant Material and Treatments
Mature S. miltiorrhiza seeds were collected from the Danshen cultivation base of the Shaanxi Tasly plant medicine Co. Ltd. (Shangluo, China). These seeds were utilized to obtain sterile plantlets as previously reported (Yan and Wang, 2007). Sterilized seedlings were cultured on an MS medium (pH 5.8) that contained 7% agarose. The hairy roots of S. miltiorrhiza were derived from these plantlets, infected with Agrobacterium rhizogenes (ATCC15834 strain), and sub-cultured every 30 days.
The seeds were sown in plugs and hydroponically grown at 25 • C, under continuous light, for 2 months. Fresh roots, stems, and leaves were separately harvested from the ensuing seedlings, which were used as samples to determine expression levels of SmSnRK2.3/2.6 and SmAREB1 in different tissues of S. miltiorrhiza.
For the ABA treatment, the ensuing seedlings were sprayed with 100 µM ABA solution. The samples were harvested at 0, 0.5, 1, 3, 6, 9, 12, 24, and 48 h after treatment. All of these collected samples were immediately frozen in liquid nitrogen and stored at −80 • C prior to analysis.

Total RNA and DNA Extraction
Total complete RNA was isolated from the frozen S. miltiorrhiza samples by using the RNAprep Pure Plant Kit (TIANGEN, China). The RNA was then reversely transcribed to generate the cDNA, according to manufacturer's instructions of the PrimeScript TM RT reagent Kit with gDNA Eraser (Takara, Japan). To isolate the genomic DNA, an improved cetyltrimethylammonium bromide (CTAB) method was used. The quality and concentration of the genomic DNA and RNA were determined by using 0.8% agarose gel electrophoresis and a nucleic acid spectrometer (NanoDrop ND-1000, Thermo Scientific).

Gene Isolation and Bioinformatics Analysis
Based on a previously established S. miltiorrhiza transcriptome database (Shao et al., 2016), local BLAST analyses were performed using three subclass III SnRK2 members of Arabidopsis thaliana and an SmAREB1 fragment, which was amplified by using degenerate primers, as queries. According to the obtained sequences, we designed three pairs of specific primers to perform the PCR amplification with the pfu DNA polymerase (Thermo, United States). Amplification templates were genomic DNA and cDNA reversely transcribed via RNA, both of which were extracted from the S. miltiorrhiza hairy roots. The procedure for the PCR reaction system followed the manufacturer's instructions. The PCR products were gel purified by a DNA Gel Extraction Kit (OMEGA, United States) and then cloned into the pEASY-Blunt Simple cloning vector (TransGen Biotech, China) for sequencing. The primer sequences are shown in Supplementary  Table S1.
Both the ORF-finder 1 and the GENSCAN Web Server 2 were used to confirm the open reading frame (ORF) of each gene. Amino acid sequences of the genes were deduced with the ProtParam tool 3 and submitted to the NCBI database 4 for BLASTP searches. The DNAMAN software was used to perform the multiple alignments. Sequence alignment results were combined with an online motif scan tool 5 to analyze the functional domain of proteins. Phylogenetic trees were generated with the MEGA7 software program, by employing the neighborjoining method, with 1000 bootstrap replicates. All amino acid sequences of the other species were downloaded from the NCBI database 6 . The intron and exon distributions of the genes were analyzed by the online GSDS2.0 software 7 (Hu et al., 2015).

Quantitative Real-Time PCR Analysis
Total RNA was extracted from the S. miltiorrhiza hairy roots or S. miltiorrhiza seedlings, according to the method described above. The RNA samples (1 µg) were then reverse transcribed to 20 µL cDNA using the PrimeScript TM RT reagent Kit with gDNA Eraser (Takara, Japan), according to manufacturer's instructions. The resulting cDNA was diluted to 150 ng/µL with DNase/RNase-Free H 2 O. Quantitative real-time PCR (qRT-PCR) was performed in the CFX96 Real-Time PCR System (Bio-Rad, United States). The 20 µL reaction mixture contained 10 µL of 2 × SYBR R Premix Ex Taq TM II (Perfect Real Time, Takara, Japan), 1.6 µL of a forward/reverse primer (10 µM), 1.6 µL of a cDNA template, and 15.2 µL of DNase/RNase-Free H 2 O. The PCR procedure went as follows: at 95 • C for 30 s, then 40 cycles of 95 • C for 5 s and 60 • C for 30 s, this program was followed by a melting curve analysis (65-95 • C with temperature increment of 0.5 • C every 5 s). In this study, the qRT-PCR data for the genes responding to the ABA treatment were normalized to β-actin (Yang et al., 2012;Wei et al., 2017), while the other qRT-PCR data were normalized to β-actin and ubiquitin (Yang et al., 2010;Xiao et al., 2011). All the primers used for the qRT-PCR analysis are listed in Supplementary Table S3.

Subcellular Localization Analysis
The modified-vector pCAMBIA1301 containing eGFP was used in this study. The coding regions of the SmSnRK2.3, SmSnRK2.6, and SmAREB1 genes (without a stop codon) were successfully amplified and cloned upstream of the eGFP gene in the BamHIcleaved pCAMBIA1301 binary vector, to generate the GFP fusion vectors SmSnRK2.3-p1301, SmSnRK2.6-p1301, and SmAREB1-p1301. Subsequently, the SmAREB1-p1301 construct and the pCAMBIA1301 empty vector were transformed into onion epidermal cells via gene gun, and the transformed onion cells were later examined under an A1 confocal microscope (Nikon, Japan). To locate nuclei, the tissues were incubated with 4, 6diamidino-2-phenylindole dihydrochloride (DAPI). A transient expression assay in Nicotiana benthamiana leaves was performed to determine the subcellular localization of SmSnRK2.3 and SmSnRK2.6. The Agrobacterium tumefaciens (EHA105 strain) suspension cultures, harboring either the construct SmSnRK2.3-p1301 or SmSnRK2.6-p1301, were, respectively, infiltrated into leaves of N. benthamiana along with a plasma membrane marker, following the method described by Zhang H. et al. (2014). The plasma membrane marker was based on a full-length fusion of the aquaporin PIP2A to red fluorescent proteins (Nelson et al., 2007). To serve as the control, the A. tumefaciens (EHA105 strain) suspension cultures containing the empty vector pCAMBIA1301 were infiltrated into the leaves of N. benthamiana. After 3-4 days, the N. benthamiana leaves were harvested and subjected to a chimeric fluorescence signal analysis under an A1 confocal microscope (Nikon, Japan). The primers used for the subcellular localization analysis are listed in Supplementary Table S4.

Analysis of SmAREB1 Transcriptional Activity
The transactivation activity of SmAREB1 was examined by the yeast one-hybrid assay, a method widely used to measure the transactivation ability of transcription factors (Shukla et al., 2006;Hu et al., 2013;Shen et al., 2014). The complete coding sequence of SmAREB1 was amplified with specific primers (Supplementary Table S5). The PCR product was cloned into EcoRI and BamHI sites in the pGBKT7 vector, which contains the Trp reporter gene and the DNA binding domain of the transcription factor GAL4, to generate a GAL4 DNA-BD-SmAREB1 fusion plasmid, named SmAREB1-BD. The AtMYB15-pGBKT7 plasmid was constructed as a positive control. These two recombinant plasmids and the negative control pGBKT7 plasmid were used, respectively, to transform the Saccharomyces cerevisiae AH109 strain which carriers the His3, Ade2, and LacZ reporter genes under the GAL4 promoter, following the manufacturer's protocol (Clontech, United States). The ensuing transformants were verified via a yeast colony PCR and then plated onto synthetic dropout (SD)/-Trp and SD/-Trp-His-Ade plates cultured at 30 • C for 3 days. The transcriptional activation activity was evaluated according to their growth status.

Yeast Two-Hybrid (Y2H) Assays
The coding sequence of SmSnRK2.3/2.6 was cloned into the pGADT7 vector, enzyme digested by EcoRI and BamHI, thus producing the construct SmSnRK2.3/2.6-pGADT7 (SmSnRK2.3/2.6-AD). The primer sequences, including the cleavage site used for the amplification, are listed in Supplementary Table S5. The plasmid combinations of SmSnRK2.3/2.6-AD and SmAREB1-BD, SmSnRK2.3/2.6-AD and pGBKT7, pGADT7 and SmAREB1-BD, as well as that of pGBKT7 and pGADT7, were co-transformed into AH109, following the manufacturer's protocol (Clontech, United States). The AH109 cells carrying the SmSnRK2.3/2.6-AD + pGBKT7, pGADT7 + SmAREB1-BD, or pGBKT7 + pGADT7 plasmids were used as negative controls. The transformed colonies were shaken in an YPDA liquid medium until the OD 600 of their cell density was approximately 0.6, then, they were serially diluted to fractions of 1/10, 1/100, and 1/1000 by using sterilized double-distilled water. From these yeast cell dilutions, 2 µL were taken and spotted onto SD/-Trp-Leu and SD/-Leu-Trp-His-Ade plates and incubated at 30 • C for 3-5 days. Next, to test for possible interactions, the X-β-Gal staining assay was carried out following the methodology of Zhang H. et al. (2014).

Bimolecular Fluorescence Complementation (BiFc) Analysis
To confirm the protein interactions in vivo, two plant expression binary vectors, pSPYNE-35S and pSPYCE-35S, were used in BiFc analysis (Walter et al., 2004). On the basis of the SmSnRK2.3/2.6-p1301 and SmAREB1-p1301 constructs, the pSPYCE-35S and pSPYNE-35S vectors were used to replace the pCAMBIA1301 vector at the BamHI restriction enzyme site to generate the recombinant plasmids SmSnRK2.3/2.6-pSPYNE and SmAREB1-pSPYCE. For the transient expression assay, these plasmids and an empty vector (negative control) were co-transformed into the A. tumefaciens strain EHA105, in combination with the p19 strain to infiltrate 6-week-old N. benthamiana leaves, as done by Zhang H. et al. (2014). The chimeric fluorescence signals of the expressed fusion proteins were detected 3-5 days after infiltration under a Fluoview FV1000 confocal microscope (Olympus, Japan).
Purification of the SmSnRK2.3/2.6-GST Fusion Protein in Escherichia coli The coding sequence of SmSnRK2.3/2.6 was cloned into the pGEX-6p-1 vector, it harbored a glutathione S-transferase (GST) tag, by using specific primers (Supplementary Table S6). The Escherichia coli cells (Rosetta strain) harboring SmSnRK2.3/2.6-GST and the empty vector pGEX-6p-1 were induced by adding 0.5 mM isopropy-β-D-thiogalactoside (IPTG) for 16 h at 22 • C. The E. coli cells were harvested via centrifugation, resuspended in pre-cooled PBS buffer, and then broken with the ultrasonic method. The soluble GST fusion proteins were purified using the ProteinIso GST Resin (TransGen Biotech, China). Then, these purified proteins were verified through the Western blotting technique.

Identification of Phosphorylation Sites via Liquid Chromatography Tandem MS (LC-MS/MS)
To identify the phosphorylation sites of the SmSnRK2.3/2.6 protein, we separated the purified protein samples via sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE). Specifically, the targeted protein was excised from the Coomassie Brilliant Blue (CBB) staining gel, and then it was digested via filter-aided sample preparation (FASP), as previously described (Wisniewski et al., 2009). The peptide mixtures from the gel slice were injected onto a Zorbax 300 SB-C18 peptide trap (Agilent Technologies, United States), to desalt them in an autosampler, and then they were separated by reverse phase capillary high performance liquid chromatography (HPLC), using a RP-C18 column (0.15 mm × 150 mm, Column Technology Inc., United States). Prior to this, the chromatographic column had been balanced using a 95% solution A, and the mobile phases were solution A (0.1% formic acid in water) and solution B (0.1% formic acid and 84% acetonitrile in water). Samples were eluted with the following linear gradient: solution B increased from 4 to 50% in the first 30 min, then increased to 100% in the next 4 min, after which it was kept constant at 100% for 1 min.
After separation and desalination via HPLC, samples were analyzed by tandem mass spectrometry, performed on a Q Exactive mass spectrometer (Thermo Fisher, United States) equipped with an electrospray interface and operated in the positive ion mode. The mass spectrometer used one full MS scan, followed by 10 MS/MS scans on the 10 most intense ions from the MS spectrum, under the following settings: repeat count of 2; repeat duration of 30 s; and an exclusion duration of 90 s. Raw MS/MS spectra were subjected to the MASCOT engine (Matrix Science, United Kingdom) against the sequence of the SmSnRK2.6 protein.
For protein identification, the following parameters were used: Enzyme = Trypsin, Missed cleavage = 2, Peptide mass tolerance = 20 ppm, MS/MS tolerance = 0.1 Da, Mascot score ≥ 20, the ESI ion trap was selected for the instrument type, Carbamidomethyl (C) was the fixed modification, while Oxidation (M), Phospho (ST), and Phospho(Y) were used as the variable modifications.

Construction of Plant Expression Vectors and Acquisition of Positive Transgenic Hairy Roots
The coding regions of SmSnRK2.3, SmSnRK2.6, and SmAREB1 were amplified and cloned into the restriction site speI of the pCAMBIA1304 binary vector, under the control of the CaMV35S promoter and the NOS terminator, respectively. The recombinant plasmids SmSnRK2.3-1304, SmSnRK2.6-1304, and SmAREB1-1304 were transformed into A. rhizogenes ATCC15834, by using the empty pCAMBIA1304 vector as a vector-only control. The transformation of leaf explants from the sterile plantlets of S. miltiorrhiza, which followed a previously described method (Kai et al., 2011), acquired the hairy root lines; when these reached a certain biomass, their genomic DNA was extracted for PCR identification and to screen for the positive transgenic lines. These latter lines were used for RNA and phenolic acid extraction and they were regularly sub-cultured (every 30 days). All of the primers used for the overexpression vector construction and in the PCR identification of transgenic lines are listed in Supplementary  Table S7.

HPLC Analysis of Phenolic Acid Contents
Rosmarinic acid (RA) and salvianolic acid B (Sal B) are the main phenolic acids in S. miltiorrhiza. while Sal B has been designated as a marker component of S. miltiorrhiza in the official Chinese Pharmacopoeia, RA is the synthetic precursor material (Di et al., 2013). The contents of Sal B and RA in the transgenic lines and empty vector controls (EVs) were verified by an HPLC analysis, according to the general method in our laboratory (Xing et al., 2015;. The only minor difference is that a 20 mg sample of powder was dissolved in 4 mL of 70% methanol.

Statistical Analysis
In this study, irrespective of the HPLC analysis of phenolic acid contents, or the qRT-PCR analysis of gene expression levels, all the experiments were performed in triplicate. The results are presented as means ± standard deviation (SD). Gene relative expression levels were calculated using the delta-delta Ct (2 − Ct ) method (Livak and Schmittgen, 2001;Wang et al., 2015). All data were analyzed using Statistical Package for Social Science (SPSS v.16.0) software. Significant difference (p < 0.05) of mean values was compared using Tukey's multiple comparison test or student's t-test (indicated in the Figure legends).

Isolation and Bioinformatics Analysis of S. miltiorrhiza Genes
SmSnRK2.3 contained a 1,068 bp ORF, encoding a protein of 355 amino acids with a predicted molecular mass of 40.327 kDa. SmSnRK2.6 contained a 1,098 bp ORF, encoding a protein of 365 amino acids with a predicted molecular mass of 41.340 kDa. The amino acid sequence analysis and multiple alignments revealed that both SmSnRK2.3 and SmSnRK2.6 had a highly conserved N-terminal kinase domain, in addition to C-terminal regulatory domain that consisted of domains I and II ( Figure 1A). To visually examine their evolutionary origins, the two SmSnRK2s, along with 10 SnRK2 proteins from A. thaliana and 10 SAPK proteins from O. sativa, were used to construct a phylogenetic tree. As shown in Figure 2A, both SmSnRK2.3 and SmSnRK2.6 were clustered within subclass III of the SnRK2. However, SmSnRK2.3 showed a higher degree of similarity with AtSnRK2.3 than with AtSnRK2.2. SmSnRK2.6 shared the same branch and had the highest degree of similarity with AtSnRK2.6. Hence, we named these two genes SmSnRK2.3 and SmSnRK2.6.
Comparing the genomic and cDNA sequences revealed that SmSnRK2.3 contained seven introns and eight exons; the SmSnRK2.6 gene contained eight introns and nine exons; and SmAREB1 contained three introns and four exons (Figure 1C).

Isolation and Analysis of Promoters
The promoter sequences of SmSnRK2.3, SmSnRK2.6, and SmAREB1 were identified via cloning and sequencing. The PlantCARE analysis showed that they contained several ciselements which are related to phytohormone response, abiotic and biotic stresses, and plant development, with the exception of core cis-acting elements, such as the TATA box and CAAT box ( Table 1). Moreover, more than two ABREs were detected in all of their promoters, likely because the expression of ABA responsive genes requiring multiple ABREs or the combination of an ABRE with one of several coupling elements (Fujita et al., 2013). These results indicated that all the three genes may be involved in the S. miltiorrhiza responses to ABA and stresses. Expression Analysis of Genes in S. miltiorrhiza Tissues and in Response to ABA As Figure 3A shows, these three genes were ubiquitously expressed in the roots, stems, and leaves of S. miltiorrhiza. No clear differences were found in the expression levels of SmSnRK2.3 among the tissues. However, the expression levels of SmSnRK2.6 and SmAREB1 in the leaves were significantly higher (p < 0.01) than those in the roots and stems, while that in the roots and stems were very similar.
As Figure 3B shows, when ABA treatment began, the expression level of SmSnRK2.3 declined slightly and then remained steady until 6 h, after which it rose significantly (p < 0.01), peaking at 24 h, followed by sharp decrease to a minimum at 48 h. A similar pattern was seen under the ABA treatment, in that after a brief and slight decline at 0.5 h, the expression of SmSnRK2.6 and SmAREB1 began to increase, peaking at 3 h, but then declined sharply to a minimum at 12 h (with the exception of a small rise at 9 h), followed by another significant peak (p < 0.01) at 24 h before it declining again at 48 h (Figures 3C,D).

Subcellular Localization of SmSnRK2.3, SmSnRK2.6, and SmAREB1
A detailed investigation of the subcellular distribution of target proteins should enhance our understanding of their functions. As shown in Figure 4A, the N. benthamiana epidermal cells, which were infected by A. tumefaciens and harbored the empty vector pCAMBIA1301, showed a ubiquitous fluorescent distribution. However, both SmSnRK2.3-p1301 and SmSnRK2.6-p1301 fusion proteins generated GFP signals not only in the cell membrane, since it overlapping with the plasma membrane marker, but also in the cytoplasm and nucleus of the N. benthamiana epidermal cells. These results indicated that SmSnRK2.3 and SmSnRK2.6 were located in the cell membrane, cytoplasm, and nucleus, which are consistent with the subcellular localization of TaSnRK2.4/2.7/2.8, and ZmSAPK8 Zhang H. et al., 2010;Ying et al., 2011;Zhang et al., 2011). The SmAREB1-p1301 fusion protein was localized in the nucleus exclusively, since overlapping with the DAPI nuclear dye ( Figure 4B).

Transcriptional Activation Analysis of SmAREB1
The AH109 yeast cells containing pGBKT7, AtMYB15-pGBKT7, and SmAREB1-pGBKT7 grew well on the SD/-Trp medium. However, on SD/-Trp-His-Ade medium, the yeast cells harboring both the SmAREB1-pGBKT7 and negative control pGBKT7 constructs were unable to grow. By contrast, those cells harboring the positive control, AtMYB15-pGBKT7, did grow well ( Figure 5). Together, these results demonstrated that SmAREB1 did not activate transcription in the AH109 yeast cells. 1 | Potential cis-acting regulatory elements in the SmSnRK2.3, SmSnRK2.6, and SmAREB1 promoter with the exception of core cis-acting elements (TATA box and CAAT box).

Cis-element
Number  Data represent means ± SD of three replicates. Asterisks denote significant differences compared to control samples at * P < 0.05, * * P < 0.01, according to Tukey's test. Physical Interaction between SmSnRK2.3/2.6 and SmAREB1 To investigate whether the protein kinase SmSnRK2.3/2.6 interacts with the SmAREB1 transcription factor, Y2H and BiFc assays were simultaneously utilized for in vitro and in vivo verification. Yeast cells co-transformed by SmSnRK2.3/2.6-AD + SmAREB1-BD not only grew well on SD/-Trp-Leu medium, but also grew on the SD/-Trp-Leu-His-Ade medium, in addition to their ability to turn blue in the X-β-Gal staining assay. However, all the yeast cells that harbored the negative controls could only grow on the SD/-Trp-Leu medium, but not on SD/-Trp-Leu-His-Ade medium ( Figure 6A). The BiFc analysis revealed strong YFP fluorescent signals in the N. benthamiana leaf epidermal cells infected by the A. tumefaciens strain EHA105, which harbored the SmSnRK2.3/2.6-pSPYNE + SmAREB1-pSPYCE plasmids. However, no YFP fluorescent signals were observed in any of the negative controls ( Figure 6B). In Combination, these analyses indicated that the SmSnRK2.3/2.6 protein could interact with the SmAREB1 transcription factor.

Protein Purification and LC-MS/MS Analysis of SmSnRK2.3/2.6
Following the Coomassie Brilliant Blue R250 staining, the SDS-PAGE results revealed that the molecular weights of the SmSnRK2.3/2.6 (fused with GST-tag protein) agreed with their previously predicted size (including the 26 kDa of the GSTtag protein) (Supplementary Figures S1A, 2A). Through the induced expression and ultrasonic breaking of the E. coli cells, purified proteins were obtained, under native conditions without denaturation, which could be verified via Western blotting (Supplementary Figures S1B, 2B), and analyzed by the LC-MS/MS system. The LC-MS/MS analysis revealed that this method achieved a 37.78% sequence coverage of the SmSnRK2.3 fusion protein, which was almost half that of SmSnRK2.6 (73.15%). Accordingly, we found several phosphorylated peptides and 14 nonredundant phosphorylation sites in SmSnRK2.6, but only 3 non-redundant phosphorylation sites were found in SmSnRK2.3 ( Table 2). In this study, positive transgenic hairy root lines overexpressing SmSnRK2.3, SmSnRK2.6, and SmAREB1 (i.e., SmSnRK2.3-OEs, SmSnRK2.6-OEs, and SmAREB1-OEs) were obtained and identified via PCR, to evaluate the contribution of SmSnRK2.3/2.6 and SmAREB1 to the regulation of the phenolic acid metabolism in S. miltiorrhiza. The contents of Sal B and RA in the SmSnRK2.3-OEs were apparently very similar to those in the empty-vector control lines (EVs) (results not shown). However, we found a remarkable increase (p < 0.05) of Sal B and RA contents in SmSnRK2.6-OE13, and these contents were significantly enhanced (p < 0.01) in SmSnRK2.6-OE2, SmAREB1-OE4, SmAREB1-OE28, and SmAREB1-OE29, when compared with those in the EVs. In SmSnRK2.6-OE29, whereas only the RA content was significantly higher (p < 0.01), the Sal B content did not obviously increase (p > 0.05) relative to the EVs (Figures 7A,B).

Effects of SmSnRK2.6 and SmAREB1 Overexpression on Expression Levels of Structural Genes
In this study, SmSnRK2.6-OE2, SmSnRK2.6-OE13, SmAREB1-OE28, and SmAREB1-OE29 were selected as representatives to investigate the molecular mechanism enabling SmSnRK2.6 and SmAREB1 overexpression to increase the contents of phenolic acids in the S. miltiorrhiza hairy roots.
The qRT-PCR was performed with two aims: First, to analyze the expression level of structural enzyme genes, namely phenylalanine ammonia lyase 1 (SmPAL1), cinnamic acid 4hydroxylase (SmC4H), 4-coumaric acid CoA-ligase 1 (Sm4CL1), tyrosine aminotransferase (SmTAT), 4-hydroxyphenylpyruvate reductase (SmHPPR), and RA synthase (SmRAS), which participate in the phenolic acid biosynthetic pathway in S. miltiorrhiza. Second, to check the expression level of the key enzyme genes located in the entry point to the other side-branch pathways, namely that of chalcone synthase (SmCHS, key enzyme in the flavonoid pathway), cinnamoyl-CoA reductase and caffeic acid O-methyltransferase (respectively, SmCCR and SmCOMT, key enzymes in the lignin pathway), in addition to 4-hydroxyphenylpyruvated dioxygenase (SmHPPD) which competes for the same substrate as SmHPPR.

DISCUSSION
ABA plays key regulatory roles in plant growth and development processes, such as seed dormancy and germination, fruit ripening, and stomatal closure (Chen et al., 2016). It is also considered a stress hormone, because its rapid accumulation in plants under stressful conditions can function to protect them against various environmental stresses (Tuteja, 2007). Exogenous ABA treatment could mimic the effects of osmotic stresses and cold stress (Roychoudhury et al., 2013). In our previous work, we found that exogenous ABA could promote the accumulation of phenolic acids in S. miltiorrhiza hairy roots . Increasing evidence shows that the members of subclass III SnRK2 may be strongly activated when treated with ABA, and they have been found to function as the main positive regulators of ABA-dependent signal transduction (Fujii et al., 2011). Therefore, we speculate that members of subclass III SnRK2s in S. miltiorrhiza will be good candidate genes for applied use in genetic engineering approaches, whether to improve the ability of S. miltiorrhiza to resist stresses or to boost its content of active ingredients, or both.
In this study, we first identified two genes SmSnRK2.3 and SmSnRK2.6, belonging to subclass III of SnRK2, from the S. miltiorrhiza hairy roots. PlantCARE analysis showed that their promoters contained ABA and stress response elements   Values are mean ± SD, n = 3. Asterisks indicate significant differences between transgenic lines and EV control, obtained via Student's t-test ( * P < 0.05, * * P < 0.01).
( Table 1); these results supported the qRT-PCR results, which confirmed that SmSnRK2.3 and SmSnRK2.6 were strongly induced by the exogenous ABA treatment (Figures 3B,C). Further, the analysis of their amino acid sequences showed that they both contained the myristyl N-terminal myristoylation region (Figure 1A), which is essential for protein function in mediating membrane targeting and signal transduction in plant responses to environmental stress (Podell and Gribskov, 2004). Based on the above results, we suggest that both SmSnRK2.3 and SmSnRK2.6 are involved in the ABA and stress response of S. miltiorrhiza.
Identifying the phosphorylation sites of protein kinases will provide strong evidence for a kinase activity assay, in addition to valuable information for elucidating and understanding the operation of signaling networks based on phosphorylation. The LC-MS/MS analysis of the SmSnRK2.3/2.6-GST fusion protein found 14 non-redundant phosphorylation sites in SmSnRK2.6 and 3 non-redundant sites in SmSnRK2.3 ( Table 2); a plausible reason for this result may be the low abundance of SmSnRK2.3 protein phosphorylation or a low ionization efficiency of the phosphorylated peptides. Among the identified non-redundant phosphorylation sites in SmSnRK2.3/2.6, we found that Thr178 was in the kinase activation loop of SmSnRK2.3, while Ser168, Ser170, Ser171, Ser175, Ser179, Thr180, and Thr183 were in the kinase activation loop of SmSnRK2.6. The kinase activation loop is reportedly essential for the kinase activity of SnRK2s (Umezawa et al., 2009;Vlad et al., 2010). These results suggest that SmSnRK2.3 and SmSnRK2.6 might be autophosphorylated during recombinant expression and purification, which is an interpretation consistent with other findings (Vlad et al., 2010;Xie et al., 2012).
by the exogenous ABA treatment ( Figure 3D). SmAREB1 was deemed a characteristic bZIP transcription factor, based on the results of our amino acid sequence analysis ( Figure 1B) and subcellular localization assay of SmAREB1 ( Figure 4B). However, SmAREB1 had no activity in the transcriptional activation assay (Figure 5). According to literature, the activation of AtAREB1 reportedly requires the ABA-dependent posttranscriptional phosphorylation of Ser/Thr residues in the AtAREB1 conserved regions (Furihata et al., 2006). Given that SmAREB1 contains four conserved domains very similar to those of AtAREB1 (Figure 1B), we speculated that the SmAREB1 protein requires phosphorylation by SnRK2 family proteins, or it depends upon regulation by other proteins to perform its transcriptional regulation. The result that the SmSnRK2.3/2.6 protein interacted with the SmAREB1 protein (Figures 6A,B) lends support to this conjecture: i.e., SmSnRK2.3/2.6 might directly phosphorylate SmAREB1 to "switch on" the activity of SmAREB1. This phosphorylation role still requires validation and could be verified through an in-gel kinase assay in future research.
Furthermore, we investigated whether or not SmSnRK2.3, SmSnRK2.6, and SmAREB1 participated in the regulation of phenolic acid synthesis. To do this, overexpression vectors of SmSnRK2.3/2.6 and SmAREB1 were built under the control of the CaMV35S promoter to obtain SmSnRK2.3-OEs, SmSnRK2.6-OEs, and SmAREB1-OEs. In this respect, the hairy root system we used is considered optimal for studying the metabolic regulation of effective components in S. miltiorrhiza, as it has the following advantages: high-level productivity of secondary metabolites, stable hereditary, rapid growth under simple conditions, and inclusion of the characteristic secondary metabolic pathway of the parent plants (Chen et al., 1999;Guillon et al., 2006;Kai et al., 2011).
Overexpression of SmSnRK2.3 did not significantly increase the contents of Sal B and RA, whereas overexpression of SmSnRK2.6 and SmAREB1 significantly increased both (Figures 7A,B). Two plausible explanations for these results are as follows: (1) The qRT-PCR analyses revealed that SmSnRK2.3 was ubiquitously expressed in roots, stems, and leaves of S. miltiorrhiza, but to the same degree in all tissues. However, the greatest expression levels of SmSnRK2.6 and SmAREB1 were detected in the leaves. A larger accumulation of phenolic acids has been reported in S. miltiorrhiza leaves (Hang et al., 2008). So overexpression of SmSnRK2.6 and SmAREB1 are more conducive to promote substantial accumulation of phenolic acids, compared to that of SmSnRK2.3. (2) The LC-MS/MS analysis of the SmSnRK2.3/2.6-GST fusion protein indicated that SmSnRK2.3 had a lower phosphorylated abundance, when compared with SmSnRK2.6, which suggests that SmSnRK2.3 cannot effectively activate SmAREB1 to regulate downstream gene expression. As a drought-sensitive plant species, S. miltiorrhiza resists oxidative stress via the accumulation of phenolic acids (Bettaieb et al., 2011). Therefore, we speculate that the overexpression of SmSnRK2.6 and SmAREB1 would increase the contents of phenolic acids, thus enhancing the antioxidant activity of S. miltiorrhiza (leading to an increased tolerance of osmotic stresses).
It has been reported that rosmarinic acid synthase (RAS) is the most specific enzyme in the phenolic acid biosynthetic pathway, because it couples products from the phenylpropanoid and tyrosine-derived pathways (Di et al., 2013;Wang et al., 2015). Irrespective of how the expression levels were regulated of the other structural genes participating in phenolic acidbranched and side-branched pathways, the RAS transcripts were significantly enhanced in the SmSnRK2.6-OEs and SmAREB1-OEs. Our qRT-PCR results revealed that the overexpression of SmSnRK2.6 and SmAREB1 ultimately promoted more metabolic flux to the phenolic acid-branched pathway, by regulating the expression levels of structural enzyme genes that participate in the phenolic acid-branched and competition bypass-branched pathways (Figures 8A,B). Besides, the expression levels of SmCHS in the SmSnRK2.6-OEs were significantly higher than those in the EVs. SmCHS is the key enzyme in the entry point to the flavonoid pathway; these results indicate that an overexpression of SmSnRK2.6 might also promote the accumulation of flavonoids, which have long been recognized as playing multiple roles in the responses of higher plants to a wide range of environmental stresses (Agati and Tattini, 2010). Prior studies reported the existence of ABRE cis-acting elements in the promoters of key enzyme genes involved in the biosynthetic pathway of phenolic acid (Huang et al., 2008;Song et al., 2012). In the future, we plan to conduct electrophoretic mobility shift assay (EMSA) and other experiments to verify whether SmAREB1 regulates the expression of these key enzyme genes by directly binding to the ABRE cis-acting elements in their promoter regions.

CONCLUSION
Three novel genes named SmSnRK2.3, SmSnRK2.6, and SmAREB1 were cloned from S. miltiorrhiza hairy roots and functionally characterized. Comparing to SmSnRK2.3, overexpression of SmSnRK2.6 significantly promotes the accumulation of RA and Sal B in the S. miltiorrhiza hairy roots, by effectively regulating the transcription activity of SmAREB1. Moreover, it is suggested that SmSnRK2.6 was involved in ABA and stress response of S. miltiorrhiza. Therefore, this study provides one candidate gene, SmSnRK2.6, for breeding transgenic S. miltiorrhiza lines featuring improved tolerance to abiotic stresses and increased active ingredients. Furthermore, these results provide a theoretical foundation to elucidate the molecular mechanism underlying the ability of exogenous ABA to increase the content of phenolic acids in S. miltiorrhiza hairy roots.

ACCESSION NUMBERS
The cDNA sequences of SmSnRK2.3, SmSnRK2.6, and SmAREB1 in S. miltiorrhiza cloned in this study were deposited in GenBank under the accession No. MF185314, No. MF185316, and No. MF185317.

AUTHOR CONTRIBUTIONS
ZL designed the research and led this project. YJ, ZB, TP, and KD carried out the experiments and analyzed the results. YJ wrote the manuscript. YG provided scientific advice and revised the manuscript. All authors have read and approved the final manuscript.