The NAC-like transcription factor CsNAC7 positively regulates the caffeine biosynthesis-related gene yhNMT1 in Camellia sinensis

Abstract Caffeine is an important functional substance and is abundant in tea plant, but little is known about how its biosynthesis is regulated by transcription factors. In this study, the NAC-like transcription factor-encoding gene CsNAC7, which is involved in caffeine synthesis, was isolated from a Yinghong 9 cDNA library using a yeast one-hybrid assay; this gene comprises 1371 bp nucleotides and is predicted to encode 456 amino acids. The expression of CsNAC7 at the transcriptional level in tea shoots shared a similar pattern with that of the caffeine synthase gene yhNMT1 in the spring and summer, and its expressed protein was localized in the nucleus. Assays of gene activity showed that CsNAC7 has self-activation activity in yeast, that the active region is at the N-terminus, and that the transient expression of CsNAC7 could significantly promote the expression of yhNMT1 in tobacco leaves. In addition, overexpression or silencing of CsNAC7 significantly increased or decreased the expression of yhNMT1 and the accumulation of caffeine in transgenic tea calli, respectively. Our data suggest that the isolated transcription factor CsNAC7 positively regulates the caffeine synthase gene yhNMT1 and promotes caffeine accumulation in tea plant.


Introduction
Tea [Camellia sinensis (L.) O. Kuntze] is rich in functional metabolites and is highly consumed worldwide. Constituting the main functional metabolites, alkaloids include caffeine, theobromine and theophylline. Caffeine is the main component of tea alkaloids, accounting for 2%-4% of tea dry weight, and has been verified to have many beneficial effects, such as causing refreshed feelings, reducing high blood pressure and reducing anxiety [1][2][3]. Therefore, caffeine has received much more attention due to its unique physiological activity, and various related studies, including studies on its biosynthesis mechanism, have also been conducted in depth.
Caffeine was first found in tea plant in the early 1820s, and its main synthesis pathway is xanthosine (XR) → 7-methylxanthosine (7-mXR) → 7-methylxanthine (7-mX) → theobromine (Tb) → caffeine (Cf); N-methyl transferase (NMT) plays a catalytic role in the threestep methylation reaction in this pathway [4]. The first NMT gene cloned from tea leaves, TSC1, exhibits twostep transmethylation activity, including catalysis of 7-mX to Tb and Tb to Cf [5]. Subsequently, more NMT genes, including TCS2, PCS1, PCS2, ICS1, ICS2 and yhNMT1, have been isolated from Camellia ptilophylla, Camellia irrawadiensis and Camellia sinensis respectively [6,7]. Previous studies have shown that NMT expression in tea plant not only is closely related to caffeine content but also has typical spatiotemporal specificity, suggesting that NMT gene expression is complex and may be regulated by specific factors [8,9].
Transcription factors (TFs), important regulatory proteins, have been proven to play a key role in the metabolism of alkaloids, and many TFs, including AP2/ERF, WRKY, bHLH, Myb-like, bZIP, TFIIIA zinc finger and AT hook TFs, are involved in the biosynthesis of alkaloids by regulating the expression of related synthetase-encoding genes [10]. In Catharanthus roseus, the biosynthesis of vincristine alkaloids has been verified to be regulated by TFs targeting STR and TDC gene expression, including CrWRKY1, CrMYC1, ORCA2 and ORCA3 TFs [11][12][13][14]. In tobacco, nicotine biosynthesis has been confirmed to be regulated by TFs including NbbHLH1, NtMYC2a and NtMYC2b targeting synthaseencoding genes [15,16]. NAC, a plant-specific TF, has been found in dozens of plant species, such as maize, apple and Coffea canephora [17][18][19][20][21]. Studies have shown that NACs are widely involved in plant metabolism, including basic physiological development, secondary metabolite synthesis and environmental stress responses [22][23][24][25]. In apple, MdNAC52 promotes the biosynthesis of anthocyanins and proanthocyanidin [19]. In tobacco, NtNAC-R1 has been shown to regulate the biosynthesis of nicotine [26]. Recently, transcriptome analysis of tea varieties with different caffeine contents showed that NACs are related to the biosynthesis of purine alkaloids [27]. Nevertheless, no studies have revealed how the synthesis of caffeine is regulated by transcription factors in tea plant.
In this study, Yinghong 9 tea plants were selected, and the transcription factor CsNAC7 was cloned using yeast one-hybrid (Y1H) assays via the promoter of yhNMT1 used as bait. The cloned TFs showed 99.12% homology with the NAC gene registered in the National Center for Biotechnology Information (NCBI, https://www.ncbi.nlm. nih.gov/) GenBank database for tea. To determine the function of CsNAC7 in caffeine accumulation by which the expression of the caffeine synthase gene yhNMT1 is regulated in tea plant, the spatiotemporal expression of CsNAC7 and yhNMT1 was determined by real-time fluorescence quantitative PCR (qRT-PCR), and the transcriptional activation of CsNAC7 was analyzed by using both transient expression and yeast two-hybrid technology. Moreover, CsNAC7 was overexpressed and silenced in tea calli, and the changes in yhNMT1 expression and caffeine content in transgenic calli were also determined. Our work and results are of great significance to further understand the molecular mechanism of caffeine metabolism in terms of TFs in tea plant.

cDNA library construction and candidate TFs of yhNMT1 via Y1H assays
To reveal the expression of the caffeine synthase gene yhNMT1 regulated by TFs in tea plant, a cDNA library was constructed, followed by TF screening via Y1H assays. For construction of the cDNA library, the promoter of yhNMT1 was amplified, digested with Hind III and Xho I, cloned, and then inserted into the pAbAi vector pretreated with Hind III and Xho I. The recombinant P NMT1 -AbAi vector was selected as a bait vector and transformed into the Y1H Gold strain. Total RNA was extracted from the shoots of Yinghong 9 plants and subjected to SMART RT-PCR. Then, SMART-obtained cDNA was separated by a Chroma column, which yielded more than 500 bp of ds-cDNA. The purified ds-cDNA and the linearized pGADT7-Rec prey vector were cotransformed into a new host yeast strain using a ClonExpress II One Step Cloning Kit (Takara, Tokyo), and a cDNA library was generated. The calculated capacity of 2.7 × 10 [6] CFU, which was higher than the required 1.0 × 10 [6] CFU, verified the success of the cDNA library.
Moreover, to identify TFs by screening the constructed cDNA library, an experiment to determine the minimum concentration of AbA inhibiting the growth of a new host yeast strain with the bait vector P NMT1 -AbAi was conducted. Our results showed that the new developed host yeast strain could not grow on SD/−Leu media plus 150 ng.mL −1 AbA (Fig. 1A); thus, a higher concentration of AbA (200 ng.mL −1 ) was selected for addition to SD/−Leu media for TF screening from the cDNA library (Fig. 1B). As the screening results showed, one colony grew on the selected media.
Analysis of the TF-encoding gene CsNAC7 screened from the cDNA library Colonies growing in SD/−Leu plus AbA were selected, and the recombinant pGADT7 plasmid was isolated followed by sequencing. Our data showed that the cDNA that was integrated into the pGADT7 vector had a complete open reading frame (ORF) containing 1371 bp nucleotides and encoding a predicted 456 amino acids, with a molecular weight of 49 Fig. 2A). Further phylogenetic analysis showed that CsNAC7 from Yinghong 9 exhibited a high evolutionary relationship with NAC proteins from Gramineae and woody plant species registered in the NCBI GenBank database, especially Actinidiaceae (A. chinensis), which was assigned to Group 1 ( Fig. 2B and Supplemental Table S2).

CsNAC7 Acts as a Transcription Activator in Yeast AH109
To determine whether the CsNAC7 gene has selfactivation activity and which region of the gene has an activation effect, assays of transcriptional activators were conducted in yeast AH109 cells. Based on the results of our bioinformatics analysis, full-length CsNAC7 and two truncated fragments, including a 900 bp N-terminal sequence and 468 bp C-terminal sequence fused to the GAL4-binding domain, were amplified and inserted into the pGBKT7 vector, and the recombinant vectors were designated as NAC7, NAC7 N and NAC7 C (Fig. 3A). The three recombinant vectors and the negative control vector pGBKT7 were transformed into yeast AH109 cells, which were then inoculated onto the designated media (Fig. 3B). The results showed that all the strains could grow in SD/−Trp media; however, only two strains with NAC7 and NAC7 N could grow in SD/−Trp-His-Ade media, and the color of the colonies turned blue when X-a-gal was added to the media. Our data confirmed that the CsNAC7 gene has self-activation activity and that the activation region is located at the N-terminus.

Transient expression of CsNAC7 in tobacco reveals its TF role
To explore whether CsNAC7 has a regulatory effect on yhNMT1, a reporter vector P NMT1 -GUS and effector construct pCaMV35S-CsNAC7 were constructed, and then the two constructed vectors and positive-control plasmid pCAMBIA1301 (pCaMV35S-GUS) were transformed into Agrobacterium EHA105, which were subsequently injected into tobacco leaves (Fig. 4A). The results of GUS histochemical staining and GUS activity determination of the injected leaves showed that GUS activity was the highest in the positive control; moreover, the GUS activity in leaves cotransformed with the reporter vector (P NMT1 -GUS) and effector construct (pCaMV35S-CsNAC7) was significantly higher than that of the reporter vector alone ( Fig. 4B-C), which indicated that the expression of CsNAC7 driven by the yhNMT1 promoter had a significant effect on the expression of GUS in tobacco leaves.

Subcellular localization of CsNAC7
To determine the subcellular localization of the transcription factor CsNAC7, the vector pGreen-C18-GFP (35S:GFP) was selected, and the full ORF of CsNAC7 fused to GFP driven by the 35S promoter was inserted into a 35S:GFP vector to generate the recombinant vector 35S:CsNAC7-GFP (Fig. 5A). The recombinant vectors 35S:CsNAC7-GFP and 35S:GFP were transformed into Agrobacterium strain GV3101. GV3101 containing the 35S:GFP or 35S:CsNAC7-GFP vector was injected into tobacco leaves (Nicotiana benthamiana) and observed by fluorescence microscopy (Fig. 5B). GFP fluorescence was observed in the nucleus and cytoplasm of 35S:GFPtransformed leaves, but GFP fluorescence was observed only in the nucleus of the 35S:CsNAC7-GFP-transformed leaves, which indicated that the expressed protein of CsNAC7 was located in the nucleus of tobacco.

Expression analysis of CsNAC7 and yhNMT1 in tea shoots
Different parts of tea shoots (including the buds, the first leaf and the second leaf) in different seasons (spring, summer and autumn) were collected (Fig. 6A), and the expression of CsNAC7 and yhNMT1 was measured by qRT-PCR (Fig. 6B). Our results showed that CsNAC7 and yhNMT1 exhibited similar expression changes in different parts of tea shoots in the spring and summer. However, in autumn, the expression pattern of CsNAC7 was different from that of yhNMT1, and the expression of yhNMT1 gradually decreased from the buds to the second leaf; nevertheless, the expression of CsNAC7 in the first leaf decreased significantly from summer to autumn, and the expression levels in the buds, first leaf and second leaf were the same, with no significant difference.

Overexpression and silencing of CsNAC7 affects caffeine accumulation in tea Calli
To further elucidate the effect of the TF CsNAC7 on the expression of yhNMT1 and caffeine accumulation in tea plant, recombinant vectors pCAMBIA1031-35SN-CsNAC7 and RNAi-CsNAC7 for overexpression and silencing of CsNAC7, respectively, were constructed by inserting the CsNAC7 gene into pCAMBIA1031-35SN and pYLRNAi-35S, respectively. Genetic transformation was conducted via Agrobacterium infection of tea calli, which involved the use of newly generated calli resistant to hygromycin ( Fig. 7A-D). Genomic DNA was extracted from the resistant calli and used as template for HPT gene amplification, and the PCR results verified the success of the transgene. Then, the transgenic calli were designated as P7-1 and R7-1, which corresponded to overexpression and silencing of CsNAC7, respectively (Supplemental Figure S2).
Total RNA was extracted from the transgenic calli, and the expression levels of yhNMT1 and CsNAC7 were detected by qRT-PCR (Fig. 7E). Our results showed that the expression of CsNAC7 in P7-1 was 6 times higher than that in non-transgenic CK calli, and the expression level of the caffeine synthase gene yhNMT1 was increased by 2.46 times compared with that in the control, indicating that overexpression of the CsNAC7 gene in P7-1 transgenic calli increased the expression of the yhNMT1 gene.
In contrast, the expression of CsNAC7 and yhNMT1 in R7-1 was decreased by 60% and 46%, respectively, compared with that in the control, indicating that the silencing of CsNAC7 in transgenic calli decreased the expression of yhNMT1. Moreover, caffeine accumulation in calli was determined by HPLC (Supplemental Figure S1), and the content of caffeine in P7-1 increased to 1398 μg.g −1 and  reached 1.28 times that in the control (1094 μg.g −1 ); nevertheless, the content of caffeine in R7-1 decreased to 358 μg.g −1 and was reduced by 67% (Fig. 7F). Therefore, overexpression of CsNAC7 promoted the expression of yhNMT1 and the accumulation of caffeine, whereas inhibition of CsNAC7 decreased the expression of yhNMT1  The expression levels of the CsNAC7 and yhNMT1 genes in the spring buds were set to 1 for comparative analysis. The different letters show significant differences between values, with p < 0.05. The capital letters represent significant differences in CsNAC7, and the lowercase letters represent significant differences in yhNMT1. and the accumulation of caffeine in the calli of Yinghong 9. In conclusion, the transcription factor CsNAC7 positively regulates the yhNMT1 gene and promotes caffeine accumulation in calli.

Discussion
Alkaloids are important secondary chemical substances in plants, and their biosynthesis is mainly catalyzed by synthetase genes. Previous studies have shown that many transcription factors are involved in the regulation of plant alkaloid synthase genes. The biosynthesis of terpenoid indole alkaloids in C. roseus was inhibited by the transcription factors CrGBF1 and CrGBF2 by inhibiting the expression of the strictosidine synthase gene [28]. A similar phenomenon was found in C. roseus, in which light-induced biosynthesis of vindoline was regulated by the transcription factor CrPIF1 by inhibiting CrGATA1 gene expression [29]. On the other hand, transcription factors also promote the synthesis of plant alkaloids. In C. roseus, overexpression of the transcription factors ORCA3 and G10H significantly increased the accumulation of monoindole indole alkaloids [14]; overexpression of CrERF5 can increase the expression of key genes in the biosynthesis pathway of monoterpenoid indole alkaloids and upregulate the synthesis of bisindole alkaloids [30]. In tobacco, the transcription factors ERF189, ORCA3 and AtERF13 have been found to activate PMT genes related to nicotine and tropane alkaloid biosynthesis [31]; the transcription factor NtWRKY-R1 integrates JA and IAA signals and regulates the expression of nicotine synthesis-related genes [32]. In addition, transcriptome analysis showed that 549 transcription factor genes belonging to the MYB, bHLH, NAC, bZIP, WRKY, GRAS and other families were involved in the regulation of nicotine biosynthesis in tobacco after topping was performed [33]. Caffeine is one of the most important alkaloids in tea plant, and little is known about the regulation of caffeine biosynthesis by transcription factors. Transcriptome analysis of four tea varieties with high, normal and low caffeine contents showed that NAC, bHLH, WRKY, GRAS and MYB TFs were related to the biosynthesis of purine alkaloids, but the regulatory network of caffeine biosynthesis in tea is still unclear [27]. In this study, we cloned the NAC TF CsNAC7 from tea plant and confirmed its regulatory role in the expression of the caffeine synthase gene. Overexpression or silencing of CsNAC7 in tea calli leads to a significant increase or decrease, respectively, in yhNMT1 expression and accumulation.
As plant-specific transcriptional regulators, NACs are widely involved in the regulation of plant secondary metabolite synthesis. In A. chinensis, three transcription factors, AaNAC2, AaNAC3 and AaNAC4, are involved in monoterpene production by binding to the promoter of AaTPS1 [34]. In V. vinifera, two NAC genes probably participate in the regulation of anthocyanin accumulation [35]. In peach, PpNAC1 acts as a transcriptional activator of PpMYB10.1, regulating the synthesis of anthocyanin pigmentation [36]. In Arabidopsis thaliana, VaNAC17 is involved in jasmonic acid biosynthesis by upregulating the transcription of the LOX3, AOC1 and OPR3 genes [37], and overexpressing ANAC046 significantly increases the expression of suberin biosynthesis-related genes [38]. Interestingly, the NAC-like transcription factor CsNAC7 isolated from Yinghong 9 tea plant in our study proved to be involved in the caffeine biosynthesis regulatory network by activating yhNMT1 expression.
How to isolate and scientifically determine the functional activity of TFs is of great significance for understanding the synthesis of metabolites in organisms. Y1H assays combined with transient expression technology is generally used to study metabolic regulation. In sweet cherry, Y1H assays revealed that the transcription factor PaMADS7 could directly bind to the PaPG1 promoter, which is involved in fruit softening [39]. In apple, MdLUX and MdPCL-like transcription factors were found via Y1H assays to bind to the promoter of the anthocyanin biosynthesis-related gene MdF3H [40]. In Taraxacum antungense, a Y1H assay proved that the transcription factor TaWRKY14 binds to the W-box of the TaPAL1 promoter, which is related to the accumulation of chlorogenic acid [41]. In this study, Y1H technology was also adopted to screen TFs from Yinghong 9 tea plant with the use of the promoter of yhNMT1 as bait, and as a result, the transcription factor CsNAC7 was successfully screened on the basis of its ability to bind to the promoter. On the other hand, transient expression using GUS as a reporter is often used to reveal the regulatory function of transcription factors. In soybean calli, GUS activity analysis showed that the TF GmERF6, a transcriptional repressor, downregulated the expression of AtPDF1.2 and AtPR4 [42]. In Nicotiana benthamiana leaves, GUS activity analysis revealed that the TF CrGATA1 isolated from C. roseus could strongly activate the expression of D4H [29]. In tobacco, GUS activity analysis demonstrated that the TFs ORCA3 and AtERF13 could promote the expression of PMT2 [31]. In our study, transient expression in tobacco leaves was successfully adopted, and the TF CsNAC7 was verified via GUS staining and GUS activity analysis to activate the expression of yhNMT1.
Transgenic plant technology provides a good platform for gene functional research. However, some species are difficult to regenerate, which limits the application of transgenic technology to reveal gene function. Callus, which is induced in vitro and exhibits metabolic characteristics similar to those of cultivated plants [43][44][45], provides a new platform for revealing the function of genes and has been successfully used in sugarcane, apple, grapevine, carica papaya and melon [46][47][48][49][50]. TFs are important proteins in plants, and functional research via callus platforms in which transgene technology was used has also been conducted. In chrysanthemum, the regulatory gene eIF5B1 was introduced into calli, and transgenic calli with high temperature tolerance [51] were obtained. In blueberry, overexpression of the transcription factor VcMYB4a in calli enhanced the sensitivity of transgenic calli to salt, drought, cold, freezing, and heat stress [52]. Similarly, in grapefruit, overexpression of the transcription factor CsPIF8, which was isolated from Citrus sinensis calli, increased the tolerance of transgenic calli to cold [53]. Moreover, reports have shown that overexpression of the transcription factor CmAGL11 in Chinese chestnut calli could enhance somatic embryogenesis [54]. Tea plants are perennials used for beverages, and regeneration in vitro is difficult due to the abundance of secondary metabolites, which conversely hinders functional research on tea genes [55]. To date, the function of only the TCS1 gene isolated from tea plant has studied through transgenic plant regeneration technology in tea [56]. Fortunately, our previous study verified that transgenic calli induced from tea leaves could be employed to reveal gene function [57]. In the present study, a transgene was generated in tea calli induced from Yinghong 9 by using overexpression and silencing technology, and it was found that the transcription factor CsNAC7 could positively affect yhNMT1 expression and the accumulation of caffeine in the transgenic calli, which verified the regulatory function of CsNAC7 in caffeine metabolism in tea. In addition, the established transgenic overexpression and silencing methods in our research will help to reveal the functions of other genes involved in the development and metabolism of tea plant. Overall, our data demonstrated that the synthesis of caffeine is regulated by transcription factors as well as by caffeine synthase genes, which indicates the complexity of the molecular mechanism underlying caffeine metabolism in tea plant (Fig. 8).

Plant materials and growth conditions
Camellia sinensis Yinghong 9 is a tea variety bred from a single plant selected from an inbreeding population of large-leaf Yunnan species. The caffeine content of Yinghong 9, a common tea variety, is similar to that of other tea plant varieties, but Yinghong 9 is rich in tea polyphenols, is suitable for the production of black tea and is widely planted in Guangdong Province [58]. New shoots of Yinghong 9 were collected from tea plantations at South China Agricultural University, Guangzhou, China. Tea calli were induced from Yinghong 9 as previously reported [45].

Construction of a complementary DNA (cDNA) library and screening candidate TFs via Y1H
The pAbAi vector (Clontech) was selected for construction of the bait plasmid. The promoter of yhNMT1 was amplified from the cloned plasmid with the specific primers P NMT1 -F and P NMT1 -R together with the action of restriction endonuclease Hind III and Xho I recognition sequences at the 5 end, respectively (Supplemental Table S1). The amplified product was treated with Hind III and Xho I and ligated into the pAbAi vector, which was pretreated with Hind III and Xho I, transformed into the yeast Y1H Gold strain (Clontech) and grown on synthetic dropout (SD)/-Ura media; the strains were transformed with the bait recombinant vector P NMT1 -AbAi, yielding bait yeast/P NMT1 -AbAi strains. Then, the strains resistant to aureobasidin A (AbA) were grown in SD/-Ura media plus 0 ng.mL −1 , 25 ng.mL −1 , 50 ng.mL −1 , 75 ng.mL −1 , 100 ng.mL −1 , 125 ng.mL −1 and 150 ng.mL −1 AbA. Moreover, a total of 800 mg of leaves was crushed in liquid nitrogen, and total RNA was extracted using Plant RNA Purification Reagent (Takara, Tokyo) according to the manufacturer's instructions. First-strand cDNA was synthesized using a SMARTer ® PCR cDNA Synthesis Kit, and amplification was conducted using the Advantage R2 PCR Kit (Clontech) according to the manufacturer's instructions to obtain SMART cDNA, followed by separation with a Chroma spin TE400 column and purification of fragments greater than 500 bp. The purified cDNA was mixed with the pGADT7-Rec prey vector, transformed into the bait yeast/P NMT1 -AbAi strain using a ClonExpress II One Step Cloning Kit (Vazyme Biotech Co., Ltd.), and used to construct a cDNA library. The capacity of the cDNA library was calculated by growing the bacteria on SD/−Leu media, and then the candidate transcription factors were screened by inoculating the library solution into SD/−Leu media plus 200 ng.mL −1 AbA. Positive colonies were identified by DNA sequencing, and candidate transcription factors related to yhNMT1 were identified.

Sequence analysis of CsNAC7
The candidate transcription factor gene CsNAC7 was analyzed using online tools of the NCBI database. The isoelectric point and molecular weight of the CsNAC7deduced protein were predicted using ProtParam, multiple sequence alignment was analyzed by DNAMAN, and a phylogenetic tree comprising CsNAC7 and other registered NAC transcription factor genes from related plant species was constructed by MEGA X.

Analysis of the transcriptional activation function of CsNAC7 in yeast AH109
Based on the analysis of the conserved domains of the CsNAC7 gene, the full-length ORF (1371 bp), N-terminal 900 bp sequence and C-terminal 468 bp sequence of CsNAC7 were amplified from the recombinant plasmid by three specific pairs of primers (Supplemental Table S1). Then, the three purified PCR products were cloned into a linearized pGBKT7 vector (Clontech) pretreated with EcoR I and BamH I using a ClonExpress II One Step Cloning Kit (Vazyme Biotech Co., Ltd.), and recombinant vectors designated as pCsNAC7, pNAC7 N and pNAC7 C were obtained. The recombinant vectors were transformed into the yeast strain AH109 and grown on SD/−Trp media. The transformants growing on SD/−Trp media were selected, diluted in 100 μL of water and inoculated onto SD/−Trp-His-Ade plates. After culture at 30 • C for 3-5 days, 4 mg/mL X-a-gal solution was added to the growing yeast colonies until selected single colonies were covered by the X-a-gal solution. Then, the color change of selected single colonies was observed to evaluate the transactivation activity.

Transient expression of CsNAC7 in tobacco leaves
The promoter of yhNMT1 was amplified from the cloned plasmid with the specific primers p1301-F and p1301-R (Supplemental Table S1), which had restriction endonuclease Hind III and Nco I recognition sequences at the 5 end. The amplified product was treated with Hind III and Nco I and ligated into the pCAMBIA1301 vector, which was pretreated with Hind III and Nco I, to obtain the reporter vector P NMT1 -GUS. In addition, full-length CsNAC7 was amplified with the specific primers pRI101-F and pRI101-R (Supplemental Table S1), cloned and ligated into the pRI101 AN vector pretreated with EcoR I and Sal I using a ClonExpress II One Step Cloning Kit (Vazyme Biotech Co., Ltd.) to obtain the effect vector pCaMV35S-CsNAC7. The effect vector and reporter vector were transformed into Agrobacterium EHA105. The infection solution with the two strains of Agrobacterium tumefaciens at a ratio of 1:1 and an OD600 of 0.6 (adjusted by dilution of a solution consisting of 10 mmol·L −1 MES, 150 μmol·L −1 AS and 10 mmol·L −1 MgCl 2 ) was prepared and injected into tobacco leaves, and GUS histochemical staining and determination of GUS enzyme activity were conducted on the injected leaves 3 days later in darkness. For GUS staining, the tobacco leaves were cut into 0.5 cm 2 pieces and soaked in GUS staining solution for 24 hours at 37 • C, followed by washing with 75% ethanol. The GUS enzyme activity was detected with a plant GUS activity assay kit from FCNCS Technology Laboratories. Untreated tobacco leaves were used as negative controls, and tobacco leaves transformed with pCaMV35S-GUS were used as positive controls.

Analysis of subcellular localization
The full-length ORF of CsNAC7 was amplified from the recombinant plasmid by a specific pair of primers (GFP-F and GFP-R) to which the base sequence of pGreen-C18-GFP (35S:GFP) was added to the 5 end (Supplemental Table S1), and the purified PCR products were cloned and inserted into a linearized pGreen-C18-GFP vector pretreated with Hind III and Bam HI; a ClonExpress II One Step Cloning Kit (Vazyme Biotech Co., Ltd.) was used. The recombinant vector was identified and designated as pGreen-C18-CsNAC7-GFP (35S:CsNAC7-GFP), in which the CsNAC7 coding DNA sequence (CDS) was fused to GFP under the control of the 35S promoter. Then, the pGreen-C18-CsNAC7-GFP recombinant vector was transformed into A. tumefaciens strain GV3101. Transient fusion expression of CsNAC7-GFP in tobacco leaf epidermal cells via GV3101 infection was performed as previously described [59], and GFP fluorescence observations were conducted using confocal laser scanning microscopy (Zeiss).

Analysis of the CsNAC7 gene expression pattern in tea plant via qRT-PCR
To determine the expression pattern of CsNAC7 and its relationship with the caffeine synthase gene yhNMT1, the spatiotemporal expression levels of CsNAC7 and yhNMT1 in the shoots at different leaf positions and in different seasons were determined by qRT-PCR, and the GAPDH gene was selected as an internal reference. The specific primers used for amplification are shown in Supplemental Table S1, among which qGADPH-F and qGADPH-R were used for amplification of the GAPDH gene; additionally qCsNAC7-F and qCsNAC7-R were used for amplification of CsNAC7, and qNMT1-F and qNMT1-R were used for amplification of the yhNMT1 gene. The buds, the first leaf and the second leaf of Camellia sinensis Yinghong 9 shoots were selected in different seasons (spring, summer and autumn), and total RNA was extracted and used to synthesize complementary DNA (cDNA). qRT-PCR analysis was conducted by using the ABI StepOne Plus system (Applied Biosystems, CA, USA) as previously reported 45 . Three biological replicates were included for every experiment.

Genetic transformation of CsNAC7 in tea callus using gene overexpression and silencing technology
The CDS of CsNAC7 was amplified from the recombinant plasmid using the primers 35SN-F and 35SN-R (Supplemental Table S1), and the purified PCR products were cloned and ligated into the pCAMBIA1031-35SN vector, which was pretreated with BamH I and Sal I using a ClonExpress II One Step Cloning Kit (Vazyme Biotech Co., Ltd.). The overexpression recombinant vector was designated as pCAMBIA1031-35SN-CsNAC7, among which the CsNAC7 gene under the control of the 35S promoter was inserted into the BamH I and Sal I sites. In addition, amplification of the R1 and R2 specific fragments of CsNAC7 was conducted by using a combination of primers (Supplemental Table S1), and the primary recombinant plasmid NAC7-R1 was obtained by inserting R1 into the pYLRNAi-35S vector at the sites digested by Bam HI and Hind III (Takara). Then, the recombinant vector RNAi-CsNAC7 was inserted into the NAC7-R1 plasmid at the site digested by Pst I and Mlu I.
The recombinant vectors pCAMBIA1031-35SN-CsNAC7 and RNAi-CsNAC7 were transformed into A. tumefaciens strain EHA105. Colonies identified by PCR were selected and inoculated into Luria-Bertani (LB) liquid media that included 50 mg.L −1 kanamycin (Kana) and 20 mg.L −1 rifampin (Rif) antibiotics and incubated at 28 • C for 12 h. The culture media were centrifuged at 5000 × g for 2 min to collect Agrobacterium and prepared as an infection solution by resuspension of the MS media to an OD600 value of 0.6. The induced tea calli were cut into pieces approximately 0.5 cm × 0.5 cm in size and immersed in Agrobacterium infection solution plus 100 μmoL.L −1 acetosyringone (AS) for 20 min. The infected calli were dried on sterile filter paper and grown on MS solid media plus 100 μmoL.L −1 AS for 3 days in the dark. Afterward, the calli were washed three times with 400 mg.L −1 carbenicillin solution, washed three times with sterile water, and then inoculated on MS solid media plus carbenicillin (200 mg.L −1 ) and hygromycin (35 mg.L −1 ); the media was replaced every 3 weeks. Approximately 60 days later, newly reproduced calli resistant to hygromycin were obtained. The newly reproduced calli were then subjected to PCR and sequencing analysis to identify transgenic calli.

Nucleic acid isolation from Calli and gene expression analysis via qRT-PCR in transgenic tea Calli
Genomic DNA and total RNA were isolated from newly reproduced calli resistant to hygromycin and nontransgenic calli (as CKs) using an All-in-One DNA/RNA Mini Prep Kit (Sangon Biotech). The screening marker gene HPT, which is resistant to hygromycin B, was amplified by the specific primers Hpt-F and Hpt-R, with the isolated genomic DNA used as template, and then sequencing of the PCR products was conducted to identify the resistant calli, which were transgenic. cDNA was prepared from isolated RNA of CsNAC7, yhNMT1 and GAPDH and subjected to qRT-PCR analysis as described in Section 2.7. Nontransgenic calli were used as negative controls, and the relative gene expression was set as 1. Each experiment was repeated 3 times with biological replicates.

Caffeine content determination in tea Calli
Tea calli were cut into small granules and dried at 50 • C to a constant weight. Callus samples (0.5000 g) were accurately weighed, and 20 mL of boiling water was added for 40 min. The extract was filtered immediately and allowed to cool naturally, after which it was added to ddH 2 O (25 mL total volume). The content was determined by an Agilent 1200 HPLC series instrument. Isolation was accomplished on a Poroshell 120 Bonus-RP column (4.6 mm × 50 mm, 2.7 μm) maintained at 30 • C. The mobile phase consisted of (A) 100% acetonitrile and (B) water plus 0.05% trifluoroacetic acid. The elution conditions were as follows: 0 ∼ 8 min, A from 0% to 9%, B from 100% to 91%; 8 ∼ 17 min, A from 9% to 17%, B from 91% to 83%; and 17 ∼ 26 min, A from 17% to 28%, B from 83% to 72%. The flow rate was 0.8 mL·min -1, and the injection volume was 5 μL. The peak of caffeine was verified according to the retention time of the standard. Figure S1. Caffeine content determination in transgenic calli and nontransgenic calli by HPLC Figure S2. Transgene identification in resistant calli by PCR amplification of the HPT gene Table S1. Primers used in this study Table S2. Accession numbers of genes selected from the NCBI website (https://www.ncbi.nlm.nih.gov) for phylogenetic analysis

Supplementary data
Supplementary data is available at Horticulture Research online.

Statistical Analysis
All the results are presented as the means ± Standard Deviations of three independent biological experiments conducted three times. SPSS software was used for t-tests and one-way analysis, and the difference was significant at least when p < 0.05.