Metabolic Characterization of the Anthocyanidin Reductase Pathway Involved in the Biosynthesis of Flavan-3-ols in Elite Shuchazao Tea (Camellia sinensis) Cultivar in the Field

Anthocyanidin reductase (ANR) is a key enzyme in the ANR biosynthetic pathway of flavan-3-ols and proanthocyanidins (PAs) in plants. Herein, we report characterization of the ANR pathway of flavan-3-ols in Shuchazao tea (Camellia sinesis), which is an elite and widely grown cultivar in China and is rich in flavan-3-ols providing with high nutritional value to human health. In our study, metabolic profiling was preformed to identify two conjugates and four aglycones of flavan-3-ols: (−)-epigallocatechin-gallate [(−)-EGCG], (−)-epicatechin-gallate [(−)-ECG], (−)-epigallocatechin [(−)-EGC], (−)-epicatechin [(−)-EC], (+)-catechin [(+)-Ca], and (+)-gallocatechin [(+)-GC], of which (−)-EGCG, (−)-ECG, (−)-EGC, and (−)-EC accounted for 70–85% of total flavan-3-ols in different tissues. Crude ANR enzyme was extracted from young leaves. Enzymatic assays showed that crude ANR extracts catalyzed cyanidin and delphinidin to (−)-EC and (−)-Ca and (−)-EGC and (−)-GC, respectively, in which (−)-EC and (−)-EGC were major products. Moreover, two ANR cDNAs were cloned from leaves, namely CssANRa and CssANRb. His-Tag fused recombinant CssANRa and CssANRb converted cyanidin and delphinidin to (−)-EC and (−)-Ca and (−)-EGC and (−)-GC, respectively. In addition, (+)-EC was observed from the catalysis of recombinant CssANRa and CssANRb. Further overexpression of the two genes in tobacco led to the formation of PAs in flowers and the reduction of anthocyanins. Taken together, these data indicate that the majority of leaf flavan-3-ols in Shuchazao’s leaves were produced from the ANR pathway.


Introduction
It is estimated that more than 2 billion people drink tea every day. More than 50 countries cultivates tea [Camellia sinensis (L.) O. Kuntze] plants. China has cultivated tea for more than 2000 years. During the long cultivation history, numerous elite tea varieties have been cultivated for health benefits. To date, approximately 246 cultivars, such as Longjing and Shuchaozao, are grown for materials to produce tea products. Enzyme analysis showed that crude protein of fresh leaves converted cyanidin and delphinidin to (−)-EC and (−)-EGC as major products and (−)-Ca and (−)-GC as minor products, respectively. Two cDNAs, namely CssANRa and CssANRb, were cloned from young leaf tissues and expressed in E. coli to induce recombinant proteins. Enzymatic analysis demonstrated that CssANRa catalyzed anthocyanidins to 2S,3R-2,3-trans, 2R,3R-2,3-cis-flavan-3-ols, and 2S,3S-2,3-cis-flavan-3-ols such as (−)-Ca, (−)-EC and (+)-EC. CssANRb converted anthocyanidins to 2S,3R-2,3-trans and 2S,3S-2,3-cis-flavan-3-ols such as (−)-Ca and (+)-EC. Transgenic analysis further demonstrated that the overexpression of either of the two genes led to the formation of PAs in transgenic tissues. These data are informative to regulate flavan-3-ol biosynthesis in this elite tea for producing high quality tea products. Biosynthetic pathways of flavan-3-ols and proanthocyanidins starting from leucoanthocyanidins in tea plant.

Flavan-3-ol Profile in New Leaves
In tea industry, buds and leaves at the top four positions, which are developed on new branches of Shuchazao shrub (Figure 2a) in the early spring, form primary materials to produce different quality tea products. Accordingly, new leaves at these positions were collected to characterize flavan-3-ol profiles.  (h) contents of six metabolites in leaves and buds (on each colored line, points labeled with different low case letters mean significant difference, p-value less than 0.05, while labeled with the same ones mean insignificance, p-Value higher than 0.05. p-values were calculated with Student's t-test).
Among the five positional samples, the buds contained the highest total contents of flavan-3-ols, approximately 9.6% (g/g, DW). The 4th position leaves contained the lowest total content, approximately 5.7% (g/g, DW). Among six metabolites, except for (−)-EGC and (−)-EC, four metabolites showed an apparent decrease trend from the 2nd leaf to the 4th leaf. This result provides evidence indicating that different quality of tea products is associated with these sample positions.

ANR Activity in New Leaf Samples
Crude enzyme analysis has been reported being an effective method to demonstrate the ANR pathway in plants [13]. To understand whether the formation of the six flavan-3-ol molecules is associated with an ANR activity, we pooled new leaves from the five positions ( Figure 2a  To further confirm the chirality of enzymatic products separated by reverse phase column, the major product derived from the incubation of crude ANR extract and cyanidin was isolated on TLC plates according to our previous reports [9]. Based on the (−)-EC standard, analysis of HPLC coupled with a normal phase-chiral column identified that this metabolite was (−)-EC (Figure 3f,h). However, (+)-EC was not found from enzyme reaction products. This result supported that the 2R,3R-2,3-cis stereo types, such as (−)-EC and (−)-EGC, accounted for most of flavan-3-ol metabolites synthesized in leaves described above.

Field-Grown Shuchazao Leaves Express Two ANR Homologs
We previously submitted GenBank two ANR sequences (GenBank: KY615701.1 and GenBank: KY615702.1), which have been curated by NCBI. Based on these two sequences, we isolated two ANR cDNA sequences from new leaves of Shuchazao. In addition to ours reported here, two ANR homologs, namely CsANR1 and CsANR2, were reported from the blister blight-resistant tea cultivar TRI2043 in Sri Lanka [17]. To distinguish from these two terms, we named ours as CssANRa and CssANRb, in which Css resulted from abbreviation of Camelina sinensis cv. Shuchazao.
To date, in addition to two ANR cDNAs from TRI2043 as well as CssANRa and CssANRb reported here, four additional ANR cDNA sequences from other tea cultivars are currently curated at the National Center for Biotechnology Information (NCBI). Their GenBank accession numbers are KY615701.1, KY615702.1, HM003282.1, GU992402.1, KF879515.1, and AY169404.1. In addition, the corresponding GenBank accession numbers for amino acid sequences of these four cDNAs are ASU87432.1, ASU87433.1, ADZ58166.1, AAO13092.1, ADZ58168.1, and AHJ11240.1. Amino acid sequences deduced from eight tea ANR cDNAs including ours were aligned to compare their sequence identity and difference. In addition, a grape (Vitis vinifera) ANR homolog, VvANR (GenBank: CAD91911), was used as reference for alignment. The sequence length of VvANR includes 337 amino acids. The aligning result revealed a feature of two groups of tea ANR sequences (Figure 4a), each of which was composed of four sequences. It is interesting that four members in the first group are composed of 336 amino acids, 10 shorter in the N-terminus than the four members in the second group consisting of 346 ones. In addition, the two groups have 51 additional different amino acid residues, 47 of which are completely conserved in the first group, while 48 of which are conserved in the second group, respectively ( Figure 4a). In addition to these interesting differences, a few of scattered amino acid residue differences were observed in eight sequences ( Figure 4a). Sequences of the G-rich NADPH and NADH binding domain and three catalytic amino acid residues (Ser, Tyr, and Lys) are conserved in all homologs ( Figure 4a).
To understand the phylogenetic relationship, amino acid sequences of four tea ANR homologs (CssANRa, CsANR1, CssANRb, and CsANR2) and 12 additional other plant ANR homologs were used to build a tree. The resulting unrooted tree showed that four tea ANR homologs, VvANR, and DkANR were grouped in the same clade ( Figure 4b).
Based on the reported crystal structure (2rh8A) of VvANR [19], potential three-dimensional structures for CssANRa and CssANRb were predicted using the Swiss Model software (http: //swissmodel.expasy.org/). The resulting models characterized structural similarity between VvANR and CssANRa and CssANRb (Figure 4c  Those different amino acids are highlighted using a small red rectangle. The yellow frame shows G-rich NADPH and NADH binding domain; (b) a phylogenetic tree established using 16 ANR homolog amino acid sequences; (c,d) Three dimensional structure prediction for CssANRa (c) and CssANRb (d) using grape ANR template (2rh8A), gray: 2rh8A, red: CssANRa and yellow: CssANRb.

Expression Profiles of CssANRa and CssANRb
Quantitative RT-PCR was carried out to characterize expression profiles of CssANRa and CssANRb in seven tea varieties, five types of tissues, three types of pigmented leaves, and two types of treatments. The resulting data showed that CssANRa and CssANRb had different expression levels among seven varieties. Their expression levels were high in Shuchazao, Pingyangtezao, and Jinfenghuang, but low in Huangdan and Dayewulong (Figure 5a). The expression pattern of CssANRa and CssANRb in five tissues of Suchazao from the highest to the lowest levels were in the order of the 2nd, 1st leaves, old stems, bud, and young stems (Figure 5b). In three types of pigmented leaves, their expression levels were the highest in purple leaves, followed by green, and then yellow leaves (Figure 5c). The expression levels of CssANRa was reduced by both NaCl and sucrose treatments, while the expression levels of CssANRb was decreased by NaCl treatment but increased by sucrose treatment (Figure 5d).

Catalytic Activity of Recombinant CssANRa and CssANRb
The ORFs of CssANRa and CssANRb were cloned to a pET31a + vector to obtain pET31a + -CssANRa and -CssANRb plasmids to induce recombinant enzymes. After induction with IPTG, E. coli harboring the pET31a + -CssANRa plasmid expressed a soluble recombinant protein, which was shown by SDS-PAGE analysis ( Figure 6a). However, no recombinant protein was induced from E. coli containing the pET31a + -CssANRb plasmid. To induce a recombinant CssANRb, its ORF was cloned to the T7 SUMO vector to obtain a recombinant plasmid, T7 SUMO-CssANRb. After induction with IPTG, E. coli harboring this plasmid expressed a recombinant protein (Figure 6a). SDS-PAGE analysis showed that total soluble recombinant protein from each induction contained excessive other proteins. Accordingly, the MagnetHis Protein Purification System (Promega, Madison, WI, USA) was used to reduce those excessive proteins from total crude proteins. SDS-PAGE analysis showed that this treatment obviously reduced certain types of excessive soluble proteins ( Figure 6a). Therefore, recombinant CssANRa and CssANRb proteins were partially purified for catalytic analysis.  (Figure 6m). In addition, a third metabolite peak was also observed from the CssANRb assay and demonstrated to be (+)-EC (Figure 6l).
In addition, the optimum temperature values of recombinant CssANRa and CssANRb were 40 • C (Supplementary Figure S1a,b). The optimum pH values of recombinant CssANRa and CssANRb were 6.5 and 5.5, respectively (Supplementary Figure S1c,d).

Overexpression of CssANRa and CssANRb in Tobacco Flowers Reduces Anthocyanins and Produces PAs
CssANRa and CssANRb controlled by a 35S promoter were introduced to tobacco plants, respectively. After selection using antibiotics, 19 and 17 positive transgenic plants were obtained for CssANRa and CssANRb transgenes. Transgenic and wild-type plants were grown in a glass house to develop flowers, during the growth period of which no phenotypic difference was observed. Flower buds (stage 1, unopened), slightly opening flowers (stage 2), and fully opening flowers (stage 3) were collected to analyze anthocyanins and PAs. Reduction of red-pigmentation was observed on stage 3 transgenic flowers compared with-type flowers (Figure 7a,b) although effects of the two transgenes on red pigmentation were different. Semi-quantitative RT-PCR analysis performed using fully opened flowers showed that the CssANRa and CssANRb transgenes were expressed in transgenic flowers but not in wild-type flowers (Figure 7c To further characterize PA formation in transgenic flowers, both anthocyanins and PAs were measured during flower development from unopened buds to fully opened flowers. The resulting data showed that the absorbance of anthocyanins were significantly lower in transgenic flowers than in wild-type flowers at stages 2 and 3 ( Figure 7g). The reaction of DMACA and PAs led to blue color. The bluish color resulted from PAs in transgenic flowers was much deeper than that from wild-type flowers. Further measurement at 640 nm showed that the absorbent values were higher from transgenic flowers than wild-type flowers at stages 2 and 3. These results indicated that the increase of PAs was a tradeoff consequence of the decrease of anthocyanins. In (e,f), bars labeled with different low case letters for anthocyanin or PAs mean significant difference (p-value less than 0.05), while with the same ones mean insignificance (p-value higher than 0.05). In (g,h), points labeled with different low case letters mean significant difference (p-value less than 0.05), while with the same ones mean insignificance (p-value higher than 0.05). p-Values were calculated with Student's t-test.

Discussion
Shuchaozao is an elite commercial tea cultivar. The ANR pathway in Shuchaozao leaves for tea products lacks an appropriate characterization, although, to date, this pathway in numerous other plant species has been solidly substantiated in genetics, biochemistry, metabolic engineering, and evolution [12,13,13,19,20], and two ANR homologs have also been reported in a blister-resistance tea cultivar. In this study, we focused on characterization of the ANR pathway in field-grown plants' leaves that are used for commercial tea products. Metabolic profiling revealed that young leaves  (Figure 3a-d). Chiral analysis further showed that crude leaf ANR extracts produced (−)-EC but not (+)-EC (Figure 3f-h). Therefore, these data reveal that the ANR pathway in leaves of Shuchaozao is actively responsible for the high total contents of flavan-3-ols. Furthermore, these data indicate that the ANR pathway is closely associated with flavan-3-ols based high quality products.
To further understand the role of the ANR pathway in this cultivar, gene specific primers were designed to clone full length of cDNAs. Two ANR homologs, CssANRa and CssANRb, were isolated from young leaves. Results from qRT-PCR analysis showed that the expression levels of CssANRa were more apparently variable than those of CssANRb in different tea varieties and different leaf development stages (Figure 5a,b). By contrast, qRT-PCR analysis characterized that the expression levels of CssANRb was more obviously variable than those of CssANRa in differently pigmented leaves and in those leaves under NaCl and sucrose treatments (Figure 5c,d). In addition, it was interesting that heterogeneous expression in E. coli revealed that CssANRa and CssANRb exhibited different plasmid preferences to produce recombinant proteins. The pET31a + plasmid was appropriate to induce a recombinant CssANRa but not CssANRb. Instead, the recombinant CssANRb was induced when the pET31 SUMO plasmid was used (Figure 6a). In addition, recombinant CssANRa and CssANRb showed different pH optima. By contrast, the temperature optima of the two recombinant enzymes were the same. As those results obtained from the CsANR1 and CsANR2 overexpression reported by Pang et al. [17], the overexpression of CssANRa and CssANRb also led to the formation of PAs in anthocyanin-producing flowers and reduction of anthocyanins during flower development (Figure 7). These transgenic data provide evidence that CssANRa and CssANRb are involved in the biosynthesis of PAs.
In our experiments, we observed an interesting different result between leaf ANR and recombinant ANR assays. (+)-EC formed from the recombinant ANR assay (Figure 6k,m) was not observed from the leaf ANR assay (Figure 3f,h). The biochemical mechanism behind this observation is unclear. One possibility might be associated with the difference of native leaf ANR extract and recombinant ANR. Native ANRs were extracted from Lotus corniculatus, Desmodium uncinatum, Hordeum vulgare, Vitis vinifera, Vitis bellula, Parthenocissus heterophylla, Cerasus serrulata, and Dryopteris pycnopteroides [13]. Enzymatic analysis demonstrated that native ANRs from all of these plant species catalyzed cyanidin to (−)-EC as a major product and (−)-Ca as a minor product. Another interesting possibility may be associated with recombinant vectors. Xie et al. reported that both recombinant AtANR and MtANR fused with pMAL converted cyanidin to (−)-EC and (−)-Ca but no (+)-EC [21]. Although Xie et al. [11] did not use a chiral column in their study, they identified chirality via a dichroism spectrum analysis, which resolved the absolute configurations of flavan-3-ols. It was interesting that (+)-EC was observed in reactions catalyzed by His-tag fused ANR. Gargouri et al. demonstrated that a recombinant V. vinifera ANR fused with a His-Tag performed a dual catalytic activity, reduction and epimerization [19]. The recombinant VvANR converted cyanidin to (−)-EC, then to (−)-Ca and (+)-EC. Pang et al. [17] also fused CsANR1 and CsANR2 with a His-Tag, which resulted in the formation of (+)-EC. In our report herein, CssANRa and CssANRb were also fused with His-Tag. These interesting differences indicate that the final understanding of absolute configurations of flavan-3-ols needs more studies of both native plant ANRs and recombinant ANRs.
To enhance general understanding of ANR expression in tea tissues, we further mined ANR transcripts (without separate two paralogs) in tissues of two other elite green tea varieties, "Pingyangtezao (PYTZ)" and "Ruixue". Five transcriptomes were sequenced for buds, the 1st leaf, the 2nd leaf, young stem, and old stem of PYTZ, respectively. We obtained transcripts of ANR from all five transcriptomes. In addition, we annotated 1420 potential transcription factor (TF) cDNAs. The transcript of each TF was also obtained from all five transcriptomes. The transcripts of ANR and each TF were used for association analysis of expression profile trend in five tissues. The resulting data revealed the coupled expression of 885 TFs cDNAs and ANR in each tissue. A further expression trend analysis using 885 TF genes and ANR resulted in 20 expression profiles in buds, the 1st leaf, the 2nd leaf, young stem, and old stem (Figure 8). The transcript trend of ANR was the same as those of 23 TF genes, which was characterized in profile 18 (Figure 8k). Further annotation analysis characterized the 23 TFs in 13 families (Supplementary Figure S2), which included four bHLH and 4 MYB members. These data indicate that the expression of ANR is positively associated with bHLH and 4 MYB members. Meanwhile, we analyzed gene expression association between ANR and TF cDNAs in flowers. Three transcriptomes were sequenced for three developmental stages of flowers of Ruixue. We also obtained transcripts of ANR from all three transcriptomes. Moreover, we annotated 1325 TF genes from the three flower transcriptomes and obtained their transcripts in each transcriptome. Based on transcripts of these TF genes and ANR, we performed expression association analysis to obtain 1100 TF genes, the expressions of which were coupled with that of ANR. Based on transcript of each TF gene and ANR, eight expression trend profiles were established ( Figure 9). The expression trend of ANR was the same as those of 205 TF genes, which was characterized in profile 0 ( Figure 9c). The 205 TF genes were annotated to 29 families, which included 74 MYB (MYB family) and 12 bHLH (bHLH family) members (Supplementary Figure S3). These data indicate that the transcript trend of ANR is closely associated with those of MYB and bHLH gene members. All of these data will be useful for us to further study the regulation of the ANR pathway in Suchazao and other green tea varieties.

Field Growth of Shuchazao and Other Varieties for Sampling
Shuchazao (Camellia sinensis (L.) O. Kuntze) plants are grown in an experimental tea garden (research station) at Anhui Agricultural University (latitude 31.86 • N, longitude 117.27 • E, altitude 20 m above mean sea level) and in the farming field in Hefei, Anhui, China. Plants grown in the tea garden is used for numerous research purposes, while plants grown in the field are used to harvest leaves to produce tea products for sale. Plants were grown 10 years when we performed all experiments reported here. Buds, the 1st leaf, the 2nd leaf, the 3rd leaf, and the 4th leaf (Figure 2a) of new shoots were collected from plants at these two locations. When samples were harvested, they were immediately frozen in liquid nitrogen, transported to laboratory, and then stored in a −80 • C freezer until use for gene cloning and metabolic analysis. In addition, leaves were collected from one year old Shuchazao seedlings and then separated into three groups, which were immediately treated 4 h with 200 mM NaCl and 200 mM sucrose (dissolved in double deionized water) and dd-water as control. Leaf samples treated were immediately frozen in liquid nitrogen and then stored in a −80 • C freezer until RNA isolation and qRT-PCR analysis described below.
In addition, 15 other tea varieties with 10 years old are grown in the same garden for numerous research purposes. Seven were selected to comparatively understand ANR gene expression. These are C. sinensis cv. Pingyangtezao (PYTZ), C. sinensis cv. Jinfenghuang (JFH), C. sinensis cv. Dayewulong (DY), C. sinensis cv. Huangdan (HD), C. sinensis cv. Zhengdayin (ZDY), C. sinensis cv. Shuchazao (SCZ), C. sinensis cv. Queshe (QS). The 2nd leaf (Figure 2a) was sampled from new shoots of these varieties. Eight other varieties were used to compare ANR expression in three different types of pigmented leaves, green, yellow, and purple. The green group included C. sinensis cv. LongJing, and C. sinensis cv. Tieguanyin. For sample collection, Shuchazao was added to form three varieties in this green group. The yellow group included C. sinensis cv. Huangjinya, C. sinensis cv. Zhonghuangyihao, and C. sinensis cv. Zhonghuangerhao. The purple group consisted of C. sinensis cv. Zijuan, C. sinensis cv. Ziyan, and C. sinensis cv. Sunrouge). The 2nd leaf of new shoots (Figure 2a) was also sampled from each variety in each group for qRT-PCR analysis.

Extraction and Analysis of Flavan-3-ols in Leaves from Field-Grown Shuchazao Plants
Fresh samples, including buds, the 1st leaf, the 2nd leaf, the 3rd leaf, and the 4th leaf (Figure 2a), were grounded into fine powder in liquid nitrogen. A powdered sample (100 mg) was suspended in 1 mL extraction solution (methanol: ddH 2 O, 80:20) in a 1.5 mL tube and sonicated for 10 min at room temperature. After centrifugation at 4000× g for 15 min, the resulting upper clear phase was pipetted to a new 1.5 mL tube. The remaining pellet was extracted a second time. Supernatants were combined and the final volume was adjusted to 2 mL. All extracts were filtered through a 0.22 µm membrane for HPLC analysis.

Extraction of Crude Enzyme from Field-Grown Tea Leaves and Enzyme Assay
Frozen tea leaves (approximately 25 g, a mixture of the 1st-4th leaves) were ground into fine powder in liquid nitrogen. The powdered sample was completely suspended in 200 mL phosphate buffer (0.1 M, pH 7.4) supplemented with 25 g polyvinyl polypyrrolidone (PVPP) and 5 mM β-mercaptoethanol in a 250 mL centrifugation tube. The mixture was placed on ice for 1 h and then was centrifuged at 12,000× g for 15 min at 4 • C. The resulting supernatant (80 mL) was pipetted into a new tube, followed by addition of ammonia sulfate (from 0% to 40%) to precipitate proteins. The mixture was centrifuged at 12,000× g for 15 min at 4 • C and the resulting supernatant was disposed to a waste container. The remained residue was dissolved in 5 mL phosphate-buffer (0.1 M, pH 7.0) and centrifuged at 12,000× g for 10 min at 4 • C. The resulting supernatant containing crude proteins was used for ANR activity assay described below.
Enzyme assay for crude enzyme extract was carried out by following a method reported by Xie et al. [21] and Zhang et al. [22] with slight modifications. In brief, enzyme assay was carried out in 15 mL capped polypropylene tubes. The reaction volume was 3.5 mL consisting of 100 mM phosphate buffer (pH 6.5), 2 mM NADPH, 0.5 mM substrate (cyanidin or delphinidin), and 300 µg crude enzyme extract. Boiled crude enzyme extract was used to replace crude enzyme in the reaction as one control. In addition, reactions without adding substrates were carried out as the other control. The enzymatic reactions were incubated for 40 min at 40 • C. All reactions were stopped by addition of 1 mL ethyl acetate (EA) to reaction tubes and vortexing, followed by centrifugation at 5000× g for 5 min. The resulting ethyl acetate supernatant phase was pipetted into a new tube. This step was repeated two times. Three times of EA extractions was pooled together to obtain 3 mL extract. EA was dried off via a rotary evaporation at room temperature. The remained residues were completely suspended in 100 µL HPLC-grade methanol, followed by centrifugation at 10,000× g for 2 min. The methanol phase was transferred to a new tube for HPLC analysis, which was as described above in the analysis of flavan-3-ols in leaves from field-grown Shuchazao plants. (−)-EC, (−)-Ca, (−)-GC, and (−)-EGC standards were used to identify metabolites from enzymatic reactions.

Isolation of Shuchazao ANR cDNA (CssANR) and Quantitative Reverse Transcription-Poly Chain Reaction (qRT-PCR) Analysis
Total RNA was isolated from mixed samples of buds and leaves (Figure 2a) using the RNAiso-mate for Plant Tissue Kit according to the manufacturer's protocol (Takara, Tokyo, Japan). The first strand cDNA was synthesized using PrimeScript ® RT reagent Kit according to the manufacturer's protocol (Takara, Tokyo, Japan). According to two cDNA sequences, KY615701.1 (CssANRa) and KY615702.1 (CssANRb), which we previously submitted to GenBank, the first pair of primers consisting of forward 5'-ATGGAAGCCCAACCGACA-3' and reverse 5'-TCAATTCTTCAAAATCCC3' was designed to amply for CssANRa. The second pair of primers consisting of forward 5'-ATGGCAATGGCAATGGCAACAAC-3' and reverse 5'-TCAGTTCTGCAAAAGCCCCTTAG-3' was designed to amplify CssANRb. The open reading frame (ORF) of CssANRa and CssANRb were amplified using a thermal gradient program that was composed of 5 min at 94 • C, 30 cycles of 30 s at 94 • C, 30 s at 62 • C, and 1 min at 72 • C, followed by a 10-min extension at 72 • C. The products of PCR were gel-purified using Takara MiniBEST Agarose Gel Extraction Kit (Takara) and ligated into the pMD18-T vector (Takara, Tokyo, Japan). The ligated products were transformed into E. coli strain DH5α competent cells using electroporation. Transformed cells were streaked on LB medium supplemented with 50 mg/L ampicillin. Positive colonies were selected to isolate recombinant plasmids. The resulting plasmids were termed pMD18-T-CssANRa and pMD18-T-CssANRb for sequencing at BGI (http://www.bgitechsolutions.com/).
It was used for normalization. The primer pair used for this gene consisted of forward 5'-TTGGCATCGTTGAGGGTCT-3' and reverse 5'-CAGTGGGAACACGGAAAGC-3'. Values were normalized against the expression level of GAPDH [23]. Relative expression values were calculated with the 2 − Ct method. Values of gene expression were means of four replicates.

Analysis of ANR Sequences
Available C. sinensis ANR sequences curated at NCBI was identified and their amino acid sequences were aligned using ClustalX. In addition, four C. sinensis and 12 other species' ANR amino acid sequences were used to construct a phylogenetic tree via Molecular Evolutionary Genetics Analysis version 5.0 (MEGA5.0; MegaSoftware, Tempe, AZ, USA) [24] using a neighbor-joining statistical method (with bootstrapping 1000 replicates) [25].

Induction of Recombinant CssANRa and CssANRb in E. coli
The pMD18-T-CssANRa and -CssANRb vectors containing the ORFs of CssANRa and CssANRb were digested using XbaI and SnaBI restriction enzymes. The resulting ORFs of CssANRa and CssANRb were separated on agarose gel using electrophoresis and then purified using the MiniBEST Plasmid Purification Kit (Takara) according to the manufacturer's protocol. In addition, the pET31a + vector was digested using the same restriction enzymes. Then, the ORFs of CssANRa and CssANRb were separately ligated to pET31a + overnight using T4 DNA Ligase (Takara) under 15 • C. The ligation products were introduced to BL21 (DE3) competent cells using a common electroporation method.
Transformed BL21 cells were streaked on agar-solidified medium containing 50 mg/L ampicillin. Antibiotic-resistant colonies were selected to confirm the presence of ORFs in the plasmids and positive colonies were used to induce protein expression. The new plasmids were named as pET31a + -CssANRa and -CssANRb, respectively. Furthermore, the pET31a + vector was also introduced into BL21 cells as protein induction control. In addition to pET31a + , the ORF of CssANRb was ligated to the T7 SUMO vector (Invitrogen, K300-01) using the same ligation method. The ligation products were introduced to One Shot Mach1™-T1R Competent Cells (Invitrogen, K300-01) to select positive colonies on LB medium supplemented with 50 mg/L kanamycin. Kanamycin resistant colonies were selected to confirm the presence of CssANRb. The resulting new plasmid was named as T7 SUMO-CssANRb. In addition, T7 SUMO was introduced to the same competent cells as vector control for induction of recombinant protein expression.
Five colonies each containing one of pET31a + -CssANRa, pET31a + -CssANRb, T7 SUMO-CssANRb, pET31a + , and T7 SUMO plasmids were used to induce protein expression. Each colony was inoculated into 10 mL LB broth containing 50 mg/L antibiotics (ampicillin for pET31a + -CssANRa and pET31a + vector, and kanamycin for T7 SUMO-CssANRb and T7 SUMO vector) in an E-flask. All flasks were placed on a shake with a speed of 250 rpm at 37 • C. When cells were grown to a concentration indicated by an absorbent value about 0.6 that was recorded at 600 nm on an UV spectrophotometer, each flask was added isopropyl β-D-1-thiogalactopyranoside (IPTG) to a final concentration of 0.4 mM. Each flask was continuously shaken 3 h at 37 • C. Cells were harvested by centrifugation at 4000× g for 15 min at 4 • C. The remained cell pellet was suspended in 10 mL 100 mM phosphate buffer (pH 6.5) by vortexing. A T4 lyase was added to the suspension mixture with a final concentration of 100 µM. Then, the mixture was sonicated for 1 min on ice, followed by 20 min of centrifugation at 10,000× g at 4 • C. The supernatant was transferred to a new tube. Crude protein extracts were further treated using the MagnetHis Protein Purification System (Promega, Madison, WI, USA) to remove excessive proteins of E. coli according to the manufacturer's protocol. Partially purified protein dissolved in 100 mM phosphate buffer (pH 6.5) was measured to estimate concentrations with the Bio-Rad Protein Assay system (Bio-RAD, California, CA, USA) and were stored in −80 • C until enzyme assay described below.

Enzyme Assays
Partially purified recombinant CssANRa and CssANRb (Figure 6a) were assayed using the method reported by Xie et al. [21] to optimize temperature and pH values. In all assays, experiments were carried out in 1 mL volume. Three replicates were completed for each reaction. The incubation time was 40 min. Boiled enzyme was used as control. In addition, reactions without addition of substrates were used as another control.
For optimization of temperature, enzyme reactions were carried out in 2 mL volume including 100 mM phosphate buffer (pH 6.5), 2 mM NADPH, 0.5 mM cyanidin, and 150 µg recombinant proteins. Methods for stopping all reactions, extracting metabolites, and HPLC analysis were as described for crude enzyme assay.

Chiral Analysis of Products from Enzyme Reactions
HPLC using a normal-phase chiral column (catalog No. 80325; Chiral Technologies; 5 µm, 4.6 × 250 mm; Tokyo, Japan) was performed to determine the chirality of enzyme reaction products from cyanidin substrate according to a method reported by Pang et al. [17]. Elution buffer was composed of solvent A (hexane:acetic acid, 99.5:0.5, v/v,) and solvent B (ethanol:acetic acid, 99.5:0.5, v/v). Isomers were separated using a gradient program, which was composed of 0-20 min with 20% solvent B, 20-23 min with 20-50% B, 23-38 min with 50% B, and 38-40 min with 50-20% B. The flow rate was set at 1 mL/min. UV spectrum was recorded at 280 nm to characterize isomers. (−)-EC and (−)-Ca standards were used as control.

Construction of Binary Vectors, Tobacco Transformation, and Flower Sample Harvest
The Gateway ® Cloning System (Invitrogen, New York, NY, USA) was used to construct binary vectors according to a method reported by Lei et al. [26]. In brief, pairs of primers containing attB Gateway sequence were designed to amplify CssANRa and CssANRb to develop binary vectors. The primer pairs were CssANRa-attB (attB1 and attB2) and CssANRb-attB (attB1 and attB2) ( Table S1). The pMD18-T-CssANRa and -CssANRb plasmids were used as template for PCR. The thermal program used for PCR was composed of 95 • C × 2 min, 35 cycles of 94 • C × 1 min, 56 • C × 1 min, and 72 • C × 1.5 min, and 72 • C for 10 min extension. The resulting PCR products were purified and then cloned to the entry vector pDONR207 with Gateway ® BP Clonase ® Enzyme mix according to the manufacturer's instructions (Invitrogen). The resulting ligation products were introduced to competent DH5α cells, which were selected on agar-solidified LB plates containing gentamycin. Antibiotic-resistant colonies were selected and cultured to isolate recombinant entry vector. The positive entry vectors were incubated with the Gateway plant transformation destination vector pCB2004 using Gateway ® LR ClonaseTM enzyme (Invitrogen) according to the manufacturer's protocol. The incubation mixtures were then introduced into competent DH5α cells, which were further selected on agar-solified LB medium containing 50 µg/mL kanamycin. Positive colonies were selected to isolate plasmids. Finally, two plasmids, namely pCB2004-CssANRa and -CssANRb, were created, in which two genes were controlled by a 35S promoter. These two vectors and empty pCB2004 (as control) were introduced into competent Agrobacterium tumefaciens strain EHA105 by electroporation, respectively. One positive colony was obtained for each vector. Positive colonies, namely EHA105-pCB2004-CssANRa, EHA105-pCB2004-CssANRb, EHA105-empty pCB2004, were selected on agar-solified LB medium containing 50 µg/mL kanamycin for genetic transformation.
A single colony, each for EHA105-pCB2004-CssANRa, EHA105-pCB2004-CssANRb, and EHA105-empty pCB2004, was inoculated into 20 mL liquid LB medium containing 50 mg/L kanamycin and 50 mg/L spectinomycin in 100 mL E-flask. Three flasks were placed on a rotary shaker at 200 rpm in the dark at 28 • C for overnight until OD value was about 0.6 at 600 nm. Cultured cells were transferred to a 50 mL sterile polypropylene centrifuge tube, followed by centrifugation at 4500× g for 10 min and disposal of supernatant. The remaining pellet was suspended in 20 mL liquid MS medium containing 100 µM acetosyringone (Sigma-Aldrich, St. Louis, MO, USA R40456), followed by centrifugation as above. This step was repeated once and the pellet was suspended in 20 mL liquid MS medium containing 100 µM acetosyringone to obtain activated cells for genetic transformation.
Sterile seedlings of tobacco (Nicotiana tabacum 'G28 ) plants were grown on agar-solidified MS medium. Leaves of seedlings were cut into 1 × 1 cm discs. Infection of leaf discs with activated A. tumefaciens strain EHA105 described above and selection of transgenic plants on regeneration medium containing 25 mg/L phosphinothricin followed a protocol reported previously [27]. Gene specific primer pairs (CssANRa-test and CssANRb-test) were designed to screen transgenic plants (Table S1). RT-PCR as described above for expression profiling analysis was carried out to identify positive transgenic candidates. Positive transgenic vs. wild-type tobacco plants were planted into pot soil and then grown in a glass house, which was provided with natural light and maintained at 22-25 • C. Flower pigmentation was photographed every day. Flowers were collected from different developmental stages and then frozen in liquid nitrogen immediately. Frozen samples were stored at −80 • C until PA and anthocyanin analysis described below. performed the experiments and analyzed the data; P.-Q.W. contributed to graph and data analysis; L.Z., D.-Y.X. and L.-P.G. wrote the paper. All authors approved the version to be published; and agreed to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Conflicts of Interest:
The authors declare no conflict of interest.