Analysis of accumulation patterns and preliminary study on the condensation mechanism of proanthocyanidins in the tea plant [Camellia sinensis]

In the present study, proanthocyanidins were qualitatively and quantitatively identified using hydrolysis and thiolysis assays, NP-HPLC, HPLC-ESI-MS, MALDI-TOF-MS, 1H-NMR, and 13C-NMR techniques in different organs of tea plants. The results showed that in leaves, the tri-hydroxyl, cis- and galloylated flavan-3-ols were the main monomeric catechins units, and (epi)catechin was found to be the major unit of polymeric flavan-3-ols when the degree of polymerization was greater than five. In roots, the PAs were found to be abundant, and epicatechin formed the predominant extension unit of oligomeric and polymeric PAs. In order to understand the mechanism of proanthocyanidins polymerization, auto-condensation of the flavan-3-ols was investigated. The results showed that the same trimers (m/z 865) were detected in the extracts of tea plants and in the non-enzymatic in vitro assay, in weak acid as well as weak alkaline solutions at room temperature, when the substrates used were either procyanidin B2 and monomeric flavan-3-ols (epicatechin or catechin), or only procyanidin B2. This suggested that procyanidin B2 not only released carbocation as electrophilic upper units, but also could be used as nucleophilic lower units directly itself, to form the procyanidin trimer in vitro or in vivo.

In the present study, proanthocyanidins were qualitatively and quantitatively identified using hydrolysis and thiolysis assays, NP-HPLC, HPLC-ESI-MS, MALDI-TOF-MS, 1 H-NMR, and 13 C-NMR techniques in different organs of tea plants. The results showed that in leaves, the tri-hydroxyl, cisand galloylated flavan-3-ols were the main monomeric catechins units, and (epi)catechin was found to be the major unit of polymeric flavan-3-ols when the degree of polymerization was greater than five. In roots, the PAs were found to be abundant, and epicatechin formed the predominant extension unit of oligomeric and polymeric PAs. In order to understand the mechanism of proanthocyanidins polymerization, auto-condensation of the flavan-3-ols was investigated. The results showed that the same trimers (m/z 865) were detected in the extracts of tea plants and in the non-enzymatic in vitro assay, in weak acid as well as weak alkaline solutions at room temperature, when the substrates used were either procyanidin B2 and monomeric flavan-3-ols (epicatechin or catechin), or only procyanidin B2. This suggested that procyanidin B2 not only released carbocation as electrophilic upper units, but also could be used as nucleophilic lower units directly itself, to form the procyanidin trimer in vitro or in vivo.
The condensation mechanism of proanthocyanidins has gained much attention since the 1970s. Haslam et al. hypothesized that the condensation mechanism of procyanidin involved two distinct units, the nucleophilic flavan-3-ols, and the electrophilic C-4 carbocation 17,18 . There is still a lack of understanding of the PA biosynthesis and its regulation, despite the significant progress made in comprehending the biochemistry and molecular genetics 19 , the PA transport forms 20 , and transcription regulatory mechanisms since the 21st century [21][22][23][24][25][26][27][28] . For PA biosynthesis mechanism, the general view perceived is that the flavan-3-ols, (1)-catechin and (2)-epicatechin, are the building blocks of proanthocyanidins; and at first, PA polymerizes into cytoplasmic vesicle-like structures and then the latter fuses with the central vacuole 29 . The biosynthesis of flavan-3-ols, such as (1)-catechin and (2)-epicatechin, first occurs on the cytosolic side of the endoplasmic reticulum (ER) surface, then the (2)epicatechin is catalyzed and transformed into epicatechin 39-Oglucoside by a glycosyltransferase 20 . Epicatechin 39-O-glucoside has been proven to be a substrate for a vacuolar multidrug and toxic compound extrusion (MATE) transporter 30 . However, in the vacuoles, the mechanism involving condensation of (2)-epicatechin or epicatechin 39-O-glucoside into PA oligomer in vivo still remains a mystery 19 . Some biochemical and molecular genetic studies offer theories about the polymerization of flavan-3-ol [(1)-catechin and (2)epicatechin] units to form proanthocyanidin. For example, procyanidin B3 may be obtained from the oxidative reaction of catechin, catalyzed by a tyrosinase 31 ; and the A-type linkages (the C7 hydroxyl to C2) can result from polyphenol oxidase-mediated oxidation, when epicatechin is incubated with banana extracts 32 . Recently, Pourcel et al. cloned a TT10 gene, which expressed a laccase-like enzyme that may be involved in the oxidative polymerization of flavonoids in Arabidopsis seed coat 33 (Fig. 1). From their results, it could be concluded that the laccase-like enzyme is involved only in the polymerization of PA oligomers into polymers, because the tt10-2 mutant contained 4.6-fold more soluble PAs (PA oligomers) than the wildtype control.
There are some reports available on the chemical oxidation of flavan-3-ols. For instance, treatment of procyanidin B1 or B2 with the oxidizing agent, 1,1-diphenyl-2-picrylhydrazyl radical, resulted in the formation of procyanidin A1 or A2 34 . The flavan-3-ols, including (1)-catechin, (2)-epicatechin, and epigallocatechin gallate, could be oxidized in vitro by PPO into theaflavins (TFs) and thearubigins (TRs), which exist only in processed tea, such as black tea, and not in fresh tea plants 35-37 . In the present study,we find that the same procyanidin trimers (m/ z 865) could be detected in the extracts of tea plants as well as in the non-enzymatic in vitro assay, when either procyanidin B2 with epicatechin (EC) or catechin (C), or only procyanidin B2 were used as a substrate in acid and alkaline solutions, respectively. This suggested that procyanidin B2 not only released the carbocation as electrophilic upper units, as expected, but also as nucleophilic lower units, to form  the procyanidin trimer in vitro or in vivo. This phenomenon may contribute to a greater understanding of the PAs biosynthesis and regulatory mechanisms.

Results
Identification of PAs in Different Organs of Tea Plants. RP-LC-ESI-MS Assay. Liquid chromatography (LC) combined with electrospray ionization mass spectrometry (ESI-MS) are particularly useful for the qualitative analysis of compounds 6,38,39 . As presented in Table 1 and Fig. 2, eight monomers and 12 oligomers of PAs existing in the roots, stems, and leaves were identified through LC-TOF-MS and UPLC-QQQ-MS/MS techniques, based on various standards reported in the literature, including t R , lmax, , and major fragment ions in tea or other plants 8,[40][41][42][43] . Among these compounds, eight monomers and 10 PAs oligomers were detected in leaves and stems, with high contents of monomers (including catechin, epicatechin, gallocatechin, epigallocatechin, epicatechin gallate and epigallocatechin gallate) as compared to their oligomers (Table 1, Fig. 2A and 2B). In contrast, only one monomer (epicatechin) and a few oligomers were detected in roots, which suggested that the oligomers were dimers and trimers of flavan-3-ol with di-hydroxyl groups in the B-ring (Table 1, Fig. 2C). Unfortunately, it was not possible to perform accurate qualitative or quantitative analysis of PAs with molecular weights more than 1000, including oligomers (DP 5 4-10) and polymers (DP . 10).
Hydrolysis Assay. PAs were hydrolyzed into colored anthocyanidins in butanol-HCl solution upon heating 44 . The hydrolyzed products were analyzed by RP-HPLC at 530 nm to identify the hydroxyl groups in the B-ring of the extension units of PAs. The data showed three product peaks in leaves and stems ( Fig. S2A and S2B), but only one peak was detected in roots (Fig. S2C). According to the standards, peaks 1, 2, and 3 represented delphinidin, cyanidin, and pelargonidin, respectively. This suggested that the extension units of PAs with mono-, di-, and tri-hydroxyl groups in the B-ring existed simultaneously in leaves and stems; while in roots, only the extension units with di-hydroxyl groups in the B-ring occurred. These results were consistent with the findings of LC -MS.
Thiolysis Assay. DP of PAs was determined by thiolysis. As shown in Table 2, the DP increased from leaves to stems and roots. In addition, the DP in different developmental stages of leaves and roots did not vary substantially. For example, the DP in annual and biennial leaves was determined as 1.03 6 0.02 and 1.02 6 0.03, respectively, and in fibrous and tap roots, the values were found to be 5.85 6 0.05 and 5.41 6 0.11, respectively. However, significant differences in the values of DP were observed in different developmental stages of stems (Table 2). Interestingly, the PAs showed a significantly higher DP in the biennial stem as compared to the annual stem. In summary, the DP of PAs increased gradually from the upper morphological region of the tea plant to the lower part. The average value of DP, obtained through thiolysis assay, explained the molecular weight distribution characteristics of PAs in different organs of a tea plant.
NP-HPLC Assay. Various researchers have used normal-phase-HPLC to quantify the oligomeric and polymeric PAs in foods 1,5,14,45 . For instance, Gu et al. 1,5 in their assay project, applied NP-HPLC-MS/MS to screen for PAs in 88 different kinds of foods, and PAs up to decamers were separated according to the DP values. Polymers with DP . 10 were eluted as a single peak at the end of the chromatogram. In the current study, oligomeric and polymeric PAs in leaves, stems, and roots of tea plants were identified using the same methodology (Fig. 3A, 3B and 3C). The results showed a great variation in the composition of oligomeric and polymeric PAs between leaves and roots, while only monomers were detected in the leaves (Fig. 3A). A series of oligomeric PAs were detected, and a large peak of polymers (DP . 10) was present at the end of the chromatogram (Fig. 3C).
These observations were in agreement with the results of the thiolysis assay. The oligomeric PAs with DP less than ten were separated and qualitatively analyzed by NP-HPLC.
MALDI-TOF-MS Assay. Matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF-MS) is appropriate for the analysis of a mixture of biomacromolecules. PAs in leaves and roots were analyzed by MALDI-TOF-MS using 2,5dihydroxybenzoic acid (DHB) as a matrix in reflection positive ion mode ( Table 3, Fig. 4). As presented in Fig. 4A, a series of peaks with ions corresponding to cesium adducts [M 1 Cs] 1 distributed from m/z 995.1264 to m/z 2728.4940, were reviewed in the mass spectrum    Fig. 4A and 4B); thereby suggesting the presence of a series of oligomers in leaves, with DP ranging from 3 to 9.
Remarkably, only one obvious signal peak was observed when DP was greater than five. The cesium adducts of the peak with different DP increased at a distance of 288 Da, corresponding to the catechin/ epicatechin monomer. The results indicated that the PAs with DP . 5 comprised of epicatechin or catechin monomer units in tea leaves. It was proved that the PAs in leaves mainly comprised of the B-type with C-C bond linkages. The A-type was also detected in PAs with a DP of 3, with one or two additional C2 -O7 ether bonds; for instance, at m/z 1179.0103, m/z 1181.0724, and m/z 1183.1965 (Table 3,   Table 1.   TOF mass spectrum of PAs in tea roots. The distance between each adjacent ion peak was 288 Da, corresponding to one catechin/epicatechin monomer (Table 3, Fig. 4D). This suggested the presence of a series of oligomers and polymers in roots, with DP ranging from 3 to 12. The catechin and/or epicatechin monomers were the predominant structural units of PAs. Moreover, only B-type PAs with C-C bond linkage were detected in roots. However, MALDI-TOF-MS failed to detect higher polymers of PAs (DP . 12). Also, the technique could not identify the epimerization of PA units, or assist with quantitative determination of PAs. The C-2 region occurred at 75-85 ppm, with 80 ppm as the dividing line between cis-and trans-structures. The signals at 75-77 ppm were attributed to the top extension units of the C2-cis form. The peak at 77-79 ppm was ascribed to C2-cis isomer in terminal units or monomeric catechins. A chemical shift of more than 80 ppm was the characteristic signal of the C2-trans configuration. The monomeric catechins and the polymeric PAs could not be distinguished by C-2 region. The spectrum presented in Fig. 5, also indicated the predominance of cis form, irrespective of the location. According to the area ratio of C2 (75-80 ppm) to [C2 (75-80 ppm) 1 C2 (80-85 ppm)], leaves contained 95% cis structure, while the roots comprised 100% C2-cis isomers.
The DP could be determined from the ratio of the peak areas of [C3 top and extension units to C3 terminal and monomeric units] 1 1. The shifts for C3 terminal and monomeric units, and C3 top and extension units occurred at 65.0-69.5 ppm and 69.5-75.0 ppm, respectively. As illustrated in Fig. 5, the signals for C3 occurred in leaves mainly at a concentration of 65.0-69.5 ppm, showing that the DP of PAs was 1.03. In roots, the C3 top and extension units were primarily present and the DP of PAs was estimated as 5.44. These results were also consistent with the thiolysis data.
When PAs were linked by C4-C8, the chemical shifts of C8 in extension and terminal units occurred at 106-107 ppm. The C4 in extension and terminal units occurred at 37-38 ppm. As shown in  www.nature.com/scientificreports These results, therefore, indicated co-existence of di-and trihydroxyl groups in the B-ring in leaves. C2-cis units were most commonly found. Polyphenols included various monomeric catechins. In roots, the units consisted only of di-hydroxyl groups in the B-ring, and C2-cis, with a DP . 5.
Purification and 1 H-NMR Characterization of PAs in Roots. PAs from roots of tea plants were purified and identified to facilitate their structural characterization with 1 H-NMR assay. The rough extracts of PA were separated into three sections (spot A, B, and C) by GF254 silica gel plate (Fig. S3 left). RP-HPLC analysis showed that the relative shift values (RF) of spot A (0.71) and spot B (0.29) were in accordance with the standard substances epicatechin (EC) and procyanidin B2, respectively ( Fig. S3A and 7B). Spot C was found to localize at the origin, which made it difficult to separate (Fig. S3C).
Compared with the silica gel plate, silica gel H column improved the purification efficiency owing to the large amounts of sample used. The rough extracts of PAs were further separated into six fractions (S1 to S6) on silica gel H column and stained with Vanillin-H 2 SO 4 and DMACA-HCl reagents (Fig. 6A). The DP of PAs increased gradually from S1 to S6 fractions via TLC and NP-HPLC ( Fig. 6B  and 6C). The fractions S1 and S2 were relatively simpler, and the single peak in fraction S6 suggested the existence of polymers (Fig. 6C). The S1, S2, and S6 fractions were purified by preparative HPLC.
The compounds obtained from S1, S2, and S6 fractions were identified by 1 H-NMR interpretation (Fig. 7). The results suggested that the compounds in S1 and S2 fractions were EC and procyanidin B2 (structures shown in Fig. 1), respectively, by comparing with the reference compounds in published literature 47,48 . Chemical shifts were expressed in ppm (d) and were as follows. EC  (Fig. 7A and 7B). H-2 and H-3 chemical shifts in the top unit (C-ring, d 5.07 and 3.91) of B2 were also shifted, although the regularity was not as obvious as that of H-4 (Fig. 7B), which meant that the linkage was at position 4 in the C-ring.
S6 fraction was the polymer of mixed PAs. PAs with higher DP were difficult to purify due to their easy oxidation and isomerization. Compared with EC and B2 (Fig. 7A and 7B), the corresponding peaks of S6 (Fig. 7C) were broader, which is a feature of the polymers. In particular, the relative abundance of H-4 in terminal units and the significant decrease in H-6 and H-8 suggested the presence of C4-C6 or C4-C8 linkages of the polymeric PAs in the roots of tea plants.
Based on the 1 H-NMR spectrum of EC, B2, and polymeric mixed PAs, coupled with the results of hydrolysis, MS, and 13 C-NMR, it was concluded that the PAs of roots were homogeneous.
Quantitative Determination of PAs in Roots of Tea Plants. Plant PAs are inherently complex with different linkages, diverse monomer units, and variable chain lengths 2,3 . The lack of appropriate standards and efficient analytical methods pose several challenges to the accurate quantitative determination of PAs. In the present study, we found the accumulation of PAs in tea roots, with a relatively simple structure (Fig. S2C, 5B, and 6B). Therefore, root PAs were the first to be analyzed quantitatively. After separation and purification, 1222 mg freeze-dried powders were obtained from 15 g dry-weight roots. It was estimated that the roots contained 8.1% of PAs. In addition, re-extraction and re-purification yielded 366.7 mg PAs from 400 mg of purified PAs, with a recovery rate of 91.7%. Therefore, purification could be accomplished for the quantitative estimation of PAs in roots, even though the technique was complicated. Spectrophotometry is a simple and effective method for the quantitative determination of PAs. In previous studies 44,49,50 , Vanillin-H 2 SO 4 and DMACA-HCl methods were found to be more appropriate in the absence of interference by monomeric catechins. Furthermore, the butanol-HCl protocol was specifically  used for the analysis of oligomeric and polymeric PAs, in the absence of anthocyanin and leucocyanidin in the samples 44 . Hence, a standard must be selected carefully, due to varying degrees of polymerization. For instance, in our study, epicatechin (monomeric PAs), B2 (dimeric PAs), and polymeric PAs (the mixture) exhibited different levels of sensitivity to Vanillin-H 2 SO 4 , DMACA-HCl, and butanol-HCl reagents (Fig. S4). Nevertheless, the results with different approaches varied even against identical standards. For example, when purified root PAs were used as standard, the PA content was 121.77 mg/g, 104.32 mg/g, and 85.06 mg/g by Vanillin-H 2 SO 4 , DMACA-HCl, and butanol-HCl protocols, respectively. The results of the butanol-HCl assay closely matched the results of the purification method. In addition, the accumulation of insoluble PAs was rare, and only a tenth of the soluble PAs were found in tea roots (data not shown).
In Vitro Assay for the Auto-condensation of Flavan-3-ols. Nonenzymatic in vitro assay with different monomeric flavan-3-ols and procyanidin B2 was carried out for investigating the auto-condensation mechanism of PAs. Whether in acid or alkaline solution, EC could be condensed to form multiple dimers under mild conditions in vitro (25uC). In this study, the auto-condensation of EC showed pHdependent and time-related features. Although the types of dimers of EC were consistent, their contents were different when assayed at different pH (Fig. 8A). However, the time-depent feature of autocondensation was one of the most easily neglected aspects, since it was difficult to observe the process within a short time (less than 3 d) in mild reaction conditions (at pH 3.0), or in a protracted period of time (more than 2 d) due to the inevitable oxidation reaction (at pH 7.5).
As expected, the dimers formed in vitro did not match with those found in tea plants, in light of the retention time, fragments of MS/ MS spectrum, and collision energy (Fig. 8A and 8B). For collision energy, a stronger energy (26 eV) was needed to collide the dimers of EC formed in vitro than that of procyanidin B2 (18 eV) in tea plants (Fig. 8B); thereby showing that the condensation pattern of the dimers formed in vitro was different from procyanidin B2 (4-8 linkage). It was speculated that the dimers formed from commercial EC in vitro were possibly the oxidation products, which did not exist in fresh tea plants.
Moreover, the same trimers (m/z 865) were detected in the negative mode when a combination of procyanidin B2 and monomeric flavan-3-ols (EC or C), or procyanidin B2 alone was used as a substrate, in acid as well as alkaline solutions, but especially in the acid solution (Fig. 9). Based on the retention time and the fragments of MS/MS spectrum, the trimer obtained at peak T1 in vitro revealed the same compound as represented by peak T2 in vivo (Fig. 9A, 9B, 9C and 9D). This suggested that the trimer may have contained the C4-C8 or C4-C6 linkage formed from the direct reaction. In addition to the trimers, monomers (m/z 289) (Fig. 9E) and trace tetramers (m/z 1153) (data not shown) were also detected among the reaction products with procyanidin B2 as the substrate. This indicated that procyanidin B2 was degraded into its carbocation intermediate and monomeric flavan-3-ols; and the intermediate was further processed by nucleophilic addition of B2 to initiate the formation of trimers and tetramers. The trimers, T3 and T4, T5 and T6, were the major products formed at pH 7.5, which could essentially be the oxidation products of the substrates, when procyanidin B2 and EC or C were used as substrates, respectively ( Fig. 9A and B).
Enzyme assays were designed to explore the polymerization mechanism of PAs in vitro, using crude enzymes extracted from tea plants and grape seeds, and commercial oxidases, including polyphenol oxidase (PPO) and peroxidase (POD). The reaction mixtures were observed to undergo oxidation when crude enzyme extracted from tea plants or the commercial oxidases (PPO and POD) were added. Peaks of m/z 575 and m/z 1151 were detected when procyanidin B2 was used along with the crude enzyme extracted from grape seeds (data not shown). This suggested the occurrence of an enzymatic oxidation reaction in the assays. However, no obvious oligomeric PAs were detected in weak acid solution (pH 5.5) or in alkaline solution (pH 7.5), when UPLC-QQQ-MS/MS was performed with procyanidin B2 and EC or only procyanidin B2 as the substrate (data not shown).
To sum up, the dimerization and trimerization reactions in vitro using monomeric flavan-3-ols as substrates could not reproduce the products formed in vivo. However, a trimerization reaction, which occurred in vitro with procyanidin B2 and monomeric flavan-3-ols (EC or C) or only procyanidin B2 as the substrate, was indicated to take place in vivo as well.

Discussion
The Qualitative and Quantitative Identification of PAs. Our results showed that LC-ESI-TOF-MS was particularly useful for the qualitative analysis of monomeric catechins and oligomeric www.nature.com/scientificreports PAs (DP , 3). UPLC-QQQ-MS/MS was used not only for qualitative analysis but also for quantitative analysis of the monomers and oligomers of PAs (DP , 4) based on the major fragment ions produced from MS/MS. MALDI-TOF-MS with precise molecular weights was found to be superior to NP-HPLC in terms of the qualitative analysis of oligomeric and polymeric PAs. The most important information for the structural characterization of PAs can be obtained from the 13 C-NMR spectrum; the data include the stereochemistry of the heterocyclic ring at C2-C3 position, hydroxylation pattern of the B-ring, structures of the chain terminating units, and the DP of PAs 15,16,46 . In order to obtain the 1 H-NMR spectrum, higher purity of the sample is required.
Our results additionally showed that the di-and tri-hydroxyl groups in the B-ring units, cis-and trans-C-2 units, galloylated and nongalloylated units of monomeric and polymeric flavan-3-ols co-existed in leaves. However, the tri-hydroxyl, cis-and galloylated flavan-3-ols were the main monomeric catechins units, and the (epi)catechin was the major unit of polymeric flavan-3-ols found, when the DP was greater than five. The PAs were observed to be abundant in roots, with the epicatechin unit acting as the predominant catechin of oligomeric and polymeric PAs. Comparative studies of flavonoid metabolism in tea leaves and roots will provide greater insights into the mechanisms of galloylation and polymerization. Indeed, the lack of galloylated catechins and compounds with tri-hydroxyl groups in the B-ring might be attributed to the absence of the expression of CsF3'5'H and CsEGGT and the related enzyme activity in roots 6 . It could be speculated from these results that the polymerization of the monomers was hampered by the 5'-position of hydroxyl groups in the B-ring, the galloyl group at the 3'-position of the C-ring, and the 2, 3-trans isomers. Roots were probably the ideal material for studying polymerization metabolism, due to the presence of di-hydroxyl groups in the B-ring and lack of other flavonoids.
The major limitation to accurate quantitation of PAs is the lack of availability of appropriate standards. Spectrophotometry staining methods with Vanillin-H 2 SO 4 44 , or DMACA-HCl reagents 49 16 .
PAs from plant sources are naturally complex due to the occurrence of several kinds of monomer units, different types of linkages, and variable chain lengths [1][2][3]5,15,16 . Further, the structural characterization of PAs in each of our samples was found to be varying, requiring confirmation with different standards. Compared with the relatively simple and unitary PAs in tea roots, the existence of a plethora of monomeric catechins and impurities in tea leaves complicates the purification method for quantitative analysis, rendering it impracticable. The spectrophotometry protocol entails staining with Vanillin-H 2 SO 4 44 , and DMACA-HCl 44,49 reagents, which react with the A-ring of catechins. It is the most common quantitative analytical method for PAs. In hot mineral acid solutions, PAs get converted into anthocyanidins, enabling quantitative analysis and determination of absorbance at 550 nm A 550 . However, the spectrophotometry protocols were found to be affected by a series of factors 44,51 , with undesirable effects on their stability and reproducibility.
The Auto-Condensation Mechanism of PAs. According to a former study 52 , all the rings (A, B, and C) of the flavonoid framework are reactive and can undergo three specific types of reactions. The A ring, with a meta-hydroxy substitution pattern, is prone to the polycondensation reaction. The B ring, with an ortho-dihydroxy substitution pattern, leads to the oxidation reaction. The C ring, when positively charged in an acid solution, is inclined to process direct reactions, such as proanthocyanidins condensation.
The condensation mechanism of proanthocyanidins has gained much attention since the 1970s in the last century. Based on the structure and stereochemistry of proanthocyanidins, the isotopic tracer studies on procyanidin biosynthesis, and the biogenetically patterned synthesis of natural procyanidins in the laboratory, it was hypothesized that the dimers are biosynthesized from two metabolically distinct units, the nucleophilic flavan-3-ols (constituting the lower unit), and the electrophilic C-4 carbocation (forming the upper unit), which may generate from flav-3-en-3-ol 17,53-57 .
The origin of C-4 carbocations or flav-3-en-3-ol in vivo is intriguing, and these compounds may be difficult to identify directly in plants; which in turn renders it difficult to comprehend the condensation mechanism of proanthocyanidins. However, since 1980, several progresses have been made in the biochemical and molecular fields for the understanding of flavonoid biosynthesis. Stafford et al. reported that the dihydroflavonols (dihydroquercetin and dihydromyricetin) could sequentially be reduced to form 3,4-diol and 2,3trans-(-)flavan-3-ols by two NADPH-dependent reductases, which were later reported to be DFR 58 and LAR 59 derived from tissue cultures of Douglas fir and Ginkgo biloba 60,61 . Meanwhile, they also found that the production of gallocatechin (4a-8)-catechin dimer could be detected when (1)-catechin was added to the enzymic incubation mixture with dihydromyricetin in acidic medium, suggesting that the carbocation or quinone methide intermediate might be derived from 3,4-diol (leucodelphinidin) and subsequently added to an electrophilic group, such as (1)-catechin to form dimer with 4-8 or 4-6 bond linkages 62 . Additionally, they theorized the presence of a C3-epimerase involved in the biosynthesis of 2,3-cis (-)flavan-3ols 63 . However, it was later demonstrated by Xie et al. that the BANYULS (BAN) genes from Arabidopsis thaliana and Medicago truncatula could encode anthocyanidin reductase, which converted cyanidin to 2,3-cis-(-)-epicatechin 64 ; and therefore, speculated that flav-3-en-3-ol might be an intermediate in this conversion process 65 . Further, Dixon et al. 19 provided a theoretical view that the carbocation and quinone methide intermediates might be derived from the corresponding flavylium ion of anthocyanidins in acidic conditions catalyzed by polyphenol oxidase (PPO). They alternatively proposed that EC or C could be converted to carbocation derivatives via corresponding quinones and flav-3-en-ols, which were in turn catalyzed by PPO or through non-enzymatic coupled oxidation reactions. Certain studies have shown that tyrosinase, PPO, and laccase-like enzymes might be involved in the enzymatic condensation and polymerization of PAs [31][32][33] .
Based on the aforementioned studies, leucoanthocyanidin, anthocyanidin, and flavan-3-ols were likely to be the substrates in the formation of C-4 carbocation or flav-3-en-3-ol. One of the reasons for the insufficient knowledge on the condensation mechanism of proanthocyanidins is the lack of a biogenetically patterned synthesis assay. For instance, the suitable substrates, leucoanthocyanidin and carbocation, are commercially unavailable, and the flavan-3-ols have been found to be unstable and readily oxidized in vitro. Considering the higher content of PAs in older stems and roots, and the irrelevant trend between the accumulation of PAs and the relative expression rate of LAR and ANR genes in tea plants 6 , we inferred that the condensation and polymerization of PAs occur only after the biosynthesis of monomeric flavan-3-ols.
In the present study, procyanidin B2 could not be detected, as expected, when only EC was used as a substrate in the non-enzymatic in vitro assay in acid or alkaline solution (from pH 3.0 to 7.5) (Fig. 8). Therefore, it was proved that the carbocation or flav-3-en-3-ol did not generate from EC. The primary dimers in the above-mentioned assay needed higher energy to collide as compared to the dimers with C4-C8 or C4-C6 linkages, such as procyanidin B2 (Fig. 8A and 8B); and thus, the generation of anthocyanin became difficult when sub-www.nature.com/scientificreports mitted to acid-catalyzed hydrolysis. These characteristics were consistent with that of certain dimers with biphenyl linkages (QB-QA) and biphenyl ether linkages (QB-O-QA), which usually form from the oxidation reactions of flavanols 52 .
The same procyanidin trimers could be detected in the extracts of tea plants as well as in the non-enzymatic in vitro assay, when procyanidin B2 with EC or C were used as a substrate in acid and alkaline solutions, respectively ( Fig. 9A and 9B). Interestingly, the same procyanidin trimer was also detected when only procyanidin B2 was used as the substrate in non-enzymatic in vitro assay (Fig. 9C). This suggested that procyanidin B2 not only released the carbocation as electrophilic upper units, as expected, but also as nucleophilic lower units, to form the procyanidin trimer in vitro or in vivo.
These findings led us to speculate that the mechanism of autocondensation involves the release of carbocations through certain cleavage reactions of B2; following which, the carbocations as electrophiles were involved in the direct reactions with procyanidin B2 as nucleophiles, leading to the formation of a trimer in vitro or in vivo, as found in the roots of tea plants (Fig. 10).
Nevertheless, the procyanidin trimerization in the non-enzymatic in vitro assay showed a very low conversion rate, approximately 5-10%, which did not correspond to the high concentration of PAs in tea plants. Whether or not a more efficient reaction in tea plants occurs, such as enzymatic condensation, still needs to be probed. In order to research the enzymatic condensation mechanism of PAs, we worked with procyanidin B2 as a substrate in weak acid (pH 5.5) as well as alkaline solutions (pH 7.5); however, no enzymatic product was detected. Further studies could, therefore, prove instrumental for the in-depth understanding of the mechanisms of proanthocyanidins condensation, auto-or enzymatic condensation.

Methods
Plant Materials. The samples of leaves, stems and roots were collected from mature tea plants during early summer from the Experimental Tea Garden, Anhui Agricultural University, Anhui, China. The collected samples were immediately frozen in liquid nitrogen and stored at 230uC prior to analysis.
Extraction and Purification of PAs. PAs were extracted and purified according to the method described by Gu et al 1 with some modifications. Samples (leaves, stems, or roots; 120 g) were extracted three times with an extraction solvent containing 70% acetone and 0.5% acetic acid. The extracts were combined, and the acetone was evaporated at 30uC using a rotary evaporator under vacuum. The resultant brown slurry (approximately 600 mL) was extracted three times with chloroform (500 mL each) and the aqueous fraction was applied to a 50 3 3.5 cm Sephadex LH-20 column. The column was eluted with 50% methanol, followed by elution with 70% acetone to remove all the sugars and phenolic acids. The brown acetone fraction containing PAs was collected under partial vacuum at 30uC, and then freeze-dried to obtain a fluffy rough PAs powder (from individual sources of leaves, stems, and roots), which was subsequently characterized using RP-HPLC-TOF-MS, UPLC-QQQ-MS/MS, NP-HPLC, MALDI-TOF-MS, and 13 C-NMR analyses.
The rough PAs powder extracted from the roots was further purified on thin-layer chromatography (TLC, GF254 silica gel plate) on a small scale, with elution using chloroform: methanol: formic acid (2851051, v/v) 66 , or on a larger scale on a silica gel H column chromatography (50 3 3.5 cm) with elution using chloroform: methanol: formic acid (75350.4, v/v). The PAs fractions detected qualitatively by the Vanillin-H 2 SO 4 44 , and DMACA-HCl reagent 49 were collected and freeze-dried. The S1, S2, and S6 fractions were further purified by preparative HPLC. The purified fractions were concentrated and freeze-dried for 1 H-NMR analysis of the characteristics of PAs in roots.
Identification of PAs in Tea Plants. Reversed-Phase LC-ESI-MS. Liquid chromatography (LC)-time of flight (TOF) mass spectrometer (MS) system and the ultra performance liquid chromatography (UPLC)-triple quadrupole mass spectrometry (QQQ-MS/MS) system were used for the qualitative identification of monomeric catechins and oligomeric PAs as described by Jiang et al. 6 Hydrolysis. Upon treatment with hot mineral acid solutions, PAs were depolymerized into colored anthocyanidins with absorbance maxima around 550 nm 44,51 . PAs of tea plants were assayed by the butanol-HCl method according to Jiang et al. 6 . The hydrolysis products (anthocyanidins) were further characterized using reversedphase (RP)-HPLC analysis. www.nature.com/scientificreports Thiolysis. We performed thiolysis of PAs based on the methods described by Guyot et al. 13 and Gu et al. 1 . Briefly, in a 250-mL polypropylene insert (Fisher Scientific, Boston, MA), 50 mL solution of each fraction (2 mg/mL in methanol) was mixed together with 50 mL of methanol acidified with concentrated HCl (3.3%, v/v) and 100 mL of benzyl mercaptan (5%, v/v in methanol). The inserts were transferred into a 1.5-mL vial and sealed with an inert Teflon cap (Agilent Technologies, Wilmington, DE). The reaction was carried out at 40uC for 30 min and then kept at room temperature for 10 h to ensure complete degradation. To minimize further epimerization and any side reaction, the reaction mixtures were left in the freezer (220uC) until 5 mL was injected directly for RP-HPLC analysis, for additional characterization of the products of thiolysis (catechins and their benzylthioethers).
Reversed-Phase HPLC Analysis of Products of Hydrolysis and Thiolysis. A Phenomenex Synergi 4 u Fusion-RP80 column (particle size: 5 mm, length: 250 mm, and internal diameter: 4.6 mm) was used at a flow rate of 1.0 mL min 21 . The column oven temperature was set at 25uC. The mobile phase consisted of 1% acetic acid in water and 100% acetonitrile. The gradient of the hydrolyzed products increased linearly from 13 to 30% (v/v) at 30 min, to 13% at 32 min. The DAD was set at 280 nm and 530 nm, respectively for real-time monitoring of the peak intensities. The gradient of the products of thiolysis increased linearly from 13 to 24.3% (v/v) at 20 min, to 40% (v/v) at 30 min, to 55% (v/v) at 45 min, to 80% at 55 min, to 13% at 60 min. The DAD was set at 280 nm for catechin, epicatechin, gallocatechin, epigallocatechin, epigallocatechin gallate, epicatechin gallate, and their benzylthioethers.
Normal-Phase HPLC Analysis. The analysis was conducted using an Agilent 1100 HPLC system with Phenomenex Luna 5 u Silica (2) 100 A column (particle size: 5 mm, length: 250 mm, and internal diameter: 4.6 mm). The detection wavelength was set at 280 nm and the column oven temperature at 25uC. The ternary mobile phase consisted of (A) methanol, (B) dichloromethane, and (C) acetic acid and water (151 v/v). The 57-min linear gradient based on that of Hammerstone et al. 67  MALDI-TOF-MS Analysis. MALDI-TOF-MS was performed according to the method described by Zhang et al. 16 . The MALDI-TOF-MS spectra were recorded on a Bruker Reflex III instrument (Germany). The irradiation source was a pulsed nitrogen laser with a wavelength of 337 nm, and the duration of the laser pulse was 3 ns. In the positive reflection mode, an accelerating voltage of 20.0 kV and a reflectron voltage of 23.0 kV were used. Spectra of PAs were obtained from a sum of 100-150 shots and calibrated using angiotensin II (1,046.5 MW), bombesin (1,619.8 MW), ACTHclip18-39 (2,465.2 MW), and somatostatin 28 (3,147.47 MW) as external standards. 2,5-dihydroxybenzoic acid (DHB, 10 mg/ mL aqueous solution) was used as the matrix. The sample solutions (7.5 mg/mL aqueous) were mixed with the matrix solution at a volumetric ratio of 153. The mixture (1 mL) was applied to the steel target. Amberlite IRP-64 cation-exchange resin (Sigma-Aldrich), equilibrated in deionized water, was used to deionize the analyte/matrix solution three times. Cesium chloride (1 mg/mL) was mixed with the analyte/matrix solution at 153 volumetric ratio to promote the formation of a single type of ion adduct ([M 1 Cs] 1 ) 68,69 .
NMR Analysis. A protocol described by Kraus et al. 15 was employed for nuclear magnetic resonance (NMR) spectroscopy analysis. The preliminary purified PAs of leaves and roots were analyzed by 13 C-NMR and the purified PAs of roots were identified by 1 H-NMR. The compounds were dissolved in a mixture of 0.25 ml acetone and 0.25 ml D 2 O and transferred through sintered glass filters into 5-mmdiameter NMR tubes. 1 H-NMR and 13 C-NMR spectra were obtained on an AVANCE AV 400 (400/100 MHz) spectrometer (Bruker, Fallanden, Switzerland), using inverse-gated decoupling, 45u pulse, 0.4 sec acquisition time, 2.6 sec delay, and 25,000-30,000 scans.
Extraction and Quantitative Determination of PAs in Tea Roots. The total PAs were extracted and purified from 50 g of roots. The extraction and preliminary purification were carried out according to the above described protocols. The freezedried fluffy powders were weighed to calculate the percentage of PAs. We selected 400 mg of powdered root PA with purity greater than 90%, for re-extraction and purification by Sephadex LH-20 column to measure the recovery rate of purified PAs.
Quantitative determination of PAs in root was accomplished using the Vanillin-H 2 SO 4, 44 , DMACA-HCl 49 , and butanol-HCl 6 methods; EC, B2, and purified PAs of root were used as standards, respectively.
Auto-condensation of Flavan-3-ol in Vitro. In acid solution, the reaction mixture was incubated at 25uC for 15 d, in a total volume of 0.2 mL consisting of 100 mM acetate buffer (pH 3.0), or 100 mM Mes buffer (pH 5.5), and 0.4 mM each of procyanidin B2, EC, or C. In alkaline solution, the reaction mixture were determined at 25uC with 24 h incubations, in a final volume of 0.2 mL consisting of 100 mM Tris-HCl buffer (pH 7.5), and 0.4 mM of the flavan-3-ols, as used in acid solution. The reaction products were then analyzed by UPLC-QQQ-MS/MS system as described by Jiang et al. 6 . The extraction of crude enzyme was carried out as described by Liu et al. 7 , and the reaction mixture was similar to to that of a non-enzymatic reaction.