Stacking triple genes increased proanthocyanidins level in Arabidopsis thaliana

Anthocyanins and proanthocyanidins are two important plant secondary metabolites, and they contribute to plant survival and human health. In particular, proanthocyanidins could also prevent ruminants from the damage of pasture bloat. However, the improvement of proanthocyanidins content remain unsatisfied. In this study, we attempted to improve proanthocyanidins level by gene stacking in Arabidopsis thaliana as prove-of-concept. Two proanthocyanidin pathway genes from tea plant, CsF3’5’H and CsANR2, were co-expressed in the wild type and PAP1 over-expression Arabidopsis. Over-expression of CsF3’5’H slightly affected anthocyanins level in leaves and proanthocyanidins in mature seed when expressed alone in the pap1-D line. Over-expression of CsANR2 led to an obvious decrease in anthocyanins in leaves of both wild type and pap1-D lines, but increase in proanthocyanidin level in mature seeds. Over-expression of CsANR2 in pap1-D lines lead to production of DMACA-reactive soluble proanthocyanidins in leaves, but not in wild type or pap1-D lines. Anthocyanins level was decreased in the leaves of CsF3’5’H, CsANR2 and pap1-D co-expression lines, but proanthocyanidins were increased remarkably in both leaves and mature seeds in the co-expression line. It is concluded that co-expression of CsANR2 and PAP1 in Arabidopsis produce soluble proanthocyanidins in leaves, and co-expression of CsF3’5’H, CsANR2 and PAP1 lead to a significant increase in proanthocyanidins in mature seeds. The transcript levels of endogenous CHS, DFR, ANS and ANR genes in Arabidopsis were up-regulated in the triple genes co-expression line. Based on these studies, it is possible to develop new plant germplasm with improved proanthocyanidins by co-expressing of multiple genes.


Introduction
Flavonoids are one of the most important secondary metabolite groups in plant, among them, anthocyanins and proanthocyanidins are two important branch products in flavonoid pathway. Anthocyanins contribute to plant coloration and survival [1], and proanthocyanidins provide defense against plant pathogens and other diseases [2,3]. Both anthocyanins and a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 proanthocyanidins showed numerous beneficial effects on human health [4,5]. For animal health, it has been suggested that modest level of proanthocyanidins in the leaves and stems of alfalfa and white clover could prevent pasture bloat which is lethal in ruminant animals [6].
In Arabidopsis, flavonoid biosynthetic pathway starts with the formation of chalcone by the condensation of three molecules of malonyl-CoA and one molecule of p-coumaroyl-CoA, and this step is catalyzed by chalcone synthase (CHS). Chalcone will be subsequently isomerized by chalcone isomerase (CHI) to form flavanone, and flavanone can be catalyzed by flavanone 3-hydroxylase (F3H) to form dihydroflavonol. As the substrate of dihydroflavonol reductase (DFR), dihydroflavonol can be further catalyzed to form leucoanthocyanidins, or to form anthocyanidins by the catalyzation of anthocyanidin synthase (ANS). Subsequently, anthocyanidin will be reduced by anthocyanidin reductase (ANR) to form proanthocyanidin precursors [7] (Fig 1). In Arabidopsis, dihydrokaempferol and kaempferol can be hydroxylazed by F3'H to form dihydroquercetin and quercetin on the 3' position of B-ring. The existence of F3'5'H in many other plants, such as tea plant, grape and Medicago truncatula could promotes further hydroxylation at the 5' position in the B-ring of dihydroquercetin and quercetin, to form dihydromyricetin and myricetin [8] (Fig 1). However, no F3'5'H gene exists in Arabidopsis, which could not synthesize any flavonoid derivatives with trihydroxyl groups on the B-ring as tea plant or grapes that are rich in proanthocyanidins in leave tissues.
The biosynthesis of anthocyanins and proanthocyanidins, and the conversion between them have been hotspots in the study of flavonoid biosynthesis. Arabidopsis pap1-D (production of anthocyanin pigment 1-Dominant) mutant accumulates massive anthocyanins, in which the MYB type transcription factor PAP1 is activated by using the activation tagging with the enhancer from 35S promoter [9]. A number of genes in anthocyanin biosynthesis, such as CHS, CHI, F3H, F3'H, DFR and ANS, are strongly induced by PAP1 in pap1-D line [10]. Theoretically, more anthocyanin aglycones could provide sufficient substrates for proanthocyanidin production in pathway. In a previous study, ectopic expression of PAP1 led to massive anthocyanin accumulation in tobacco, and a large amount of proanthocyanidins are accumulated in both mature and young leaves of the PAP1 and MtANR expression plants. Similar results were demonstrated in Medicago truncatula, more proanthocyanidins are accumulated in the leaves of PAP1 and MtANR over-expression plants compared with the wild type [11]. These results suggested that ectopic accumulation of proanthocyanidins in plant, in particular in leaves, may be feasible by co-expression of PAP1 and ANR genes.
ANR is a key gene for the formation of proanthocyanidin monomers from anthocyanidins, the expression level of ANR directly correlates with the accumulation of proanthocyanidins in many plants. The expression pattern of VvANR from grapevine (Vitis vinifera L. cv Shiraz) was positively correlated with the level of proanthocyanidins in grape fruit; the expression level of VvANR determined the accumulation of proanthocyanidins in berry development [12]. In blackberry (Rubus spp.) fruit, anthocyanins level was gradually increased and proanthocyanidins level was gradually decreased during maturation, the transcript levels of RuANS, and RuANR were coordinated with the changing patterns of anthocyanins and proanthocyanidins, in which RuANS was gradually up-regulated and RuANR were down-regulated during the ripening of blackberry fruit [13].

PLOS ONE
have been characterized to be functional by us in a previous study [20], therefore, it is feasible to re-construct the pathway with genes from the proanthocyanidin-rich tea plant, in another model plant Arabidopsis as prove of concept.
In this study, we demonstrated that by the combined over-expression of CsF3'5'H, CsANR2 and PAP1 in Arabidopsis, the accumulation of proanthocyanidins could be changed not only in mature seeds but also in leaves.

Plant material and growth condition
After vernalized under 4˚C for 2 days, the wild type Arabidopsis (Col-0) and all transgenic lines were germinated and grown in mix soil (vermiculite: nutrient soil = 1:1). The plants were grown under long-day condition (16 h light and 8 h dark) in growth chamber with constant temperature of 23˚C.
For the observation of anthocyanins, to avoid obstruction on chlorophyll, Arabidopsis were planted on 1/2 MS medium supplied with 100 μM Norflurazon and grown for 4 days after vernalization (4˚C for 2 days).

Cloning and vector constructs
The amplification of the ORF fragment of CsANR2 (NCBI cDNA accession number: GU992400) and the vector construction of pB2GW7-CsANR2 were described in our previous study [20]. The ORF fragment of CsF3'5'H (NCBI cDNA accession number: DQ194358) was amplified from cDNA by reverse transcription from a tea cDNA library [20], using primers: CsF3'5'H-CF: 5'-CACC ATG GCCCTAGACACCGTCTTCCTGC-3'(the start codon was boxed), CsF3'5'H-R: 5'-TTA AGCAGCATAAGCATTTGGAGGC-3'(the stop codon was boxed), PCR products were purified and cloned into the Gateway Entry vector pENTR/ D-TOPO and confirmed by sequencing. The Entry vector was then transferred into Gateway plant transformation vector pK2GW7 to generate pK2GW7-CsF3'5'H by LR reaction.

Agrobacterium-mediated transformation and the crossing of Arabidopsis
The pK2GW7-CsF3'5'H, and pB2GW7-CsANR2 were individually transformed into Agrobacterium tumefaciens strain GV3101, a single colony for Agrobacterium-mediated transformation of Arabidopsis was confirmed by PCR. After inoculated into LB liquid medium containing 50 mg/L spectinomycin and 50 mg/L rifampicin, cell suspension was cultured until OD 600 = 0.6. The cells were collected by centrifugation at 4000 g for 10 min, then resuspended in MS liquid medium for plant transformation. The transformation was performed by using floral-dip method [21].
Before cross-fertilizing, petals and anthers of well-grown female parent plant were gently removed before blooming, then pollenated by the anthers from male parent plant. The pollinated flower was covered by a transparent plastic bag to prevent pollution from other anthers until the silique is formed. For the crossing of CsF3'5'H × pap1-D and CsANR2 × pap1-D overexpression lines, pap1-D was used as male parent, the CsF3'5'H and CsANR2 over-expression lines were used as female parent, for the crossing of CsF3'5'H × CsANR2 × pap1-D overexpression lines, CsANR2 over-expression lines were used as male parent, and the CsF3'5'H × pap1-D over-expression lines were used as female parent.

Metabolite analysis
To analyze total flavonoids and flavonols, aerial part of eh 21-day-old seedlings and mature seeds were collected and then grounded in liquid nitrogen. After lyophilized at -50˚C for 18 h, 1 ml 80% methanol were firstly added to the samples (200 mg for seedlings and 50 mg for seeds) and sonicated for 30 min, then centrifuged at 10000 g for 30 min after placed at 4˚C for 12 h. The absorption of the supernatant was measured at 254 nm. Absorbance values were converted into flavonoids equivalents by using the standard curve of quercetin. The flavonoid extracts were also run on an HPLC 1260 (Agilent) system with an Eclipse XDB-C18 reverse phase column (4.6 × 150 mm, particle size 5 μm) with 150 μL of the supernatant, a linear eluting gradient were consisted of solvent A (0.1% formic acid in water) and solvent B (acetonitrile): 0 min, 10% B; 10 min, 20% B; 15 min, 20% B; 25 min, 50% B; and 30 min, 60% B. A flow rate of 0.8 ml/min were used for the separation of the compounds, the UV-visible absorption from 190-600nm was detected by using a DAD detector (Agilent). Peak areas were converted and calculated into flavonols equivalents by using the standard curve of kaempferol.
For the analysis of anthocyanins, leaves of 30-day-old plants were collected and then lyophilized, then the samples (200 mg) were extracted with methanol (contains 0.1% hydrochloric acid, by vol.) and sonicated for 30 min. After centrifuged at 10000 g for 30 min, the supernatant was extracted and chloroform/water were added to remove chlorophyll. The absorption of the products was measured at 550 nm. Data were collected from three biological replicates. The anthocyanins level was calculated by using the formula: C anthocyanins = OD 550 /ε × V/ m × 10 6 ; in which ε = 4.62 × 10 6 , V = 4mL, m = 200 mg.
To analyze the soluble proanthocyanidins, leaves of 30-day-old plants and mature seeds were collected and grounded in liquid nitrogen and lyophilized in -50˚C for 18 h. DMACA (pdimethylaminocinnamaldehyde) is an aldehyde dye, and proanthocyanidins can react with DMACA to form a blue stain. For the DMACA detection method, 500 μL 70% acetone (contain 0.5% acetic acid by vol.) were firstly added to the samples (200 mg for leaves and 50 mg for mature seeds) and sonicated for 30 min, and then centrifuged at 3000 g for 10 min, the extraction was repeated twice and the supernatants were all collected. After extracted by chloroform and n-hexane to remove the chlorophyll, 760 μL of 0.2% (w/v) DMACA solution were added into 40 μL of the supernatants and the absorption were measured at 640 nm within 15 min; for the butanol-HCl reaction, 760 μL of the butanol-HCl (95%-5%, by vol.) reagent was added into 40 μL of the supernatants and the absorption 1 (A1) of the samples were measured at 550 nm, then the samples were boiled at 95˚C for 1 h and measured at 550nm again as absorption 2 (A2), the practical absorption of the samples was calculated as A2 minus A1. Absorbance values were converted into proanthocyanidins equivalents by using the standard curve of procyanidin B1. The remained residues of these steps were all collected for insoluble proanthocyanidins measurement.
To analyze the insoluble proanthocyanidins, the residues described above were dried using a vacuum drier, then 1 mL of the butanol-HCl (95%-5%, by vol.) reagent was added and sonicated for 60 min. After centrifuged at 3000 g for 10 min, the absorption (A1) of the samples were measured at 550 nm. Then the samples were boiled at 95˚C for 1 h and measured at 550 nm again as absorption 2 (A2). The practical absorption of the samples was calculated as A2 minus A1. Absorbance values were converted into proanthocyanidins equivalents by using the standard curve of procyanidin B1.

Analyses of anthocyanins in transgenic Arabidopsis
To verify the functions of CsF3'5'H and CsANR2 in vivo, we individually transformed CsF3'5'H and CsANR2 into the wild type Arabidopsis (Columbia-0), 42 CsF3'5'H over-expression lines and 45 CsANR2 over-expression lines were obtained, respectively. The over-expression lines of these two genes were verified by both RT-PCR and qRT-PCR, one line with the highest expression level of 42 CsF3'5'H over-expression lines or 45 CsANR2 over-expression lines were identified, respectively, and further used for the following cross-fertilization assays. We pollinated the stigma of the CsF3'5'H over-expression line or the CsANR2 over-expression line with pap1-D line to obtain CsF3'5'H × pap1-D and CsANR2× pap1-D double over-expression lines, respectively. At last, the stigma of one CsF3'5'H × pap1-D over-expression line was pollinated with pollens from one CsANR2 over-expression line to obtain the CsF3'5'H × CsANR2 × pap1-D triple over-expression lines. The expression levels of the three genes CsF3'5'H, CsANR2 and PAP1 in all the transgenic plants were further confirmed by RT-PCR and qRT-PCR (Fig 2).
First, the accumulation of anthocyanins was analyzed in the 4-day-old seedlings, we found that the CsF3'5'H over-expression lines showed light purple cotyledons as the wild type, but the cotyledons of the CsANR2 over-expression lines were pale and slightly purple on the edge. The CsF3'5'H × pap1-D over-expression lines showed the same deep purple as pap1-D. Leave color of the CsANR2 × pap1-D over-expression lines and the CsF3'5'H × CsANR2 × pap1-D over-expression lines was lighter than that of the wild type (Fig 3A).
In the leaves of 30-day-old seedlings of the CsF3'5'H and CsF3'5'H × CsANR2 × pap1-D over-expression lines, the petioles were light purple and the blades were green, which was nearly the same color as the wild type. The petioles and the blades were all green in the leaves of CsANR2 over-expression lines. Leaves in CsF3'5'H × pap1-D over-expression lines and pap1-D were intensively purple than the wild type. The leaves of CsANR2 × pap1-D overexpression lines were slightly purple (both petioles and blades) than the wild type, but not as deep as the CsF3'5'H × pap1-D over-expression lines or pap1-D lines (Fig 3B and 3C), indicating that anthocyanins might be converted to proanthocyanidin by CsANR2.
To further quantitatively determine anthocyanins content, anthocyanins in wild type and the other six representative transgenic lines were extracted and measured. As shown in Fig 3D, anthocyanins level in CsANR2 over-expression lines was only half of the wild type, and anthocyanins level in CsF3'5'H over-expression lines was nearly the same as the wild type. However, anthocyanins level in pap1-D was 30-fold higher than the wild type, anthocyanins level inCsF3'5'H × pap1-D line was decreased by 25-fold compared to the wild type. Anthocyanins level in CsANR2 × pap1-D over-expression lines decreased more obviously. When CsF3'5'H and CsANR2 were co-expressed in pap1-D, anthocyanins level was dramatically decreased to only 1.4-fold of the wild type ( Fig 3D).
Taken together, anthocyanins level in CsF3'5'H over-expression lines showed no difference from the wild type; the anthocyanins level of CsF3'5'H × pap1-D over-expression lines was obviously reduced compared to the pap1-D line, and the over-expression of CsANR2 could significantly reduce the level of anthocyanins in pap1-D. More importantly, anthocyanins level could be largely reduced in CsF3'5'H × CsANR2 × pap1-D over-expression lines, indicating that co-expression of CsF3'5'H and CsANR2 could significantly reduce anthocyanin levels in the presence of PAP1.

Analyses of proanthocyanidins in leaves of transgenic Arabidopsis
Proanthocyanidins are accumulated mainly in Arabidopsis seeds [23], and the co-expression of MtANR and PAP1 lead to a huge increase of proanthocyanidins in tobacco leaves [11]. To exam whether co-expression of CsF3'5'H, CsANR2 and PAP1 could affect proanthocyanidins biosynthesis in leaves, leaves of transgenic Arabidopsis were collected for proanthocyanidins measurement.
The DMACA-reactive soluble proanthocyanidins were only found in CsANR2 × pap1-D over-expression lines and no detectable soluble proanthocyanidins were detected in the other six lines by using DMACA method. When detected by using butanol-HCl method, the content of soluble proanthocyanidins in CsANR2 × pap1-D and CsF3'5'H × CsANR2 × pap1-D overexpression lines was approximately 3-fold (2.11 mg/g DW and 2.28 mg/g DW, respectively) higher than in the wild type (0.74 mg/g DW) and the level of butanol-HCl reactive proanthocyanidins in other 4 lines stayed unchanged (Fig 4A).

PLOS ONE
The insoluble proanthocyanidins level in the CsF3'5'H over-expression lines, CsANR2 overexpression lines showed no obvious difference (1.91 mg/g DW and 1.70 mg/g DW, respectively) with the wild type (2.02 mg/g DW). Over-expression of CsANR2 in pap1-D led to a 20% increase in insoluble proanthocyanidins compared to the pap1-D line (5.65 mg/g DW and 4.64 mg/g DW, respectively), and obvious increase was found in the CsF3'5'H × pap1-D overexpression lines (4.33 mg/g DW). Insoluble proanthocyanidins level in the CsF3'5'H× CsANR2 × pap1-D co-expression line was significantly up-regulated to 9.11 mg/g (DW), which was 4-fold higher than the wild type and 2-fold than the pap1-D line (Fig 4B).
In conclusion, proanthocyanidins were significantly increased in leaves of pap1-D when CsANR2 was over-expressed, and the over-expression of CsF3'5'H alone or pap1-D did not increase proanthocyanidin accumulation, but co-expression of CsF3'5'H and CsANR2 significantly increased proanthocyanidins level in leaves of the pap1-D line.

Analysis of proanthocyanidins in mature seeds of the transgenic Arabidopsis
Owing to the increase of proanthocyanidins in 30-day-old leaves of the transgenic Arabidopsis, we further detected the accumulation of proanthocyanidins in mature seeds.
The level of soluble proanthocyanidins in CsF3'5'H × CsANR2 over-expression lines was slightly increased (1.14-fold and 1.32-fold with DMACA method, 1.10-fold and 1.03-fold with butanol-HCl method, respectively) compared to the wild type. The level of proanthocyanidins were slightly increased by the over-expression of CsF3'5'H using butanol-HCl method (1.70-fold) in pap1-D, while other lines stayed the same level compared to pap1-D that was 1.36-fold of the wild type. In addition, the level of soluble proanthocyanidins was substantially increased in CsF3'5'H × CsANR2 × pap1-D over-expression lines, the DMACA-reactive proanthocyanidins were 4.31-fold higher than the wild type, and the butanol-HCl-reactive proanthocyanidins was 2.75-fold higher than the wild type (Fig 5A).
We further detected the level of insoluble proanthocyanidins and found that the accumulation of insoluble proanthocyanidins was slightly increased in CsF3'5'H over-expression lines and CsANR2 over-expression lines (1.06-fold and 1.26-fold, respectively) than in the wild type. The level of insoluble proanthocyanidins in CsF3'5'H × pap1-D over-expression lines did not show obvious change as in pap1-D, and the CsANR2 × pap1-D over-expression lines had nearly 30% increase in insoluble proanthocyanidins. The accumulation of insoluble proanthocyanidins was significantly increased in CsF3'5'H × CsANR2 × pap1-D over-expression lines, which was approximate 85% increase compared to the wild type and nearly 50% increase compared to the pap1-D line (Fig 5B).
In summary, in mature seeds, the over-expression of CsF3'5'H or CsANR2 has no strong effect on the level of soluble proanthocyanidins, the level of insoluble proanthocyanidins could be up-regulated by CsANR2. The co-expression of CsF3'5'H and CsANR2 in pap1-D led to a huge increase in both soluble and insoluble proanthocyanidins levels.

Analysis of total flavonoids and flavonols in transgenic Arabidopsis
To exam whether total flavonoids of the transgenic lines were affected, leaves of all the transgenic lines and the wild type were collected for analysis. The level of total flavonoids in all six transgenic lines did not show significant changes compared to the wild type (Fig 6A).
Flavonols of all six transgenic lines and the wild type were detected on UPLC-MS (Fig 7). The level of total flavonols did not show any obvious changes in all six transgenic lines ( Fig  6B), but the composition of flavonols was obviously changed in all over-expression lines when PAP1 was over-expressed. As revealed in Fig 7C,

Analyses of transcript level of key genes in the transgenic Arabidopsis
PAP1 has been demonstrated as a key transcription factor in flavonoid pathway and the transcript level of many pathway genes could be up-regulated by PAP1 [10]. We further investigated whether the expression level of other flavonoid pathway genes was affected by the overexpression of CsF3'5'H, CsANR2 and/or PAP1.

PLOS ONE
As the first gene of the flavonoid pathway, CHS, was strongly up-regulated when PAP1 was expressed. The transcript level of CHS in pap1-D was nearly 10-fold higher than of the wild type, and it was increased by approximately 7-fold in CsF3'5'H × pap1-D, CsANR2 × pap1-D and CsF3'5'H × CsANR2 × pap1-D over-expression lines. The transcript level of CHS in the CsF3'5'H and CsANR2 over-expression lines did not show significant difference with the wild type (Fig 8A).
DFR and ANS are critical genes for anthocyanins biosynthesis. The expression level of DFR was increased almost 200-fold in pap1-D compared to the wild type, as the transcript level of DFR in CsF3'5'H × pap1-D, CsANR2 × pap1-D and CsF3'5'H × CsANR2 × pap1-D over-expression lines was approximately increased 150-fold, which is mainly due to the presence of PAP1. The transcripts of DFR in CsF3'5'H over-expression line or in CsANR2 over-expression line did not show obvious difference from the wild type. The transcript level of ANS were elevated more than 100-fold in all 4 lines due to the over-expression of PAP1, especially more than 170-fold in CsF3'5'H × pap1-D over-expression line. In CsF3'5'H over-expression line and CsANR2 over-expression line, the transcript level of ANS did increase significantly (1.36-fold and 1.78-fold, respectively, Fig 8B and 8C).
The transcript level of FLS is directly related to the level of flavonols in Arabidopsis. As shown in Fig 8D, the transcript level of FLS in all the transgenic lines was almost unchanged compared to the wild type, which was positively correlated to the total flavonols level in all transgenic lines.
ANR is one of the most important genes in the biosynthesis of proanthocyanidins. The transcript level of endogenous ANR gene in CsANR2, CsANR2 × pap1-D and CsF3'5'H × CsANR2 × pap1-D over-expression lines was slightly increased (2.08-fold, 2.32-fold and 4.18-fold, respectively) compared to the wild type, but the transcripts of endogenous ANR in CsF3'5'H, pap1-D, and CsF3'5'H ×pap1-D over-expression lines did not show significant change (Fig 8E).
These data indicated that the transcript level of endogenous CHS, DFR and ANS was strongly up-regulated by PAP1, which is correlated with anthocyanin levels (Fig 3D). The transcript level of endogenous ANR slightly increased while CsANR2 existed, but the expression of FLS was not affected by CsF3'5'H, CsANR2 or PAP1.

Discussion
The redirection of metabolic flux to proanthocyanidin pathway Anthocyanidins could be catalyzed by ANR to form proanthocyanidin monomers; presumably, in the presence of ANR, more proanthocyanidins will be produced when there is massive anthocyanidins accumulated, but the competition between the biosynthesis of proanthocyanidins and anthocyanins could be a block for the engineering of proanthocyanidins in plant as they share the same early biosynthetic pathway (Fig 1) [17,24]. Anthocyanidins could be reduced to form proanthocyanidin monomers, meanwhile, they could also be modified to form anthocyanins as conjugates. The ANR mutant ban in Arabidopsis accumulated more anthocyanins compared to the wild type, and the mutant of anthocyanidin UDP-glucosyltransferase gene UGT78D2 accumulated more proanthocyanidins [25].
In our study, total flavonoids level did not change in all transgenic lines compared to the wild type (Fig 6A), but the levels of two main final products-anthocyanins and proanthocyanidins obviously changed. Anthocyanins level in pap1-D was much higher than in the wild type, but decreased by the introduction of CsF3'5'H or CsANR2 (Fig 3), which were mainly ascribed to the heterologous expression of CsF3'5'H and CsANR2 since the expression of endogenous AtANS and AtANR were not significantly affected (Fig 8C and 8E). Therefore, by the over-expression of CsF3'5'H and CsANR2, proanthocyanidins level in leaves and mature seeds were increased at various degree (Figs 4 and 5).

PLOS ONE
It should also be pointed out that though CsF3'5'H and CsANR2 were both driven by the 35S promoter, proanthocyanidins were affected distinctly in different parts. Proanthocyanidins level was increased in leaves of the CsANR2 × pap1-D over-expression lines (Fig 4), levels of butanol-HCl reactive soluble proanthocyanidins in mature seeds of the CsANR2 × pap1-D over-expression lines were not changed and those of the CsF3'5'H × pap1-D over-expression lines were increased (Fig 5A). These results indicated that, total flavonoids level could not be affected by the over-expression of CsF3'5'H or CsANR2, but their components were changed by redirecting the metabolic flux from anthocyanins to proanthocyanidins. Therefore, it is possible that the ectopic expression of the specific genes could redirect the metabolic flux in Arabidopsis.

The synergistic effects of CsANR2 and CsF3'5'H in Arabidopsis
It is well known that massive proanthocyanidins were accumulated in tea leaves but none in Arabidopsis leave [14,18], the different accumulation pattern between these two plants was interesting for investigation. The functions of CsANR2 and CsF3'5'H were previously identified as two structural genes in flavonoids biosynthesis in tea plant [19,26]. It is reported that the presence of CsANRs in enzyme preparations could catalyze cyanidins and delphinidin into epicatechin and epigallocatechin in tea plants [26], the ectopic over-expression of CsANR2 in tobacco leaves caused the reduction of anthocyanins and the increase of proanthocyanidins. Meanwhile, by the over-expression of CsANR2, the level of soluble proanthocyanidins were up-regulated in the hairy roots of Medicago truncatula alone, with the decrease of anthocyanins [20]. Heterologous expression of CsF3'5'H in yeast showed that naringenin was converted into both 3'4'-and 3'4'5'-forms, the over-expression of CsF3'5'H in tobaccos could produce intensively purple flowers by increasing the content of anthocyanins [19].
However, previous work was mainly focused on the functions of these genes alone but not their synergistic effects, therefore, we further investigated the effect of their co-expression. In our study, we found that the synergistic effect of CsF3'5'H and CsANR2 appeared to be very strong by co-expression in Arabidopsis, the effects on both the decreasing of anthocyanins and the increasing of proanthocyanidins were more significant compared to the expression of individual CsF3'5'H or CsANR2, it might be interesting to further investigate the in-depth mechanism of their synergy in future work.
In spite of the obvious effects of CsF3'5'H and CsANR2 on anthocyanins and proanthocyanidins, it is unfortunately that we did not detect trihydroxylated flavonoids even by mass spectrum (Fig 7), which may be due to the transit conversion of trihydroxylated intermediates to other products, or they were all accumulated in form of possible insoluble proanthocyanidin (trihydroxylated monomers) that is hard for detection. Nevertheless, the heterologous expression will still be a big bottleneck for the genes to be fully functional in plant.

Engineering of proanthocyanidins in plant leaves
The co-expression of PAP1, TT2 and maize TF Lc resulted in detectable proanthocyanidins in the leaves of Arabidopsis, but the plant could only survive no more than 2 weeks after germination [27]. In tobacco leaves, the level of anthocyanins in PAP1 and MtANR co-expression lines was decreased compared to the PAP1 expression lines but higher than wild type, and there are detectable proanthocyanidins in both leaves and corollas [11]. In our study, the expression of CsF3'5'H, CsANR2 or PAP1 alone did not show a significant effects on proanthocyanidin levels, but the co-expression of PAP1 and CsANR2 in Arabidopsis led to a massive accumulation of DMACA-reactive proanthocyanidins in leaves (Fig 4A), and the plants grew normally and generate progeny. As the PAP1 and CsANR2 co-expressing lines accumulated DMACA-reactive proanthocyanidins in leaves, we did not find any accumulation of DMACA-reactive proanthocyanidins in CsF3'5'H × CsANR2 × pap1-D co-expressing lines (Fig 4A). It should be pointed out that the content of anthocyanidins in leaves may obstruct the measurements of butanol-HCl reactions and weakly react with DMACA.
The manipulation of proanthocyanidins in Arabidopsis by co-expression of CsF3'5'H, PAP1 and CsANR2 caused changes at different levels and in different tissues, especially in aerial parts of Arabidopsis. Although many genes, biosynthetic pathway genes or regulator genes were used to increase proanthocyanidins content in aerial parts of plants, in particular legume crops like alfalfa and white clover, but the results remain still unsatisfactory. Therefore, our study could offer a novel solution for the improvement of proanthocyanidins in a heterologous plant species. Nevertheless, more studies are required to discover more efficient genes for high proanthocyanidins engineering in forage crops.