Differential response of the biosynthesis of resveratrols and flavonoids to UV-C irradiation in grape leaves

Abstract

Resveratrols and flavonoids have important beneficial roles in plants and to humans. The grape family (Vitaceae) is one of several families that can biosynthesise both resveratrols and flavonoids. Previous research has shown that UV-C may induce the biosynthesis of resveratrols in the fruit of grape. But little is known about the UV-C-induced biosynthesis of resveratrols and flavonoids in grape leaves. This study demonstrated that the biosynthesis of resveratrol in grape leaves strongly increased in response to UV-C irradiation. The largest contribution to total resveratrols was trans-resveratrol. The expression of related genes, including PAL, C4H, 4CL and STS, increased and peaked 6–12 h after treatment, earlier than the peak of total resveratrol. In contrast, total flavonoid content and the expression of the key gene CHS was not affected by UV-C radiation, staying at a low level after UV-C irradiation. In summary, there is a differential response to UV-C radiation for the biosynthesis of stilbenes and flavonoids in grape leaves.

Keywords:

• flavonoids
• grape
• leaves
• resveratrols
• UV-C

Introduction

Phenylpropanoid metabolism is comprised of a complex series of branching biochemical reactions that provide the plant with a host of important phenolic compounds. The general phenylpropanoid pathway leads from phenylalanine to coumaroyl-CoA and is catalysed by the enzymes phenylalanine ammonia-lyase (PAL), cinnamate-4-hydroxylase (C4H) and coumaroyl-CoA ligase (4CL). From coumaroyl-CoA, there are two branches, one to flavonoids, whose first committed step is conducted by chalcone synthase (CHS). The other branch of the pathway is to trans-resveratrol (trans-res), which is catalysed by stilbene synthase (STS). Flavonoids include anthocyanin, flavone, flavonol, isoflavone and proanthocyanidin. Trans-res may be isomerised to cis-resveratrol (cis-res) or transformed to trans-piceid (trans-pd) and cis-piceid (cis-pd) by resveratrol 3-O-β-glycosyltransferases (O-3-GT) (Donnez et al. 2009). Trans-res and its derivatives form stilbenes.

Almost all plants can synthesise flavonoids; however, only a few plants can synthesise stilbenes. In stilbene-synthesising plants, STS competes with CHS in the synthesis of resveratrol (Lanz et al. 1991). Due to their antioxidant activity, flavonoids and stilbenes from plants are used as food additives, as dietary products and for medicinal use. One of the most relevant and extensively studied stilbenes is trans-res, which has been reported to have a number of beneficial health properties, such as high antioxidant capacity, cardiovascular protective effects, antimutagenic properties, as well as oestrogenic and cancer chemopreventive activities (Jang et al. 1997; Lu & Serrero 1999; Gusman et al. 2001). In addition to trans-res, trans-pd also shows high inhibitory activity against tumours and metastasised carcinoma (Kimura & Okuda 2000). The grape family Vitaceae is one of several families that can biosynthesise both resveratrols and flavonoids. In grapevines, accumulation of stilbenes can be induced by stresses such as fungal attack (Langcake 1981) and wounding (Langcake & Pryce 1976).

UV-B radiation (280–320 nm), which filters through the ozone layer, makes up only a small amount of the radiation that reaches the Earth's surface. UV-C radiation (200–280 nm), which is the most hazardous type of UV light, is physiologically insignificant, because these wavelengths are almost completely absorbed by the atmosphere (Coohill 1989; Frohnmeyer & Staiger 2003). Moderate levels of supplemental UV-B can stimulate transcription of genes involved in protective responses (Brosché & Strid 2003), and relatively high levels of solar UV-B can enhance the accumulation of UV-absorbing compounds such as stilbenes (Berli et al. 2008). These compounds are thus postulated to reduce UV-B transmittance and protect the photosynthetic apparatus in the leaf mesophyll (Burger & Edwards 1996; Mazza et al. 2000). However, little is known about the response of flavonoid biosynthesis to UV-C. Grapevines accumulate stilbenes in response to UV-C irradiation (Douillet-Breuil et al. 1999; Adrian et al. 2000). The influence of UV-C depends on the dose, the sensitivity of the plant species and its ability to attenuate the irradiation (Lavola et al. 2003). Grape leaves are an important possible source of stilbenes and flavonoids. Liu et al. (2013) reported that grape leaves of about 75 cultivars (or species) have stilbenes. It would be very useful to produce stilbenes and flavonoids from full grape leaves, by either physical or chemical methods. In the past, studies of UV-C have mainly focused on the fruit of grape. The response of stilbene and flavonoid biosynthesis in grape leaves to UV-C irradiation is not well understood. In this study, the response of biosynthesis of stilbenes and flavonoids to UV-C irradiation in grape leaves was investigated.

Materials and methods

Plant material and treatment

Vines of Vitis vinifera ‘Hongbaladuo’ (table grape) were used in our experiments and were grown in the vineyard at the Institute of Botany, Chinese Academy of Sciences, Beijing. As in Wang et al. (2013), mature (30-day-old), healthy leaves of similar size were detached from the shoot between 0800 h and 0900 h; the leaf petioles were immediately inserted into water, recut under water into 6 cm pieces and rapidly transferred from the water to triangular flasks containing ddH₂O. All leaves were incubated in the dark at 25 °C for 30 min, and then the leaf abaxial surfaces were exposed to 10 min of 6 W m⁻² UV-C irradiation from a UV-C lamp (Model ZW30S26W, Beijing Lighting Research Institute, Beijing, China). The distance between the leaf blade and the UV-C lamp was 40 cm. The leaves remained in the flasks until sampling. Control leaves were not irradiated. Samples were collected at 0, 3, 6, 12, 24 and 48 h after the initiation of treatments. All treatments and controls were replicated three times, and each replication consisted of six leaves. After sampling, the leaves were ground into powder in liquid nitrogen and stored at –80 °C until analysis.

Extraction and determination of total resveratrol

The resveratrol in leaves was extracted as in Liu et al. (2013). Briefly, 1 g of leaf tissue was extracted with 15 mL of methanol/ethyl acetate (1:1, v/v) for 24 h at room temperature in the dark. After centrifugation at 10,000 g at 4 °C for 10 min, the supernatant was evaporated at 40 °C until dry and then dissolved in 2 mL methanol. The extract was filtered through a 0.45 μm PTFE membrane before HPLC analysis.

All samples were analysed using a Dionex P680 HPLC system (Dionex Corporation, Sunnyvale, CA, USA) equipped with a reverse-phase C18 column of Atlantis® T3 (5 µm particle sizes, 4.6 × 250 mm ID; Waters Corporation, Milford, MA, USA) and a C18 Nova Pack guard pre-column (Waters). The injection volume was 10 μL and the column temperature was 30 °C. Cis-isomers were detected at 288 nm and trans-isomers at 306 nm, and photodiode array spectra were recorded from 240 to 600 nm. Separation was performed at a flow rate of 1.0 mL min⁻¹ with the mobile phase consisting of acetonitrile (A) and ddH₂O (B). The solvent gradient was as follows: 0 min 10% solvent A; 5 min 17% solvent A; 12 min 18% solvent A; 22 min 22% solvent A; 30 min 33% solvent A; 45 min 38% solvent A; and 58 min 100% solvent A. The fluorimetric detection for cis-isomers was at 288 nm, while that for trans-isomers was at 306 nm. The maximum excitation wavelength was 240 nm and emission was measured at 600 nm.

Extraction and determination of total flavonoids

Flavonoids in leaves were extracted as in Chen et al. (2012). Briefly, 1 g of fresh weight frozen leaves was extracted with 30 mL of methanol/water (70:30, v/v) at 4 °C for 36 h in the dark. The extract was centrifuged at 20,000 g for 10 min, the supernatant was collected, and the tissues were re-extracted as above two additional times. The combined supernatant solutions were evaporated until dry in a rotary vacuum at 35 °C and then reconstituted in 2 mL of 100% methanol. After centrifugation at 10,000 g for 10 min, 50 μL of the supernatant was diluted to 500 μL for further analysis. All samples were analysed using the ultraviolet visible spectrophotometer (UV-VIS) (SPECORD 200, Analytik Jena, Jena, Thuringia, Germany) at 350 nm.

Total RNA isolation and RT-PCR analysis

Total RNA was extracted from leaves using Trizol reagent (Invitrogen, Carlsbad, CA, USA) according to the manufacturer's instructions, and was treated with DNase I (Promega Corporation, Fitchburg, WI, USA) to avoid DNA contamination. The quantity and quality of total RNA was determined by spectrophotometry and by 1% formaldehyde denaturing gel electrophoresis. The samples with bright bands of ribosomal 28S to 18S RNA in a ratio >1.5:1 were used for RT-PCR analysis. 1 μg of RNA was reverse transcribed using Superscript II reverse transcriptase (Invitrogen) with an oligo(dT)15 primer according to the manufacturer's instructions (Tiangen Biotech, Beijing, China). qRT-PCR experiments were conducted using Real Master Mix (SYBR Green) (Tiangen Biotech). Reactions were performed on a StepOnePlus Real-Time PCR system (Life Technologies Corporation, Carlsbad, CA, USA). The following standard thermal profile was used for all PCR experiments: −94 °C for 5 min; 40 cycles of 95 °C for 15 s; and 60 °C for 60 s. Fluorescence signals were captured at the end of each cycle, and a melting curve analysis was performed from 68 to 95 °C to check the specificity of the PCR reaction. Gene-specific primers were designed using the Primer 5 software (Table 1). The enzyme STS is pivotal to the biosynthesis of resveratrol. Analyses of the preliminary grape genome sequence confirmed the large size of the multigene families with an estimated number of 48 STS genes (Vannozzi et al. 2012). The primers used here matched 13 STS mRNA sequences. They included VvSTS7, VvSTS9, VvSTS10, VvSTS15, VvSTS21, VvSTS27, VvSTS29, VvSTS33, VvSTS37, VvSTS41, VvSTS45, VvSTS47 and VvSTS48. This may reflect the expression of the STS family. PAL, CHS and O-3-GT are also multi-gene families, and have 10, three and three members, respectively, based on PN40024 12X V1 coverage (Table 2). The primers used for qRT-PCR in this study matched VvPAL1-10, VvCHS1-2 and VvO-3-GT1-2.

Table 1 Primers used in RT-PCR.

Genes	Forward primer	Reverse primer	Tm (°C)
Actin 1	CTTGCATCCCTCAGCACCTT	TCCTGTGGACAATGGATGGA	60
PAL	CAACCAAGATGTGAACTCCTT	TTCTCCTCCAAATGCCTC	60
C4H	GGCAAGCACAAAGAGCACAGAT	TTCTTCTGGATGTGAGGGTGGTT	60
4CL	CGAAGAACCCGATGGTGGAGA	CACGAGCCGGACTTAGTAGGA	60
STS	TTAGAAACGCTCAACGTGCCAAGGG	AATCAGCATAATCAGACTGGTAGAC	60
3-O-GT	GGCTTCAAAGGGCTTGCTTGTG	GGCGTTGGTTGGTCAGTGAT	60
CHS	GTCTGAAGGAAGAGAAACTGAGAG	CCAGGATAAACAACACGCAT	60

Table 2 PAL, C4H, 4CL, CHS members identified based on the grapevine PN 40024 12X V1 coverage.

Proposed nomenclature	Chr	PN40024 12X V1 Location	Genoscope annotation	ORF predicted
VvPAL1	16	649922 … 652537	GSVIVT01024299001	Yes
VvPAL2	16	706775 … 713137	GSVIVT01024306001	Yes
VvPAL3	16	688556 … 691178	GSVIVT01024303001	Yes
VvPAL4	16	699979 … 702603	GSVIVT01024305001	Yes
VvPAL5	16	777034 … 779523	GSVIVT01024315001	Yes
VvPAL6	16	607611 … 610206	GSVIVT01024293001	Yes
VvPAL7	16	595690 … 601170	GSVIVT01024292001	Yes
VvPAL8	16	619737 … 622170	GSVIVT01024295001	Yes
VvPAL9	16	613592 … 616135	GSVIVT01024294001	Yes
VvPAL10	16	724490 … 724106		No
VvC4H	6	8882429 … 8886021	GSVIVT01024554001	Yes
Vv4CL	16	2123299 … 2128332	GSVIVT01024400001	Yes
VvCHS1	14	24673432 … 24675024	GSVIVT01032968001	Yes
VvCHS2	14	24687028 … 24688331		Unsure
VvCHS3	5	13597404 … 13599563	GSVIVT01000521001	Unsure
VvO-3-GT1	3	6124465 … 6126180	GSVIVT01003157001	Yes
VvO-3-GT2	3	6282001 … 6283513	GSVIVT01003143001	Yes
VvO-3-GT3	3	6338114 … 6339894	GSVIVT01037921001	Yes

The predicted open reading frame (ORF) prediction was assigned as follows: Yes, gene encoding a complete ORF; Unsure, it was not possible to determine with certainty whether the gene encode was for a full length or truncated ORF based on available sequence information from the PN40024 genotypes; No, gene encoding for a truncated ORF.

The amplification curves were analysed with the LinRegPCR software and the amplification efficiency of the primers was at 90%–110%. The actin gene was used as the internal control for all of the qRT-PCR data. Analyses of qRT-PCR data used the classic (1 + E)^−ΔΔCT method (C_T is the threshold cycle of one gene, E is the amplification efficiency). ΔC_T is equal to the difference in threshold cycles for target (X) and reference (R) (C_T,_X–C_T,R), while the ΔΔC_T is equal to the difference of ΔC_T for the control (C) and the treatment (T) (ΔC_T,T–ΔC_T,C). The amplification system (e.g. primer and template concentrations) was properly optimised, and the efficiency was close to 1. Therefore, the amount of the target product, scaled to an endogenous reference and relative to a calibrator, was given by the equation:

Statistical analysis

Statistical analyses were performed using SPSS 14.0 and were tested for statistical significance using the paired Student's t-test. Differences were considered significant at P < 0.05.

Results

Accumulation of resveratrol and flavonoids in leaves after UV-C irradiation

Our results showed that trans-pd, cis-pd and trans-res were detectable, but cis-res was undetectable. Total resveratrol (sum of the above three forms) in the control leaves showed no significant (P < 0.05) change and were at trace levels (0–10 μg g⁻¹). In contrast, the total amount in UV-C-irradiated leaves changed significantly. Before treatment, the concentration of total resveratrol was very low (0.97 μg g⁻¹ FW). At 3 h after treatment, the total amount of resveratrol began to increase, and accumulated slowly until 6 h post-irradiation. Thereafter, total resveratrol content sharply increased and peaked (18.18 μg g⁻¹ at 6 h post-irradiation and 467.62 μg g⁻¹ at 24 h post-irradiation). Total resveratrol at the peak was 480.8 times higher than levels in control leaves. Levels then declined to 199.55 μg g⁻¹ by 48 h post-irradiation. The contribution of the three forms of resveratrol to the total amount differed significantly. Trans-res made the greatest contribution and showed a similar trend as total resveratrol. Concentration of trans-pd continually rose, reaching 18.72 μg g⁻¹ FW at 48 h after UV-C irradiation. Concentration of cis-pd always maintained lower levels after UV-C irradiation (Table 3). In contrast to the strong response of resveratrol to UV-C, biosynthesis of flavonoids did not respond significantly in the 48 h after UV-C irradiation (Table 3).

Table 3 The content of resveratrols and flavonoids in grape leaves treated by UV-C irradiation.

		Time after treatment (h)
	0	3	6	12	24	48
Total res (µg/g FW)	Treatment	0.97 ± 0.17aD	1.21 ± 0.37 aD	18.18 ± 1.58 aD	130.90 ± 10.00aC	467.62 ± 15.95aA	199.55 ± 24.22aB
	Control	0.97 ± 0.17 aC	1.19 ± 0.50 aBC	2.06 ± 0.32bBC	6.41 ± 1.25bAB	9.11 ± 2.02 bA	5.96 ± 1.75 bABC
Trans-res (µg/g FW)	Treatment	0.13 ± 0.08aD	0.75 ± 0.18aD	16.90 ± 1.42aD	127.30 ± 10.01aC	456.03 ± 16.21aA	178.65 ± 23.89aB
	Control	0.13 ± 0.08aC	0.32 ± 0.32aBC	1.13 ± 0.31bBC	2.97 ± 1.00bAB	4.15 ± 1.04bA	1.17 ± 0.36bBC
Trans-pd (µg/g FW)	Treatment	0.08 ± 0.08aC	0.11 ± 0.07aC	0.58 ± 0.07aC	3.06 ± 0.06aC	10.56 ± 0.62aB	18.72 ± 2.99aA
	Control	0.08 ± 0.08aA	0.08 ± 0.08aA	0.07 ± 0.07bA	0.41 ± 0.19bA	0.51 ± 0.13bA	0.61 ± 0.15bA
Cis-pd (µg/g FW)	Treatment	0.76 ± 0.10aA	0.35 ± 0.21aA	0.69 ± 0.19aA	0.53 ± 0.07aA	0.64 ± 0.14aA	0.38 ± 0.10aA
	Control	0.76 ± 0.10aA	0.78 ± 0.27aA	0.66 ± 0.09aA	0.63 ± 0.09aA	0.64 ± 0.09aA	0.25 ± 0.05aA
Flavonoids (µg/g FW)	Treatment	2.77 ± 0.07aA	2.85 ± 0.02aA	2.63 ± 0.20aA	2.65 ± 0.07aA	2.80 ± 0.03aA	2.66 ± 0.10aA
	Control	2.77 ± 0.07aA	2.85 ± 0.00aA	2.61 ± 0.01aA	2.42 ± 0.03aA	2.68 ± 0.04aA	2.45 ± 0.01aA

Means followed by the same lower case letters between the treatment and the control for each compound at different times after UV-C treatment indicates that there is no significant difference between them (P < 0.05). Means followed by the same upper case letters at different times after UV-C treatment for each compound indicates that there is no significant difference between them (P < 0.05).

Expression of the general phenylpropanoid pathway genes, STS and CHS

In order to analyse the relationship between total resveratrol and total flavonoid accumulation in the leaves after treatment and the expression of related biosynthetic genes, we performed qRT-PCR analyses of PAL, C4H, 4CL, CHS, STS and O-3-GT (Fig. 1). After UV-C irradiation, PAL and 4CL exhibited a similar trend. Their expression first increased, peaked and then declined. The expression of PAL and 4CL peaked 6 h after UV-C irradiation, but the expression of C4H peaked 12 h after UV-C irradiation. The expression of PAL reached a higher peak compared with C4H and 4CL. We designed the primers in the conserved region of 20 STS mRNA sequences, which may reflect the expression of the STS family. STS expression was low in the control leaves (Fig. 1). At 3 h after UV-C irradiation, the expression of STS slowly increased, then rose sharply, and peaked 12 h after UV-C irradiation. The relative expression at its peak was more than 5000 times that before irradiation. After the peak, the expression continued decreasing, and declined to a low level by 48 h. In contrast, the relative expression of CHS showed no significant difference between the leaves of the treatment and the control (Fig. 1). The expression of O-3-GT gradually rose after UV-C irradiation, plateaued from 12 to 24 h, and then declined.

Figure 1 The expression of genes related to resveratrol and flavonoid biosynthesis to UV-C irradiation.

Discussion and conclusion

UV-C induced the accumulation of resveratrol and the expression of related genes in leaves

Previous studies showed that salicylic acid, methyl jasmonate (JAMe), cyclodextrin, mechanical damage and plant diseases can promote the accumulation of resveratrol (Bavaresco et al. 1997; Krisa et al. 1999; Vitrac et al. 2002; Bru et al. 2006; Vezzulli et al. 2007; Kiselev et al. 2010). Moreover, some studies have investigated transcription levels of resveratrol synthesis-related genes. Application of JAMe resulted in an increase in the expression of PAL, C4H, 4CL and STS in calluses of berries and leaf petioles (Tassoni et al. 2005; Belhadj et al. 2008; Lijavetzky et al. 2008). In addition, the expression of PAL and STS was induced by salicylic acid in a cell suspension of grape leaves (Kiselev et al. 2010). Previous studies of the synthesis of UV-C-induced resveratrol were focused on fruits of grapes at different developmental stages and from different cultivars (Adrian et al. 2000; Cantos et al. 2002, 2003). Grape leaves have a relatively remarkably high content of resveratrol (Petit et al. 2009; Wang et al. 2010). However, less attention has been paid to the influence of abiotic and biotic factors on the biosynthesis of resveratrol in leaves. This study showed that irradiation by UV-C induced an increase in resveratrol content in grape leaves, and the expression of PAL, C4H, 4CL and STS was increased. The expression of STS sharply increased, with a peak value at about 5000-fold of the control. Trans-res is the product of the STS enzyme. The start and peak times of trans-res accumulation were later than those of STS expression (Fig. 1 and Table 3). Resveratrol O-glucosyltransferase (O-3-GT) may convert trans-res into trans-pd (Jeandet et al. 1997; Belhadj et al. 2008). From 3 to 24 h after UV-C treatment, trans-pd content and the expression of O-3-GT constantly increased, which indicates that O-3-GT has an important role for trans-pd accumulation. Leaf tissue may make a better study system than fruits for studying the mechanism of synthesis of resveratrol. Leaves are easily treated with UV-C, and using leaves excludes the possibility of transportation of resveratrol between the skin and pulp in fruit when studying the accumulation of resveratrol.

UV-C did not induce the synthesis of flavonoids and the expression of related genes in leaves

Flavonoids and resveratrol are two branches of the phenylpropanoid pathway, which involves the enzymes PAL, C4H and 4CL. STS and CHS use the same substrate. Therefore, there should be some relationship between the synthesis of resveratrol and flavonoids. UV-B has also been found to increase flavonoids in apple skins (Solovchenko & Schmitz-Eiberger 2003) and in fruits of grape (Schultz 2000; Cantos et al. 2001; Kolb et al. 2003). Some key enzymes involved in the phenylpropanoid (PAL) and flavonoid (CHS) biosynthetic pathways are up-regulated by UV-B (Pontin et al. 2010). Moreover, UV-B induced an increase in resveratrol content in grape skin (Cantos et al. 2001; Berli et al. 2008). Similar to effects seen from UV-B, UV-C induced the synthesis of resveratrol in the present study as well as in other studies. Different from effects from UV-B irradiation, UV-C irradiation did not result in a significant change in total flavonoid content and the expression of CHS in this study (Fig. 1 and Table 3). Zhang et al. (2012) reported that 1.8 kJ m⁻² UV-C exposure slightly increased the level of various anthocyanins in the skins of the 11 w grape berries. Erkan et al. (2008) reported that doses of 0.43 and 2.15 kJ m⁻² UV-C irradiation had little effect on anthocyanin accumulation, and 4.30 kJ m⁻² showed markedly reduced concentrations in strawberry fruit. The effects of UV-C irradiation on anthocyanins were found to depend on dosage and cultivar (Perkins-Veazie et al. 2008). Li et al. (2014) reported that 4.1 kJ m⁻² UV-C irradiation suppressed anthocyanin and flavonoid accumulation in strawberry plants. The effects of UV-C irradiation on flavonoids were found to depend on dosage and cultivar (Perkins-Veazie et al. 2008).

A dramatic increase in the expression and activity of PAL after UV-C irradiation was also reported by Pombo et al. (2011). This increase was found to be beneficial to increasing total phenolic content. A study by Li et al. (2014) showed that 4CL activity was inhibited in strawberry plants by UV-C radiation, and the effects of UV-C radiation on C4H were not obvious. In the current study, PAL gene expression was strongly up-regulated after UV-C treatment, whereas C4H and 4CL gene expression was only relatively weakly up-regulated. This should provide enough substrates for enzymes of STS and CHS. High STS expression is an important cause of resveratrol accumulation. However, relatively stable CHS expression is also an important cause of resveratrol accumulation; this is because of the decline in substrate competition. Vannozzi et al. (2012) found that UV-C resulted in a dramatic increase in STS transcription, but the expression of all three CHS genes was strongly suppressed relative to the untreated leaf discs. Our results support this result for the expression of CHS genes and total flavonoid content (Fig. 1 and Table 3). UV-B and UV-C only differ in wavelength, but their effects differ greatly. This may be associated with different signal transduction components between UV-B and UV-C, and is worthy of further investigation.

In summary, these results demonstrated that the biosynthesis of resveratrol in grape leaves strongly increased in response to UV-C irradiation. Total resveratrols peaked 24 h after UV-C treatment. The biggest contributor to total resveratrols was trans-resveratrol. The expression of related genes, including PAL, C4H, 4CL and STS, increased and reached a peak at 6–12 h after treatment, which is earlier than the peak of total resveratrol. In contrast, total flavonoids and the expression of CHS were not affected by UV-C irradiation. In conclusion, there is a differential biosynthetic response to UV-C irradiation for stilbenes and flavonoids in grape leaves.

Additional information

Funding

This work was supported by the National Natural Science Foundation of China (No. 31171918).