Extraction of Proanthocyanidins and Anthocyanins from Grape Skin by Using Ionic Liquids

Proanthocyanidins (condensed tannins) and anthocyanins are the main phenolic compounds of grape skin, belonging to the fl avonoid family. In recent years, these compounds have att racted great att ention since they are responsible for major grape and wine sensory properties like mouth-feel (1–3) and colour (4), but also for their protective eff ect against cardiovascular risk factors due to many biological activities (5). Structurally, grape skin proanthocyanidins are polymers consisting of (+)-catechin, (−)-epicatechin, (−)-epicatechin gallate and (−)-epigallocatechin units; highly more polymerized and less galloylated than seed proanthocyanidins (3,6). Anthocyanins of Vitis vinifera grape cultivars are primarily based on fi ve major anthocyanin monoglucosides, namely delphinidin, cyanidin, petunidin, peonidin and the most abundant mal vidin; with lower portions of their acetic, caff eic and p-coumaric acid esters (7,8).


Introduction
Proanthocyanidins (condensed tannins) and anthocyanins are the main phenolic compounds of grape skin, belonging to the fl avonoid family.In recent years, these compounds have att racted great att ention since they are responsible for major grape and wine sensory properties like mouth-feel (1-3) and colour (4), but also for their protective eff ect against cardiovascular risk factors due to many biological activities (5).Structurally, grape skin proanthocyanidins are polymers consisting of (+)-catechin, (−)-epicatechin, (−)-epicatechin gallate and (−)-epigallocatechin units; highly more polymerized and less galloylated than seed proanthocyanidins (3,6).Anthocyanins of Vitis vinifera grape cultivars are primarily based on fi ve major anthocyanin monoglucosides, namely delphinidin, cyanidin, petunidin, peonidin and the most abundant mal vidin; with lower portions of their acetic, caff eic and p-coumaric acid esters (7,8).
The conventional extraction of phenolic compounds most commonly implies the application of volatile, toxic and fl ammable organic solvents; and procedures oft en employing mixtures of solvents and multiple repetitions, which can be time-consuming (9).For instance, methanol, ethanol, acetone, water and their combinations are commonly used in extraction of proanthocyanidins from various plant materials (10).In addition, the two-step procedure using aqueous acetone followed by aqueous methanol represents the most common procedure used in the ex-ISSN 1330-9862 scientifi c note doi: 10.17113/ft b.55.03.17.5200 traction of overall grape proanthocyanidins (2,3,6), since methanol has been shown to have great extraction efficiency towards fl avan-3-ol monomers and oligomers (11), and acetone towards polymers.On the other hand, due to the stability of red fl avylium cation in highly acidic media (12), the extraction of anthocyanins is commonly performed with slightly acidifi ed organic solvents (10), namely methanol or ethanol, which are found to be more eff ective than water (13).
Recently, the application of ionic liquids (ILs) as greener alternatives to conventional organic solvents has been proposed (14,15).ILs are organic salts in the liquid state that consist of organic cations paired with organic or inorganic anions.The unique physicochemical properties of ILs that make them interesting from technological point of view are negligible vapour pressure (reduced air pollution), non-fl ammability (process safety), good stability (reusability) and possibility to design an optimal IL for a specifi c purpose (16,17).Properties like thermal stability and miscibility are mainly aff ected by the structure of the anion, while others, such as viscosity, surface tension and density depend on the length of the alkyl chain in the cation and/or its shape or symmetry (18,19).Reports on the ecotoxicologycal profi le of ILs have pointed out that toxicity of commonly used imidazolium and pyridinium ILs ranges from low to hazardous; however, environmental impact is strongly infl uenced by the IL structure, and as a result a great eff ort is being made to identify the factors that control IL behaviour.Also, the major benefi t of using ILs as solvents results from their low vapour pressure and non-fl ammability, which surely makes them preferable to volatile and fl ammable organic solvents (20).
Previous studies have demonstrated promising applicability of imidazolium-based ILs in the extraction of diff erent phenolic compounds such as phenolic acids (gallic, ellagic, chlorogenic and caff eic acids) (21,22), stilbenes (trans-resveratrol) (21,23), fl avonols (quercetin, my ri cetin, kaempferol and rutin) (21,22,24,25) and proanthocyanidins (26,27) from various plant materials.In addition, physicochemical properties of ILs, which can be moderated and dictated by their structural characteristics of anion and cation, were also found to signifi cantly aff ect extraction yields of target analytes (21,22,27).For instance, the solubility of phenols in ILs was shown to depend on the ability of phenols to form intra-and intermolecular bonds, as well as the polarity of ILs (28).However, we have not observed in these reports that authors studied the application of ILs in the extraction of fl avonoid compounds from the complex grape skin matrix, which requires further understanding.
The aim of this work is to study the potential of imidazolium-type ILs in the extraction of main grape skin fl avonoids (proanthocyanidins and anthocyanins) by compar ing them to conventional organic solvent extraction.The eff ect of IL anions and cations, and of their concentration on the extraction yield of grape skin fl avonoids was examined.

Grape skin samples
Grapes of the native Croatian red grape cultivar, Vitis vinifera cv.Plavac mali, originating from Dalmatia were harvested at their technological maturity in October 2012.The mass of 2 kg of randomly selected grapes was used for the study.Skins were immediately manually separated from the pulp, freeze-dried for three days at -40 °C and stored at -20 °C before the analysis.

Proanthocyanidin and anthocyanin extracts
Grape skin proanthocyanidins and anthocyanins were extracted using conventional organic solvents and ILs.The conventional extraction of proanthocyanidins was conducted according to the procedure described by Chira et al.The extraction of anthocyanins and proanthocyanidins with ILs was conducted by the modifi ed procedure of IL-based maceration extraction described by Liu et al. (24).A mass of 0.2 g of freeze-dried grape skin powder was weighed in an extraction tube (15 mL) and 5 mL of IL were added.The sample was vortexed for 10 s and placed in a shaker (Thermo Fisher Scientifi c, Waltham, MA, USA) for 4 h of extraction, at room temperature.Aft er extraction, the sample was centrifuged for 15 min at 5000×g (model KL2, Fischer Bioblock Scientifi c, Schwerte, Germany), while the obtained supernatant was fi ltered.Filtered extracts of IL were further directly manipulated.

Spectrophotometric analysis of total proanthocyanidins and anthocyanins
The content of total proanthocyanidins in skin extracts was determined by the method of proanthocyanidin interfl avan bond cleavage by acid hydrolysis (32).The content of total anthocyanins in skin extracts was determined by the bisulfi te bleaching procedure (33).All measurements were performed on spectrometer GENESYS™ 10S (Thermo Fisher Scientifi c, Madison WI, USA).

HPLC analysis of free anthocyanins
The HPLC analysis was conducted on a Varian Pro Star Solvent Delivery System 230 (Varian) with a Photodi-ode Array (PDA) detector Varian Pro Star 330 (Varian), and on the Agilent 1100 Series LC-MSD system (Agilent Technologies, Waldbronn, Germany) with a PDA detector and single quadrupole mass detector equipped with electrospray ionization interface (model G1946D), where the latt er was used for confi rmation of peak identifi cation.Separation was performed using a Phenomenex Nucleosil C18 column (250 mm×4.6 mm, 5 μm; Phenomenex, Torrance, CA, USA) with column temperature set at 45 °C.The mobile phases used were solvent A: water/formic acid (95:5, by volume), and solvent B: acetonitrile/formic acid (95:5, by volume) applied at a fl ow rate of 1 mL/min under the gradient conditions previously described by Lorrain et al. (31) with small modifi cations: 10-35 % B linear from 0-25 min, 35-100 % B linear from 25-26 min, 100 % B isocratic from 26-28 min, 100-10 % B linear from 28-29 min, with the re-equilibration of the column from 29-35 min under the initial gradient conditions.UV-Vis spectra were measured in the wavelength range from 200 to 600 nm.The detection was performed at λ=520 nm.The identifi cation and peak assignment of anthocyanins were based on the comparison of their retention times, UV-Vis and mass spectral data with those of the standards and published data (3,31).Quantitative determinations were performed using calibration curves of anthocynin-3-Omonoglucosides at the wavelength of maximum absorbance (520 nm) determined by PDA spectra.Results are converted to mg per g of skin dry mass.

Data analysis
All analyses were conducted in triplicate.Statistical data analysis was carried out with analysis of variance (ANOVA) using STATISTICA v. 7 soft ware (34).Tukey's honestly signifi cant diff erence test was used for the comparison when samples diff ered signifi cantly aft er ANO-VA (p<0.05) for the analysis of polyphenols.

Results and Discussion
Structural characteristics of ionic liquids (ILs) have a signifi cant eff ect on their physicochemical properties (19).Structure of anions and cations, as well as the concentration of ILs can greatly aff ect the extraction yields of phenolic compounds (21)(22)(23)(25)(26)(27)(28)35).In order to evaluate IL performance in the extraction of main grape skin fl avonoids (proanthocyanidins and anthocyanins), eight diff erent imidazolium-based ILs varying in concentration (Table 1) were tested.

Eff ect of ionic liquids on the extraction of grape skin proanthocyanidins
Signifi cant diff erences in the mass fraction of total proanthocyanidins obtained with selected ILs demonstrated great diversity in extraction effi ciency among the studied ILs (Table 2).This was particularly pronounced for [C 4 mim][Br] and [C 10 mim][Br], representing the most and the least eff ective IL, respectively; where the mass fraction obtained in the former was 95 % higher than in the latt er.It is interesting to note that this fi nding is in agreement with the studies of Liu et al. (26) and Yang et al. (27), where [C 4 mim][Br] was also selected as the optimal IL in the extraction of proanthocyanidins, while [C 10 mim] [Br] showed to be the least suitable one.
A series of 1-alkyl-3-methylimidazolium ILs with Br - and alkyl chains ranging from C 2 to C 10 showed a remarkable impact of the alkyl chain length of cations on the extraction yield of grape skin proanthocyanidins (Table 2).Similarly as previously observed by other authors (26,27), increasing the alkyl chain length from ethyl to butyl slightly enhanced the extraction effi ciency of proanthocyanidins, while on the contrary, further increase from butyl to decyl caused a drastic decrease.Moreover, this eff ect was also noticed in the extraction yield of other phenolic compounds from various plant matrices (21,22,35).It has been shown that the increase in the alkyl chain length of 1-alkyl-3-methylimidazolium cation decreases the surface tension and increases the viscosity and hydrophobicity of ILs (19).Therefore, it is possible that more hydrophilic ILs like [C 4 mim][Br] interact more strongly with proanthocy-anidins and thus extract the higher mass fractions of these compounds.However, among protic-based ILs (Table 2), extraction yields of total proanthocyanidins decreased in the order: [sC 4   The comparison of ILs with conventional proanthocyanidin extraction procedures including organic solvents was also performed (Table 2).It is important to note that signifi cantly higher mass fractions of total proanthocyanidins (p<0.05) were extracted with 1. ] than with aqueous methanol.However, the mass fraction of total proanthocyanidins extracted in the two-step conventional procedure using aqueous acetone followed by aqueous methanol was signifi cantly higher (p<0.05)than with methanol and all ILs.These results confi rmed the superiority of the two-step conventional procedure and the necessity to combine diff erent organic solvents in the extraction of proanthocyanidins.Nevertheless, further possibilities of replacing the two-step conventional procedure including organic solvents with the single-step extraction using ILs should be examined.

Eff ect of ionic liquids on the extraction of grape skin anthocyanins
Herein, the capacity of ILs to extract anthocyanins was shown to be structure-dependent, similarly as earlier estimated for proanthocyanidins.The evident eff ect of the alkyl chain length of imidazolium cation, as well as the eff ect of the IL anion structure on the extraction yield of grape skin anthocyanins was observed (Table 3).For instance, the elongation of cation alkyl chain from ethyl to decyl led to a signifi cant decrease in the mass fraction of extracted total anthocyanins.These results might be attributed to two important factors occurring with the increase of chain length: fi rst, the increase of hydrophobicity (19), causing poorer diff usion capacity of the solution; and second, the increase of pH value up to 9.00 inducing the degradation of anthocyanins (12).The lack of increase in the extraction effi ciency of total anthocyanins from C 2 to C 4 of 1-alkyl-3-methylimidazolium cation, contrary to observations previously reported by numerous authors on various phenolic compounds (21,22,26,27,35) and contrary to our results for total proanthocyanidins, is probably associated with higher pH values of [C 4 mim][Br] compared to [C 2 mim][Br].In fact, anthocyanins can be found in diff erent chemical forms depending on the pH of the solution (4,12).At pH=1 they are predominantly present in the form of red fl avylium cation, and at pH between 2 and 4 as blue quinoidal species.At pH between 5 and 6 colourless carbinol pseudobase and chalcone were observed, while at pH values higher than 7 anthocyanins are degraded.Knowing that the pH has great importance for anthocyanin equilibrium forms and stability, it was no surprise that among 1-alkyl-3-methylimidazolium ILs, [C 2 mim][Br] with the lowest pH value obtained the highest mass fraction of total anthocyanins.In addition, this IL showed the best extraction performance among all, confi rming the dominance of Br -towards anthocyanins, similarly as earlier indicated for proanthocyanidins.However, mass fractions of total anthocyanins obtained in [sC 4 mim] [HSO 4 ] and particularly in [mim][HSO 4 ] characterized by very low pH values were similar or only slightly lower than those in [C 2 mim][Br] but higher than the ones found in [C 4 mim][Br].Therefore, these two protic ILs containing [HSO 4 ] anion showed to be more suitable for the extraction of anthocyanins than proanthocyanidins.
The concentration of ILs was found to be a less important parameter aff ecting extraction yields of anthocyanins from grape skins than the structure of cation or anion.In general, the increase in the IL concentration range from 0.5 to 2.5 mol/L only slightly contributed to extrac-  tion yields of anthocyanins (Table 3).This eff ect was to a much lesser extent than shown in our results for proanthocyanidins.The less effi cient ILs, [mim][CF 3 CO 2 ] and especially [C 10 mim][Br], even showed a decrease in the function of concentration, the latt er possibly due to higher viscosity (19).Our results indicate that ILs can constitute a useful matrix to incorporate fl avylium compounds (36,37).Moreover, selected ILs could be used as an alternative for volatile organic solvents in extraction of anthocyanins from grape skin.-e), their sum (Fig. 1f), as well as the sum of their major acylated derivatives (acetylmonoglucosides and p-coumaroylmonoglucosides of peonidin and malvidin) (Figs.1g and h) were determined in selected extracts.Some trends observed among diff erent solvents concerning individual anthocyanin-3-O-monoglucosides (Fig. 1a-e) were very similar to the ones noticed earlier for total anthocyanins.For example, 2.  1a) and petunidin-3-O-monoglucoside (Fig. 1c) was much higher than that of peonidin-3-O-monoglucoside (Fig. 1d) and malvidin-3-O-monoglucoside (Fig. 1e) since the mass fractions of the fi rst two aforementioned free anthocyanin-3-O-monoglucosides extracted by 2.5 mol/L of [sC 4 mim][HSO 4 ] were signifi cantly higher (Figs.1a and c).Also, 2.5 mol/L of [C 2 mim][Br] and 2.5 mol/L of [mim][HSO 4 ] always yielded higher mass fractions, particularly of petunidin-3-O-monoglucoside (Fig. 1c) and the most abundant malvidin-3-O-monoglucoside (Fig. 1e), increasing it in the overall sum of anthocyanin-3-O--monoglucosides (Fig. 1f).Moreover, it is interesting to note that the selectivity of 2.5 mol/L of [mim][HSO 4 ] towards delphinidin-3-O-monoglucoside (Fig. 1a) was greater than that of 2.5 mol/L of [C 2 mim][Br], while the reverse was found for all other monoglucosides.Furthermore, the solution of meth anol showed bett er extraction effi ciency than 2.5 mol/L of [C 4 mim][Br], where signifi cantly higher mass fractions of delphinidin-3-O-monoglucoside, cyanidin-3 --O-monoglucoside and petunidin-3-O-monoglucoside (Figs.1a-c), and of the overall sum of anthocyanin-3-O-acetyl--monoglucosides (Fig. 1f) were found.The diff erences in selectivity were also observed between 2.5 mol/L of [sC 4 mim] [HSO 4 ] and the methanol solution, since the former extracted higher mass fractions of all determined monoglucosides, except malvidin-3-O-monoglucoside and the overall sum of anthocyanin-3-O-monoglucosides.Surprisingly, acylated derivatives of anthocyanins (Figs.1g and h) followed a very diff erent patt ern from the one observed in anthocyanin-3-O-monoglucosides.For instance, the extraction effi ciency of 2.5 mol/L of [mim] [HSO 4 ] and the acidifi ed aqueous solution of methanol was signifi cantly lower (p<0.05)than of other tested ILs.Despite the high effi ciency in the extraction of total anthocyanins and individual anthocyanin-3-O-monogluco-sides, 2.5 mol/L of [mim][HSO 4 ] showed signifi cantly lower yields than all solvents in the extraction of anthocyanin-3-O-acetylmonoglucosides (Fig. 1g).A possible explanation for these results may be a very high acidity of this solvent, which can cause partial hydrolysis of malvidin-3-O-acetylmonoglucoside during the extraction of grape anthocyanins and lead to changes in their relative content (38).More interestingly, all ILs demonstrated higher selectivity for p-coumaroylmonoglucosides than the conventional methanol solution.Interestingly, 2.5 mol/L of [sC 4 mim][HSO 4 ] that showed moderate extraction capacity of total anthocyanins was shown to be particularly effi cient in the extraction of two acetylmonoglucosides (Fig. 1g), and together with 2.5 mol/L of [C 2 mim][Br] in the extraction of two p-coumaroylmonoglucosides (Fig. 1h).

Conclusions
A series of imidazolium-based ionic liquids (ILs) tested in our study showed the diff erences in the extraction effi ciency of proanthocyanidins and anthocyanins from grape skin, depending on the anion and cation structure and concentration.The superiority of Br -and alkyl chain elongation of cation up to butyl and ethyl was established in the extraction yield of proanthocyanidins and anthocyanins, respectively, while the increase in their mass fraction aff ected the former more than the latt er.The highest mass fraction of proanthocyanidins was extracted with 2.5 mol/L of [C 4 mim][Br].Some ILs extracted higher concentrations of total proanthocyanidins than the methanol solution, but could not compete with the two-step conventional acetone-methanol extraction.On the other hand, 2.5 mol/L of [C 2 mim][Br] followed by 2.5 mol/L of [mim][HSO 4 ] extracted signifi cantly higher mass fractions of total and free anthocyanin-3-O-monoglucosides than the conventional acidifi ed methanol solution.Surprisingly, a diff erent patt ern was noticed for acylated derivatives, where affi nity towards anthocyanin-3-O-acetylmonoglucosides decreased in the order 2.
(3) using: (i) two-step extraction with solvent A: acetone/water (80:20 by volume) (fi rst step), followed by extraction with solvent B: methanol/water (60:40, by volume) (second step), where centrifugation, supernatant separation and collection were applied aft er each extraction step; and (ii) single-step extraction with methanol/water (60:40 by volume) following the same protocol.The conventional extraction of anthocyanin extracts was conducted according to the procedure described by Lorrain et al. (31) using acidifi ed aqueous solution of methanol (methanol/water/12 M HCl of 70:29.5:0.5 (by volume), pH=1.51).
mim][HSO 4 ]>[mim][HSO 4 ]>[mim][CF 3 CO 2 ].These protic-based ILs generally extracted lower mass fractions of total proanthocyanidins than the two most effi cient ILs containing Br -: [C 2 mim][Br] and particularly [C 4 mim][Br].This was especially pronounced in [mim] [CF 3 CO 2 ], where the obtained values were very close to the lowest ones determined in [C 10 mim][Br].These results were in accordance with the earlier research (26,27), which has likewise shown that ILs based on Br -were also the most effi cient in the extraction yield of proanthocyanidins.Furthermore, extraction yields of total proanthocyanidins were signifi cantly enhanced (p<0.05) when concentrations of all ILs were increased from 0.5 to 2.5 mol/L, except for [mim][CF 3 CO 2 ], when this increase was negligible.The same trend was noticed earlier by other authors, indicating that the extraction effi ciency of proanthocyanidins from the Larix gmelinii bark with [C 4 mim] [Br] increased in the concentration range of 0.25-1.25 5 and 2.5 mol/L of [C 4 mim][Br] as well as 2.5 mol/L of [C 2 mim][Br], [C 5 mim][Br] and [sC 4 mim][HSO 4

Table 1 .
List of the studied ionic liquids

Table 2 .
Eff ect of conventional solvent and ionic liquids on the extraction of total proanthocyanidins from grape skin

Table 3 .
Eff ect of conventional solvent and ionic liquids on the extraction of total anthocyanins from grape skin