Exploration of the anthocyanin and proanthocyanidin profile of Greek red grape skins belonging to Vradiano, Limnio, and Kotsifali cultivars, analyzed by a novel LC‐QTOF‐MS/MS method

Abstract

Introduction

Winegrape varieties Kotsifali, Limnio, and Vradiano

Objective

The aim of this study was to develop a liquid chromatographic quadrupole time‐of‐flight tandem mass spectrometric (LC‐QTOF‐MS/MS) method for the investigation of the anthocyanin and proanthocyanidin content of Greek grape varieties employing target and suspect screening strategies.

Methodology

A novel LC‐QTOF‐MS/MS method was developed and validated to assess the anthocyanin content of Kotsifali, Limnio, and Vradiano grape varieties. Sixteen grape samples were collected from the main growing areas of each variety in Greece. The influence of the grape variety on the anthocyanin and proanthocyanidin composition of three Greek winegrapes was investigated using chemometrics.

Results

Excellent linearity (R ² > 0.99) was achieved for all the target analytes, and recoveries ranged between 90.1% and 119.1%. The limits of quantification (LOQs) and limits of detection (LODs) were calculated over the range of 0.020–0.40 mg/g and 0.010–0.13 mg/g, respectively. The RSD% was lower than 9.1% and 7.3% for intra‐day and inter‐day studies, respectively, indicating satisfactory trueness and precision. Target and suspect screening resulted in the identification of 5 and 26 anthocyanins, respectively.

Conclusions

Kotsifali variety exhibited a higher concentration of anthocyanins compared with Vradiano and Limnio. Higher levels of mean degree of polymerization (mDp) and different percentage levels of prodelphinidins (%P) were established among the varieties.

Short abstract

A novel LC‐QTOF‐MS/MS method was developed and validated for the analysis of anthocyanins and proanthocyanidins in Greek grape varieties: Kotsifali, Limnio, and Vradiano. The method exhibited excellent linearity (R2 〉 0.99) and high recoveries (90.1–130.8%). Limits of quantification (LOQs) and limits of detection (LODs) were calculated over the ranges 0.020–0.40 mg/g and 0.010–0.13 mg/g, respectively. Chemometrics revealed the effects of grape variety on anthocyanin and proanthocyanidin composition. Kotsifali exhibited the highest anthocyanin levels, while different levels of proanthocyanidin composition were found among the varieties.

1. INTRODUCTION

Considering the need for more accurate and truer labeling, food authenticity has become crucial in recent years. A product is authentic as long as the label appropriately describes it and complies with the laws in the country where it is sold. ¹ Wine authenticity and traceability have attracted a lot of interest worldwide, and this has necessitated the development of analytical methods to guarantee wine traceability. ²

The phenolic composition of grapes is a critical factor in the assessment of wine quality and authenticity. Anthocyanins and condensed tannins, commonly known as proanthocyanidins constitute the two main classes of phenolic compounds that could be used to assess the quality of red wine. ³ Anthocyanins are the pigmented phenolic compounds responsible for the color of red grapes and wines. ³ , ⁴ Five glycosylated anthocyanins are found in red Vitis vinifera varieties, and their relative amount is characteristic for each variety, thereby providing a potential tool for assessing wine authenticity. ⁵ , ⁶ Flavan‐3‐ols exhibit high concentrations in grapes and have a significant impact on the color, astringency, bitterness, stability, and aging of wine, as well, depending on their amount, composition, and degree of polymerization. ⁶ , ⁷ , ⁸ Flavan‐3‐ols are found in various forms such as monomers, which are the constitutive units of oligomers and polymers (proanthocyanidins or condensed tannins), with degrees of polymerization ranging from 2 to over 100 depending on the matrix. ⁹ Proanthocyanidin profile has been also used for taxonomical studies in grapes, sensory characterization, structure–function studies, and other technological objectives, as well. ⁶ , ⁷ , ¹⁰

High‐pressure liquid chromatography (HPLC) coupled to various detectors such as UV–Vis, diode array (DAD), or mass spectrometric detectors (MS) has been successfully applied in wine fingerprinting studies, as it has already been reviewed. ¹¹ High‐resolution mass spectrometric techniques (HRMS) have revolutionized the field of wine characterization and authenticity. ¹² Untargeted analysis, without a doubt, offers a comprehensive strategy with several applications documented on the wine metabolome. ¹³ The use of target and suspect screening workflows has also been applied to investigate patterns revealing varietal markers that could be used in wine authentication studies. Additionally, the interpretation of HRMS data using chemometric tools offers new perspectives on wine analysis and authenticity, as it is possible to classify the samples according to the grape cultivar. ¹⁴

The bottleneck in grape and wine authenticity studies is the absence of validated analytical methods that combine careful sample collection and preparation, and the use of HRMS techniques that enable the accurate determination of anthocyanins and the tentative identification of their derivatives, as well. ¹⁵ Among the studied cultivars, Greek autochthonous cultivars ¹⁶ have received a lot of attention recently owing to the wine industry's need for adaptation to climate change. Many of the autochthonous Greek varieties are part of the country's viticultural legacy and constitute a genetic pool for future adaptation of viticulture due to their higher resilience to warm and dry conditions. ¹⁷ The majority of these cultivars have not been characterized, yet. ¹⁸

In this work, a novel liquid chromatographic time‐of‐flight mass spectrometric (LC‐QTOF‐MS/MS) method was developed and validated for the determination of anthocyanins in winegrape skins. The method was applied for the first time in the chemical characterization and subsequent discrimination of three Greek red winegrape (V. vinifera) varieties based on the anthocyanin profiles using both targeted and suspect screening workflows combined with chemometrics. The monomeric and polymeric proanthocyanidin fractions of skins were also analyzed using HPLC‐DAD and LC–MS/MS to identify their composition and their relative mean degree of polymerization (mDp) respectively.

2. EXPERIMENTAL

2.1. Sample collection

In this study, 16 red grape samples from the native Greek Vitis Vinifera cultivars Kotsifali, Limnio, and Vradiano were collected at the stage of full ripeness. The grapes were cultivated in different geographical regions of Greece and were harvested in 2020. Samples of 50 berries were collected for the grape maturity analysis. A table depicting sample origin and classical analysis results can be found in Table S1 of Supporting Information A. For the analysis of phenolics, another 50 berries were collected from each vineyard to form a bulk sample. Grape samples were stored in boxes with dry ice and were brought to the laboratory and stored in a deep freezer (−80°C) until further treatment. ¹⁹

2.2. Sample preparation

Skins were removed by hand from the grapes, washed in distilled water, and freeze‐dried for 2 days. The frozen skins were ground in a ball grinder before starting lyophilization, and then the freeze‐dried skins were stored in a deep freezer (−25°C) until analysis. ¹⁹

2.3. Chemicals and reagents

Acetone (ACE) and Acetonitrile (ACN) HPLC grade were purchased from Merck (Zedelgem, Belgium). Methanol (MeOH) (HPLC‐Ultra LC–MS grade) and water (H₂O) (HPLC‐Ultra LC–MS grade) were purchased from HiPerSolv CHROMANORM (VWR Chemicals BDH, The Netherlands). Formic acid (99%) and hydrochloric acid (37%) for analysis were purchased from Carlo Erba (Chaussée du Vexin, France), and trifluoroacetic acid for LC–MS was purchased from Fluka (Buchs, Switzerland). Sodium Phosphate Dibasic Heptahydrate and Sodium Phosphate Monobasic Monohydrate were purchased from Sigma Aldrich (St. Louis, MO, USA). Τhe analytical standards of delphinidin‐3‐O‐glucoside chloride (99.5%) (Dlp‐3‐O‐glu), cyanidin‐3‐O‐glucoside chloride (96.2%) (Cyn‐3‐O‐glu), petunidin‐3‐O‐glucoside chloride (97.5%) (Pt‐3‐O‐glu), peonidin‐3‐O‐glucoside chloride (96.6%) (Pn‐3‐O‐glu), and malvidin‐3‐O‐glucoside chloride (96.7%) (Mlv‐3‐O‐glu) were obtained from Extrasynthese (GenayCedex, France). Catechin (98%), epicatechin (97%), epigallocatechin (99.5%), epicatechin gallate (98%), gallocatechin (98%), epigallocatechin gallate (95%), procyanidin B1 (90%), and procyanidin B2 (90%) were purchased from Sigma‐Aldrich (Darmstadt, Germany).

2.4. Anthocyanins extraction and analysis

2.4.1. Extraction of anthocyanins

The optimum extraction parameters were carried out as they have already been reported in our previous work. ¹⁹ Specifically, 0.014 g of freeze‐dried skin was weighed, and 200 μL of MeOH was added. Then, 1.4 mL of ACE/H₂O/TFA (70:29.95:0.05) was added. The analytes were extracted under sonication for 10 min and stirred at 40 rpm for 20 min, repeated twice. The extraction was carried out at 4°C in the dark, and then the extract was centrifuged for 10 min at 10,000 rpm, with the temperature constant at 4°C. A fraction of 0.5 mL of the supernatant was dried under nitrogen pressure. The concentrate was reconstituted with a mixture of 250 μL MeOH and 750 μL H₂O 0.134% formic acid. Afterward, the solution was led to an ultrasonic device again at 4°C in the absence of light for 30 min. Then, the extract was centrifuged and thermostated for 15 min at 4°C, 14,000 rpm, and the supernatant was filtered through 0.22 μm RC syringe filters (RC) (Captiva, Agilent Technologies), and injected into the LC‐QTOF‐MS/MS. All analyses were performed in triplicate (n = 3).

2.4.2. Instrumentation used for anthocyanin extraction and analysis

All samples were freeze‐dried in an Alpha 2–4 LD freeze dryer acquired from Martin Christ Gefriertrocknungsanlagen GmbH (Osterode am Harz, Germany) that was equipped with a two‐stage vacuum rotary pump RZ 2.5 Vacuubrand (condenser temperature: −80°C, max flow 2.3/2.8 m3 h⁻¹, ultimate vacuum 4 × 10⁻⁴ mbar). A 5804 R centrifuge system with rotor F‐45‐30‐11 was purchased from Eppendorf AG (Germany). Water was purified in a Direct‐Q® 3 UV Water Purification System, which was acquired from Merck KGaA, Darmstadt, Germany. ME 25 ST 0.45 μm membrane filters (Schleicher and Schuell, W. Germany) were employed to filter the aqueous mobile phase. For solvent evaporation under nitrogen gas, a TurboVap LV workstation was used by Caliper Life Sciences (Hopkinton, MA, USA). A Stuard‐SB3 stirrer and an ultrasonic bath RK 100H (Bandelin Sonorex, Berlin, Germany) were used for the extraction. An HPLC‐QTOF‐MS/MS system consisting of an Exion LC AD system (solvent degasser, two pumps, autosampler, column oven, and controller) and an X500R quadrupole‐Time of Flight (QTOF) mass spectrometer (SCIEX, Framingham, MA, USA) were used for analysis. The analytes were introduced into the mass spectrometer via an ESI turbo VTM source operated in positive ion mode.

2.4.3. LC‐QTOF‐MS/MS analysis

TOF–MS and TOF–MS/MS data were acquired using data‐dependent acquisition (IDA) electrospray ionization mode. Nebulizer gas was set at 55 psi, heater gas at 50 psi, and curtain gas at 30 psi. Ionspray voltage was 4500 V, and the declustering potential was 80 V. Turbo spray temperature was 550°C, accumulation time 0.2 s, and 7 CAD gas. Mass calibration was performed daily using the appropriate calibration solution provided by SCIEX. Mass calibration was also performed during batch for internal calibration. Mass spectra were collected in the range of 100–1000 Da. MS/MS spectra were obtained at a collision energy of 35 V and collision energy spread of 15 V. IDA was used for the collection of MS/MS data at an accumulation time of 0.08 s for the five most abundant precursor ions for each MS scan. Sample acquisition and processing were controlled by SCIEX OS software. The chromatographic separation was carried out in the C18 column (2.1 × 100 mm, 2.6 μm) acquired from Fortis (Cheshire, United Kingdom), which was preceded by a pre‐column C18 (10 × 2 mm, 2.6 μm) also acquired from Fortis (Cheshire, United Kingdom). Mobile phase (A) consisted of 90% H₂O, 10% MeOH with 0.2% formic acid, and mobile phase (B) 100% MeOH with 0.2% formic acid. For the analysis of anthocyanins, gradient elution started with 1% of the organic phase (B) (flow rate 0.2 mL min⁻¹) for 1 min, gradually increasing to 39% for the next 4 min, and then increasing to 95% (12–15 min) and remaining constant for the following 3 min (flow rate 0.4 mL min⁻¹). Then, the organic phase increased gradually to 99% at a flow rate of 0.2 mL min⁻¹, within 1 min, and remained constant for another 5 min (16–21 min). Finally, the system was returned to its initial conditions (1% B–99% A) and was restored within 0.1 min (flow rate decreased to 0.2 mL min⁻¹) to re‐equilibrate the column for 5 min before the next injection. Column temperature was set at 40°C, and the injection volume was 5 μL. The method was adapted from Kalogiouri et al. ²⁰

A quality control (QC) sample was used throughout the batch to evaluate sample stability during analysis. A typical QC sample was prepared by combining aliquots of the same volume from each sample. ¹ As part of the analytical run, the QC sample was injected at regular intervals to produce a set of data from which the repeatability and the reproducibility of the batch could be evaluated. The relative standard deviations (RSDs) of the retention times (t _R ) and the peak areas for the glucoside derivatives of anthocyanins (delphinidin, cyanidin, peonidin, petunidin, and malvidin) are presented in Supporting Information A (Table S2), demonstrating the good performance of the analytical system. Specifically, the %RSDs for the peak areas of the standard compounds were less than 5% (n = 6), and the retention time shift was in the range of 0.3–0.4% RSD (n = 6).

2.4.4. Screening strategies

Target screening

The identification and quantification of target anthocyanins was performed using analytical standards. A target list of compounds was created from the literature and included five significant anthocyanins that have been determined in grapes and wines according to the literature. ¹⁶ , ¹⁸ The target list is presented in Table S3 in Supporting Information A. For each target compound, extracted ion chromatograms (EICs) of the precursor ions were created using the SCIEX OS Software (Data Processing–Analytics). Target screening was performed using the following parameters: mass accuracy of the precursor ion and the MS/MS fragments with a selection window of 5 ppm, retention time tolerance (t _R  < 0.2 min), a response peak area threshold of above 1000, and peak intensity of at least 800.

Suspect screening

A suspect list was created using compounds identified in grapes, berries, and wines, retrieved from several public databases, including Phenol Explorer (http://phenol-explorer.eu; accessed on February 5, 2023), FooDB (http://foodb.ca; accessed on February 7, 2023), and Grape Cyc (https://www.plantcyc.org, accessed on February 8, 2023). The in‐house suspect database consisted of 97 anthocyanins and is presented in Table S4 in Supporting Information A. The initial suspect list includes the molecular formulas, the CIDs of PubChem, CAS numbers, and the simplified molecular‐input line‐entry system (SMILES).

The presence of the suspect compound was determined by careful analysis of the isotopic fit, comparison of the MS/MS fragments with those in mass spectral libraries, or by in silico fragmentation tools such as MetFrag. MetFrag was employed using the neutral exact mass (with a mass error of 5 ppm) and the appropriate ionization mode. For the identification of the suspect compounds, the following parameters were defined: mass accuracy of monoisotopic peak (5 ppm), isotopic fit values (≤50), ion intensity (>1000), and peak area (>2000). Semi‐quantification of the suspect compounds was performed using the calibration curves of target compounds having similar structures. ¹ In accordance with that, the estimated concentration of malvidin derivatives and pinotin A were expressed as malvidin‐3‐O‐glucoside equivalent (mg/100 g f.w.), compounds with cyanidin in their structure and luteolinidin as cyanidin‐3‐O‐glucoside equivalent (mg/100 g f.w.), and the same for petunidin, peonidin, and delphinidin derivatives as their glucoside equivalents (mg/100 g f.w.).

2.4.5. LC‐QTOF‐MS/MS method validation

Method validation was performed to evaluate trueness and precision, estimate linearity, selectivity, limits of detection (LODs), and limits of quantification (LOQs), using freeze‐dried samples (0.014 g of freeze‐dried skin). LODs were calculated as 3.3 signal‐to‐noise ratio (S/N), and the formula LOQ = 10 S/N was employed for the calculation of the LOQs. ²¹ Standard stock solutions of anthocyanins were diluted in LC‐MS grade MeOH with 0.1% HCl to a final concentration of 1000 mg/L. Then, they were stored at −20°C in dark brown glass bottles to prevent photodegradation. The stock solutions were gradually diluted in MeOH:H₂O (50:50, v/v) to create the working solutions. Spiked calibration curves were constructed using 0.014 g of freeze‐dried skins that were spiked with working solutions over the range LOQ to 1.0 mg/g to create 7‐point calibration curves for all the target analytes. Relative recoveries (%R) were determined using means of recovery percentage by comparing three concentration levels (0.05, 0.1, 1.0 mg/g) of the studied analytes and added (mean concentration found/added concentration *100%) to assess trueness. For the estimation of matrix effect (%ME), the response of each analyte in a post‐extraction spiked sample was divided by the response in the standard solution and then subtracting 1 from the quotient. Positive values imply signal enhancement and negative ones imply signal suppression. ²² Finally, repeatability (inter‐day precision) and within laboratory reproducibility (intra‐day precision) were expressed as %RSD values of five replicate analyses (n = 5) on the same day and three replicate analyses on three consecutive days (n = 3 × 3), respectively.

2.4.6. Chemometric analysis

The statistical differences between the species based on their concentration were calculated with ANOVA at a 95% confidence level (p < 0.05), in Microsoft Excel (Microsoft, WA, USA) utilizing the Data Analysis tool. Partial least discriminant analysis (PLS‐DA) was applied as a supervised chemometric tool to classify the samples according to the variety based on the anthocyanin profile. The PLS‐DA model was developed using the SIMCA package (version 14.1.; Umetrics, Sweden). Variable importance in the projection algorithm (VIP) was used to evaluate the importance of each feature, applying a cut‐off value of 1. ²³ The PLS‐DA model was validated, and R2Y (total sum of variation in Y explained by the model and Q2Y indicates the goodness of prediction) and R2X (estimates the fraction of the variation in X explained by the model) cross‐validation parameters were calculated.

2.5. Proanthocyanidins extraction and analysis

2.5.1. Preparation of stock solutions and calibration standards

Standards stock solutions of proanthocyanidins were diluted in MeOH (LC–MS grade) to a final concentration of 1000 mg/L and kept in dark brown bottles at −20°C. For quantitative studies, calibration standards were created by diluting the stock solutions with a diluent (MeOH:H₂O [50:50, v/v]). Calibration curves were constructed over the range of 1–10 mg/kg for the monomeric/oligomeric analysis and in different ranges for the polymeric analysis, depending on the compound, as shown in Tables S5 and S6 in Supporting Information A, respectively. Each sample was analyzed in triplicate (n = 3).

2.5.2. Extraction of proanthocyanidins

Sample extraction was adapted from previously published methods ¹⁹ , ²⁴ , ²⁵ with slight modifications. The mass‐to‐volume ratio of the grape powder was optimized until the ultimate extraction process was reached. Seventy milligrams of freeze‐dried skin was weighed and extracted using 1.4 mL of ACE/H₂O/TFA (70:29.95:0.05) and 200 μL of MeOH. The solutions were extracted under sonication for 10 min and stirred at 40 rpm for 20 min during two cycles (2 × 30 min). The extraction was carried out at 4°C in the absence of light, and the extract was centrifuged and thermostated at 4°C, 10,000 rpm, for 10 min. Then, a fraction of 1 mL of the supernatant was dried with nitrogen under pressure.

2.5.3. Fractionation of skin proanthocyanidins

Using C‐18 SPE cartridges (TELOS C18 500 mg/6 mL), skin extracts were cleaned up according to Bordiga et al. ²⁵ The concentrates from the previous step were redissolved in 15 mL of phosphate buffer pH 7.0 (67mM). Solid phase extraction was followed with C‐18 SPE cartridges. Before sample loading, the SPE cartridges were conditioned with 10 mL of MeOH, 20 mL of H₂O, and 10 mL of phosphate buffer. Then, the sample is introduced into the SPE column, where the analytes are retained by the packing material. The flow rate during conditioning and sample loading was 2 mL min⁻¹. Phenolic acids were first eluted with 10 mL of phosphate buffer. After the cartridges were dried, the elution of monomeric and oligomeric flavan‐3‐ols was carried out with 20 mL of ethyl acetate, followed by the elution of polymeric proanthocyanidins with 12 mL of MeOH. Then, monomeric/oligomeric and polymeric solutions were dried under a nitrogen atmosphere.

2.5.4. Analysis of monomeric and oligomeric fractions

The dry monomer residues were re‐dissolved in 1 mL of MeOH:H₂0 (50:50, v/v). Subsequently, the solutions were led to an ultrasonic device at 4°C in the dark for 30 min. The solutions were filtered through 0.22 μm RC syringe filters (RC) (Captiva, Agilent Technologies) before injection in the chromatograph.

A SpectraSYSTEM chromatographic system (Thermo Separation Products, Austin, TX, USA) equipped with a diode array detector (UV6000LP) was used. The HPLC/DAD system consisted of a P2000 secondary solvent pump and an AS3000 autosampler provided with a 100 μL injection loop. Separation was performed on a reversed‐phase chromatographic column (Nucleosil 100‐5 C18, 250 × 4.6 mm, 5 μm, Macherey–Nagel, Düren, Germany), and gradient elution of an aqueous mobile phase acidified with 1% acetic acid (solvent A) and ACN (solvent B) was applied under a constant flow rate at 1 mL min⁻¹. The elution program used was as follows: 5%–10% B linear from 0 to 5 min, 10%–12% B from 5 to 15 min, 12%–17% B linear from 15 to 25 min, 17%–95% B from 25 to 30 min, 95% B isocratic from 30 to 40 min, 95%–5% B linear from 40 to 41 min, and re‐equilibration of the column from 41 to 50 min under initial gradient conditions. The injection volume was 10 μL, the column temperature was 35°C, and the detector signal was recorded at 280 nm using ChromQuest 5.0 software (Thermo Fisher Scientific Inc.).

2.5.5. Phloroglucinolysis reaction

The phloroglucinolysis reaction was used to determine the proanthocyanidins mDP for skin extracts in polymeric tannins fractions. Until the reaction took place, the methanolic derivatives were stored in a freezer at −20°C. Τhe phloroglucinol process was based on older methods and more specifically on the works of Kenedy and Pinasseau. ²⁵ Τhe phloroglucinolysis reaction was conducted on the methanolic concentrates. After optimization, 350 μL of the phloroglucinol solution (50 g/L phloroglucinol, 10 g/L ascorbic acid, 0.2 mol/L HCl in MeOH) was added. The solution was heated in a water bath for 20 min at 50°C, after being solubilized using an ultrasonic bath with ice for 30 min. The phloroglucinolysis process was stopped by placing the sample in ice and adding 350 μL of ammonium formate solution (12.6 g/L). Prior to the chromatographic analysis, the extract was centrifuged (10 min 4°C, 14,000 rpm), and the supernatant was filtered through 0.22 μm RC syringe filters (RC) (Captiva, Agilent Technologies) before injection.

2.5.6. Analysis of polymeric flavan‐3‐ols

An LC‐ESI‐MS/MS system consisting of a Surveyor LC pump, an in‐line degasser, an autosampler, and a TSQ Quantum Discovery Max triple quadrupole mass spectrometer (Thermo Electron Corporation, Waltham, MA, USA) was used for the quantification of the reaction products. A HyPURITY C18 analytical column (150 mm × 2.1 mm i.d., 5 μm particle size) was used for the chromatographic separation (Thermo Scientific, Waltham, MA, USA). The mobile phase consisted of 0.1% (v/v) formic acid in H₂O purity LC–MS (solvent A) and 0.1% (v/v) formic acid in ACN purity LC–MS (solvent B). The gradient program was as follows: isocratic for 17 min with 5% mobile phase B, 5%–75% mobile phase B (at 17 min), 75%–95% mobile phase B (at 18 min), isocratic for 18 min–23 min with 95% mobile phase B, 95%–5% mobile phase B (at 23.01 min), and isocratic for 23.01 min to 36 min with 5% mobile phase B. The mobile phase flow rate was 0–23.1 min: 0.25 mL min⁻¹, 24–32: 0.5 mL min⁻¹, and 33–36: 0.25 mL min⁻¹. The oven temperature was set at 25°C, the injection volume was 4 μL, and the total run time was 36 min.

A triple quadrupole mass spectrometry system equipped with an electrospray ionization (ESI) source operating at the negative ion mode was used. The operating conditions of ESI were as follows: sheath gas pressure was 50 arbitrary units; auxiliary gas pressure was 10 arbitrary units; spray voltage was 4000 V; the capillary temperature was 375°C. The collision gas pressure was 1.5 mTorr. The acquisition was made in the selected reaction monitoring mode. All the data were acquired and processed by the Trace Finder software (Thermo Scientific).

3. RESULTS AND DISCUSSION

3.1. LC‐QTOF‐MS/MS validation results

The method was validated in terms of linearity, selectivity, LODs, LOQs, trueness, and precision, and the validation results are presented in Table 1. The method presented excellent linearity (R ² > 0.990) over the range LOQs–1.0 mg/g. The analytes showed satisfying recovery efficiency (90.1%–119.1%) for all the anthocyanins. LOQs ranged between 0.020 and 0.40 mg/g, and LODs were calculated over the range of 0.010–0.13 mg/g, respectively. The RSD was <9.1% and 7.3% for intra‐day and inter‐day studies, respectively. Matrix effect values demonstrate slight signal suppression (as low as −19.4% for peonidin) for all the studied compounds.

TABLE 1.

LC‐QTOF‐MS/MS method validation results.

Compound	Equation	r 2	0.05 mg/g				0.1 mg/g				1.0 mg/g				LOD (mg/g)	LOQ (mg/g)
			%R	ME%	Intra‐day %RSD (n = 5)	Inter‐day %RSD (n = 3 × 3)	%R	ME%	Intra‐day %RSD (n = 5)	Inter‐day %RSD (n = 3 × 3)	%R	ME%	Intra‐day %RSD (n = 5)	Inter‐day %RSD (n = 3 × 3)
Dlp‐3‐O‐glu	y = (−44,493 ± 80,739) + (1,046,352 ± 16,385)x	0.998	117.1	−7.8	4.8	7.3	107.3	−8.2	8.2	5.6	119.1	−7.7	9.1	4.6	0.13	0.40
Cyn‐3‐O‐glu	y = (390,074 ± 201,133) + (1,097,719 ± 40,816)x	0.991	112.2	−8.5	1.2	2.5	114.2	−7.4	2.1	0.9	115.1	−7.6	2.5	3.1	0.010	0.030
Pn‐3‐O‐glu	y = (1,548,059 ± 61,549) + (664,279 ± 303,299)x	0.990	108.1	−18.2	6.7	5.2	90.6	−19.4	4.4	3.8	101.2	−19.2	4.9	6.2	0.030	0.10
Pt‐3‐O‐glu	y = (299,190 ± 198,074) + (1,388,250 ± 40,196)x	0.994	107.3	−11.5	1.8	3.6	105.1	−14.1	5.8	5.1	104.2	−15.3	2.3	2.7	0.010	0.030
Mlv‐3‐O‐glu	y = (468,631 ± 211,758) + (1,269,566 ± 42,972)	0.993	106.2	−18.1	7.6	5.3	90.1	−16.2	1.9	3.2	103.1	−18.1	3.6	5.5	0.010	0.020

3.2. Target screening results

The concentrations of the target compounds were quantified in all samples, taking into consideration that quantitative analysis is essential to provide a thorough overview of the anthocyanin content of red grapes. The concentration of the non‐acylated anthocyanidins was quantified according to their correspondingcalibration curves, prepared with freeze‐dried samples (0.014 g of freeze‐dried skins) spiked over the range LOQ to 1.0 mg/g. The calculated quantification results of freeze‐dried skins were converted to the corresponding fresh weight equivalents, and the results are presented as mg/100 g of fresh weight (f.w.) in Table 2. Target screening results and the MS/MS fragments of the compounds are summarized in Table S7 in Supporting Information A.

TABLE 2.

Target compounds quantification results expressed as mean values (n = 3) of mg/100 g fresh weight (f.w.).

Variety	Sample name	Dlp‐3‐O‐glu	Cyn‐3‐O‐glu	Pt‐3‐O‐glu	Pn‐3‐O‐glu	Mlv‐3‐O‐glu
Kotsifali	KAP	6.2 ± 0.1	9.7 ± 0.2	4.6 ± 0.1	15.2 ± 0.2	10.9 ± 0.1
	KAR	4.0 ± 0.1	6.2 ± 0.1	3.1 ± 0.1	10.2 ± 0.5	9.0 ± 0.6
	KD	4.0 ± 1.6	6.5 ± 0.2	3.9 ± 0.1	11.7 ± 0.6	9.7 ± 0.3
	KI	2.7 ± 0.1	4.9 ± 0.1	2.7 ± 0.1	8.3 ± 0.1	8.6 ± 0.2
	KKA	3.2 ± 0.0	3.7 ± 0.1	3.2 ± 0.1	10.6 ± 0.3	11.1 ± 0.1
	KT	4.0 ± 0.0	3.9 ± 0.1	3.9 ± 0.1	12.3 ± 0.5	12.3 ± 0.5
Limnio	LAO	1.4 ± 0.1	0.9 ± 0.0	2.0 ± 0.1	6.5 ± 0.2	11.2 ± 0.6
	LE	0.66 ± 0.1	0.5 ± 0.0	1.2 ± 0.1	4.1 ± 0.3	8.7 ± 0.2
	LI	2.2 ± 0.1	0.7 ± 0.1	3.6 ± 0.1	5.3 ± 0.4	21.6 ± 0.3
	LL	2.1 ± 0.1	2.1 ± 0.3	2.8 ± 0.1	8.7 ± 0.6	13.7 ± 0.5
	LSI	1.7 ± 0.3	1.0 ± 0.1	2.2 ± 0.3	5.5 ± 0.5	12.7 ± 1.1
	LSR	0.8 ± 0.0	0.5 ± 0.0	1.3 ± 0.0	2.5 ± 0.1	10.6 ± 0.6
	LX	0.6 ± 0.0	0.3 ± 0.0	1.1 ± 0.0	2.4 ± 0.2	9.1 ± 0.2
Vradiano	VG	2.4 ± 0.1	1.9 ± 0.1	2.6 ± 0.1	9.0 ± 0.2	8.9 ± 0.3
	VI	1.6 ± 0.0	1.5 ± 0.1	1.8 ± 0.1	7.9 ± 0.3	8.8 ± 0.2
	VIS	1.4 ± 0.0	1.9 ± 0.0	1.6 ± 0.0	8.7 ± 0.3	6.3 ± 0.2

Representative qualifier ions from the discovered compounds were cross‐checked with fragments from previous studies. Delphinidin‐3‐O‐glucoside shows characteristic fragmentation at m/z 303.0496 and cyanidin‐3‐O‐glucoside at m/z 287.0531. The qualifier ions of the glycosylated derivatives of peonidin, petunidin, and malvidin are m/z 301.0683, 317.064, and 331.0789, respectively. ²⁶ , ²⁷ The EIC of the identified target compounds is presented in Figure 1.

FIGURE 1.

Extracted ion chromatogram of the target anthocyanins in a standard solution at 1.0 mg/g concentration level.

Concentration values reported in Table 2 are in general accordance with ranges published in our previous work. ¹⁹ The anthocyanin profiles of the examined cultivars varied greatly, as each one is distinguished by a particular set of anthocyanins. ²⁸ The predominant glucoside was Mlv‐3‐O‐glu, which is the most stable of the five anthocyanins. Mlv‐3‐Ο‐glu was the dominant anthocyanin in the Limnio variety (in a range of 51.3%–67.5%), as it has been previously reported for a large number of native and international varieties. ⁴ , ¹⁶ The percentage of Mlv‐3‐Ο‐glu in Kotsifali variety was calculated over the range 23.5%–34.9%, whereas, in Vradiano, it ranged between 31.7% and 40.6%. Meanwhile, Kotsifali and Vradiano varieties exhibited comparable levels of Mlv‐3‐Ο‐glu and Pn‐3‐Ο‐glu (Figure S1, Supporting Information A), and the percentage of Pn‐3‐Ο‐glu in Kotsifali between ranged between 30.7% and 33.9%, and between 36.1% and 43.7% in Vradiano, respectively. Moreover, tri‐oxygenated derivatives (Dlp‐3‐Ο‐glu, Pt‐3‐Ο‐glu, Mlv‐3‐Ο‐glu) exhibited higher concentrations compared to the di‐oxygenated ones (Cyn‐3‐Ο‐glu and Pn‐3‐Ο‐glu) in Limnio variety (Figure S2, Supporting Information A), which is confirmed by the stable color of Limnio wines. ¹⁹

According to the literature, the most prominent compound is Mlv‐3‐O‐glu followed by the corresponding glucoside of peonidin. ⁴ Kotisfali and Vradiano exhibited similar amounts of Mlv‐3‐O‐glu, whereas the Limnio variety exhibited two times higher concentration of this compound. Specifically, the concentration of Mlv‐3‐O‐glu in Kotsifali varied between 8.6 and 12.3 mg/100 g f.w., 8.7–21.6 mg/100 g f.w. in Limnio, and 6.3–8.9 mg/100 g f.w. in Vradiano. According to the total anthocyanin quantification results (Figure S2, Supporting Information A), the Kotsifali variety grapes presented the highest total anthocyanin content (46.6 mg/100 g f.w.), whereas the Limnio variety presented the lowest (13.5 mg/100 g f.w.) concentration.

Following an individual analysis of each mono glucoside (Figure S3, Supporting Information A), it was found that Cyn‐3‐O‐glu presented the lowest mean content (0.8 mg/100 g f.w. for Limnio and 1.8 mg/100 g f.w. for Vradiano), followed by Dlp‐3‐O‐glu (1.3 mg/100 g f.w. in Limnio, and 1.8 mg/100 g f.w. in Vradiano). At this point, the variety Kotsifali stood out from the others because it had remarkably high values of di‐oxygenated anthocyanin Cyn‐3‐O‐glu (5.8 mg/100 g f.w.), compared to the other varieties. The anthocyanins in Limnio grapes, on the other hand, contributed almost exclusively to Mlv‐3‐O‐glu, whereas Vradiano's di‐oxygenated fraction consists almost exclusively of Pn‐3‐O‐glu. The results of this research are in agreement with the information provided by other studies. ⁴ , ¹⁶

3.3. Suspect screening results

Through suspect screening, 26 compounds from the in‐house suspect screening database were tentatively identified in grape samples. Table S8 in Supporting Information A presents the suspect screening results, providing information for each compound. All the compounds were identified based on their molecular ions [M + H]+, and their MS/MS spectra were compared and analyzed using MetFrag ²⁹ and literature data.

The initial suspect list mainly consisted of the anthocyanins that are presented in natural products. Many derivatives of cyanidin were identified. Especially the precursor ions of the aglycone molecule of cyanidin, cyanidin 3,5‐O‐glucoside, cyanidin 3‐(6″‐acetyl‐galactoside), cyanidin 3‐(6‐p‐caffeoyl) glucoside, and cyanidin 3‐O‐sophoroside were detected at m/z 288.0588, ³⁰ , ³¹ , ³² m/z 612.1691, ³² , ³³ m/z 492.1262, ³¹ , ³² m/z 612.1440, ³⁴ and m/z 612.1698, ³¹ respectively. Cyanidin 3‐O‐galactoside elutes at 5.64 min and shows two qualifier ions at m/z 137.0239 ²⁰ and m/z 287.0529, ²⁰ , ³¹ corresponding to [C₇H₅O₃]⁺ and [C₁₅H₁₀O₆]+H⁺, respectively.

Petunidin 3,5‐O‐diglucoside among its fragments appears one qualifier ion which is the same as petunidin 3‐O‐(6″‐acetyl‐glucoside) at m/z 317 corresponding to [C₁₆H₁₂O₇]+H⁺ that has reported by Colombo et al. ³² The fragment m/z 287.0532 of petunidin 3‐O‐arabinoside, corresponding to [C₁₅H₉O₆+H]+H⁺ was reported by Oancea et al. ³⁵ This is the first report of this compound in grape as it is mainly known as a typical anthocyanin of blueberries. ³¹ For petunidin, 3‐galactoside, three qualifier ions were identified: m/z 274.0472, ²⁰ m/z 302.0426, ²⁰ and m/z 317.064, ²⁰ , ³¹ corresponding to [C₁₄H₁₁O₆‐H]⁺, [C₁₅H₉O₇]+H⁺, and [C₁₆H₁₂O₇]+H⁺ respectively.

The suspect compounds were semi‐quantified according to Kalogiouri et al. ¹ All the derivatives of cyanidin, delphinidin, petunidin, peonidin, and malvidin were semi‐quantified using the corresponding glucoside standards. Table S9 in Supporting Information presents the semi‐quantification results of all the suspect compounds. The EICs, MS, and MS/MS spectra are presented in Figures B1–B31 in Supporting Information B.

3.4. Chemometric analysis

The PLS‐DA analysis of the target and suspect quantification and semi‐quantification results revealed the discrimination of the cultivars based on their anthocyanin content with an explained variance of 70.8%. The PLS‐DA scores plot is presented in Figure 2. The loading plot is presented in Figure S4 of Supporting Information A. The predictive value of the model was satisfactory since values for goodness of fit (R2Y = 0.785) and predictability of the model (Q2 = 0.731) were acceptable. The performance of the PLS‐DA model was evaluated using a receiver operating characteristic (ROC) curve. The area under the ROC curve (AUC) showed that the PLS‐DA model classified all samples with 100% accuracy (AUC = 1, in all cases). Permutation test statistics with 200 random permutations were calculated and the results confirmed that all permuted R2s and Q2s were lower than the original values, suggesting valid model fitting (Figure S4b, Supporting Information A).

FIGURE 2.

PLS‐DA scores plot presenting the discrimination of Kotsifali (marked in green), Limnio (marked blue), and Vradiano grapes (marked in red).

The variables' importance in the projection (VIP values) of the PLS‐DA model was evaluated (VIP analysis), considering the VIP scores for each phenolic compound (suspect and target) detected. The VIP score summarizes the contribution a variable makes to the model, and it is calculated as a weighted sum of the squared correlations between the PLS‐DA components and the original variables. ³⁶ The most important features were selected according to the VIP algorithm (Figure 3), considering only variables with VIP scores above 1. ³⁷ Thirteen features were selected as most significant, responsible for the classification of the samples according to the variety. Those were cyanidin‐3‐O‐glucoside, cyanidin 3‐O‐galactoside, peonidin‐3‐O‐glucoside, delphinidin‐3‐O‐glucoside, and petunidin 3‐O‐arabinoside, which were more abundant in Kotsifali samples compared with the other varieties. Specifically, petunidin 3‐O‐arabinoside exhibited high concentration in KAP and was selected as a marker for differentiation of grape samples affected by the variety. Its high concentration in the sample KAP produced in Creta, reveals that it could be used as a marker for the geographical origin, as well. This is the same case for peonidin 3‐O‐galactoside, malvidin 3‐arabinoside, and petunidin‐3‐O‐glucoside, which all exhibited high concentrations in Kotsifali grapes cultivated in Creta (sample KAP).

FIGURE 3.

Discriminant anthocyanins identified by VIP analysis (variable importance in projection) on grape cultivars according to their geographical origin.

Cyanidin 3‐(6″‐malonyl glucoside), peonidin 3‐O‐(6‐O‐acetyl‐beta‐D‐glucoside), malvidin 3‐O‐(6‐p‐coumaroyl)‐glucoside, and petunidin 3,5‐O‐glucoside were distinctive for Limnio grapes cultivated in Athens (sample LI) (Table S9, Supporting Information A). Another anthocyanin that was revealed as marker was petunidin 3‐galactoside, due to its high content in grapes of Kotsifali and Limnio from Creta and Athens, respectively (samples KAP and LI). According to the literature, some of these markers, especially cyanidin‐3‐O‐glucoside, peonidin‐3‐O‐glucoside, petunidin‐3‐O‐glucoside, and malvidin 3‐O‐(6‐p‐coumaroyl) glucoside have been identified as characteristic metabolites of red wines. ¹²

3.5. Proanthocyanidins composition

3.5.1. Monomeric and oligomeric proanthocyanidin profile

The monomeric (C, EC, ECG, and EGC) and oligomeric (B1 and B2) proanthocyanidins in grape skin extracts were calculated in the range of 0.2–1.1 mg/100 g of f.w. grapes, were presented in the skins (Table S10, Supporting Information A), with the Limnio variety showing the highest amount. The distribution of flavan‐3‐ols of the skins consists equally of all of the six components, with slight variations between cultivars (Figure S5, Supporting Information A). Specifically, skins belonging to Vradiano are composed of the flavan‐3‐ols B1 (31.1% on average), catechin (27.5% on average), and B2 (24.7% on average), whereas the contribution of the other three monomers is smaller and varies between samples. The proanthocyanidin profile in the skins of the other two varieties was formed by the participation of all the monomeric/oligomeric components, with the B2 oligomer being found in a greater percentage in both Kotsifali and Limnio (41.0% and 43.5%, respectively). Figure S6 in Supporting Information A presents a characteristic chromatogram of a real sample.

The number of studied samples was small to classify the grape varieties according to their flavanol composition. However, the presented values imply that at least some of the cultivars appear a specific pattern of flavanols. A greater number of grape samples is needed to extend the classification of grape proanthocyanidins, which is required to describe their qualitative profile in grapes. ⁸ As far as we know, this is the first time that the monomeric and oligomeric composition of Vradiano grapes are exploited. Kotsifali grapes have been studied only in terms of the average degree of polymerization. ¹⁶

3.6. Identification of grape flavan‐3‐ols and their associated phloroglucinol adducts

As demonstrated in the characteristic chromatogram of a polymerized sample of skin (Figure S7, Supporting Information A), the analysis LC–MS allows the rapid and effective separation of flavan‐3‐ol monomers and their corresponding phloroglucinol adducts in 36 min, which is very short in contrast with the conventional methods. ³⁸ Standards solutions in H₂O:MeOH 50:50 (v/v) supplied the fragmentation pattern and the appropriate collision energy for the target analyte standards to be detected as efficiently as possible (C_term, EC_term, ECG_term). For the phloroglucinol adducts that are not available, standards additional experiments have been performed and according to the contemporary literature determined the optimum ionization and the fragmentation parameters. All compounds received singly charged negative precursor ions through ESI. For each component, two MRM ion transitions were chosen (Table S11, Supporting Information A), one for the quantification (quantifier ion) and the other one used for identification. The developed method was based on previously described protocols, ⁷ , ²⁴ adopted with small modifications.

The changes in the extractable polymeric fraction that was isolated by phloroglucinolysis of three different grape varieties during ripening are presented in Table 3. Catechin‐phloroglucinol (Cup), epicatechin‐phloroglucinol (ECup), epicatechin gallate‐phloroglucinol (ECGup), and epigallocatechin‐phloroglucinol (EGCup) were identified as extension proanthocyanidin units. Catechin (C), epicatechin (EC), and epicatechingallate (ECG) were detected as terminal units, with a significantly higher proportion of catechin, which is referred to in many corresponding works. ⁷ , ²⁴ The quantification of these compounds included the calculation of the contribution of each proanthocyanidin in each cultivar, the mean degree of polymerization (mDp), the percentage of galloylation (%G), and the percentage of prodelfinidin units (%P).

TABLE 3.

Structural characteristics and composition (percent in moles) of skin tannin extract.

	Skin extract
		Extension units (%)				Terminal units (%)			mDp	%G	%P
Sample	Grape variety	Cup	ECup	ECGup	EGCup	Cterm	ECterm	ECGterm
KAP	Kotsifali	1.5 ± 0.2	64.9 ± 0.1	0. ± 0.0	33.1 ± 0.1	85.6 ± 0.1	13.0 ± 0.3	1.4 ± 0.1	37.5 ± 0.6	0.5 ± 0.1	32.3 ± 0.2
KAR		2.0 ± 0.1	84.9 ± 0.0	1.8 ± 0.5	11.3 ± 0.0	93.1 ± 0.1	5.1 ± 0.1	1.8 ± 0.1	20.6 ± 0.1	1.8 ± 0.1	10.8 ± 0.1
KD		1.2 ± 0.1	61.8 ± 0.1	1.0 ± 0.1	36.0 ± 0.0	74.6 ± 0.1	21.9 ± 0.1	3.5 ± 0.1	55.4 ± 0.1	1.0 ± 0.1	35.5 ± 0.1
KI		1.9 ± 0.0	60.5 ± 0.1	1.8 ± 0.0	35.8 ± 0.1	90.8 ± 0.1	7.1 ± 0.1	2.1 ± 0.1	25.2 ± 0.0	1.8 ± 0.0	34.4 ± 0.0
KKA		2.6 ± 0.1	82.9 ± 0.0	1.5 ± 0.0	13.0 ± 0.1	87.2 ± 0.1	11.6 ± 0.0	1.2 ± 0.1	27.6 ± 0.0	1.8 ± 0.0	11.1 ± 0.1
KT		1.2 ± 0.3	71.4 ± 0.1	0.5 ± 0.1	26.9 ± 0.1	84.7 ± 0.1	12.9 ± 0.0	2.4 ± 0.1	38.6 ± 0.2	0.6 ± 0.1	26.3 ± 0.1
Kotsifali average (n = 6)		1.7 ± 0.1	71.1 ± 0.1	1.2 ± 0.1	26.0 ± 0.1	86.0 ± 0.1	11.9 ± 0.1	2.1 ± 0.1	34.1 ± 0.2	1.2 ± 0.1	25.1 ± 0.1
LAO	Limnio	3.9 ± 0.0	90.5 ± 0.3	2.5 ± 0.1	3.1 ± 0.2	93.2 ± 0.0	5.7 ± 0.0	1.1 ± 0.0	11.8 ± 0.1	2.3 ± 0.0	2.9 ± 0.0
LE		2.2 ± 0.1	94.8 ± 0.0	2.0 ± 0.1	1.0 ± 0.1	92.6 ± 0.0	5.5 ± 0.2	1.9 ± 0.1	14.7 ± 0.1	2.0 ± 0.1	1.0 ± 0.1
LI		1.0 ± 0.0	96.7 ± 0.1	0.1 ± 0.0	2.3 ± 0.0	91.4 ± 0.1	8.6 ± 0.1	0.1 ± 0.2	12.9 ± 0.1	0.1 ± 0.2	1.7 ± 0.1
LL		3.2 ± 0.1	92.3 ± 0.1	2.2 ± 0.1	2.3 ± 0.1	92.3 ± 0.1	6.1 ± 0.0	1.6 ± 0.0	14.8 ± 0.0	2.1 ± 0.1	2.2 ± 0.0
LSI		2.4 ± 0.0	95.5 ± 0.2	0.1 ± 0.1	2.0 ± 0.1	92.2 ± 0.0	6.8 ± 0.0	1.0 ± 0.0	10.0 ± 0.0	0.2 ± 0.0	1.9 ± 0.1
LSR		3.0 ± 0.0	88.5 ± 0.0	1.7 ± 0.1	6.8 ± 0.1	92.8 ± 0.2	6.1 ± 0.0	1.1 ± 0.0	14.0 ± 0.1	1.7 ± 0.1	6.3 ± 0.0
LX		1.9 ± 0.2	82.5 ± 0.1	0.1 ± 0.1	15.5 ± 0.0	93.2 ± 0.0	5.5 ± 0.1	1.3 ± 0.1	16.2 ± 0.1	0.1 ± 0.0	14.6 ± 0.2
Limnio average (n = 7)		2.5 ± 0.1	91.5 ± 0.1	1.2 ± 0.1	4.7 ± 0.1	92.5 ± 0.1	6.3 ± 0.1	1.1 ± 0.1	13.5 ± 0.1	1.2 ± 0.1	4.4 ± 0.1
VG	Vradiano	3.4 ± 0.0	67.7 ± 0.1	1.2 ± 0.0	27.7 ± 0.1	93.7 ± 0.1	5.5 ± 0.0	0.8 ± 0.0	13.1 ± 0.1	1.2 ± 0.0	25.6 ± 0.1
VI		1.8 ± 0.1	48.9 ± 0.1	0.7 ± 0.0	48.6 ± 0.1	92.3 ± 0.1	6.5 ± 0.0	1.2 ± 0.0	17.0 ± 0.2	0.8 ± 0.1	45.9 ± 0.1
VIS		3.8 ± 0.0	71.2 ± 0.1	0.4 ± 0.1	24.6 ± 0.0	94.0 ± 0.0	6.0 ± 0.1	0.0 ± 0.0	11.6 ± 0.1	0.3 ± 0.0	22.1 ± 0.1
Vradiano average (n = 3)		3.0 ± 0.0	62.6 ± 0.1	0.8 ± 0.0	33.6 ± 0.1	93.3 ± 0.1	6.0 ± 0.0	0.7 ± 0.0	13.9 ± 0.1	0.8 ± 0.0	31.2 ± 0.1

Grape skin tannins consisted of a mixture of procyanidins and prodelphinidins. ⁸ In the skin polymer extracts, EC was the main extension subunit in all of the studied varieties, especially reaching a maximum of 96.7% in a sample LI from the Limnio variety (Figure S8, Supporting Information A). This is followed by the EGCup extension unit to appear a maximum of 48.6% and fluctuates in a high range from 1.0 (in the Limnio variety) to 48.6% (in the Vradiano variety), that is according to the literature. ³⁸

In terms of the terminal unit values in skins, catechin demonstrated in all samples a substantially higher proportion than the other two, particularly of ECG, which was within 1%. Catechin was the dominant terminal unit with levels ranging from 74.6% (a Kotsifali sample) to a maximum of 94% in a sample from the Vradiano variety. The terminal units of epicatechin were higher in Kotsifali skin samples with an average participation rate of 11.9%, and their contribution in Limnio and Vradiano samples appeared stable at 6% on average in both of the two varieties. The derived data are also in agreement with the results obtained by Cohen et al. ³⁹ who reported that ECup was the main extension unit, whereas C was the main terminal in skin extracts of Merlot grapes.

Skin proanthocyanidins differed from seed proanthocyanidins in a lower content of galloylated derivatives and a higher mDP, as previously observed. ²⁵ Skin proanthocyanidins showed a double mDp in the variety of Kotsifali (mDp around 34.1), and the other varieties had similar lower values (between 12.1 to 13.9 on average for Limnio and Vradiano varieties, respectively) (Figure 4). The findings are in accordance with results previously reported by Kyraleou et al., ¹⁶ but slightly higher. The percentage of prodelfinidin units (%P) was highly variable, ranging from 1% to 45.9%, with the lower amount appearing at the Limnio variety (4.4% average). Meanwhile, %G values in skins were uniformly low in all of the studied samples and ranged at the low level from 0.1% to 2.3%. According to the results, the higher values of mDp in the skins of the Kotsifali variety may indicate that it will lead to more astringent wines, but the low values of %G could offset this result.

FIGURE 4.

Mean polymerization degree (mDp) and % G of skins proanthocyanidins from the varieties Kotsifali, Limnio, and Vradiano.

The use of target and suspect screening resulted in the determination of 5 target and 26 suspect anthocyanins in grape skins, respectively. Kotsifali variety exhibited higher concentrations of anthocyanins, with the primary contribution coming from the dihydroxylated anthocyanins Pn‐3‐O‐glu and Cyn‐3‐O‐glu. The anthocyanins in Limnio grapes were mainly attributed to Mlv‐3‐O‐glu, whereas Vradiano presented an intermediate anthocyanin profile. Proanthocyanidin characteristics, such as greater levels of mDP and lower levels of %P, were investigated among the studied cultivars. Kotsifali was characterized by a greater number of polymerized skin proanthocyanidins, and the extension units consist mainly of (−)‐epicatechin.

Supporting information

Table S1 ‐ Grape maturity analysis and origin of grape samples

Table S2 ‐ Quality control results

Table S3 ‐ List of target compounds

Table S4 ‐ List of suspect compounds

Table S5 ‐ HPLC‐DAD calibration parameters

Table S6 ‐ Calibration parameters in LC–MS method

Table S7 ‐ Target screening results

Table S8 ‐ Identified compounds from suspect list

Table S9 ‐ Semi‐quantification results of suspect compounds (in mg/100 g f.w., results expressed as mean values [n = 3])

Table S10 ‐ Monomeric and oligomeric proanthocyanidin composition of skin extracts in red grapes varieties Kotsifali, Limnio, and Vradiano

Table S11 ‐ List of quantified compounds and LC–MS parameters (Retention time, precursor and product ions and ion mode)

Figure S1‐ Percentage of individual anthocyanins (Dlp‐3‐O‐glu, Cyn‐3‐O‐glu, Pt‐3‐O‐glu, Pn‐3‐O‐glu and Mlv‐3‐O‐glu) per total anthocyanin concentration in red grapes varieties of Kotsifali, Limnio and Vradiano

Figure S2 ‐ Total anthocyanins of Kotsifali, Limnio and Vradiano red grape varieties (in mg/100 g f.w., results expressed as mean values [n = 3])

Figure S3 ‐ Box and whisker plots for the glucoside anthocyanins in red grapes varieties Kotsifali, Limnio, and Vradiano

Figure S4 – Chemometrics

Figure S5 ‐ Percentage of monomeric and oligomeric proanthocyanidins in the skins of the grape varieties Kotsifali, Limnio and Vradiano (A) and total skin flavanols (B)

Figure S6 ‐ Characteristic chromatogram of a monomeric/oligomeric fraction of a skin extract

Figure S7 ‐ Characteristic chromatogram of flavan‐3‐ols and their corresponding phloroglucinol adducts in grape skin extract

Figure S8 ‐ Proanthocyanidin composition in percentage of extension units % (A) and proanthocyanidin composition in percentage of terminal units % (B)

Data S1. Supporting Information

Karadimou C, Petsa E, Ouroumi N‐A, et al. Exploration of the anthocyanin and proanthocyanidin profile of Greek red grape skins belonging to Vradiano, Limnio, and Kotsifali cultivars, analyzed by a novel LC‐QTOF‐MS/MS method. Phytochemical Analysis. 2024;35(8):1781‐1793. doi: 10.1002/pca.3400

DATA AVAILABILITY STATEMENT

Data will be made available on request.