Variety–compound–quality relationship of 12 sweet cherry varieties by HPLC-chemometric analysis

Abstract

In this study, 12 sweet cherry varieties grown in Shandong (China) were compared based on their quality indices and high performance liquid chromatography (HPLC) fingerprints, along with chemometric discrimination by similarity analysis, hierarchical clustering analysis and principal component analysis. The established sample preparation and analysis methods represented a non-targeted and practical approach with reasonably high sensitivity, precision, stability and reproducibility, providing comprehensive evaluation of the quality of sweet cherries while allowing the tentative classification of cherry varieties. The results revealed the variety-dependent nature of the HPLC fingerprints of sweet cherries, and the fruit groupings according to their chemical composition (three groups based on phenolics, or four groups if organic acids were of main interest). All cherries were rich in organic acids, hydroxycinnamic acids, hydroxybenzoic acids, anthocyanins, flavan-3-ols, flavonols and other flavonoids, with malic acid, cyanidin-3-O-rutinoside, p-coumaroylquinic acid, neochlorogenic acid and rutin as the characteristic compounds (in the range of 0.57–1.80, 0.002–1.390, 0.370–3.083, 0.38–1.87 and 0.020–0.167 mg g⁻¹ fruit fresh weight, respectively). While anthocyanin and flavonol patterns were useful for a varietal assignment of cherries, cyanidin-3-O-rutinoside could represent a preliminary index for grouping similar cultivars and colours.

Graphical Abstract

Open in new tabDownload slide

Introduction

Sweet cherry (Prunus avium L.) is one of the most popular and highly valued fruits due to its good taste, appealing colour, and high nutritional and health-promoting properties associated with its complex profile of organic acids, sugars, volatile compounds and antioxidants (especially phenolics such as anthocyanins, non-anthocyanin flavonoids and hydroxycinnamic acids) (Chockchaisawasdee et al., 2016; Commisso et al., 2017; Martini et al., 2017; Mirto et al., 2018). Both in vitro assays and in vivo trials have demonstrated the health-promoting properties of sweet cherries, which include anticarcinogenic, immunity-boosting, anti-inflammatory and anti-neurodegenerative effects of cherry constituents and derived metabolites (Kang et al., 2003; Kim et al., 2005; Kelley et al., 2013; Sotelo et al., 2018). However, the lack of consistent and complete methods for determining the multivariate association between chemical composition, sensory and health properties of various cherry varieties make it difficult to assess cherry quality and applicability to support fruit farming and variety selection. Owing to the chemical complexity of fruit matrices, it is inaccurate to assess and compare the quality of various cherry varieties merely based on one or two markers or a single class of bioactive constituents (Li et al., 2010; Liu et al., 2013). Accordingly, establishing a specific pattern of recognition for sweet cherry varieties via a non-targeted approach to obtain metabolomic fingerprints corresponding to all the coexisting chemical components is imperative.

High performance liquid chromatography-photodiode array detection (HPLC-PDA) is particularly useful in this regard, as it allows the analysis of natural products containing different classes of compounds with characteristic chromophores through recording multiple detector wavelengths at the same time during a single analysis (Larsen & Hansen, 2008). Moreover, chromatography fingerprint techniques combined with chemometrics, such as similarity analysis (SA), hierarchical clustering analysis (HCA) or principal component analysis (PCA) can characterise both the marker compounds and unknown components in a complex system, thereby allowing meaningful assessment of all chemical components present in a chromatographic profile (Kong et al., 2008; Meléndez et al., 2013; Wang et al., 2014; Olivas-Aguirre et al., 2017).

To our knowledge, no prior study has combined chromatography fingerprint analysis and chemometrics for pattern recognition along with variety–compound–quality relationship, considering the characteristic UV absorption wavelengths for the key compounds in sweet cherries grown in China. Therefore, in this study, a HPLC-PDA method was established in conjunction with chemometric methods including SA, HCA and PCA, for differentiation of sweet cherry varieties grown in Taian City of Shandong Province (China).

Materials and methods

Materials and chemicals

Thirty-six batches of mature sweet cherry fruits (80 berries per batch), corresponding to 12 varieties, were collected randomly on May 31st, 2018, from cherry trees growing in the cherry germplasm resource base of Shandong Institute of Pomology in Tai'an, Shandong, China. The harvested cherries were sorted to remove undesirable fruits and unwanted substances (including dirt), placed in labelled food-grade paper bags, packed in polystyrene boxes (about 5 kg of each variety), and transported to our laboratory within 20 min (without exposure to direct sun light and high heat). A portion of samples from the 12 cherry varieties were subjected to immediate assessment of fruit quality attributes. To prevent oxidation, the remaining fruits were stored in a freezer (-80 °C) for later chemical analyses, including HPLC profiling over the next few days.

HPLC grade methanol was purchased from Shandong Yuwang Industrial Limited (Shandong, China). Analytical-reagent grade phosphoric acid and ethanol (95%) were purchased from Tianjin Yongda Chemical Reagent Co., Ltd. and Tianjin Kaitong Chemical Co., Ltd. (Tianjin, China), respectively. Commercial bottled purified water (Wahaha brand) was used throughout the study.

Physicochemical characteristics and antioxidant activities of the 12 varieties of sweet cherries

Preliminary sensory analysis was performed according to the method of Vavoura et al. (2015) with slight modifications. Fifty healthy panelists (female: 26; male: 24; age: 20–50; non-smokers), who consumed fresh cherries frequently, were recruited locally. The to-be-analyzed fruits were washed with food grade water and a set of the 12 varieties were randomly selected for each panelist. The panelists evaluated samples in a sensory room under controlled conditions (relative humidity 60%; temperature 25 °C; slight positive pressure; natural lighting). They were instructed to evaluate visually the cherry colour and shape, before the intake of cherries for evaluating the fruit flavour (sweetness/sourness). They were asked to cleanse their palates via consuming a plain cracker (unsalted and unseasoned) and sipping mineral water between samples. The evaluation results were recorded.

The average fruit weight for each cherry variety was estimated according to the method of Grafe & Schuster (2014).

Fruit firmness (as g) was estimated by averaging the firmness values of 30 randomly chosen whole berries (with skin) measured using a TA-XT2i texture analyzer (Stable Micro Systems Ltd., Godalming, Surrey, UK) equipped with a P50 probe (Zhang et al., 2008). The pre-test speed, test speed and post-test speed were fixed at 1.0 mm s⁻¹, and the pressed distance was 5 mm.

The fruit shape index was determined according to the method described by (Li et al., 2016). Ten berries were chosen randomly, and their polar diameter and equatorial diameter were measured using a Vernier caliper. The fruit shape index was calculated using the following equation:

The juice yield was determined by de-pitting the cherries to produce cherry pulp (2 kg), followed by juicing with an electric juicer. The juice was then subjected to centrifugation at 6149 g and 4 °C for 5 min using a high-speed refrigerated centrifuge (TGL-20bR, Shang Hai, China) and the supernatant collected and weighed. The juice generated was stored at 4 °C for immediate analyses, or at −80°C for later analyses. The juice yield was estimated using the equation below:

The pH of the aforementioned cherry juice was measured in triplicate at 25 °C using a pH meter (PHS-3C, Hang Zhou, China).

The total soluble solid content (SSC) of the cherry juice was measured (Crisosto et al., 2003; Raiola et al., 2017), using a handheld digital refractometer (MASTER-α, Atago Co. Tokyo, Japan) and expressed as a mass percentage (%).

The titratable acidity (TA) of the juice was measured according to the method of Raiola et al. (2017). Three titrations per sample were performed, and the TA results expressed as the mass percentage of malic acid (which is the major organic acid in cherries) (Grafe & Schuster, 2014). The SSC/TA ratio was then calculated to estimate the taste of the cherries.

The total reducing sugar content of the cherry juice was determined based on the Nelson–Somogyi method as described by Green et al. (1989), using a UV spectrophotometer (UV-8000, Shanghai, China), and expressed as a mass percentage (%).

The vitamin C content (VC) of the cherry juice was determined via titration based on the method of the Association of Official Analytical Chemists (AOAC, 2016).

The rate of the edible portion of cherries was estimated in triplicate by choosing randomly 30 berries for measurements of the mass of the whole fruit (g) and the mass of the stone (g); the respective mean values were used to estimate the edible portion using the following equation:

The total monomeric anthocyanin content (TMA) was determined following the procedure described by Lee et al. (2005), using a rotary evaporator (SB-1000, Shanghai, China) and a UV spectrophotometer (UV-8000, Shanghai, China). The total phenolic content (TPC) was determined in triplicate based on the Folin–Ciocalteu method described by McDonald et al. (2001). Total flavonoid content (TFC) values were determined in triplicate based on the aluminum chloride colorimetry method described by Chang et al. (2002).

The antioxidant activity of cherries was evaluated by measuring the 2,2’-diphenyl-l-picrylhydrazyl (DPPH) free radical scavenging activity (using the DPPH assay) and reducing ability (using ferric reducing antioxidant power (FRAP) assay) (Sun-Waterhouse et al., 2013).

HPLC analysis

Apparatus and chromatographic conditions

An Agilent 1260 HPLC system (Agilent Corp, Palo Alto, CA, USA) equipped with a quaternary pump, a thermostat column compartment and a PDA were used for HPLC analyses. Agilent Empower 2 software was used for data collection and analysis. In order to achieve useful chemical information and better separation, preliminary experiments were conducted to identify the best type of column (Anasil AQ-C₁₈ column or Agilent Zorbax SB-C₁₈ column, both having dimensions 5 μm, 4.6 × 150 mm), best column temperature (25 °C, 30 °C or 35 °C), composition of the mobile phase (acetonitrile-water, methanol-water, methanol-0.1% (v/v) formic acid-water, methanol-0.1% (v/v) phosphoric acid aqueous solution, or methanol-0.5% (v/v) phosphoric acid aqueous solution), elution velocity (flow rate of 0.6, 0.8 or 1.0 mL min⁻¹), and run time (up to 120 min). From these preliminary experiments, the optimised conditions of cherry juice analysis by HPLC were established: the chromatographic separations reported herein used an Agilent Zorbax SB-C₁₈ column (5 μm, 4.6 × 150 mm) operating at 30 °C. The mobile phase consisted of methanol (solvent A) and 0.5% phosphoric acid in water (solvent B) with a linear gradient elution: 0–10 min, 5%A; 10–18 min, 5–8%A; 18–40 min, 8–40%A; 40-50 min, 40–55%A; 50–53 min, 55–65% A; 53–60 min, 65–100%A. The flow rate was 1.0 mL min⁻¹, with a run time of 80 min and an injection volume of 10 μL. The HPLC mobile phase was prepared fresh daily, with all the solutions being filtered through a 0.45 μm membrane (Schleicher & Schuell, Dassel, Germany) and then degassed prior to the use in HPLC analyses. HPLC spectra were acquired in the range of 200–600 nm, with the following wavelengths being monitored during a run: 210, 280, 330, 370 and 520 nm.

Preparation of sample solutions for HPLC analyses

For HPLC analysis of the phenolic compounds in cherries, extraction conditions of phenolics were optimised via single factor experiments with the following variables: extraction solvents (methanol, ethyl acetate, petroleum ether, or 95% aqueous ethanol), extraction solution volume (25, 50, 100, 150, 200 or 250 mL), and evaporation temperature (35, 45, 55 or 65 °C). From these experiments, an optimised extraction procedure was established: portions (10 g) of each variety of sweet cherries were accurately weighed, destoned, homogenised with 250 mL of 95% ethanol, and steeped at room temperature (22 ± 2 °C) for 30 min in capped flasks. Each of the homogenates was then subjected to an ultrasonic treatment (50 kHz for 0.5 h), using an ultrasonicator (see above), followed by cooling to the room temperature. The solid and liquid were then separated using a conventional funnel. The resulting supernatant was then evaporated to dryness using a rotary evaporator at 45 °C. The dried sample was subsequently reconstituted in methanol to a volume of 10 mL in a 10-mL volumetric flask, before filtration through a 0.45-μm membrane and subsequent injection onto the HPLC system for analyses. For the analysis of organic acids in the cherries, the juice generated in Section Physicochemical characteristics and antioxidant activities of the 12 varieties of sweet cherries was used, filtered through a 0.45-μm membrane, and injected into the HPLC system for analysis at 210 nm. The organic acids and phenolic compounds were tentatively identified using external standards and the database for standard compounds of the Shandong Institute of Pomology based on retention times and absorption characteristics. External standards malic acid, gallic acid, chlorogenic acid, quercetin and cyanidin-3-glucoside were used to estimate the concentrations of the compounds identified.

HPLC method validation and chemometric analyses

The HPLC method was validated in terms of precision, stability and reproducibility using Similarity Evaluation System for Chromatographic Fingerprint of Traditional Chinese Medicine (Version 2004A). Intra-day precision was evaluated based on the repeatability/closeness in agreement between results of successive measurements of the same analyte, and expressed as the relative standard deviation (RSD) of five repeated runs of the same sample solution on the same instrument by the same operator on the same day. Sample stability in this study was assessed on the basis of one variable factor, that is, sample storage time (measurements were performed on the same sample solutions that had been stored at room temperature (22 ± 2 °C) for 0, 6, 12, 24 and 48 h). The reproducibility can be evaluated by calculating separately the intra-day and inter-day repeatability, and in this study, was determined through calculating the RSD value obtained by analysing five replicate samples prepared independently for the same cherry variety for one day or two consecutive days (Crupi et al., 2015). The limit of detection (LOD) and the limit of quantification (LOQ) were determined as a signal-to-noise ratio of 3:1 and 10:1, respectively.

HPLC chromatographic data of the 12 varieties of sweet cherries were integrated automatically and exported as AIA (*.cdf) format files for further processing. First, the format files were imported into the software programme Similarity Evaluation System for Chromatographic Fingerprint of Traditional Chinese Medicine (Version 2004A) which was developed by Chinese Pharmacopoeia Committee. Then a representative reference chromatogram was generated by this software from a group of chromatograms for each sweet cherry variety. Similarity Analysis (SA) values for each chromatogram of a cherry variety were calculated. The relative retention time (RRT) and relative peak area (RPA) were obtained at the same time by correlative coefficient calculation for the fingerprint chromatograms. HCA and PCA were performed using Data Lab 2.7 and SPSS19.0, respectively.

Statistical analysis

Data were analyzed using one-way analysis of variance (Anova). Data statistical analysis and processing were performed using Excel 2010, origin software, SPSS 20.0 software (Chicago, IL, USA), and Duncan's method was used to compare the mean values. In all cases, differences were regarded as statistically significant at a probability level of P < 0.05.

Results and discussion

Physico-chemical and sensory characteristics of the 12 varieties of cherries

The 12 varieties of cherries differed visually in colour, shape, and size (Fig. 1, Table 1). Cherry weight differed significantly among varieties, with Tai Zhu being the heaviest (12.53 g) while Red Honey having the lowest weight (6.00 g) (Table 1). Fruit shape index of the current cherries ranged from 0.79 (Aaron Brooks) to 0.94 (Red Honey), and such differences resulted from the differences in fruit development (which is mainly determined by fruit genotype) (Li et al., 2016). The rates of edible portion for the 12 cherry varieties were quite high and comparable (94.24–96.40%), but their juice yields varied greatly. Red Light (60.45%) had the highest juice yield followed by Rock Candy Cui (59.43%), with Bright Pearl having the lowest juice yield (21.52%). It is well known that the fruit juice yield is affected by a number of factors such as fruit variety, maturity, and juice extraction process. For the same fruit variety, greater fruit maturity generally leads to a higher juice yield (Silva et al., 2018). As expected, the brix to acid ratio differed significantly among the 12 cherry varieties, that is, ranging from 19.25 (Pleasurable Red) to 42.75 (13–33), which contributed to the differences in fruit flavour. The sweet cherries also showed significant differences in the reducing sugar content, that is, from the highest with Rock Candy Cui (15.1%) to the lowest with Pleasurable Red and Rainier (9.40%). Thus, Rock Candy Cui is characterised by its high reducing sugar content.

In general, a higher TSS value indicates greater sweetness. High TSS and TA levels in cherries often lead to a relatively high consumer acceptability, although it ultimately depends on the TSS/TA ratio (Kalyoncu et al., 2009); for example, an unusually high TA value may negatively affect consumer acceptability (Chang & Chang, 2010); In this study, Rock Candy Cui (24.80%) had the highest total TSS followed by 13–33 (19.17%), with Shandong Jade having the lowest TSS (13.60%). Juice pH and TA were slightly different among the varieties studied. On the basis of TSS and TA results, 13–33 and Rock Candy Cui should have a sweet–sour taste superior to that of Pleasurable Red, Red Light, Beautiful Red or Reigyoku. However, the overall sweetness / sourness perceived via fruit intake did not follow this trend. Therefore, there may exist some other factors influencing the sweet–sour balance of cherries, for example, the bitterness imparted by fruit polyphenols (Jaeger et al., 2009). Sun-Waterhouse & Wadhwa (2013) pointed out that polyphenolic compounds with low molecular weight make an important contribution to bitterness, while those of high molecular weight such as tannins generally impart astringency sensations. They further pointed out that the ultimate taste perception would result from the interactions among the food components, interplay among taste qualities and sensations, as well as the release characteristics of various flavour compounds from foods.

In this study, Rock Candy Cui had the highest firmness, with Red Honey having the lowest one (Table 1). The different firmness values among the cherries reflected the differences in their cellular arrangements and compositions, skin characteristics, internal turgor pressure and/or respiration rates (which were closely related to fruit variety, growth conditions, maturity and ripening stage, as well as harvesting, handling, processing, storing and transporting conditions). Furthermore, significant differences in TPC, TFC and TMA were found among the cherries. Shandong Jade had the second highest TPC and the highest TFC (147.08 mg GAE/100 g and 2.12 mg QE /100 g, respectively), while 13–33 having the lowest TPC and TMA values (66.06 mg GAE/100 g and 4.08 mg CGE/100 g, respectively). The differences in fruit firmness and the contents of TSSC, TA, TPC, TFC and TMA all contribute to the differences in taste perceived via fruit intake such as Shandong Jade and 13–33.

Antioxidant capacity

Differences in antioxidant activity (by either the DPPH or FRAP assay) were found among the different cherry varieties, and the changing trends detected by the two assays were not fully the same (Table 1). In the DPPH assay, Red Light had the highest antioxidant activity, followed by Shandong Jade, with Rainier having the lowest radical scavenging power. In the FRAP assay, Red Light and Pleasurable Red had the highest and lowest antioxidant activity, respectively. Thereby, from matching the results of the two assays Red Light and Shandong Jade showed the highest and second highest antioxidant activity, respectively, while Rainier and Pleasurable Red exhibited the lowest one. The detected antioxidant activities of the cherries were affected by the evaluation methods, and resulted from the antioxidant contents and other co-existing substances in cherries. Different antioxidant activity assays have different working mechanisms, and using more than one method to evaluate the antioxidant activity of foods allow a more reasonable conclusion (Tabart et al., 2009). The detected antioxidant activities were previously found to correlate with the phenolic content of sweet cherries (Kelebek & Selli, 2011), although the vitamin C content also made a contribution. In this study, Pleasurable Red had the highest VC value (4.02 mg/100 g) followed by 13–33 and Reigyoku (both ~3.15 mg/100 g) (Table 1).

Methodology validation

As shown in Figs S1-S5, similarities were found in the HPLC chromatographic fingerprints of the 12 varieties of cherry. For each detection wavelength, one of the characteristic peaks in the fingerprint chromatograms with good resolution and relatively high peak area was chosen for the calculations of the RRT and RPA, and to represent other peaks at that wavelength: malic acid for 210 nm, catechin for 280 nm, p-coumaroylquinic acid for 330 nm, rutin for 370 nm, and cyanidin-3-O-rutinoside for 520 nm. The RRT and RPA values of the characteristic peaks (Table 2) were used to gauge the measurable expression of the sweet cherry samples. In terms of intra-precision, most of the RSD values for RRT were not greater than 0.40%, except for Rock Candy Cui at 210 and 280 nm (0.54%, 0.42%), Tai Zhu at 210 nm (0.43%), Shangdong Jade at 280 and 520 nm (0.52%, 0.43%), and Pleasurable Red at 520 nm (0.41%). Most of the RSD values for RPA were not more than 2.87%, suggesting that the precision was reasonably good for both the RRT and RPA of the representative peaks for the cherries. Small variations in the retention time for the same compound among the cherry varieties reflected the differences in fruit matrices, especially pH. An increase in pH would promote ionisation of some compounds, decreasing the retention time in a reversed-phase HPLC separation system (Nogata et al., 1994). The small variations in peak areas and positions were acceptable given the fact that the chromatographic fingerprints were produced from a large number of fresh fruits collected from the different positions on the same cherry tree or different trees of the same cherry variety. In terms of stability during the 48 h storage at 22 ± 2 °C, most RSD values of RRT were less than 0.30% except for Red Honey at 210 nm (0.50%), Red Light at 280 nm (0.44%), 13–33 at 330 nm (0.54%), Tai Zhu at 330 and 370 nm (0.38% and 0.34%) and Aaron Brooks at 520 nm (0.33%). Most RSD values of RPA were less than 2.00%, with only a few at 330, 370 and 520 nm falling in the range 2.00–2.50%. These results indicate that the cherry samples were reasonably stable over 48 h at 22 ± 2 °C, and hydroxycinnamic acids and flavonoids (including anthocyanins) might be slightly more sensitive than other compounds to environmental factors (in good accordance with common knowledge about the stability of these compounds). In terms of reproducibility, most RSD values of RRT and RPA were not more than 0.16% and 1.59%, indicating good reproducibility for preparing and analysing independently prepared replicate samples for a particular cherry variety on the same day. The calculated LOD and LOQ values for these five representative compounds were in the range 0.001–0.002 mg g⁻¹ and 0.003–0.005 mg g⁻¹ respectively. Accordingly, the sample preparation and analysis method established in this study was suitable for the simultaneous quantitative evaluation of organic acids and phenolic compounds in the 12 cherry varieties.

Optimisation of phenolic extraction and HPLC analysis conditions to obtain HPLC fingerprints of the 12 cherry varieties

According to the previously published research (Crupi et al., 2018b), the optimal phenolic extraction conditions were established in this study based on comparative experiments. It was found that 95% ethanol was the most efficient extraction solvent, as the total peak areas of the resultant phenolic compounds identified at 280, 330, 370 and 520 nm were generally higher, compared to those for extracts in methanol, ethyl acetate and petroleum ether. Similarly, the optimal extraction sample to extraction solvent ratio was 10 g per 250 mL 95% ethanol and the optimal extraction temperature was 45 °C. Accordingly, all the sample solutions for HPLC analysis were prepared via ultrasonic extraction (50 kHz) of cherries in 250 mL of 95% ethanol at 45 °C for 0.5 h, followed by heating to dryness at 45 °C.

The optimal HPLC analysis conditions (i.e. use of the Agilent Zorbax SB-C18 column at 30 °C, methanol-0.5% / phosphoric acid in water as the mobile phase and flow rate at 1.0 mL min⁻¹) were also established from comparative investigations, using the highest total peak areas of identified peaks and the highest peak areas of the peaks of interest, the best peak separation, least interfering peaks and smoothest baseline to decide the optimal testing conditions. A run time of 2 h was initially used, though no chromatographic peaks corresponding to a phenolic compound or organic acid were detected after 80 min. Accordingly, an elution time of 80 min were generally used in the HPLC analyses.

The selection of wavelengths for HPLC analysis was based on the absorption maxima of the analytes of interest, also considering the mobile phases UV cutoff. Methanol has a UV cut-off λ_max at 205 nm, thus it would not interfere with the UV–Vis absorption at 210, 280 nm, 330, 370 and 520 nm for the compounds in the cherries. Following previous studies (Sun-Waterhouse et al., 2013), the phenolic compounds in the 12 cherries were monitored using four wavelengths: 280 nm (for quantification of flavanones and flavan-3-ols), 330 nm (for quantification of cinnamic acid derivatives), 370 nm (for quantification for flavones and flavonols) and 520 nm (for quantification of anthocyanins). A linear absorbance–concentration relationship was found between the UV absorbance at 210 nm and the malic acid concentration under the HPLC testing conditions used in this work. The regression equation: y = 1424.3x+4.5267 had high correlation coefficient (R² = 0.999), where y is the absorbance at 210 nm and x is the malic acid concentration (g L⁻¹). A number of other peaks appeared in the chromatogram at 210 nm for each cherry. The flat baselines and good linear absorbance–concentration relationship suggested that 210 nm was a suitable detection wavelength for assessment of sweet cherry quality.

HPLC analysis of the cherry varieties

Under the aforementioned optimised phenolic extraction and HPLC analysis conditions, chromatographic fingerprints of the 12 types of sweet cherries were obtained (Figs S1–S5) and matched using Similarity Evaluation System for Chromatographic Fingerprint of Traditional Chinese Medicine (Version 2004A). All peaks present in the HPLC fingerprint chromatograms were quite well-resolved.

A number of organic acids were detected (Table 3) in the cherry juices, which could be identified by running standards or with reference to HPLC data previously reported for cherries (Mahmood et al., 2012). The organic acids were eluted within the first 8 min. Their retention times are presented as a narrow time range, because slight shifts were detected in their retention (which was caused by the differences in fruit matrix among cherry varieties). A relatively high oxalic acid content was found in all the cherries. This may be due to the naturally occurring metabolic processes that exert physiological functions (such as induction of systemic resistance against plant diseases), or result from the pre-harvest application of oxalic acid to maintain fruit quality and bioactive contents (Wagner & Bladt, 2001; Wagner et al., 2009; Valero et al., 2011; Martínez-Esplá et al., 2014). However, in the present experiment, no treatments with oxalic acid were performed, and thus, the high oxalic acid content fount in these cherry cultivars should be attributed to their genetic characteristics. Malic acid and tartaric acid were abundant in all 12 varieties of cherry, ranging in concentration from 0.57 (Red Honey) to 1.80 mg g⁻¹ fruit fresh weight (Tai Zhu), and 0.47 (Shandong Jade) to 1.30 mg g⁻¹ fruit fresh weight (Rock Candy Cui), respectively. The detected ascorbic acid content by HPLC was in the range 0.18–0.42 mg g⁻¹ fruit fresh weight for the 12 cherry varieties.

The HPLC analyses showed that the 12 types of cherries contained hydroxycinnamic acids, hydroxybenzoic acids, anthocyanins, flavan-3-ols, flavonols and other flavonoids like taxifolin-O-deoxyhexosyl hexoside. Their amounts are listed in Table 3. The same compounds at comparable concentrations in cherries were reported previously. Crupi et al. (2018a) reported that p-coumaroyl and caffeoyl quinic acid derivatives are widely present in cherries, with methyl coumaroyl quinates at 6.1–56 mg/100 g FW, chlorogenic acid at 150 mg/100 g FW and methyl caffeoyl quinates at 24–110 mg/100 g FW. The highest value of the total hydroxycinnamic acid content (mg g⁻¹ fresh weight) was more than four times the lowest value, whereas, the highest value of the total anthocyanin content (mg g⁻¹ fresh weight) was 390 times the lowest value. A dark red colour is closely associated with the anthocyanin content of the cherries. The finding that cyanidin-3-O-rutinoside was the most abundant anthocyanin in the 12 types of cherries was in good agreement with prior studies (Usenik et al., 2008; Crupi et al., 2015). These results were obtained in other sweet cherry varieties. The highest value of the total flavonoids content (mg/g fresh weight) was six times the lowest value. Red Light and Rock Candy Cui had the highest amounts of catechin (~0.43 mg g⁻¹ fruit fresh weight), which was 14 times that of the variety with the lowest content (0.031 mg g⁻¹ fruit fresh weight, Tai Zhu). The total phenolic acid content (mg/g fruit fresh weight) was reasonably similar among the cherry varieties, with the highest value being only 2.7 times the lowest value. The salicylic acid content in the cherries might be endogenous (i.e. a growth regulator for fruit development) or from the pre-harvest and/or post-harvest application of salicylic acid to enhance fruit quality and bioactive contents (Hayat et al., 2010; Valero et al., 2011).

Quality assessment by SA, HCA and PCA

Similarity analysis

HPLC-PDA fingerprinting of 12 batches of cherry samples was carried out using the Chinese Medicine Chromatographic Fingerprint Similarity Evaluation System (Version 2004A) recommended by the State Food and Drug Administration (SFDA). The software is primarily used for synchronisation and quantitative comparisons between different samples to calculate the correlation coefficient of the entire chromatogram of the sample. The software was used to generate a simulated average chromatogram, and the similarity between the different chromatographic patterns was compared with the average chromatogram between the test samples, then the SA values between the chromatograms of the cherry samples and the reference chromatograms were calculated separately. Moreover, the relative retention time (RRT) and relative peak area (RPA) of each characteristic peak associated with the reference peak were also calculated to quantify the chemical properties in the chromatogram of the cherry. SA can reveal not only the similarity but also the dissimilarity among the HPLC fingerprints. The closer the similarity index or correlative coefficient to 1, the more similar the two HPLC chromatograms (Wei et al., 2010; Feng et al., 2014). In this study, the correlative coefficients among the HPLC chromatographic fingerprints of the sweet cherry samples at each wavelength varied with the cherry variety (ranging from 0.011 to 1.000, Table 4), which reflected the wide range of chemical compositions in the 12 cherry varieties. The characteristics shown by SA results were largely consistent with those revealed by visual observations (Fig. 1), physico-chemical analyses (Table 1), and chromatographic profiling (Figs S1–S5). Correlation coefficients larger than 0.9 (the appropriate threshold value for qualification) indicate a high similarity in the basic chemical composition (especially organic acids and phenolic compounds). Correlation coefficients < 0.9 suggest marked differences in chemical composition. Results allow the 12 cherry varieties to be clustered into groups for further analyses.

Hierarchical cluster analysis

Hierarchical cluster analysis (HCA) is a clustering method that transforms a complex set of sample structures into a unique, mutually exclusive group of subjects that, for some characteristics, are similar to each other (Zhou et al., 2015; Granato et al., 2018). After a suitable Centroid Euclidean distance was chosen, dendrograms at different wavelengths were generated (Fig. 2) to reveal the relationship among the 12 sweet cherry varieties. The varieties with high similarities essentially have shorter distances between one another, and vice versa. As shown in Table 5, the HCA results were consistent with the results of the SA. The grouping results of the 12 varieties of cherry depended on the detection wavelength, confirming that the cherry varieties showed differences in the classes of compound detected at each wavelength.

HCA at 210 nm revealed that Reigyoku, Tai Zhu and Bright Pearl differed (likely in organic acid composition), with these three varieties being different to the rest of the cherry varieties. Some varieties were grouped into the same cluster at both 210 and 280 nm: 13–33, Rainier, Red Light, Aaron Brooks and Beautiful Red; Shandong Jade and Pleasurable Red; Red Honey and Rock Candy Cui. Cherries that grouped together at both 210 nm and 330 nm were: Red Light, Shandong Jade, Aaron Brooks and Beautiful Red; 13–33, Rainier, Red Honey and Rock Candy Cui. Cherries that grouped together at both 210 and 370 nm were as follows: Red Light, Red Honey, Aaron Brooks, Shandong Jade and Pleasurable Red, 13–33, Beautiful Red and Rock Candy Cui. Cherries that grouped together at both 210 and 520 nm were 13–33 and Pleasurable Red. These results imply a relationship between the chromatograms at 210 nm and those at the other four wavelengths. Accordingly, the clustering results at 210 nm did not exclude contributions from other non-organic acid species.

The cherries that grouped in the same cluster at both 280 and 330 nm were as follows: 13–33, Rainier and Bright Pearl; Reigyoku, Red Honey and Rock Candy Cui; Red Light, Aaron Brooks and Beautiful Red. The cherries that grouped together at both 280 nm and 370 nm were 13–33 and Beautiful Red; Rainier and Bright Pearl; Tai Zhu, Shandong Jade and Pleasurable Red; Reigyoku and Red Honey. The cherries that grouped together at both 280 nm and 520 nm were 13–33 and Bright Pearl; Tai Zhu and Shandong Jade; Rainier, Red Light and Beautiful Red. Accordingly, the clustering results at 280 nm were associated with those at 330, 370 and 520 nm, as a range of phenolic compounds (including phenolic acids, flavanones, flavan-3-ols and anthocyanins) absorb at 280 nm.

The cherries that grouped together at both 330 and 370 nm were as follows: Red Light, Shandong Jade and Aaron Brooks; 13–33 and Rock Candy Cui; Rainier and Bright Pearl; Tai Zhu, Reigyoku and Red Honey. The cherries that grouped together at both 330 and 520 nm were Red Light and Beautiful Red; Shandong Jade and Aaron Brooks; 13–33 and Bright Pearl; Tai Zhu and Reigyoku; Rainier, Red Honey and Rock Candy Cui. Furthermore, the cherries that grouped together at both 370 and 520 nm were Tai Zhu, Reigyoku, Aaron Brooks and Shandong Jade; Red Light and Red Honey; Beautiful Red, Rock Candy Cui. Accordingly, the clustering results at 330, 370 and 520 nm were related, containing contributions from cinnamic acid derivatives and other compounds including certain flavonoids that absorb at 330 nm. The cherries likely share some common anthocyanins and non-anthocyanin flavonoids. Interestingly, Red Light which had dark red coloured skin was not grouped in the same cluster as the other four types of cherries with dark red coloured skins. No varieties appeared in the same group at the five wavelengths analyzed, suggesting the 12 cherry varieties showed a degree of difference in their organic acid and phenolic profiles. The complexity of the clustering results indicate that the cherries showed similarities and differences in their fruit chemical composition, in particular differences in their organic acid and phenolic compound profiles. The similarities and differences in the cherry fruit profiles also manifests in their sensory attributes, including appearance and flavour, and is also expected to depend somewhat on the degree of fruit maturity (fruit type and growing conditions including unavoidable differences such as growing location of cherry trees and position of the fruit on the cherry tree).

Principle component analysis

Given the varied contents of organic acids and phenolic compounds in the 12 cherry varieties, PCA is a method of feature extraction and dimensionality reduction, which makes the difference in the sample magnified and easily distinguishes the sample(Chen et al., 2008; Da et al., 2012). In this study, the PCA revealed the variation patterns of cyanidin-3-O-rutinoside, p-coumaroylquinic acid, neochlorogenic acid and malic acid (Fig. 3). Three different classes were separated based on phenolic profiles at 280, 330, 370 and 520 nm, while four classes were separated at 210 nm (a result in agreement with the HCA finding). A clear discrimination among the cherries at 210 nm was found, with the three obvious outliers, 9 (Reigyoku), 11 (Bright Pearl) and 12 (Tai Zhu) spreading in different clusters while the rest was grouped closely in the same cluster. This result indicated that the cherry varieties differed mainly in organic acid composition, especially malic acid. Compared with the plots at 280 and 370 nm, a slightly less scattered pattern was found at 330 nm, suggesting smaller variation in the hydroxycinnamic acid contents. The PCA results obtained at 280, 330 and 370 nm all showed three pairs of cherries: 8 (Rainier) and 11 (Bright Pearl); 3 (Aaron Brooks) and 5 (Red Light); and 6 (Red Honey) and 9 (Reigyoku). These pairings indicate a similar non-anthocyanin phenolic profile for each pair. However, all these pairs were discriminated in the PCA at 520 nm, suggesting the varieties possessed very different anthocyanin profiles. Cyanidin-3O-rutinoside seemed to be an index for grouping similar varieties with similar colours, although Red Light with a dark-red colour was classified as an outlier (probably due to its low anthocyanin content compared with the other dark-red coloured cherries). A large drift was observed in certain cherry varieties, implying a considerable difference in the phenolic composition compared to other varieties.

Conclusion

The 12 cherry varieties used in this study differed widely in their physico-chemical characteristics, juice yields and chemical composition. These cherries can be classified into four groups based on their skin colour. The cherry varieties also differed in their Brix to acid ratio. Tai Zhu and 13–33 showed the highest and the lowest TPC, respectively. Shandong Jade had the highest TFC, while Rock Candy Cui had the lowest. Red Light had the highest TMA with Beautiful Red having the lowest.

The SA, PCA and HCA results revealed quality similarities and differences among the 12 cherry varieties. There may be approximately 20 phenolic compounds and 10 organic acids in the 12 cherry varieties. Among these compounds, malic acid, cyanidin-3-O-rutinoside, p-coumaroylquinic acid, neochlorogenic acid and rutin appeared to be the characteristic compounds for the sweet cherries. The anthocyanins and flavonol patterns could be useful for varietal assignment of sweet cherries, and cyanidin-3-O-rutinoside, as the major anthocyanin, might be a preliminary index for grouping similar cultivars and colours. To define the exact and easy-to-identify markers for supporting cherry quality control, further studies are necessary to investigate the effects of pre-harvest, harvest and post-harvest variables including agro-climatic factors on the compositions of sweet cherries.

Acknowledgements

The authors acknowledge the general assistance from A./Professor Feng Li, and the support from National Key Research and Development Program (2016YFE0130900).

Conflict of interest

The authors declare no conflict of interest.

References