Chemicals and reagents

High-performance liquid chromatography (HPLC) grade methanol and acetic acid were purchased from Merck Co. (Darmstadt, Germany). Standard phenolic compounds: coumaroylquinic acid, protocatechuic acid, neo-chlorogenic acid, chlorogenic acid, catechin, epicatechin, caffeic acid, p-cumaric acid, ferulic acid, phloretin, phlorizin, quercetin, quercitrin, quercetin-3-O-galactoside, procyanidin B1 of HPLC grade and bought from Sigma-Aldrich^® Co. (St. Louis, Missouri, USA). Folin-Ciocalteu reagent, analytical hydrochloride, sodium hydroxide obtained from Aladdin^® Industrial Corporation (Fengxian, Shanghai, China). Ultra-pure water provided by ULUP-Ⅰ microanalysis system (Xi’an, China) was used in all the experiment.

Apple samples

Apples of 7 varieties were collected in October 2017, which were provided by the national apple breeding research base of China in Xingcheng, Liaoning Province (N40°33’ E120°45’). The selected seven varieties include FS, QG, QP, JS, CMO, GL, and HH. The apples were packed in plastic atmosphere bags in shock proof cartons, and stored at 4°C, and 85% relative humidity refrigerator until use. For each apple variety, 5 fruits were segmented into three parts of core, pulp, and peel (seed was not considered in present study), which were instantly freezed in liquid nitrogen, and subsequently smashed into powder. The core, pulp, and peel freeze powder were then stored at –80°C for further analysis.

Sample preparation for metabolomic analysis

About 10 g frozen powder of the core, pulp, and peel of each variety of apple was taken, mixed with 1.0 mL 2% NaCl solution and 10 mL 80% methanol solution. The samples were extracted by using ultrasonic under 45 kHz, < 30°C, for 20 min. (Ultrasonic machine, SB 25-12 DTN, Scientz Biotechnology Co.,Ltd, Ningbo, China) . The supernatant was centrifuged (10,000 rpm, 4°C, 10 min) by using centrifuge (5804R, Eppendorf, Hamburg, Germany) before being collected. The residue part was then added 50 µL of 37% HCl and 10 mL of 80% methanol for ultrasonic-assisted extraction and collected the supernatant after centrifugation. The residue was further added 100 µL 5 mol/L NaOH, and 10 mL of 80% methanol for ultrasonic-assisted extraction, and collect the supernatant after centrifugation. Combined the neutral, acidic, and alkaline supernatant extracts, and dried by nitrogen blow at the temperature of less than 30°C. The dry matter was then dissolved in 5 mL methanol and stored at –80°C for further analysis.

Total polyphenol content determination

The Folin-Ciocalteu method was used for total polyphenol concentration determination, based on the method established previously [18] with some modifications. Gallic acid was used as standard and ultra-pure water was used as blank. Briefly, 1.0 mL of diluted apple extracts (apple extracts were diluted by distilled water for 100-fold) was mixed with 0.5 mL of Folin-Ciocalteu reagent before adding 3 mL of sodium carbonate solution (10%, weight/volume). Water was added to bring the volume to 10 mL. The absorbance was read at 760 nm after 30 min reaction in dark place. Results were expressed as mg gallic acid equivalent (GAE) per gram sample. Each sample was tested in triplicate.

Polyphenol composition determination

HPLC analysis of polyphenol composition was performed on a Waters^® 2695 HPLC system with ultraviolet-photodiode array (UV-PDA) detector (HLPC-PAD, Waters, Milford, USA), the wavelength of which was set at the range of 200–600 nm. The system equipped with a Promosil^® C18 column (4.6 mm × 250 mm, 100 A, 5 µm, Agilent Technologies, China) operated at 30°C. Phase A was 2% acetic acid in ultra-pure water, and phase B was 100% methanol. The flow was set as 1.0 mL/min, and the injection volume was 10 µL. The gradient program was the optimized settings: 20 min, 5–25% B; 15 min, 25–40% B; 5 min, 40% B; 5 min, 40–95% B; 5 min, 95% B; 3 min, 95–5% B; and finally, 2 min, 5% B, which referred the previous method [19]. The system was controlled by Breeze software. Quantification was conducted by external standards using HPLC grade standard polyphenol compounds.

HPLC-ESI-Q-TOF-MS analysis

Analysis was performed on an ultra-performance liquid chromatography (ACQUITY UPLC^®, Waters Corp., Milford, MA) equipped with an Acquity UPLC^® BEH column C18 1.7 μm 2.1 mm × 100 mm, Waters Corp.) connecting to a quadrupole time of flight mass spectrometer (Xevo^® G2-S QTof, Waters Corp.). The detailed protocol was established in the reference [19]. The mobile solutions were water with 0.04% acetic acid (A) and methanol containing 0.04% acetic acid (B). The gradient program was set as follows, from 5–95% B, 11 min; 95% B, 1 min; 95–5% B, 0.1 min; 95% B, 2.9 min. The column was kept at 40°C and the flow rate was 0.3 mL/min. Mass spectra were acquired using electrospray ionization under positive mode over the range of m/z 100–1,700. The desolvation gas temperature was 550°C, the cone gas flow was 50 L/h, and the desolation gas flow was 1,000 L/h. The accurate mass and composition of the precursor were calculated and sequenced using the MassLynx^™ software (Waters Corp., Milford, MA). The stability of the method was tested by performing 6 repeated injections of solutions prepared from authentic reagents catechins and chlorogenic acid every 2 h. The relative standard deviation of the retention times was below 2% and the mass error was below 5 ppm. Quality control (QC) samples were prepared by mixing all the sample extracts, then divided into four QC samples and analyzed using the same method as the experimental samples. The QC samples were run every six samples throughout the analytical run to make sure the repeatability could be assessed.

Data processing and statistical

Chromatographic data acquired from the HPLC coupled with electrospray ionization quadrupole time-of-flight tandem mass spectrometry (HPLC-ESI-Q-TOF-MS) (waters Xevo^® G2) were directly imported to Progenesis QI V2.0 data processing platform, which adopts an intuitive workflow approach to perform data analysis. Automatic data processing was available in the form of licensing individual data files, running and reviewing alignment before experiment design, in which two ways of grouping (group by apple varieties and group by peel, pulp, and core of apple fruit) were used. Furthermore, parameter was set for compounds ions peak picking, and data normalization method was established to review normalization. Then we review and edit the various adduct and forms of a compound, and identify compound using progenesis MetaSCOPE^®, combined with database of NCBI, PUBMED filtered by conditions of ANOVA P value ≤ 0.05, variable importance in the projection (VIP) > 1, max fold change ≥ 20. The possible compounds were further analyzed by their identity and expression between conditions. Multivariate statistical analysis was conducted using principal component analysis (PCA), partial least-squares discriminant analysis (PLS-DA), and orthogonal partial least squares discriminant analysis (OPLS-DA) performed on EZinfo 3.0 (Umetrics^®, Umea, Sweden). Regression model was constructed by PLS to examine the relation between metabolite profiling and the score of evaluation.

Chromatographic data acquired from the LC-Q-TOF-MS (Waters Xevo^® G2) were imported to Waters UNIFI^® for metabolites identification and binary comparison. The analysis was based on the natural products database, which was established by collecting the polyphenols and related metabolites information from the recently published literatures [10, 14, 20, 21]. The analysis was conducted by matching the imported data file and the data processing method, in which blank, standard, QC, and experimental tests were labeled and set for three replicates. The characteristic components screening and difference constitutes comparison were processed automatically.

Morphological difference between the seven apple cultivars

Apple (Malus domestica) is the mostly consumed, cultivated, and investigated fruit, which is distinct in both the morphology and flavor among different apple varieties. Even though the seven apple varieties were cultivated in the same research base with the same cultivation approach, harvested time, and exactly the same storage method, the appearances of the same maturity of apple fruit were remarkably different from those in different apple varieties (Figure 1A). It has been reported that different apple varieties with distinctive morphology, texture, taste, flesh color, and also their polyphenol content and composition [18].

Display full size

Figure 1. 

(A) The morphology and cross-section color of 7 apple varieties; (B) total polyphenol distribution in core, pulp, and peel of apple fruit. Data are presented as mean ± SD of three independent experiments, different letters upon error bars indicate significant difference. P < 0.05

Identification and quantification of significant phenolic components

Polyphenols are the main secondary metabolites in apple fruit, the content varied in different apple cultivars and in different part of the apple. In present study, there was a large variation in total polyphenol content across the peel, pulp, and core of the seven apple cultivars (Figure 1B), ranging from 368.6 μg/g to 684.0 μg/g fresh weight (FW) in core; from 202.5 μg/g to 423.5 μg/g FW in pulp; and from 401.6 μg/g to 952.9 μg/g FW in apple peel. These results are in line with the total polyphenol content (363.9–2,516.9 μg/g FW) of peel in 145 apple cultivars for two years [22]. The difference was proposed to be induced by different sampling strategies, quantification approaches, as well as climate variation, agricultural practices, and location. As comparable to the previous research that provided the polyphenol level in the pulp of 66 apple varieties from the UK [23], 58 apple varieties from China [2], and 20 apples from Jena German [7]. Moreover, total polyphenol level was a significant difference among different cultivars. Total polyphenol of apple peel was found in descending order in GL, HH, QP, CMO, JS, FS, and QG, while total polyphenol level of pulp was in the descending order of GL, HH, JS, CMO, QP, FS, QG; and total polyphenol level of core was in descending order of GL, JS, QP, CMO, QG, HH, FS.

In the present work, the distribution of four groups of polyphenols, including phenolic acid, flavanols, dihydrochalcones, and flavonols in apple peel, pulp, and core were separately identified and quantified by using high-performance liquid chromatography-diode array detection (HPLC-DAD). Phenolic acid, including gallic acid, p-coumaric acid, protocatechuic acid, caffeic acid, ferulic acid, neochlorogenic acid, and chlorogenic acid were listed in Table 1. Flavanols including catechin, and epicatechin; two dihydrochalcones including phloretin and phloretin-2’-O-glucuronide, and four flavonols including quercetin, quercetin-3-O-rhamnoside, quercetin-3-O-galactoside, quercetin-3-O-rutinoside level were showed in Table 2. These results are in good agreement with the research that chlorogenic acids were then the main phenolic acids in different apple varieties [3, 8, 24].

Metabolomics profiling in three parts of apple fruit

Polyphenol composition difference among different apple cultivars is one of the major characteristics for distinguishing varieties, as discussed in previous literature [25]. Given that other components of apple Metabolic profiles differences in peel, pulp, and core of 7 apple varieties were evaluated by multivariate statistics analysis. PCA is a widely used chemometric tool for extracting and rationalizing information from any multivariate description of a biological system, and it is used to reveal the internal structure of several variables by a few principal components [26]. In the PCA plot (Figure 2) the QC samples grouped together, revealing that the QC samples had similar metabolic profiles and the entire analysis was stable and repeatable. PCA analysis indicated that there are 9 components that explained 80% of the total variance upon 10,259 variables. The samples of peel, pulp, and peel part of apple fruit grouped in 3 separate areas in the plot, among which the metabolic profiles of pulp and core were partially overlapped, indicating that peel had notably different metabolic profile from that of pulp and core, while the metabolic profile of the latter two shared some similarities.

Display full size

Figure 2. 

Multivariate modeling of variations using PCA, PLS-DA, and OPLS-DA of peel, pulp, and core of the 7 apple varieties. Score plots of the first t(1) and second t(2) predictive components are illustrated. On the score plot, the percentage of total variation explained by the component is indicated in parentheses. The colored ellipses include 95% of the multivariate normal distribution of the peel, core, and pulp samples respectively. On the loading plots of PCA, the variables with extreme values in each direction of principal component p(1) and p(2) were focused. On the plot of coefficients versus VIP, the variables with VIP value higher than 10 were focused. On the scores plot scatter (S)-plot, the predictive of P value (loadings) in the range of |P| > 0.1 were focused

Partial least-square analysis (PLS-DA) was used to generate pairwise comparisons of the metabolite contributions [27]. In present work, 7 components established for metabolic profile of peel, pulp, and core were compared to identify the metabolites that were responsible for the observed difference. As much as 88% of predictability (Q²) and strong goodness of fit (96%) on R²Y were observed in the PLS-DA models. Thereinto metabolic profiles of peel, pulp, and core were well separated from each other in three quadrants of the rectangular coordinates.

The OPLS-DA extends a regression of the PCA and uses the class membership to maximize the variation and introduce an orthogonal signal correction filter to separately handle the systematic variation correlated to or uncorrelated to the Y variable. Therefore, it has a better discriminant ability for the samples with larger within-class divergence than PCA, as has been illustrated by Bylesjö [28]. In the present study, 5 components were selected for samples from three parts of an apple that were clearly separated in the OPLS-DA score plot. The variance explained R²Y indicating 98% of the variation in the dataset could be modeled by the selected component. The variance prediction Q² reached 96%, indicating a high goodness of fit.

Metabolomics profiling in 7 different apple varieties

The metabolomics profile of 7 apple varieties (FS, QG, QP, JS, CMO, HH, GL) were comparatively analyzed (Figure 3). In PCA plot, with 80% variance explanation upon 10,259 variables, 9 components were found. Based on the predictive component 1, t(1) and 2, t(2), HH, GL, and QP were distinctively separated from FS and QG. Besides, CMO and JS showed the most overlap. In PLS-DA modeling, 2 predictive components were established and only 17% variance explanation and 6% variance prediction were found. While 7 apple varieties were grouped into 3 parts on predictive component 1. It suggests the different metabolomic profiles of peel, pulp, and core of GL were different from that of HH and QP, and also that of the rest apple varieties. In respect of OPLS-DA model, there were 5 components, explaining 98% variance and possessing 96% variance prediction between the two groups (one with GL, HH, QP, and the other including FS, QG, JS, CMO), indicating the metabolomics profile difference and relationship among the 7 apple varieties.

Display full size

Figure 3. 

Diagnostic of multivariate modeling (PLS-DA). The diagnostic plot shows the distances within and orthogonal to the selected score plan for PLS modeling

Study of key metabolite in peel, pulp, and core of apple

The loading plot (Figure 3) indicated the variables responsible for the difference of the metabolomic profiles of peel, pulp, and core of apple fruit, which contained 27 compounds. The variables’ importance distribution along the PLS regression coefficient showed in the plot of coefficients versus VIP, among which 11 variables that had VIP > 10, coefficient > 0.00005 were investigated. The important variables distribution along the OPLS regression coefficients were displayed in S-plot, in which 9 variables were located at the extreme value in two direction of predictive component 1 (Table 3).

The significantly different phenolic components between core and pulp were hyperin, there is more hyperin found in core as compared to the pulp. While the melilotoside, isorhamnetin-3-O-rutinoside, flavonoid glycoside, and rutin were the significant different phenols found between core and peel, among which melilotoside and flavonoid glycoside were higher in core than that in peel, isorhamnetin-3-O-rutinoside and rutin were lower in core than that in peel. quercetin-3-O-rutinoside and naringenin-7-O-glucoside were the phenolic compound found different in peel and pulp, in which Quercetin-3-O-rutinoside were more in pulp and naringenin-7-O-glucoside were in peel.

In respect of metabolomics profile difference attributed to apple varieties, the variables located in the area related to the metabolomics profile difference among seven varietals of apple slices on the basis of the loading plot. While the variables’ importance distribution along PLS regression coefficient of CMO, HH, and JS showed that 5 variables had VIP > 1 coefficient > 0.00005. There were 21 important variables distribution along the OPLS regression coefficients as shown in S-plot (Table 4).

The remarkable phenolic compound to differentiate GL and HH was 4-O-caffeoylquinic acid, in which GL possessed higher this specific phenol. Melilotoside, Chlorogenic acid, kaempferol, isorhamnetin were the significant phenols to differentiate the GL and QP, and only kaempferol was found higher in QP than in GL. 4-O-p-coumaroylquinic acid, chlorogenic acid, neochlorogenic acid showed different in GL and FS, in which neochlorogenic acid was higher in FS. Similarly, chlorogenic acid, neochlorogenic acid, 4-O-caffeoylquinic acid, and flavonoid glycoside were significant different in GL and QG, and only flavonoid glycoside was remarkably higher in QG than GL. Moreover, isorhamnetin was the significant phenolic compounds to distinguish GL and CMO. Whereas polyphenol could not necessarily differentiate all the apple varieties, which is closely dependent on the apple varieties.