Relationships between sensory properties and metabolomic profiles of different apple cultivars

Highlights

• •Metabolomic analysis was performed using GC − MS and HPLC − UV(−RI) in apples.
• •Sensory evaluation was conducted using general consumers and trained panels.
• •Through statistical correlation analysis, key metabolites linked to the flavor quality of apples were identified.
• •Collected information will be used as preliminary data for the quality control of apples.

Abstract

Flavor is a critical factor in apple quality. To better understand apple flavor, this study aimed to identify the relationships between sensory attributes and the chemical composition (volatiles and non-volatiles) of apples using a combined metabolomic and sensory evaluation. Sensory results showed the positive (apple, fruity, pineapple, sweetness, sourness) and negative (cucumber) flavor attributes of apples. A metabolomic analysis with statistical correlations revealed significant metabolites related to the flavor attributes of apples. Volatile esters (e.g., hexyl acetate and 2-methylbutyl acetate for apple and fruity notes) and non-volatile sugars and acids (total sugars, tartaric acid, and malic acid for balanced sweet and tart flavors) were associated with the apple flavor preferred by consumers. Some aldehydes and alcohols (e.g., (E)-2-nonenal) contributed to a negative hedonic perception (cucumber). The collected information demonstrated the roles of key chemical compounds in apple flavor quality, and may be applicable to quality control.

1. Introduction

Apple (Malus domestica Borkh.) is one of the most popular fruits in the world because of its high sensory quality and nutritional value (Harker et al., 2003). A pleasant combination of sweet and tart flavors is a major factor in apples being consumed globally (with a world production of more than 80 million tons) as fresh fruit, juice, dried products, alcoholic beverages, or processed products such as candies and desserts (FAO, 2018, Harker et al., 2003). The flavor quality of an apple is ultimately driven by the chemical composition (metabolic makeup) of a given fruit, which can provide an important clue for the quality control of apples (e.g., preservation of a desirable flavor during storage), and their breeding selection. The sweetness and sourness of an apple primarily depend on the amount of sugars and organic acids (F. R. Harker et al., 2002). The characteristic aroma of an apple is attributed to a variety of volatile compounds and their mixtures (Espino-Díaz et al., 2016). With more than 3,500 apple cultivars currently reported, there is great diversity in apple flavor traits, including sweetness, acidity, and aroma, which affect the overall quality of apples (Volk & Henk, 2016).

With the rapid improvement of instrument technologies, analytical methods such as high-performance liquid chromatography (HPLC) with ultraviolet (UV) and refractive index (RI) detection (for non-volatiles, e.g., sugars, organic acids), and headspace solid-phase microextraction (HS-SPME) combined with gas chromatography − mass spectrometry (GC − MS; for volatiles) have been increasingly used for chemical analyses in food and agricultural science (Robbins, 2003). With the application of these instrumental analyses, sensory-related compounds in fruits have largely been unveiled, such as sugars in banana, mangosteen, kiwifruit and cherry, organic acids in citrus fruits, and volatile compounds (e.g., esters, terpenes, aldehydes, alcohols) in various different fruits (strawberry, grape, banana, peach, pineapple, citrus, litchi, mango, etc.) (Yun et al., 2022). The chemical composition of apple fruits has also been investigated, including their volatile and non-volatile compounds (Aprea et al., 2011, Ceci et al., 2021, Chitarrini et al., 2021, Cuthbertson et al., 2012, Heinmaa et al., 2017, Lee et al., 2017, Roberts and Spadafora, 2020, Xie et al., 2021). More than 300 volatile compounds alone have been identified in apples to date (Espino-Díaz et al., 2016). A chemical approach called “metabolomics” has significantly contributed to the characterization of these components because it aims at the comprehensive profiling of the chemical compounds (metabolites) in an organism (Aprea et al., 2011, Ceci et al., 2021, Cuthbertson et al., 2012, Xie et al., 2021, Zhou et al., 2012). With these achievements, some of the identified compounds were found to be involved in health effects, the olfactory sense (aroma), and other quality properties (storage, texture, geographic regions, post-harvest processing, etc.) of apples (Ceci et al., 2021, Cuthbertson et al., 2012, Heinmaa et al., 2017, Lee et al., 2017, Roberts and Spadafora, 2020, Xie et al., 2021). In relation to apple flavors, the majority of previous research focused on volatile compounds and related aroma qualities (Chitarrini et al., 2020, Roberts and Spadafora, 2020). However, the linkage between the overall chemical composition (volatiles and non-volatiles) and sensory attributes of apples has remained unclear, although their unique flavor profile is likely derived from the integration of multiple components in apples, including sugars and acids (non-volatiles) with volatile emission (Espino-Díaz et al., 2016; F. R. Harker et al., 2002). Indeed, many volatile aroma compounds are generated from and correlated with non-volatile compounds (flavor precursors), and some non-volatiles (e.g., sugars, organic acids) are taste molecules that directly influence the flavor quality of apples (Espino-Díaz et al., 2016; F. R. Harker et al., 2002). Hence, a metabolomic approach covering a diverse range of these metabolites would be helpful in better understanding the complex flavor traits and related quality of apples.

The aim of this work was to identify the relationships between the sensory attributes and chemical profiles (volatiles and non-volatiles) of apples using a joint metabolomic approach and sensory evaluation. We hypothesized that there are key chemical compounds responsible for specific sensory (flavor) properties of apples, which influence their quality and consumer acceptance. To prove this, (1) a sensory evaluation using general consumers and trained panels was conducted to determine the hedonic perception and flavor attributes of three apple cultivars. (2) A metabolomic analysis was performed to assess the major chemical constituents of apples, including volatiles and non-volatiles (sugars, organic acids). Lastly, (3) a statistical correlation analysis was utilized to elucidate the roles of these chemical components in the hedonic perception and flavor quality of apples.

2. Materials and methods

2.1. Chemicals

Volatile reference standards (ethanol, butyl acetate, hexanal, 2-methylbutyl acetate, butanol, amyl acetate, geranyl propionate, butyl butyrate, 2-methyl-1-butanol, butyl 2-methylbutyrate, (E)-2-hexenal, 2-pentylfuran, hexyl acetate, 2-methylbutyl 2-methylbutyrate, acetoin, amyl hexanoate, hexanol, nonanal, (E)-2-hexenol, hexyl 2-methylbutyrate, cyclohexanol, (E)-2-hexenyl butyrate, acetic acid, furfural, decanal, octanol, (E)-2-nonenal, 2,3-butanediol, hexyl hexanoate, furfuryl alcohol, estragole, 5-hydroxymethylfurfural, and 5-acetoxymethyl-2-furaldehyde) and non-volatile reference standards (glucose, fructose, sorbitol, sucrose, malic acid, oxalic acid, tartaric acid, and citric acid) were purchased from MilliporeSigma (St. Louis, MO, USA). Acetonitrile, isopropyl alcohol, and water were HPLC grade and purchased from Thermo Fisher Scientific (Fair Lawn, NJ, USA). All other reagents were of analytical grade.

2.2. Apple samples

“Gamhong,” “Yangwang,” “Hongro,” and “Fuji” are the major apple cultivars grown in Korea (KOSIS, 2016). All of the apple samples were grown in the same province (Gyeongsangbuk-do) of South Korea. The fruits of three of these apple cultivars (“Gamhong,” “Yangwang,” “Hongro”) were harvested in the 2020 and 2021 growing seasons from September to October from orchards in Mungyeong-si, Yecheon-gun, and Yeongdeok-gun, respectively. A globally consumed cultivar (“Fuji”) was harvested as a reference material in November 2020 and November 2021 from orchards in Yeongdeok-gun. In each harvest, a total of six fruits were randomly collected from three different trees (two harvest times and three biological replicates; n = 12) for each cultivar. All of the trees were mature (greater than5 years) and of similar size (a tree height of 3–4 m). The fruits were selected for their uniform size and freedom from defects (Fig. S1). All the fruits were picked at commercial maturity, when the fruit stalks were easily turned and separated from the tree by hand. The samples were shipped on that day to the Department of Food Science and Biotechnology, Andong National University (Andong, 36729, Gyeongsangbuk-do, South Korea). The samples were washed with sodium bicarbonate to remove contaminants from the surface, and were stored at 4 °C with 90–95% relative humidity until the sensory evaluation. On the day before the sensory evaluation, the apples were washed, peeled, and cut into uniform cubes (1 × 1.5 × 1.5 cm). The cubes from same cultivars were treated with an antioxidant solution (0.02% citric acid, 0.02% ascorbic acid, and 0.05% calcium chloride) and stored at 4 °C until the sensory analysis (Corollaro et al., 2013). For the physicochemical and metabolomic analyses, the samples were quenched by adding liquid nitrogen, and homogenized into pastes using a blender. For the HPLC-based metabolomic analysis, the samples were quenched by liquid nitrogen, and then freeze-dried for 48 h using a freeze dryer (FDTA-4504, Operon Co. Ltd., Korea). The freeze-dried flesh was ground using a mortar and pestle under liquid nitrogen. All the samples for the physicochemical and metabolomic analyses were kept at −80 °C until used.

2.3. Sensory evaluation

The study protocol and consent procedure received ethical approval from the Andong National University Institutional Review Board (IRB approval number; 1040191–202006-HR-010–01).

2.3.1. Consumer study

A total of 100 general consumers (ages 37 ± 17 years; 55 males and 45 females) were recruited in 2020 and 2021. The tests took place in designed sensory booths under room temperature (22 °C) and artificial light. The panelists received instructions about the scaling systems and procedures before the test, and rated samples on anchored line scales (general labeled magnitude scale, gLMS) (Linda M. Bartoshuk et al., 2005, Kalva et al., 2014). For the hedonic rating, the scales assessed the overall liking, color liking, and flavor liking of apples in the context of all pleasure/displeasure experiences (−100 = strongest disliking of any kind experienced; 0 = neutral; and + 100 = strongest liking of any kind experienced). For the sensory intensity rating, the scales assessed the taste (sweetness and sourness), aroma (apple flavor), and texture (firmness, juiciness, and crispness) sensations for apples in the context of all imaginable sensory experiences (0 = no sensation and 100 = strongest sensation of any kind experienced) (L. M. Bartoshuk et al., 2003). “Fuji” (reference material) was not included in the 2020 consumer study. All the cultivars were evaluated in the 2021 consumer study.

2.3.2. Quantitative description analysis (QDA)

Thirteen trained panelists (ages 23 ± 3 years; 3 males and 10 females) participated in training and test sessions in 2020 and 2021. The training sessions familiarized the panelists with the descriptions used for flavor attributes. Ten flavor attributes (taste: sweetness and sourness; and flavor: pineapple, apple, pear, honey, cucumber, green, floral, and fruity) were chosen, and reference standards (Table S2 in Supporting Information file) were given for their proper evaluation. Three trial replicates were tasted on three separate tasting days (two harvest times, three trial replicates; n = 6). In the test sessions, the trained panelists assessed the flavor attributes of samples using a 15-point scale (0 = low intensity, 15 = high intensity) (L. M. Bartoshuk et al., 2003).

2.4. Physicochemical analysis

The color of the apple flesh was measured according to the L*a*b* coordinate system (L: lightness, a: green to red, b: blue to yellow) using a calibrated colorimeter (Spectrophotometer; CM-3500d, Minolta Co., Ltd., Osaka, Japan). The °Brix (total soluble solids, TSS) and titratable acidity (TA, percentage malic acid) were determined using a refractometer (Digital Hand-held Pocket Refractometer; PAL-1, Atago, Tokyo, Japan).

2.5. Metabolomic analysis

2.5.1. Volatile profiling

The volatiles of apple paste samples were extracted using HS-SPME and analyzed using GC–MS and GC–MS/olfactometry (O). Seven grams of the apple paste sample were placed into a headspace vial containing an internal standard (cyclohexanone, 20 μg/mL in methanol) and 1 mL of distilled water, and incubated in a thermostated autosampler tray at 50 °C for 10 min before the HS-SPME. The headspace volatiles were captured on a divinylbenzene/polydimethylsiloxane (DVB/PDMS) fiber (65 μm film thickness, Supelco, Bellefonte, PA) for 20 min, and thermally desorbed at 240 °C for 15 min in the GC injector port (splitless mode). The GC − MS (QP 2020, Shimadzu, Kyoto, Japan) system was operated in the electron impact (EI, ionization energy: 70 eV) ion mode with a scan range of m/z 30–300. The volatiles were separated on an HP-FFAP capillary column (50 m × 0.32 mm i.d., 0.5 μm film thickness; Agilent, Santa Clara, CA, USA). The GC oven temperature program was as follows: start at 40 °C for 2 min, increase from 40 °C to 64 °C at a rate of 2 °C/min (after a 1 min hold), then increased from 64 °C to 200 °C at a rate of 4 °C/min (after a 10 min hold), and finally hold at 230 °C at a rate of 10 °C/min for 13 min. Helium gas was used as a carrier gas with a constant flow rate of 1.0 mL/min. The transfer line temperature was kept at 230 °C. GC–MS/O with a sniffing port was also used to measure the quality and intensity of the aroma emitted by the apple samples. The GC–MS/O system consisted of an Agilent 7890A/5977B MSD Series (Agilent Technologies, Santa Clara, CA, US) equipped with a GERSTEL MPS 2XL autosampler (Mülheim an der Ruhr, Germany and the Olfactory Detection Port (ODP4, Gerstel). The instrument conditions (column, temperature gradient profile, etc.) were set the same as the GC–MS conditions described above. A series of n-alkane (C7-C30) standards was employed to determine the linear retention index (RI) of each compound. The volatiles were identified by matching the mass spectra from the NIST library (ver.14) and RIs of authentic standards. When standards were not available, experimental RIs from the literature were used instead. The relative metabolic responses were obtained by calculating the peak area ratio between an analyte and the internal standard in each sample.

2.5.2. Non-volatile profiling

For a sugar analysis, 200 mg of freeze-dried apple powder was mixed with 10.0 mL of 50% acetonitrile. Sugars were extracted by vortexing (5 min), followed by ultrasonic-assisted extraction (30 min at 50 °C). After centrifugation (3,000 rpm, 15 min), the supernatant was filtered through a 0.45 μm nylon filter, and an aliquot (10 μL) was injected into an HPLC − RI system (Waters, Milford, MA, USA). The sugars were separated on an Asahipak NH2P-50 column (250 × 4.6 mm, 5 μm particle size; Shodex, Tokyo, Japan) using acetonitrile/water (75:25, v/v) as a mobile phase, with a flow rate of 1 mL/min and a column temperature of 40 °C. The sugars were identified by matching the retention times with authentic standards.

For the organic acid analysis, 200 mg of the freeze-dried apple powder was mixed with 10.0 mL of phosphoric acid (pH = 2) in water. The organic acids were extracted by vortexing (5 min) followed by ultrasonic-assisted extraction (20 min) on wet ice. After centrifugation (12,000 rpm, 20 min), the supernatant was filtered through a 0.2 μm nylon filter, and an aliquot (20 μL) was injected into an HPLC − UV system (Osaka soda, Osaka, Japan) detector at 220 nm. The organic acids (non-volatiles) were separated on a C18 column (cadenza CD-C18, 250 × 4.6 mm, 3 μm particle size; Imtakt Corp., Kyoto, Japan) using phosphoric acid/water (pH 2) as a mobile phase. The flow rate was 0.6 mL/min, the run time was 30 min, and the column temperature was maintained at 30 °C. The organic acids were identified by matching the retention times with authentic standards.

2.6. Statistical analysis

All the experimental (from the sensory and metabolomic analyses) data were analyzed by a one-way analysis of variance (ANOVA) followed by a Tukey honestly significant difference (HSD) test using a statistical analysis system (SAS 9.4 Institute Inc., Cary, NC, USA). Variables with a p-value < 0.05 (from ANOVA) were considered potential markers, which could discriminate different groups. Pearson’s correlations were determined and visualized as a heatmap using Jamovi 1.6.23 software (The Jamovi Project, Sydney, Australia). A principal component analysis (PCA) was conducted using SIMCA-P ver. 17 (Umetrics, Umea, Sweden) software to assess the apple cultivars based on the obtained data and similarities or differences between sample profiles (Chong & Jun, 2005). The performance of the PCA model was evaluated using R² and Q² coefficients, which were defined as the proportions of the variance in the data explained and predictable by the model, respectively (Bro & Smilde, 2014). R² values close to one indicated an excellent model, and Q² greater than 0.5 was considered to indicate good predictability (Lin et al., 2011). A partial least-squares (PLS) regression analysis was performed using SIMCA-P ver. 17 software to identify the relationships between the metabolic response (independent variables, x) and sensory attributes (dependent variables, y) in the apple cultivars (Xue et al., 2010). The PLS model was validated using cross-validation parameters (R² and Q²) from a random permutation test (n = 200).

3. Results

3.1. Physicochemical characteristics of apple cultivars

The physicochemical properties of different apple cultivars (“Gamhong,” “Yangwang,” “Hongro,” and the reference material “Fuji”) harvested in 2020 and 2021 (2 years; n = 6) are presented in Table 1. The flesh color of the apples was typically white-yellowish. Overall, there was no significant color parameter difference between the cultivars, although “Hongro” showed a relatively intense red flesh color (a* value) compared to other cultivars such as “Gamhong” and “Yangwang.” The °Brix (total soluble solids, TSS) and titratable acidity (TA, percentage malic acid) values, and their ratio (TSS/TA), are often used as indicators of apple juice quality, as well as fruit maturity (Musacchi & Serra, 2018). In our study, “Gamhong” had the highest °Brix value (15.26), and “Hongro” showed a markedly lower °Brix value (12.44) than the other cultivars (Table 1, Fig. S1B).

Table 1.

Physicochemical characteristics of four apple cultivars.

Parameter	Cultivar (Mean ± standard deviation)
	‘Gamhong’	‘Yangwang’	‘Hongro’	‘Fuji’ (reference)
TSS (oBrix)	15.26 ± 1.07a	13.37 ± 0.65bc	12.44 ± 0.69c	14.24 ± 0.67ab
TA (%)	0.22 ± 0.03ab	0.25 ± 0.03ab	0.21 ± 0.01b	0.26 ± 0.03a
TSS/TA ratio	69.94 ± 9.72a	54.27 ± 7.71b	60.96 ± 3.14ab	55.79 ± 6.51b
L* valuens	76.34 ± 8.08	69.23 ± 15.73	82.07 ± 5.65	73.11 ± 5.96
a* value	2.00 ± 1.57b	2.00 ± 0.88b	6.78 ± 4.87a	2.62 ± 1.92ab
b* valuens	17.90 ± 4.49	19.40 ± 4.37	24.10 ± 13.78	22.71 ± 4.85

Different letters (a-c) in the same row indicate statistically significant differences (Tukey’s HSD, n = 12, p-value < 0.05); ^nsNot significant.

Fig. 1.

Radar chart showing the flavor profiles of apple cultivars (A) and PCA scatter plot of relationships between sensory attributes and cultivars (R²X = 0.763, Q² = 0.500) (B) based on the quantitative description analysis (trained panels, n = 13).

3.2. Sensory characteristics of apple cultivars

Sensory data for each cultivar were collected through the general consumer study in 2020 and 2021 (2 years; n = 100). The results of the consumer evaluations in 2020 and 2021 are presented in Table S2 and Table 2, respectively. “Gamhong” achieved the highest scores for hedonic liking (flavor liking), texture (juiciness), taste (sweetness), and flavor (apple flavor) in both 2020 and 2021, while “Hongro” consistently obtained low scores for evaluations such as hedonic liking (flavor liking, color liking), texture (juiciness), taste (sweetness, sourness), and flavor (apple flavor). Based on the above results, the overall liking of the apple cultivars (2020 and 2021) was ranked in the following manner: “Gamhong” > “Yangwang” > “Hongro” (p < 0.05 between “Gamhong” and “Hongro”) (Table S2, Table 2). A heatmap showing the relationships between the sensory variables based on Pearson’s correlations demonstrated that the overall liking of the apples was positively correlated with the color liking (0.61), flavor liking (0.79), sweetness (0.7), and apple flavor (0.63), supporting the consumer preference results for the different cultivars (Fig. S2). Notably, the overall liking of “Gamhong” was comparable to that of “Fuji” (reference material), which is globally recognized as an apple cultivar of high quality (Veberic et al., 2007).

Table 2.

Sensory characteristics of apple cultivars assessed by consumer test in 2021.

Sensory variables	‘Gamhong’	‘Yangwang’	‘Hongro’	‘Fuji’ (reference)
Hedonic liking
Color liking	41.14a	42.04a	22.90b	37.06a
Flavor liking	40.83a	26.99b	27.39b	46.53a
Overall liking	50.52a	31.66b	26.79b	52.52a
Texture
Firmnessns	37.56	42.76	37.65	40.27
Crispnessns	44.68	44.02	40.85	48.21
Juiciness	48.33a	37.61b	38.16b	51.28a
Taste
Sweetness	45.59a	29.68b	32.94b	49.11a
Sourness	20.19bc	22.89ab	15.53c	27.11a
Flavor
Apple flavor	42.20a	32.45b	32.53b	47.89a

Different letters (a-c) in the same row indicate statistically significant differences (Tukey’s HSD, n = 100, p-value < 0.05); ^nsNot significant.

The sensory attributes of the apples were evaluated and collected by trained panelists (n = 13) through the quantitative description analysis (2020 and 2021; 2 years). The sensory results are presented in Fig. 1. The results of the description analysis provided detailed information on the apple flavors liked/disliked by consumers. The radar chart for flavor profiles of four apples showed that the consumer-preferred cultivars (“Gamhong” and “Fuji”) were characterized by intense sweet, sour, apple, pineapple, and honey flavors, with a weak cucumber-like flavor (Fig. 1A). “Hongro” showed the opposite results. “Hongro” was characterized by an intense cucumber-like flavor, with weak sweet, sour, apple, pineapple, fruity, and honey flavors. Based on these results, statistically significant differences (p < 0.05) in the attributes (sweetness, sourness, apple, pineapple, cucumber, honey, etc.) were observed in the different cultivars. A PCA scatter plot of the description analysis results clearly illustrates the relationships between the sensory attributes and apple cultivars (Fig. 1B). The plot was built with the first two principal components (PCs), including PC1 and PC2, of the total variance (PC1 = 65% and PC2 = 11.3% in Fig. 1B).

3.3. Chemical (metabolomic) diversity of apple cultivars

The volatile and non-volatile (sugars, organic acids) compounds of the apple cultivars were identified and quantified using a metabolomic approach. The results of the volatile aroma analyses in 2020 and 2021 (2 years; n = 12) using HS-SPME followed by GC − MS and GC − MS/O are listed in Table 3. A total of 61 of the most abundant volatiles were found in the apple cultivars. These compounds included 26 esters, 14 alcohols, 8 aldehydes, 4 ketones, 3 furans, 3 acids, 1 terpene, and 2 others. The consumer-preferred cultivars (“Gamhong” and “Fuji”) mainly contained higher amounts of esters (hexyl acetate, amyl acetate, hexyl hexanoate, hexyl butyrate, butyl 2-methylbutyrate, butyl butyrate, butyl acetate, 2-methylbutyl acetate, 2-methylbutyl isovalerate, isobutyl butyrate, isoamyl isobutyrate, amyl butyrate, and 2-methylbutyl butyrate) compared to the other cultivars (Table 3). Among these, hexyl acetate, amyl acetate, 2-methylbutyl acetate, isoamyl isobutyrate, and 2-methylbutyl butyrate were exclusively found in these two cultivars (“Gamhong” and “Fuji”), and the aroma (fruity, apple flavor) of hexyl acetate and 2-methylbutyl acetate was clearly detected via the GC − MS/O analysis (Table 3). Meanwhile, “Hongro” had relatively higher amounts of other class volatiles, such as (E)-2-nonenal, 2-methyl-1-butanol, isobutanol, furfuryl alcohol, and furfural (Table 3). Among these, (E)-2-nonenal and isobutanol were only detected in “Hongro.”.

Table 3.

Volatile compositions of apple cultivars.

Compounds	RIA	Aroma descriptorB	Mean ± standard deviation, μg/kg				IdentificationC
			‘Gamhong’	‘Yangwang’	‘Hongro’	‘Fuji’ (reference)
Esters
Ethyl butyrate	1047	Apple	ndD	nd	nd	0.006 ± 0.020	MS, TI
Butyl acetate	1105	Pear	0.105 ± 0.092ab	ndb	0.005 ± 0.019b	0.064 ± 0.086a	FI, MS, TI
Isoamyl isobutyrate	1130	Fruity, sweet	0.061 ± 0.114	nd	nd	0.018 ± 0.046	MS, TI
Propyl butyrate	1131	Fruity, pineapple	0.012 ± 0.040	nd	0.008 ± 0.028	0.026 ± 0.072	MS, TI
Propyl 2-methylbutyrate	1146	Fruity, sweet	nd	nd	nd	0.005 ± 0.009	MS, TI
2-Methylbutyl acetate	1154	Fruity	0.157 ± 0.192ab	ndb	ndb	0.365 ± 0.584a	FI, MS, TI, O
Isobutyl butyrate	1165	Sweet, pineapple, apple	0.001 ± 0.003	nd	nd	nd	MS, TI
Methyl hexanoate	1197	Fruity, pineapple	nd	nd	0.006 ± 0.012	0.005 ± 0.012	MS, TI
Menthyl isovalerate	1237	Fruity, sweet	nd	nd	nd	0.003 ± 0.010	MS, TI
Butyl 2-methylbutyrate	1266	Fruity, pineapple	0.042 ± 0.040ab	ndb	0.014 ± 0.018b	0.060 ± 0.066a	FI, MS, TI
2-Methylbutyl butyrate	1272	Fruity, sweet	0.009 ± 0.012a	ndb	ndb	0.000 ± 0.001b	MS, TI
Amyl acetate	1275	Banana	0.018 ± 0.028a	ndb	ndb	0.009 ± 0.017ab	FI, MS, TI
2-Methylbutyl isovalerate	1299	Apple, fruity	0.001 ± 0.003	nd	nd	nd	MS, TI
Geranyl propionate	1305	Tropical, floral	0.004 ± 0.012	nd	nd	nd	FI, MS, TI
2-Methylbutyl 2-methylbutyrate	1314	Apple, fruity	0.024 ± 0.045	nd	0.009 ± 0.024	0.003 ± 0.010	FI, MS, TI
Butyl butyrate	1320	Fruity, pineapple	0.178 ± 0.139a	ndb	0.007 ± 0.025b	0.059 ± 0.062b	FI, MS, TI
Amyl butyrate	1323	Pineapple, sweet	0.008 ± 0.012a	ndb	ndb	ndb	MS, TI
Isoamyl hexanoate	1324	Fruity	nd	nd	0.001 ± 0.002	nd	MS, TI
Hexyl acetate	1379	Apple, fruity	0.101 ± 0.141	nd	nd	0.201 ± 0.353	FI, MS, TI, O
Amyl hexanoate	1386	Fruity	nd	nd	nd	0.006 ± 0.013	MS, TI
Methyl octanoate	1397	Fruity, citrus	nd	nd	nd	0.001 ± 0.002	MS, TI
Hexyl butyrate	1454	Apple, fruity, pineapple	0.297 ± 0.281a	0.026 ± 0.040b	0.002 ± 0.008b	0.088 ± 0.149b	MS, TI
Hexyl 2-methylbutyrate	1463	Apple, fruity	0.119 ± 0.070	0.270 ± 0.349	0.081 ± 0.080	0.120 ± 0.133	FI, MS, TI, O
Methyl nonanoate	1499	Coconut, sweet	0.001 ± 0.002	nd	nd	nd	MS, TI
(E)-2-hexenyl butyrate	1587	Fruity	0.034 ± 0.084	nd	nd	nd	FI, MS, TI
Hexyl hexanoate	1651	Apple peel	0.015 ± 0.025	0.003 ± 0.009	nd	0.010 ± 0.023	FI, MS, TI
Alcohols
Ethanol	969	Sweet	nd	nd	nd	0.012 ± 0.042	MS, TI
Isobutanol	1105	Wine-like	nd	nd	0.002 ± 0.008	nd	MS, TI
Butanol	1184	Fruity	0.041 ± 0.049ab	0.084 ± 0.066a	0.045 ± 0.019ab	0.030 ± 0.048b	MS, TI
2-Methyl-1-butanol	1242	Wine, fusel oil-like	0.051 ± 0.067b	0.173 ± 0.191ab	0.217 ± 0.208a	0.083 ± 0.054ab	FI, MS, TI
Pentanol	1261	Wine-like	nd	0.004 ± 0.014	nd	0.000 ± 0.001	MS, TI
Hexanol	1389	Green, apple	0.431 ± 0.215	0.928 ± 0.544	0.677 ± 0.461	0.735 ± 0.738	MS, TI, O
3-Octanol	1396	Earthy	0.016 ± 0.020a	ndb	0.010 ± 0.015ab	0.003 ± 0.006ab	MS, TI
(E)-2-hexenol	1451	Green	ndb	ndb	0.022 ± 0.038ab	0.062 ± 0.064a	MS, TI
Cyclohexanol	1461	Camphorous	nd	nd	0.008 ± 0.026	0.044 ± 0.154	MS, TI
Octanol	1595	Moss, green	0.003 ± 0.009	0.002 ± 0.006	0.001 ± 0.003	0.006 ± 0.008	FI, MS, TI
2,3-Butanediol	1634	Creamy	nd	nd	0.002 ± 0.006	0.026 ± 0.081	MS, TI
Nonanol	1661	Green	nd	nd	0.002 ± 0.008	0.003 ± 0.007	MS, TI
Furfuryl alcohol	1719	Burnt	nd	0.017 ± 0.050	0.025 ± 0.051	nd	MS, TI
Phenethyl alcohol	1948	Floral, sweet	nd	nd	nd	0.002 ± 0.007	MS, TI
Aldehydes
Hexanal	1122	Green, apple	1.025 ± 0.298ab	1.487 ± 0.437a	0.832 ± 0.623b	0.518 ± 0.610b	FI, MS, TI, O
(E)-2-hexenal	1275	Fresh green	0.452 ± 0.153	0.348 ± 0.122	0.222 ± 0.189	0.256 ± 0.342	MS, TI, O
Nonanal	1445	Green	nd	nd	0.009 ± 0.017	0.004 ± 0.013	FI, MS, TI
(E)-2-octenal	1466	Green	0.012 ± 0.015	0.006 ± 0.012	0.004 ± 0.010	0.003 ± 0.010	MS, TI
Decanal	1554	Orange, floral	0.002 ± 0.006	nd	0.006 ± 0.014	0.002 ± 0.008	FI, MS, TI
(E)-2-nonenal	1604	Cucumber	nd	nd	0.002 ± 0.005	nd	FI, MS, TI
(E)-2-dodecenal	1779	Orange, citrus	nd	nd	nd	0.002 ± 0.006	MS, TI
5-Acetoxymethyl-2-furaldehyde	2284	Sweet	nd	0.002 ± 0.007	nd	nd	MS, TI
Ketones
3-Octanone	1272	Herbal	nd	nd	nd	0.001 ± 0.002	MS, TI
Acetoin	1353	Butter, creamy	nd	nd	0.020 ± 0.047	0.040 ± 0.075	MS, TI
Geranylacetone	1912	Floral	nd	0.005 ± 0.012	0.001 ± 0.004	nd	MS, TI
Furans
2-Pentylfuran	1266	Green bean	0.001 ± 0.002	0.020 ± 0.031	0.003 ± 0.009	0.008 ± 0.021	FI, MS, TI
Furfural	1544	Almond	nd	0.064 ± 0.159	0.064 ± 0.115	nd	MS, TI
5-Hydroxymethylfurfural	2061	Caramellic	nd	0.059 ± 0.115	0.501 ± 1.133	nd	MS, TI
Acids
Acetic acid	1504	Sour	nd	0.029 ± 0.087	0.044 ± 0.076	nd	MS, TI
Butyric acid	1641	Acidic	0.004 ± 0.012	nd	nd	nd	MS, TI
Hexanoic acid	1849	Pungent	nd	nd	0.003 ± 0.012	0.004 ± 0.010	MS, TI
Terpenes
d-limonene	1210	Citrus, sweet	0.050 ± 0.102	0.068 ± 0.140	0.017 ± 0.046	0.027 ± 0.058	MS, TI, O
Others
Estragole	1745	Licorice	0.171 ± 0.109a	0.029 ± 0.022b	0.014 ± 0.023b	0.016 ± 0.018b	FI, MS, TI
α-farnesene	1787	Woody	0.010 ± 0.021b	0.101 ± 0.104ab	0.205 ± 0.148a	0.112 ± 0.151ab	MS, TI

Different letters (a-b) in the same row indicate statistically significant differences (Tukey’s HSD, n = 12, p-value < 0.05); ^ARetention indices were determined on HP-FFAP capillary column using n-alkanes (C7 − C30) as external references; ^BAroma descriptors from the Cornell University’s Flavornet (https://www.flavornet.org/flavornet.html) and Good Scents Company (https://www.thegoodscentscompany.com/index.html); ^CIdentification with FI (fully identified using authentic standard), MS (mass spectrum consistent with that from the NIST library), TI (tentatively identified based on the NIST library and literature), and O (Odor quality perceived at the sniffing port); ^DNot detected.

The results of the non-volatile (sugars and organic acids) analyses in 2020 and 2021 (2 years; n = 12) using HPLC − RI and UV are listed in Table 4. The total sugar content of the consumer-preferred cultivar “Gamhong” was significantly higher than those of the other cultivars such as “Yangwang” and “Hongro,” coinciding with the °Brix values (Table 1). The low amount of total sugars in “Hongro” (the least preferred cultivar) was likely attributable to the low level of sucrose in it. The sucrose content of “Hongro” was <50% of that in “Gamhong” (Table 4). For organic acids, the tartaric acid and malic acid contents were higher in “Gamhong” and “Fuji” than in “Hongro,” while the citric acid content was inversely higher in “Hongro.”.

Table 4.

Non-volatile compositions of apple cultivars.

	Mean ± standard deviation
	‘Gamhong’	‘Yangwang’	‘Hongro’	‘Fuji’ (reference)
Sugars (g/kg dried sample)
Fructose	403.72 ± 56.66b	414.86 ± 32.66b	478.29 ± 32.62a	464.13 ± 30.28a
Sorbitol	24.76 ± 9.37b	27.94 ± 6.03b	24.27 ± 5.76b	37.27 ± 8.10a
Glucose	172.66 ± 39.24ab	103.32 ± 18.00c	193.38 ± 26.24a	160.74 ± 21.71b
Sucrose	258.28 ± 47.75a	290.49 ± 31.40a	126.67 ± 35.29c	179.05 ± 26.57b
Total sugars	859.42 ± 41.37a	836.62 ± 16.03ab	822.61 ± 26.53b	841.19 ± 36.71ab
Organic acids (mg/kg dried sample)
Oxalic acid	303.76 ± 245.60b	337.80 ± 195.02ab	486.10 ± 69.77a	478.54 ± 66.54ab
Tartaric acid	383.22 ± 157.29a	302.07 ± 41.01a	39.87 ± 45.67b	312.30 ± 95.27a
Malic acid	23637.92 ± 2396.64a	24880.90 ± 2727.94a	19129.22 ± 2387.89b	22699.17 ± 1447.33ab
Lactic acidns	nd	43.31 ± 106.09	nd	nd
Citric acid	313.19 ± 271.94b	696.25 ± 274.24a	752.61 ± 97.71a	212.20 ± 120.19b
Total organic acids	24638.08 ± 1988.30a	26260.32 ± 3020.82a	20407.80 ± 2498.42b	23702.21 ± 1569.70ab

Different letters (a-b) in the same row indicate statistically significant differences (Tukey’s HSD, n = 12, p-value < 0.05).

3.4. Relationships between sensory attributes and chemical (metabolomic) profiles of apple cultivars

A PLS regression analysis was performed to confirm the relationships between the sensory properties and chemical compositions of the apple cultivars. The regression model proved the contributions of volatile (x variables, n = 61) and non-volatile compounds (x variables, n = 11) to the taste and flavor attributes (y variables, n = 10) of the different apple cultivars (“Gamhong,” “Yangwang,” “Hongro,” and the reference material “Fuji”). The consumer-preferred cultivars (“Gamhong” and “Fuji”) were clearly distinguished from “Hongro” (the least preferred cultivar) on the PC1 axis (Fig. S3), with markedly different sensory attributes and chemical profiles (Fig. 2). The PLS-regression models were validated by a 200-permutation test with comparing Q² values obtained from the original dataset, and the distribution of Q² values calculated when original y values are randomly assigned to the individuals (criterion: Q² and R² values on the permutated data sets should be lower than those on the actual data set) (Triba et al., 2015). The permutation test showed overall good predictability with no overfitting of the model on the relationships between volatile compounds and sensory attributes (e.g., apple, pear, cucumber, floral, green, honey) (Fig. S3), indicating these attributes are significantly correlated with the volatile profile of apples. Likewise, the PLS model for the relationships between non-volatile compounds and sensory attributes (taste parameters) was also validated in the same way (permutation test) (Fig. S4).

Fig. 2.

PLS regression plot showing relationships between sensory attributes and chemical profiles of apple cultivars. (A) The loading plot for four apple cultivars with volatile compounds (R²X = 0.274, R²Y = 0.523, Q² = 0.223) and (B) loading plot for four apple cultivars with non-volatile compounds (R²X = 0.557, R²Y = 0.620, Q² = 0.446).

“Yangwang” was closer to “Hongro,” but this cultivar was not completely separated from either the consumer-preferred group (“Gamhong” and “Fuji”) or the least preferred group (“Hongro”), revealing its ambiguous characteristics in terms of sensory and chemical properties. “Gamhong” (the consumer-preferred cultivar) was mainly characterized by pleasant aroma (apple, pineapple, fruity, honey, floral, pear) and taste (sweetness, sourness) attributes, which were grouped with the “ester” class volatiles, especially acetate and butyrate derivatives, including hexyl acetate (apple, fruity), hexyl hexanoate (apple peel), hexyl butyrate (apple, fruity, pineapple), butyl 2-methylbutyrate (fruity, pineapple), butyl butyrate (fruity, pineapple), butyl acetate (pear), 2-methylbutyl acetate (fruity), 2-methylbutyl isovalerate (apple, fruity), isobutyl butyrate (sweet, pineapple, apple), isoamyl isobutyrate (fruity, sweet), amyl butyrate (pineapple, sweet), and 2-methylbutyl butyrate (fruity, sweet), as well as non-volatiles such as the total sugars (sweetness), sorbitol, and tartaric acid (sourness). On the other hand, “Hongro” (the least preferred cultivar) was characterized by unpleasant aroma (mostly a cucumber flavor) attributes that were grouped with volatiles such as (E)-2-nonenal (cucumber-like), 2-methyl-1-butanol (wine, fusel oil-like), isobutanol (wine-like), furfuryl alcohol (burnt), and furfural (almond).

4. Discussion

Fruit quality is a dynamic concept that changes on the basis of consumer needs and perceptions with sociocultural evolution. According to a food and health survey in 2019, flavor was the top-ranking driver of food/beverage purchasing decisions, inferring a recent trend of consumer insights and the importance of sensory properties (FoodInsight, 2019). A balance between sweet and tart flavors is generally considered to be the primary factor in consumer acceptance of apples, but other factors (e.g., texture, color) may also be involved. In this study, four different apple cultivars (“Gamhong,” “Yangwang,” “Hongro,” and “Fuji”) were selected based on their distinct quality profiles, including their variability in taste, flavor, and texture. “Gamhong” has good eating quality with a strong sweet taste and flavor. “Yangwang” has flesh with a rough texture. “Hongro” is an early-maturing variety with slight fruity flavor. “Fuji” is extensively harvested around the world and has excellent quality and sensory characteristics (Echeverría et al., 2004, Koh et al., 2009, Park and Yoon, 2012, Yoo et al., 2021).

In relation to the physicochemical characteristics of the apple cultivars (“Gamhong,” “Yangwang,” “Hongro,” and the reference material “Fuji”), “Gamhong” had the highest °Brix value (15.26), while “Hongro” had quite a low °Brix value (12.44), with a relatively intense red flesh color compared to the other cultivars. The flesh color of “Hongro” might be related to the content of anthocyanins, which are pigments responsible for the color of apple fruits (Iglesias et al., 2012). It was reported that consumers prefer apples with °Brix values in the range of 13–16 (Kajikawa, 1998). All the cultivars (“Gamhong,” “Yangwang,” and “Fuji”) except “Hongro” had °Brix values within the consumer-preferred range (13.37–15.26). These results (flesh color, °Brix value) linked to the consumer acceptance of apples were further confirmed through the sensory evaluations.

In the general consumer studies conducted over 2 years, “Gamhong” achieved the overall highest scores for the consumer-preferred sensory parameters (e.g., flavor liking, juiciness, sweetness, apple flavor), while “Hongro” obtained low scores for parameters such as flavor liking, color liking, juiciness, sweetness, sourness, and apple flavor (Table S2 and Table 2). The favorable sensory rating of “Gamhong” was likely related to its high °Brix (TSS) value and balanced °Brix/acid (TSS/TA) ratio, which contribute to the sweet-tart flavor of apples (Table 1). On the other hand, the low sensory rating of “Hongro” may be related to its low °Brix value, which was outside of the consumer-preferred range (Table 1) (Kajikawa, 1998). In addition, the low color liking (flesh) of “Hongro” was likely associated with its reddish flesh color, which is not commonly expected for the flesh color of apples (Table 1, Table S2, Table 2). Our results suggested that consumers do not prefer apples with red flesh (internal color), although they may like the red peel (external color) of apples. Meanwhile, it turned out that the sensory quality of the apples was not always proportional to physicochemical parameters such as the °Brix value or °Brix/acid ratio. For example, although “Fuji” had a much lower °Brix/acid ratio than “Gamhong” (even lower than “Hongro,” the least preferred cultivar), “Fuji” was a top consumer-preferred cultivar like “Gamhong,” with high scores for the hedonic values (flavor liking, overall liking, etc.) (Table 1, Table 2). This indicates that there must be other factors (e.g., chemical composition) contributing to the sensory quality of apples, beyond the parameters discussed above.

The description analysis using trained panels (2 years) provided detailed information on the flavors of apples liked/disliked by consumers. The consumers preferred intense sweet, sour, apple, pineapple, and honey flavors, but did not like flavors such as a cucumber-like aroma (Fig. 1). The PCA plot clearly distinguished the consumer-preferred cultivars (“Gamhong” and “Fuji”) from “Hongro” (the least preferred cultivar) with these attributes (Fig. 1B). The PCA results suggested that a combination of taste (sweetness, sourness) and flavor (apple, pineapple, fruity, honey) attributes may be a key for the positive hedonic perception of apples, while attributes like a cucumber flavor might break this balance, leading to a negative hedonic perception of apples.

Next, a metabolomic analysis was performed to identify and characterize the volatile and non-volatile profiles of apples (Table 3, Table 4). The chemical composition of a product is important to understand its flavor quality, because flavor perception involves a complex neurophysiological process triggered by various factors (odor, taste, chemesthesis, etc.) via chemical stimuli from food components (e.g., metabolites: volatiles and non-volatiles) that could determine the hedonic value of food (Laing & Jinks, 1996). As previously mentioned, an apple’s flavor quality is primarily driven by a diverse set of chemicals, including volatiles, sugars, acids, and their mixtures (Espino-Díaz et al., 2016; F. R. Harker et al., 2002). Our results revealed that the consumer-preferred cultivars (“Gamhong” and “Fuji”) were mainly characterized by high levels of esters (volatiles), sugars (total sugars), and organic acids (tartaric acid, malic acid), while the least preferred cultivar (“Hongro”) contained high amounts of other classes of volatiles and citric acid. This indicated the potential role of these metabolites in consumer preferences for apples (e.g., esters for consumer-liked aroma attributes, sugars, and organic acids for balanced sweet and tart flavors). Although existing in relatively high levels in “Hongro,” citric acid did seem to affect the sourness of the apples as much as other acids (tartaric acid, malic acid). A previous study demonstrated that citric acid had less effect on sourness perception than tartaric or malic acid in pair tests (Noble et al., 1986). Another study showed that malic acid has a higher relative sourness, compared to citric acid on a pound-weight basis, supporting this assumption (Buechsenstein & Ough, 1979).

Finally, the relationships between the sensory properties and chemical compositions of the apple cultivars were confirmed by the PLS regression analysis (Fig. 2). The pleasant sensory attributes (apple, pineapple, fruity, floral, honey, sweetness, and sourness) of apples were linked to the metabolites, including ester class volatiles, sugars, and tartaric acid. In detail, for aroma, the pineapple and fruity notes of the apples were highly correlated with butyl 2-methylbutyrate and hexyl butyrate (Fig. 2A, Table 3). The sweet and fruity notes of the apples were likely attributable to isoamyl isobutyrate and 2-methylbutyl butyrate, which have these aroma properties. We also found that two flavor compounds (hexyl acetate and 2-methylbutyl acetate) were mainly responsible for the typical fruity note of apples. Their aroma profiles were confirmed by the GC − MS/O analysis (Table 3). These compounds, including 2-methylbutyrate, hexyl butyrate, isoamyl isobutyrate, 2-methylbutyl butyrate, hexyl acetate, and 2-methylbutyl acetate were found in relatively large amounts (or exclusively detected) in the consumer-preferred cultivars “Gamhong” and “Fuji” (Table 3). Acetate and butyrate-derived esters, including hexyl acetate, were proven to be impact odorants of apples in a previous study (Reis et al., 2009), partially supporting our results. For taste, although there was some variation in the concentrations of individual sugars (e.g., sucrose) in the different cultivars, the total sugars were shown to have a major impact on the sweetness perception of the apples (Fig. 2B, Table 4). As mentioned earlier, the sourness of apples is considered to be derived from tartaric and malic acids (Buechsenstein and Ough, 1979, Noble et al., 1986). These sugars, acids, and their combination are likely associated with the consumer acceptance of apples with balanced sweet and tart flavors. Overall, the features (favorable attributes and metabolites) discussed above were well grouped with the consumer-preferred cultivars (Fig. 2). Meanwhile, a metabolite (E)-2-nonenal was found to be involved in unpleasant sensory attributes (e.g., the cucumber-like aroma) (Fig. 2A, Table 3). (E)-2-nonenal was reported to be an off-flavor of fruits such as watermelon (Guler et al., 2013, Yang et al., 2020). (E)-2-nonenal, which only existed in “Hongro,” may have negatively affected the consumer acceptance of this cultivar by weakening the original flavor profile of the apples. Furfuryl alcohol (burnt flavor) and furfural (almond flavor) were predominantly detected in “Hongro” and are also known to be responsible for off-flavors in apples and oranges (Amann, 2016, Rouseff et al., 1992), possibly leading to the low sensory rating of “Hongro.” In addition, the active biosynthesis of furfural and furfural alcohol might have depleted the sugars in the fruits, because sugars are metabolic precursors of these furfural derivatives (Noomsiri & Lorjaroenphon, 2018). This may partially explain why “Hongro” was characterized by a low amount of sugars, with high levels of furfural and furfural alcohol. We also found volatiles with a wine-like flavor, such as 2-methyl-1-butanol and isobutanol, which were potentially associated with the negative sensory value of the apples. This wine-like flavor is typically derived from fermentation, which is not expected from fresh apple fruits (Styger et al., 2011). This needs to be further confirmed with a sensory evaluation using wine flavor as an attribute. Nonetheless, the above results effectively demonstrated the correlations between the sensory attributes and chemical profiles of the apple cultivars (e.g., hexyl acetate and 2-methylbutyl acetate for the fruity flavor of “Gamhong”; (E)-2-nonenal for the cucumber flavor of “Hongro”).

5. Conclusions

In the present work, the interactions between the sensory attributes and chemical compositions of apple cultivars were elucidated using a combined metabolomic analysis and sensory evaluation. Significant volatile and non-volatile compounds that are responsible for the flavor quality of apples were identified based on the statistical correlations of the sensory and metabolomic data. Volatile esters and non-volatile sugars and acids were found to be associated with consumer preferences (flavor: apple, fruity, pineapple, sweetness, sourness, etc.) for apples, whereas some aldehydes and alcohols were majorly linked to a negative hedonic perception (e.g., cucumber-like flavor) of apples. Our approach based on a joint metabolomic and sensory analysis with the help of statistics provided novel insights and a framework in the area of fruit studies to better understand the complex hedonic perceptions and sensory qualities of fruits via chemical information. In future apple research, we expect that the sensory qualities and related chemistry of apple peels (which are also important in markets) could be investigated using this platform. We also expect that the information collected on the key flavor compounds of apples could be used to control the quality of the apple fruits in the apple industry, and by laboratories and regulatory authorities.

CRediT authorship contribution statement

Keono Kim: Conceptualization, Investigation, Methodology, Formal analysis, Visualization, Writing – review & editing. Ik-Jo Chun: Conceptualization, Investigation, Resources. Joon Hyuk Suh: Conceptualization, Supervision, Visualization, Writing – original draft, Writing – review & editing. Jeehye Sung: Conceptualization, Investigation, Supervision, Funding acquisition, Project administration, Writing – review & editing.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following are the Supplementary data to this article:

Data availability

Data will be made available on request.