Volatile Organic Compounds Profiles to Determine Authenticity of Sweet Orange Juice Using Head Space Gas Chromatography Coupled with Multivariate Analysis.

An efficient and practical method for identifying mandarin juice over-blended into not from concentrate (NFC) orange juice was established. Juices were extracted from different cultivars of sweet orange and mandarin fruits. After being pasteurized, the volatile organic compounds (VOCs) in the juice samples were extracted using headspace solid-phase microextraction, and qualitatively and quantitatively analyzed using gas chromatography-mass spectrometry detection. Thirty-two VOCs contained in both the sweet orange juice and mandarin juice were used as variables, and the identification model for discriminating between the two varieties of juice was established by principal component analysis. Validation was applied by using common mandarin juices from Ponkan, Satsuma and Nanfengmiju cultivars blended at series of proportions into orange juices from Long-leaf, Olinda, and Hamlin cultivars. The model can visually identify a blending of mandarin juice at the volume fraction of 10% or above.


Introduction
Orange juice is one of the most popular fruit juices because of its high vitamin C content, rich flavor, and balanced sweetness and sourness. According to the United Nations Commodity Trade Database, in 2017, the global export volume and value of citrus juices reached 5.853 million tons and 6769 million US dollars, respectively, in which orange juice accounted for more than 85% of the total amount. Among all common juice products, not from concentrate (NFC) orange juice has been considered as being of the best quality, accounting for an important share of orange juice consumption in developed countries, and its share in emerging countries' markets has also increased year by year [1,2]. Orange juice is generally allowed to be mixed with a small amount of other citrus juice or juice cells, according to Codex 247-2005 of the Codex Alimentarius Commission, to adjust its taste and flavor before sterilized packaging. However, excessive addition will reduce the quality of orange juice and affect the sensory experience of consumers (data from FAO, 2019) [3].
Preparation of NFC juice: Fresh and mature citrus fruits were manually peeled, and juices were extracted using a Hurom model H-100-DWBIA0 rotary juice extractor, respectively. The juice was filtered through 100-mesh gauze and collected in a food-grade stainless steel pot. The pot was heated by immersion in a thermostatic water bath at 65 • C and stirred at 240 rmp. The juice was heated for 30 min to mimic thermal processing conditions used in the juice industry. After heating, the juice was bottled and immediately put into an ice bath to quickly cool to 4-5 • C. Total soluble solids of the NFC juices were determined by using a Pal-1 handheld Brix meter (ATAGO, Tokyo, Japan).
Preparation of mixed juices for authenticity test: Sweet oranges and mandarins commonly used for juice were selected and extracted, and their juices were mixed in different volume proportions (v:v = 5:95, 10:90, 15:85, 20:80, 25:75, and 30:70). Three groups of mixed juice, Olinda Valencia orange and Xinshengxi No.3 Ponkan, Hamlin orange and 2003-4 Satsuma, and Long-leaf orange and Nanfeng mandarin, were prepared to test the differentiation ability of the discriminant model.

Analysis of VOCs by HS-SPME-GC-MS
Five milliliters of shaken fruit juice was put into a 20 mL glass vial. Three microlitres of the internal standard were accurately added into the glass vial, and the headspace was filled with nitrogen. The sample was equilibrated for 20 min at 40 • C in water. Solid-phase microextraction was performed by exposing a divinylbenzene/carboxen/polydimethylsiloxane (DVB/CAR/PDMS; Supelco, Bellefonte, PA, USA) StableFlex fiber (1 cm, 50/30 µm) in the vial headspace of the sample at a constant depth at 40 • C for 30 min. The samples were gently vortexed during equilibration and extraction at 240 rpm, using a magnetic stirrer. After extraction, the fiber was removed from the vial and immediately inserted into the injection port of the GC for desorption at 250 • C for 5 min. All analyses of the volatile compounds were performed in triplicate.
The gas chromatography analyses were carried out using Agilent-7890B gas chromatography (GC) equipped with a mass selective detector (MSD). Volatile compounds were separated on an Agilent DB-5 MS column (30 m × 0.25 mm, 0.25 µm film thickness), with helium as the carrier gas at a constant flow of 1.2 mL/min. The oven temperature conditions were 35 • C for 0 min, increasing at 7 • C min −1 to 98 • C, 3 • C min −1 to 161 • C, 10 • C min −1 to 241 • C, and then held for 16 min. Electronic ionization was used (70 eV); the ion source and transfer line temperatures were 230 • C and 280 • C, respectively. Detection was performed in scan mode in the range between 40 and 350 m/z.

Determination of Detection Limit and Quantitative Limit
Model mandarin juices were prepared using the formulation used by PELEG's method [29] with slight modification. The basic model juice contained fructose (3.0 g), glucose (3.0 g), sucrose (6.0 g), and citric acid (1.0 g), dissolved using Milli-Q water and diluted to a final volume of 100 mL. Vitamin C was added to the simulated juice at a concentration of 0.3 g/L. Using model mandarin juices as a blank sample and cyclohexanone as an internal standard, the determination of extremely small amounts of VOCs was carried out following the requirements of the International Council for Harmonization (ICH) for the Registration of Technical Requirements for Human Drugs and the Chinese Pharmacopoeia. Different concentrations of internal standard samples were tested, and the noise ratio calculation formula was used to calculate the detection limit (pk-pk S/N = 3) and the quantitative limit (pk-pk S/N = 10) of this experimental method. The detection limit was 9.28 × 10 −5 µg/mL, and the limit of quantification was 2.78 × 10 −4 µg/mL, which ensured that the VOCs needed for identification could be detected and quantified.

Identification of Volatile Organic Compounds
The mass spectrometry and retention index of the compound were matched and compared with the Agilent W10N14.l mass spectrometry library and retention index database, respectively. The retention index database was provided by the University of Florida (http://www.flavornet.org/flavornet.html) [30]. The retention index of the mixture was calculated with N-alkane standard (C5-C20) under the same conditions as those of the sample, and the difference between the calculated retention index and the database retention index was less than 5%, which ensured the accuracy of the results.
Quantitation was performed according to Shui et al.'s quantitative method [31]. The internal standard method was used, and the internal standard was cyclohexanone (14.25 µg/L). The semi-quantitative analysis was performed by comparing the peak area of each component with that of the internal standard, and the content unit was µg/mL. Samples of each variety were prepared three times and injected, respectively. All samples used the same preparation method. The results are expressed as the mean of three samples.

Statistical Analysis
Chromatograms and spectra were recorded and processed using Masshunter Qualitative Workflows B.08.00 software. The data analysis software Unscrambler X 10.4 was implemented. Data processing was performed using Microsoft Excel 365. Image processing was performed with Origin 2018.
The principal idea of principal component analysis is to reduce a quantity set that includes many interconnected variables and keep as many useful variables as possible in the quantity set. In this study, we used PCA to recombine VOCs that needed to be analyzed in sweet orange juice and mandarin juice, forming a new comprehensive index that reflected the characteristics of the sample.
PCA analysis steps: The category of VOCs was taken as the row of the matrix, and the sample to be tested was taken as the column of the matrix. Data were inserted, and principal component analysis was performed. A maximum of four components was selected for principal component analysis.  Table S1. Volatile organic compounds in 27 sweet orange and 19 mandarin heated juices were determined, and their typical total ion chromatograms (TIC) are shown in Figure 1 (above is Olinda Valencia orange juice, and below is Xinshengxi No.3 Ponkan juice). Orange and mandarin juice TICs were mirrored for a clear comparison. Among all VOCs, D-limonene (39) had the highest concentration. The VOCs of both varieties of juice were composed of monoterpenes, sesquiterpenes, aldehydes, esters, alcohols, and ketones. The chromatographic elution can be roughly divided into four sections. In Section 1, low carbon alcohols and aldehydes flowed out in 5.0-8.0 min. In Section 2, monoterpenes flowed out in 8.0-12.5 min. In Section 3, alcohols, aldehydes and esters with higher molecular weight flowed out in 12.5-22.0 min. In Section 4, sesquiterpenoids flowed out in 22.0-32.0 min. It can be seen from the Figure 1. that there are differences in the types and concentrations of VOCs in sweet orange juice and mandarin juice. This is also consistent with the existing research results [28].  Table 1 and Table 2. Each type of substance is shown in different colors: red: monoterpenes; yellow: sesquiterpenes; brown: aldehydes and ketones; green: alcohols; and blue: esters. IS: internal standard.

Characteristic VOCs in Sweet Orange Juice and Mandarin Juices
After comparison, the characteristic VOCs, which were exclusively contained in NFC sweet orange juice or NFC mandarin juice, respectively, are shown in Table 1. According to the table, among the 27 varieties of sweet orange juices and 19 varieties of mandarin juices, there were 20 VOCs unique to sweet orange juice, including four sesquiterpenoids, three aldehydes, four alcohols, six esters, and three ketones. There were also nine VOCs unique to mandarin juice, including four sesquiterpenoids, one aldehyde, two alcohols, and two ketones. Neither of the characteristic volatile organic compounds in sweet orange juice or mandarin juice contained monoterpenoids. Esters have been well-known to be a major aroma contributor to citrus juice [32]. However, in the characteristic VOCs, sweet orange juice contained six esters, which were not contained in the mandarin juice. At the same time, the aroma threshold of volatile ester was lower. This may be one reason why sweet orange juice has a better overall aroma than mandarin juice.   Tables 1 and 2. Each type of substance is shown in different colors: red: monoterpenes; yellow: sesquiterpenes; brown: aldehydes and ketones; green: alcohols; and blue: esters. IS: internal standard.

Characteristic VOCs in Sweet Orange Juice and Mandarin Juices
After comparison, the characteristic VOCs, which were exclusively contained in NFC sweet orange juice or NFC mandarin juice, respectively, are shown in Table 1. According to the table, among the 27 varieties of sweet orange juices and 19 varieties of mandarin juices, there were 20 VOCs unique to sweet orange juice, including four sesquiterpenoids, three aldehydes, four alcohols, six esters, and three ketones. There were also nine VOCs unique to mandarin juice, including four sesquiterpenoids, one aldehyde, two alcohols, and two ketones. Neither of the characteristic volatile organic compounds in sweet orange juice or mandarin juice contained monoterpenoids. Esters have been well-known to be a major aroma contributor to citrus juice [32]. However, in the characteristic VOCs, sweet orange juice contained six esters, which were not contained in the mandarin juice. At the same time, the aroma threshold of volatile ester was lower. This may be one reason why sweet orange juice has a better overall aroma than mandarin juice.
The characteristic VOCs of mandarin juice can be used to determine if mandarin juice is added to the orange juice. Orange juice samples that contained mandarin-specific VOCs could be preliminarily determined as not 100% orange juice. The sample juice was then further evaluated for how much mandarin juice was blended.

Common VOCs in Both Sweet Orange Juice and Mandarin Juices
By further analysis, the common VOCs in sweet orange juice and mandarin juice are shown in Table 2. In both the sweet orange juices and mandarin juices, terpenes were the most common, including 15 monoterpenes and 11 sesquiterpenes; the number of alcohols, aldehydes, and esters were all approximately 10, and ketones were the least, with three detected. A total of 61 volatile substances were shared in both groups, proving that sweet orange juice and mandarin juice have most of the same VOCs constituent.
By eliminating D-limonene, which was contained in almost all citrus species with a high concentration and the components with a content of 0 or below the quantitation limit, 32 common VOCs of sweet orange juice and mandarin juice were selected and shown in Tables S2-S5. Concentrations of the selected common monoterpenes in each juice samples are shown in Table S2. The contents of β-pinene (34), α-terpinene (38), and γ-terpinene (41) in mandarin juice were significantly higher than those in sweet orange juice (p < 0.005). β-trans-ocimene (40) and L-carvone (89) in sweet orange juice were significantly higher than that in mandarin juice (p < 0.005). However, the overall concentrations of common monoterpenes in mandarin juice were significantly higher than those in sweet orange juice, and the content of β-myrcene was higher than that of other monoterpenes in both sweet orange juice and mandarin juice. According to Table S3, the content of β-selinene (51), valencene (52), and α-selinene (53) in sweet orange juice was significantly higher than that in mandarin juice (p < 0.005). As most sweet orange juice does not contain γ-muurolene (49), this may be related to the variety and cultivation environment. Also, the overall concentrations of common sesquiterpenes in sweet orange juice were significantly higher than those in mandarin juice. However, linalool (70) was one of the most abundant alcohols in both citrus juices. This is also consistent with the results of previous studies [28]. In addition, the total amounts of common alcohols, aldehydes, and esters in sweet orange and mandarin juices showed no significant difference in Tables S4-S5. It can be seen that the main differences in the concentration of common VOCs in sweet orange juice and mandarin juice exist in terpenoids. These characteristics in Tables S2-S3 also coincide with the distribution characteristics of these common VOCs in Figure 1, and can help to establish the identification model.

Principal Component Analysis of VOCs in Samples
To further elucidate the difference of common VOCs between sweet orange and mandarin juice, all collected citrus juices were analyzed using principal component analysis. The 32 common volatile components were used as variables, and the results are shown in Figure 2. As seen from the two-dimensional biplot (Figure 2A), the cluster of sweet orange juice samples was well separated from that of mandarin juice samples. The red dots represent variables whose position distribution indicates their contribution to the sample scores, in which γ-terpinene (41), β-myrcene (35), and valencene (52) had more significant contributions to samples. Figure 2B is a correlation loading plot, which provides a scale independent assessment of the VOCs and a clearer indication of variable correlations. In Figure 2B, the contribution of γ-terpinene (41), α-thujene (30), β-pinene (34), α-terpinolene (42), decanal (62), and α-pinene (31) to PC-1 reached between 0.8 and 1.0, while that of valencene (52), α-selinene (53), β-selinene (51), and β-myrcene (35) to PC-2 reached 0.8−0.9. Among these, most of the major contributions in the X+ direction were monoterpenoids, and β-myrcene (35) contributed to both the X+ and Y-directions, which means that it made a great contribution to clusters of sweet orange and mandarin. In addition, the substances that contributed significantly to the X-and Y-directions were sesquiterpenes (valencene (52), α-selinene (53), and β-selinene (51)). Combining with Figure 2A,B, valencene (52) and γ-terpinene (41) made the largest contribution to clusters of sweet orange juice and mandarin juice, respectively. These VOCs play a key role in differentiating between sweet oranges and mandarin oranges. At the same time, this also shows that the spatial relationship between the scores and loadings biplot and the correlation loading plot correspond with each other.
The above results show that the distribution of VOCs in sweet orange juice and mandarin juice have characteristic distribution in the model space, and that the two types of juice can be distinguished by PCA of their VOC concentrations. Figure 3A shows that the Hotelling's T2 test had only a few outliers, far fewer than the number of samples, and none of them appeared in the large error area (Hot T2 > 11.32, F Res > 13.75). Moreover, in Figure 3B, the contributions of the four groups of principal components to the calibration and validation variances were 99.3% and 98.9%, respectively. The calibration variance curve was largely consistent with the validation variance curve. The above results show that this model can reflect the characteristics of the original sample sufficiently. The above results show that the distribution of VOCs in sweet orange juice and mandarin juice have characteristic distribution in the model space, and that the two types of juice can be distinguished by PCA of their VOC concentrations. Figure 3A shows that the Hotelling's T2 test had only a few outliers, far fewer than the number of samples, and none of them appeared in the large error area (Hot T2 > 11.32, F Res > 13.75). Moreover, in Figure 3B, the contributions of the four groups of principal components to the calibration and validation variances were 99.3% and 98.9%, respectively. The calibration variance curve was largely consistent with the validation variance curve. The above results show that this model can reflect the characteristics of the original sample sufficiently.

Identification of Sweet Orange Juice Mixed With Mandarin Juice
The obtained discriminant model was used to identify and verify sweet orange juice mixed with mandarin juice. The presented VOCs and their contents in the mixed juices were identified according to the experimental method, and the data of each substance were substituted into the PCA discrimination space for identification. As shown in Figure 4, the points of the mixed juices with low adding proportions were close to the boundary of the sweet orange juice group. With the increase in the proportion of mandarin juice, the sample points of blended orange juice became closer to the mandarin juice group. When the mixing ratio was 10%, samples No. 1 and No. 2 were significantly

Identification of Sweet Orange Juice Mixed With Mandarin Juice
The obtained discriminant model was used to identify and verify sweet orange juice mixed with mandarin juice. The presented VOCs and their contents in the mixed juices were identified according to the experimental method, and the data of each substance were substituted into the PCA discrimination space for identification. As shown in Figure 4, the points of the mixed juices with low adding proportions were close to the boundary of the sweet orange juice group. With the increase in the proportion of mandarin juice, the sample points of blended orange juice became closer to the mandarin juice group. When the mixing ratio was 10%, samples No. 1 and No. 2 were significantly away from the orange juice group. When the mixing ratio was 5%, sample No. 3 was still beyond the orange juice group. Therefore, blended orange juice can be visually identified with a blending proportion of ≥10%.
away from the orange juice group. When the mixing ratio was 5%, sample No. 3 was still beyond the orange juice group. Therefore, blended orange juice can be visually identified with a blending proportion of ≥10%. In addition, it can be seen that the distance between the respective points of samples No. 1 and No. 2 is significantly shorter than the distance between the points in sample No. 3. In other words, the farther the distance between sweet orange (the yellow points) and mandarin (the blue points) cultivars in the PCA discriminant model, the larger the distance between sample points of the mixed orange juice with different mixing proportions of the combination, and the more obvious the effect. This method suits for determination if sweet orange juice was added with botanically pure mandarin juice, but was not designed to deal with adulteration using orange or mandarin hybrids, such as the Shiranui tangor, Murcott tangerine, Kiyomi tangor, etc. The distribution characteristics of volatile organic compounds of these hybrids may be between the sweet orange and mandarin juices, and the sample scores of hybrid oranges in the above discriminant space may fall between the clusters of sweet orange and mandarin. However, because these hybrids normally have a nice fresh taste, their fresh fruits are sold at a high price, and they are not likely to be used for juicing and blending.
In this study, we collected almost all the major varieties currently used for juicing in sufficient quantities, and established their VOCs database differentiating model. Types and contents of VOCs in the sample citrus only needed to be determined, then the target VOCs input into the identification model, to identify the sample citrus juice. This model used the difference of VOCs between sweet orange juice and mandarin juice to distinguish the two species and quickly identify whether orange juice was mixed with mandarin juice. In China and worldwide markets, mandarin juice is often added to NFC sweet orange juice to reduce costs, and sometimes the product is fraudulently labeled as 100% sweet orange juice. There are no big differences in color and taste between the mixed juice and the pure orange juice. At present, there is no official detection method specifically used to distinguish between sweet orange juice and mandarin juice. The establishment of this method could be an option to solve the adulteration problems. In addition, it can be seen that the distance between the respective points of samples No. 1 and No. 2 is significantly shorter than the distance between the points in sample No. 3. In other words, the farther the distance between sweet orange (the yellow points) and mandarin (the blue points) cultivars in the PCA discriminant model, the larger the distance between sample points of the mixed orange juice with different mixing proportions of the combination, and the more obvious the effect. This method suits for determination if sweet orange juice was added with botanically pure mandarin juice, but was not designed to deal with adulteration using orange or mandarin hybrids, such as the Shiranui tangor, Murcott tangerine, Kiyomi tangor, etc. The distribution characteristics of volatile organic compounds of these hybrids may be between the sweet orange and mandarin juices, and the sample scores of hybrid oranges in the above discriminant space may fall between the clusters of sweet orange and mandarin. However, because these hybrids normally have a nice fresh taste, their fresh fruits are sold at a high price, and they are not likely to be used for juicing and blending.
In this study, we collected almost all the major varieties currently used for juicing in sufficient quantities, and established their VOCs database differentiating model. Types and contents of VOCs in the sample citrus only needed to be determined, then the target VOCs input into the identification model, to identify the sample citrus juice. This model used the difference of VOCs between sweet orange juice and mandarin juice to distinguish the two species and quickly identify whether orange juice was mixed with mandarin juice. In China and worldwide markets, mandarin juice is often added to NFC sweet orange juice to reduce costs, and sometimes the product is fraudulently labeled as 100% sweet orange juice. There are no big differences in color and taste between the mixed juice and the pure orange juice. At present, there is no official detection method specifically used to distinguish between sweet orange juice and mandarin juice. The establishment of this method could be an option to solve the adulteration problems.

Conclusions
The volatile organic compounds in citrus juice were determined by HS-SPME-GC-MS. It was found that 20 and nine characteristic VOCs were identified in sweet orange juice and citrus juice, respectively. Then 32 common VOCs were identified in both species of juices, and their concentrations were different between the two species. By determining whether an orange juice sample contains the characteristic VOCs of mandarin juice, we can preliminarily determine whether the sample was blended with mandarin juice. The identification model of blending sweet orange juice with mandarin juice was established by 32 selected common VOCs, of which concentrations were used in the principal component analysis. The model could visually identify blended orange juice with a 10% or higher volume fraction of mandarin juice.
Supplementary Materials: The following are available online at http://www.mdpi.com/2304-8158/9/4/505/s1, Table S1: Citrus sample information, Table S2: Relative percentage concentration of common monoterpenes volatile substance in sweet orange and mandarin compared to the internal standard express as %, Table S3: Relative percentage concentration of common sesquiterpenes volatile substance in sweet orange and mandarin compared to the internal standard express as %, Table S4: Relative percentage concentration of common alcohols volatile substance in sweet orange and mandarin compared to the internal standard express as %, Table S5: Relative percentage concentration of common aldehydes and ketone volatile substance in sweet orange and mandarin compared to the internal standard express as %.