Comparative Analysis of Volatiles of 15 Brands of Extra-Virgin Olive Oils Using Solid-Phase Micro-Extraction and Solvent-Assisted Flavor Evaporation

Aroma profiles, key aroma compound quantification, and cluster analysis of 15 brands of extra-virgin olive oils (EVOOs) from three countries (Spain, Italy, and Greece) were investigated in the current study. Aroma compounds were isolated from the oil by using solvent-assisted flavor evaporation (SAFE) and solid-phase micro-extraction (SPME) and analyzed by gas chromatography-olfactometry mass spectrometry (GC-MS/O). A total of 89 compounds were screened by SPME/SAFE-GC-MS/O with chromatographic columns in 15 brands of samples. Eighty and 54 compounds were respectively identified by SPME- and SAFE-GC-MS/O. Of those, 44 compounds were detected by both methods. Undecanol, (Z)-4-decenal, (E)-2-dodecenal, and 2-nonanone extracted by SAFE were not found in EVOOs before. Eight classes of aroma compounds were identified, including 17 alcohols, 22 aldehydes, 9 ketones, 4 acids, 14 esters, 5 aromatics, 12 alkene, and 6 others. Eleven compounds were identified as the key aroma compounds in alternative brands of EVOOs by SAFE-aroma extract dilution analysis (AEDA). Hexanal, (E)-2-hexenal, (E)-3-hexenol, acetic acid, and (E)-2-heptenal were the common key aroma compounds by AEDA and odor activity values (OAVs). From the cluster analysis of the heatmap, the aroma compounds of all the Spain EVOOs were similar, and there were some differences from the samples of Italy and Greece. It suggested that both the amount and concentration of aroma compounds determine the similarity of aroma in EVOOs.


Introduction
Extra-virgin olive oil (EVOO) is a kind of high-grade edible vegetable oil obtained by cold mechanical extraction from fresh olives (Olea europaea) without use of solvents or refining methods [1]. It can be consumed in original form without refining, possessing good stability as well as nutritional and healthy features with respect to other vegetable oils [2]. Thanks to its excellent nutritional and organoleptic properties, EVOO has become one of the most popular oils in the world. The demand for it has increased rapidly in the past decade. In 2014, 3.05 million tons of olive oil was produced worldwide, and the production of olives increased to 19.27 million tons in 2016 [3].
Aroma is an important criterion for EVOO. Its characteristic flavor is one of the main f distinguishing it from the other edible vegetable oils [4]. Meanwhile, the identification of the compounds contributing to the aroma is considered to be a key for quality and authentication control. The aroma compounds of EVOO derive from the enzymatic reactions and autoxidation of unsaturated fatty acids; they occur by the lipoxygenase (LOX) pathway comprising mainly the actuation of LOX and hydroperoxide-lyase enzymes [5] after crushing and during malaxation, a basic step of EVOO extraction process, aimed to improve oil yield and quality [6,7]. From the qualitative and quantitative results, C6 aldehydes and alcohols and the corresponding esters are considered to be the most crucial aroma compounds of EVOO [8,9]. Reasonable amounts of various classes of C5 compounds are also contained in the aroma of EVOO [10].
The selective adsorption coating of SPME results in the incomplete extraction of aroma compounds. The high temperature of SDE affects the accuracy during the extraction process. Solvent-assisted flavor evaporation (SAFE) has become a preferred method because of the low extraction temperature and high vacuum characteristics (about 10 −5 Pa), which can extract more low-boiling compounds. SAFE allows the fast and careful isolation of volatiles from either solvent extracts of foods, oil samples, or even fruit pulps. Application of SAFE to model solutions of selected aroma compounds resulted in higher yield from both solvent extracts and fatty matrices (50% fat) compared to previously used techniques. Besides its efficiency in aroma isolation, the use of the equipment saves time and reduces costs due to the stability of the compact distillation unit [17]. The volatile profiles of pomelo flower, leaf, peel, and juice were comparatively analyzed using HS-SPME and SAFE [18]. To our knowledge, SAFE has not been applied to the aroma extraction of EVOO until now. In addition, the difference among EVOO brands from different countries was not evaluated previously. From this viewpoint, in this study, (1) SAFE coupled with aroma extract dilution analysis (AEDA) was applied to identify the key aroma compounds of 15 brands of EVOO from Spain, Italy, and Greece; (2) the most contributing compounds were quantified by standard curve method using selective ion monitoring (SIM) mode by SPME; (3) the EVOOs were classified based on their volatile compounds.

Aroma Compounds of 15 Brands of EVOOs Extracted by SPME and SAFE
Eighty-nine compounds were screened by SPME/SAFE-GC-MS/O with chromatographic columns of DB-Wax and DB-5 (Table 1). Eight classes of flavor compounds were identified in the 15 brands of EVOOs, including 17 alcohols, 22 aldehydes, 9 ketones, 4 acids, 14 esters, 5 aromatics, 12 alkene, 6 others. The important aroma compounds in EVOOs are the C6 compounds, which were generated from lipoxygenase pathway through a series of enzymatic acting on fatty acids, such as linoleic and linolenic acids initiated by the tissue disruption [5]. The hydroperoxide-lyase enzyme produces aldehydes, subsequently reduced to alcohols by the alcohol dehydrogenase enzyme [19]. In this study, 17 alcohols were identified from the 15 brands of EVOOs (Table 1). Among them, hexanol, (E)-3-hexenol, (Z)-3-hexenol, and phenylethyl alcohol existed in all the 15 EVOOs. Compared to previous studies, these four compounds were also detected in virgin olive oils produced in Iran, Turkey, Spain, Greece, Tunisia, and Italy [2,8,11,12,15,16,[20][21][22][23]. (E)-3-Hexenol and (Z)-3-hexenol are derived from the reduction of (E)-3-hexenal and (Z)-3-hexenal by the action of an alcohol dehydrogenase (ADH). They present the most obvious green, fresh, and grass note. However, these aroma compounds were not thought to be responsible for the significant effect on olive oil odor, owning to their high odor threshold values in oil [19,24]. 2-Ethyl hexanol, linalool, octanol, 1-nonanol, undecanol, and benzyl alcohol were found in most of the 15 brands of EVOOs. 2-Ethyl hexanol, octanol, 1-nonanol, and benzyl alcohol were also reported in the previous studies about EVOOs from Turkey, Spain, Greece, Tunisia [2,8,16,23]. Nonanol presents fat or green note, which also derives from lipoxygenase pathway of unsaturated fatty acids [19]. Linalool with the flower and lavender note was detected in the samples of Turkey [2], while undecanol was not found in previous studies on EVOO flavor.

Acids
In this study, 4 acids were identified from the 15 brands of EVOOs, including acetic acid, propionic acid, hexanoic acid, and nonanoic acid (Table 1). These 4 acid compounds were not detected in all the 15 brands of EVOOs. They were all found in the study on Turkish olive oils as well [23]. Many kinds of virgin olive oils did not contain acid compounds [8,11,15,16,20,22,28]. Acetic acid was reported in the olive oil samples of Turkey, Tunisian [2,12,29]. Propionic acid and hexanoic acid existed in Iranian olive oil [11]. Nonanoic acid and acetic acid were related to fusty and winey-vinegary defects, respectively [26]. Negative aromas such as rancid, fusty, winey-vinegary, and frozen are sensory attributes of defective virgin olive oil recognized by the International Olive Council. In the next research, the panel test will also be applied to confirm if the 15 brands of EVOOs show olfactory defects due to nonanoic acid and acetic acid, as the oil quality will be compromised by the perceived intensity of the defect and lead to a downgrading.

Esters
In this study, 14 ketones were identified from the 15 brands of EVOOs (Table 1). Among these compounds, hexyl acetate, (Z)-3-hexenyl acetate, methyl benzoate, and methyl salicylate were found in all the 15 brands of EVOOs. They were also detected in the virgin olive oil from Italy, Greece, and Tunisia [8,9]. Besides, hexyl acetate and (Z)-3-hexenyl acetate were reported in the sample of Portugal, Turkey, Tunisia, and Spain, which possess the fruit and green note [2,12,13,15,20,22,23]. Methyl benzoate in the samples of Iran and Spain, and methyl salicylate in the ones of Tunisia and Turkey presented the herb and peppermint note [11,12,16,23]. Ethyl benzoate existed in several kinds of EVOOs with chamomile or flower note, which was reported in the samples from Iran [11].

Others
In this study, 6 other compounds were identified from the 15 brands of EVOOs (Table 1). All these compounds (2-pentylfuran, benzyl methyl ether, dimethyl sulfoxide, naphthalene, 4-ethylphenol and 4-allylanisole) only existed in some brands of samples. They had no contribution to the aroma of all the samples by FDs. Based on the studies before, there was no report about the compounds above in EVOOs. These six compounds were not the important aroma compounds in EVOOs.

Comparison of Extraction Effect of SPME and SAFE
Eighty compounds in EVOOs were identified by SPME-GC-MS/O and 54 by SAFE-GC-MS/O method (Table 1). Forty-four compounds were detected by both methods. Two extraction methods (SPME and SAFE) had different extraction efficiency for different kinds of volatile compounds (Figure 1). method (Table 1)

100
It can be seen from Figure 1 that more alcohols, aldehydes, ketones, esters, terpenes, and others 101 were detected by SPME compared to SAFE. The same numbers of acids and aromatics were 102 extracted by these two methods. The reasons can be the following: formyl groups of aldehydes and 103 carbonyl groups of ketones were unstable, which could get oxidized or reduced in organic solvent.

104
Alcohols in EVOOs are derived from the reduction of aldehydes by dehydrogenase, for example, 105 from (E)-2-pentenal to 1-pentene-3-ol, from (E)-2-hexenal to (E)-2-hexenol, from octanal to octanol, 106 which resulted in the more alcohols by SPME [31]. This result was in accordance with that of 107 watermelon juice [32]. It indicated that SPME had better extraction effect on aldehydes, ketones, and 108 alcohols just because of the different systems of EVOOs and watermelon juice.

109
Similarly, SPME could extract more esters than that by SAFE, while the result was just opposite 110 in natto [33]. In natto, all the esters were ethyl esters, and there were more complicated constitutes of   It can be seen from Figure 1 that more alcohols, aldehydes, ketones, esters, terpenes, and others were detected by SPME compared to SAFE. The same numbers of acids and aromatics were extracted by these two methods. The reasons can be the following: formyl groups of aldehydes and carbonyl groups of ketones were unstable, which could get oxidized or reduced in organic solvent. Alcohols in EVOOs are derived from the reduction of aldehydes by dehydrogenase, for example, from (E)-2-pentenal to 1-pentene-3-ol, from (E)-2-hexenal to (E)-2-hexenol, from octanal to octanol, which resulted in the more alcohols by SPME [31]. This result was in accordance with that of watermelon juice [32]. It indicated that SPME had better extraction effect on aldehydes, ketones, and alcohols just because of the different systems of EVOOs and watermelon juice.

l c o h o l s A l d e h y d e s K e t o n e s A c i d s E s t e r s A r o m a t i c s T e r p e n e s O t h e r s
Similarly, SPME could extract more esters than that by SAFE, while the result was just opposite in natto [33]. In natto, all the esters were ethyl esters, and there were more complicated constitutes of esters in EVOOs including ethyl, methyl, hexyl, and linalyl ester, which might result in the different extraction effect. SPME also had better extraction effect to terpenes. The previous studies showed that SPME with the fiber of DVB/CAR/PDMS (divinylbenzene/carbon/polydimethylsiloxane) had better extraction effect to terpenes compared to the other fibers (100 µm PDMS, 85 µm PA (polyacrylate), and 7 µm PDMS) or method (hydrodistillation), which confirmed the result in this study to some extent [34,35]. In conclusion, SPME had better extraction efficiency to the aroma compounds in EVOOs.

Identification and Quantification of Key Aroma Compounds of 15 Brands of EVOOs
Eleven compounds were identified as the key aroma compounds in alternative brands of EVOOs by SAFE-AEDA (Table 2)  Among these key aroma compounds identified by SAFE-AEDA, 1-octen-3-one could not be quantified because it had no peak in total ion chromatogram. 2-Octanone could be detected only in four samples by SAFE, so it could not be quantified by SPME with external standard; nonanal, 2-nonanone, (E)-2-nonenal and methyl benzoate could only be detected in several samples with the weak contribution (lower FD). Therefore, they were not quantified precisely. Except for the compounds above, all the other five compounds were quantified by external standard. Besides, (Z)-3-hexenyl acetate and 6-methyl-5-hepten-2-one were also identified as the key aroma compounds by SPME with internal standard, although they could not be sniffed by SAFE. Hence, all these seven compounds were quantified precisely (Table 3). As shown in Table 4, (Z)-3-hexenyl acetate showed the highest odor activity value (OAV), indicating the strong contribution to the flavor of EVOOs. (E)-2-heptenal, hexanal, acetic acid, and (E)-2-hexenal were also identified as the key aroma compounds based on OAVs by SPME. It was in accordance with the results by SAFE-AEDA. While (E)-3-hexenol and 6-methyl-5-hepten-2-one had little aroma contribution due to the lower OAVs, there were some differences between the results based on AEDA and OAV. Hence, the key aroma compounds could be identified precisely by both methods.

Comparison of Aroma Compounds of 15 Brands of EVOOs
As shown in Table 5, the detection frequency of compounds was compared based on the kinds of EVOOs and producing areas. The detection frequency of alcohols in EVOOs of Italy (subtotal: 54) and Greece (subtotal: 58) were higher than that of Spain (subtotal: 39) generally. In particular, there were only five and seven kinds of alcohols in PL and BDS. There was a big difference on that of esters among the samples from these three countries, EVOOs of Spain showed the highest detection frequency (39). Similarly, for the other compounds, there was also the highest detection frequency from samples of Spain. While the ones of aldehydes, ketones, acids, aromatics, and terpenes were very close among the samples of these three areas, for the brand of sample, the range of detection frequency was from 41 to 59. The highest one (OL: 59) and the lowest one (BLN: 41) were both from Spain. The impact factor of aroma compound of EVOOs included the cultivar of olive, growing environment, producing technology, storage condition, and so on. From this aspect, these factors might impact EVOO aroma of Spain greatly.
Grouping different types of oil is meaningful by flavor because it is an important criterion for EVOO. In the previous studies, cluster analyses were applied to distinguish the different types of edible oil based on their flavor. Sesame oils, soybean oils, and peanut oils could be completely classified using cluster analysis of volatiles [36]. Also, hierarchical cluster analysis showed similarities between EVOO, VOO and LOO (lampante olive oil) samples based on HS-GC-IMS fingerprints [37]. It can be seen from Figure 2, EVOOs were clustered in the heatmap based on their aroma compounds extracted by SPME and SAFE. From Figure 2A, EVOOs of BLN, OL, BDS, OLWL, and YGY had the similar aroma components, all these samples were from Spain except for OLWL. In addition, the flavor was similar between the samples from Italy (ALCF, MNN, AN, and OS) and Greece (DMDN and AGL). From Figure 2B, the aroma compounds of five brands of EVOOs from Spain (PL, OL, BDS, YGY, and BLN) were alike, and they were analogous between the ones from Italy (OS, AN, and MNN) and Greece (DMDN, AGL, and MSWN). Based on these results by SPME and SAFE, the flavor of Spain EVOOs was similar, and that of some of samples from Italy and Greece was alike. It indicated that there was big difference between samples of Spain and the other countries. These results seemed to contradict with that of detection frequency, but it just suggested that the similarity depended not only on the number of compounds but also on the content of ones.

Samples
Fifteen varieties of EVOOs from the three largest export countries of olive oil in the world were purchased from the import supermarket. All the samples were authenticated by "Inspection and Quarantine Certificate of Entry Goods" from Entry-exit Inspection and Quarantine Bureau of the People's Republic of China. EVOO samples were stored at 15 °C in dark. Specific brands were used in this study as follows, Spain: PL (MUELOLIVA), BDS (BETIS), OL (EURO GOLD), BLN

Samples
Fifteen varieties of EVOOs from the three largest export countries of olive oil in the world were purchased from the import supermarket. All the samples were authenticated by "Inspection and Quarantine Certificate of Entry Goods" from Entry-

Aroma Extraction of EVOO by SPME
The method was in accordance with Wang et al. with minor modification [38]. After being selected, a manual SPME (Supelco, Inc., Bellefonte, PA, USA) with a 50/30 µm divinylbenzene/ carboxen/polydimethylsiloxane (DVB/CAR/PDMS) SPME fiber was used for volatile extraction after the fiber had been conditioned at 250 • C for 30 min. Ten milliliter of EVOO sample was quickly transferred into a 40 mL vial, 1 µL of 4-methyl-2-pentanol was added as an internal standard at the concentration of 401 µg/µL. After the equilibrium of 60 • C for 20 min, a stainless steel needle, housing the SPME fiber, was placed through the hole to expose the fiber at the position of 1 cm over the liquid surface for 40 min. The vials were sealed tightly with screw caps fitted with a Teflon/silicon septum. Vials were continuously swirled during SPME exposure with an agitation speed of 100 rpm.

Aroma Extraction of EVOO by SAFE
The method was in accordance with Usami et al. with minor modification [39]. Fifty milliliters of EVOO was transferred into 250 mL Teflon bottle, then 150 mL of dichloromethane and 1 µL of 4-methyl-2-pentanol was added as an internal standard at the concentration of 401 µg/µL. The bottle was placed in a shaker for 8h at 4 • C and 180r/min. The volatile compounds were separated from the solvent extracts using SAFE [17]. The filtrate was vacuum distilled using SAFE apparatus as previously described [40].

Gas Chromatography-Olfactometry-Mass Spectrometry (GC-MS/O) Analysis
The method was in accordance with Nuzzi et al. with minor modification [41]. The qualitative and quantitative analyses of the volatile compounds were conducted using Agilent 7890A gas chromatograph coupled with an Agilent model 7000B series mass spectrometer (GC-MS) and desorbed for 7 min in a split/splitless GC injection port, which was equipped with an inlet linear specific for SPME use (Agilent Technologies, Wilmington, DE, USA). The GC-MS was equipped with a sniffer 9000 Olfactometer (Brechbühler, Switzerland). The volatiles were separated on DB-5 and DB-Wax (30 m × 0.25 mm i.d. × 0.25 µm, J&W Scientific), a type of fused silica capillary columns.
The oven temperature was initially at 40 • C, held for 3 min, ramped at 5 • C/min to 200 • C, then ramped at 10 • C /min to 230 • C and held for 3 min, then baked at 250 • C for 3 min. The injection port and ionizing source were kept at 250 and 230 • C, respectively; the carrier gas was helium at 1.2 mL/min. The injector mode was splitless. Electron-impact mass spectra were generated at 70 eV, with an m/z scan range from 35 to 350 amu. Compounds were identified according to NIST 14.0 mass spectra libraries installed in the GC-MS equipment.
A sniffing port (Sniffer 9000) coupled to a GC-MS instrument was used for odor-active compound characterization. At the exit of the capillary column, the effluents were split 1:1 (by volume) into a sniffing port and a MS detector by employing the Agilent capillary flow technology; the transfer line to the GC/O sniffing port was held at 280 • C. GC/O was performed by three experienced panelists.

Aroma Extract Dilution Analysis (AEDA)
The highest sample concentration after SAFE was assigned with a FD factor of 1. The volatile components were stepwise diluted at the ratio of 1: 2 with dichloromethane, and aliquots of the dilutions (1 µL) were subjected to analysis. The process was stopped when aromas ceased to be detected by the evaluators. The result was expressed as the FD factor, which was the ratio of initial and final concentration of the odorant in the sample.

Identification of Volatile Aroma Compounds
The method was in accordance with Xu et al. with minor modification [42]. The chemical identification was performed using a mass spectrum database, the linear retention index (LRI), and aromatic characteristics by sniffing. Some important aroma compounds were identified by comparison with standard compounds. LRI was calculated using normal alkane series and compared with the references. Mass spectra identification was performed based on the NIST 2.0 mass spectra libraries. The formula for the calculation of LRI was (1).
where "n" represents the number of carbon atoms of n-paraffins; "t n " represents the retention time of n-paraffins C n , "t n+1 " represents the retention time of n-paraffins C n+1 , "t a " represents the retention of an unknown compound in the sample time (to be satisfied "t a " is between "t n " and "t n+1 ").

Quantification of Volatile Aroma Compounds
SPME was used to extract the volatile aroma components in olive oil for quantification analysis. At the same time, two quantitative methods were used, including an internal standard method for total aroma compounds and an external standard method in SIM mode for the key aroma compounds. In the internal standard method, 4-methyl-2-pentanol (401 µg/µL in hexane) was added as an internal standard to the sample to calculate the target compound concentration.
where "C IS " represents the concentration of internal standard; "A IS " represents the peak area of the internal standard; "C X " represents the concentration of the target compound; "A X " represents the peak area of the target compound. Hexanal, (E)-2-hexenal, (Z)-3-hexenyl acetate, (E)-2-heptenal, 6-methyl-5-hepten-2-one, (Z)-3-hexenol, and acetic acid were quantified using SIM mass spectrometry by standard curve method. The solutions of the mixture of 4-methyl-2-pentanol and reference compounds at different concentrations were prepared and analyzed by GC-MS. The standard curves were prepared by plotting the ratio of the peak areas of the reference compound relative to 4-methyl-2-pentanol against their concentration ratio.

Calculation of OAV
OAV was calculated using the following equation: where C i is the concentration of the compound in the watermelon juice and OT i is its odor threshold. Compounds with OAV equal to or greater than 1 actually contribute to aroma as an odor-active compound because they are above their odor threshold, whereas those with OAV smaller than 1 may not.

Statistical Analysis
Analysis of variance (ANOVA) were carried out by using the software SAS (SAS Institute Inc., Cary, NC, USA). The ANOVA test was performed for all experimental runs, to determine the significance at 95% confidence interval. All experiments were performed in triplicate. The cluster analysis was applied to the concentration of volatile compounds data, and it was conducted by using the software of Morpheus online (https://software.broadinstitute.org/morpheus/). In this process, hierarchical clustering was selected, "Euclidean distance" and "Average" were applied to calculate the distance of samples and groups.

Conclusions
Eighty-nine compounds were screened by SPME/SAFE-GC-MS/O with chromatographic columns of DB-Wax and DB-5 in 15 brands of EVOOs. Eighty compounds were identified by SPME-GC-MS/O and 54 by SAFE-GC-MS/O method. Forty-four compounds were detected by both methods. Undecanol, (Z)-4-decenal, (E)-2-dodecenal, and 2-nonanone were found in EVOOs for the first time only by SAFE. 2-Nonanone was the key aroma compounds identified by AEDA. SPME had better extraction efficiency to aroma compounds in EVOOs, while SAFE could extract the most effective components (seven key aroma compounds). Hence, it is feasible and advantageous that SAFE is applied to the aroma extraction of EVOOs.
Eight classes of flavor compounds were identified in the 15 brands of EVOOs, including 17 alcohols, 22 aldehydes, 9 ketones, 4 acids, 14 esters, 5 aromatics, 12 alkene, and 6 others. Eleven compounds were identified as the key aroma compounds in alternative brands of EVOOs by SAFE-AEDA. Key aroma compounds authenticated by AEDA had some differences from those by OAVs. The aroma compounds of Spain EVOOs were similar, and those of some of samples from Italy and Greece were alike.
Author Contributions: Q.Z. collected samples, drafted and revised the manuscript; S.L. carried out the laboratory work and performed statistical analysis; Y.L. conceived, designed and coordinated the study; H.S. helped to edited and revised the manuscript. All authors gave final approval for publication.

Conflicts of Interest:
The authors declare no conflict of interest.