From Extra Virgin Olive Oil to Refined Products: Intensity and Balance Shifts of the Volatile Compounds versus Odor

To explore relationships between the volatile organic compounds (VOCs) of different grades of olive oils (OOs) (extra virgin olive oil (EVOO), refined olive oil (ROO), and pomace olive oil (POO)) and odor quality, VOCs were measured in the headspace of the oils by proton transfer reaction quadrupole ion guide time-of-flight mass spectrometry. The concentrations of most VOCs differed significantly between the grades (EVOO > ROO > POO), whereas the abundance of m/z 47.012 (formic acid), m/z 49.016 (fragments), m/z 49.027 (fragments), and m/z 115.111 (heptanal/heptanone) increased in that order. Although the refined oils had considerably lower VOC abundance, the extent of the decline varied with the VOCs. This results in differences in VOCs proportions. The high VOC abundance in the EVOO headspace in comparison to ROO and POO results in a richer and more complex odor. The identified C5–C6 compounds are expected to contribute mainly to the green odor notes, while the identified C1–C4 and C7–C15 are mainly responsible for odor defects of OOs. Current results reveal that processing strongly affects both the quantitative and relative abundance of the VOCs and, therefore, the odor quality of the various grades of OOs.


Introduction
Olive oils (OOs) are very popular with customers due to their pleasant flavor and odor, as well as their health benefits. The concentrations of the volatile compounds (odor profile) in OOs are affected by many factors, including the cultivar [1], environmental factors [2], olive fruit maturity [3], the technical processing [4], as well as storage of the fruit (long time storage may be responsible for odor defects) [5], or storage of OOs (oxidative degradation) [6]. Therefore, OOs come with great variation in odor quality. Among those factors, processing methods (e.g., cold pressing and refining steps) cause dramatic effects on the concentrations of the volatile organic compounds (VOCs) in different grades of OOs [7], such as extra virgin olive oil (EVOO), refined olive oil (ROO), and pomace olive oil (POO).
The VOC molecules comprising 5 and 6 carbon atoms (called C5 and C6 compounds), which are mainly produced through enzymatic reactions leading to degradation of polyunsaturated fatty acids during processing [8], are considered the most important VOCs for the green odor notes of EVOO [9,10]. These compounds are more likely to be formed under cold-pressed conditions. Some oils are subjected to refining and most of C5 and C6, as well as many other VOCs are removed during the deodorization step in the refining process [11,12]. VOCs of OOs have been studied before [13][14][15], but so far, the differences in headspace concentrations of the VOCs of different OO grades has not yet been studied

PTR-QiToF-MS Spectral Profile
Two hundred OOs were subjected to PTR-QiToF-MS analysis, and 295 mass peaks in the range of m/z 18.033 to m/z 207.204 were obtained for each sample. The 10 log transformed mass spectra for three grades, i.e., EVOO (n = 140), ROO (n = 45), and POO (n = 15), are presented in Figure 1. For each average spectral profile in Figure 1, the summed observed concentration of the mass peaks ranging from m/z 18.033 to m/z 207.204 in the headspace of the samples were calculated. The value for EVOO (782 parts per million by volume (ppmv)) was about 3 times higher than the summed value for ROO (276 ppmv) and 5 times higher than that value for POO (127 ppmv). This indicates that EVOO is richer in VOCs than the other lower grades of OOs. Furthermore, the concentration of most mass peaks measured in the headspace of the EVOO samples is considerably higher than the other two grades. It further supports the idea that large amounts of the VOCs are removed during the refining process [11,26].

Measured
Superscript letters a and b in a row indicate significant differences (p < 0.05). A C1-C7 indicate that the VOCs with 1-7 carbon atoms in the molecule, respectively. B The OT of trans-2-pentenal. C The OT of pentanal. D The OT of trans-2-hexenal. E The OT of hexanal. F The OT of 3-octen-2-one. G The OT of 1-octen-3-ol. H The OT of propyl butyrate. I 2981, 5465, 119 refer to the OTs of α-pinene, β-pinene, limonene, respectively. J The OT of tran-2-nonenal. K The OT of hexyl acetate. L 2,4-Decadienal. Green indicates that the odor activity value (OAV, the average concentration of the volatile compound of the oils divided by its OT) is more than 2. Yellow indicates that the OAV is between 1 and 2. Red indicates that the OAV is less than 1. Rows without color indicate that no OT was found for this compound.
Molecules 2020, 25, 2469 7 of 18 by volume (ppmv)) was about 3 times higher than the summed value for ROO (276 ppmv) and 5 times higher than that value for POO (127 ppmv). This indicates that EVOO is richer in VOCs than the other lower grades of OOs. Furthermore, the concentration of most mass peaks measured in the headspace of the EVOO samples is considerably higher than the other two grades. It further supports the idea that large amounts of the VOCs are removed during the refining process [11,26].

VOCs with Higher Concentrations in the EVOO Headspace
Among the 295 mass peaks, 287 mass peaks were present in significantly higher concentrations in the headspace of the EVOO samples than for the other OOs. It confirms that most of the VOCs are removed during the refining process [11]. Similarly, for 51 out of the 54 tentatively identified VOCs, significantly higher concentrations were observed for EVOO than for the other oils (Table 1).
Except for the mass peak m/z 47.012 (formic acid), significantly higher headspace concentrations were determined for all identified C1-C4 compounds in EVOO compared to ROO/POO. Mass peak m/z 33.033 (methanol) was the most abundant compound in the EVOO headspace, followed by mass peaks m/z 45.034 (acetaldehyde), m/z 47.049 (ethanol), m/z 43.018 (esters), m/z 57.033 (2-propenal), and m/z 61.028 (acetic acid). However, the detection of methanol is rarely reported, and/or methanol is often found to be present at low concentrations using gas chromatography-mass spectroscopy-based techniques. There are several possible reasons for the detection of high concentrations of the VOCs with low molecular mass in the headspace of OOs using PTR-QiToF-MS. One of the possible reasons is that this mass may be a mixture of a small amount of methanol and a large amount of fragments from higher molecular masses. It may also be due to differences in the set-up between the current method and gas chromatography approaches. For instance, the way of sampling and injection differ considerably.
Furthermore, significantly higher concentrations of 28 out of 30 identified C7-C15 compounds were determined in the headspace of EVOO than for the other OO counterparts. Among these 28 identified compounds, the concentrations of mass peaks m/z 93.070 (254 ± 223 ppbv, toluene, C1-benzene), m/z 105.09 (324 ± 247 ppbv, ethenyl benzene, C2-benzene) and m/z 107.086 (338 ± 275 ppbv, ethyl benzene, C2-benzene) found in the headspace of EVOO were higher than the other compounds. It is reported that oils can be easily contaminated by these potentially harmful VOCs because of their lipophilic nature and widely distribution [45]. The presence of these compounds in OOs is likely due to the contamination by gasoline vapors in the oil mill [46,47]. These compounds could also originate from the packaging materials [48]. The contamination of these compounds in OOs deserves a special consideration in the future due to their potential harm to human body [49]. Furthermore, significantly higher concentrations of those three compounds in the EVOO headspace compared to the other OOs is likely due to removal during the refining process [48].

VOCs with Higher Concentrations in the ROO/POO Headspace
It is interesting to note that four mass peaks (m/z 47.012, m/z 49.012, m/z 49.027, and m/z 115.111) were present in significantly higher headspace concentrations in ROO and POO compared to EVOO ( Figure 1 and Table 1).
Mass peak m/z 47.012 was tentatively identified as formic acid [31,32]. The headspace concentration of formic acid in EVOO (3 ± 3 ppmv) was significantly lower than in ROO (12 ± 11 ppmv) and POO (17 ± 16 ppmv). One of the possible pathways contributing to the formation of formic acid (m/z 47.012) in the lower grades of OOs is oxidation during storage [31], such as the decomposition of unstable volatiles (2,4-(E-E)-decadienal) [31,50]. Another possible pathway is microbial metabolism during storage [51]. The other two mass peaks m/z 49.012 and m/z 49.027 (non-identified) were also present in significantly lower concentrations for EVOO compared to ROO and POO. They are most likely fragments of higher molecular masses. In addition, the concentration of mass peak m/z 115.111 (51 ± 31 ppbv, heptanal/heptan-2-one) for EVOO was significantly lower than that for ROO (81 ± 65 ppbv) and POO (77 ± 63 ppbv). This mass peak was tentatively identified as heptanal or heptan-2-one (Table 1) [9,39], which originates from the decomposition of linoleic acid [43]. The formation of this compound in ROO and POO most likely occurs during storage, because the steam deodorization step before storage would have removed such organic compounds [12,52].
Taken together, the concentrations of most VOCs were significantly lower in the headspace of OOs that have been subjected to a refining step, whereas the concentrations of four mass peaks, i.e. m/z 47.012 (formic acid), m/z 49.016 (non-identified), m/z 49.027 (non-identified) and m/z 115.111 (heptanal/heptan-2-one), presented a reversed trend.

Odor Implications
Some groups of scientists relate the odor contribution of a certain VOC to the human perceivable odor not only to be related to its concentration, but also to its OT [8,9,36,39]. This approach allows some ranking of the VOCs in terms of their relevance to the odor. When the odor activity value (OAV, the average concentration of the volatile compound of the oils divided by its OT) of the volatile compound is greater than one, the odor of this compound is expected to contribute to the odor of the oils according to this theory [53]. In this study, we looked into the odor relevance of compounds using the OTs.
Considering the average concentrations of the identified C1-C4 compounds in OOs and their OTs in Table 1, the odor of those compounds (acetaldehyde, OAV = 185; 2-propenal, OAV = 19; propanal, OAV = 24; acetic acid, OAV = 43; dimethyl sulfide, OAV = 16; propanoic acid, OAV = 12; dimethyl sulfoxide, OAV = 68; butanoic acid, OAV = 143. OAV is calculated based on the average concentration for the OO grade with lowest intensity and its OT in Table 1) are considered to contribute strongly to the odor of the oils due to their high OAVs. Surprisingly, most of those compounds are associated with odor defects.
Regarding the identified C5 compounds, trans-2-pentenal (m/z 85.064), associated with green-fruity odor note [1,9], was present with an OAV of 16 for EVOO. Regarding the identified C6 compounds, trans-2-hexenal (m/z 99.081), associated with a green-fruity odor note [1,9], was present with an OAV of 13000 for EVOO. Hexanal (m/z 101.095), associated with a green-sweet odor note [9], was present with an OAV of 570 for EVOO. Butyl acetate, ethyl butyrate, and ethyl isobutyrate (m/z 117.091), associated with a green-sweet-fruity note [1,9], were present with OAVs of 23, 82, and 131 for EVOO, respectively. Therefore, trans-2-pentenal, trans-2-hexenal, hexanal, butyl acetate, ethyl butyrate, and ethyl isobutyrate might be the relevant contributors to the green odor notes of EVOO. This is in agreement with previous studies [9,10] reported that the C5-C6 compounds were described as the most important VOCs in terms of the contribution to the green odor notes for EVOO. Moreover, trans-2-hexenal and hexanal are most likely the most important contributors to the green odor notes of the EVOO due to the highest OAV (13000 and 570, respectively) compared to the other compounds. These results agree with those in a previous study, which reported that the identified C6 aldehydes (especially trans-2-hexenal and hexanal) contribute to the green odor notes in European EVOO [37]. In addition, the OAVs of those compounds mentioned above for EVOO were higher than its lower grade counterparts. Furthermore, it is reported that a great amount of the VOCs associated with the green odor notes have been found in high-quality/grade OO (EVOO) [3,54]. Therefore, the odor of those compounds most likely contributes to the differences in perception of the green odor notes between the premium grade EVOO and the lower grades of OOs.
Although the identified C7-C15 compounds have relatively low concentrations in OOs compared to the identified C1-C6 compounds, they were also components of the volatile odor fraction in OOs, especially the identified C7-C10 compounds [10]. The OAVs of some of these compounds were over one in OOs (2, . However, most of those compounds are associated with odor defects. Hexyl acetate (m/z 145.122), associated with a green-fruity note, was present with an OAV less than one in OOs, which support previous research [55]. This indicates that this compound might not be a relevant contributor to the green odor notes of OOs. Thus, those minor compounds are more likely related to odor defects of OOs due to their high OAV value and related odor notes.
Summarizing, the identified C5-C6 compounds mainly possess the green odor notes, while the identified C1-C4 and C7-C15 compounds are mainly associated with odor defects. EVOO has 31 volatile compounds exceeding an OAV of one, which is more than ROO (30 volatile compounds) and POO (26 volatile compounds). EVOO is also present with higher OAV values for 29 out of these 31 compounds compared to ROO and POO. Thus, most likely, these VOCs contribute to the richer and more complex odor of EVOO compared to ROO and POO. This is similar to the result in Section 2.2 that EVOO were present with significantly higher headspace concentrations of the VOCs in comparison to ROO and POO.
Consumers' preference in OOs is mainly related to the odor descriptors qualified with the 'green' note [56]. Therefore, the green notes are fairly important sensory traits. In Table 1, trans-2-hexenal (m/z 99.081), hexanal (m/z 101.095), butyl acetate, ethyl butyrate, and ethyl isobutyrate (m/z 117.091) are expected to contribute to the green odor notes of OOs, since their OAVs are greater than one. In order to compare the full sets of samples, the scatter plots of the 10 log transformed concentrations of m/z 99.081, m/z 101.095, and m/z 117.091 are presented for all samples in Figure 2. The plots show distinct clustering of the three grades of OOs. EVOO (located in the upper right corner in Figure 2) grouped separately from the lower grades of OOs (widely spread in the lower left corner). This indicates that EVOO was present with consistently higher concentrations of these compounds with green notes, and with OAV values >1, in the headspace of EVOO, which is in agreement with previous studies [3,54].
compare the full sets of samples, the scatter plots of the 10 log transformed concentrations of m/z 99.081, m/z 101.095, and m/z 117.091 are presented for all samples in Figure 2. The plots show distinct clustering of the three grades of OOs. EVOO (located in the upper right corner in Figure 2) grouped separately from the lower grades of OOs (widely spread in the lower left corner). This indicates that EVOO was present with consistently higher concentrations of these compounds with green notes, and with OAV values >1, in the headspace of EVOO, which is in agreement with previous studies [3,54].

Relative Concentration Differences of the VOCs
To explore the VOCs proportions of OO grades, the relative average concentrations and standard deviation of 54 tentatively identified VOCs in the headspace of each grade (EVOO, ROO, and POO) are shown in Table 2. Table 2. Tentative identification volatile organic compounds (VOCs), average relative concentrations, standard deviation (SD), and statistical comparisons (Kruskal-Wallis tests and Mann-Whitney U-tests, p <
In total, the VOCs proportions were different in the headspace of different OO grades, which is most likely due to the different processing methods. It is known from the literature that absolute concentrations of the VOCs are important for the odor traits of products, but the balance of the VOCs is just as important [57].

Samples Preparation
For this study, 240 OOs were gathered from producers, traders, and retailers across Europe in 2016 and 2017. The authenticity of the 180 EVOO samples was verified by fatty acid fingerprints combined with chemometrics [58], ultraviolet-visible spectra analysis [59], and evaluated by 2/3-monochloropropane-1,2-diol and glycidyl esters analysis [60]. A total of 40 EVOO samples did not meet the requirements for one or more of these methods and were therefore removed from the sample set. In total, 200 OOs were used in this study, which consisted of 140 EVOO, 45 ROO, and 15 POO. Prior to analysis, all the oils were sealed and stored in the dry and dark environment at 18 ± 1 • C. Sampling was completed within 6 months. To avoid long-term storage, the analyses were carried out within two weeks after sampling of each sample.
The sample preparation method was similar to our previous study [25] with minor modifications. Firstly, 5 mL of oil was transferred into a 250 mL flask. Then, the closed flask was kept in a water bath at 30 • C for 30 min to equilibrate the headspace before instrumental analysis.

PTR-QiToF-MS Analysis
PTR-QiToF-MS (Ionicon Analytik GmbH, Innsbruck, Austria) was operated with a drift voltage of 999 V, a drift temperature of 61 ± 1 • C, a drift pressure of 3.803 mbar, and an E/N value of 134 ± 1 Townsend. The laboratory air was measured for the first 10 s as a blank before each sample. Then, the VOCs in the headspace of the flask were transported to the PTR-QiToF-MS through a peek inlet tube with a temperature of 60 ± 0.5 • C. The flow rate of the air in the tube was 61 ± 2 mL min −1 .
The measurement time was 30 s. The acquisition rate was 1 spectrum per second. On each of two different days, one independent replicates per sample was measured. Samples were analyzed in a random sequence to avoid any order bias. Results were stored in the system automatically.

VOCs Data Pre-Processing
All the raw data obtained from the PTR-QiToF-MS machine were integrated by PTRwid software (Utrecht University, Utrecht, the Netherlands; http://www.staff.science.uu.nl/~{}holzi101/ptrwid/) [61]. The unified mass list with the ion count per second (cps) of each sample were provided after the autonomous mass scale calibration, as described by Holzinger [61]. The average of the 30 sample scans and the average of the 10 blank scans were calculated separately. The VOC concentrations (molecules per cm 3 ) were calculated from cps according to Equation (1) [62]: where t is the residence time of the primary ions in the drift tube, k is the coefficient of the reaction rate with a value of 2 × 10 −9 cm 3 /s, [VOC·H + ] measured is the ion count rate of the protonated VOC, [H 3 O + ] measured is the ion count rate of the protonated water at m/z 21.022, 487 is the intensity ratio of the protonated water at m/z 19.018 (100%) to the protonated water at m/z 21.022 (0.2055%) [63], (m/z) H3O+ and (m/z) VOC·H+ are the molecular weight of protonated water and protonated VOC. Subsequently, the unit of molecules per cm 3 was converted to ppbv, on the basis of ideal gas using Equation (2) [64]: where P (Pa) is the pressure, V (cm 3 ) is the volume, n (mol) is the number of moles, R (J K −1 /mol) is the gas constant, and T (K) is the temperature. After unit conversion, the average of each sample's 10 blank cycles were subtracted from the sample's averaged scan. The replicates of each sample were checked using autocorrelation [65], and the sample will be removed when the correlation value was below 0.9. In this study, the correlation values of all the samples were over 0.9. Finally, sample averages were calculated from the data of two replicate measurements.
Sample independent ions, such as N 2

Relative Concentration
The relative concentration (C, %) of each mass of each sample was calculated by the intensity of the single mass (I s , ppbv) per sample and the total mass intensity (I t , ppbv) per sample using Equation (3):

Data Analysis
Significant differences of the (relative) concentrations of tentatively identified VOCs for different grades of OOs were assessed using non-parametric Kruskal-Wallis tests (p < 0.05). Mann-Whitney U-tests were used to perform pairwise comparisons between OO grades. These analyses methods were performed using SPSS (version 23, IBM, Chicago, IL, USA).

Odor Threshold in Air
The OTs of the identified compounds were collected from several publications [29,39,[66][67][68][69]. Firstly, the OT in oil were converted into the OT in air using Equation (4): where C a1 (µg/L) is the OT in air, C o (µg/kg) is the OT in oil, K is the air/liquid partition coefficient of compound [67], and ρ is the density of olive oil (0.916 kg/L). Then, the unit of the OT in air was converted using Equation (5) [64]: where C a2 (ppbv) is the OT in air, V m is the standard molar volume of ideal gas at 1 bar and 298 K with a value of 24.77 L/mol [70], and M (g/mol) is the molecular weight of the compound. Subsequently, the lowest OT of each compound was used in this study. The calculated OTs in air from various literature sources are provided in the Supplementary Material (Table S1). The OAV was calculated using the average concentration of the volatile compound for each of the type of OO divided by the corresponding OT.

Conclusions
Significant differences in VOC headspace concentrations were determined for the different grades of OOs. Most of the VOCs were present in significantly higher concentrations for EVOO than for ROO/POO. However, significantly higher concentrations of mass peaks m/z 47.012 (formic acid), m/z 49.016, m/z 49.027, and m/z 115.111 (heptanal/heptanone) were found for the lower grades of OOs (ROO/POO) compared to EVOO. Furthermore, significant differences of the VOCs proportions were observed indicating a distinct change in the balance of the VOCs across OO grades. Thus, EVOO and ROO/POO not only differ quantitatively (concentrations of compounds) but also qualitatively (proportions of compounds). Comparison with OAVs of the compounds revealed the expected change in contribution to the odor of the OOs. Our results underpin the well-known richer and more complex odor of EVOO by the elevated contribution of the VOCs with green notes exceeding the minimal OAV. Furthermore, the consistent differences in VOCs concentrations between EVOO and other grades of OO may provide potential for verification of the identity of OOs.