Determination of volatile compounds for the differentiation of PDO fortified wines with different ageing methods as a tool for controlling their authenticity

Food Research International 173 (2023) 113320 Available online 26 July 2023 0963-9969/© 2023 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/bync-nd/4.0/). Determination of volatile compounds for the differentiation of PDO fortified wines with different ageing methods as a tool for controlling their authenticity Cristina Ubeda, David Cortejosa, M. Lourdes Morales , Raquel M. Callej´ on , Rocío Ríos-Reina * Departamento de Nutrici´ on y Bromatología, Toxicología y Medicina Legal. Facultad de Farmacia, Universidad de Sevilla. Sevilla, Spain ARTICLE INFO Keywords: Fortified Wines Protected Designation of Origin Ageing Volatile Compounds SPME ABSTRACT The aim of this work was to study the differentiating volatile profiles of the Spanish protected designation of origin (PDO) fortified wines obtained by headspace solid phase microextraction in conjunction with gas chromatography-mass spectrometry and powerful chemometric tools, to finally identify the marker volatile compounds most related to fortified wine types. Results revealed a satisfactory discrimination, for the first time, of the different types of PDO fortified wines, involving only a reduced number of volatile compounds selected by chemometrics. Thus, 28 volatile compounds were responsible for the differentiation according to ageing type (biological, oxidative, or mixed) resulting useful markers for the identification of each specific type of fortified wine. Among them, some esters were strongly related to biological ageing, aldehydes and acids to oxidative ageing, and lactones to mixed ageing. These volatile molecules involved in their differentiation could explain the unique organoleptic characteristics or attributes of these PDO fortified wines. 1. Introduction Spain is one of the main producers of high-quality wines. Particularly the fortified wines, which are elaborated in Andalusia, have acquired great prestige for being unique due to their production in a specific geographical area with traditional methods, the grape variety used, the climate and the soil. Such is their distinguishing feature achieved that they have been protected by the European Union with the indication ‘Protected Designation of Origin’ (PDO). Thus, there are four PDO of fortified wines in Andalucía (‘Condado de Huelva’, ‘Jerez-X´ er` es-Sherry’, ‘Manzanilla-Sanlúcar de Barrameda’, and ‘Montilla-Moriles’). Furthermore, within each PDO, there are recognised different categories according to their particular winemaking conditions such as the ageing process (Fino and Manzanilla, Oloroso, Amontillado and Palo Cortado) (BOJA N◦34 de 16/02/2018a, BOJA N◦34 de 16/02/2018b, BOJA N◦ 70 de 12/04/2018). The production of these wines is divided in two phases: firstly, production of the base wine through total or partial fermentation of the grape must and, secondly, wine ageing, which is the phase of ageing in wood barrels to which these wines are subjected after the addition of alcohol to raise their alcoholic degree (a process known as ‘fortification’) and to achieve the particular organoleptic qualities of their respective types of wine. The other particularity is that the ageing period, in addition to be mainly performed by the traditional system of ‘criaderas and solera’, can be of three kinds: biological ageing, by means of which the Fino and Manzanilla wines are obtained; oxidative ageing, which is the one that takes place in the Oloroso wines; and the combination of both ageing methods from which Amontillado and Palo cortado wines are obtained. The wines submitted to biological ageing are fortified until they reach an alcoholic content of 15.5◦. This enables the formation of the socalled ‘velo de flor’, which is a film of veil yeast that grows spontaneously on the wine surface, under which the wines age, preventing oxidation. This phenomenon is undoubtedly one of the main reasons for the uniqueness of these wines, as the action of the metabolism of these yeasts causes significant changes in the wine and, therefore, in its definitive organoleptic characteristics. In oxidative ageing, the wines are fortified until they reach an alcohol content of at least 17◦, above which biological activity becomes impossible, even for the veil yeasts, and they aged exposed to the direct action of oxygen, which can be seen visually by the gradual darkening of the wine’s colour (BOJA N◦34 de 16/02/ 2018a, BOJA N◦34 de 16/02/2018b, BOJA N◦70 de 12/04/2018). There is also the mixture of both ageing process from which Amontillado and Palo Cortado wines are produced. In the case of Amontillado wine, it is * Corresponding author. E-mail address: [email protected] (R. Ríos-Reina). Contents lists available at ScienceDirect Food Research International journal homepage: www.elsevier.com/locate/foodres https://doi.org/10.1016/j.foodres.2023.113320 Received 2 May 2023; Received in revised form 28 June 2023; Accepted 24 July 2023 Food Research International 173 (2023) 113320 2 well established that there is a first step of biological ageing under the ‘velo de flor’ that can entail several years (usually 5), followed by an oxidative step after the fortification of the wine and stoppage of the flor yeast growth (usually 2–3 years). However, Palo Cortado is a wine that, despite going through both ageing methods, the first step, biological ageing, used to last only between 6 months and 2 years because the oenologists of the winery detect its potential for Palo Cortado and then add alcohol to 17 alcoholic degrees avoiding the growth and starting oxidative ageing (BOJA N◦34 de 16/02/2018a, BOJA N◦34 de 16/02/ 2018b, BOJA N◦70 de 12/04/2018). This genuine and rare production process has always been surrounded by a bunch of mystery, which has brought with it, together with the high prices, several falsifications, and adulterations. Surprisingly, there are scarce research that studied the volatile profile of Palo Cortado wines, and even less with the application of chemometric techniques. The great diversity of high-quality wines on the market and the increase in their demand makes it is necessary to characterize them to establish quality and authenticity control parameters, thus protecting and assuring consumers that the product they are purchasing has the quality and characteristics declared. Although a multitude of wine parameters can be analysed, one of the most important, whose characterisation can be used to guarantee its authenticity, is the aroma. Wine in general, and particularly fortified wines, are very complex products, containing a wide variety of organic and inorganic substances in addition to water and alcohol. The focus on the aroma has been object of study for the characterization of these products since it is considered one of the most relevant quality criteria for wine (Zea et al., 2001, 2010a). It depends on the volatile compounds profile of the wine, having a primary origin, the grape used; secondary origin, fermentation by yeast and production process; and finally, tertiary origin, the ageing process. Therefore, by studying them, it is possible to characterise and differentiate wines based on the main sources of variation (raw material, production process and ageing). Even though, some authors have previously studied the volatile profile of some of these fortified wines (García-Moreno et al., 2021; Moyano et al., 2002; Zea et al., 2010b; Zea et al., 2001), they have some limitations. Hence, some of them considered a limited number of samples, which were sometime not controlled samples or without PDO, or studied only one type of fortified or PDO wine and not all at the same time, or even they did not employ chemometric tools or multivariate data analysis. Gas chromatography-mass spectrometry (GC–MS) has been the most widely-used technique for analysing volatile compounds in wine and its derivates, after a proper extraction method, being headspace-solid phase microextraction (HS-SPME) one of the employed one for analysing the volatile profile of wine and wine vinegars (Morales et al., 2020; Panighel & Flamini, 2014; Ríos-Reina et al., 2019; Ubeda et al., 2017). It offers important advantages such as the non-use of extraction solvents and its capability of carrying out the extraction and concentration steps simultaneously in a short time. As wine is a complex matrix and its analysis by GC–MS provides a large amount of information, the treatment of the generated data is a difficult task. To deal with it, the combination of GC–MS with chemometric techniques such as PARAFAC2 has demonstrated to reduce the problems associated with GC–MS analysis of complex mixtures and to obtain the maximum information of the volatile profile for distinguishing between samples (Johnsen et al., 2017; Ríos-Reina et al., 2023). Thus, different software and algorithms have been developed that apply chemometric techniques useful to process complex data sets from GC–MS analysis quickly, easily and efficiently, such as the Deconvolution and Identification System called PARADISe® (Johnsen et al., 2017). In this context, the aim of this work was to study and compare the characteristic volatile profile of the different fortified wines from each Spanish PDO by a reliable and simple method such as headspace solid phase microextraction (HS-SPME) in conjunction with gas chromatography-mass spectrometry (GC–MS) and powerful chemometric tools, to differentiate the Andalusian PDO fortified wines, being the first time that all the fortified wine types and PDOs are studied together. For that, a strategy based on different steps was developed: the processing of the profile or volatile fingerprint using chemometric techniques such as Parallel Factor Analysis (PARAFAC2), followed by principal component analysis (PCA) and partial least squaresdiscriminant analysis (PLS-DA), to finally selecting and identifying the marker volatile compounds most related to the each PDO fortified type of wine. 2. Materials and methods 2.1. PDO fortified wines under study A total of 104 samples of fortified wines from the different Andalusian PDOs were collected from the Regulatory Councils of each PDO: ‘Condado de Huelva’, ‘Jerez-X´ er` es-Sherry’, ‘Manzanilla -Sanlúcar de Barrameda’, and ‘Montilla-Moriles’. Within each PDO, the different types of fortified wines produced and marketed were considered: Fino (FI), Manzanilla (MA), Amontillado (AM), Palo Cortado (PC) and Oloroso (OL). Table 1 shows the number of samples of each type of wine belonging to each PDO, and the different types according to ageing process, as well as the codes used in this study. The number of samples for each PDO and type included in the study was limited according to the authentication and certification of the samples by the corresponding Regulatory Councils. 2.2. Reagents and solvents As internal standard (IS) a solution of a commercial standard 4- methyl-2-pentanol in 10 mL methanol was used, both from Merck (Darmstadt, Germany), preserved at −20 ◦C until use. Sodium chloride (NaCl) from Sigma-Aldrich (Madrid, Spain) was also used to increase the salting out in the vial and facilitate the passage of volatile compounds into the headspace of the vial. The linear retention index (LRI) was calculated by injecting a series of straight-chain n-alkanes C10 to C40 (50 mg/L in n-hexane) purchased from Fluka (Madrid, Spain). Moreover, available standards of volatile compounds used for identification were bought from different commercial sources (Merck, Darmstadt, Germany and Sigma-Aldrich, Madrid, Spain). 2.3. Headspace extraction of volatile compounds by solid phase microextraction and analysis by gas chromatography and mass spectrometry (HS-SPME-GC–MS) The basic conditions of sample extraction were based on previous reported validated method for wines (Ubeda et al., 2017), which was slightly adapted to increase the efficiency of the method. The method was as follows: 7.5 mL of wine, 1.5 g of NaCl and 10 µL of internal Table 1 Samples of each type of wine considered in this study according to each PDO. Ageing type Wine type Code DOP Number of samples Oxidative Oloroso OL Jerez 9 Condado de Huelva 8 Montilla-Moriles 8 Biological Fino FI Jerez 9 Condado de Huelva 5 Montilla-Moriles 34 Manzanilla MA Sanlúcar de Barrameda 5 Biological þ Oxidative Amontillado AM Jerez 10 Montilla-Moriles 9 Palo Cortado PC Jerez 5 Montilla-Moriles 2 C. Ubeda et al. Food Research International 173 (2023) 113320 3 standard (IS) were placed into a 20 mL glass vial which was positioned in the thermostatized autosampler tray at 20 ◦C for HS-SPME sampling. Then, samples were incubated for 5 min at 40 ◦C with an agitation speed of 300 rpm. After the incubation, the volatile compounds were extracted from the headspace of the vial by exposing a 1 cm Carboxen/DVB/PDMS 50/30-µm SPME fibre (Supelco, Bellefonte, USA) during 40 min at 45 ◦C, and desorbed using the splitless mode with an injector temperature of 250 ◦C for 180 s. GC–MS analysis was carried out using an 8890 Agilent GC system coupled with an Agilent 5977B quadrupole mass spectrometer (Agilent, Santa Clara, CA, US)). A J&W CPWax-57CB capillary column (50 m × 0.25 mm) and with a film thickness of 0.25 μ m was employed (Agilent, Santa Clara, CA, US), with helium carrier gas at a flow rate of 1 mL/min. The oven temperature program was as follows: 35 ◦C for 1 min and subsequently raised to 160 ◦C at 2.5 ◦C/min (holding for 1 min), and then to 220 ◦C at 5 ◦C/min, being the final GC time of 64 min. The electron ionization mass spectra (29 to 300 amu) were acquired in full scan mode at 70 eV. The samples were analysed in duplicate, and blank runs using empty glass vial were performed after each analysis. All data were recorded using MS ChemStation software (Agilent). 2.4. Data processing, identification of volatile compounds and statistical analysis The analysis of the chromatographic profiles was carried out in several steps, starting with the use of Parallel Factor Analysis (PARAFAC2) applied by the software PARADISe® to integrate the areas of volatile compounds and identified them, followed by classification analysis by partial least squares discriminant analysis (PLS-DA) for feature selection using the variables with importance in the projection (VIPs) obtained. Moreover, as PLS-DA was not used with a classificatory purpose due to the limited number of samples in some of the categories, exploratory analyses using principal component analysis (PCA) were performed with the volatile compounds selected to better explore the grouping and relationship with the volatiles. PARADISe® is a powerful methodology for analysing complex chromatographic data based on PARAFAC2, which offers the advantage of being effective in the simultaneous deconvolution of the pure mass spectra of peaks and co-eluting compounds while robustly calculating their peak areas for all samples, as well as in the identification of these peaks using the resolved mass spectra and the NIST MS library (Johnsen et al., 2017). It is also effective in correcting for baseline and noise. These advantages provide an opportunity to identify minor compounds in complex volatile profiles such as those of fortified wines, as well as it facilitates and speeds up statistical analyses. For using the program, different steps need to be taken. First, the chromatographic data were converted to netCDF format and exported to AIA format by MSD ChemStation (version F.01.01.2317). Then, selection of intervals along the chromatogram is needed. In this study, 176 intervals were selected to run the modelling of each interval using the PARAFAC2 algorithm, setting the maximum number of components to 7 per interval and applying the non-negativity constraint. To select the correct number of components for each model, and to check that the model is correct, two parameters were carefully optimized, the percentage of fit and the percentage of core-consistency, trying to achieve values as close as possible to 100% for each parameter. After modelling, the area value of a total of 346 peaks were obtained and their tentatively identification was provided by the NIST MS Search database (v.2.0). Then, the peak area values provided for each compound were normalized dividing them by the area value of the internal standard (i.e., obtaining the relative areas). In this study quantification of peaks was not considered since the aim was a non-targeted approach for a comparison and differentiation of the samples, leading to a somewhat more targeted approach by selecting relevant peaks. Thus, the following step was to reduce the number of compounds due to within this set of compounds, there were volatile compounds, but there were also possible contaminants, noise, etc., which must be eliminated. Moreover, as the aim of this work was the differentiation of fortified wines and the selection of possible volatile markers, and not their complete characterization per se, already found in the literature, it was decided to carry out a methodology for data reduction and compound selection strategy. It was based on the selection of variables with importance in the projection (VIPs) using Partial Least Squares Discriminant Analysis (PLS-DA) according to the types of fortified wines. Hence, PLS-DA was not developed directly for a classification approach, due to the limited number of samples in some of the cases, but it was applied to study variables with predictive importance (VIP) for data reduction and selection. For that, a sequential PLS-DA classification models were developed by selecting in each one the volatile compounds with high VIP value, due the VIPs are scaled in such a way that all predictors that have a VIP ≥1 are considered relevant (Mehmood et al., 2012). In this study, to greatly reduce the large number of compounds obtained, variables whose VIP was ≥2 (VIPs ≥2) were selected in each of the PLS-DA models. PLS-DA models were validated by means of crossvalidation (venetian blinds). The identification was assessed by using version 2.0 of the standard NIST library and the linear retention index (LRI) of reference standards, when available, or data in the literature. Significant differences between data were obtained by analysis of variance (ANOVA) followed by a post hoc comparison test (Tukey’s test) using INFOSTAT software (FCA, Universidad Nacional de C´ ordoba, Argentina), principal component analysis (PCA) and PLS-DA were conducted using PLS Toolbox 7.9.5 (Eigenvector Research Inc., Wenatchee, WA, USA) working in a MATLAB 2016a environment (Mathworks) and circle packing heatmap was plotted by https://www.bioinformatics. com.cn/en. Prior to any modelling, the data were autoscaled. 3. Results and discussion 3.1. Data reduction and selection of the differential volatile compounds for each type of PDO fortified wine After analysing the total set of samples, the chromatograms of all the samples were processed by the software PARADISe®. As it was described in the materials and method section, a data matrix of 208x346 was provided by PARADISe®, including the area values of the volatile compounds and a tentative identification. After normalizing the area values by the IS, a process of feature or variable reduction using VIP scores was repeated two times, hence obtaining three different PLS-DA models: one with the total deconvoluted peaks (346), called from now on PLS-DA346; another with 101 volatile compounds which were the VIP compounds in the PLS-DA346 model, called from now on PLSDA101; and a third one with 28 volatile compounds defined as VIP in the PLS-DA101 model, called from now on PLS-DA28. The objective of this strategy of sequential PLS-DA models was to evaluate how many variables or features could be removed from the model without reducing the discriminant performance achieved. For that reason, it should be noted that only compounds relevant to the differentiation were selected in the selection of the VIPs. Therefore, other compounds that were present in the samples but did not make a difference between the types of fortified wine were not considered, since, as indicated, the study was not merely a characterization of each type of wine separately, but rather a differentiation of samples. Hence, although the volatile profile of these wines has been studied by other authors (Moyano et al., 2002; Valc´ arcel-Mu˜ noz et al., 2022; Zea et al., 2001), to our knowledge, this is the first time that this strategy, i.e., going from untargeted to targeted approach by means of the use of VIP values and PLS-DA, has been carried out with the volatile composition of all the types of fortified wines together, and with the aim of selecting possible volatile markers with the VIP approach. Moreover, although classification was not the aim of the PLS-DA model, the classification and cross-validation results obtained for all the models showed promising classification results (Table 2), presenting for the first time the C. Ubeda et al. Food Research International 173 (2023) 113320 4 possible usefulness of the proposed methodology to classify fortified wines according to the fortified type. Once the first PLS-DA model (PLS-DA346) was performed with the 346 peaks for the selection of the compounds which VIP values ≥2, whose classification results are shown in Table 2, the number of compounds was reduced from 346 to 101 VIPs. Moreover, by performing the second PLS-DA with the 101 VIPs, it can be also seen that the reduction of volatile compounds from 346 to 101 clearly improved the classification results in most of the cases by increasing the percentage of sensitivity and specificity of both calibration and cross-validation (e.g., from 84.2% of sensitivity in AM to 100%) and decreasing the percentage of error of each wine type (e.g., from 10.4% to 1.47% in calibration of AM) (Table 2). The identification of these 101 selected volatile compounds was checked by their LRI and they were grouped into their corresponding chemical families. The mean and standard deviation of their relative peak area values according to each type of wine, as well as Tukey test results, are shown in Table 3, whose discussion will be done in the following section. In addition, when a PCA was performed with this dataset (208x101), this already showed a grouping of the samples according to the ageing type (Fig. S1). However, with the aim of further reducing and selecting those volatile compounds that could potentially be markers of each type of fortified wine and to move towards a more targeted approach, subsequently, a second PLS-DA analysis was carried out with the 101 compounds selected in the PLS-DA101 model, and once again, the volatile compounds with a VIP ≥2 were selected, being in this case 28. Then, to check that the reduction of compounds did not worsen the differentiation, thus finding the best and simplest classification model with fewer variables, a new PLS-DA model (PLS-DA28) was built with the 28 volatile compounds previously selected. Once again, an improvement in the classification results was obtained by this second reduction of compounds (Table 2), i.e., with only the 28 selected compounds, especially for the AM, MA, OL, and PC categories, which means that despite the important reduction in features, the statistical performance of the resulting model was not affected. The LRI, chemical families and odour description of these 28 volatile compounds is presented in Table S1. This process of variable reduction using VIP scores by repetition of PLS-DA models, and the selection of VIPs as markers, has been successfully applied in other previous works (Rodríguez-Hern´ andez et al., 2023). This approach has been often used for the analysis of data such as metabolome or volatilome (Freire et al., 2021), with the aim of selecting features from multiple potential compounds which could be considered as important markers. 3.2. Study of the differences between each type of fortified wine according to the possible volatile markers selected Among the 28 volatile compounds selected as possible markers of each wine (highlighted in Table 3 and summarized in Table S1), different chemical families were distinguished: esters, alcohols, aldehydes, acids, ketones, and lactones, and a group called others that included acetals, terpenes and norisoprenoids, among others. Some differences in the chemical families could be observed among wine types by studying the percentage of relative areas for each one (Fig. 1) and by studying the circle packing plots (Fig. 2), which show the differences in the volatiles dominating in relative area content between the different wine types. On the one hand, samples that have been submitted to oxidation (AM, PC and OL) showed a higher presence of differentiating compounds of the aldehydes’ family in the proportion of the 28 volatile compounds selected (Fig. 1, Fig. 2). The relevance of aldehydes in the differentiation of wines with oxidative ageing was expected since aldehydes are formed because of alcohol oxidation (Moyano et al., 2002; Waterhouse et al., 2016; Zea et al., 2001). On the other hand, wines with only biological ageing, such as FI and MA, showed a higher percentage of relative areas for ester compounds within these 28 selected volatile compounds (Fig. 1, Fig. 2). Flor yeasts have been shown to increase the content of ethyl esters (Zea et al., 1995) being consistent with our results that are differentiation markers of biological ageing. Moreover, interesting differences could be also observed between the FI and MA samples, mainly with respect to the acid family, being higher in FI samples, and in the alcohol family, in higher amounts in MA samples. Differences among the group of samples with oxidative ageing was also observed, such as the higher relative areas of acids in OL samples than in those samples with the combination of both ageing types, and slightly higher presence of esters in AM than in PC samples (Fig. 1, Fig. 2), highlighting their previous biological ageing. These compositional differences and similitudes can be easily observed by building a principal component analysis (PCA) with the 28 variables selected (Fig. 3). Fig. 3a shows the distribution of the analysed samples (scores) depicted in the function of their ageing type for PC1 and PC2, which accounted for 30.0% and 19.6% of the variance, respectively. As can be seen in Fig. 3a, the scores plot of the first two principal components (PCs) showed a grouping of the samples according to their type of ageing: biological (FI and MA, in green and blue colour) mainly in the negative side of the PC1; oxidative (OL, in light blue colour) in the negative side of PC2; and biological and oxidative (AM and PC) mainly in the positive part of PC1. As can be observed, wines submitted to two different kinds of ageing (oxidative and biological) were arranged in the plane more dispersed than the other types of wines that have only undergone one type of ageing. This fact difficulted the interpretation but despite this, it seems that samples were distributed among the y-axis (PC2). Thus, PC2 distributed samples from the oxidative ageing with high positive values y-axis of OL (in light blue colour) to the samples with mixture ageing type, with PC (pink colour) situated closer to OL and transitioning to AM (red colour), that was in the middle of the plot. At the bottom of the y-axis, with negative PC2 values, fortified wines with biological ageing are placed. Very similar samples such as FI and MA (in green and blue colour respectively) did not achieve a good separation, for that reason, the wines were studied separately to deeply understand the results. Thus, three groups were Table 2 Classification results for calibration (Cal.) and cross-validation (CV) obtained by developing a PLS-DA model with the total of 346 resolved peaks provided by PARADISe (4VL, 48.63% explained variance); a PLS-DA model with 101 VIP volatile compounds (6 LVs and 76.56% explained variance) and a PLS-DA model with 28 VIP volatile compounds (8 LVs and 58.57% explained variance). PLS-DA346 a (346 volatile compounds) PLS-DA101 (101 volatile compounds) PLS-DA28 (28 volatile compounds) Types of fortified wines according to their ageing AM FI MA OL PC AM FI MA OL PC AM FI MA OL PC Cal % Sens. b 84.2 96.8 100 94.0 100 100 96.9 100 96.0 100 94.7 96.9 100 98.0 100 % Spec. 94.1 92.0 84.3 94.9 95.4 97.1 94.6 93.4 93.7 96.4 97.6 94.9 94.9 97.5 97.9 % Error 10.8 5.58 7.82 5.53 2.32 1.47 4.76 3.28 5.16 1.80 3.80 5.13 2.52 2.26 1.03 CV c % Sens. 84.2 95.8 100 94.0 100 94.7 95.8 100 94.0 100 94.7 96.9 100 96.0 100 % Spec. 94.1 92.0 85.9 92.4 93.8 96.5 94.6 92.4 93.7 95.9 97.6 92.9 93.4 96.8 96.9 % Error 10.8 6.10 7.07 6.79 3.09 4.39 4.76 3.78 6.16 2.06 3.80 5.13 3.28 3.58 1.54 Note: a The best classification values for each wine type were highlighted in bold. b Sens: sensitivity; Spec: specificity. c CV: Venetian blinds cross-validation. C. Ubeda et al. Food Research International 173 (2023) 113320 5 Table 3 Means and standard deviations of the relative areas of the 101 selected volatile compounds according to wine type, linear retention index (LRI), identification and chemical family, as well as results of the Tukey statistical analysis. Biological Oxidative Biological þOxidative FI d MA OL AM PC LRI b ID c Chemical family Volatile compounds a mean ±SD e WT f mean ±SD WT AT g mean ±SD WT AT mean ±SD WT mean ±SD WT AT 1274 B acid 3-Cyclohexene-1-carboxylic acid 0.59 0.30 b 0.81 0.15 c b 0.11 0.20 a a 0.16 0.22 a 0.07 0.06 a a 1451 A acid Acetic acid 1.85 2.01 a 0.19 0.20 a a 5.75 2.31 bc b 4.84 2.53 b 7.00 1.31 c b 1485 C acid 2-Thiopheneacetic acid 0.11 0.06 a 0.11 0.03 a a 0.08 0.07 a a 0.20 0.12 b 0.13 0.04 a b 1802 C acid 2-Phenoxypropionic acid 0.41 0.26 a 0.51 0.26 a a 1.06 0.39 b b 1.23 0.58 b 1.83 0.63 c c 1824 C acid 4-Amino-1,5-pentandioic acid 0.02 0.02 a 0.01 0.01 a a 0.11 0.06 b b 0.10 0.08 b 0.10 0.04 b b 1837 C acid Ethanediamide 0.05 0.06 a 0.07 0.05 ab a 0.42 0.35 c b 0.27 0.31 bc 0.86 0.45 d b 1986 A acid Heptanoic acid 0.08 0.04 b 0.03 0.01 a a 0.06 0.03 ab a 0.06 0.03 ab 0.07 0.03 b a 2155 A acid Sorbic acid 0.17 0.19 b 0.00 0.00 a b 0.43 0.27 c c 0.04 0.04 ab 0.03 0.07 ab a 2213 A acid Decanoic acid 1.50 1.45 ab 0.61 1.21 a a 3.75 2.46 c b 1.24 1.06 ab 2.31 1.04 b a 1157 A alcohol 1-Butanol 1.06 0.45 ab 0.76 0.12 a b 0.76 0.38 a a 1.49 0.87 b 1.24 0.46 b c 1311 B alcohol 2-Ethylbutanol 0.22 0.20 b 0.26 0.08 b b 0.06 0.04 a a 0.31 0.26 b 0.17 0.07 ab b 1331 A alcohol 3-Methyl-1-pentanol 0.31 0.12 b 0.23 0.05 ab b 0.17 0.05 a a 0.29 0.17 b 0.24 0.07 ab b 1386 A alcohol (Z)-3-Hexen-1-ol 0.16 0.12 a 0.34 0.09 b a 0.18 0.09 a a 0.17 0.15 a 0.14 0.09 a a 1463 A alcohol 1-Heptanol 0.15 0.07 a 0.22 0.07 b a 0.18 0.09 ab a 0.22 0.12 b 0.20 0.04 ab b 1487 B alcohol 4-Nonanol 0.09 0.08 b 0.07 0.03 ab b 0.03 0.02 a a 0.10 0.11 b 0.06 0.03 ab b 1518 B alcohol (S)-3-Ethyl-4-methylpentanol 0.13 0.18 a 0.42 0.11 b a 0.17 0.14 a a 0.25 0.24 a 0.27 0.15 ab b 1527 B alcohol 2-Nonanol 0.05 0.03 a 0.05 0.02 a a 0.20 0.08 c b 0.15 0.07 b 0.32 0.13 d b 1547 C alcohol E-2-Caren-4-ol 0.04 0.08 a 0.09 0.05 ab a 0.12 0.15 b b 0.05 0.05 ab 0.06 0.05 ab a 1672 B alcohol 1-Nonanol 0.28 0.13 a 0.32 0.14 ab a 0.61 0.39 c c 0.39 0.19 ab 0.49 0.19 bc b 1731 A alcohol Methionol 0.23 0.09 b 0.31 0.12 c b 0.07 0.07 a a 0.07 0.08 a 0.06 0.04 a a 1749 B alcohol 4-Methyl-4-heptanol 0.05 0.05 ab 0.03 0.01 a b 0.02 0.02 a a 0.12 0.11 c 0.08 0.02 bc c 1773 B alcohol 1-Decanol 0.07 0.07 ab 0.03 0.02 a a 0.17 0.12 c b 0.08 0.04 ab 0.12 0.04 bc a 1385 A aldehyde Nonanal 0.21 0.12 a 0.41 0.16 b a 0.40 0.17 b b 0.43 0.26 b 0.43 0.18 b b 1453 A aldehyde Furfural 2.31 2.65 a 2.43 0.90 a a 25.62 9.16 b b 32.15 13.09 b 42.59 13.39 c c 1514 A aldehyde Benzaldehyde 1.56 1.98 a 1.15 0.25 a a 29.02 12.20 b b 29.87 11.66 b 47.65 12.82 c c 1567 A aldehyde 5-Methylfurfural 0.21 0.16 a 0.13 0.05 a a 3.82 1.43 b b 3.34 1.55 b 5.44 2.49 c b 1605 B aldehyde 1-Ethyl-1H-pyrrole-2-carbaldehyde 0.03 0.03 a 0.04 0.01 a a 0.13 0.07 b b 0.18 0.16 b 0.35 0.16 c c 1702 C aldehyde 3-(2,6,6-Trimethyl-1-cyclohexen-1-yl)-2- propenal 0.29 0.19 a 0.33 0.18 a a 0.74 0.31 bc b 0.68 0.37 b 0.96 0.27 c b 1779 B aldehyde Benzenebutanal 0.03 0.04 ab 0.00 0.00 a a 0.07 0.08 bc b 0.11 0.11 c 0.01 0.00 a b 1956 B aldehyde (E)-3-(2,6,6-Trimethyl-1-cyclohexen-1- yl)-2-propenal 0.14 0.09 a 0.10 0.04 a a 0.15 0.05 a a 0.32 0.16 c 0.23 0.08 b b 1058 A ester Ethyl butyrate 3.84 2.09 a 2.95 0.90 a a 2.85 1.67 a a 6.91 4.45 b 4.47 0.85 a b 1070 A ester Ethyl 2-methylbutyrate 2.86 2.11 a 1.92 0.84 a b 1.36 0.92 a a 5.61 4.35 b 3.13 1.55 a c 1081 A ester Ethyl isovalerate 6.86 4.26 ab 4.65 1.90 ab b 3.78 1.95 a a 11.92 8.48 c 7.84 3.06 b c 1191 B ester Isopentyl isobutyrate 0.44 0.31 bc 0.34 0.18 abc b 0.15 0.09 a a 0.54 0.49 c 0.26 0.09 ab b 1227 A ester Ethyl hexanoate 34.62 15.18 a 45.38 17.72 a a 35.70 18.79 a a 43.84 27.70 a 32.92 9.03 a a 1252 A ester Ethyl pyruvate 0.02 0.01 a 0.02 0.00 a a 0.12 0.05 b b 0.14 0.08 b 0.13 0.03 b b 1289 A ester 2-Methylbutyl isovalerate 0.19 0.13 a 0.11 0.09 a a 0.11 0.08 a a 0.39 0.37 b 0.23 0.14 a b 1298 B ester Methyl 2-hydroxy-2-methylbutanoate 0.02 0.02 a 0.03 0.01 a a 0.04 0.03 a b 0.09 0.04 b 0.12 0.07 c c 1325 A ester Ethyl heptanoate 0.25 0.14 a 0.24 0.06 a a 0.35 0.13 a b 0.62 0.27 b 0.81 0.27 c c 1381 A ester Methyl octanoate 0.07 0.05 a 0.10 0.06 b a 0.14 0.10 ab b 0.09 0.05 ab 0.09 0.03 ab ab 1400 B ester Ethyl 2-hydroxybutanoate 0.02 0.03 a 0.01 0.00 a a 0.03 0.02 a a 0.13 0.10 b 0.12 0.09 b b 1413 B ester Ethyl glycolate 0.01 0.01 a 0.01 0.00 a a 0.07 0.03 b b 0.08 0.07 b 0.19 0.11 c c 1421 B ester Ethyl 2-hydroxyisovalerate 0.65 0.68 a 0.46 0.19 a a 0.35 0.25 a a 1.76 1.21 b 1.44 0.49 b b 1433 A ester Ethyl octanoate 55.90 27.74 a 89.29 43.15 b a 104.75 53.29 b b 83.56 48.84 ab 102.50 30.23 b b 1484 B ester Ethyl diethoxyacetate 0.01 0.01 a 0.00 0.00 a a 0.09 0.04 b b 0.11 0.04 b 0.18 0.05 c c 1505 B ester Methyl 3-methoxypropionate 0.10 0.09 a 0.11 0.05 a a 0.51 0.40 b b 0.39 0.43 b 1.12 0.44 c b (continued on next page) C. Ubeda et al. Food Research International 173 (2023) 113320 6 Table 3 (continued) Biological Oxidative Biological þOxidative FI d MA OL AM PC LRI b ID c Chemical family Volatile compounds a mean ±SD e WT f mean ±SD WT AT g mean ±SD WT AT mean ±SD WT mean ±SD WT AT 1533 B ester Ethyl 2-hydroxy-4-methylvalerate 1.36 1.47 a 0.86 0.24 a a 0.67 0.50 a a 3.55 2.76 b 2.74 0.91 b b 1543 B ester Ethyl dl-2-hydroxycaproate 1.77 1.51 ab 2.04 0.39 ab b 0.98 0.39 a a 3.28 2.01 c 2.84 0.93 bc c 1576 C ester Diethyl ethylmalonate 0.04 0.06 a 0.02 0.02 a a 0.03 0.02 a a 0.15 0.16 b 0.11 0.06 b b 1606 B ester Ethyl levulinate 0.03 0.02 a 0.02 0.00 a a 0.29 0.24 b b 0.27 0.29 b 0.20 0.08 b b 1612 B ester Ethyl 2-furoate 0.80 0.57 a 0.91 0.14 a a 1.46 0.55 b b 2.13 0.82 c 2.56 1.03 c c 1631 B ester Ethyl methyl succinate 0.07 0.03 a 0.04 0.01 a a 0.19 0.08 b b 0.20 0.14 b 0.22 0.07 b b 1638 A ester Ethyl decanoate 1.36 0.77 b 0.53 0.09 a a 1.24 0.35 b a 2.38 1.14 c 2.37 0.65 c b 1661 A ester Ethyl benzoate 0.75 0.43 a 0.56 0.16 a a 2.58 0.96 b b 3.38 1.31 b 6.03 2.64 c c 1682 B ester Diethyl succinate 49.15 17.82 a 43.29 8.61 a a 91.93 21.15 b b 107.74 49.88 b 151.34 38.71 c c 1712 C ester Ethyl 1-methyl-4-oxo-2-cyclohexene-1- carboxylate 0.01 0.01 a 0.01 0.00 a a 0.04 0.04 ab b 0.05 0.05 b 0.17 0.13 c c 1755 C ester Butyl isobutyl succinate 0.02 0.02 a 0.01 0.01 a a 0.20 0.09 b b 0.37 0.37 c 0.67 0.38 d c 1759 B ester Butyl ethyl succinate 0.03 0.02 a 0.02 0.01 a a 0.11 0.04 b b 0.15 0.06 c 0.23 0.08 d c 1778 C ester Ethyl 2-formyl-2-methyl-4-pentenoate 0.01 0.01 a 0.01 0.00 a a 0.11 0.10 b b 0.12 0.13 b 0.47 0.28 c c 1782 B ester Diethyl glutarate 0.61 0.32 a 0.45 0.20 a a 1.06 0.34 b b 1.93 0.97 c 2.48 0.64 d c 1858 C ester Diethyl 2-hydroxy-3-methylsuccinate 0.25 0.16 a 0.21 0.17 a a 1.31 0.58 b b 1.65 1.00 b 3.12 1.65 c c 1993 C ester Ethyl 3-hydroxy-3-methylbutanoate 0.07 0.06 a 0.03 0.01 a a 0.28 0.15 b b 0.51 0.33 c 0.87 0.36 d c 2034 B ester Isopropyl myristate 0.18 0.14 b 0.07 0.03 a b 0.37 0.17 c c 0.07 0.04 a 0.10 0.04 ab a 2047 B ester Diethyl malate 0.26 0.18 ab 0.21 0.18 a a 4.54 2.27 c c 1.52 1.44 b 5.96 3.30 d b 2051 B ester Methyl 3-hydroxydecanoate 0.00 0.00 a 0.00 0.00 a a 0.04 0.03 b b 0.03 0.04 b 0.11 0.06 c b 2121 B ester Ethyl pentadecanoate 0.00 0.01 a 0.00 0.00 a a 0.01 0.01 a a 0.03 0.04 b 0.00 0.00 a b 2182 B ester Ethyl 5-oxooxolane-2-carboxylate 0.37 0.18 a 0.46 0.34 a a 1.21 0.44 b b 1.00 0.55 b 2.07 0.87 c b 2238 B ester (+)-Diethyl L-tartrate 0.11 0.11 a 0.12 0.12 a a 0.60 0.46 b b 0.86 0.57 b 1.17 0.63 c c 1350 B ketone 3-Nonanone 0.01 0.01 a 0.01 0.00 a a 0.17 0.21 b c 0.09 0.06 ab 0.13 0.07 b b 1383 B ketone 2-Nonanone 0.16 0.14 a 0.48 0.66 b a 0.70 0.51 b c 0.49 0.24 b 0.65 0.19 b b 1395 C ketone 5,6-Dihydro-4-(2,3-dimethyl-2-buten-1- yl)–2H-pyran-2-one 0.17 0.15 a 0.19 0.15 a a 0.14 0.11 a a 0.43 0.40 b 0.24 0.15 a b 1489 C ketone 2-Hydroxy-1-phenylbutan-1-one 0.11 0.05 ab 0.12 0.02 ab a 0.10 0.04 ab a 0.09 0.04 a 0.12 0.03 b a 1498 B ketone Acetylfuran 0.22 0.13 a 0.15 0.02 a a 0.53 0.22 b b 0.90 0.41 c 1.09 0.50 c c 1528 B ketone 4-Undecanone 0.01 0.01 a 0.01 0.00 a a 0.06 0.03 b b 0.12 0.10 c 0.12 0.05 c c 1618 C ketone 1,1-Diethoxypropan-2-one 0.11 0.10 a 0.12 0.04 a a 0.39 0.25 b b 0.33 0.28 b 0.77 0.29 c b 1646 B ketone 3-Nonen-5-one 0.02 0.04 a 0.02 0.03 a a 0.90 1.20 b c 0.29 0.29 a 0.83 0.38 b b 1647 A ketone Acetophenone 0.08 0.05 a 0.10 0.05 a a 0.45 0.26 b b 0.45 0.42 b 0.74 0.25 c b 1692 B ketone 2-Nonen-4-one 0.01 0.01 a 0.01 0.01 a a 0.11 0.09 b c 0.05 0.04 a 0.11 0.03 b b 1756 C ketone 7-Ethyl-2-methyl-4-undecanone 0.10 0.08 abc 0.09 0.02 ab b 0.04 0.02 a a 0.16 0.12 c 0.12 0.04 bc c 1789 C ketone Benzocyclobuten-1(2H)-one 0.03 0.04 a 0.05 0.03 a a 0.12 0.12 ab b 0.17 0.21 b 0.62 0.33 c c 1629 A lactone γ-Butyrolactone 0.64 0.27 a 0.50 0.08 a a 0.59 0.30 a a 1.10 0.56 b 1.18 0.30 b b 1724 B lactone 4-Ethoxy-γ-butyrolactone 0.01 0.00 a 0.01 0.00 a a 0.12 0.12 b b 0.15 0.16 b 0.26 0.12 c b 1727 B lactone γ-Ethoxybutyrolactone 0.01 0.01 a 0.01 0.01 a a 0.11 0.07 b b 0.09 0.09 b 0.23 0.11 c b 2037 B lactone γ-Nonalactone 0.01 0.01 a 0.01 0.00 a a 0.03 0.03 a b 0.04 0.05 a 0.15 0.10 b c 2062 B lactone Solerone 0.04 0.06 a 0.01 0.00 a a 0.07 0.07 a a 0.23 0.28 b 0.05 0.01 a b 1156 C other 4,5-Dimethyl-2-pentadecyl-1,3-dioxolane 0.36 0.32 a 0.28 0.07 a a 2.39 1.38 b b 4.22 3.07 c 5.59 3.15 c c 1243 C other 1,1-Diethoxy-2-methylpropane 0.06 0.05 a 0.05 0.01 a a 0.47 0.28 b b 0.61 0.35 b 0.87 0.34 c c 1278 C other LRI1278 1.49 0.53 a 1.52 0.31 a a 2.50 0.72 b c 1.55 0.55 a 3.27 0.95 c b 1308 B other 2,2-Diethyl-1,3-dioxolane 0.02 0.02 a 0.01 0.00 a a 0.15 0.07 b b 0.36 0.33 c 0.27 0.07 bc c 1420 C other Valeric anhydride 0.05 0.06 a 0.03 0.01 a a 0.11 0.05 b b 0.18 0.08 c 0.24 0.12 d c 1445 A other E-Linalool oxide (furanoid) 0.19 0.14 ab 0.33 0.05 c b 0.11 0.07 a a 0.27 0.20 bc 0.31 0.16 c c 1467 B other Nerol oxide 0.06 0.08 a 0.22 0.08 b a 0.08 0.07 a a 0.08 0.11 a 0.08 0.07 a a 1522 B other Vitispirane 4.89 3.54 ab 7.71 3.96 b ab 6.14 5.57 ab b 4.05 3.48 a 2.99 2.17 a a 1549 B other Linalool 0.11 0.14 b 0.04 0.03 ab b 0.01 0.02 a a 0.01 0.01 a 0.01 0.01 a a 1582 B other Methyl thymol 0.01 0.03 a 0.14 0.21 b a 0.02 0.02 a a 0.02 0.03 a 0.03 0.04 a a (continued on next page) C. Ubeda et al. Food Research International 173 (2023) 113320 7 organized: 1) wines produced by biological ageing (FI and MA); 2) wines produced by oxidative ageing (OL); 3) wines produced by mixed ageing (AM and PC). Furthermore, looking at the loadings plot (Fig. 3b), it could be seen that one of the reasons of the differentiation of the samples is that those with oxidative ageing, either only oxidative or biological and oxidative, seemed to show a richer volatile profile, i.e., showing a higher relation of most of the volatile compounds than wines produced only biologically, being most of the selected volatile compounds in the positive side of PC1. Moreover, it could be also seen in the loadings plot the higher relation of esters and aldehydes with OL, AM and PC types, and acids with biological samples, being in line with the above discussion. 3.2.1. Differentiation of wines produced by biological ageing: Fino and Manzanilla The most extensive literature on fortified wines belongs to studies on the Fino wine type (Zea et al., 2007, 2010a, 2010b). Our results highlighted, as shown in Fig. 1 and Fig. 2, that the esters group has a higher presence in the fortified wines produced by biological ageing. This is mainly because of the flor yeasts metabolic activity and because they may undergo autolysis, releasing fatty acids among other molecules to the wine and contributing to esters synthesis (Charpentier et al., 2004). However, by looking at the total mean of relative areas and not the percentage of each type, wines with mixed ageing showed higher mean relative area for the selected esters (Fig. S2). This was mainly due to the huge relative area that presented some esters from fixed or non-volatile acids (Table 3). In contrast, several other esters achieved a VIP score above 2 in the PLS-DA model and could be defined as specific markers of biological ageing (highlighted in bold letters in Table 3, and summarized in Table S1), which were mainly derived from volatile acids. This is the case of ethyl 2-methyl butyrate, ethyl isovalerate, and isopropyl myristate, showing higher contents with respect to the oxidative ageing, however, lower amounts than the mixed ageing. This could be explained by the fact that those compounds, mainly formed by biological ageing, are concentrated during the oxidative period of ageing that the mixed fortified wines suffer. These compounds have been described with a fruity odour contribution (Table S1), which is an aroma that has shown to be strongly related to the biological ageing process due to the flor yeast activity (Zea et al., 2007). On the contrary, some of the esters with a VIP ≥2, stood out because of their relatively low amounts in the biological samples compared to the other types of wines. This was the case of ethyl pyruvate, ethyl diethoxyacetate and diethyl malate, which presented relatively higher areas for wines with oxidation ageing (Fig. 2), and which could contribute to the wine with caramellike notes (Table S1). These results agree with those from Zea et al. (2001) who determined diethyl malate and pyruvate in much lower quantities in Fino-type wines compared to Oloroso-type wines (Zea et al., 2001). In fact, in another work carried out also by Zea et al. (2007), evaluating volatile compounds in Fino wines along ageing time, ethyl pyruvate was found with an odour activity value (OAV) <0.1 in all the ageing times analysed (Zea et al., 2007), meaning that the concentration of that compound in Fino wines was relatively low compared to its odour threshold, that is, it might not significantly contribute to the overall aroma or flavor perception of these wines. As was previously discussed, oxidative ageing seems to enhance the esterification of non-volatile acids, giving rise to these esters, in contrast to the majority of esters in biological samples, which are derived from volatile acids. Regarding the group of alcohols, FI and specially MA, presented a relatively highlighted presence of alcohols in comparison to the other wines (Fig. 1, Fig. 2, Table 3). This had been previously assigned due to the flor yeast metabolic activity and autolysis releasing amino acids, precursors of higher alcohols (Zea et al., 1995). Among the alcohols with VIPs ≥2 in the PLS-DA2, it was found that heptanol, (Z)-3-hexen-1-ol, (S)-3-ethyl-4-methylpentanol and methionol were present in MA in significantly higher amounts than in the rest of the wines (Table 3). It Table 3 (continued) Biological Oxidative Biological þOxidative FI d MA OL AM PC LRI b ID c Chemical family Volatile compounds a mean ±SD e WT f mean ±SD WT AT g mean ±SD WT AT mean ±SD WT mean ±SD WT AT 1593 B other Isothymol methyl ether 0.04 0.14 a 1.17 2.41 b a 0.03 0.03 a a 0.08 0.15 a 0.07 0.09 a a 1739 A other TDN (1, 1, 5-Trimethyl-1, 2- dihydronaphthalene) 3.01 2.73 ab 4.81 3.07 ab a 6.19 7.12 b b 2.61 2.36 a 2.37 1.18 a a 1819 B other (E)-1-(2,3,6-trimethylphenyl) buta-1,3- diene (TPB) 0.09 0.10 a 0.08 0.03 a a 0.25 0.19 b b 0.06 0.06 a 0.27 0.07 b a 1835 C other 1,1-Diethoxyacetone 0.03 0.03 a 0.01 0.00 a a 0.25 0.22 c c 0.21 0.30 bc 0.09 0.04 ab b 1936 C other (1-Ethyl-1-propenyl) benzene 0.02 0.02 a 0.02 0.01 a a 0.51 0.23 b b 0.41 0.25 b 0.90 0.30 c b 2246 B other 2,5-Dihydrothiophene 2.38 2.33 ab 2.22 1.35 ab b 0.66 0.61 a a 3.52 3.42 b 2.73 1.22 b c Notes: a Volatile compounds in bold are the 28 volatile compounds with VIP value ≥2 obtained by the PLS-DA101 model. b LRI: linear retention index. c ID: identification: A, LRI agreed with standards, and mass spectrum with NIST mass spectral data base; B, mass spectrum agreed with NIST mass spectral data base and LRI agreed with the literature; C, mass spectrum agreed with mass spectral data base and LRI was not described in literature for polar column or did not match. d FI: Fino, MA: Manzanilla, OL: Oloroso, AM: Amontillado, PC: Palo Cortado. e SD: standard deviation.. f WT and g AT: Tukey test results. Different lower-case letters in different “WT” columns indicate significant differences according to Tukey’s test (p <0.05) among fortified wine types. Different cursive lower-case letters in different “AT” columns indicate significant differences according to Tukey’s test (p <0.05) among ageing types. C. Ubeda et al. Food Research International 173 (2023) 113320 8 should be noted that, although higher alcohols tend to have very high perception thresholds (above 100 mg/L) and are therefore unlikely to contribute directly to wine flavour, these specific alcohols have been shown to have the lowest odour thresholds among alcohols, between 0.5 and 2.5 mg/L (e.g., for methionol and heptanol, respectively) (Zea et al., 2007). Among them, heptanol and (Z)-3-hexen-1-ol have been described with green and fresh odour notes (Table S1), while the methionol is considered an off-flavor with attributes of potato and cauliflower giving Fig. 1. Volatile profile of each type of fortified wine with respect to the mean of relative areas (%) considering the 28 VIP volatile compounds selected as differentiators after the second PLS-DA analysis. FI: Fino; MA: Manzanilla; OL: Oloroso; AM: Amontillado; PC: Palo Cortado. Fig. 2. Circle packing plots of the volatile profiles of each type of wine obtained by including only the 28 VIP volatile compounds selected from the second PLS-DA model. FI: Fino; MA: Manzanilla; OL: Oloroso; AM: Amontillado; PC: Palo Cortado. C. Ubeda et al. Food Research International 173 (2023) 113320 9 vegetal notes to the wine (Zea et al., 2007). Thus, this compound was also found in high relative areas in FI samples with respect to the other types of wine, being significatively higher its presence in biological samples than in the rest (Table 3). This compound has shown to present concentrations in Fino wines above its odour threshold (Moyano et al., 2002), appears after 2.5 years of ageing, and continuously increases in a great extent during this stage of the production process (Zea et al., 2007). Our results revealed this compound as the strongest marker of biological ageing (Table 3) since the rest of the wines, in which oxidation has taken place, it was present in significantly lower amounts. The high production of methionol in biologically aged wines is expected since the flor yeasts have the difficult task of prevailing in a medium poor in available nitrogen, this results in the yeasts having to consume amino acids present in the wine to survive, methionine in this case, giving rise to methionol through the Ehrlich pathway (Panighel & Flamini, 2014). Likewise, the low presence of methionol in oxidized wines could be due to the oxidation to methional (Waterhouse et al., 2016), or to the esterification of acetates. It is also very dependent on the yeast strain employed to produce Fino wines (Morales et al., 2020). Similar results had been shown in the study of Moyano et al. (2010) after analysing the volatile profile during the sequential biological and oxidative ageing typical of the Amontillado wine, showing that after the fortification and beginning of the oxidative stage, methionol was not detected (Moyano et al., 2010). Conversely to these logical results, Zea et al. (2001), reported that Oloroso and Amontillado wines had higher significant quantities of methionol than Fino wines (Zea et al., 2001). Regarding the other chemical families, although the group of acids and the group of others showed a high percentage in the volatile profile of biological samples (Fig. 2), the volatile compounds included in those two groups were more relevant in the differentiation of FI and MA, as will be discussed later. Thus, considering that the elaboration process of Fino and Manzanilla is basically the same with the origin of the grape and the location of the cellar as the only variables, Jerez de la Frontera and Sanlúcar de Barrameda, respectively, located both in C´ adiz (Andalusia), there has been an increased interest to define chemical differences among them. Surprisingly, as far as we know, no studies have been carried out to deepen this aspect. For this purpose, a PCA was performed with the 28 selected volatile compounds including only wines from these two types (Fig. 4a and b). It was found that plotting of PC2 vs PC4 in the scores plot, it could be seen a trend of grouping of the samples into both types of wines (Fig. 4a). Manzanillas were in the upper left quadrant defined by (Z)-3-hexen-ol, methyl octanoate, ethyl decanoate, methyl thymol, isothymol methyl ether, among others. These last two compounds caught our attention among the others. Methyl thymol and isothymol methyl ether have been described in the essential oil of plants such as Thyme (Asllani & Toska, 2003; Jan et al., 2020), as well as in some wines (Ruiz-Bejarano, et al., 2016). Their aroma description is herbal/green (Table S1) and are structurally related to the monoterpene thymol and this last, in turn, with p-cymene and carvacrol, which have been found in grapes and wines (Perestrelo et al., 2014; Ruiz-Bejarano et al., 2013; Zea et al., 2007). They have been also described as volatile compounds that can suffer a migration from the cork to the wine (Díaz-Maroto, et al., 2023). These compounds would be among the more useful to differentiate among these two types of fortified wines (Fig. 2), showing significant differences in their relative values (Table 3). Moreover, two esters resulted suitable for this differentiation since the significantly higher presence of ethyl decanoate in FI and methyl octanoate, again with a green odour character (Table S1), in MA were notably useful (Table 3). In addition, two acids also showed significant differences between FI and MA, heptanoic acid and sorbic acid, with relative areas significantly higher in FI wines. 3.2.2. Differentiation of wines produced by oxidative ageing: Oloroso The analysis of wines produced exclusively with oxidative ageing presented significant differences with respect to the wines biologically aged. Among the 28 VIPs selected, decanoic and sorbic acids showed significatively higher relative values in OL samples than in the other types (Table 3). It should be highlighted that OL samples are usually aged during longer periods of time than biological samples, whose ageing time is shorter due to the early disappearance of the ‘velo de flor’, and this could explain the higher presence of some compounds in this type of samples. Thus, as expected, some aldehydes also increased with ageing, mainly due to the oxidation of alcohols, among other factors such as the degradation of the wood lignin that takes place during the ageing, as well as the concentration phenomena during ageing (García- Moreno et al., 2021). The higher presence of aldehydes in OL samples represented one of the main differences with respect to biological ageing Fig. 3. Scores (a) and loadings (b) plots of the PCA model performed with the total of samples and the 28 VIP volatile compounds obtained from the second PLS-DA model. Samples coloured according to their type of ageing. AM: Amontillado (red); FI: Fino (green); MA: Manzanilla (dark blue); OL: Oloroso (clear blue); PC: Palo Cortado (pink). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) C. Ubeda et al.