Lactic Acid Bacteria and Yeast Inocula Modulate the Volatile Profile of Spanish-Style Green Table Olive Fermentations

foods Article Lactic Acid Bacteria and Yeast Inocula Modulate the Volatile Profile of Spanish-Style Green Table Olive Fermentations Antonio Benítez-Cabello 1, Francisco Rodríguez-Gómez 1, M. Lourdes Morales 2,* , Antonio Garrido-Fernández 1, Rufino Jiménez-Díaz 1and Francisco NoéArroyo-López 1 1Departamento de Biotecnología de Alimentos, Instituto de la Grasa (CSIC), Ctra. Utrera km 1, Edificio 46, Campus Universitario Pablo de Olavide, 41013 Sevilla, Spain 2Área de Nutrición y Bromatología, Dpto. Nutrición y Bromatología, Toxicología y Medicina Legal. Facultad de Farmacia, Universidad de Sevilla, C/P. García González, nº2, 41012 Sevilla, Spain *Correspondence: [email protected] Received: 4 July 2019; Accepted: 22 July 2019; Published: 24 July 2019   Abstract: In this work, Manzanilla Spanish-style green table olive fermentations were inoculated with Lactobacillus pentosus LPG1, Lactobacillus pentosus Lp13, Lactobacillus plantarum Lpl15, the yeast Wickerhanomyces anomalus Y12 and a mixed culture of all them. After fermentation (65 days), their volatile profiles in brines were determined by gas chromatography-mass spectrometry analysis. A total of 131 volatile compounds were found, but only 71 showed statistical differences between at least, two fermentation processes. The major chemical groups were alcohols (32), ketones (14), aldehydes (nine), and volatile phenols (nine). Results showed that inoculation with Lactobacillus strains, especially L. pentosus Lp13, reduced the formation of volatile compounds. On the contrary, inoculation with W. anomalus Y12 increased their concentrations with respect to the spontaneous process, mainly of 1-butanol, 2-phenylethyl acetate, ethanol, and 2-methyl-1-butanol. Furthermore, biplot and biclustering analyses segregated fermentations inoculated with Lp13 and Y12 from the rest of the processes. The use of sequential lactic acid bacteria and yeasts inocula, or their mixture, in Spanish-style green table olive fermentation could be advisable practice for producing differentiated and high-quality products with improved aromatic profile. Keywords: table olives; starter cultures; GC-MC analysis; volatile composition 1. Introduction Table olives are a fermented vegetable with a pronounced influence on the Mediterranean diet and culture. Nowadays, worldwide production exceeds 2.5 million tons/year [ 1 ]. Due to the presence of oleuropein, the fresh fruits are strongly bitter and, therefore, should be appropriately conditioned before consumption. The most common processing styles are: (a) alkali-treated green olives (Spanish-style); (b) ripe olives, obtained by oxidation in an alkaline medium (Californian style); and (c) directly brined olives (Greek style) [2]. Lactic acid bacteria (LAB) are the main beneficial microorganisms found in the fermentation of Spanish-style green table olives [ 3 ], but yeasts are also always present and provide exciting technological and probiotic features [ 4 ]. Together, they form stable biofilms on the olive surface [ 5 – 8 ] which are ingested by consumers. As a result, the interest in multifunctional starters with adequate technological properties and probiotic potential, as well as the efforts for finding synergy between these two groups of microorganism, has strongly increased. LAB stabilise the product through the production of lactic acid, which lows the pH, whereas the production of enzymes such as lipase and esterase may contribute to the biological hydrolysis of bitter compounds [ 7 , 9 ]. Simultaneously, yeasts improve organoleptic Foods 2019,8, 280; doi:10.3390/foods8080280 www.mdpi.com/journal/foods Foods 2019,8, 280 2 of 17 quality [ 4 ]. Then, their coexistence provides the table olives with an attractive sensory appeal in which the presence of volatile organic compounds (VOCs) plays an unquestionable role. Such substances are produced, in both the fruit matrix and brine, by the action of endogenous enzymes, such as lipoxygenases, or exogenous, released by the microorganisms. There are several studies related to the determination of VOCs in table olives. Sabatini and Marsilio [ 10 ] studied the volatile profile in Spanish-style, Greek-style and Castelvetrano-style green olives of the Nocellara del Belice cultivar. The twenty-two VOCs formed during this olive fermentation were significantly affected by the type and time of processing. L ó pez-L ó pez et al. [ 11 ] studied the sensory profile and volatile composition of 24 samples of Spanish-style green table olives, identifying a total of 133 VOCs and finding a trend to separate samples according to sampling time whereas the segregation by olive cultivar was poor. However, none of these studies associated the presence of VOCs with the use of inoculum. Recently, De Angelis et al. [ 12 ] identified 47 different VOCs during fermentation of directly brined Bella di Cerignola table olives, reporting differences between uninoculated and inoculated treatments with Lactobacilli and Wickerhanomyces anomalus. Tufariello et al. [ 13 ] studied the influence of the type of inoculation on VOCs production in directly black table olives belonging to two Italian (Leccino and Cellina di Nard ò ) and two Greek (Conservolea and Kalam à ta) cultivars, using sequential inoculation of native yeasts and selected LAB starter. De Castro S á nchez et al. [ 14 ] reported the influence of the inoculation with Lactobacillus on the volatile composition of Manzanilla green olives. Among a considerable number of new VOCs, a remarkable amount of 4-ethylphenol was detected in inoculated olives compared to the uninoculated processes. The same group related the formation of some VOCs with the presence of Propionibacterium and Clostridium genera in spoilt Spanish-style green table olives [ 15 ]. Pino et al. [ 16 ] determined the influence of the inoculation with Lactobacillus plantarum and Lactobacillus paracasei on the VOCs composition of directly brined Sicilian table olives, finding differences between spontaneous and inoculated processes. All these studies show that the addition of starter cultures could have a marked influence on the VOC composition of fermented olives. The working hypothesis of this study was to support that the VOCs profile of olive fermentation may be modulated by the addition of starter culture. For this purpose, gas chromatography-mass spectrometry (GC-MS) analysis was used for the analysis of VOCs and diverse multivariate statistical techniques were applied for studying the results. 2. Materials and Methods 2.1. Olive Fermentations Fermentations were carried out in the 2017/2018 season using olives from Manzanilla variety, processed according to the Spanish-style in cylindrical fermentation vessels (9.5 kg olives/5 L liquid). To hydrolyse the oleuropein, fruits were treated with a solution containing 32.4 g/L NaOH, 21.9 g/L NaCl and 0.89 g/L CaCl 2 (97% purity), for 7 h, until NaOH penetrated 2/3 pulp. To remove the excess of alkali, fruits were washed in tap water for 5 h. Then, olives were placed in fermentation brines containing 120 g/L (w/v) NaCl, 1.3 g/L CaCl 2 and 0.012 L de HCl. After performing all these operations in the industry, the fermentation vessels were transported to the pilot plant of the Instituto de la Grasa (CSIC, Sevilla, Spain) for their inoculation, fermentation and analysis. 2.2. Experimental Design Two strain of L. pentosus (LPG1, Lp13), one of L. plantarum (Lpl15) and the yeast strain W. anomalus Y12, all of them previously isolated from the biofilm of table olives, were used for single and co-inoculation experiments. Their selection was based on their technological and probiotic properties determined in previous studies [ 7 ]. The experimental design consisted of four individual inoculations of each organism (T1, for LPG1; T2, for Lp13; T3, for Lpl15; T4, for Y12), a combination of all them Foods 2019,8, 280 3 of 17 (T5, for Y12+LPG1+Lp13+Lpl15), and a spontaneous process (T6). All experiments were performed in duplicate. Previously to the inoculation, LAB strains were grown at 37 ◦ C overnight on Man Rogosa and Sharpe (MRS) broth medium (Oxoid, Basingstoke, Hampshire, England) whereas the yeast was grown on YM broth (Difco, Le Pont de Claix, France) at 28 ◦ C during 48 h. To favour the acclimation of inoculum, culture media were supplemented with 4% NaCl. Previous to inoculation, to remove the medium, cultures were washed and re-suspended in 0.9% sterile saline buffer. Inoculation was executed 1 day after brining for yeasts and at the 9th day for LAB to reach 5 log 10 CFU/mL and 6 log 10 CFU/mL in the cover brine, respectively. In the mixed treatment (T5), yeast and LAB were inoculated sequentially after the same periods and population levels, but using 1/3 for each LAB strain in the case of LAB. 2.3. Control Points of Fermentations At the moment of LAB inoculation (9 days), and at the end of fermentation (65 days), brine were analysed to determine their main physicochemical parameters (pH, salt, free and combined acidity). LAB, yeast and Enterobacteriaceae populations were also determined in brine at the end of fermentation and, in the case of LAB, also before inoculation. These parameters were determined according to procedures described in Ben í tez-Cabello et al. [ 7 ]. Rep-PCR with GTG 5 primer and clustering analysis were used to determine LAB inoculum imposition at 19 days of fermentation when the LAB populations were at the highest level. For this purpose, 10 colonies from each treatment were randomly picked from the highest dilution and their fingerprinting compared with LPG1, Lp13, and Lpl15 profiles according to the protocol described in Ben í tez-Cabello et al. [ 7 ]. Each fermentation vessel was individually analysed. 2.4. Olive Brines’ Sequential Extraction and GC-MS Analysis At the end of fermentation, 100 mL of brines were removed from each treatment and stored at 4 ◦ C until further analysis. The brines’ volatile fraction was submitted to a sequential sorptive extraction with Twisters ® (Gerstel, Müllheim an der Ruhr, Germany). The sequential extraction procedure was performed using two polydimethylsiloxane Twisters ® in each sample, i.e., first in immersion (SBSE) and then in the headspace (HSSE) [ 17 ]. Six mL of the olive brine was placed in a 20 mL vial, and 1.8 gr of NaCl (30%) plus 8 µ L of the internal standard 4-methyl-2-pentanol were added (1,044 mg/L final concentration). A special device called Twicester ® was used. This device enables to position the Twister magnetically on the wall of a sample vial and, in this way, to keep it immersed and prevent it from brushing against the vial wall. Extraction by immersion was performed for 1 h, and the sample was stirred with a conventional magnetic stir bar (non-coated stir bar) at 200 rpm at room temperature during the extraction process. The headspace extraction was performed by placing a new Twister ® in an open glass insert inside the vial and heating the sample in a water bath at 62 ◦ C for 1 hour. In both cases, after extraction, the Twister ® was removed with tweezers, rinsed with Milli-Q water, and dried with a lint-free tissue paper. Both Twisters ® were then introduced into the same desorption tube and thermally simultaneously desorbed in a gas chromatograph/mass spectrometer (GC-MS). Analyses were conducted using an Agilent 6890 GC system coupled up to an Agilent 5975 inert quadrupole mass spectrometer (Agilent, Santa Clara, CA, US) equipped with a Gerstel Thermo Desorption System (TDS2) and a Cooling Injector System CIS-4 PTV inlet (Gerstel, Müllheim an der Ruhr, Germany). The desorption temperature program was the following: the temperature was held at 35 ◦ C for 0.1 min, was ramped at 60 ◦ C/min to 250 ◦ C and held for 5 min. The temperature of the CIS-4 PTV injector, with a Tenax TA inlet liner, was held at − 35 ◦ C using liquid nitrogen for the total desorption time and was then raised at 10 ◦ C/s to 260 ◦ C and held for 4 min. The solvent vent mode was used to transfer the sample to the analytical column. A J&W CPWax-57CB column with dimensions 50 m × 0.25 mm and a 0.20 µ m film thickness (Agilent, Santa Clara, CA, US) was used, and the carrier gas was He at a flow rate of 1 mL/min. The oven temperature program was the following: the temperature was 35 ◦ C for 4 min and was then raised to 220 ◦ C at 2.5 ◦ C/min Foods 2019,8, 280 4 of 17 (held 15 min). The quadrupole, source and transfer line temperatures were maintained at 150 ◦ C, 230 ◦ C and 280 ◦ C, respectively. The electron ionization mass spectra in the full-scan mode were recorded at 70 eV with the electron energy in the range of 29 to 300 amu. Compound identification was based on mass spectra matching using the standard NIST 98 library and the linear retention index (LRI) of authentic reference standards. LRIs were calculated by injecting an n-alkanes mixture (C 10 –C 40 ) under identical conditions as the samples. We considered identified compound the one which mass spectrum and LRI value matched with those of standards, tentatively identified (TI) when mass spectrum matched with those from NIST mass spectral library and LRI value with literature LRI, compound with identification not confirmed when only the mass spectrum of compound matched with those from NIST library, as unknown compounds we include the compounds which mass spectrum reached a low value of probability of right identification in library search report. 2.5. Statistical Analyses The values of relative peak area of the diverse VOCs found in the treatments were first subjected to analysis of variance (ANOVA) according to treatments. Only those who showed a significant difference between at least two fermentations conditions (Fisher’s LSD post-hoc test) were used later for studying the influence of inoculation. The contribution of the diverse inocula to the selected VOCs was also modelled by ANOVA, using treatments as explicative factors and VOCs as dependent variables (tolerance =0.0001 and confidence interval for p=0.05), with an =0 constrain (that is, considering T6, the spontaneous process, as a standard or control). The treatments’ contributions were assessed by the corresponding standardised coefficients of the explicative factors. When the contribution was positive, it was estimated that the treatment significantly contributed to the formation of the corresponding VOCs over the levels reached in the spontaneous. On the contrary, treatments with negative coefficients mean that they decreased the presence of the volatile below the level in the spontaneous process (T6). The relationship between main microbial population or final physicochemical characteristics with the volatile profile was achieved by PLS-R, using a fast algorithm, automatic stop conditions, Jackknife (LOO) validation, as well as centred and reduction of variables. For the study, the final microbial population (LAB and yeast) and the physicochemical characteristics (pH, titratable and combined acidity) were used as independent and the VOCs as dependent variables. To notice that the physicochemical characteristics corresponded to the final conditions when the samples for the volatile profiles were taken, but they did not represent necessarily those in which the compounds were formed, although both may, in some way, summarize the overall fermentation process. The relationships between independent and dependent variables were measured by the respective standardized coefficients of the first for each VOC (the independent variable). Positive (negative) coefficients mean that the independent and dependent variables changed in the same (oppose) direction. Besides, the relationships between treatments and volatile profile were also analysed by biplots and bicluster graphs. Biplots are an exciting tool to study simultaneously the relationship between cases and variables since they can represent both (scores and loadings) in the same plot. Both covariance (more appropriate to study the relationships among variables) and form (more useful for segregating cases) biplots were studied. Also, bicluster was suitable for simultaneously clustering observations and VOCs, therefore providing a map of their relationship. The statistical analysis was achieved using XLSTAT v2018 (Addinsoft, Paris, France), for ANOVA and PLS analysis, and R package Multbiplot v 2018 [18], for biplot, clustering, and biclustering. Foods 2019,8, 280 5 of 17 3. Results 3.1. Fermentation Process Table 1shows the main physicochemical characteristics of the brines at the moment of the LAB inoculation (9 days after olive brining). LAB and yeast strains were inoculated in a pH ranging from 6.19 to 6.33, titratable acidity of 0.09–0.14%, combined acidity of 0.12–0.15 Eq/L, and a salt concentration of 6.61–6.77%. At the 19th day of fermentation, LAB inoculum imposition was determined by molecular methods, finding that the frequency of isolation of the strains LPG1, Lp13 and Lpl15 in their inoculated treatments (T1, T2, and T3, respectively) were 100%. However, Lp13 also was detected (100% frequency) in the rest of the treatments (T4, T5, and T6), showing that this strain has a high ability for brine colonization. All treatments developed safe final pH values (<4.5). However, T4, inoculated only with the yeast Y12, had a particular performance since its pH was higher than the values observed for the rest of the inocula, although the final difference was only significant regarding T1. However, its titratable acidity was significantly lower than the other treatments. On the contrary, treatment inoculated with LPG1 was the most technologically efficient, reaching the significantly lowest pH value (see Table 1). NaCl concentration was similar in all treatments (7.47 ±0.21% average). Regarding LAB growth, they were not detected in any treatment before inoculation. At the VOCs sampling time, they have reached similar populations in all inoculated treatments, although their average value (6.94 log 10 CFU/ml) in the spontaneous fermentation was the highest at the end of fermentation (65 days). On the contrary, no significant differences between treatments were found in the yeast population at the end of the process (5.99 ± 0.10 log 10 CFU/mL average value). Enterobacteriaceae were never detected during the process. 3.2. ANOVA Analysis A total of 131 VOCs were determined in the brines from the 12 fermentation trials (6 treatments in duplicate). Results were expressed as relative area values respect to the internal standard (see Table S1 in Supplementary Materials). The chemical group with the highest number of compounds was alcohols (32) followed by ketones (14), aldehydes (9) and volatile phenols (9). A first ANOVA screening of the VOCs according to treatments (Table S1) showed that the levels of only 71 compounds were significantly different between at least two fermentation processes. Therefore, 60 VOCs were produced regardless of the process and represent a common profile which, at least in the current fermentation conditions, could always be found and included both identified and not assigned formula components. Because in this study the interest was focused exclusively on those which presence could be attributed to the inocula, the compounds not significantly different among treatments were not considered for further analysis. To investigate the relationships between the inoculated starter cultures and the initially significant VOCs, the ANOVA was again repeated with only these response variables, but using T6, the spontaneous process, as reference. As a result, the contribution of each treatment to the concentration of each volatile (versus the spontaneous) was evaluated through the standard coefficient of the respective models. Due to the large number of VOCs remaining in the study, only a few examples of treatments contributions to the formation of volatile will be illustrated graphically (Figure 1) while the rest are summarised (including only significant coefficients) in Table 2. Foods 2019,8, 280 6 of 17 Table 1. Microbiological and physicochemical analysis of the treatments at the moment of LAB inoculation (I, 9 days) and at the end of fermentation (F, 65 days). Inoculum imposition was determined by rep-PCR with GTG5primer at 19 days of fermentation. Treatment Inoculum Strain Inoculum Imposition TA-I * (%) TA-F (%) pH-I pH-F CA-I ** (Eq/L) CA-F (Eq/L) Salt-I (%) Salt-F (%) LAB-F (Log10 cfu/mL) Yeasts-F (Log10 cfu/mL) Enterobacteriaceae-F (Log10 cfu/mL) T1 LPG1 100%-LPG1 0.14 (±0.06) a 0.77 (±0.03) a 6.29 (±0.06) a 4.06 (±0.11) a 0.15 (±0.00) a 0.17 (±0.02) a 6.67 (±0.02) a 7.60 (±0.25) a5.97 (±0.02) a5.95 (±0.08) aNd T2 Lp13 100%-Lp13 0.09 (±0.01) a 0.79 (±0.00) a 6.31 (±0.11) a 4.18 (±0.00) a,b 0.15 (±0.00) a 0.15 (0.00) a 6.44 (±0.37) a 7.30 (±0.09) a5.98 (±0.25) a5.93 (±0.10) aNd T3 Lpl15 100%-Lpl15 0.18 (±0.11) a 0.78 (±0.04) a 6.19 (±0.16) a 4.21 (±0.02) a,b 0.15 (±0.01) a 0.16 (±0.01) a 6.61 (±0.02) a 7.21 (±0.03) a6.22 (±0.06) a6.11 (±0.04) aNd T4 Y12 100%-Lp13 0.10 (±0.00) a 0.58 ( ± 0.05) b 6.29 (±0.00) a 4.38 (±0.09) b 0.15 (±0.00) a 0.14 (±0.00) a 6.62 (±0.06) a 7.80 (±0.10) a6.23 (±0.01) a6.01 (±0.17) aNd T5 Y12+Lp13+ LPG1+Lpl15 100%-Lp13 0.11 (±0.01) a 0.76 (±0.03) a 6.33 ( ± 0.02) b 4.13 (±0.01) a,b 0.15 (±0.01) a 0.15 (±0.01) a 6.77 (±0.09) a 7.50 (±0.17) a6.15 (±0.08) a6.04 (±0.06) aNd T6 Control 100%-Lp13 0.12 (±0.06) a 0.81 (±0.01) a 6.22 (±0.05) a 4.18 (±0.06) a,b 0.12 (±0.06) a 0.15 (±0.00) a 6.63 (±0.15) a 7.43 (±0.31) a6.94 (±0.07) b5.92 (±0.04) aNd * TA. Titratable Acidity, ** CA. Combined Acidity, Nd. Not detected. LAB were absent before inoculation. Different superscript letter, within the same column, are significantly different (p≤0.05) according to post-hoc comparison test. Table 2. Contribution of treatments (inoculation with different LAB and yeast species) to the production of the different VOCs found in brine at the end of the fermentation as assessed by their standardized effects. Only 71 significant compounds (from a total of 131) with differences between at least two treatments were used for these analyses. Contribution of treatments for the different VOCs was compared with respect to the spontaneous fermentation (T6). T1 stands for treatment inoculated with LPG1, T2 with Lp13, T3 with Lpl15, T4 with Y12, and T5 with Y12+LPG1+Lp13+Lpl15. Contributor (ANOVA) and Sign (Standaridised Coefficient) PLS-R Analysis. Significant Coefficients Compound Code T1 T2 T3 T4 T5 Pooled SD TA * pH CA ** Methyl acetate A - - - 0.710 0.703 0.155 −0.351 (±0.092) - −0.171 (±0.046) Ethyl acetate B - - - 0.957 0.470 0.053 −0.464 (±0.199) 0.345 (±0.114) −0.226 (±0.104) cis-3-Hexenyl acetate C - −0.542 - - - 0.226 −0.187 (±0.061) - −0.091 (±0.028) 2-Phenylethyl acetate D 0.596 - - 0.970 0.625 0.024 - - - 3-Methylbutanoic acid E - - - - 0.725 0.185 - - - Methanol F−0.341 - 0.771 −0.242 - 0.112 - - - Ethanol G - - - 0.750 - 0.147 −0.459 (±0.144) 0.341 (±0.135) −0.223 (±0.072) 2-Butanol H 0.688 - - 0.377 0.764 0.132 - - - 2-Methyl-1-propanol I - −0.819 - - −0.819 0.152 - - - 1-Butanol J 0.501 0.413 0.351 1.208 0.753 0.048 - 0.308 (±0.103) - 2-Methyl-1-butanol K - −0.346 - 0.798 - 0.090 - 0.348 (±0.118) −0.227 (±107) 3-Methyl-1-butanol L - −0.430 - 0.738 - 0.101 −0.449 (±0.179) 0.334 (±0.121) −0.219 (±0.088) 3-Methyl-3-buten-1-ol M 0.684 −0.200 - 0.587 0.650 0.079 - - - Foods 2019,8, 280 7 of 17 Table 2. Cont. Contributor (ANOVA) and Sign (Standaridised Coefficient) PLS-R Analysis. Significant Coefficients 1-Pentanol N - - - 0.816 0.436 0.123 −0.351 (±0.159) - −0.171 (±0.069) cis-2-Penten-1-ol O0.774 - - 0.760 0.765 0.046 - - - 2-Methyl-2-buten-1-ol P 0.613 - - 0.631 0.502 0.160 - - - 1-Hexanol Q 0.399 - - 0.637 0.539 0.146 - - - cis-3-Hexen-1-ol R0.564 −0.301 - 0.464 0.604 0.108 - - - 2-Methyl-3-hexanol S 0.611 - - 0.881 0.389 0.142 - - - 1-Heptanol T - −0.428 - 0.744 - 0.117 −0.394 (±0.160) +0.293 (±0.126) −0.192 (±0.071) 6-Hepten-1-ol U - - - 0.749 - 0.165 −0.421 (±0.164) +0.313 (±0.133) −0.205 (±0.081) cis-5-Octen-1-ol V-−0.385 - 0.615 - 0.178 −0.376 (±0.132) +0.280 (±0.120) −0.183 (±0.062) Benzyl alcohol W 0.283 −0.282 - 0.788 0.433 0.090 - +0.237 (±0.099) - 2-Phenylethanol X 0.558 −0.339 −0.122 0.665 0.466 0.038 - - - 2-Ethenyl-2-butenal Y - - 0.370 −0.412 −0.412 0.147 - −0.221 (±0.064) - Isoxylaldehyde Z - - 0.670 - - 0.212 - - - Dimethyl Sulfoxide AA - - - - - 0.212 - - - β-Damascenone AB −0.667 - - - - 0.201 −0.348 (±0.147) - −0.169 (±0.068) Ethyl lactate AC - - - 0.470 - 0.191 - - −0.111 (±0.045) Ethyl 5,6-dimethylnicotinate AD - - 0.129 0.981 - 0.045 - +0.369 (±0.122) - Unknown ester (m/z 88) AE −0.732 - - −0.767 - 0.178 - - - Furfuryl methyl ether AF - −0.349 - - −1.031 0.103 - - - Acetoin AG −0.358 - −0.328 −0.319 0.567 0.143 - - - 6-Methyl-3,5-heptadien-2-one AH - - 0.363 0.495 - 0.183 −0.344 (±0.120) +0.256 (±0.084) −0.167 (±0.066) Purpurocatecho AI -0.151 −0.251 - −0.887 −0.887 0.065 - -0.259 (±0.077) - Iridomyrmecine AJ - −0.496 - 0.324 −0.688 0.118 - - - Methyl lactate AK - - - −0.864 - 0.132 - -0.328 (±0.108) - Methyl hydrocinnamate AL - - - −0.673 - 0.217 - - - Methyl 4(methylamino)benzoate AM - - - 0.786 - 0.136 −0.373 (±0.142) +0.278 (±0.108) −0.182 (±0.076) 3-Ethylpyridine AN −0.457 - - - −0.526 0.182 - - - 4-Methylguaiacol AO 0.347 −0.483 - 0.662 - 0.100 - - - 4-Ethylguiacol AP 0.705 −0.263 −0.196 0.608 0.216 0.067 - - - 4-Ethylphenol AP - - 0.632 −0.407 - 0.149 - - - Isovanillic acid AR −0.239 −0.551 - 0.260 −0.739 0.111 - +0.184 (±0.081) - Coumaran AS −1.093 −0.898 −0.571 −7.460 −0.903 0.109 - - - 5-tert-Butylpyrogallol AT - −0.610 −0.455 0.781 −0.371 0.097 - - - Methoxyeugenol AU −0.247 −0.430 - 0.366 −0.723 0.105 - +0.220 (±0.092) - Vainillin AV - - −0.692 −0.799 −0.529 0.187 +0.320 (±0.075) - +0.156 (±0.036) Foods 2019,8, 280 8 of 17 Table 2. Cont. Contributor (ANOVA) and Sign (Standaridised Coefficient) PLS-R Analysis. Significant Coefficients α-Isophorone AW −0.481 −0.584 - - −0.513 0.200 - - - α-Terpineol AX - - - - 0.572 0.198 - - - Geraniol AY 0.712 −0.297 - - 0.546 0.127 - - - Unknown A (m/z 71-59) AZ - - - - −0.627 0.187 - - - Unknown B (m/z 123-138-96) BA - −0.402 - 0.590 - 0.167 - +0.279 (±0.107) - Unknown C (m/z 83-112-97) BB 0.270 - - 0.974 - 0.098 - +0.300 (±0.137) - Unknown D (m/z 55-93-108) BC 0.151 - - 1.005 0.506 0.037 - +0.329 (±0.116) - Unknown E (m/z 111-198) BD - −0.326 - 0.808 - 0.125 - +0.309 (±0.115) - Unknown F (m/z 95-154-110) BE 0.261 - - 0.940 0.665 0.085 −0.376 (±0.159) −0.183 (±0.072) - Unknown G (m/z 138) BF −0.463 −0.664 - - −0.716 0.153 -0.250 (±0.106) −0.121 (±0.051) - Unknown H (m/z 113-81-153) BG 0.548 −0.457 - 0.518 - 0.127 - - - Unknown I (m/z 99-139-67-81) BH - −0.413 - −0.807 - 0.179 - −0.296 (±0.129) - Unknown K (m/z 93-79) BI - −0.414 0.289 - −0.754 0.123 - - - Unknown L (m/z 222-43-85-177) BJ −0.822 −0.612 - −0.576 −0.776 0.218 - - - Unknown M (m/z 138-120) BK −0.629 −0.799 - - −0.710 0.186 - - - Unknown N (m/z 151-43) BL - - - −0.894 - 0.157 - −0.344 (±0.113) - Unknown O (m/z 95-110-138) BM −0.704 −0.859 - −0.721 −0.370 0.195 - - - Unknown P (m/z 138) BN −0.669 −0.644 - - −0.694 0.172 - - - Unknown Q (m/z 102-55-69) BO −0.208 −0.194 - 0.866 0.235 0.059 −0.497 (±0.185) +0.70 (±0.130) −0.242 (±0.097) Unknown S (m/z 167-121) BP 0.364 −0.613 - - - 0.157 - - - Unknown T (m/z 70-55-82) BQ - - - −0.590 - 0.217 - −0.245 (±0.065) - Unknown U (m/z 119-159-192) BR −0.457 - - −0.928 −0.528 0.181 - - - Unknown W (m/z 121-136-161) BS - −0.761 - - - 0.193 - - - Notes: LAB and yeast columns were removed from the PLS-R information since these variables were never significant. * TA, titratable acidity; ** CA, combined acidity. In parenthesis standard errors. Foods 2019,8, 280 9 of 17 Foods 2019, 8, x FOR PEER REVIEW 11 of 19 Figure 1. Standardised coefficients obtained after the ANOVA analysis (an = O, equivalent to stablish T6 treatment, spontaneous fermentation, as standard) for two selected VOCs: (A) 2-phenylethyl acetate, and (B) 3-methyl-1-butanol. In the case of 2-phenylethyl acetate (Figure 1A), the treatments 4 and 5 promoted the formation of this compound due to the presence of the yeast in the inoculum. Interestingly, this compound was stimulated not only by the presence of the yeast but also LPG1 led to an important contribution and, therefore, the presence of this LAB did not interfere with its possible formation but even stimulated it. However, this effect was not observed in the treatments inoculated with Lp13 or Lpl15, since their 2-phenylethyl acetate contents were similar to those in the spontaneous process. Hence, the behaviour of LPG1 differs from Lp13 and Lp115 regarding the formation of 2- phenylethyl acetate. Similarly, 3-methyl-1-butanol was promoted by the presence of the yeast (Figure 1B); however, Lp13 had a marked negative effect on its formation, which is also reflected in T5, where the joint presence of Lp13 and Y12 has practically prevented its presence. The effects of inoculation treatments on these two volatile substances are reflected in Table 2 with a sign (+ means promotion or increase versus the spontaneous process whereas - indicates prevention or decrease) and the coefficient value (the large the value the most important the effect while the absence of data means not significant contribution). A similar methodology was also followed for the other 69 T1 T2 T3 T4 T5 T6 -0.2 0 0.2 0.4 0.6 0.8 1 1.2 Standardized coefficients Treatment 2-Phenylethyl acetate / Standardized coefficients (95% conf. interval) A) B) -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 1.2 Standardized coefficients Treatment 3-Methyl-1-butanol / Standardized coefficients (95% conf. interval) T1 T2 T3 T4 T5 T6 Figure 1. Standardised coefficients obtained after the ANOVA analysis (a n =O, equivalent to stablish T6 treatment, spontaneous fermentation, as standard) for two selected VOCs: ( A ) 2-phenylethyl acetate, and (B) 3-methyl-1-butanol. In the case of 2-phenylethyl acetate (Figure 1A), the treatments 4 and 5 promoted the formation of this compound due to the presence of the yeast in the inoculum. Interestingly, this compound was stimulated not only by the presence of the yeast but also LPG1 led to an important contribution and, therefore, the presence of this LAB did not interfere with its possible formation but even stimulated it. However, this effect was not observed in the treatments inoculated with Lp13 or Lpl15, since their 2-phenylethyl acetate contents were similar to those in the spontaneous process. Hence, the behaviour of LPG1 differs from Lp13 and Lp115 regarding the formation of 2-phenylethyl acetate. Similarly, 3-methyl-1-butanol was promoted by the presence of the yeast (Figure 1B); however, Lp13 had a marked negative effect on its formation, which is also reflected in T5, where the joint presence of Lp13 and Y12 has practically prevented its presence. The effects of inoculation treatments on these two volatile substances are reflected in Table 2with a sign (+means promotion or increase versus the spontaneous process whereas - indicates prevention or decrease) and the coefficient value (the large the value the most important the effect while the absence of data means not significant contribution). A similar Foods 2019,8, 280 16 of 17 that the type of inoculum modulates the volatile composition of the final product significantly. The inclusion of yeast in the inoculum increases the production of VOCs while the presence of Lactobacillus alone, in general, decrease the concentrations of some compounds or keep them at the same levels than in the spontaneous process. This lack of impact on the VOCs by Lactobacillus strains may be explained because the emphasis when selecting starters was mainly focused on the acidification and the pH lowering characteristics. However, as the process is better known from the microbiological point of view, the introduction of genomic methodologies and the application of more accurate and sophisticated methods for the identification of metabolites formed during the process could make possible the design of inocula with wider and better-identified characteristics, including their aromatic profile. Therefore, to enhance the organoleptic characteristics of final products, the inclusion of yeasts in the inoculum appears as a promising alternative. By studying in detail, the relationships between the VOCs formed and the sensory characteristics, appropriate selection of yeast could be achieved. Besides, the relationships found in this work between starter cultures and VOCs may facilitate further studies on the numerous metabolic transformation occurring in table olive fermentations. Supplementary Materials: The following are available online at http://www.mdpi.com/2304-8158/8/8/280/s1, Table S1: Table S1. Volatile composition determined by GC-MS analysis in the brines of the different treatments assayed at the end of fermentation. T1 stand for treatment inoculated with LPG1, T2 inoculated with Lp13, T3 inoculated with Lpl15, T4 inoculated with Y12, and T5 inoculated with Y12+LPG1+Lp13+Lpl15. Author Contributions: Conceptualization: F.R.-G., R.J.-D., F.N.A.-L., and A.G.-F.; Methodology: A.B.-C., F.R.-G., and M.L.M.; Software: A.G.-F. and M.L.M.; Validation: F.R.-G. and M.L.M.; Formal Analysis: A.G.-F. and M.L.M.; Investigation: A.B.-C., F.R.-G., F.N.A.-L. and A.G.-F.; Resources: R.J.-D. and F.N.A.-L.; Data Curation: M.L.M. and A.G.-F.; Writing—Original Draft Preparation: A.B.-C., M.L.M., F.N.A.-L., and A.G.-F.; Writing—Review & Editing: M.L.M., A.G.-F., and F.N.A.-L.; Visualization: M.L.M. and F.N.A.-L.; Supervision: F.N.A.-L.; Project Administration: R.J.D. and F.N.A.-L.; Funding Acquisition: R.J.D. and FNA.-L. Funding: The research was funded by the Spanish Government (Project OliFilm AGL-2013-48300-R: www.olifilm. science.com.es) AB-C thanks the Spanish Ministry of Economy and Competitiveness for their FPI grant. Acknowledgments: AGF thanks the CSIC for his “Ad honorem” appointment. Conflicts of Interest: The authors declare no conflict of interest. References 1. International Olive Oil Council (IOC). World Table Olive Figures. 2019. Available online: http://www. internationaloliveoil.org/estaticos/view/132-world-table-olive-figures (accessed on 24 July 2019). 2. Garrido-Fern á ndez, A.; Fern á ndez-D í ez, M.J.; Adams, R.M. Table Olives Production and Processing; Chapman & Hall: London, UK, 1997. 3. Hurtado, A.; Requant, C.; Bordons, A.; Rozes, N. Lactic acid bacteria from fermented olives. Food Microbiol. 2012,31, 1–8. [CrossRef] [PubMed] 4. Arroyo-Lopez, F.N.; Romero-Gil, V.; Bautista-Gallego, J.; Rodriguez-Gomez, F.; Jimenez-Diaz, R.; Garcia-Garcia, P.; Querol, A.; Garrido-Fernandez, A. Yeasts in table olive processing: Desirable or spoilage microorganisms? Int. J. Food Microbiol. 2012,160, 42–49. [CrossRef] [PubMed] 5. Arroyo-L ó pez, F.N.; Bautista-Gallego, J.; Dom í nguez-Manzano, J.; Romero-Gil, V.; Rodr í guez-G ó mez, F.; Garc í a-Garc í a, P.; Garrido-Fern á ndez, A.; Jim é nez-D í az, R. Formation of lactic acid bacteria-yeasts communities on the olive surface during Spanish-style Manzanilla fermentations. Food Microbiol. 2012,32, 295–301. 6. Ben í tez-Cabello, A.; Romero-Gil, V.; Rodr í guez-G ó mez, F.; Garrido-Fern á ndez, A.; Jim é nez-D í az, R.; Arroyo-L ó pez, F.N. Evaluation and identification of poly-microbial biofilms on natural green Gordal table olives. Antonie Van Leeuwenhoek 2015,108, 597–610. [CrossRef] [PubMed] 7. Ben í tez-Cabello, A.; Calero-Delgado, B.; Rodr í guez-G ó mez, F.; Garrido-Fern á ndez, A.; Jim é nez-D í az, R.; Arroyo-L ó pez, F.N. Biodiversity and multifunctional features of lactic acid bacteria isolated from table olive biofilms. Front. Microbiol. 2019,10, 836. [CrossRef] [PubMed] 8. Grounta, A.; Panagou, E.Z. Mono and dual species biofilm formation between Lactobacillus pentosus and Pichia membranifaciens on the surface of black olives under different sterile brine conditions. Ann. Microbiol. 2014,64, 1757–1767. [CrossRef] Foods 2019,8, 280 17 of 17 9. Behera, S.S.; Ray, R.C.; Zdolec, N. Lactobacillus plantarum with functional properties: An approach to increase safety and shelf-life of fermented foods. BioMed Res. Int. 2018. [CrossRef] 10. Sabatini, N.; Marsilio, V. Volatile compounds in table olives (Olea Europaea, L., Nocellara del Belice cultivar). Food Chem. 2008,107, 1522–1528. [CrossRef] 11. L ó pez-L ó pez, A.; S á nchez, H.; Cort é s-Delgado, A.; de Castro, A.; Montaño, A. Relating sensory analysis with SPME-GC-MS data for Spanish-style green table olive aroma profiling. LWT Food Sci. Technol. 2018 , 89, 725–734. [CrossRef] 12. De Angelis, M.; Campanella, D.; Cosmai, L.; Summo, C.; Rizzello, C.G.; Caponio, F. Microbiota and metabolome of un-started and started Greek-type fermentation of Bella di Cerignola table olives. Food Microbiol. 2015,52, 18–30. [CrossRef] 13. Tufariello, M.; Durante, M.; Ramires, F.; Grieco, F.; Tommasi, L.; Perbellini, E.; Falco, V.; Tasioula-Margari, M.; Logrieco, A.F.; Mita, G.; et al. New process for production of fermented black table olives using selected autochthonous microbial resources. Front. Microbiol. 2015,6, 1007. [CrossRef] [PubMed] 14. De Castro, A.; S á nchez, A.H.; Cort é s-Delgado, A.; L ó pez-L ó pez, A.; Montaño, A. Effect of Spanish-style processing steps and inoculation with Lactobacillus pentosus starter culture on the volatile composition of cv. Manzanilla green olives. Food Chem. 2019,271, 543–549. [CrossRef] [PubMed] 15. De Castro, A.; S á nchez, A.; L ó pez-L ó pez, A.; Cort é s-Delgado, A.; Medina, E.; Montaño, A. Microbiota and metabolite profiling of spoiled Spanish-style green table olives. Metabolites 2018 ,8, 73. [CrossRef] [PubMed] 16. Pino, A.; Vaccalluzzo, A.; Solieri, L.; Romeo, F.; Todaro, A.; Caggia, C.; Arroyo-L ó pez, F.N.; Bautista-Gallego, J.; Randazzo, C. Effect of sequential inoculum of beta-glucosidase positive and probiotic strains on brine fermentation to obtain low salt Sicilian table olives. Front. Microbiol. 2019,10, 174. [CrossRef] [PubMed] 17. Ubeda, C.; Callej ó n, R.M.; Troncoso, A.M.; Peña-Neira, A.; Morales, M.L. Volatile profile characterisation of Chilean sparkling wines produced by traditional and Charmat methods via sequential stir bar sorptive extraction. Food Chem. 2016,207, 261–271. [CrossRef] [PubMed] 18. Vicente Villal ó n, J.L. MULTIPLOT. A Package for Multivariate Analysis Using Biplots; Departamento de Estad í stica, Universidad de Salamanca: Salamanca, Spain, 2016; Available online: http://biplot.usal.es/ ClassicalBiplot/index.htlm (accessed on 24 July 2019). 19. Panagou, E.Z.; Tassou, C.C. Changes in volatile compounds and related biochemical profile during controlled fermentation of cv. Conservolea green olives. Food Microbiol. 2006,23, 738–746. [CrossRef] [PubMed] 20. Cosmai, L.; Campanella, D.; De Angelis, M.; Summo, C.; Paradiso, V.M.; Pasqualone, A.; Caponio, F. Use of starter cultures for table olives fermentation as possibility to improve the quality of thermally stabilized olive-based paste. LWT Food Sci. Technol. 2018,90, 381–388. [CrossRef] 21. Bonatsou, S.; Tassou, C.C.; Panagou, E.Z.; Nychas, G.J.E. Table olive fermentation using starter cultures with multifunctional potential. Microorganisms 2017,5, 30. [CrossRef] [PubMed] 22. Sabatini, N.; Perri, E.; Marsilio, V. An investigation on molecular partition of aroma compounds in fruit matrix and brine medium of fermented table olives. Innov. Food Sci. Emerg. Technol. 2009 ,10, 621–626. [CrossRef] 23. Viana, F.; Gil, J.V.; Genoves, S.; Valles, S.; Manzanares, P. Rational selection of non-Saccharomyces wine yeasts for mixed starters based on ester formation and enological traits. Food Microbiol. 2008 ,25, 778–785. [CrossRef] 24. Cavin, J.F.; Andioc, V.; Etievant, P.; Divi è s, C. Ability to wine lactic acid bacteria to metabolize phenol carboxylic acids. Am. J. Enol. Vitic. 1993,44, 76–80. 25. Couto, J.A.; Campos, F.M.; Figueiredo, A.R.; Hogg, T.A. Ability of lactic acid bacteria to produce volatile phenols. Am. J. Enol. Viticul. 2006,57, 166–171. 26. Randazzo, C.L.; Todaro, A.; Pino, A.; Pitino, I.; Corona, O.; Caggia, C. Microbiota and metabolome during controlled and spontaneous fermentation of Nocellara Etnea table olives. Food Microbiol. 2017 ,65, 136–148. [CrossRef] [PubMed] © 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).