In the present study, olives of Nocellara Etnea cultivar, kindly provided by local companies, situated in Adrano and Paternò, Catania, Sicily, were processed at industrial scale, following the Sicilian style method, without the addition of sodium hydroxide solution. After harvesting (September–October, 2019), drupes were pretreated according to Pino et al. (2018a) and directly placed into brines, containing 5 and 8% of sea salt.

The L. plantarum C11C8, F1.16, and F3.5 strains, previously isolated from brine samples of natural Sicilian table olives at 5% of NaCl (Pino et al., 2019) and characterized for the ability to grow and to degrade the oleuropein under stress conditions (Vaccalluzzo et al., 2020b), were used as starter cultures. In detail, the two starter cultures were set up as follows: starter 1: with L. plantarum F1.16 and F3.5 strains and starter 2: with L. plantarum C11C8, F1.16, and F3.5 strains. The starter cultures were inoculated to a final cell density of 7 log CFU/mL, directly after brining. As displayed in Figure 1, four different fermentations were carried out: O1: 5% of NaCl, with the addition of starter 1; O2: 5% of NaCl, with the addition of starter 2; and C5 (5% of NaCl) and C8 (8% of NaCl) without the addition of any starters, both used as control. All fermentations were carried out at room temperature (18 ± 2°C) and monitored for 80 days. In addition, salt was periodically added into the samples, in order to maintain the initial sodium chloride concentration. Each fermentation was carried out in triplicate.

The pH of brine was measured, during the fermentation process at times 0,7,15, 24, 30, 60, and 80 days, using a MettlerDL25 pH meter (MettlerDL25; Mettler–Toledo International Inc.). The NaCl salt content was monitored following the method proposed by Benítez-Cabello et al. (2020). Microbiological analyses were performed on both brine and olive samples after 0, 15, 30, 60, and 80 days of fermentation following the method reported by Pino et al. (2018a,b, 2019). In detail, the drupes were carefully separated from the brine samples, drained, pitted (25 g), transferred into a sterile stomacher bag, 10-fold diluted with Ringer’s solution (Sigma–Aldrich, Milan, Italy), and homogenized for 2–5 min in a stomacher (Lab-Blender, Seward, London). Both brine (10 mL) and olive samples, homogenized as reported above, were serially diluted, using sterile quarter-strength Ringer’s solution, and plated on the following agar media and conditions: plate count agar, incubated at 32°C ± 2°C for 48 h, for total mesophilic bacteria; de Man-Rogosa-Sharp agar, supplemented with cycloheximide (5 mL/L), anaerobically incubated at 32°C for 24–48 h, for LAB count; Sabouraud dextrose agar, supplemented with chloramphenicol (0.05 g/L), incubated at 25°C for 4 days, for yeast count; violet red bile glucose agar, aerobically incubated at 37°C for 24 h, for Enterobacteriaceae count; mannitol salt agar, incubated at 32°C for 48 h, for staphylococci enumeration; MacConkey incubated at 32°C for 24–48 h for Escherichia coli determination; and sulfite polymyxin sulfadiazine agar, anaerobically incubated at 37°C for 24–48 h, for the detection of sulfite-reducing clostridia, as Clostridium perfringens species. All media were purchased from Liofilchem (Roseto degli Abruzzi, Italy). All analyses were performed in triplicate. Results were expressed as log10 CFU/mL for brine and log10 CFU/g for olive samples.

Olives from all samples, at 15 and 80 days of fermentation, were subjected to total DNA extraction using the Dneasy Mericon Food Kit (Qiagen, Milan, Italy) with some modifications. In detail, 12 g of olive samples were diluted with 30 mL of sterile Ringer’s solution and incubated for 2 h at 37°C under constant shaking. After incubation, samples were homogenized in a stomacher apparatus for 3 min at room temperature. The suspension was collected and centrifuged at 10,000 g for 10 min at 20°C, and the pellet was washed twice with 30 mL of phosphate-buffered saline, pH 7.4. The pellet was resuspended in 400 μL of Tris-EDTA buffer (10 mM Tris-HCl, 1 mM EDTA, pH 8.0) 1X, and the suspension was transferred into tube containing 0.3 g of zircon beads, added with 150 μL of phenol solution, and homogenized with Precellys Evolution Homogenizer (Bertin Technologies) at 10,000 rpm for 5 min. The obtained suspension was subjected to DNA extraction, following the manufacturer’s instructions. DNA concentration was determined using the fluorimeter Qubit 4.0 (Invitrogen, Carlsbad, CA, United States) before storage at −20°C until use.

Partial 16S rRNA gene sequences were amplified from extracted DNA using the primer pair Probio_Uni and Probio_Rev, targeting the V3 region of the 16S rRNA gene sequence (Milani et al., 2013). 16S rRNA gene amplification and amplicon checks were carried out as previously described (Milani et al., 2013), and 16S rRNA gene sequencing was performed using a MiSeq (Illumina) according to Milani et al. (2013).

The VOC profile of all samples at 15 and 80 days of fermentation was investigated by GC–MS. In detail, olives (approximately 100 g), from three different replicates, were pitted and homogenized, and 3.3 g of pulp was placed in a 20-mL glass vial. After the addition of 10 mL NaCl solution (300 g/L) and 10 μL of 2-methyl-4-pentanol (final concentration 75 μg/g) as internal standard, the vial was closed, and extraction by HS-SPME was performed with subsequent analysis by GC–MS, according to the method reported by de Castro et al. (2019). Compound identification was based on mass spectra matching with the standard NIST 08 MS library, on the comparison of retention indices sourced from the NIST Standard Reference Database and from authentic reference standards when available. All analyses were made in triplicate. A PAL COMBI-xt autosampler (CTC combiPAL; CTC Analysis AG, Zwingen, Switzerland) was used to standardize the extraction procedure. The olive samples were kept at 60°C for 15 min. The divinylbenzene/carboxen/polydimethylsiloxane (Supelco, Bellefonte, United States) fiber was exposed to the sample headspace for 60 min (Sánchez et al., 2018). VOC injection was made under splitless mode into the port at 230°C, equipped with a Merlin Microseal. A Clarus 680 (PerkinElmer, Beaconsfield United Kingdom) GC equipped with an Rtx-WAX column (30 m × 0.25-mm internal diameter, 0.25-μm film thickness) (Restek Superchrom, Milano, Italy) was used to thermally desorb and to separate the headspace VOCs. The column temperature was set initially at 35°C for 8 min and then increased to 230°C at 4°C/min and held for 15 min (Montemurro et al., 2020). Helium was used as carrier gas at flow rate of 1 mL/min. A single-quadrupole mass spectrometer Clarus SQ8MS (PerkinElmer) was used to detect the different compounds; the source and transfer line temperatures were 250 and 230°C, respectively. The MS detector system operated in scan mode with mass-to-charge ratio interval 30–350 Da.

One-way analysis of variance (ANOVA) followed by Tukey post hoc multiple-comparisons test was applied to the pH values and microbiological and VOC data from three biological replicates. Differences were considered statistically significant at p < 0.05. The differences among fermentation processes of the detected genera were assessed with STAMP software (Parks et al., 2014) using a two-sided Welch t-test with a Benjamini–Hochberg false discovery rate correction (p < 0.05). In order to correlate the experimental and control samples with the volatile compounds, the data obtained at 15 and 80 days of fermentation were subjected to principal component analysis (PCA). Similarities between the volatile profiles of the inoculated and control olive samples were carried out using the PermutMatrix software. The permutation analysis of the significantly different VOCs was evaluated in the drupe samples with (O1–O2) and without (C5–C8) addition of starter during the fermentation process (T15–T80). All statistical analyses were performed using STATISTICA software (version 7.0 for Windows; TIBCO Software, Palo Alto, CA, United States). In addition, Principal coordinate analysis (PCoA) was performed on the UniFrac distance matrices to show the differences among samples. Correlation analysis between cell density in different microbial groups and VOCs evaluated at 15 and 80 days of fermentation was performed. The Spearman rank correlation was computed in R by using the cor test package¹ and plotted by using the corrplot package [Wei and Simko (2017); R package ‘‘corrplot’’: Visualization of a Correlation Matrix (version 0.90)].² Results for genera showing significant correlation (p < 0.05) were visualized as a correlation matrix (R package “corrplot,” version 0.90) (see text footnote 2).

FastQ files were processed using a custom script based on the QIIME2 software suite (Caporaso et al., 2010). Paired-end read pairs were assembled to reconstruct the complete Probio_Uni/Probio_Rev amplicons. Quality control retained sequences with a length between 140 and 400 base pairs (bp) and mean sequence quality score > 20, whereas sequences with homopolymers > 7 bp and mismatched primers were omitted. In order to calculate downstream diversity measures (alpha and beta diversity indices, UniFrac analysis), 16S rRNA Amplicon sequence variants (ASVs) were defined at 100% sequence homology using DADA2 (Callahan et al., 2016); ASVs not encompassing at least two sequences of the same sample were removed. Notably, this approach allows highly distinctive taxonomic classification at single-nucleotide accuracy (Callahan et al., 2016). All reads were classified to the lowest possible taxonomic rank using QIIME2 (Caporaso et al., 2010; Bokulich et al., 2018) against SILVA database release 138 (Quast et al., 2013).

Biodiversity within a given sample (alpha diversity) was calculated with Shannon and Chao1 in QIIME2.

Similarities between samples (beta diversity) were calculated by weighted UniFrac (Lozupone and Knight, 2005). The range of similarities is calculated between values 0 and 1. PCoA representations of beta diversity were performed using QIIME2 (Caporaso et al., 2010; Bokulich et al., 2018).

Table 1 shows pH values of O1, O2, C5, and C8 brine samples measured during the fermentation process. At the beginning of the fermentation, significant differences were found between the C5 and the C8 samples, with values of 6.04 and 6.24, respectively, and the inoculated samples (O1 and O2) exhibited a value of approximately 6.14. As expected, at the beginning of the fermentation, the pH value decreased significantly, reaching a value of approximately 4.5 in the inoculated samples after 24 days, whereas in the controls after 60 days of fermentation. With the exception of the control at 5% of NaCl (C5), a pH < 4.3 was detected at the end of fermentation in all samples, falling within the recommended critical threshold, which would guarantee the microbiological safety of the final product.

Table 2 and Supplementary Table 1 show microbial counts detected both in olives and brines, respectively, after 0, 15, 30, 60, and 80 days of fermentation. Overall, in all olive samples (Table 2), a significant decrease in Enterobacteriaceae counts was observed starting from the 30th day of fermentation, in inoculated samples, especially in O2 sample. Similar trend was revealed by coagulase negative staphylococci, whereas coagulase-positive staphylococci counts were below the threshold limit of detection (<1 absent in 25 g) starting from the 15th day of fermentation, in both inoculated samples (Table 2). Viable mesophilic bacteria exhibited different count levels among samples up to 30 days of fermentation, whereas at the 60th day of fermentation, the inoculated samples registered count values of 4.5, with the O2 sample reaching the value of 4.3 log CFU/g, at the end of fermentation. Higher mesophilic bacteria counts were detected in control samples, which showed, at the 60th and 80th days of fermentation, mean values of 6.20 and 5.5 log CFU/g, respectively. Regarding LAB count, no significant difference was found among samples at the beginning of fermentation, whereas a decrease and an increase of approximately 1 log unit, in control samples and in inoculated samples, respectively, were observed at the end of fermentation. Regarding the yeast population, a significant increase (approximately 6 log units) was observed at the 15th day of fermentation, in all samples, especially in C8 samples, reaching a value of 8 log CFU/g. Starting from the 60th day of fermentation, the yeast population decreased, reaching mean values of 3.9 and 5.2 CFU/g in inoculated and control samples, respectively. Furthermore, the presence of sulfite-reducing species and E. coli was never detected in any samples at any times.

Regarding microbiological analyses of brine samples, at the same sampling times mentioned previously, results are reported in the Supplementary Table 1. Enterobacteriaceae counts showed an initial mean value of 3.6 log CFU/mL, with a significant decrease to a mean value of 2.6 log CFU/mL at the 30th day of fermentation. From an initial value of approximately 5.0 log CFU/mL, coagulase-negative staphylococci decreased significantly through the fermentation, reaching a value below the threshold limit of detection (<1 absent in 25 g) from the 60th day of fermentation in inoculated samples. With the exception of C5 sample, low values of coagulase-positive staphylococci were detected in all samples at the beginning of fermentation, which decreased significantly starting from the 30th day. Mesophilic bacteria decreased through the fermentation in all samples, reaching a mean final value of 6.1 log CFU/mL. LAB counts were higher in inoculated samples than in controls, showing a constant trend up to the end of fermentation. Yeast population showed a reduction of almost 3 log units, with the exception of inoculated sample O1, where a 2 log units’ decrease was observed. The presence of E. coli was never detected in sample O2; its complete inhibition was revealed in O1 and C8 samples starting from the 30th day and from the 60th day in C5 sample. Sulfite-reducing species were not detected (absent in 25 g) in any tested samples.

Inoculated and control table-olives samples both at 15 and 80 days of fermentation were subjected to sequencing of the V3 region of the 16S rRNA gene. Per sample number of reads, denoised reads, Shannon and Chao 1 indices, are reported in Supplementary Table 2. Overall, the analysis allowed obtaining a total of 449,654 bacterial sequences with an average of 56,207 sequences for each analyzed sample. The Chao1 index increased from 15 to 80 days of fermentation except in the control sample C8, in which a decrease in Chao1 index was observed during the fermentation process (Supplementary Table 2). A satisfactory coverage of the bacterial diversity was found for all the analyzed samples with Good’s coverage values greater than 99% (Supplementary Table 2), which was confirmed by rarefaction curves analysis (Supplementary Figure 1, only Richness data are shown). The bacterial biota of the analyzed olive samples was covered by 3 phyla (Figure 2A), 7 families (Figure 2B), and 10 genera (Figure 2C) (relative abundance > 0.5%). In detail, Firmicutes and Proteobacteria were the phyla mainly detected. Among these, Firmicutes showed the highest relative abundance in all samples, with the exception of C5 sample at 80 days of fermentation. The Bacteroidetes phylum was found, even at low abundance, only in C8 sample at the beginning of fermentation (Figure 2A). Zooming on the microbiota profile at family level (Figure 2B), Lactobacillaceae family was mainly detected in inoculated samples at both 15 and 80 days of fermentation, whereas Leuconostocaceae family was revealed in control samples. Enterobacteriaceae family showed the highest occurrence in C5 sample at 80 days of fermentation (Figure 2B). The prevalence of 10 genera (relative abundance > 0.5%) within table-olives samples is depicted in Figure 2C. At the genus level (Figure 2C), Lactobacillus (according with the old nomenclature) dominated the O1 and O2 olive samples at both 15 and 80 days of fermentation. Weissella was mainly detected in C8 sample both at the beginning of fermentation (85.09%) and at the end of the process (78.34%), and in C5 sample at 15 days of fermentation (49.05%). High occurrence of Enterobacter genus was revealed only in C5 sample. The genera Bacteroides, Faecalibacterium, Klebsiella, and Raoultella were detected only in C8 sample at 15 days of fermentation (Figure 2C). By comparing the different fermentation processes, between inoculated and control samples, with a two-sided Welch test, genera Enterobacter, Lactobacillus, and Weissella resulted to be statistically different (Supplementary Figure 2). Looking at group statistics, Enterobacter was statistically higher (p = 0.030) in C5 when compared with C8. The control samples with NaCl 8% were proven to contain significantly higher levels of Weissella (compared with O1 and O2), whereas a significantly lower amount of Lactobacillus was found when compared with O2. Figure 3 shows the unweighted UniFrac analysis based in PCoA of 16S sequences. The PCoA allowed to group samples based on both treatment and fermentation time along the Axes 1 (PC1) that explained more than 75% of total variance. High similarity among C8 olive samples at 15 and 80 days of fermentation was recorded. In addition, based on the PCoA, olive samples O2, at both 15 and 80 days of fermentation, and olive sample O1 at 80 days of fermentation grouped together.

An objective comparison of volatile metabolic profile in table-olives samples at 15 and 80 days of fermentation was performed based on qualitative and quantitative differences in VOCs using HS-SPME GC–MS methodology. Fifty-two volatile compounds were identified and grouped according to chemical classes, that is, alcohols (9), esters (21), aldehydes (7), phenols (5), ketones (2), organic acids (2), terpenes (3), and others (3). The two principal factors of PCA analysis (PC1 and PC2), explaining 63% of the total variance, showed that drupes were clearly distributed according to the time of fermentation (Figure 4). Thirty-one of 52 volatile compounds showed a significant difference among the uninoculated and inoculated table olives during the fermentation (Supplementary Table 3). The significant different compounds, resulting from ANOVA, were used for the permutation analysis (Figure 5). The permutation analysis clearly showed that table olives were grouped into three clusters according to the fermentation process and time. Indeed, cluster I grouped all inoculated table olives at 15 days, cluster II included the uninoculated samples at the same time, whereas cluster III encompassed samples at 80 days of fermentation. It should be noted that O2 sample at 15 days of fermentation was characterized by the highest content of 2-butanone-3-hydroxy (acetoin), ethyl acetate, and lactic acid ethyl ester. The same sample after 80 days of fermentation was characterized by the highest content of acetic acid, 3-methyl-1-butanol and its derivatives esters as well as phenylethyl alcohol, acetic acid 2-methyl ester, 2-heptanal, and benzene propanoic acid methyl ester. It is interesting to note that salt content did not discriminate the VOC profile between control samples at the end of fermentation. In detail, compared with inoculated samples (O1 and O2) at 80 days of fermentation, C5 and C8 were characterized by the higher content (p < 0.05) of 4-ethyl-phenol and the 2-methoxy-phenol. Focusing on alcohols, it is possible to assert that after 80 days of fermentation, a significant decrease in this class of compounds, except for phenylethyl alcohol, 3-methyl-1-butanol, and benzyl alcohol, was observed. A similar trend was detected for aldehydes, with butanal-3-methyl and hexanal showing a significant decrease during the fermentation. An overall increase for acetic acid and derivative esters in all samples was detected during the fermentation. Of note, ethyl acetate decreases in the inoculated olives (O1 and O2) after 80 days of fermentation, whereas an opposite trend was assessed in the uninoculated (C5 and C8) samples.

Statistically significant correlations between microbial groups and VOCs are shown in Figure 6. LAB cell density values positively correlated with acetic acid (p = 0.017) and its derivatives ethyl and methyl esters (acetic acid methyl ester, ethyl acetate, 1-butanol 3-methyl-acetate), as well as with lactic acid ethyl ester (p = 0.009) and 2-butanone-3-hydroxy (acetoin) (p = 0.006). Looking at mesophilic bacteria group, a positive correlation was detected with phenol-2-methoxy and phenol-4-ethyl, which was negatively correlated with the Enterobacteriaceae group (p = 0.011). Moreover, the latter microbial group showed a positive correlation with alcohols (ethanol, 1-hexanol), esters (butanoic acid ethyl ester, hexanoic acid ethyl ester, benzoic acid 2-hydroxy ethyl ester), aldehydes (hexanal and benzeneacetaldehyde), and dimethyl sulfide and a negative correlation with acetic acid (p = 0.012), 1-butanol 3-methyl- (p = 0.015), and phenylethyl alcohol (p = 0.039). Regarding yeasts, a positive correlation with benzoic acid 2-hydroxy-ethyl ester was revealed, whereas 1-butanol-3-methyl-acetate, acetic acid-2-phenylethyl ester, and 6-methyl-5-hepten-2-one were negatively correlated. The above detected remarks were partially confirmed by the correlation analysis between relative abundances of bacterial genera and VOCs (Supplementary Figure 3). In detail, Lactobacillus was positively correlated (p < 0.05) with derivative esters of acetic and lactic acid (ethyl acetate, acetic acid methyl ester, and lactic acid ethyl ester) and 2-butanone-3-hydroxy. On the other hand, this group was significantly negatively correlated with 2–3 butanediol, 1-octanol 3,7-dimethyl-, muurolene, and phenol-2-methoxy. An opposite trend was observed for Weissella, showing a negative correlation (p < 0.05) with ethyl acetate, acetic acid methyl ester, butanoic acid-2-methyl, and 2-butanone-3-hydroxy, whereas positive correlations were found with 1-octanol 3,7- dimethyl-, muurolene, and phenol, 4-ethyl-2-methoxy. Enterobacter was negatively correlated (p < 0.05) with 2.3-dihydro-benzofuran.