Cyanidin-3-O-rutinoside (aka keracyanin), cyanidin-3-O-glucoside (aka kuromanin) and luteolin were purchased from Extrasynthese (Genay Cedex, France). 3-O-caffeoylquinic acid (aka neochlorogenic acid), 4-O-caffeoylquinic acid (aka cryptochlorogenic acid), 5-O-caffeoylquinic acid (aka chlorogenic acid), kaempferol, quercetin-3-O-rutinoside hydrate (aka rutin hydrate), quercetin dihydrate, DL-3-phenyllactic acid, 3,4-dihydroxybenzoic acid (aka protocatechuic acid), caffeic acid and toluene were purchased from Sigma-Aldrich (St. Louis, MO, USA). DL-p-hydroxyphenyllactic acid was from Santa Cruz Biotechnology (Dallas, TX, USA). HPLC-grade acetonitrile was from Sigma-Aldrich, while HPLC-grade water and LC-MS grade formic acid were purchased from VWR International (Milan, Italy).

Four strains of L. casei belonging to the Food and Drug Department, University of Parma (Italy) were used as adjunct in elderberry juice. L. casei 2057, isolated from Grana Padano cheese and L. casei 2306, 2240, and 2246 from Parmigiano Reggiano cheese were maintained as frozen stocks (-80°C) in de Man Rogosa Sharpe (MRS) medium (Oxoid, Milan, Italy) supplemented with 15% glycerol (v/v). The cultures were propagated three times with about 3% (v/v) inoculum in MRS and incubated in anaerobiosis (AnaeroGen, Oxoid) at 37°C for 15 h.

Viable cells to be added to elderberry juices were prepared cultivating the strains until the late exponential phase (ca. 15 h), harvesting the cell by centrifugation (10,000 × g for 10 min at 4°C), washing twice with Ringer's solution (Oxoid, Milan, Italy), and finally re-suspending in sterile distilled water to a final concentration of 9.0 Log CFU/mL. CFE from the same strains were prepared by sonication of microbial cultures using SONOPULS Ultrasonic homogenizers (Bandelin, Germany) In brief, three cycles of sonication at 12 W for 30 s, with 60 s of pause between the cycles, were carried out. MRS plate agar count was performed in duplicate to verify the consistency of sonication. The efficacy of sonication was calculated as follow: E = 100- (n_t/n₀)^*100 where n₀ = CFU/mL of L. casei cultures and n_t = CFU/mL of L. casei cultures after sonication treatment. The CFE were separated from the membrane residues by subsequent centrifugation at 10,000 × g for 15 min at 4°C (Centrifuge 5810 R, Eppendorf). The supernatant was sterilized by filtration and the absence of growth was evaluated in TSB 0.6% YE (tryptone soya broth, yeast extract) after incubation at 37°C for 48 h. All the samples were conserved at −80°C. Moreover, the protein content of CFE were determined by fluorometric assay, using Qubit protein assay kit and the measurements were carried out with Qubit Fluorometer (Invitrogen, Thermo Fisher Scientific, Massachusetts, USA).

A commercial pasteurized non-filtered elderberry juice was added with viable cells and CFE of L. casei. Viable cells of L. casei 2057, 2306, 2240, 2246, prepared as described previously, were inoculated into elderberry juice in order to reach 7 Log CFU/mL. The juice was fermented at 37°C for 48 h. Microbial counts before and after fermentation were performed in triplicate using MRS Agar medium (Oxoid, Italy) and incubating at 37°C for 48–72 h, under anaerobiosis. Elderberry juice was also added with CFE from the same strains, incubated at 37°C for 48 h and after this period the absence of cultivable cells was evaluated using MRS plate count agar. The eight experiments were carried out in triplicate and elderberry juice not added of L. casei was incubated at 37°C for 48 h and used as control. All the samples were maintained at −80°C until the analysis.

The volatile profile of elderberry juice samples added with viable inocula and CFE of L. casei, incubated for 48 h at 37°C was fully characterized by means of HS-SPME/GC-MS technique following the protocol reported by Ricci et al. (2018). Briefly, 2 mL of elderberry juice were placed in a 20 mL glass vial adding 5 μL of an aqueous toluene standard solution (100 μg/mL in 10 mL). Head space solid phase micro-extraction was performed for 30 min at 40°C after 15 min of equilibration time. For each analysis a SPME fiber coated with 50/30 μm of Divinylbenzene–Carboxen–Polydimethylsiloxane (DVB/Carboxen/PDMS) was used (Supelco, Bellefonte, PA, USA). Before the analyses, the fiber was conditioned by insertion into the GC–MS injector at 230°C for 2 min, analogously, the desorption of volatiles was accomplished exposing the fiber into the GC injector for 2 min at 230°C. GC–MS analyses were performed on a Thermo Scientific Trace 1300 gas chromatograph coupled to a Thermo Scientific ISQ single quadrupole mass spectrometer equipped with an electronic impact (EI) source. All samples were injected in splitless mode, maintaining the valve closed for 2 min. Helium was used as carrier gas with a total flow of 1 mL/min. The separation was performed on a SUPELCOWAX 10 capillary column (Supelco, Bellefonte, PA, USA; 30 m × 0.25 mm × 0.25 μm) at programmed temperature starting from 50°C for 3 min, increasing of 5°C per min to reach 200°C and maintaining this final temperature for 12 min. The transfer line temperature was 250°C. The signal acquisition mode was full scan (from 41 m/z to 500 m/z). The main volatile compounds of elderberry juices were identified on the basis of their mass spectra compared with the library (NIST 14) mass spectra. To obtain a more confident identification, the linear retention indices (LRIs) were calculated on the basis of the retention times of a solution of C8–C20 alkanes analyzed under the same conditions applied for sample analyses. The semi-quantification of all detected gas-chromatographic signals was performed on the basis of the use of an internal standard (toluene).

All the juices added with L. casei and the control were analyzed with an Accela UHPLC 1250 equipped with a linear ion trap-mass spectrometer (MS) (LTQ XL, Thermo Fisher Scientific Inc., San Jose, CA, USA) fitted with a heated-electrospray ionization probe (H-ESI-II; Thermo Fisher Scientific Inc., San Jose, CA, USA). Analyte separation was performed using an Acquity UPLC HSS T3 (2.1 × 100 mm) column coupled with a pre-column Acquity UPLC HSS T3 VanGuard (2.1 × 5 mm) (Waters, Milford, MA, USA). The injection volume was 5 μL, the autosampler temperature was set at 10°C and column compartment temperature was set at 40°C. For the determination of phenolic profile was followed the protocol reported by Zaupa et al. (2015) with some modification. Briefly, acetonitrile (phase A) and water (phase B) both acidified with formic acid (0.1% v/v) were used as eluents applying a gradient of elution. In particular, the initial conditions were set at 95% B and 5% A and maintained for 0.5 min, then the phase A turned up at 51% in 9 min. After this step, at 9.5 min phase A was increased at 80%, the column was flashed under these conditions for 2 min and then, at 12 min, initial proportions were re-established. The total run time was 17 min at a flow rate of 0.3 mL/min.

The anthocyanin detection was carried out in positive ionization mode, with a capillary temperature of 275°C, and a source heater temperature of 300°C. The sheath gas (nitrogen) flow was 40 units, while auxiliary gas (nitrogen) was equal to 5 units. The source voltage was 4.5 kV, while the capillary and tube lens voltages were 20 and 95 V, respectively. Untargeted preliminary analyses were carried out in order to evaluate differences in the anthocyanin profile among the investigated samples. In detail, the mass spectrometer worked in full scan mode, data-dependent MS³ scanning from m/z 220–900, with CID equal to 30. Once performed the identification, the anthocyanins have been quantified in Selective Reaction Monitoring (SRM) mode using a Collision Induced Dissociation (CID) equal to 30. Cyanidin-3-O-glucoside was quantified through its authentic standard compound, while cyanidin-O-sambubioside, cyanidin-O-sambubioside-O-hexoside and cyanidin-3-O-rutinoside were quantified by using authentic standard of cyanidin-3-O-rutinoside.

The other polyphenolic compounds were analyzed in negative mode, applying the same capillary and source temperatures used before. The sheath gas (nitrogen) flow was 40 units, while auxiliary gas (nitrogen) was equal to 5 units. The source voltage was 4 kV, while the capillary and tube lens voltages were −33 and −98 V, respectively. Untargeted preliminary analyses were carried out in order to evaluate differences in the polyphenolic profile, as cited above for anthocyanin profile. In detail, the mass spectrometer worked in full scan mode, data-dependent MS³ scanning from m/z 100–2,000, with CID equal to 35. Once performed the identification through untargeted analyses, the aglycone flavonoids have been quantified in SIM mode, while the other negatively charged polyphenols were quantified in SRM mode using a CID equal to 35. Pure helium gas was used for CID. Data processing was performed using Xcalibur 2.2 software from Thermo Fisher Scientific Inc. (San Jose, CA, USA). Where possible, polyphenols were quantified trough external calibration curves by using the reference standard compound. When such standards were not available, polyphenols were quantified using a structurally related compound. In detail, isorhamnetin has been quantified as quercetin dihydrate equivalents, coumaroylquinic acid I and coumaroylquinic acid II were quantified as 3-O-Caffeoylquinic acid and 5-O-Caffeoylquinic acid, respectively. Quercetin-3-O-glucoside, isorhamnetin-O-hexoside, kaempferol-O-rutinoside, isorhamnetin-O-rutinoside and quercetin-O-rutinoside-O-hexoside were quantified as rutin hydrate equivalents.

One way ANOVA was used to compare the different results obtained in juices added with viable and CFE L. casei strains, performed with IBM SPSS statistics package v. 21 (IBM Corp., Armonk, NY, USA). To evaluate the normal distribution for every group of independent samples Shapiro-Wilk test was used; multiple comparisons were performed with Tukey's test, differences were considered statistically significant when p < 0.05. Data obtained from the analyses, such as volatile compounds amounts and polyphenols concentration were analyzed by Principal Component Analysis (PCA) using the packages ade4 and FactoMineR on R environment (http://www.r-project.org). The principal coordinates were selected depending on their weight on each of the principal components, and a threshold value was set at ± 0.75 (Table S1). Volatile compounds and polyphenols concentrations were Z-transformed and used to build heatmaps, by using the cim function in mixOmics package; radarplot graphs were built using package fmsb, both running on R software v. 3.4.1.

The elderberry juice was added, singularly, of four cultures containing viable cells of L. casei (2057, 2306, 2240, and 2246) and of the respective CFE. The growth ability of the four viable strains was evaluated after 48 h of incubation by plate count in MRS agar, revealing two different behaviors. Strains 2057 and 2306 shown zero-growth, meaning that cell densities remained unchanged from the original inocula (about 7 Log CFU/mL). On the other hand, strains 2240 and 2246 grew of about 2 Log CFU/mL. During the incubation of the started juices the pH only moderately decreased from an initial value of 3.84 ± 0.07 to 3.73 ± 0.05 (mean value). Unstarted juice (control sample) didn't show microbial growth after 48 h.

The efficacy of sonication in L. casei cellsresulted equal to 99%. To obtain a sterile sample the absence of bacterial growth was evaluated in TSB added with 0.6% of yeast extract after filtration step. In order to analyse the protein content of CFE, fluorometric assay was used. The protein content obtained was 0.81 ± 0.10 mg/mL (zero-growth strains) and 2.59 ± 0.05 mg/mL (grow strains) according with cellular concentration. CFE was added in elderberry juice to obtain comparable concentrations with viable cells (9 Log CFU/mL, grow strains, 2240 and 2246; 7 Log CFU/mL, zero-growth strains, 2057 and 2306) and the absence of bacterial growth was tested after 48 h of incubation.

Alcohols, terpenes and norisoprenoids, acids, ketones and esters (67 compounds) were the main volatiles detected and semi-quantified in elderberry juices added with viable cells, CFE and control (Tables 1, 2). Phenolic acids and flavonoids (Tables 3, 4) were also identified and quantified resulting in a total of 23 compounds. PCA considering the identified volatiles and polyphenols was performed to highlight the differences among the samples analyzed. Considering the 90 variables, the total variance explained by PCA accounted to 64.8% and a good clustering of the samples was obtained according to the growth profile of the viable cells, either growing or zero-growing, and of juices added with CFE (data not shown). Nevertheless, due to the high number of variables used for the classification, 34 of them belonging to volatile compounds (acids, alcohol, esters, ketones, terpenes) and polyphenols (anthocyanins, flavonol glycosides, hydroxycinnamic acids) were selected according to their contribution either to dimension 1 or dimension 2 of the PCA (Table S1). The new PCA analysis explains 77.3% of the variance, with PC1 and PC2 describing 48.3 and 29% of variance, respectively (Figures 1A,B).

Elderberry juices added with viable cells and CFE were grouped in three distinct cluster, separated from the control (Figure 1A). Samples where bacterial growth occurred were separated from the control in the upper right quadrant of PCA, strongly positively associated with all the acids (acetic acid, caproic acid, 2-hexenoic acid (E) acid, caprilic acid), isoamyl isovalerate and two terpenes.

On the other hand, the variables with strongly negative values on the first component (quercetin, ethyl acetate, methyl isovalerate) were determinant to group the elderberry juices added with CFE.

Interestingly, elderberry juice added with L. casei showing zero-growth appeared well separated from the control in lower right quadrant of PCA, associated with variables with positive values on component 1 and negative value on component 2, as many terpenes and alcohols.

In accordance with the PCA analysis, the heatmap describing the volatile compounds showed a separation among the three type of samples highlighting the peculiarities of specific strains (Figure 2). Indeed, elderberry juice added with CFE of L. casei 2057, differently from the other CFE samples and the control, has a profile characterized by a higher production of volatile compounds, especially esters and alcohols. Despite esters were overall less abundant among the samples (Table 2) they were associated to fruit notes contributing positively to the aromatic profile. The heatmap also clearly displays that samples showing either a growth or a zero-growth phenotype were marked by different volatiles, that were mainly acid, esters and ketones for the former, and alcohols and terpenes for the latter. Most of the acids, acetic in particular, increased in juice fermented with L. casei growing strains 2240 and 2246 (Table 2), whereas isovaleric acid resulted higher in the juice added with CFE of L. casei 2057 compared to the control (p < 0.05), where it reached 0.68 ± 0.15 μg/mL.

Moreover, after 48 h of incubation the alcohol content increased, especially in samples with zero-growth, with the highest production found in elderberry juice with L. casei 2057 (4.93 ± 0.73 μg/mL) (Table 3). Notably, hexanol (V-AL4), and (E) 2- hexen-1-ol (V-AL8), associated with green aromatic notes, were the most abundant and their concentration significantly differs from control (p < 0.05). Similarly, terpenes and norisoprenoids, such as β-damascenone, β-linalool and α-terpineol, increased mainly in juice added with zero-growth strains L. casei 2057 and L. casei 2306 (Table 2).

The polyphenolic profiles drive a different clusterization of samples in comparison with volatile compounds (Figure 3). CFE and growth strains grouped separately, instead samples with zero-growth showed a different potential. To note, despite the absence of growth, L. casei 2057 was able to significantly increase many polyphenolic compounds, especially anthocyanins (Table 4).

The (poly)phenolic profile of samples analyzed was mainly characterized by hydroxycinnamic acids, phenyllactic acids, flavone, flavonols, flavonol glycosides, anthocyanins and hydroxybenzoic acid. Samples shown a different trend of hydroxycinnamic acids. An increase, in comparison with the control, was observed for all the juice added with CFE, but especially in juice added with L. casei 2057 zero-growth (Table 4). Conversely, juice fermented with growing L. casei showed a reduction in these compounds after 48 h or no differences with unfermented juice (Table 4). About phenyllactic acids only 3 samples have shown differences in comparison to the control and among these, fermentation with 2240 growing strain, lead to the highest concentration of phenyllatic acid (6.08 ± 0.35 μg/mL) and p-hydroxyphenyllactic acid (9.10 ± 0.39 μg/mL) (Table 4). Flavonoids identified comprise flavone, flavonols (mainly found as glycosides) and anthocyanins. Total flavonols (not glycosylated), as quercetin, kaempferol and isorhamnetin, resulted highest in the juice added with CFE of L. casei 2306 (20.29 ± 2.40 μg/mL) (Table 4). However, a higher concentration of these compounds was observed in their glycosylated forms: quercetin as quercetin-3-O-rutinoside, quercetin-3-O-glucoside, kaempferol as kaempferol-O-rutinoside and isorhamnetin as isorhamnetin-O-hexoside or isorhamnetin-O-rutinoside. Quercetin-3-O-rutinoside resulted the highest especially in juice added with L. casei 2057 zero-growth and quercetin-3-O-glucoside in CFE of 2306 (both significantly different when compared to the control, p < 0.01). The last and the most interesting results were about the anthocyanins (Table 4). Surprisingly, we observed a huge increase of these compounds in juice added with L. casei 2057 zero-growth, where their concentration reached 115.57 ± 8.58 μg/mL, in comparison of 9.62 ± 0.54 μg/mL of the control. In particular, the most abundant anthocyanins were cyanidin-3-O-sambubioside and cyanidin-3-O-glucoside, while cyanidin-3-O-rutinoside and cyanidin-O-sambubioside-O-hexoside were almost negligible.

In order to better highlight the differences among juices added with viable cells and CFE we focused on variables that significantly differ among the samples, belonging to the most abundant classes of detected compounds which are alcohols and terpenes for the volatiles and anthocyanins and flavonol glycosides for polyphenols. Moreover, we have considered phenyllactic acids because they are often found in fermented food as a consequence of microbial metabolism and exert an interesting role as antimicrobials.

The radar plot (Figure 4) shows how viable cells and CFE from the same strain differently impact on the variables considered. Firstly, the viable cells lead to increase the relevant polyphenols and volatiles more than CFE, except for strain 2246. It can also be observed that viable cells of growing strains, in particular the strain 2240, were able to especially increase the content of phenyllactic acids while viable cells of non-growing strains strongly impact on more variables, especially alcohol and terpenes. Finally, the strain 2057, even in absence of growth, shows the greatest impact to increase the polyphenolic content (especially anthocyanins) and the volatile compounds associated with typical aroma of elderberry juice (hexanol, β-linalool, α-terpineol, β-damascenone and (E)-2-hexen-1-ol).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.