Modifications of Phenolic Compounds, Biogenic Amines, and Volatile Compounds in Cabernet Gernishct Wine through Malolactic Fermentation by Lactobacillus plantarum and Oenococcus oeni

: Malolactic fermentation is a vital red wine-making process to enhance the sensory quality. The objective of this study is to elucidate the starter cultures’ role in modifying phenolic compounds, biogenic amines, and volatile compounds after red wine malolactic fermentation. We initiated the malolactic fermentation in Cabernet Gernishct wine by using two Oenococcus oeni and two Lactobacillus plantarum strains. Results showed that after malolactic fermentation, wines experienced a content decrease of total flavanols and total flavonols, accompanied by the accumulation of phenolic acids. The Lactobacillus plantarum strains, compared to Oenococcus oeni , exhibited a prevention against the accumulation of biogenic amines. The malolactic fermentation increased the total esters and modified the aromatic features compared to the unfermented wine. The Lactobacillus plantarum strains retained more aromas than the Oenococcus oeni strains did. Principal component analysis revealed that different strains could distinctly alter the wine characteristics being investigated in this study. These indicated that Lactobacillus plantarum could serve as a better alternative starter for conducting red wine malolactic fermentation.


Introduction
Malolactic fermentation is a vital red wine-making process after alcoholic fermentation since it can significantly improve the taste quality of red wine through converting the tart-taste malic acid into softer-taste lactic acid [1]. It also plays an important role in enhancing the aromatic complexity of wine via metabolism of nutrients by lactic acid bacteria [2][3][4]. It has been known that indigenous lactic acid bacteria naturally present in grapes must initiate spontaneous malolactic fermentation in wine [5,6]. This could trigger the production of undesirable metabolites (such as volatile acids and ethoxymethylenemalonate (derivatization reagent) and external standards of phenolic compounds, biogenic amines, and volatile compounds were all purchased from Sigma-Aldrich (St. Louis, MO, USA). 4-Methyl-2-pentanol (internal standard for volatiles extraction) was also a product from Sigma-Aldrich (St. Louis, MO, USA).

Analysis of Phenolic Compounds
An Agilent 1200 Series high-performance liquid chromatography coupled with an Agilent 6410 triple quadrupole mass spectrometer (HPLC-QqQ-MS/MS, Agilent Technologies, Santa Clara, CA, USA) was used to analyze phenolic compounds in the wine samples according to a published method [21]. The wine sample (1.0 mL) was filtered through a 0.45-µm polyether sulphone membrane, and then the filtered wine sample (1.0 µL) was directly injected to HPLC. Phenolic compounds in the wine samples were separated on a Poroshell 120 EC-C18 column (150 × 2.1 mm, 2.7 µm, Agilent Technologies, Santa Clara, CA, USA) with a flow rate at 0.4 mL/min. Mobile phase consisted of (A) 0.1% v/v formic acid in water and (B) 0.1% v/v formic acid in acetonitrile:methanol (1:1, v/v). The gradient elution program was as follows: 0-28 min, 10%B to 46%B; 28-29 min, 46%B to 10%B; and 29-34 min, 10%B isocratic. The column was maintained at 55 °C during the gradient elution program. A negative electrospray ionization was used on QqQ-MS/MS with a spray voltage at 4 kV, gas temperature at 350 °C, and nebulizer pressure at 35 psi. The ion source temperature was set at 150 °C. Multiple reaction monitoring mode for the transition of the precursor to product ion was used to identify phenolic compounds in the wine samples. The external standard (procyanidin B1, procyanidin B2, procyanidin C1, (+)-catechin, (−)-epicatechin, (−)-epigallocatechin and gallocatechin, myricetin, myricetin-3-O-glucoside, quercetin-3-O-galactoside, quercetin-3-O-glucuronide, dihydroquercetin, syringetin-3-O-glucoside, gallic acid, protocatechuic acid, caffeic acid, and 4hydroxycinnamic acid) were used to quantify flavanols, flavonols, and phenolic acids, respectively. Each wine sample was analyzed in duplicate.

Analysis of Biogenic Amines
A published derivatization-liquid chromatography method was used to analyze biogenic amines in these wine samples with minor modifications [44]. In brief, the wine sample (500 µL) was mixed with 10 µL of 1 g/L L-2-aminoadipic acid (internal standard), 375 µL of methanol, 15 µL of diethyl ethoxymethylenemalonate, and 875 µL of 1 mol/L borate buffer (pH 9.0). The mixture was sonicated for 30 min and then heated at 70 °C in a water bath for 2 h. After the mixture was cooled down to the room temperature, the mixture was filtered through 0.22 µm nylon filters and then 20 µL filtered wine sample was injected to HPLC. A Shimadzu LC-20AT LC system (Shimadzu, Kyoto, Japan) with a Venusil XSB C18 column (4.6 × 250 mm, 5 µm, Bonna-Agela Technologies Co. Ltd., Tianjin, China) was used to analyze the biogenic amine derivatives. The mobile phase was comprised of (A) acetonitrile:methanol (4:1, v/v) and (B) 25 mM acetate buffer (pH 5.8) containing 0.02% sodium azide. The flow rate was set at 0.9 mL/min with a column temperature maintained at 16 °C during the elution program. The elution was programed as follows: 0-20 min, 90%B isocratic; 20-30.5 min, 90%B to 83%B; 30.5-33.5 min, 83%B isocratic; 33.5-65 min, 83%B to 73%B; 65-73 min, 73%B to 28%B; 73-78 min, 28%B to 18%B; 78-82 min, 28%B to 0%B; 82-85 min, 0%B isocratic; 85-90 min, 0%B to 90%B; and 90-93 min, 90%B isocratic. A model wine matrix, consisting of 14% v/v ethanol and 5 g/L of tartaric acid in water (pH adjusted to 3.8 using 5 M NaOH), was mixed with the external biogenic amine standards. The standards wine matrix was analyzed under the same derivatization-liquid chromatography method. The quantitation was performed using the regression curves generated through the peak ratio of external over internal standards versus the concentrations of the external standards. Each sample was analyzed in duplicate.

Analysis of Volatile Compounds
Volatile compounds in these wine samples were extracted using headspace solid-phase microextraction and then analyzed using gas chromatography-mass spectrometry [45]. Briefly, the wine sample (5.0 mL) was mixed with 1 g NaCl, 10 µL of 1.0018 g/L 4-methyl-2-pentanol (internal standard) in a 15-mL glass vial containing a magnetic stirrer. The vial was capped tightly with a PTFE-silicon septum. The mixture was equilibrated at 40 °C for 30 min under an agitation at 300 rpm. Afterwards, a 50/30 µm divinylbenzene/carboxen/polydimethylsiloxane fiber (DVB/CAR/PDMS; Supelco, Bellefonte, PA, USA) was inserted to the headspace of the vial for 30 min to adsorb volatiles at 40 °C under agitation at 300 rpm. Afterwards, the fiber was removed from the headspace of the vial and immediately inserted to the GC injector to release the volatiles at 250 °C for 8 min. An Agilent 6890 gas chromatography coupled with an Agilent 5975 mass spectrometry (Agilent Technologies, Santa Clara, CA, USA) was used to analyze the volatile compounds in the wine samples. An HP-Innowax capillary column (60 m × 0.25 mm × 0.25 µm, J&W Scientific, San Francisco, CA, USA) was used to separate the volatile compounds under a carrier gas (helium) flow rate at 1 mL/min. The oven temperature program was as follows: held at 50 °C for 1 min, increased to 220 °C at a 3 °C/min rate, and held at 220 °C for 5 min. The temperature set at the interface and the ion source on the GC-MS were 280 and 230 °C, respectively. All mass scan from m/z 30 to m/z 350 under a selective ion mode was conducted with the electron impact set at 70 eV. A C6 to C24 n-alkane series (Supelco, Bellefonte, PA, USA) was analyzed using the same chromatography condition to calculate the retention indices. Volatile compounds with their reference standard available were identified by comparing their mass spectrum and retention indices with their corresponding standard. Volatile compounds without the available standard were tentatively identified through comparing their retention indices and mass spectrum with the NIST11 library and the retention indices of references in our local library. Regarding the quantitation of volatile compounds, the external volatile standards were dissolved in ethanol and then diluted to 15 successive levels in the wine model matrix containing 14% v/v ethanol and 5 g/L of tartaric acid in water (pH adjusted to 3.8 using 5 M NaOH). The extraction and analysis of these standard samples followed the same volatiles analysis procedure as the wine samples. Standard curve of each volatile was established through the peak ratio of target volatile over internal standard versus the concentration of target volatile. For volatile compounds with the reference standard, the quantitation was achieved using their corresponding reference standard. Volatiles without the available standard were quantified using the standard that exhibited the same functional group and similar C atom numbers as the volatile. Each wine sample was analyzed in duplicate.

Odor Activity Value and Aroma Profile
Odor activity value (OAV) is a critical indicator to reflect the aromatic contribution of a volatile. The OAV is determined by the volatile concentration over its perception threshold reported in the literature. A volatile compound with OAV greater than 1 (esters with OAV above 0.1) significantly contributes its aroma features. Aroma profile in wine is divided into seven series, including fruity, floral, herbaceous, caramel, balsamic, chemical, and fatty aromas. The overall intensity of each aroma series was calculated through summing up the OAV of the volatiles that have significant aromatic contributions.

Statistical Analysis
Data were expressed as the mean ± standard deviation. One-way analysis of variance (ANOVA) was carried out to compare the means under Tukey's honestly significant difference at a significant level of 0.05 using SPSS Statistics 23 (IBM, New York, NY, USA). Principal component analysis was conducted using XLSTAT (Addinsoft, New York, NY, USA). Only the variables with a significant difference between wine samples were included.

Phenolic Compounds
A total of 17 phenolic compounds including seven flavanols, six flavonols, and four phenolic acids were found in these wine samples ( Table 1). The total phenolic compounds before the malolactic fermentation was 95.72 mg/L. Although the total concentration of phenolic compounds was not altered by any strain, a significant decrease in total flavanols was found, where the sample with C8-1 displayed the lowest. A slight reduction in total flavonols and similar content of total flavonols were also observed after the malolactic fermentation. It is worth noting the dramatic increase of gallic acid and protocatechuic acid in the sample with Lp39. The caffeic acid and 4-hydroxycinnamic acid were not quantified before the malolactic fermentation. However, a significant accumulation of these two phenolic acids was observed in three fermented samples except C8-1. The caffeic acid and 4hydroxycinnamic acid in wines with O. oeni had doubled concentrations compared to samples with Lp39. It has been confirmed that lactic acid bacteria could release numerous enzymes, such as gallate decarboxylase, p-coumaric acid decarboxylase, benzyl alcohol dehydrogenase, oxidoreductase, decarboxylase, and demethylases, in wine during malolactic fermentation. These enzymes could metabolize phenolic compounds, resulting in the reduction of phenolic compounds in wine after malolactic fermentation [27,[46][47][48][49]. This could explain why phenolic compounds, especially flavanols and flavonols, exhibited a concentration decrease in the Cabernet Gernischt wines after the malolactic fermentation in the present study. It should be noted that lactic acid bacteria also have capacity of releasing tannases, and these enzymes could cleave the ester bonds in hydrolysable tannins to yield their corresponding phenolic acid monomers [50,51]. This could increase the accumulation of phenolic acids in wine, such as caffeic acid and 4-hydroxycinnamic acid in the present study.

Biogenic Amines
The Cabernet Gernischt wine before the malolactic fermentation contained total biogenic amines at 7.29 mg/L ( Table 2). Except C8-1, all other wine samples exhibited a significant increase in the total biogenic amines. Regarding the individual biogenic amine, histamine was the dominant one before or after the malolactic fermentation. Its accumulation was observed in three samples expect C8-1. Spermidine and putrescine also existed at a moderate amount. The O. oeni strains slightly reduced the content of putrescine after the malolactic fermentation, whereas the spermidine remained unchanged. Biogenic amines negatively influence the quality of wine since these metabolites could potentially bring safety concerns to consumers [36]. In wine, the amino acids are the major precursors of biogenic amines and these molecules could be metabolized by lactic acid bacteria thus to yield biogenic amines [52]. Therefore, lactic acid bacteria strains have been investigated and screened as starter cultures in the wine industry in terms of their capacity of producing biogenic amines [14]. Different lactic acid bacteria species possess different abilities of metabolizing amino acids to yield biogenic amines, and the O. oeni strains have been reported with high production of biogenic amines [37,38]. For example, the malolactic fermentation in Falanghina wine by O. oeni doubled the histamine compared to that by L. plantarum [39]. Our results were consistent with this report. Meanwhile, L. plantarum strains have been reported to lack some enzymes as O. oeni, which convert amino acids into biogenic amines [9,11,12,40,41]. This might explain why the C8-1 fermented Cabernet Gernischt wine did not increase biogenic amines after malolactic fermentation in the present study.

Volatile Compounds
A total of 71 volatile compounds were detected in these Cabernet Gernischt wine samples (Table  3), including 24 esters, 14 higher alcohols, six aldehydes/ketones, seven fatty acids, 10 terpene derivatives, and 10 other volatiles. The most abundant groups were found to be higher alcohols, followed by esters, aldehydes/ketones, and then fatty acids. Terpene derivatives and other volatiles appeared to exist in these wine samples at a low-level range.

Esters
Esters are the fruity scented volatile compounds in grapes, which are mainly extracted from the maceration process and contribute the varietal aromatic feature to wine [29,53]. Alcoholic fermentation could also result in the formation of esters via the activity of esterase released from yeasts [54]. Esters can be divided into acetate esters, ethyl esters, and other esters according to their structural natures [45].
Acetate esters appeared to be the dominant esters in the Cabernet Gernischt wine before the malolactic fermentation. They represented over 90% of the total esters content (Table 3). It has been reported that ethyl acetate exhibited the pineapple, fruity, solvent, and balsamic scents [55]. Its flavor notes could be significantly incorporated to the overall aroma in the Cabernet Gernischt wine due to its high OAV (Table 4). Isoamyl acetate had a concentration below 1 mg/L before malolactic fermentation (Table 3), but it still significantly contributed its fruity note to the overall aroma in the wine due to its low odor threshold [45]. The malolactic fermentation resulted in a decrease in ethyl acetate and isoamyl acetate ( Table 3). It should be pointed that the wines treated with the L. plantarum strains maintained a higher content of these acetate esters than those with O. oeni strains. Even though a concentration reduction in acetate esters occurred after the malolactic fermentation, they were still higher than their odor thresholds, thus significantly affecting the overall aroma of wine.
The Cabernet Gernischt wine contained total ethyl esters of 6.49 mg/L before the malolactic fermentation (Table 3). A dramatic content increase occurred after the malolactic fermentation by these lactic acid bacteria strains. The sample with C8-1 exhibited the greatest accumulation in ethyl esters (120.76 mg/L). Regarding the individual ethyl ester, ethyl lactate, a volatile with fruity and buttery notes [45], experienced a significant increase after the malolactic fermentation, and its level was found to be 116.95 mg/L in the sample with C8-1. By contrast, ethyl butanoate had suprathreshold concentration before the malolactic fermentation, which contributed to banana, pineapple, and strawberry scents [45] (Table 3). However, the malolactic fermentation via all these strains significantly reduced its concentration below the odor threshold. This indicated that its aromatic contribution was no longer effective then. The ethyl nonanoate and diethyl succinate were also affected by the malolactic fermentation in terms of their concentrations. However, these volatiles exhibited the limited aromatic contribution to the wine due to their low OAVs. Regarding different strains, the L. plantarum strains delayed the content decrease of ethyl butanoate, ethyl hexanoate, ethyl octanoate, and ethyl isopentyl succinate in the wine. Meanwhile, ethyl lactate and diethyl succinate appeared to be accumulated faster with the L. plantarum strains. It was worth noting that 2hydroxyisovaleric acid ethyl ester was only quantified in the wine with C8-1 strain. Among other ester volatiles, the wines fermented by L. plantarum strains showed more other esters than those with O. oeni strains. Isoamyl hexanoate significantly contributed its fruity, banana, apple, pineapple, and green scents (Table 3) [55], which was found to present a higher concentration in L. plantarum fermented wine samples. Both isoamyl lactate and methyl salicylate increased their concentrations two times after the malolactic fermentation.
It has been reported that both O. oeni and L. plantarum strains could release alcohol acyltransferase and reverse esterase [67]. These enzymes cleave alcohols to form esters in wine, and their activities were strain-dependent [35,[67][68][69][70][71]. Esterase is another critical enzyme released by lactic acid bacteria to affect the metabolism of esters during wine malolactic fermentation [19]. Lactic acid produced by malolactic fermentation could be further esterified under the activity of esterase to yield lactate-related esters, such as ethyl lactate and isoamyl lactate [69,71]. This could explain the significant accumulation of lactate-related esters in the present study. In our previous work, we noticed that the malolactic fermentation by C8-1 exhibited the longest duration (14 days for C8-1, 10 days for Lp39, 4 days for Oenos, and 6 days for CiNe), which potentially extended the esterification of lactic acid. This resulted in a greater accumulation of ethyl lactate and isoamyl lactate in the wine with the C8-1 strain (Table 3). In addition, esterase could also exhibit the capacity of hydrolyzing esters in wine during malolactic fermentation [19]. In the present study, isoamyl acetate, ethyl butanoate, ethyl hexanoate, and ethyl octanoate decreased their concentrations in the wine by all these strains (Table 3). Our results were consistent with the previously reported studies [13,35]. It should be noted that the L plantarum strains had higher concentrations of these esters than those with the O. oeni strains, indicating that the L. plantarum could delay the aromatic loss in wine after malolactic fermentation.

Higher Alcohols
Wine alcoholic fermentation plays a vital role in generating higher alcohols [72]. Higher alcohols have been reported to contribute the desirable complexity when their concentration was below 300 mg/L; when exceeding 400 mg/L, they negatively affect the aromatic quality of wine [29]. In the present study, the malolactic fermentation did not significantly alter the level of higher alcohols (Table 3). Isoamyl alcohol, 2-phenylethanol, 1-hexanol, and isobutanol were found to be the dominant higher alcohols in all these samples. Isoamyl alcohol was described as the solvent, sweet, alcohol, and nail polish notes [55], whereas 2-phenylethanol possessed rose and honey scents [45]. 1-Hexanol could provide the herbaceous, grass, and woody flavors [45]. The flavor scents of these higher alcohols contributed to the overall aroma due to their high OAVs (Table 4). It should be noted that although 3-methylpentanol, (Z)-3-hexen-1-ol, 2,3-butanediol, and benzyl alcohol also exhibited high concentrations in these wine samples, they played quite limited roles in the overall aroma due to their high odor thresholds. Among strains, two L. plantarum strains were found to benefit the accumulation of (E)-3-hexen-1-ol, 2-nonanol and 2,3-butanediol after the malolactic fermentation (Table 3). It has been reported that odorless glycosidic precursors in wine could be cleaved into aromatic aglycones, such as 1-hexanol and (E)-2-hexen-1-ol, with L. plantarum during malolactic fermentation [9]. We speculated that the accumulation of (E)-3-hexen-1-ol and 1-hexanol in our study mainly resulted from the hydrolysis of their precursors.

Aldehydes/Ketones
The oxidation of unsaturated fatty acids in wine could lead to the formation of aldehydes. During the biosynthesis of fatty acids, some activated fatty acids could be condensed to yield ketones [73]. In the present study, acetaldehyde was found to be the major aldehyde/ketone before and after the malolactic fermentation (Table 3). An increase of 2,6-dimethyl-4-heptanone and 3,4dimethylbenzaldehyde occurred in the wine after the malolactic fermentation. The accumulation of 3,4-dimethylbenzaldehyde was significantly greater in wine with Oenos. It should be noted that these aldehydes/ketones could exhibit limited aromatic contribution due to their sub-threshold concentrations.

Fatty Acids
Alcoholic fermentation plays a primary role in yielding fatty acids in wine [74]. Aliphatic acids are formed during the yeast fermentation, whereas the aldehyde oxidation could lead to the formation of branched-chain fatty acids [73]. n-Decanoic acid was found to be the dominant fatty acid in the wine before the malolactic fermentation in this study (Table 3), while the rest of the fatty acids exhibited a moderate content in the wine. Butyric acid, hexanoic acid, and octanoic acid were found to have concentrations higher than their odor thresholds, indicating that their featured aromas (cheese and fatty notes from butyric acid and hexanoic acid; rancid, cheese, and fatty acid scents from octanoic acid) were incorporated into the wine [45,62]. The malolactic fermentation did not significantly alter the level of fatty acids in all the wine samples. Only a significant content reduction of octanoic acid was found in the wine with CiNe. It has been reported that lipase released from lactic acid bacteria could metabolize lipids, which elevates the concentration of volatile fatty acids in wine [75]. For instance, octanoic acid has been reported to increase after malolactic fermentation [42,76,77]. However, some studies reported that malolactic fermentation could reduce octanoic acid in wine [3,34,76]. The wine fermented by CiNe strain had a decrease level of octanoic acid, which was similar as these reports. We speculate that different wine matrix could regulate the secretion of lipase from lactic acid bacteria, which further alter the metabolism of fatty acids in wine [3,76].

Terpene Derivatives
Terpenes are important grape secondary metabolites with distinctive varietal aroma profiles. They are extracted into wine during fermentation [72]. In this study, the malolactic fermentation did not significantly alter the total terpene derivatives (Table 3). In terms of the individual terpene derivative, α-ionone, β-damascenone, and β-ionone were found to be the key components contributing their flavors notes, because of their suprathreshold concentrations before malolactic fermentation. A significant decrease in β-ionone occurred after the malolactic fermentation. However, the wine with the L. plantarum strains exhibited higher content of β-ionone than those with O. oeni strains. Our result was as opposite to a published study where β-ionone was accumulated after malolactic fermentation by O. oeni or L. plantarum [3]. It has been accepted that oxidative stress in wine could reduce the level of terpenes [78,79]. Meanwhile, L. plantarum strains have been reported to enhance the accumulation of acetaldehydes during malolactic fermentation [43]. Therefore, we speculated acetaldehydes produced in samples with L. plantarum strains could effectively reduce the oxidative stress and thus delayed the oxidation of β-ionone. Additionally, O. oeni strains have been reported to release β-glucosidase, and this enzyme could hydrolyze the precursor of linalool and αterpineol to yield these two terpene derivatives in wine malolactic fermentation [4,32,33,[80][81][82][83]. In the present study, a significant accumulation of linalool and α-terpineol was observed in the samples with O. oeni strains (Table 3). On the contrary, the L. plantarum strains in the present study did not exhibit a significant release of linalool and α-terpineol, indicating that both strains might lack βglucosidase due to either their genes or the matrix stress during malolactic fermentation [17,18]. It should be noted that linalool and α-terpineol exhibited their concentrations below odor thresholds, limiting their flavor contributions to the wine overall aroma.

Others
Dihydro-2-methyl-3(2H)-thiophenone appeared to be the dominant other volatile before the malolactic fermentation (Table 3). The malolactic fermentation reduced its level, and the wine fermented with O. oeni strains exhibited lower content compared to those with L. plantarum strains. A significant increase on 4-ethylphenol was found in the wine with C8-1. It has been reported that 4hydroxycinnamic acid could be decarboxylated to 4-ethylphenol by lactic acid bacteria, and 4ehtylphenol could provide wine with a phenolic off-flavor [84][85][86]. In the present study, the C8-1 strain fermented wine did not contain detectable 3-hydroxycinnamic acid (Table 1), which indicated that a thorough conversion from 4-hydroxycinnamic acid to 4-ehtylphenol might take place in the wine due to the enzymes released by the C8-1 strain. 2,3-Dihydrobenzofuran was found to disappear in these wine samples after the malolactic fermentation (except C8-1). The flavor contribution from all these other volatiles was limited due to their low concentrations.

Aroma Profile
A total of 17 volatile compounds were found to significantly contribute their flavor notes to the overall aroma in the Cabernet Gernischt wine due to their high odor activity values ( Table 4). The wine before the malolactic fermentation primarily featured fatty, floral, caramel, and fruity aromas. The malolactic fermentation resulted in a modification of most of the aromatic features except the chemical and fatty (Figure 1). A decrease in the fruity aroma was observed for both strains, but a greater decrease observed for O. oeni strains than the L. plantarum strains. In addition, the wine with O. oeni strains showed a significant reduction in the intensity of floral aroma, whereas the L. plantarum strains retained a similar floral intensity as the wine before the malolactic fermentation. It should also be noted that the malolactic fermentation by the O. oeni strains also caused a decrease in the intensity of the herbaceous and balsamic series. The C8-1 strain fermented wine exhibited a similar balsamic aromatic strength as the wine before the malolactic fermentation. These indicated that malolactic fermentation using the L. plantarum strains could better preserve the aromatic feature in wine than the O. oeni strains.

Fruity
Floral HerbaceousCaramel Balsamic Chemical Fatty  Table 4. Different letters within the same series indicate significant difference at p < 0.05.

Principal Component Analysis
Principal component analysis was performed to characterize the overall quality of these wine samples by including phenolic compounds, biogenic amines, and volatile compounds as the variables (Figure 2a). The first and second components represented 50.94% and 26.44% of the total variance, respectively. More variables contributed to the positive F1 rather than negative F1. Most phenolic compounds appeared to correlate with the positive F1 and negative F2, while no specific distribution was found in biogenic amines. The Cabernet Gernischt wine before malolactic fermentation was positioned at the positive side of the F1, while the malolactic fermentation resulted in a negative shift in F1 for all strains. Strain-specific features were observed because the samples with L. plantarum strains moved to the positive axis of F2, while the samples with O. oeni assembled in its negative axis. Similar spatial distances were found between two L. plantarum strains and two O. oeni strains.
Specific volatile compounds with significant contribution to aroma profile were separately analyzed in Figure 2b. The first two components represented 68.58% and 15.60% of the total variance, respectively. The strains and variables showed similar clustering and positioning as Figure 2a. Most volatiles pointed to the positive F1, while only two esters to the negative F1. The balsamic, fruity, and caramel series with ethyl acetate, ethyl heptanoate, octanoic acid, ethyl butanoate, ethyl hexanoate, isoamyl acetate, and methyl salicylate differentiated the samples before or after malolactic fermentation. The herbaceous and floral series with 2-phenylethanol, 1-hexanol, β-ionone, isoamyl hexanoate, and ethyl lactate better distinguished the strains' effects from L. plantarum and O. oeni.

Conclusions
In conclusion, the malolactic fermentation resulted in a decrease in the content of the total flavanols and total flavonols in wine, whereas an increase in the phenolic acids content was found in all the samples after malolactic fermentation. Cabernet Gernischt wine after malolactic fermentation with the L. plantarum strains exhibited a lower content of biogenic amines than those with the O. oeni strains. Malolactic fermentation using these strains also resulted in a significant increase in the total esters content. Modifications in the fruity, floral, herbaceous, caramel, and balsamic aroma series in Cabernet Gernischt wine was observed after malolactic fermentation. Compared to the O. oeni strains, wine fermented with the L. plantarum strains exhibited a better preservation of its aromatic features. Principal component analysis indicated that malolactic fermentation by different strains could alter the overall quality of wine towards specific directions. Some key volatile components either indicated the effects from malolactic fermentation or differentiated the specific strains.