Preparation and Application of Directed Vat Set Indigenous Freeze-Drying Lentilactobacillus hilgardii Q19 Starter in Winemaking

In order to prepare a better direct vat set for malolactic fermentation (MLF) in high ethanol and low pH wines, the high-ethanol- and low-temperature-tolerant strain Lentilactobacillus hilgardii Q19, which was isolated from the eastern foothill of the Helan Mountain wine region in China, was used to prepare a direct vat set by vacuum freeze-drying. A superior freeze-dried lyoprotectant was obtained to create the starting culture by selecting, combining, and optimizing numerous lyoprotectants with higher protection for Q19 by using a single-factor experiment and response surface approach. Finally, the Lentilactobacillus hilgardii Q19 direct vat set was inoculated in Cabernet Sauvignon wine to carry out MLF on a pilot scale, with commercial starter culture Oeno1 as control. The volatile compounds, biogenic amines, and ethyl carbamate content were analyzed. The results showed that a combination of 8.5 g/100 mL skimmed milk powder, 14.5 g/100 mL yeast extract powder, and 6.0 g/100 mL sodium hydrogen glutamate offered better protection; with this lyoprotectant, there were (4.36 ± 0.34) × 1011 CFU/g cells after freeze-drying, and it showed an excellent ability to degrade L-malic acid and could successfully finish MLF. In addition, in terms of aroma and wine safety, compared with Oeno1, the quantity and complexity of volatile compounds were increased after MLF, and biogenic amines and ethyl carbamate were produced less during MLF. We conclude that the Lentilactobacillus hilgardii Q19 direct vat set could be applied as a new MLF starter culture in high-ethanol wines.


Introduction
Malolactic fermentation (MLF) is the process by which lactic acid bacteria (LAB) decarboxylate L-malate into L-lactic acid and release CO 2 . It usually occurs after the completion of alcohol fermentation (AF). This process reduces the acidity of the wine and improves the complexity of volatile compounds, the wine's microbial stability, and quality [1,2]. Spontaneously initiated MLF, which depends on the bacteria present in the wine environment, is usually challenging to control and cannot even start MLF in some cases [3,4]. Commercial LAB is primarily used for MLF in wines, also known as a directed vat set (DVS). DVS refers to a highly concentrated and standardized starter culture that has been prepared by vacuum-packing after the target strain has been cultured at high density, cultured, isolated, mixed well with protective agents, and vacuum freeze-dried. [5]. It is widely used in MLF because of its advantages, such as convenient inoculation, high viable count, easily controlled fermentation process, and long shelf life. Recently, several emerging drying technologies have been developed for the production of lactic acid bacteria DVS starters, such as low-temperature spray-drying technology [6], freezing granulation drying

Vacuum Freeze-Drying
The microbial suspension was prepared by mixing protective agents with each microbial strain in a 2:1 ratio. This suspension was stored in a refrigerator at 4 • C for 2 h for pre-freezing, producing cold-stress proteins to improve survival [19,20]. The microbial suspension was then placed in a freeze-drying dish with a thickness of 1.5 ± 0.1 cm and later freeze-dried using a vacuum freeze-dryer. The bacteria suspension was pre-frozen at −45 • C for 4 h to ensure that the suspension was in a solid state, and then the temperature was increased. The vacuum was pumped for sublimation drying and resolution drying to remove the free water and bound water in the suspension. The heating rate was maintained at 5 • C/min during this process to reduce the damage to the suspension caused by the formation of ice crystals, and the vacuum level at the sublimation stage was maintained at 0.2-0.4 mbar to improve the efficiency of sublimation drying [21,22]. Until the surface of the bacterium cake cracked, after which it was squeezed to form a powder, to gain lyophilized bacterium powder, this process needed to be kept sterile. Thus, a lyophilizer was used under UV conditions for more than 30 min after wiping with 75% alcohol before use.

Microbial Cell Count
The freeze-dried starter culture was redissolved in sterile water according to the original volume and kept at room temperature for 20 min. Then, a suitable gradient of the diluted bacterial solution was spread on culture media and incubated at 28 • C for 48 h. Finally, we counted the plates after they had grown bacteria. The experiment was repeated twice with three parallel sets for each experiment. The freeze-dried survival rate of L. hilgardii Q19 was calculated using the following formula: survival rate/% = (N f /N 0 ) × 100 (1) Note: N f represents bacterial survival after freeze-drying (CFU/mL), and N 0 represents bacterial survival before freeze-drying (CFU/mL).
The optimal concentration screening was performed for the screened skimmed milk powder and yeast extract powder concentrations and were set to 0, 5, 10, 15, and 20 g/100 mL. L-Glutamic acid sodium salt concentrations were set to 0, 5, 7.5, 10 and 12.5 g/100 mL. The central point of the response surface was determined on this basis, and the experiment was repeated three times during the freeze-drying process.

Response Surface Optimization Experimental Design
Based on the results of the single-factor experiment, the contents of skimmed milk powder, yeast extract powder, and L-Glutamic acid sodium salt were selected as independent variables, and the freeze-drying survival rate of L. hilgardii Q19 was taken as the response value. The Box-Behnken Design method in Design-Expert.V.8 was used to conduct response surface analysis. The experimental factors and levels are shown in Table 1. In order to explore the applicability of L. hilgardii Q19 starter in the actual production of wine, the starter was inoculated to the end of AF, and the wine without MLF was subjected to pilot-scale and low-temperature-and high-alcohol-resistant pilot-scaled fermentation experiments. The fundamental indexes of wine used for MLF are listed in Table 2. Fermentation agent: Q19 starter culture; viable bacteria count was (4.36 ± 0.34) × 10 11 CFU/g. Oeno1 (O. oeni) was used as the control agent with a viable count of (2.13 ± 0.19) × 10 11 CFU/g. Fermentation conditions: Q19 starter and Oeno1 starter were inoculated in the fermenter at 1 g/100 L, respectively, and the temperature was controlled for MLF.
MLF: Samples were taken every four days during the MLF process to detect the change in L-malic acid content. When the content reached 0.1 g/L, it was regarded as the end of fermentation. Potassium sulfite was added at a concentration of 60 mg/L to terminate the fermentation, and the contents of volatile components and biogenic amines were measured before and after fermentation.  [23]. The method is consistent with 2.4.

L-Malic Acid Determination
Wines without MLF were used for the MLF assay. After filtration with a 0.22 µm membrane, 1 g/100 L of starters were inoculated into the wines. Wines without MLF were used as controls. Samples were collected every four days to determine changes in L-malic acid content. L-malic acid was measured enzymatically with Analyzer Y15 (Biosystems, Food Quality, Barcelona, Spain).

Volatile Composition Analysis
Volatile components were determined by referring to the method of Bai [10] for GC−MS instrumentation. The volatile compounds of wines were identified using gas chromatography-mass spectrometry (HS-SPME-GC-MS) (Agilent 7890B gas chromatography in tandem with an Agilent 7000D mass spectrometer) (Agilent Technologies, Santa Clara, CA, USA) with an autosampler system (PALRSI 85) (CTC Analytics AG, Zwingen, Switzerland). The separation was performed on a DB-WAX capillary column (30 m × 0.25 mm × 0.25 µm, Agilent, USA). Next, 5 mL of wine, 1.5 g NaCl, and 10 µL 4methyl-2-pentanol (2.01 mg/L) were placed into a 20 mL headspace vial, and the headspace vial was placed in the autosampler. The temperature was kept at 45 • C for 5 min, followed by extraction at 45 • C for 35 min. The injection temperature was 230 • C, and the analysis was carried out for 8 min. The injection mode was splitless injection. The carrier gas was high-purity helium and the flow rate was 1 mL/min. The GC-MS temperature program was as follows: An initial temperature of 40 • C was maintained for 5 min and increased by 3 • C/min to 130 • C/min, followed by an increase to 144 • C at a rate of 2 • C/min, then 5 • C/min to 240 • C, and then held for 10 min. The MSD transfer line heater was set to 240 • C. The temperature of the ion source was 230 • C. The mass detector was operated in full scan mode (m/z 40-300) with electron ionization (EI) mode at 70 eV. All compounds were analyzed using the NIST08 standard mass spectral library. 4-methyl-2-pentanol was used as an internal standard to calculate the relative content of compounds.

Biogenic Amine Test
According to the National Standard for Food Safety: Determination of Bioamines in Food (GB/T 5009.208-2016), biogenic amine content in wine before and after MLF was determined [24].

Statistical Analysis
The data were analyzed statistically by one-way analysis of variance (ANOVA) with SPSS Statistics 22 (Chicago, USA), and the LSD method was compared afterwards. Duncan's multiple range test compared means. Differences with p values <0.05 were considered statistically significant. Design-Expert.V.8 software (Beijing, China) was used for response surface analysis, and GraphPad Prism 9 and Origin 2021 software were used for mapping.

Selection of the Optimal Concentration of Protective Agents
This research used seven protective agents (skimmed milk powder, yeast extract powder, D-trehalose dihydrate, D-sorbitol, glycerol, sucrose, and L-Glutamic acid sodium salt) were set at two levels, and the protective effects of lyophilization were investigated by single-factor experiment. The high level was 1.5 times that of the low level. Sterile water was used as the control group. As shown in Table 3, the order of the protective effects of the seven protective agents on L. hilgardii Q19 was skimmed milk powder > L-Glutamic acid sodium salt > yeast extract powder > sucrose > D-trehalose dehydrate > D-sorbitol > glycerol > sterile water. The protective agent could provide physical support for bacteria and reduce biochemical damage caused by the freeze-drying process. Sterile water without a protective agent showed a low freeze-drying survival rate, so adding a protective agent significantly affected the freeze-drying survival rate of Q19. Among them, skimmed milk powder, yeast extract powder, and L-Glutamic acid sodium salt had the most apparent protective effect: the lyophilized survival rate of these three protective agents was above 60% at both high and low levels of concentration, so these three protective agents were selected for further analysis. Note: "*" indicates a significant effect on the results (p < 0.05); "**" indicates a highly significant effect on the results (p < 0.01); "***" indicates a highly significant effect on the results (p < 0.001). a , b , c , d , e , f and g indicate significance analysis. The populations were tested before and after freeze-drying and tested within the column of the freeze-drying survival rate.
There are different enumeration methods for viable LAB cells and the freeze-drying survival rate. Skimmed milk powder is a kind of polymer compound protective agent that can provide an excellent protective effect when used alone or separately [25]. The protective mechanism can be explained by the fact that during freeze-drying, whey protein can wrap around the surface of bacterial cells, protect the cell membrane, and fix freeze-drying enzymes [26]. The L-Glutamic acid sodium salt is an antioxidant for starters during storage. The mechanism of protection for lactic acid bacteria is the reaction between the amino group of sodium glutamate and the carboxyl group of the cell protein of the bacterium, which stabilizes the structure of the cell protein of the microorganism. It can retain more residual water to maintain cell-life activity [27]. Yeast extract powder was less involved in previous freeze-drying experiments. This study found that during vacuum freeze-drying, the surface of the cake with yeast immersion powder as the protective agent was easy to crack, which was conducive to the evaporation of the internal water of the bacterial solution. The yeast immersion powder also made it easy to freeze-dry the bacterial solution and maintained a high survival rate after freeze-drying. In addition, since the yeast extract powder was added to the MRS medium for Q19 culture and re-added as a protective agent before freeze-drying, both could adapt well and play a better protective role. This is similar to the study of Tymczyszyn et al. [28], who added sucrose into MRS medium when cultivating lactic acid bacteria to improve the protective effect of sucrose.
In conclusion, skimmed milk powder, yeast extract powder, and L-Glutamic acid sodium salt were selected for further single-factor experiments, and the central point of the response surface was selected.

The Response Surface
Through single-factor experiment, the effects of different concentrations of skimmed milk powder, yeast extract powder, and L-Glutamic acid sodium salt on the survival rate of lyophilized L. hilgardii Q19 were investigated, and the optimal concentrations of the three protective agents were determined. As shown in Figure 1, the three protective agents had significant effects on the freeze-drying survival rate of Q19 (p < 0.05). The maximum freeze-drying survival rate of L. hilgardii Q19 appeared when the amount of skimmed milk Foods 2023, 12, 1053 7 of 23 powder was 10 g/100 mL. The lyophilized survival rate of Q19 remained above 60% during the change of L-Glutamic acid sodium salt addition between 5, 7.5, 10, and 12.5 g/100 mL, indicating that L-Glutamic acid sodium salt had an excellent protective effect on the vacuum freeze-drying of Q19, which is consistent with the study of Dimitra et al. [21]. When the supplemental level of L-Glutamic acid sodium salt was 5 g/100 mL, the freeze-drying survival rate of Q19 reached the maximum; there was no statistically significant difference compared with 10 g/100 mL. The freeze-drying survival rate of Q19 increased with the dosage of yeast extract powder, and the maximum value was found when the dosage of yeast extract powder was 15 g/100 mL.
freeze-drying survival rate of L. hilgardii Q19 appeared when the amount of skimmed milk powder was 10 g/100 mL. The lyophilized survival rate of Q19 remained above 60% during the change of L-Glutamic acid sodium salt addition between 5, 7.5, 10, and 12.5 g/100 mL, indicating that L-Glutamic acid sodium salt had an excellent protective effect on the vacuum freeze-drying of Q19, which is consistent with the study of Dimitra et al. [21]. When the supplemental level of L-Glutamic acid sodium salt was 5 g/100 mL, the freezedrying survival rate of Q19 reached the maximum; there was no statistically significant difference compared with 10 g/100 mL. The freeze-drying survival rate of Q19 increased with the dosage of yeast extract powder, and the maximum value was found when the dosage of yeast extract powder was 15 g/100 mL.
According to comprehensive analysis, the content of Q19 in bacterial powder per unit mass decreased due to the increase in protective agent dosage. Therefore, considering the total cost and bacterial powder activity, the additional levels of skimmed milk powder, L-Glutamic acid sodium salt, and yeast extract powder were determined to be 10, 5, and 15 g/100 mL, respectively, and this was taken as the central point of the response surface for response surface analysis.  According to comprehensive analysis, the content of Q19 in bacterial powder per unit mass decreased due to the increase in protective agent dosage. Therefore, considering the total cost and bacterial powder activity, the additional levels of skimmed milk powder, L-Glutamic acid sodium salt, and yeast extract powder were determined to be 10, 5, and 15 g/100 mL, respectively, and this was taken as the central point of the response surface for response surface analysis.

Response Surface Experiment and Result Analysis
The result of the Box-Behnken experiment showed in Table 4 , and the results of variance analysis are shown in Table 5. Note: "*" indicates a significant effect on the results (p < 0.05); "**" indicates a highly significant effect on the results (p < 0.01). Table 5 showed that the model established in this experiment was highly significant (p < 0.01); the misfit term was not significant (p > 0.05), so the reliability of the model was high. From the primary term of the F-value, it can be seen that skimmed milk powder, yeast extract powder, and L-Glutamic acid sodium salt had significant effects on the lyophilization survival rate of Q19, with L-glutamate having a highly significant effect. From the F-value interaction term, the effect on the lyophilization survival rate of Q19 was highly significant when skimmed milk powder and L-Glutamic acid sodium salt were combined. Furthermore, R 2 = 0.9759 indicated that the fit between the predicted and true values was good. The lower coefficient of variation (CV) indicates the higher reliability of the experiment, and the coefficient of variation in the test was 1.31%, proving that the experimental results were credible. Collectively, the model can target the lyophilization survival rate of Q19 for prediction.
Response surface plots and contour plots can show the interaction between two factors; steeper response surfaces indicate that the interaction between the two factors has a greater degree of influence on the response value, and contour line that are closer to the ellipse indicate that the interaction between the two factors is more significant [29]. The response surface and contours obtained from the experiment are shown in Figure 2. When the L-Glutamic acid sodium salt was kept constant at 5 g/100 mL, the lyophilized survival rate of Q19 showed a trend of increasing and then decreasing as the amount of skimmed milk powder and yeast extract powder was increased. When the amount of skimmed milk powder was between 8.42 and 10.7 g/100 mL, and the amount of yeast extract powder was between 13.35 and 16 g/100 mL, the contours showed a closed curve, indicating a maximum value in this interval. When the amount of yeast extract powder was kept constant at 15 g/100 mL, the lyophilization survival rate of Q19 showed a trend of increasing and then decreasing with the amount of skimmed milk powder and L-Glutamic acid sodium salt. The optimal value occurred when the amount of skimmed milk powder was 8.0-10.9 g/100 mL and L-Glutamic acid sodium salt was 4.7-6.5 g/100 mL. The contour lines showed an elliptical shape, and the steep response surface indicated the significant interaction between skimmed milk powder and L-Glutamic acid sodium salt. When the amount of skimmed milk powder was held constant at 10 g/100 mL, the lyophilized survival rate of Q19 increased and then decreased with the addition of yeast extract powder and monosodium L-Glutamic acid sodium salt, with optimal values appearing in the contour closures of 13.04-16.5 g/100 mL for yeast extract powder and 4.72-6.5 g/100 mL for monosodium L-Glutamic acid sodium salt. The analysis results in Figure 2 were consistent with the significance in Table 5.

Validation Test
The results were brought into the model for statistical analysis using Design-Expert.V8 software, and the maximum survival rate occurred when 8.62 g/100 mL skimmed milk powder, 14.51 g/100 mL yeast extract powder, and 6.09 g/100 mL L-Glutamic acid sodium salt were used as the protective agent formulation. The predicted survival rate of lyophilized Q19 in this case was 87.58%. Three lyophilization replicate experiments were carried out to obtain the actual lyophilization survival rate of Q19 as 87.85 ± 1.19%, and the obtained results were close to the predicted values. The protective agent formulation was optimized to 8.5 g/100 mL skimmed milk powder, 14.5 g/100 mL yeast extract powder, and 6 g/100 mL L-Glutamic acid sodium salt. The results illustrated the reliability of the experimental results: in this case, the number of live bacteria in the Q19 starter culture was (4.36 ± 0.34) × 10 11 CFU/g.

Q19 Starter Storage Performance
The ability of starter cultures to be stored for an extended period is a prerequisite for commercialization. Table 6 shows the change of viable LAB cells of the Q19 starter during storage. At storage temperatures of 4 • C and −20 • C, the number of viable LAB cells shows a decreasing trend with the increase in storage time. Among them, the number of viable bacteria decreased slowly at −20 • C, and the survival rate remained above 70% after 180 d of storage, which is consistent with the findings of Ren et al. [30]. This showed that the Q19 starter could maintain good activity at −20 • C.
Note: "*" indicates a significant effect on the results (p < 0.05); "**" indicates a highly significant effect on the results (p < 0.01).   3.5. Pilot-Scale MLF 3.5.1. L-Malic Acid Conversion Rate during MLF L-malic acid content changes during MLF are a crucial sign of whether or not MLF has begun. L-malic acid was examined in the process of MLF of Cabernet Franc and Yan 73 to evaluate the degradation rate of L-malic acid by two lactic acid fermenters, Q19 and Oeno1, and the results are shown in Figure 3. Along with the access of Q19 and Oeno1, the L-malic acid content showed a significant decreasing trend, indicating that the accessed Q19 and Oeno1 played a major MLF role in the MLF process. In addition, wine samples inoculated with Oeno1 completed MLF after 24 d, and those inoculated with Q19 completed MLF after 34 d. Oeno1 was able to complete MLF faster compared to Q19. of storage, which is consistent with the findings of Ren et al. [30]. This showed that the Q19 starter could maintain good activity at −20 °C.

Pilot-Scale MLF
3.5.1. L-Malic Acid Conversion Rate during MLF L-malic acid content changes during MLF are a crucial sign of whether or not MLF has begun. L-malic acid was examined in the process of MLF of Cabernet Franc and Yan 73 to evaluate the degradation rate of L-malic acid by two lactic acid fermenters, Q19 and Oeno1, and the results are shown in Figure 3. Along with the access of Q19 and Oeno1, the L-malic acid content showed a significant decreasing trend, indicating that the accessed Q19 and Oeno1 played a major MLF role in the MLF process. In addition, wine samples inoculated with Oeno1 completed MLF after 24 d, and those inoculated with Q19 completed MLF after 34 d. Oeno1 was able to complete MLF faster compared to Q19.

Volatile Components Analysis
The volatile components produced by Q19 and control were analyzed after MLF, and the results showed in Table A1. Forty-nine volatile substances were detected in the wine samples without MLF, totaling 7.4693 mg/L. Inoculated with Q19, 66 volatile aroma substances were detected after MLF, including 15 alcohols, 27 esters, 14 acids, 5 terpenes, 3 aldehydes and ketones, and 2 others, with a total content of 9.8493 mg/L. In the wine samples inoculated with control, 55 volatile aroma substances were detected after MLF, including 14 alcohols, 21 esters, 10 acids, 4 terpenes, 2 aldehydes and ketones, and 4 others, with a total content of 7.2088 mg/L. Compared to non-MLF, inoculation of LAB with MLF enhanced the type and content of terpenoids; the wine samples inoculated with Q19 produced more types and content of esters, acids, and terpenoids compared to the control.
The alcohol concentration was 1.101 mg/L in the wines that were inoculated with Q19 after MLF, higher than the control. Alcohols are mainly produced from the metabolism of amino acids during AF. Higher alcohols can significantly contribute to a wine's aromatic

Volatile Components Analysis
The volatile components produced by Q19 and control were analyzed after MLF, and the results showed in Table A1. Forty-nine volatile substances were detected in the wine samples without MLF, totaling 7.4693 mg/L. Inoculated with Q19, 66 volatile aroma substances were detected after MLF, including 15 alcohols, 27 esters, 14 acids, 5 terpenes, 3 aldehydes and ketones, and 2 others, with a total content of 9.8493 mg/L. In the wine samples inoculated with control, 55 volatile aroma substances were detected after MLF, including 14 alcohols, 21 esters, 10 acids, 4 terpenes, 2 aldehydes and ketones, and 4 others, with a total content of 7.2088 mg/L. Compared to non-MLF, inoculation of LAB with MLF enhanced the type and content of terpenoids; the wine samples inoculated with Q19 produced more types and content of esters, acids, and terpenoids compared to the control.
The alcohol concentration was 1.101 mg/L in the wines that were inoculated with Q19 after MLF, higher than the control. Alcohols are mainly produced from the metabolism of amino acids during AF. Higher alcohols can significantly contribute to a wine's aromatic profile, adding fruit and floral aromas and aromatic complexity. However, when the content of higher alcohols is higher than 400 mg/L, they can produce a pungent and unpleasant aromatic profile that can negatively affect the wine aroma [31]. According to Table A1, the type and concentration of esters in the wine samples inoculated with Q19 were significantly higher than those in control. In contrast to the control, the wine samples inoculated with Q19 produced isoamyl decanoate, ethyl myristate, and butyrolactone, which gave the wine distinct aromas like rose, cinnamon, butter, and caramel. Esters have a more significant impact on wine aroma than alcohols and usually impart floral and fruity aromas to wines. However, when their contents are too high, they can mask the varietal aromas and reduce the aromatic complexity of the wine, and their concentration decreases during aging due to chemical hydrolysis.

PCA Analysis of Different Fermentation Starter
In order to better distinguish the differences in volatile components between control and Q19 after MLF, principal component analysis (PCA) analysis was performed on all 41 volatile aroma components with volatile components greater than 10 µg/L using Origin 2021 software. The obtained score plot (A) and loadings plot (B) are shown in Figure 4. The cumulative variance contribution of the two extracted principal components, PC1 (40.2%) and PC2 (26.6%), was 66.8%. The experimental groups had different distributions in Figure  A due to the different inoculation of starters. The distribution of wine samples inoculated with Q19 was mainly in the positive direction of PC1, whereas the distribution of wine samples inoculated with Oeno1 was predominantly in the negative direction of PC1. The upper right of Figure B had the most volatile aroma substances, including six esters, six acids, four alcohols, and three terpenes, which give the wine its unique floral, fruity, honey, and other characteristic aromas associated with the fermentation of Q19 in Figure A. The oleic acid, pentadecanoic acid, 4-octanone, and n-noctyl ether in the lower left part of Figure B correlated with the Oeno1 fermentation in Figure A. Collectively, the wine samples inoculated with Q19 produced more volatile aroma substances than the control.

Wine Safety
To evaluate the safety of the wine samples after Q19 inoculation, the concentration of eight biogenic amines was evaluated in the wine samples before and after fermentation; the findings are shown in Table 7. It can be seen that β-phenylethylamine, putrescine, histamine, and tyramine contents were reduced in the wine samples inoculated with Q19 compared to the control. The histamine content was 6.38 mg/L, which was lower than the limits for histamine content in France, Australia, and other countries [32]. In addition, wine samples inoculated with Q19 produced spermine after MLF, but the total biogenic

Wine Safety
To evaluate the safety of the wine samples after Q19 inoculation, the concentration of eight biogenic amines was evaluated in the wine samples before and after fermentation; the findings are shown in Table 7. It can be seen that β-phenylethylamine, putrescine, histamine, and tyramine contents were reduced in the wine samples inoculated with Q19 compared to the control. The histamine content was 6.38 mg/L, which was lower than the limits for histamine content in France, Australia, and other countries [32]. In addition, wine samples inoculated with Q19 produced spermine after MLF, but the total biogenic amine content was less than the control. On balance, the Q19 fermenter produced fewer biogenic amines and was more in line with the current demand for "green" LAB starters. Note: a , b and c indicate significance analysis, and "ND" indicates that the biogenic amine component was not detected. Tested within column.

Analysis of Fermentation Performance in a Pilot Test for Low-Temperature and
High-Alcohol Resistance 3.6.1. Bacterial Density Variation during MLF As shown in Figure 5, inoculation with both Q19 and control resulted in good survival in the wine environment, with no significant order-of-magnitude decrease. Among them, Q19 maintained a live count of 10 5 CFU/mL throughout the MLF process. The wine samples inoculated with control maintained a stable live count of 10 6 CFU/mL from 0 to 20 days of fermentation. From 20 to 36 days, the viable control count showed a gradual increase, with the viable count growing from 10 6 CFU/mL to 10 7 CFU/mL; at 36 days to the end of MLF, the viable count of the control began to stabilize and did not show an increasing trend.
in the wine environment, with no significant order-of-magnitude decrease. Among them, Q19 maintained a live count of 10 5 CFU/mL throughout the MLF process. The wine samples inoculated with control maintained a stable live count of 10 6 CFU/mL from 0 to 20 days of fermentation. From 20 to 36 days, the viable control count showed a gradual increase, with the viable count growing from 10 6 CFU/mL to 10 7 CFU/mL; at 36 days to the end of MLF, the viable count of the control began to stabilize and did not show an increasing trend.

L-Malic Acid Changes during MLF
As shown in Figure 6, L-malic acid showed a slow decline followed by a linear decline after LAB inoculation, and there were some differences in the L-malic acid degradation rate between Q19 and the control rate. The comprehensive analysis showed that the MLF of Q19 started more slowly, and the MLF cycle was longer than the control.

L-Malic Acid Changes during MLF
As shown in Figure 6, L-malic acid showed a slow decline followed by a linear decline after LAB inoculation, and there were some differences in the L-malic acid degradation rate between Q19 and the control rate. The comprehensive analysis showed that the MLF of Q19 started more slowly, and the MLF cycle was longer than the control.

Analysis of Volatile Aroma Components in Wine
The volatile components of each wine sample were measured before and after MLF fermentation to compare and evaluate the effects of the two starters on the wine aroma substances. The results are shown in Table A2: 45 volatile compounds were detected before MLF with a total content of 3.6826 mg/L. A total of 49 volatile compounds were detected in the wine samples inoculated with Q19 after MLF, with a total content of 4.304 mg/L, including 12 alcohols, 21 esters, 8 acids, 3 terpenoids, and 3 hydroxyl compounds. A total of 47 volatile compounds were detected in wine samples inoculated with control, with a total content of 4.9642 mg/L, including 14 alcohols, 20 esters, 5 acids, 3 terpenes, and 3 aldehydes and ketones. After MLF, there was a significant increase in the type and content of esters, with Q19 producing a greater variety of esters, but the control produced a greater content of esters.

Wine Safety
Performing MLF at low temperatures results in longer cycles and may cause elevated biogenic amine levels. As shown in Table 8, five biogenic amines, tryptamine, β-phenylethylamine, cadaverine, histamine, and tyramine were detected before and after MLF. After the completion of MLF, the content of biogenic amines in wine samples inoculated with Q19 was 33.34 mg/L and that of wine samples inoculated with control was 40.17 mg/L. Compared with the non-MLF, the content of cadaveric amines, histamine, and tyramine in wine samples inoculated with Q19 showed different degrees of decrease, and

Analysis of Volatile Aroma Components in Wine
The volatile components of each wine sample were measured before and after MLF fermentation to compare and evaluate the effects of the two starters on the wine aroma substances. The results are shown in Table A2: 45 volatile compounds were detected before MLF with a total content of 3.6826 mg/L. A total of 49 volatile compounds were detected in the wine samples inoculated with Q19 after MLF, with a total content of 4.304 mg/L, including 12 alcohols, 21 esters, 8 acids, 3 terpenoids, and 3 hydroxyl compounds. A total of 47 volatile compounds were detected in wine samples inoculated with control, with a total content of 4.9642 mg/L, including 14 alcohols, 20 esters, 5 acids, 3 terpenes, and 3 aldehydes and ketones. After MLF, there was a significant increase in the type and content of esters, with Q19 producing a greater variety of esters, but the control produced a greater content of esters.

Wine Safety
Performing MLF at low temperatures results in longer cycles and may cause elevated biogenic amine levels. As shown in Table 8, five biogenic amines, tryptamine, β-phenylethylamine, cadaverine, histamine, and tyramine were detected before and after MLF. After the completion of MLF, the content of biogenic amines in wine samples inoculated with Q19 was 33.34 mg/L and that of wine samples inoculated with control was 40.17 mg/L. Compared with the non-MLF, the content of cadaveric amines, histamine, and tyramine in wine samples inoculated with Q19 showed different degrees of decrease, and the total biogenic amine content decreased by 2.32 mg/L. The wine samples inoculated with control showed different degrees of increase in tryptamine, βphenylethylamine, and tyramine, and the total biogenic amine content increased by 4.51 mg/L. In addition, no cadaverine, which has a high impact on health, was detected in the wine samples inoculated with Q19.

Discussion
The preparation of DVS starter cultures provides the conditions for the industrial application of desirable strains. However, there are few studies on MLF starters in China. Cheng et al. [33] used 10% skimmed milk powder, 13% sucrose, 2% sorbitol, and 0.8% tyrosine as freeze-drying protectants in the vacuum freeze-drying of L. plantarum L1, and the survival rate reached 97.4% after freeze-drying. The study showed that lyophilized milk powder, as a better protective agent, should be used in combination with other protective agents to achieve better protection. In this study, Q19 DVS starter was prepared using L. hilgardii Q19 with a formulation of 8.5 g/100 mL skimmed milk powder, 14.5 g/mL yeast extract powder, and 6.0 g/100 mL L-Glutamic acid sodium salt as the lyophilization protectant. The survival rate was 87.85% under these conditions, and the number of viable LAB cells reached 4.36 × 10 11 CFU/g. During the lyophilization process, it was found that higher or lower dosages of protective agents were not ideal for the protection of the bacteria, which was similar to the results of previous studies [34,35], in which both high osmotic pressure and rapid changes in osmotic pressure disrupted the cell membrane of the bacteria, thus reducing the survival rate. Other studies showed that the strain, growth conditions, freeze-drying technology, and protective effect of different protectants are the primary determinants of the survival rate of lyophilized strains [36]. In addition to the lyophilization protectants investigated in this work, the anti-freezing ability of the bacteria is important for the generation and preservation of highly active starters [37].
Seong Choi et al. [38] found that with the cold treatment of Lactobacillus brevis (L. brevis) WiKim0069 before vacuum freeze-drying, the expression of potential frozen surface layer protein (SLP) was promoted, increasing the storage time and survival rate of L. brevis WiKim0069 after freeze-drying. The ideal cold-treatment temperature of L. hilgardii Q19 before vacuum freeze-drying is a topic for future research to improve the survival and storage time of L. hilgardii Q19 after freeze-drying. In this study, the survival rate of L. hilgardii Q19 was 87.85% after freeze-drying, which may be related to anhydrobiosis. Related studies found that by rehydrating, anhydrobiosis can increase cellular energy generation and increase the capacity of self-repair in a hostile environment [39,40]. Wang et al. [41] found no significant change in survival with storage time when the starter was frozen and preserved at −80 • C. This indicated that the −80 • C environment is appropriate for preserving ferments and might be studied for future starter storage.
After MLF, the metabolic activity of LAB results in fuller-bodied wines with more intense flavors; this change shows the complexity of LAB in the metabolic process of wine [42]. Another study showed that Lactobacillus is more tolerant to the microbial wine environment. That Lactobacillus has more wine aroma-related enzyme genes than O. oeni and produces bacteriocins-which would enable them to combat LAB spoilage-implies that bacteria of the genus Lactobacillus have excellent potential for the development and application of future MLF starters [43,44]. During the pilot-scale MLF, the conversion rate of L-malic acid control was faster than that of Q19, which may be because O. oeni tends to be more tolerant to the wine-fermentation environment [45]. Other studies showed that O. oeni can form biofilms during MLF, and these biofilms can effectively compose MLF-tolerant cells and successfully complete MLF [46]. In the pilot-scale fermentation with low-temperature and high-alcohol-tolerance MLF, the growth of viable LAB cells in Q19 was slower than the control, and there was no significant order of magnitude increase. This is consistent with the finding of Zhang et al. [47] that suitable environmental conditions can significantly enhance the productivity of O. oeni cultures. The MLF time of Q19 was longer compared to the control, may be related to chaotropicity-mediated ethanol stress. Santos et al. [48] found that mycobacterium may have the highest ethanol tolerance so far reported for cells of bacteria or Archaea. In addition, some studies showed that chaotropic solute has an inhibitory effect on water bodies, soils, and biofuel-producing microorganisms [49,50]. Hallsworth et al. [51] found that ethanol-induced water stress eventually disrupted enzyme and membrane structure and function. Ethanol is mildly chaotropic at low concentrations and can reduce the stability of macromolecules [52]. The stress induced by ethanol can be counteracted by absorption and/or synthesis of compatible solutes (e.g., trehalose, glutamate), which can stabilize the macromolecular systems [53][54][55]. In addition, ethanol inhibits microbial metabolism by reducing water activity and causing stress [56,57]. Therefore, we speculated that it was related to no significant increase in the viable count of Q19 in the pilot-scale experiment. We will carry out more pilot-scale fermentation experiments in the future and discuss the results.
In summary, L. hilgardii Q19 was tolerant to various stresses in wine-fermentation environments. This enhanced its vitality, making it adaptable and competitive between species in a survival environment [58,59]. Terpenoids and esters play an important role in regulating wine aromas [60]. Yeast and LAB can produce esters and terpenes during winemaking, during which LAB release volatile glycosides from glucoside precursors to increase the terpene content of the wine [61,62]. In a comparative study of the winemaking characteristics of L. plantarum CX19 and O. oeni PN4 by Sun et al. [63], L. plantarum CX19 was more tolerant of high pH environments than O. oeni PN4, and wines fermented with L. plantarum CX19 had a more complex aroma composition and better taste. In this study, the pilot-scale fermentation revealed that after MLF, Q19 produced more types and levels of esters and terpenes compared to the control, indicating that the aroma composition of wine samples inoculated with Q19 was more intense and complex, similar to the results of the above study. Inoculated wine samples with Q19 produced unique volatile aroma substances such as butyrolactone, isobutyl n-octanoate, ethyl octyl succinate, and damascenone, which gave the wines floral and fruity aromas and preserved terroir characteristics while retaining indigenous microbial diversity. In the pilot-scale fermentation at low temperature and high alcoholic strength, the ester content of the fermented wine samples were all elevated after fermentation compared to non-MLF, in agreement with the finding of Iorizzo et al. [61]. Compared to the control, the wine samples inoculated with Q19 produced more terpenoids, which may be because Lactobacillus spp. strains carry more coenzyme genes related to aroma [43,44].
Another essential feature of MLF's LAB starter is that it does not produce biogenic amines during the winemaking process, which impacts the wine's safety; some biogenic amines (e.g., putrescine) can also affect the aroma of the wine [47]. Patrignani et al. [64] studied the concentration of biogenic amines in 160 bottles of Italian red wine, in which tyramine concentrations ranged from 1.58 to 10.19 mg/L and histamine concentrations ranged from 1.49 to 16.34 mg/L. Li et al. [65] studied the concentration of biogenic amines in 39 Chinese red wines and found that the maximum values of histamine, tyramine, and β-phenylethylamine were 10.51, 9.31, and 4.58 mg/L, respectively. While there are no clear limits on the acceptable amount of biogenic amines in wine in the EU, in this study, the pilot-scale fermentation wine samples inoculated with Q19 after MLF produced reduced levels of β-phenylethylamine, putrescine, histamine, and tyramine and produced less total biogenic amines compared to the control. In particular, histamine and tyramine, which have a major impact on human health, had contents of 6.38 and 2.1 mg/L, respectively, in the wine samples inoculated with Q19, which meet the current evaluation criteria for healthy wines.

Conclusions
In this study, the optimal formulation of L. hilgardii Q19 freeze-dried lyoprotectants was determined to be 8.5 g/100 mL skimmed milk powder, 14.5 g/100 mL yeast extract powder, and 6.0 g/100 mL L-Glutamic acid sodium salt, yielding a freeze-dried survival rate of 87.85 ± 1.19% and a live count of Q19 starter of 4.36 ± 0.34 × 10 11 CFU/g. Q19 starter degrades L-malic acid well in real production to suit the demands of wine MLF, and it can complete MLF under low-temperature and high-alcohol circumstances, solving the difficulties of low temperatures and high alcohol contents faced by wine MLF in China and some other production locations. After MLF, the wine samples inoculated with Q19 produced more types and levels of volatile aromatic substances, especially esters and terpenes, than the commercial control, with volatile aromatic substances such as acacia alcohol and damascenone, giving the wine unique aromatic characteristics such as honey, rose, and floral notes. In addition, the wine samples inoculated with Q19 produced lower levels of biogenic amines compared to the control. This conclusion provides a basis for preparing highly active Q19 starter and its future production, but further research is needed for L. hilgardii to carry enzyme genes related to characteristic aromas.  Note: a and b indicate significance analysis. Contents are recorded as 4-octanol. Different lowercase letters in the same row indicate significant differences (p < 0.05). "ND" indicates that the volatile component was not detected, and "\" indicates that the relevant literature was not consulted.