An Oenococcus oeni PSU-1 ATCC™ (Garvie, 1967) preculture was prepared from a frozen stock by inoculating 100-ml Erlenmeyer flasks containing an FT80 medium, at a pH of 4.8 (Guilloux-Benatier et al., 1995), and incubated at 28°C. In our previous studies, we observed that this pH allowed for the optimization of O. oeni’s growth under ethanol stress conditions. For this reason, we used this pH in the selection step (Contreras et al., 2018). FT80 was used because it imitates wine composition and allows us to evaluate the malolactic fermentation process. Growth was followed by measuring optical density at 600 nm, using a Spectronic 20™ spectrophotometer and an Infinite 200 PRO Microplate reader (Tecan, Switzerland). PSU-1 was called “parental strain” and the mutant strain obtained during this study was called “E1” and/or “mutant strain.”

Oenococcus oeni PSU-1 strain was grown in 10 mL of FT80 medium with a pH of 4.8, adding ethanol as a stressor. An ethanol concentration of 10, 15, or 20%v/v was used in the culture medium. Similarly, the strain was evaluated by adding potassium metabisulfite as a stressor at 50, 100, or 200 mg/L in the culture media, which was prepared in the moment. Every culture was inoculated with approximately 3.2 × 10⁸ UFC/mL cells to achieve an optical density of 0.2 in the flask. Cultures were sampled periodically, for 9 days. We considered growth when the cultures showed an optical density equal to or higher to 0.3.

The PSU-1 strain was grown at 28°C in 70 mL of FT80 medium to obtain an optical density of 0.2 (3.2 × 10⁸ UFC/mL). The cultures were centrifuged at 3000 rpm for 5 min, and cells were washed and suspended in 10 mL of 50 mM H₂HPO₄/K₂HPO₄ buffer (pH 7.5). Suspended cells were mixed with ethyl methane sulfonate (EMS) and incubated at 30°C, with a gentle agitation for 30 min. The PSU-1 cells were treated individually using 5% v/v of EMS, this addition corresponded to a 99.5% of lethality. At the end of the incubation process, the cell was treated using buffer sodium thiosulfate 10% w/v for 5 min. Later, they were centrifuged at 3,000 rpm for 3 min, the supernatant was eliminated, and cells were washed two times using buffer and distilled water. Finally, the cells were suspended in distilled water, and then 100 μL of suspension were plated on FT80 agar medium and incubated at 28°C for 24 h. Subsequently, around 10 colonies were taken per plate and were evaluated with a stressor using FT80.

The mutant strain selected (E1) was evaluated using synthetic wine called MaxOeno, with 15% v/v of ethanol, which was previously developed by our group (Contreras et al., 2018). MaxOeno is a fully defined culture medium, which enables the quantification of substrate consumption and metabolite production; pH was adjusted at 3.5 or 4.8 using KCl and HCl.

Mutant strain E1 was inoculated into 100 mL flasks containing 75 mL of MaxOeno at pH 3.5 or 4.8, in duplicate. The medium was inoculated at an optical density of 0.2 (at 600 nm) and the cultures were incubated at 25 ° C, without shaking. Samples were taken periodically for 10 days, determining growth (by optical density), malate consumption, and lactate production (by HPLC).

The red and white grape musts used were provided by the Cono Sur™ vineyard and obtained from the city of Chimbarongo, VI Region of Chile. Both were characterized in terms of pH, assimilable nitrogen, glucose, fructose, total SO₂, and free SO₂ obtained for each variety. The composition of the Chardonnay must was: glucose 100 g L⁻¹, fructose 100 g L⁻¹, pH 3.2; assimilable nitrogen 415 mg N L⁻¹. The composition of the Cabernet Sauvignon must was: glucose 100 g L⁻¹, fructose 100 g L⁻¹, pH 3.5, assimilable nitrogen 300 mg N L⁻¹.

Finally, the red and white grape musts showed 32 and 28 mg L⁻¹ of free SO₂, and also, 88 and 90 mg L⁻¹ of total SO₂, respectively. The total acidity was determined by alkalinity with a bromothymol blue indicator, obtaining 1.617 and 3.185 g L⁻¹ of total acidity (considering sulfuric acid) for red and white must, respectively. In addition, we determined in those musts a total acidity of 2.475 and 4.875 g L⁻¹ (considered as tartaric acid), respectively. The total acidity was determined by both methods, as described in the technical instructions for grape musts analysis (Resolución N° 8.542, 2013).

The wines were produced using EC1118 ™ (Lallemand) yeast, which was inoculated at a concentration of 1×10⁶ cells/mL in 700 mL of grape must, using 1-liter reactors.

The Chardonnay wine composition was glucose 2.3 g L⁻¹; fructose 2.3 g L⁻¹; pH 3.2; assimilable nitrogen 200 mg N L⁻¹, and 14% v/v of ethanol. The Cabernet Sauvignon wine composition was glucose 1.9 g L⁻¹, fructose 7.2 g L⁻¹, pH 3.5; assimilable nitrogen 150 mg N L⁻¹, and ethanol 15.5% v/v. In both cases, the alcoholic fermentation was concluded when less than 10 g L⁻¹ of residual sugar was detected.

Subsequently, malolactic fermentation was carried out, inoculating the E1 mutant strain or the PSU-1 strain in each wine produced. The parental strain was subjected to adaptation to ethanol and pH because it was not able to grow without this adaptation. For this, they were cultivated in MaxOeno defined synthetic wine in which the pH varied from 4.8 to 3.5, and the ethanol from 0 to 12%. The mutant strain E1 was inoculated directly. Both strains were inoculated at a 0.2 optic density (3.2×10⁸ cells/ml), in duplicate.

Samples were taken periodically to monitor their growth; the supernatants were frozen to determine the consumption and production of compounds.

Free and total SO₂ were determined by the Ripper’s method, as described by Schwarze and Edmundo (2000). Moreover, total acidity was obtained by titration, and 10 mL of grape must was placed in a 100 mL beaker, 3 drops of bromothymol blue indicator were added. Titration began with 0.1 N NaOH with constant stirring until the green indicator turned to a blue-green color.

On the other hand, pH was determined using a pH meter HI2221™ (Hanna instruments, United States). The presence of reducing sugars in the grape must was determined through a refractometer and a high-performance liquid chromatography (HPLC) with the LaChrom L-7000 HPLC system (Hitachi, Japan). The presence of reducing sugars in the wine was determined using only HPLC.

Finally, nitrogen concentration was determined by UHPLC–MS, using the Ultimate 3,000 system (ThermoScientific, United States). The compounds were separated through the RP-18 ion exchange column (LiChrospher®, Germany) as described by Contreras et al., 2018.

Samples obtained from each culture were centrifuged and the supernatant was collected, and an aliquot was injected in a Lachrom L-700 HPLC system (Hitachi, Japan) equipped with a Diode Array and a Refractive Index detector (Merck Hitachi, Japan). Organic acids, alcohols and sugars were separated using an Aminex HPX-87H ion exchange carbohydrate-organic acid column (Bio-Rad, United States), and quantified as described previously (Varela et al., 2003).

The gene expression profiles of the mutant strain and its parent were obtained using RNAseq. Massive transcriptome sequencing was performed using the HiSeq 2,500® technique, where 10 million reads per gene were considered for mapping. The transcriptome sequencing with ribosomal depletion, and the data analysis services were hired at the Molecular Research DNA LAB (Texas, United States).

Bacterial RNA was extracted using kit E.Z.N.A™ (Omega Biotek, United States) and treated with DNAsa I (Omega Biotek, United States), following the manufacturer’s instructions. All the samples were suspended on DEPC water and dried using RNAstable® tubes (Biomatrica, United States).

The concentration of total RNA was determined using the Qubit® RNA Assay Kit (Life Technologies); it was around 200 ng/μL of RNA in each case. RNA integrity value (RIN) was determined by using the Agilent RNA 6000 Nano Reagents and the RNA Nano Chips in Agilent 2,100 Bioanalyzer (Agilent Technologies), obtaining a RIN value of about 2. To remove DNA contamination 0.5–1.5 ug of total RNA was cleaned using Baseline-ZERO™ DNase (Epicentre) following the manufacturer’s instructions, followed by purification using the RNA Clean & Concentrator-5 columns (Zymo Research). DNA free of RNA samples were used for rRNA removal by using Ribo-Zero™ Magnetic Gold Kit (Epidemiology; Illumina). Final purification was performed using the RNA Clean & Concentrator-5 columns (Zymo Research). rRNA depleted samples were used for library preparation using the TruSeq™ RNA LT Sample Preparation Kit (Illumina) according to the manufacturer’s instructions. Following the library preparation, the final concentration of all the libraries were measured by using the Qubit® dsDNA HS Assay Kit (Life Technologies), and the average library size was determined using the Agilent 2,100 Bioanalyzer (Agilent Technologies), which was approximately 33 ng/μL. The libraries were then pooled in equimolar ratios of 2 nM, and 8pM of the library pool was clustered using the cBot (Illumina), and paired-end sequenced for 300 cycles using the HiSeq 2,500 system (Illumina).

Raw reads were normalized through RPKM (reads per kilo base per million reads) method using the Qseq program (Conesa et al., 2016). Publicly available transcriptome data sets were downloaded from the NCBI’s database and they were mapped using the QSeq program of DNASTAR. Fold change for each gene was calculated as:

All the information is in the NCBI database, GEO accession GSE217343 and token mberokakzfirnyl.¹

For cDNA preparation, 5 μg of RNA was incubated with 1 unit of DNase I at 37°C for 30 min. Afterwards, the instructions of the manufacturer for M-MLV were followed. The cDNA obtained was used as template in the real time PCR reaction (QPCR). The QPCR reaction was carried out in a final volume of 10 μL. The reaction mixture contained 5 μL of Brilliant II SYBR Green QPCR Master mix (Stratagene, United States), 0.2 μM of each primer (Supplementary Table S2; Supplemental material), and 10 ng/μL of cDNA. The QPCR reaction was carried out in an AriaMx Real-Time PCR equipment (Agilent Technologies, United States) under the following conditions: 50°C for 2 min, 95°C for 1 min, 45 cycles of 95°C for 3 s, 60°C for 10s, and 72°C for 20s, a melting analysis at 95°C for 30s, at 60°C for 30 s, and at 95°C for 30s. Quantification of relative gene expression was done by using the mathematic method (Livak and Schmittgen, 2001) and normalized with OEOE_005 gene. This gene codes for a dehydrogenase-like protein, and it was selected as a housekeeping gene because each time it was evaluated it showed no changes in its expression between the strains.

Differences between measurements were determined using the Mood test (Gibbons and Chakraborti, 2011) with the Statgraphics Centurion XIX statistical software. We considered differences significant when p-values were less than 0.05.

The mutant strain E1 was able to grow in 15% ethanol and a high malic acid concentration, as stress factors, showing a higher biomass formation compared to the PSU-1 strain, under the same culture conditions (Supplementary Figure S3A, Supplementary material). Furthermore, E1 had a higher consumption of malate and, therefore, a higher production of lactate (Supplementary Figures S3B,C, Supplementary material). Besides, E1 showed a higher consumption of glucose and fructose (Supplementary Figures S3D,E, Supplementary material).

We evaluated the growth of the PSU-1 and E1 strains on three wine types (Figure 2). When comparing the growth capacity of the strains, we observed that the E1 strain is better adapted to the stress conditions of each wine and grows more than the PSU-1 strain. The E1 strain was able to complete the exponential phase earlier in synthetic wine (22 h) than in natural white and red wine (80 h). However, the highest concentration of biomass produced was observed in both natural wines at the 80 h mark. It was observed that E1 had a higher biomass production in Cabernet Sauvignon (48% more) and Chardonnay (46% more) compared to biomass production in MaxOeno (Table 1). Likewise, the PSU-1 strain showed higher biomass production in Cabernet Sauvignon (44% more) than in MaxOeno, however, it produced lower biomass in Chardonnay (62% less). Among the evaluated wines, white wine (Chardonnay) had a higher impact on the growth of the PSU-1 strain. In general, it was observed that E1 had a higher biomass production, in Cabernet Sauvignon (67% more), MaxOeno (61% more) and Chardonnay (18% more), compared to PSU-1 (Table 1).

The E1 and PSU-1 strain were cultured using MaxOeno 15% v/v of ethanol at pH 3.5 (Supplementary Figures S4, S5, Supplementary material). For this, an adaptation to pH was carried out in PSU-1, as was mentioned above. The cultivated E1 strain showed a higher biomass formation and specific growth rate, compared to the PSU-1 strain (Table 1).

The PSU-1 strain was able to grow and perform MLF in red wine. Moreover, the PSU-1 strain produced a higher biomass in red wine, even more than in synthetic wine (44% more), although with a slower rate of growth (Table 1). Likewise, the E1 strain produced a higher biomass than PSU-1 in red wine (67% more) (Table 1).

Redistribution of the metabolic flux of the strains O. oeni PSU1 (blue) and O. oeni E1 (red) was evaluated in three wine culture conditions (Figures 3–5). The metabolic pathways of carbon were studied using the iSM454 model and the pathway of phosphoketolase (heterolactic fermentation), reduction of fructose, and degradation of citrate, and malolactic fermentation showed differences between the strains evaluated (Figures 3–5).

In synthetic wine (MaxOeno), the PSU-1 strain showed a higher specific rate of fructose consumption (1.54 mmol gDCW⁻¹ h⁻¹) when compared to the E1 strain (1.10 mmol gDCW⁻¹ h⁻¹) (Figure 2). However, the E1 strain showed a higher specific rate of mannitol production (0.34 mmol gDCW⁻¹ h⁻¹) than PSU-1 (0.17 mmol gDCW⁻¹ h⁻¹). Both strains showed similar specific rates of glucose consumption. Finally, the E1 strain showed a higher consumption of L-malate (0.67 mmol gDCW⁻¹ h⁻¹), and a higher (D + L)-lactate production (0.38 mmol gDCW⁻¹ h⁻¹), comparted to PSU-1 (0.19 mmol gDCW⁻¹ h⁻¹ and 0.03 mmol gDCW⁻¹ h⁻¹, respectively) (Figure 3).

In red wine (Cabernet Sauvignon), on one hand, the E1 strain showed a higher specific rate of fructose consumption (0.54 mmol gDCW⁻¹ h⁻¹) compared to the PSU-1 strain (0.07 mmol gDCW⁻¹ h⁻¹) (Figure 4), showing a high specific rate of mannitol production (0.07 mmol gDCW⁻¹ h⁻¹) as well. Moreover, the E1 strain showed a higher consumption of L-malate (0.72 mmol gDCW⁻¹ h⁻¹), and a higher (D + L)-lactate production (1.10 mmol gDCW⁻¹ h⁻¹), comparted to PSU-1 (0.26 mmol gDCW⁻¹ h⁻¹ and 0.35 mmol gDCW⁻¹ h⁻¹, respectively) (Figure 4). On the other hand, the PSU-1 strain showed a higher specific rate of erythritol (0.14 mmol gDCW⁻¹ h⁻¹) and D + L-Lactate production (0.33 mmol gDCW⁻¹ h⁻¹) compared to the E1 strain (0.09 mmol gDCW⁻¹ h⁻¹ and 0.20 mmol gDCW⁻¹ h⁻¹, respectively) (Figure 4). Finally, E1 strains showed a low specific rate of the heterolactic pathway of acetate production (0.08 mmol gDCW⁻¹ h⁻¹), and a higher specific rate of ethanol production (0.34 mmol gDCW⁻¹ h⁻¹), compared to PSU-1 strain (0.22 mmol gDCW⁻¹ h⁻¹ and 0.17 mmol gDCW⁻¹ h⁻¹, respectively).

In white wine (Chardonnay), the E1 strain preferentially consumed L-malate (0.35 mmol gDCW⁻¹ h⁻¹) and L-citrate (0.11 mmol gDCW⁻¹ h⁻¹), producing oxaloacetate and then pyruvate (0.12 mmol gDCW⁻¹ h⁻¹). In contrast, PSU-1 consumed lower nutrient compared to E1 (Figure 5).

In general, it was observed that only E1 produces mannitol and diacetyl; PSU-1 does not produce them and only produces mannitol in red wine, and diacetyl only in white wine. We observed that the consumption of L-malate for the production of L-lactate was 82%, and the consumption of fructose and glucose to generate pyruvate and D-lactate, through its heterolactic route, was 18%.

We performed a massive transcriptome sequencing to identify changes produced by random mutagenesis. Then, eight genes of known function were observed that showed differences between the mutant strain and the parental strain, and from these genes, four were selected to be analyzed by qPCR. According to their description, these genes would be related to growth and fructose consumption, which were the highest differences observed between E1 and PSU-1 during the culture analyses presented above. The OEOE_1708 gene codes for a fructokinase enzyme (EC 2.7.1.4) involved in the transformation of fructose to fructose-6-phosphate. This overexpression was determined approximately at 16 h of culture and coincides with the greatest difference between the strains, considering the consumption of fructose and the production of acetate (Supplementary Figures S5B,D, Supplementary material). These results were validated using the quantitative chain reaction (qPCR) technique in order to confirm the highest expression observed (Table 3). The OEOE_1794 (codes a universal stress protein) and OEOE_1795 (codes a Mn²⁺ and Fe²⁺ transporter) genes showed a higher number of transcripts in E1 than in PSU-1 when both transcriptomic analysis techniques were used. However, the OEOE_0522 gene (codes a transcriptional regulator, PadR family) showed a higher number of transcripts in E1 than in PSU-1 when using the HiSeq sequencing technique but it was not satisfactorily amplified when qPCR was used.

This study analyzed the capacity of the random mutagenesis technique using EMS as a strategy for the genetic improvement of strains of the Oenococcus oeni species, which, to our knowledge, has not been applied to this species before. As the last selection step, the mutagenized strain E1 and the parental strain PSU-1 were cultivated in synthetic wine (MaxOeno) at pH 4.8, and then their metabolite consumption and production capacities were compared. Based on the results, this technique proved able to generate a new and improved strain in comparison to its parental PSU-1 strain, from which it descends. It is important to point out that the growth and metabolization traits of the malate from the E1 strain were maintained by the strain through every trial performed in this work, without modifying, its behavior which indicates that the stress and selection conditions to which the cell was subjected to generated persistent modifications that benefit its survival. Some authors who have used different techniques for natural genetic improvement (without using genetic engineering) have observed that applying stress and selection rounds are key processes for maintaining a phenotype (Akcil, 2004; Li et al., 2015; Betteridge et al., 2018; Jiang et al., 2018).

To evaluate the oenological capacities of both strains, and identify their differences, we analyzed their metabolic behavior in three different wines, using the iSM456 metabolic model constructed for O. oeni. Even more, we identified genes that showed differences in expression when comparing both strains.

In our previous work, we observed that the specific growth rate linearly decreases when the ethanol content in the wine increases (Contreras et al., 2018). We observed similar trends in the present results with the PSU-1 strain cultivated in MaxOeno with 15% of ethanol (r2 = 0.96). Conversely, the mutant strain E1, in average, showed a specific growth rate that was 39% greater than PSU-1 in synthetic wine and red wine. In wine white, the PSU-1 strain showed little growth, hence, E1 strain showed a specific growth rate that was 86% greater than PSU-1. Moreover, the E1 strain was able to convert an average of 34% more of malate in comparison to PSU-1, regardless of the wine being used.

PSU-1 strain showed a greater production of total biomass in red wine (Cabernet Sauvignon), although it was less in comparison to the E1 strain, and a minimal production of biomass in white wine. It has been noted that strains that have evolved in a particular environment develop the ability to adapt and take advantage of nutritional and physical–chemical conditions (Campbell-Sills et al., 2015; Breniaux et al., 2018; Collombel et al., 2019). Although all wines contain phenolic compounds, red wines are the ones which have the highest concentration of it (Paixão et al., 2007). Among the red wines in Chile, the Cabernet Sauvignon variety has shown the highest concentration of phenolic compounds (Lutz et al., 2012). Likewise, it has been stated that phenolic compounds aid in the survival of microorganisms since they function as controlling agents of the redox potential (Rozès et al., 2003). The PSU-1 strain of O. oeni was isolated from red wine in Pennsylvania, in 1972, and one of its main advantages has been described as its capacity to induce MLF faster in red wines (Beelman et al., 1977). It is possible that the PSU-1 strain evolved in red wine grapes and that this is why phenolic compounds might help its metabolism and resistance to stress.

Other than cellular growth, among the main differences observed between E1 and PSU-1 was the differentiated intake of nutrients. Regardless of the type of wine being used, the mutant E1 strain showed a higher total consumption of fructose, malate, and citrate, which is also related to a high production of mannitol and lactate. Some authors state that the concentration of citrate and/or malte in the culture media could trigger a change in the preference of nutrient consumption (Ramos and Santos, 1996; Rozès et al., 2003). However, we observed that its consumption was mainly promoted by metabolic needs the bacteria generates to survive stress. This was observed in the results of consumption and production of compounds expressed in g/L, and, also, in the results of specific rates expressed in mmol gDCW⁻¹ h⁻¹.

We observed that the glucose consumed preferably follows the heterolactic route for the production of pyruvate. What is more, most of the consumed fructose by both strains goes into the production of pyruvate instead of mannitol. In our previous work, we observed that O. oeni generated NAD(P)⁺ cofactors through the production of mannitol and erythritol, contributing to the balance of internal pH through the production of diacetyl; results that were also observed by other authors (Saguir and De Manca Nadra, 1996; Versari et al., 1999).

Likewise, when analyzing the intracellular specific fluxes of E1, it was observed that the production of fructose-6-phosphate, which impacts the pyruvate production route, was 86% higher than the production rate of mannitol. In addition, in this strain, the consumption rate of malate is 26 and 74% higher than its consumption rate of fructose and glucose, respectively.

Mannitol production is limited and, thereby, NAD(P)⁺ production is as well. This might be due to the lack of manganese (Mn²⁺) in the cell; this element is key for malolactic enzyme activation (Acevedo et al., 2020) and for the function of numerous enzymes related to pyruvate production (Saha, 2006). Moreover, it has been observed that a low Mn²⁺ concentration inside the cell inhibits mannitol production, even when the mannitol dehydrogenase enzyme does not require cofactors to function (von Weymarn et al., 2002). Authors who have studied the proteome and the transcriptome of bacteria have observed the presence of Mn²⁺ and the response regulation to varied types of stressors, and even an impact on the physiology and metabolism of these microorganisms (Ogunniyi et al., 2010; Mbah and Isokpehi, 2013).

Interestingly, the E1 strain showed an overexpression of the OEOE_1794 gene, which codes an UspA-like protein which allows for cell survival and growth under stress conditions. The family of universal stress proteins (UspA-like) has been described as one of the main tools for allowing bacteria, archaea, fungi, and plants to be able to respond to unfavorable stress conditions so that they adapt, survive, and grow (Nachin et al., 2005; Isokpehi et al., 2011; Mbah and Isokpehi, 2013). The deletion of USP genes affect the growth of many bacteria such as Escherichia coli, Mycobacterium tuberculosis, Salmonella typhimurium, Burkholderia pseudomallei, and Listeria monocytogenes (Chatterjee et al., 2006; Liu et al., 2007; Hingley-Wilson et al., 2010; Al-Maleki et al., 2014).

This concurs with the higher number of transcripts from the OEOE_1708 gene observed in the E1 strain grown in MaxOeno. This gene codes for a fructokinase enzyme (EC 2.7.1.4) involved in the transformation of fructose into fructose-6-phosphate. This is consistent with findings described by other authors who have observed that pyruvate formation is the cell’s key strategy to maintain pH homeostasis in low pH conditions, meaning in acid stress (Fernandez et al., 2008; Zuljan et al., 2014; Qi et al., 2021). In our case, stress is increased by the presence of high concentrations of ethanol, which generates fluidization in the plasmatic membrane and the influx of ions (Da Silveira et al., 2002; Chu-Ky et al., 2005). The above would be explained because pyruvate is a key compound that binds the metabolism of carbon to other metabolic routes related to the production of amino acids and fatty acids, mainly (Liu et al., 2007; Maleki and Eiteman, 2017; Suo et al., 2021). This indicates that the cell could be activating this route to produce systems to repair the membrane like the production of fatty acids and ATP production. It has been observed that the [NAD(P)H/NAD(P)+] redox balance inside the cell has direct impact on the production rate of pyruvate. Some authors who have worked with Escherichia coli have observed that NAD(P) + increases glucose consumption and pyruvate production (Hansen and Henning, 1966; Akita et al., 2016).

We observed that a greater expression of the OEOE_1795 gene codes for a protein that transports Mn²⁺ and Fe²⁺ (Mn²⁺ and Fe²⁺ transporter of the NRAMP family). This would indicate that the E1 strain is resistant to low pH and high ethanol stresses when compared to the PSU-1 strain because it is more efficient in metabolizing malate, mannitol, and pyruvate as a result of its capacity to obtain Mn²⁺.