Transient acetaldehyde production by SO 2 -producing Saccharomyces cerevisiae promotes the survival of Oenococcus oeni during co-fermentation

Stuck or sluggish malolactic fermentation (MLF) can be problematic in stressful wine conditions, particularly white and sparkling base musts/wines. In these cases, knowledge of yeast-bacteria strain compatibility and the amount of sulfur dioxide (SO 2 ) a yeast strain produces are important considerations for successful MLF. Here, laboratory-and pilot-scale co-fermentations in Chardonnay were used to investigate the effect of yeast-derived SO 2 on Oenococcus oeni survival. Although yeast-derived SO 2 is generally inhibitory, we show that SO 2 production (to approximately 65 mg/L) can be uncoupled from O. oeni survival in the early stages of co‑fermentation. Bacterial survival in the presence of specific SO 2 -producing yeast strains was correlated with the early, transient formation of high acetaldehyde concentrations. Oenococcus oeni survival coincided with molecular SO 2 concentrations remaining below an extremely low threshold of inhibition, which exponentially increased from approximately 3–6 µg/L in the first three days of co-fermentation. Strain-dependent sensitivity of O. oeni to bound SO 2 remains a possibility, although the extent and mechanism of such inhibition by the SO 2 adduct during co‑fermentation remain unclear. The choice of co‑inoculation yeast strain also influenced wine diacetyl concentration, which was only detected in wines co-inoculated with high SO 2 -producing S. cerevisiae strains. The wines with high diacetyl concentrations were found to be distinct by a sensory panel, with comparatively high citation frequency for a buttery sensory attribute. Both the SO 2 - and acetaldehyde production capacity of yeasts are, therefore, seen as meaningful co-inoculation selection criteria. The range of yeast strains suitable for MLF induction by co-inoculation could be widened to include SO 2 -producing strains that transiently produce an early, high concentration of acetaldehyde. The effects of low, equilibrium concentrations of molecular SO 2 should also be considered in conjunction with total SO 2 as a measure of SO 2 toxicity towards O. oeni following co-inoculation.


INTRODUCTION
Malolactic fermentation (MLF) is an important winemaking process performed by certain wine lactic acid bacteria (LAB), principally Oenococcus oeni. The onset of MLF, characterised by the conversion of L-malic acid to L-lactic acid and CO 2 , commences after the malolactic bacteria population has grown to approximately 10 6 cells/mL (Wibowo et al., 1985). Consequently, the success or failure of MLF is highly dependent on wine parameters that affect the growth and survival of malolactic bacteria, particularly wine pH, the concentrations of alcohol and sulfur dioxide, and temperature. In general, wine conditions conducive to MLF include < 13 % v/v alcohol, > pH 3.4, < 8 mg/L free SO 2 , < 30 mg/L total SO 2 , 18-22 °C, while conditions outside of this range can become more challenging for MLF (Krieger-Weber, 2017). Moreover, for any given wine, the ability of wine LAB to grow and conduct MLF is highly dependent on the species and strain of LAB involved and their respective tolerances to the latter parameters.
Another major factor affecting the growth and MLF capability of wine LAB is the potential for interaction with fermentation yeast, which can encompass a diverse range of Saccharomyces cerevisiae and also non-Saccharomyces strains. Such interactions are complex. MLF outcomes can be affected by metabolite production or physical cellular interactions . Inhibitory yeast metabolites include the formation of SO 2 (Henick- Kling and Park, 1994;Larsen et al., 2003;Wells and Osborne, 2011), medium-chain fatty acids (Guilloux-Benatier et al., 1998), inhibitory peptides (Nehme et al., 2010) and succinic acid at concentrations greater than 1 g/L (Torres-Guardado et al., 2022). Further, in a study of the interactions between Leuconostoc oenos (O. oeni) PSU-1 and S. cerevisiae 522 during co-culture, King and Beelman (1986) reported that the greatest inhibition of early bacterial growth was caused not by yeast metabolites but by yeast growth itself.
Whether intrinsically derived during fermentation or added during the winemaking process, SO 2 enters a pH-dependent equilibrium. In an aqueous solution at pH 3.5, only a minor proportion (2 %) of free SO 2 exists in the molecular, and primarily antimicrobial, form (Ribéreau-Gayon et al., 2006). At lower pH, such as the pH of white wine, this proportion increases. For example, at pH 3.2, the proportion of molecular SO 2 in an aqueous solution increases to 3.91 % (Ribéreau-Gayon et al., 2006).
The molecular SO 2 concentrations reported as inhibitory towards O. oeni and other wine LAB are varied. Fia et al. (2018) reported that molecular SO 2 concentrations between 0.2-0.6 mg/L impeded growth and MLF activity of an O. oeni strain in white wine, and Ribéreau-Gayon et al. (2006) indicated 0.15-0.4 mg/L molecular SO 2 had varying, wine-dependent, inhibitory effects on LAB survival. Carr et al. (1976), however, suggest that much lower molecular SO 2 concentrations (0.032 mg/L) completely suppressed the growth of heterofermentative cocci (isolated from cider) in a nutrient medium, whereas Lactobacillus plantarum (currently Lactiplantibacillus plantarum) could tolerate up to approximately 1.5 mg/L molecular SO 2 .
Bisulfite can also readily form adducts with compounds in wine, particularly carbonyls such as acetaldehyde, pyruvic acid, and 2-ketoglutaric acid (respective K d at pH 3.3: 1.5 × 10 -6 , 5.55 × 10 -4 and 1.4 × 10 -4 ) (Burroughs and Sparks, 1973). Such bound forms of SO 2 may also exhibit inhibitory activity towards wine LAB, with inhibition in wine typically occurring at > 50 mg/L bound SO 2 (Fugelsang, 1997;Wibowo et al., 1985). However, this concentration also appears subject to variability; as little as 5 mg/L acetaldehydeand pyruvic acid-bound SO 2 was inhibitory to the growth of wine LAB in a semi-defined culture medium (Wells and Osborne, 2012), and 10 mg/L acetaldehydebisulfite restricted the growth and MLF activity of some bacterial strains in red wine (Lafon-Lafourcade, 1975). Several authors (Larsen et al., 2003;Osborne and Edwards, 2006;Wells and Osborne, 2011) have demonstrated that wines fermented by yeast strains producing higher total SO 2 concentrations were generally inhibitory to O. oeni. However, Larsen et al. (2003) and Osborne and Edwards (2006) could not demonstrate a clear relationship between yeast-derived SO 2 production and bacterial inhibition and proposed other inhibitory mechanisms.
White and sparkling-based wines present unique and difficult challenges for MLF. Compared to red wines, white wines destined for MLF generally have higher SO 2 concentration (e.g., up to approximately 50 mg/L total SO 2 ) and lower pH (e.g., pH 3.1-3.3), which approaches the limits for O. oeni survival. In such cases, increases in SO 2 from the fermentation yeast strain can render MLF extremely challenging. Consequently, the tolerance and adaptive capability of the malolactic bacterial strain to these conditions becomes critical to the success or failure of MLF.
Another important factor that can affect MLF outcome is the timing of bacterial inoculation. Of the different inoculation timings, sequential and co-inoculation are considered two major options (Bartowsky et al., 2015;Sumby et al., 2019). Sequential inoculation, whereby the starter culture is inoculated after alcoholic fermentation (AF), reduces the risk of acetic acid accumulation from carbohydrate metabolism by heterofermentative LAB. However, exposure of the starter culture to a high alcohol concentration can be detrimental to bacterial survival and MLF capability. During co-inoculation, the malolactic starter culture is typically inoculated 24 hours or more after yeast, thus enabling the bacteria to acclimatise to increasing alcohol concentration and, further, allowing greater availability of grape-derived nutrients in the early stages of AF. Importantly, this strategy has been suggested to improve overall MLF efficiency and assist in overcoming difficult wine conditions (Bartowsky et al., 2015). Some decades ago, the benefits of early inoculation of yeast and malolactic bacteria for rapid MLF induction were recognised (Beelman and Kunkee, 1985). A growing interest in co-inoculation (Bartowsky et al., 2015) has led to its successful application for MLF induction in white wine (Jussier et al., 2006;Sereni et al., 2020;Zapparoli et al., 2006), including white varieties with low pH (Guzzon et al., 2016;Knoll et al., 2012). However, the potential of co-inoculation for MLF induction in white musts/wines with challenging concentrations of SO 2 remains unexplored. Further, since the interactions between yeast and bacteria during co-inoculation are likely more complex than sequential inoculation (Zapparoli et al., 2006), a greater understanding of such interactions is required to improve MLF performance in harsh wine environments.
This study investigates the role of S. cerevisiae-derived SO 2 on MLF progress during the co-inoculated fermentation of Chardonnay. Representative yeast strains with varying SO 2 production potential were investigated for their interactions with O. oeni during co-fermentation in Chardonnay. The early, transient formation of acetaldehyde by yeast was observed to be an important factor modulating SO 2 resistance in malolactic starter cultures during co-fermentation. Finally, a pilot-scale study further explored the effect of different yeast-bacteria pairings, in which acetaldehyde production varied, on MLF progress and wine sensory properties.

Microorganisms
Experiments investigating the SO 2 production potential of S. cerevisiae wine yeast strains utilised isolates of commercial preparations and other wine strains obtained from the Australian Wine Research Institute (AWRI) Wine Microorganism Culture Collection ( Figure 1). Some yeast strains are interspecific hybrids in nature (Borneman et al., 2016). In subsequent experiments, representative yeast strains were selected to investigate the interactive effects of yeast strain SO 2 production potential on the survival of O. oeni during co-fermentation (see Table 1 for a listing and description of S. cerevisiae and O. oeni strains used in yeast-bacteria interaction studies). The yeast strains included two non-SO 2 -producing yeast strains (AWRI796, AWRI1483) and two SO 2 -producing yeast strains (AWRI1497 and AWRI2880

Fermentation conditions
Laboratory-scale fermentations were conducted in 100 mL or 500 mL in the case of Interaction experiment 1, screw-cap sealed (air-tight) glass (Schott) vessels fitted with a one-way directional-flow gas check valve and sample port assembly described by Schmidt et al. (2020). This apparatus prevented air ingress and enabled the rapid formation of inert conditions at the start of AF, which were critical in preventing metabolic interference from oxygen and oxidative reactions with SO 2 . The 100 mL and 500 mL vessels were filled with 90-100 mL and 530 mL Chardonnay juice and constantly stirred at 250 rpm and 150 rpm, respectively. AF was undertaken at 17 °C, with minimal foaming in the fermentation headspace. Yeast starter cultures prepared in grape juice were inoculated at an optical density of 0.02 absorbance units (600 nm) (approximately 1 × 10 6 cells/mL).
In the study of SO 2 production by wine yeast, fermentations were sampled to measure glucose plus fructose. Fermentations were cold settled (4 °C) and clarified by centrifugation and filtration (0.2 µm syringe filters, Millipore, Billerica, M.A., USA) after completion of AF (average glucose plus fructose concentration ≤ 1 g/L). Analysis of total SO 2 was undertaken on completion of clarification operations. All experiments were done in triplicate.
For yeast-bacteria co-fermentations, O. oeni starter cultures were inoculated 24 h after yeast inoculation at approximately 5 × 10 6 cells/mL. Co-fermentations were undertaken in triplicate (three bacteria strain replicates per yeast strain), except for Interaction experiment 2, in which yeast strain was replicated (n = 3) and the bacterial strain was also replicated (n = 3) within yeast strain ( Figure S1). Upon completion of AF, wines were incubated statically at 22 °C for the duration of MLF. Samples (1-2 mL) were taken aseptically during fermentation to determine glucose plus fructose, L-malic acid, acetaldehyde, total SO 2, and viable bacteria as required. Samples for chemical analyses were centrifuged (approximately 24,000 × g, 2 min), and supernatants were stored at -20 °C. At MLF completion (average L-malic acid concentration < 0.3 g/L), wines from Interaction experiment 1 were centrifuged (3273 × g, 30 min) and stored at -20 °C for further analyses.

Effect of exogenous acetaldehyde addition
The effects of adding excess acetaldehyde (99 mg/L; molar ratio acetaldehyde:SO 2 , 5:1) during co-fermentation of the SO 2 -producing yeast S. cerevisiae AWRI1497 on the survival of O. oeni YV Select was investigated in the same 2017 Chardonnay juice. Acetaldehyde (Sigma-Aldrich, St. Louis, MO) was added to fermentations 4 h prior to bacteria inoculation from a freshly prepared aqueous stock solution (2 % v/v distilled acetaldehyde in Milli-Q H 2 O). In preparing the latter solution, losses of acetaldehyde were minimised by liquid handling in a 4 °C cold room, using Milli-Q purified water (49 mL) pre-chilled to 4 °C and dispensing acetaldehyde (stored at -20 °C) under the aqueous surface using a pre-chilled (4 °C) gas-tight glass micro-syringe. The required volume of acetaldehyde stock solution (0.62 mL) was added to 100 mL fermentations via a fitted (Luer) sampling port and sterile syringe.  (2021). Approximately 1 tonne of Chardonnay grapes (Willunga, South Australia) from the 2021 vintage were hand-picked (non-sulfited) and transported to the Hickinbotham Roseworthy Wine Science Laboratory, Waite Campus, The University of Adelaide, South Australia, and stored in a cool room at 0 °C for several days. Grapes (non-sulfited) were crushed, destemmed, pressed and, after the addition of pectinase enzyme (Rohavin L, AB Biotek, 4 mL/100 L), the juice was cold settled (0 °C for approx. 5 days) and acid adjusted with tartaric acid (1.5 g/L). The composition of the 2021 Chardonnay grape juice (non-sulfited) is shown in Table 2. Twelve stainless steel fermentation vessels (50 L) were filled with 35 L of Chardonnay juice and warmed to 18-20 °C before yeast inoculation. AF was initiated by inoculation of the respective yeast strain preparations. These were rehydrated and inoculated (triplicate) according to the manufacturer's recommendations. After 24 h, O. oeni VP41 malolactic starter culture was rehydrated according to the manufacturer's recommendations and inoculated into each fermentation at 1.5 times the recommended rate to achieve approximately 2 × 10 6 CFU/mL. Alcoholic fermentations were maintained at approx. 17-22 °C and, after the active phase, were racked to fill 30 L stainless steel fermentation vessels. After completion of AF, wines were maintained at 20 °C for MLF. Samples were taken aseptically during co-fermentations to determine concentrations of glucose plus fructose, L-malic acid, acetaldehyde, total SO 2 , viable bacteria, and other wine components as required. After MLF, wines were incubated for an additional 7 days (approx.) at 20 °C, after which SO 2 was added (80 mg/L). Wines were then settled at 0-4 °C, subsequently racked, free SO 2 concentration adjusted to approximately 40-45 mg/L, filtered (0.45 µm pore size membrane), and bottled with a screw-cap closure. Bottled wines were stored at 15 °C for approximately 7 months prior to sensory and diacetyl analyses.

Chemical analyses
Concentrations of residual sugar and L-malic acid were determined enzymatically (Megazyme) utilising 96-well microtiter plates. Methods for quantifying glucose and fructose were based on those described by Hohorst (1965) and Vermeir et al. (2007). Reagent and enzyme concentrations used to determine L-malic acid concentration were based on a proprietary (Molecular Devices, MDS Analytical Technologies) microplate method.
Acetaldehyde concentration was determined enzymatically (Megazyme) using 96-well microplates and glycylglycine buffer similar to that described by McCloskey and Mahaney (1981) (15.8 g/L glycylglycine, 7.46 g/L KCl, pH 9, without polyvinylpyrrolidone). Acetaldehyde standards (0-100 mg/L) were prepared from an equimolar aqueous stock solution of acetaldehyde-bisulfite (4.55 mM, 250 mL). The stock solution was prepared from freshly distilled acetaldehyde (250 µL, density 0.801 g/cm 3 at 4 °C) and potassium metabisulfite (Ace Chemical Co., Camden Park, South Australia, or Ajax Finechem, Seven Hills, N.S.W., Australia; 0.5061 g) in MilliQ purified water. During stock solution preparation, losses of acetaldehyde were minimised by using the previously described procedure for the preparation of 2 % v/v aqueous acetaldehyde. After preparation, the stock solution was equilibrated by storage at 4 °C for at least 24 h. Total SO 2 concentration was determined colorimetrically (420 nm) using 5,5΄-dithio-bis-2-nitrobenzoic acid (DTNB) reagent prepared in phosphate buffer (pH 8.4) or commercial preparation (Rowe Scientific, Adelaide, Australia). Samples of juice and wines during yeast SO 2 production experiments were analysed using a Discrete Analyser (Thermo Fischer Scientific) by Affinity Labs (previously AWRI Commercial Services) [International Organization for Standardization 17025 accredited laboratory, Adelaide, SA, Australia].
For the remaining studies, SO 2 concentration was analysed using DTNB reagent and phosphate buffer (pH 8.4) in conjunction with 96-well microplates (reaction volume 260 µL). Sulfite standards (0-120 mg/L total SO 2 ) for microplate assays were prepared from the equimolar stock solution of acetaldehyde-bisulfite (4.55 mM; molecular mass SO 2 , 64.1 g/mol) described above for the analysis of acetaldehyde.
Low concentrations of free SO 2 were calculated based on the approach of Hood (1983), using the following formula (assuming acetaldehyde as the major carbonyl contributing to bound SO 2 ): where [S], [X], and [x] are molecular concentrations of, respectively, free SO 2 in any form, (total) carbonyl compound (acetaldehyde) and undissociated carbonyl bisulphite (bound SO 2 ), and K = equilibrium constant of acetaldehyde bisulphite (1.5 × 10 -6 at pH 3.3) (Burroughs and Sparks, 1973). Molecular SO 2 concentration was calculated from the respective free SO 2 concentration and initial Chardonnay juice pH.
The concentration of succinic acid was determined by high-performance liquid chromatography (HPLC) as described by Nissen et al. (1997) using a Bio-Rad HPX-87H column. Other wine compositional analyses (pH values and total acidity, ethanol, acetic acid, and diacetyl concentrations) were determined using the National Association of Testing Authorities (NATA) accredited protocols by Affinity Labs (previously AWRI Commercial Services) (Adelaide, Australia). Diacetyl concentration was determined using a stable isotope method using solid-phase microextraction (SPME) combined with gas chromatography-mass spectrometry (GC-MS), based on that described by Hayasaka and Bartowsky (1999). Diacetyl analysis was undertaken using bottled wine samples at the time of sensory assessment.

Determination of viable bacteria
Viable O. oeni cells were enumerated by dispensing 50 µL aliquots of serially diluted (0.1 % w/v bacteriological peptone (Amyl Media, Australia)) fermentation samples onto a modified de Man, Rogosa, and Sharpe (MRS, Amyl or Oxoid) agar using an automated spiral plater (WASP 2, Don Whitley Scientific, Australia). The latter medium was prepared by supplementing MRS medium with preservative-free clarified apple juice (20 % v/v) and agar (1.5-2.4 % w/v, Amyl or Oxoid); in the first interaction experiment, this medium was further supplemented with preservative-free clarified red grape juice (20 % v/v). Natamycin (50 mg/L, NataP UltraPure, Handary, Alchemy Agencies, Australia) was added to modified MRS agar media to suppress yeast growth. The agar plates were incubated aerobically (or for some experiments in a nitrogen atmosphere) at 27 °C for approximately 7-10 days, and resultant bacterial colonies were enumerated using a Protocol 3 colony counter (Synopsis, Don Whitley Scientific, Australia).

Sensory analysis
The projective mapping sensory method of Napping (N) was used to holistically evaluate the treatments and replicate Chardonnay wines from the pilot-scale study based on their perceived similarities and differences. Ultra-Flash Profiling (UFP) was also employed to provide indicative sensory (appearance, aroma, and palate) descriptors of the sample set and groupings (Pagès, 2005). A panel of 9 assessors (5 females, 4 males) with an average age of 48 years (SD = 10.0) were convened to evaluate the wine set in duplicate. All panellists were part of the external AWRI-trained descriptive analysis panel, had extensive experience in wine quantitative sensory descriptive analysis (QSDA), and had completed N/UFP tasks previously. Prior to the evaluation, panellists were familiarised with a sub-selection of the samples for tasting and general discussion as a training task. Panellists assessed the wines in duplicate for the formal evaluation, with both sets of 12 wines (4 treatments × 3 replicates) assessed on the same day. Assessors were allowed as long as required to complete their projective maps and to add descriptive attributes, with a minimum 15-minute break between duplicate assessments. Samples were provided to panellists in 30 mL aliquots, in 3-digit-coded, covered, ISO standard wine glasses at 22-24 °C, in a randomised presentation order, but assessors were allowed to revisit samples as needed.
The formal evaluation was completed in one two-hour session. Each sample was presented in a randomised William's block design generated by Compusense Cloud sensory evaluation software (Compusense Inc., Guelph, Canada). The samples were assessed, and responses were transferred directly onto a 1.5:1 digital sensory space via a tablet, with panellists able to enter free-choice comments for each sample or group of samples.

Statistical analysis
Final wine compositional data were assessed by oneway ANOVA using Minitab Statistical Software 21 (Minitab LLC, PA, USA). Tukey's pairwise comparison post hoc test was used to determine if respective means differed among yeast strain treatments (α = 0.05). All sensory data analysis was carried out using R with the SensoMineR (http:// sensominer.free.fr) (Le and Husson, 2008), factoextra (http:// www.sthda.com/english/rpkgs/factoextra) (Kassambara and Mundt, 2020), reshape2 (https://github.com/hadley/ reshape), readxl (https://readxl.tidyverse.org, https://github. com/tidyverse/readxl) (Wickham et al., 2019) and ggrepel (http://github.com/slowkow/ggrepel) (Slowikowski, 2021) packages. Multiple factor analysis (MFA) of the sample coordinates of all panellists' maps was conducted to produce a group consensus map for the 12 wines. Secondarily, UFP descriptors were incorporated as supplementary variables to the MFA consensus map through the ratio scaling of the scores of the correlations between variables and factors to the range of the group factor scores of the wines. The position of the UFP descriptors on the MFA map provides indicative sensory properties of the wines.
For finalising the UFP data, one sensory analyst with wine expertise inspected and consolidated perceptually similar attributes upon discussion with assessors, known attribute correlations, and paying particular attention to the wine descriptive terminology hierarchy proposed in Noble et al. (1987). Attributes describing the entire wine set were generally eliminated from the data, while attributes discriminating between wines were retained. Attributes with a low frequency of use were removed. This data processing aided in the clarity of the visualisation of the consensus maps produced, with the resulting maps highlighting only the attributes that discriminated the wines. Panel performance was assessed using the multivariate RV coefficient for each panellist compared to the consensus MFA map produced. All panellists were found to be performing at an acceptable level, given the overall variability in the wine set.

Yeast strain SO 2 -production potential
The SO 2 production potential of a genetically diverse subset of 94 S. cerevisiae strains from within the wine clade (Figure 1) was assessed to identify the potential for positive and negative interactions with O. oeni. From the initial SO 2 concentration of 41 mg/L in this Chardonnay (2017) juice batch, the SO 2 concentration after fermentation varied by more than tenfold from 20 to 212 mg/L. Further, approximately half of the strains either increased or decreased the SO 2 concentration relative to the initial concentration of SO 2 . The 90 th percentile of SO 2 concentration was 63 mg/L, 22 mg/L more than the starting concentration, and the 10 th percentile of SO 2 concentration was 29 mg/L, 12 mg/L less than the starting concentration.

Effect of yeast strain and SO 2 production on the survival of O. oeni during co-fermentation
The effects of two non-SO 2 -producing yeast strains (AWRI796, AWRI1483) and two SO 2 -producing yeast strains (AWRI1497 and AWRI2880) on the survival, growth and MLF performance of the commercial O. oeni strain YV Select was investigated during co-fermentation in Chardonnay. The yeast strains were selected on their ability to modulate SO 2 concentration during fermentation in Chardonnay (Figure 1). The concentrations of total SO 2 , L-malic acid, viable bacteria, and the duration of AF during co-fermentation of O. oeni YV Select with each yeast strain are shown in Figure 2.
Bacterial viable cell concentrations in co-fermentations with each S. cerevisiae strain remained relatively constant (4.4 × 10 6 (SD = 1.1 × 10 6 ) CFU/mL) over the first 24 h following bacterial inoculation (days 1-2 of AF). However, bacterial viable cell concentrations in all co-fermentations declined over the subsequent 5 days (days 2-7 of AF). The extent of the decrease in bacterial viability varied greatly regardless of whether O. oeni was inoculated with SO 2 -producing or non-SO 2 -producing yeast strains.
For the non-SO 2 -producing yeast strains, the bacterial viable cell population within 7 days decreased by >10 2 -and >10 4 -fold for AWRI796 and AWRI1483, respectively. The decrease in bacterial cell viability affected MLF duration, which increased from 36 days with AWRI796 to 64 days with AWRI1483. These co-fermentations were associated with decreases in total SO 2 concentration of 14 mg/L and 5 mg/L during AF.
Co-fermentations with the yeast strains AWRI1497 and AWRI2880 substantially increased total SO 2 concentration to 65 mg/L and 63 mg/L after 5 days (Figure 2). However, these yeast strains had different effects on bacterial survival and MLF activity despite producing similar concentrations of SO 2 . For S. cerevisiae AWRI1497, the viable bacterial population diminished to non-detectable levels (< 20 CFU/ mL) and failed to recover sufficiently to initiate MLF. Co-fermentations with S. cerevisiae AWRI2880 exhibited a comparatively small initial decline in bacterial viability to 8.1 × 10 5 CFU/mL within the first 7 days, after which it increased to > 1.0 × 10 6 CFU/mL, and ultimately completed MLF within 34 days. Moreover, although there was evidence (P < 0.05) for differences in the concentrations of some major final wine parameters (including alcohol, total acidity, acetic and succinic acids, and pH value) (Table 3), the magnitude of those were small and remained within a normal concentration range for wine.

Influence of early acetaldehyde production by SO 2 -producing yeast on O. oeni survival
Variations between yeast strains in the production of acetaldehyde may give rise to subtle variations in free SO 2 concentrations during the early stages of fermentation. Consequently, modulation of free SO 2 concentration during the crucial post-inoculation period may also affect bacterial survival. Whether variation in acetaldehyde formation by SO 2 -producing yeast strains could explain the differential post-inoculation survival rates of O. oeni was investigated in the laboratory-and pilot-scale studies. We also sought to determine whether the effect of yeast on O. oeni survival reported in Figure 2 was specific to the bacterial strain used or whether that observation could be generalised to other bacterial strains.

Laboratory scale studies
The effects of the two previously studied SO 2 -producing yeast strains, S. cerevisiae strains AWRI1497 and AWRI2880, on the viability of O. oeni YV Select and two other commercial O. oeni strains (VP41 and CH35) following co-inoculation into Chardonnay was investigated (Figure 3). These three O. oeni strains are commercially available to the Australian wine industry. As with the previous experiment (Figure 2), AWRI1497 and AWRI2880 produced similar concentrations  of total SO 2 (25-31 mg/L average maximum increase), but their effect on bacterial survival differed (Figure 3). The variation in bacterial survival was associated with the early transient formation of acetaldehyde and, consequently, MLF duration ( Figure 3). Over the first 5 days of co-fermentations with S. cerevisiae AWRI1497, each O. oeni strain exhibited a similar, linear reduction in viability from approximately 5 × 10 6 CFU/mL to 6 × 10 2 CFU/mL (Figure 3), coinciding with a slow increase in acetaldehyde concentration to a peak of 64 mg/L at day 4 ( Figure 3). From day 5, recovery of YV Select and VP41 viability was observed, increasing to > 10 6 CFU/mL at 44 days, with MLF completion at 58 days ( Figure 3). In comparison, the recovery of O. oeni CH35 was slower, achieving > 10 5 CFU/mL after 65 days (Figure 3), and although initiating L-malic acid degradation, MLF was still incomplete at 74 days ( Figure 3).
In contrast to S. cerevisiae AWRI1497, the impact of S. cerevisiae AWRI2880 on O. oeni strains YV Select and VP41 was relatively minor, their viability declining to approximately 2 × 10 6 CFU/mL after 5 days and remaining at ≥ 1 × 10 6 CFU/mL until 23 days (Figure 3). For both O. oeni strains, MLF was complete at 26 days ( Figure 3). However, for co-fermentations of S. cerevisiae AWRI2880 and O. oeni CH35, a linear decrease in bacterial viability from 5 × 10 6 CFU/mL to 3 × 10 4 CFU/mL occurred during the first 5 days, followed by recovery to ≥ 1 × 10 6 CFU/mL at 51 days and completion of MLF at 74 days ( Figure 3). The average acetaldehyde concentration in co-fermentations with S. cerevisiae AWRI2880 increased rapidly, reaching 95 mg/L on day one and a maximum of 112 mg/L on day 3, ultimately decreasing to 50-53 mg/L on day 8 (Figure 3). Compared to S. cerevisiae AWRI1497, it is noteworthy that acetaldehyde production began within the first 24 h for AWRI2880, while SO 2 production was not observed until 2 days after inoculation.
Determination of the molar ratio of acetaldehyde to total SO 2 in the early stages of fermentation revealed these values were considerably higher for AWRI2880 than AWRI1497, particularly at, and immediately following, bacterial inoculation; day one (4.0 vs. 1.6), day two (2.7 vs. 1.9) and day three (2.7 vs. 1.6). At the end of AF, the average molar ratio of acetaldehyde to total SO 2 for each yeast-bacteria strain combination was similar (1.1-1.3). The concentrations of total SO 2 and acetaldehyde were used to calculate the concentration of molecular SO 2 over time ( Figure 3). These calculations suggest that the variation in acetaldehyde concentrations produced by the two yeast strains resulted in differences in O. oeni molecular SO 2 exposure.
During co-fermentation with S. cerevisiae AWRI1497, the calculated molecular SO 2 concentration experienced by bacteria was 3.8-5.8 µg/L at days 1-3 and 7.1-15.6 µg/L at days 4-5. These molecular SO 2 concentrations were associated with a loss in bacterial viability. In contrast, the much lower calculated concentrations of molecular SO 2 during co-fermentation with S. cerevisiae AWRI2880 (i.e., approx. 0.9-1.7 µg/L at days 1-3 and 3.6-5.9 µg/L at days 4-5) had a negligible effect on the viability of O. oeni strains YV Select and VP41, although were associated with a reduction in viability of approximately 10 2 CFU/mL for O. oeni CH35. Interestingly, a subsequent increase in calculated molecular SO 2 concentration to 14.6 µg/L on day 8 in this co-fermentation did not impact bacterial viability further.  (99 mg/L) four hours before inoculation with O. oeni YV Select. Bacterial viability was monitored and compared to fermentations without an acetaldehyde addition. The concentration of acetaldehyde in treated co-fermentations decreased rapidly ( Figure S2). Within 24 h of acetaldehyde addition, the average concentration of acetaldehyde in treated and untreated ferments was indistinguishable (42-43 mg/L). The control and acetaldehydetreated cofermentations also displayed similar bacterial viability declines to 87-100 CFU/ mL at day 4 ( Figure S2).

Pilot-scale study
The relevance of these observations to commercial wine production was demonstrated in a winery-based pilot-scale coinoculation study. This study used commercial ADWY preparations of two non-SO 2 -producing and two SO 2 -producing yeast strains. Two of these were ADWY preparations of strains used in the first interaction experiment: AWRI796 (i.e., Maurivin AWRI796) and AWRI2880 (i.e., Anchor NT 50). The ADWY strain Actiflore Rosé was included as a second SO 2 -producing strain because the commercial form of AWRI1497 (i.e., Lalvin T73) was unavailable in Australia. Lalvin CY3079 was used as the second non-SO 2 -producing ADWY strain. The O. oeni strain VP41 was selected as a representative commercial malolactic bacterial strain.
Two distinct patterns of total SO 2 formation were evident in the early stages of co-fermentation (Figure 4). Two yeast strains, Actiflore Rosé and NT 50, yielded 56 and 50 mg/L total SO 2 on days 3 and 4, respectively. The other two yeast strains, AWRI796 and CY3079, yielded 18 and 21 mg/L total SO 2 on day 3 (Figure 4). Consistent with laboratory-scale experiments, the effects of the two high SO 2 -producing yeast strains on O. oeni VP41 were notably different. The bacterial population in co-fermentations with S. cerevisiae Actiflore Rosé declined approximately 10-fold to 1.9 × 10 5 CFU/mL at day 9, recovering to 9.4 × 10 5 CFU/mL at day 16, after which MLF was complete at day 37 ( Figure 4, Table 4). In contrast, co-fermentations with S. cerevisiae NT 50 were stimulatory to bacterial growth, yielding a maximum bacterial cell population of 5.1 × 10 6 CFU/mL at day 9 and rapid completion of MLF at day 13 ( Figure 4). A similar pattern of bacterial growth and L-malic acid degradation occurred during co-fermentation with S. cerevisiae CY3079 and, to a lesser extent, AWRI796, which only produced maxima of 21 and 19 mg/L total SO 2 within the first five days of AF, respectively (Figure 4).
Consistent with the laboratory-scale interaction experiment 2, the peak acetaldehyde concentration observed during the early stages of co-fermentations with S. cerevisiae NT 50 was notably higher (183-182 mg/L maximum at days 2 and 3) compared to the other yeast strains, which produced 87-98 mg/L acetaldehyde at this time ( Figure 4).
The concentrations of other chemical components of Chardonnay wines co-fermented with O. oeni VP41 and each yeast strain are shown in Table 4. There was evidence (P < 0.05) for differences in the concentrations of alcohol, total acidity, acetic acid and in pH value between wines produced by different yeast strains. Notably, acetic acid concentration increased by 0.16 g/L in NT 50 compared to CY3079 wines. Co-fermentations with the different yeast strains had a major effect on wine diacetyl concentration, which was only detected in wines coinoculated with the high SO 2 -producing yeast strains. Wines produced with NT 50 (4.9 (SD = 0.6) mg/L) contained significantly (P < 0.05) greater diacetyl concentrations than wines produced with Actiflore Rosé (1.0 (SD = 0.1) mg/L). In contrast, diacetyl was not detected in wines produced from co-fermentations with either of the non-SO 2 -producing yeast strains (CY3079 and AWRI796).  Sensory assessment of the Chardonnay wines by Napping combined with UltraFlash Profiling revealed that there were differences related to the co-inoculation yeast strain ( Figure 5). The AWRI796 co-inoculated wines were positioned to the left of Figure 5 and were described more often with the terms citrus, apple/pear, tropical fruit, floral, vanilla, flint, sourness, and sweetness. The CY3079 co-inoculated wines were separated along Dimension 1 from the AWRI796 wines, with a lower citation frequency of apple/pear and floral terms. The Actiflore Rosé and NT 50 coinoculated wines were also positioned away from the AWRI796 wines, and stone fruit, woody, butter, and bitterness terms were used more frequently for these wines. The NT 50 coinoculated wines had the highest citation frequency for butter aroma and flavour.  Saccharomyces cerevisiae strains: AWRI796 (796, purple circles), CY3079 (CY, light brown circles), Actiflore Rosé (AR, pink circles) and NT 50 (NT, green circles). Sensory descriptors from Ultra Flash Profiling (UFP) are superimposed.

DISCUSSION
As a prerequisite to examining the effects of yeast-derived SO 2 on O. oeni, the SO 2 -producing potential of 94 wine yeast strains was investigated following fermentation in a commercially prepared, pre-sulfited Chardonnay juice (2017). The SO 2 -producing capability observed supports other phenotypic surveys (Richter et al., 2013;Suzzi et al., 1985) and reflects the categorisation that most strains exhibit low SO 2 -producing ability (an increase of up to 22 mg/L SO 2 at the 90 th percentile), while some are high sulfite-producers (Eschenbruch, 1974). Although the fermentation conditions used here may not fully account for possible minor losses in SO 2 through gaseous exchange, the observation that approximately half of the yeast strains exhibited the ability to consume SO 2 in sulfited grape juice is of practical importance, yet such capability has received little attention. Wine yeasts utilise sulfite over sulfate as a source of sulfur, and can oxidise and reduce sulfite added to grape must (Eschenbruch, 1973). Recently, Ochando et al. (2020) showed that a winemaking S. cerevisiae strain (MC005) could produce or consume SO 2 in a synthetic grape juice medium and suggested that cellular uptake (diffusion) and production of SO 2 by yeast from the sulfate pool was in equilibrium with an initial threshold concentration of SO 2 in the juice. Similarly, for respective yeast strains in the current study, the final total SO 2 concentration of wines fermented using sulfited and non-sulfited juices was also in a similar range. For example, that of Chardonnay wine fermented with S. cerevisiae AWRI796 in sulfited (2017) and non-sulfited (2021 pilot-scale study) juices was 28 (SD = 1.5) mg/L and 20 (SD = 0.5) mg/L, respectively.
Representative yeast strains with differing SO 2 production exhibited wide-ranging impacts on the survival and MLF capability of O. oeni following co-inoculation in Chardonnay. This was highlighted in the first interaction experiment in which differential MLF duration was observed in ferments of non-SO 2 -producing and SO 2 -producing yeast strains. This observation supports the findings of others (Larsen et al., 2003;Osborne and Edwards, 2006) that yeast metabolites other than SO 2 are likely to affect bacterial survival. Of those yeast metabolites with the potential to cause inhibition of O. oeni, there were no major differences in the final concentrations of succinic acid (Torres-Guardado et al., 2022). This suggests that metabolites such as toxic fatty acids or bioactive peptides  could be involved in bacterial inhibition in wines co-fermented with the non-SO 2 -producing yeast strains. Further, a notable observation from this study was the highly divergent effects of the two SO 2 -producing yeast strains (AWRI1497 and AWRI2880) on the survival and growth of O. oeni. While the accumulation of yeast-derived SO 2 is generally associated with O. oeni inhibition (Larsen et al., 2003;Wells and Osborne, 2011), particularly in grape must already containing SO 2 (Henick- Kling and Park, 1994), the enhanced survival of O. oeni after co-inoculation with S. cerevisiae AWRI2880 compared to AWRI1497 appeared novel, and formed the basis of subsequent investigations.
Of the yeast metabolites capable of modulating yeast-derived SO 2 toxicity towards O. oeni, acetaldehyde was considered a primary candidate as it binds strongly with SO 2 , especially at low pH (K = 1.5 × 10 -6 at pH 3.3). Yeasts exhibit a strain-dependent ability to rapidly produce acetaldehyde during the early stages of AF, which is subsequently re-utilised as fermentation progresses (Cheraiti et al., 2010;Li and Mira de Orduña, 2017;Ochando et al., 2020). In the current study, peak acetaldehyde production was exceptionally high with the SO 2 -producing S. cerevisiae strain AWRI2880. This phenotypic feature resulted in comparatively high molar ratios of acetaldehyde:SO 2 (up to approximately 4:1) in the first few days following co-inoculation, and is proposed as a key factor contributing to the enhanced survival of O. oeni.
These observations support the findings of Carr et al. (1976), who reported that the addition of an excess quantity of acetaldehyde (or an equivalent quantity of H 2 O 2 ) alleviated SO 2 inhibition of L. plantarum. Similarly, Larsen et al. (2003) observed improved growth and MLF rate of O. oeni in Chardonnay wine fermented by S. cerevisiae V-1116 after treatments with excess amounts (50 mg/L) of acetaldehyde (or H 2 O 2 ) and concluded that inhibition of bacterial growth was due to inhibitory concentrations of free SO 2 . Under the co-fermentation conditions of the current study, however, attempts to validate the link between bacterial survival and acetaldehyde:SO 2 ratio were prevented by the rapid utilisation of added acetaldehyde by S. cerevisiae AWRI1497 (99 mg/L consumed within ≤ 24 h), resulting in no noticeable impacts on bacterial survival. The utilisation of exogenous acetaldehyde by yeast has been previously described (Ochando et al., 2020) and reflects the tightly regulated and dynamic flux of this intermediate during fermentation.
Further insight into the impact of the relative proportions of acetaldehyde:SO 2 on O. oeni survival was obtained by calculating the concentrations of free and molecular SO 2 in fermentation samples using acetaldehyde-bisulfite equilibria. This enabled estimation of free and molecular SO 2 concentration well below the detection limit of most analytical methods (i.e., < 2-3 mg/L free SO 2 ), which can occur during and after white wine fermentation. Although becoming unreliable as acetaldehyde:SO 2 molar ratios approach unity (approximately < 1.1:1.0), these data revealed that the initial survival of O. oeni starter cultures following co-inoculation with SO 2 -producing yeast strains could be linked to, and regulated by, a narrow threshold concentration of molecular SO 2 (approximately 1-35 µg/L). In particular, the second interaction experiment demonstrated that following co-inoculation with S. cerevisiae AWRI1497, initial exposure of malolactic starter cultures to only 4-6 µg/L molecular SO 2 was associated with considerable reduction in bacterial viability over the first few days of co-fermentation. On the other hand, following co-inoculation with S. cerevisiae AWRI2880, negligible loss in viability of O. oeni strains YV Select and VP41 occurred after exposure to an initial molecular SO 2 concentration of 1-2 µg/L. These findings suggest that molecular SO 2 concentrations almost 100-fold less than those previously reported by Ribéreau-Gayon et al.
(2006) (0.15-0.4 mg/L) may cause inhibition of O. oeni in wine. Further, such inhibition appears to be alleviated, and malolactic starter performance optimised, in wine conditions where molecular SO 2 is almost completely absent in the first few days of co-fermentation.
The extremely low concentrations of molecular SO 2 associated with O. oeni inhibition observed here support other reports regarding the SO 2 sensitivity of this organism. For example, Carr et al. (1976) noted that 0.032 mg/L molecular SO 2 in a culture medium (pH 4) was sufficient to completely suppress and kill 3 of 4 wild heterofermentative cocci (isolated from cider). Similarly, Hood (1983) demonstrated that very low calculated levels of free SO 2 (0.05-0.8 mg/L), which can exist in equilibria with acetaldehyde-bisulfite, caused a dose-response inhibition in L. brevis and Leuc. oenos growth in a culture medium (pH 4) containing 10, 20, and 30 mg/L total SO 2 . This author also reported that a calculated free SO 2 concentration of 0.8 mg/L (equivalent to 5 µg/L molecular SO 2 ) entirely suppressed the growth of Leuc. oenos LN6 in the same medium containing 30 mg/L total SO 2 .
The current work has also provided further insight into the adaptive capability of O. oeni to SO 2 . For example, in the pilot-scale experiment, exposure of O. oeni VP41 to an extremely low and incremental rise in molecular SO 2 concentration (up to approximately 10 µg/L) within the first 9 days of co-fermentation with S. cerevisiae NT 50 facilitated effective adaptation to molecular and total SO 2 concentrations of approximately 16 µg/L and 57 mg/L, respectively. This suggests that the adaptive capability of O. oeni towards SO 2 (Delfini and Morsiani, 1992;Guzzo et al., 1998) encompasses the molecular form of this compound. Overall, the evidence from this work suggests the survival and adaptation of O. oeni following co-inoculation with SO 2 -producing yeast strains requires that molecular SO 2 concentrations remain below a threshold inhibitory range in the early stages of co-fermentation, which exponentially increases from approximately 3-6 µg/L in the first three days of co-fermentation.
While evidence from the second interaction experiment indicates some strain-dependent reduction in O. oeni viability after exposure to ≥ 50-60 mg/L bound SO 2 , it is unclear whether such inhibition was caused by bound SO 2 itself, or extremely low concentrations (1-6 µg/L) of molecular SO 2 accumulated over the first 5 days of co-fermentation, or a combination of both. Manufacturer recommendations for O. oeni CH35 indicate its suitability for MLF induction in wines of up to 45 mg/L SO 2 , and those of O. oeni VP41 and YV Select indicate SO 2 tolerances of up to 60 mg/L and < 50 mg/L, respectively. Interestingly, Carr et al. (1976) report no indication of toxicity of acetaldehyde-bound SO 2 towards L. plantarum. However, further studies are needed to clearly define the tolerance and adaptive capabilities of O. oeni strains to molecular and bound forms of SO 2 under wine conditions.
Further insight into the potential effects of acetaldehydebound SO 2 on co-inoculated O. oeni was indirectly gained from the concentrations of acetaldehyde observed during the latter stages of co-fermentation. The toxicity of acetaldehyde-bound SO 2 , at least in part, requires the bacterial utilisation of acetaldehyde, and subsequent release of sufficient free SO 2 to affect growth and survival (Fornachon, 1963;Hood, 1983;Wells and Osborne, 2012). Moreover, the ability of wine LAB to degrade acetaldehyde and SO 2 -bound acetaldehyde (Jackowetz and Mira de Orduña, 2012;Jackowetz and Mira de Orduña, 2013;Osborne et al., 2000;Pan et al., 2011;Wells and Osborne, 2012) appears to coincide with bacterial growth (Fornachon, 1963;Hood, 1983;. In white wine, acetaldehyde catabolism can initiate after the commencement of MLF and continue for up to several weeks after its completion (Jackowetz and Mira de Orduña, 2012). However, in the current pilot-scale study, such metabolism was not evident in post-MLF stages, nor during MLF in wines co-inoculated with S. cerevisiae Actiflore Rosé. An absence of acetaldehyde catabolism may reflect differences in wine parameters, such as pH, that are critical to the growth and MLF capability of O. oeni.  reported that acetaldehyde degradation by O. oeni in white wine containing acetaldehyde-bound SO 2 occurred at pH 3.6 but not at pH 3.3. Similarly, Flamini et al. (2002) indicated that following MLF with two different O. oeni strains, substantial acetaldehyde degradation occurred in Cabernet-Sauvignon wine (pH 3.55) but not in Chardonnay (pH 3.24). Hence, it appears unlikely that acetaldehyde metabolism is sufficiently active to account for the inhibition of O. oeni immediately following co-inoculation in low pH Chardonnay wine observed in the current study.
In addition to affecting the survival and MLF capability of O. oeni during co-fermentation, the pilot-scale study demonstrated that the choice of co-inoculation yeast strain also had a major impact on the production of diacetyl and on wine sensory properties. The diketone diacetyl is an important flavour compound imparting buttery aroma and flavour attributes, and can arise from LAB metabolism during MLF (Bartowsky and Henschke, 2004). Yeast can also synthesise and degrade diacetyl (Martineau and Henick-Kling, 1995;Mink et al., 2014) and, although the diacetyl concentration in newly fermented wine is typically low (Bartowsky and Henschke, 2004), some variation may occur depending on factors including yeast lees contact, SO 2 addition and grape variety (Martineau and Henick-Kling, 1995). For Chardonnay wines without MLF, Martineau et al. (1995b) reported an average diacetyl concentration of 0.22 (SD = 0.20) mg/L. The presence of yeast during and after MLF will also influence diacetyl content (Bartowsky and Henschke, 2004;Martineau and Henick-Kling, 1995).
Since Chardonnay wines in this study were co-fermented and left in contact with yeast lees for one week after MLF, the absence of diacetyl (< 0.1 mg/L) in wines co-inoculated with the low SO 2 -producing yeast strains AWRI796 and CY3079 is not unexpected. However, the higher concentrations of diacetyl in wines coinoculated with the high SO 2 -producing S. cerevisiae strains, particularly NT 50 (4.9 (SD = 0.6) mg/L), is surprising, and may relate to diacetyl binding and retention at elevated SO 2 concentration.
The observations regarding diacetyl were also reflected in wine sensory properties, whereby wines co-inoculated with high SO 2 -producing yeast strains Actiflore Rosé and NT 50 had higher citation frequency for butter and woody attributes, and those co-inoculated with NT 50 had the highest citation frequency for butter. These results align with the sensory detection threshold for diacetyl in Chardonnay wine (0.2 mg/L, Martineau et al. (1995a)). However, while a bacterial origin of accumulated diacetyl is likely, a potential contribution from yeast (Martineau and Henick-Kling, 1995) cannot be discounted and requires investigation.

CONCLUSIONS
This study confirms that S. cerevisiae strain choice can affect the survival of O. oeni and the conduct of MLF during co-inoculation. Immediately following bacterial inoculation, the early stages of co-fermentation provide a unique and metabolically dynamic environment, the composition of which is critical for bacterial survival and MLF efficiency.
Significantly, for co-inoculation with SO 2 -producing yeast strains, laboratory-and pilot-scale studies have demonstrated a direct association between the survival of O. oeni and the molar ratio of acetaldehyde:SO 2 in the early stages of AF. Alleviation of SO 2 inhibition of O. oeni and improved MLF efficiency was associated with an early, transient accumulation of a high concentration of acetaldehyde.
By applying relevant acetaldehyde-bisulfite equilibria, the survival and adaptation to SO 2 by O. oeni starter cultures following co-inoculation appear to be regulated by an extremely low concentration range of molecular SO 2 (approximately 1-35 µg/L). Bacterial survival coincided with molecular SO 2 concentrations remaining below a minimum inhibitory range, which increased exponentially from 3-6 µg/L in the first 3 days of co-fermentation. Although some O. oeni strains may exhibit sensitivity to bound SO 2 , the mechanism of such inhibition during co-fermentation remains unclear.
Moreover, the possible role of other mechanisms, such as medium chain fatty acid formation, should not be discounted to fully explain the complex interactions that may occur between yeast and malolactic bacteria during co-fermentation. Reliance on total/bound SO 2 concentration may only provide an indirect and potentially inaccurate measure of SO 2 toxicity towards O. oeni for co-inoculation. Equilibrium concentrations of molecular SO 2 should also be considered. Knowledge of both the SO 2 -and acetaldehyde-producing potential of yeast strains are, therefore, important selection criteria for co-inoculation, particularly for white musts/wines in which elevated SO 2 concentrations are anticipated.
The choice of co-inoculation yeast strain also influenced wine diacetyl concentration and sensory properties. Co-inoculation with high SO 2 -producing yeast strains produced wines with higher diacetyl concentration. These wines had a higher citation frequency for butter attributes.

DATABASE
Access to all data from the study is provided in the following on-line repository file: https://datadryad.org/stash/share/ a7q7umBAe6yKC_zHKzd5uJfrWwyUo6c1al0LwCLzakQ