Effects of Yeast Strain and Juice Nitrogen Status on Glutathione Utilisation during Fermentation of Model Media

,


Introduction
Glutathione (GSH) is a naturally occurring antioxidant present in grapes, wines, and other living organisms such as yeast [1].It plays an important role in winemaking, preserving wine colour and aroma by reacting sacrifcially with quinones, and acting as an antioxidant [2,3].As a result, GSH has been investigated as a potential alternative to sulfur dioxide (SO 2 ) for the preservation of wine freshness and aroma intensity [4].Glutathione may contribute to the sensory attributes of wine aroma through the liberation of hydrogen sulfde (H 2 S) during fermentation [5,6], a process that is also dependent on the nitrogen status of the juice [7].Elevated residual GSH concentrations in fnished wine have also been shown to contribute to H 2 S accumulation postbottling [8,9].Especially in combination with copper, GSH creates an environment that favours the accumulation of volatile sulfur compounds (VSCs), such as H 2 S and methanethiol (MeSH) [9].Glutathione may also contribute to the sensory attributes of wine by reacting with unsaturated aldehydes to produce varietal thiol precursors, which could be cleaved during fermentation to liberate potent thiols that modulate the varietal character of wines such as Sauvignon Blanc [10].
Aside from naturally occurring GSH in grapes, other potential sources in wine include yeast-derived products, some of which are marketed as GSH-enriched inactive dried yeast preparations.Compared to grape processing interventions that can preserve GSH [11], however, the contribution of these additives to juice or wine is minimal, increasing GSH concentrations between 1 and 3.5 mg/L when used at recommended addition rates [12][13][14].Te environmental nutrient status may infuence the uptake of GSH by yeast although contradicting results for the scale of impact are found in the literature.Kritzinger et al. [15] found that increased YAN concentration did not result in changes in GSH concentration, which suggests that neither GSH synthesis nor the metabolism of GSH by the yeast was infuenced by the nutrient status of the must.However, Park et al. [16] found that GSH production was directly correlated to the total nitrogen and YAN content of the grape juice in their experiment.Te fundamental diferences in experimental design between the two studies may explain the contradicting reports, with Kritzinger et al. [15] adjusting the YAN concentration of a single must, whereas Park et al. [16] used various grape musts that contained diferent natural YAN levels.In a complementary study, Wegmann-Herr et al. [6] evaluated the efects of added diammonium phosphate (DAP) and GSH on sulfur metabolism by yeast and found that GSH addition (40-50 mg/L) did not increase VSC when a sufcient nitrogen pool was provided such that it was not necessary for the yeast to metabolise the added GSH.
In 2015, the International Organisation of Vine and Wine (OIV) passed two resolutions relating to the addition of GSH during wine production (OIV-OENO 445-2015 and OIV-OENO 446-2015).Tese resolutions state (1) OIV-OENO 445-2015 (i) It is advised to add the glutathione while obtaining the musts or at the start of alcoholic fermentation, ensuring that, prior to and during alcoholic fermentation, the assimilable nitrogen level is sufcient to avoid the metabolism of glutathione by the yeast (ii) Te dose used should not exceed 20 mg/L (2) OIV-OENO 446-2015 (i) Te addition of glutathione on storage and/or packaging, including the bottling of wine, is recommended (ii) Te dose used should not exceed 20 mg/L (iii) Te glutathione must be in a reduced form and comply with the prescriptions of the International Oenological Codex Te objectives of these resolutions are to limit the intensity of oxidation phenomena in must or wine due to the ability of GSH to act as an antioxidant, thereby preventing browning and protecting aroma compounds in wines (and particularly those of the thiol family) from oxidation.Notably, oxidised and 'reduced' aroma faults are two of the most common wine of-aromas that result in decreased shelf life and are directly infuenced by the amount of oxygen a wine is exposed to during bottle ageing [17].Depending on a wine's composition and the concentration of antioxidants such as SO 2 and GSH, trace amounts of the short-chain unsaturated aldehydes (methional and phenylacetaldehyde, among others) may be produced, imparting aromas described as sherry, honey, bruised apple, and others associated with oxidation spoilage [18].Increased oxygen exposure can also result in the loss of tropical aromas associated with varietal compounds, particularly certain polyfunctional thiols including 3-sulfanylhexan-1-ol (3-SH), 3-sulfanylhexyl acetate (3-SHA), and 4-methyl-4-sulfanylpentan-2-one (4-MSP) [19].On the other hand, depending on wine composition, 'reductive' characters may develop during bottle ageing of wines that have been completely protected from oxygen during storage [9].Tese aroma characters are mainly ascribed to the presence of H 2 S and MeSH, which impart aromas of rotten egg, sewage, burnt rubber, and putrefaction [20].As a result, the addition of GSH at diferent stages of winemaking may have profound implications on the formation of volatile compounds that contribute positively or negatively to wine sensory properties.
Considering that GSH may become a regulated additive used during vinifcation, it was important to establish an accessible method for the analysis of GSH in reduced and oxidised (GSSG) forms that can be standardised for research and commercial laboratories, and wineries.Te conventional methods for quantitating GSH and GSSG in tissue, plasma, and liquid matrices from human, animal, or plant material include high-performance liquid chromatography (HPLC) coupled to ultraviolet (UV) [21,22], fuorometric [23][24][25], electrochemical [26], and mass spectrometric [27][28][29] detectors.Enzymatic assays have also been utilised [30], but due to its enhanced sensitivity, HPLC with tandem mass spectrometry (HPLC-MS/MS) is preferred for the analysis of blood and tissue samples, where the requirements for GSH and GSSG quantifcation can be as low as a few ng/L [31].Te use of HPLC-MS/MS for the analysis of wine and juice samples has also been widely implemented [32,33].Methods relying upon MS detection have the added beneft of being able to identify unknowns and GSH reaction products (if used in full scan mode).However, HPLC with UV detection is more accessible and completely adequate for quantifying typical GSH concentrations in juice, must, or fermenting wine, and when adding GSH or GSH products to ferments or for storage/packaging according to OIV regulations.Given that GSH can oxidise to GSSG during sampling and handling, it is necessary to derivatise GSH before analysis to preserve the natural ratio of GSH to GSSG.A derivatisation step has been included in some methods [27,31,34], and several derivatisation agents have been used, including p-benzoquinone [27], N-ethylmaleimide (NEM) [34][35][36], and 4,4-dithiodipyridine (DTDP) [37,38].None of the methods described above simultaneously analysed GSH and GSSG using simple HPLC with UV detection, with initial derivatisation of GSH to prevent autoxidation to GSSG during the sample preparation process.
Tis study aimed to address a fundamental question regarding whether a level of juice nitrogen could be defned that prevented both the metabolism of GSH and the production of VSCs by yeast during alcoholic fermentation.To address this question, GSH was added to chemically defned 2 Australian Journal of Grape and Wine Research media (CDM) with increasing yeast assimilable nitrogen (YAN) concentrations, and fermentation was conducted with yeast strains that varied in their nitrogen uptake potential.A straightforward method involving NEM derivatisation and HPLC analysis with UV detection was developed to quantify GSH and GSSG simultaneously after GSH addition to juice, to monitor the changes in GSH concentration during fermentation, and to monitor for autoxidation of GSH to GSSG. ).An eight-point calibration curve was prepared, in duplicate, for the expected concentration range of GSH (0-125 mg/L; 0-0.41 mM) and GSSG (0-125 mg/L; 0-0.20 mM), and a six-point calibration curve was prepared in duplicate for high range GSH concentrations (70-350 mg/ L, 0.23-1.14mM) to quantify GSH at the levels added to the ferments in this experiment.Te matrix used to prepare the calibration levels was the same as that of the samples being analysed (i.e., calibration standards prepared in CDM for CDM and fermenting samples, and model wine for samples at the completion of fermentation).Calibration standards were prepared by adding an aliquot of CDM or model wine to 2 mL vial, followed by the addition of 50 μL of NEM solution (5 mM), and then an aliquot of GSH working solution specifc to each calibration point was added.Te range of the calibration curve was as follows: 0.00, 0.016, 0.033, 0.081, 0.16, 0.24, 0.33, and 0.41 mM.Te sample was vortexed for 15 s and left for 5 minutes to derivatise at room temperature.A volume of GSSG working solution specifc to each calibration point (0.00, 0.008, 0.016, 0.041, 0.082, 0.12, 0.16, and 0.20 mM) was then added, followed by 75 μL i-PrOH, with vortexing for 15 seconds.Te total volume of the sample was 1.5 mL.Calibration samples prepared for the high range GSH curve (GSH: 0.00, 0.23, 0.33, 0.81, 0.98, 1.14 mM) were prepared in the same way and then diluted with 300 μL of respective model solution before analysis.A sample of derivatised GSH (1 mM) was prepared and analysed with each set of samples to monitor the GSH working solution for autoxidation to GSSG.

Sample Preparation.
Model juice, fermentation, and wine samples were centrifuged for 5 min before 1450 μL of the supernatant was added to a 2 mL HPLC vial, followed by the addition of 50 μL of NEM solution (5 mM).Te sample was vortexed for 20 seconds before the addition of 75 μL i-PrOH.

Analysis.
High-performance liquid chromatography (HPLC) was performed with an Agilent 1100 instrument (Forest Hill, Victoria, Australia) equipped with a binary pump, degasser, autosampler, column oven, and an Agilent 1200 diode array detector (DAD).A Cosmosil HILIC column (250 mm × 2.0 mm ID, 5 μM, Nacalai Tesque, Inc. Kyoto, Japan) protected by a guard column packed with the same material.Te solvents were (A) 0.1% FA in MilliQ water and (B) 0.1% FA in ACN.Two gradient profles and fow rates were used based on the matrix.A sample volume of 8 μL was directly injected into the HPLC system.Te column oven temperature was 35 °C.For CDM and ferment matrices, a fow rate of 0.5 mL was used with the linear gradient for solvent B as follows: 0 min, 90%; 5 min, 90%; 10 min, 75%; 20 min, 20%; 28 min, 20%; 28.1 min, 90%, followed by 5 min of column equilibration with 90% B. For model wine, a fow rate of 0.7 mL/min was used with the linear gradient for solvent B as follows: 0 min, 90%; 5 min, 90%; 6 min, 75%; 15 min, 20%; 22 min, 20%; 22.1 min, 90%, followed by 5 min of column equilibration with 90% B. An injection volume of 4 µL was used for high-range standards and samples.Te GSH derivative and GSSG were detected at 220 and 200 nm, respectively.

Linearity, Accuracy, Recovery, and Precision.
Linearity was assessed by comparison of three calibration curves prepared on three separate occasions in CDM, juice, real wine, and model wine matrices.Intraday accuracy was Australian Journal of Grape and Wine Research determined by evaluating the percentage recovery of three replicate samples spiked with GSH and GSSG at low, medium, and high concentrations (GSH: 0.0325, 0.163, and 0.325 mM; GSSG: 0.0163, 0.0816, and 0.163 mM) in each matrix.Interday accuracy was evaluated by analysing triplicate samples prepared on three separate occasions.Precision was expressed using relative standard deviation (RSD) (Table S1, Supporting Information).Te limit of detection (LOD) and limit of quantitation (LOQ) were determined as LOD � 3.3σ/S and LOQ � 10σ/S, where σ is the standard deviation of the response, and S is the slope of the calibration curve [40].

Stability during Storage.
Calibration standards were prepared in model wine and Chardonnay juice on two separate occasions.Te model wine samples were analysed immediately after derivatisation, then stored at −20 °C for 10 days, and brought to room temperature (20 °C) and reanalysed.Nonlinear regression analysis was used to compare the curve ft of the data sets to assess whether there were any signifcant diferences between samples analysed immediately following preparation compared to samples that were stored and reanalysed after 10 days (Figure S3).Two sets of Chardonnay fermentation samples were simultaneously sampled from four ferments (prepared in triplicate, as described in the "Preparation of ferments" section below) that contained a diferent added level of GSH (0, 100, 250, 500 mg/L; 0, 0.163, 0.184, 1.627 mM) (Figure S4).One set of the four fermentation triplicate samples was derivatised along with calibration standards and analysed immediately.Te second set was derivatised with its own calibration standards, and together, these were stored at −20 °C, thawed, and analysed as a batch 10 days later (Figure S4).

Efects of pH on Derivatised GSH and GSSG
Quantitation.Model wine was adjusted to pH 3, 5, and 7 using formic acid or sodium hydroxide.A set of duplicate calibration standards of GSH (0, 0.5, 1.0, and 1.5 mM) and GSSG (0, 0.5, 1.0, and 1.5 mM) was prepared at each pH, and nonlinear regression analysis was performed to determine whether one curve could adequately ft all three datasets prepared at diferent pH levels.Additionally, four samples of derivatised GSH (1 mM) and GSSG (1 mM) were prepared at pH 3, 5, and 7 and analysed at four time points over 15 h to determine the stability of derivatised GSH and GSSG (Figure S5).

Preparation of Ferments. Saccharomyces cerevisiae
AWRI 1688 and AWRI 2861 were obtained from the Australian Wine Research Institute (AWRI) culture collection.Te CDM was prepared as described by Schmidt et al. [39] without the addition of sterols and fatty acids.Te volume of nitrogen stock mix used was adjusted to obtain the desired low, medium, and high fnal YAN concentration in the media.YAN concentrations were determined by Afnity Labs (Adelaide, Australia) to be 155 mg/L (low), 320 mg/L (medium), and 430 mg/L (high) using the methods described in Dukes and Butzke [41] and Bergmeyer and Beutler [42].L-GSH was dissolved in sterile MilliQ water and added prior to inoculation to the required concentration (0 mg/L, 20 mg/L, 100 mg/L, and 250 mg/L; 0, 0.0651, 0.325, 0.814 mM) for each YAN level and yeast strain.Te sterile medium was inoculated with mediumacclimatised liquid culture at a starting rate of 0.02 absorbance units measured at 600 nm (Beckman Coulter DU530 UV Vis spectrophotometer, Beckman, NSW, Australia).Small-scale fermentations (200 mL) were conducted in triplicate in 250 mL vessels with a sample port, a 25-2000 ppm H 2 S detection tube (Kitagawa, Japan), and a 13.8 kPa directional-fow check valve to protect against overpressure.Te vessels were stirred at 350 rpm and incubated at 18 °C.Yeast growth was monitored by measuring absorbance at 600 nm, and fermentation progress was monitored using an enzymatic assay for reducing sugars as previously described [43] with adaptations [44] to allow the measurement in a 96-well microplate format [39].Upon completion (residual sugar <1 g/L), ferments were coldsettled for 3 days at 4 °C prior to compositional analysis.

Analysis of Volatile Sulfur Compounds.
Volatile sulfur compounds were analysed directly after primary alcoholic fermentation and then after bottling, using gas chromatography with sulfur chemiluminescence detection (GC-SCD) as previously described [20].Duration of ferments as defned by the time (days) required for residual sugar concentration to reach 2 g/L was interpolated from a curve ft to the sugar concentration data.A least-squares approach was used to ft an asymmetric sigmoidal equation, weighted by 1/y 2 , and constraints: hillslope <0, top � 200, bottom � 0. Each ferment was ft independently, and the estimate of x (days) was calculated for y � 2. Two-way ANOVA of "time to completion" estimates was used to evaluate whether the duration of ferments with diferent concentrations of added GSH could be distinguished at diferent YAN concentrations.Each strain was evaluated independently.With limited interaction between GSH and YAN and limited contribution from GSH alone, Tukey's multiple comparison test was used to evaluate simple within-row efects with individual variances computed for each comparison.

Statistical
Accuracy is presented as % recovery, calculated according to the following equation: accuracy(%) � (y/x) × 100 where y � mean measured value and x � expected value.Precision is given as the relative standard deviation which is calculated as follows: precision(%) � (s/y) × 100, where s � standard deviation of the measured values and y � mean measured value.Te level of statistical evidence based on P-value is defned as follows: P ≥ 0.10, "no evidence"; P ≤ 0.10, "weak evidence"; P ≤ 0.05, "evidence"; P ≤ 0.01, "strong evidence"; P ≤ 0.001, "very strong evidence."

Analysis of GSH and GSSG in Juices and Wines by HPLC.
A robust HPLC method for quantifcation of GSH and GSSG (Figures 1(a) and 1(c)) was developed based on UV detection.Derivatisation of GSH using NEM to produce Nethylsuccinimido-S-glutathione (Figure 1(b)) was used for analyte stability, thus avoiding GSH oxidation while awaiting analysis and preventing under-estimation of GSH and over-estimation of GSSG [31].Enhanced stability was deemed useful to permit the storage of samples prior to analysis, increasing the accessibility of the method to wine analysis laboratories, analysis centres, and research laboratories.
Hydrophilic interaction chromatography (HILIC) was used for the separation of GSH and GSSG in the present study due to its strong retention of polar analytes [45].Te relationships between GSH and GSSG concentration with chromatogram peak area were evaluated in CDM, juice, and model and real wine.
Two GSH and GSSG concentration ranges were assessed, depending on the matrix being analysed.One calibration set was developed to cover the expected concentrations for GSH and GSSG in CDM and fermenting samples (5-125 mg/L) in accordance with previously reported ranges for juice and wine [13,15,24,33,46,47].A high-range calibration set covering the concentration range of 70-350 mg/L (0.228-1.139 mM) was also evaluated.Te eight-point calibration functions were linear for both concentration ranges (R 2 > 0.99) in each matrix (Table S1).Te LOD and LOQ values were similar for GSH and GSSG in the "expected" range calibration set for model wine and CDM, ranging from 2.80-3.22mg/L (0.0046-0.0053 mM) and 6.36-7.32mg/L (0.0207-0.0238 mM), respectively (Table S1).
Inter-and intraday accuracy and precision were determined for low (0.0325 mM), medium (0.163 mM), and high (0.325 mM) concentrations of GSH and GSSG (Table S1).Accuracy (99.1-103.1%)was equivalent for medium and high concentrations of GSH and GSSG in model wine and CDM.Measures of GSSG were less precise (0.7-5.0%) than GSH (0.4-1.6%) at these medium and high concentrations (Table S1).Measurement of low GSH and GSSG concentrations (0.0325 mM) in model wine and CDM was both less accurate than medium and high concentrations with GSH tending to be underestimated (accuracy 80.0-96.6%)and GSSG tending to be overestimated (accuracy 108.6-115.8%)(Table S1).Te accuracy and precision of mid and high concentrations of GSH and GSSG determined using calibration curves prepared in Chardonnay and Shiraz wine were equivalent to GSH and GSSG concentrations determined using model solutions (Table S1).However, low concentrations of GSH were underestimated when calibration curves were prepared in Chardonnay and Shiraz juice, whereas low concentrations of GSSG were overestimated when lower concentrations were prepared either in Chardonnay and Shiraz juice or wine.

Stability of Derivatised GSH and GSSG.
Te stability of derivatised GSH and GSSG in samples stored at room temperature (20 °C) was evaluated by preparing calibration standards in CDM, model wine, and Chardonnay juice and wine.Tere was no evidence (α � 0.05) that calibration samples stored for 24 h (CDM, juice, and wine) or up to 5 days (model wine) produced diferent regression models (Figures S1 and S2).However, calibration standards prepared in juice and analysed immediately showed changes in peak area relative to stored samples (Figures S3c and S3d), with strong evidence that regression models for fresh and stored samples were diferent (P < 0.0001 for both).Nonetheless, there was no evidence that storing samples resulted in a diferent calculated concentration when freshly prepared samples (with GSH added at 100 and 250 mg/L) and calibration standards in CDM were compared to a replicate set that was stored at −20 °C and analysed after 10 days (Figures S4a and S4b).
Overall, analysis of freshly prepared samples is advisable, but if that is not possible, then storing samples at −20 °C for later analysis can be considered if GSH concentrations are less than 100 mg/L (0.325 mM).Furthermore, a set of calibration standards should be prepared and stored with the samples for subsequent analysis in the same batch.
Te efect of matrix pH on the stability of the derivatised GSH and GSSG was also evaluated.Tere was no evidence that pH afected the quantifcation of GSH (P � 0.857) or GSSG (P � 0.720) when assessed within the timeframe of this experiment (Figures S5a and S5b).

Role of Juice Nitrogen Status on GSH Consumption and VSC Liberation.
According to the OIV resolution pertaining to the use of GSH in wine production, juice should have an "assimilable nitrogen level sufcient to avoid the metabolism of GSH by the yeast."In the absence of any other reports, it was unclear what a sufcient nitrogen concentration might be.As such, an evaluation of GSH consumption potential was undertaken by determining GSH and GSSG concentrations by HPLC upon fermentation of CDM prepared with three YAN concentrations (155 mg/L, 320 mg/L, and 430 mg/L) and four GSH concentrations (0, 20, 100, and 250 mg/L; 0, 0.0651, 0.325, and 0.814 mM) for each assimilable nitrogen level (total of 12 treatments).
Te total amount of GSH remaining after fermentation was considered an indicator of the degree of GSH-related yeast metabolic activity.Two yeast strains were used in this work, with AWRI 1688 (an isolate of Zymafore VL3) chosen because it contains an expanded set of oligopeptide Australian Journal of Grape and Wine Research transporter genes (FOT1 and FOT2) on a section of a horizontally transferred genomic island known as Region C. Tese transporters have been shown to permit the assimilation of complex nitrogen sources such as GSH [48].Te other yeast strain was AWRI 2861, selected as a comparison due to its genetic similarity to VL3 [49] but lack of FOT1 or FOT2.
As anticipated, YAN concentration was a primary mediator of the sugar consumption rate for both yeast strains (Figure 2).It is known that YAN concentration can range from 53-411 mgN/L (average of 175 mgN/L) and that YAN content of must correlates to maximum CO 2 production rate and thus provides information on fermentation health and duration [50].Fermentations were slow at 155 mgN/L, taking an average of 12.5 days (SD 1.2) for AWRI 1688 and 19.9 (SD 1.4) days for AWRI 2861 to reach a target sugar concentration of less than 2 g/L.Tis was consistent with the common understanding that 155 mg/L YAN is the lower limit for reliable white wine fermentations [51].Tere was strong evidence that YAN concentration was the major factor responsible for variations in fermentation time (P < 0.0001 for both strains) (Tables S2 and S3).Te addition of   Australian Journal of Grape and Wine Research either 320 or 430 mgN/L decreased the respective fermentation durations by 6.6 days, 95% CI [5.4,7.9] and 7.6 days, 95% CI [6.3, 8.8] for AWRI 1688 and by 7.8 days, 95% CI [9.9, 10.7] and 8.9 days, 95% CI [6.0, 11.8] for AWRI 2861 (P < 0.0001).Tere was no evidence for diferences in fermentation time between the two higher YAN conditions.In fermentations using yeast strain AWRI 2861, fermentation time appeared unafected by GSH addition (P � 0.321), nor was there evidence for an interaction between GSH and YAN (P � 0.152); however, with AWRI 1688, there was weak evidence that supplementing the low YAN medium with GSH afected fermentation time (P � 0.082) and strong evidence for an interaction between GSH and YAN (P � 0.002) (Tables S2 and S3).Te efects of GSH on fermentation time were only apparent at low YAN concentration (155 mg/L) and higher GSH addition rates [100 and 250 mg/L (0.325 and 0.814 mM)].Te magnitude of the efect was small, however, compared to the efects of YAN.Mean diferences in ferment duration of 2.8 days, 95% CI [1.0, 4.6], and 1.9 days, 95% CI [0. 1, 3.7] were observed between control ferments and those with 100 mg/L and 250 mg/L GSH addition, respectively.Te altered fermentation time in response to GSH addition suggested that AWRI 1688 used GSH as a nitrogen source under nitrogenlimited conditions but only when GSH concentrations were high.Te uptake of GSH by AWRI 1688 was consistent with both its genetic background, having an expanded set of oligopeptide transporters relative to AWRI 2861, and with the previously reported ability of FOT1-2 to mediate GSH uptake by yeast [48].
During these experiments, the change in total GSH concentration was monitored by quantifying and summing oxidised and reduced GSH upon completion of alcoholic fermentation.Te total "end of ferment" GSH concentration for the two yeast strains is shown in Figure 3. End of ferment GSH concentrations were compared with initial GSH concentration using one-way ANOVA and Dunnett's test (Table S4).Tis data provided strong evidence that GSH concentration decreased during fermentation regardless of the initial YAN or GSH concentration, or yeast strain used for fermentation.In contrast, small increases in GSH concentration were observed in ferments without added GSH.Tere was strong evidence for both yeast strains that the magnitude of GSH loss was greater in ferments with higher initial GSH concentrations, especially at low initial YAN concentrations (P < 0.0001).In low YAN/low GSH ferments of AWRI 1688 (155 mg/L, 0.073 mM GSH), the mean decrease in total GSH concentration was 0.043 mM, 95% CI [0.022, 0.063], whereas, in low YAN/high GSH ferments (155 mg/L, 0.798 mM GSH), the mean decrease in total GSH concentration was 0.246 mM, 95% CI [0.181, 0.311].Notably, loss of GSH during fermentation was inversely related to the initial YAN concentration.Again, in fermentations by AWRI 1688 with the highest initial GSH concentration (0.798 mM GSH), there was strong evidence (P < 0.0001) that an increase in YAN concentration from 155 mg/L to 430 mg/L was associated with a mean decrease in the postferment total GSH concentration of 0.140 mM, 95% CI [0.075, 0.206].Nevertheless, Australian Journal of Grape and Wine Research even at the highest YAN and lowest initial GSH concentrations, measurable decreases in GSH concentration were observed (mean decrease of 0.024 mM, 95% CI [0.004, 0.044], P = 0.025), such that GSH metabolism by yeast could not be ruled out.
GSH metabolism by yeast may manifest as an increase in H 2 S production during active fermentation.Te yeastspecifc production of H 2 S is shown in Figure 4 with AWRI 1688 and AWRI 2861 difering in both their overall production of H 2 S during fermentation and in their response to diferent nitrogen concentrations.Tere was evidence that high preferment concentrations of GSH (250 mg/L, 0.814 mM) increased the total production of H 2 S in low YAN conditions relative to ferments where no GSH addition was made for strain AWRI 2861 (P = 0.02, Table S5).Te imprecise nature of the tube-based H 2 S measurement rendered it impossible to draw more fnegrained conclusions about the efect of GSH treatment on H 2 S production.Nevertheless, this apparent trend was consistent with previous studies [52].Both yeast strains responded to higher YAN concentrations with decreased H 2 S production.Overall, AWRI 2861 produced less H 2 S than AWRI 1688 at any YAN level.Indeed, H 2 S production by AWRI 2861 was only detected in low YAN ferments.Te lower overall production of H 2 S by AWRI 2861 was consistent with its genetic background, carrying a heterozygous R301G mutation in MET2, a dominant allele contributing to the decrease in H 2 S production in such strains [53].
Tere is strong evidence that high GSH addition rates (100 and 250 mg/L (0.325 and 0.814 mM)) increased postferment H 2 S concentration in low nitrogen fermentations (155 mg/L) (Figure 5, Table S6), with mean H 2 S concentration increases of 35 μg/L (P � 0.013) and 42 μg/L (P � 0.002) (AWRI 1688) and 32 μg/L (P � 0.01) (AWRI 2861 with 250 mg/L GSH).Te efects of GSH addition were most evident in ferments conducted using strain AWRI 1688  Australian Journal of Grape and Wine Research although similar trends were observed with strain AWRI 2861.Tere was no evidence that a 20 mg/L GSH addition (0.0651 mM), as recommended in OIV-OENO 445-2015, increased postfermentation H 2 S concentrations.Tese data suggested that adding a low concentration of GSH to a fermentation would not be sufcient to stimulate H 2 S formation, regardless of whether the added GSH is metabolised by yeast.Postfermentation levels of MeSAc were also afected by YAN and GSH concentrations (Figure 5).Generally, higher concentrations of MeSAc were observed at the end of fermentations with the lowest YAN, and greater amounts of MeSAc were produced in response to increased GSH additions under this YAN condition.AWRI 1688 produced more MeSAc than AWRI 2861 at all YAN concentrations.Two-way ANOVA provided strong evidence (P � 0.001 and 0.002 for AWRI 1688 and AWRI 2861, respectively) for an interaction between GSH and YAN that contributed to the overall strain-specifc variation in MeSAc concentration, with YAN concentration being the dominant factor for both AWRI 1688 (P < 0.0001) and AWRI 2861 (P < 0.0001) (Figure 5, Table S7).Tere was no evidence for GSH-related alterations in MeSAc concentration at higher initial YAN concentrations.
Tese experiments do not enable conclusions to be drawn about the causes of GSH concentration decrease.Both yeast metabolic activity and reaction with medium components could have contributed to the decrease in GSH concentration.However, it is noteworthy that these experiments were undertaken in a defned medium with limited extraneous material that could interact with GSH (e.g., phenolic compounds) and that both oxidised and reduced GSH were quantifed.Terefore, it is plausible that contrary to the stipulation in the OIV recommendation, no YAN condition exists that can avoid the metabolism of GSH by yeast during fermentation, especially considering that yeast produce GSH through their own metabolic action, which may be accompanied by the formation of undesirable VSCs.Depending on yeast strain and YAN status of must or juice, the best-case scenario may be that sufciently low concentrations of GSH do not stimulate the production of VSCs at concentrations that are detrimental to wine quality.

Conclusion
A fundamental hypothesis was tested in this work regarding whether a level of juice nitrogen could be defned that prevented the metabolic consumption of GSH by yeast.To Australian Journal of Grape and Wine Research address this gap and provide evidence or otherwise for the adoption of GSH addition during winemaking, GSH was added to defned media with increasing YAN concentrations, and the change in GSH concentration was monitored.Facilitating this, a robust HPLC method with UV detection was developed to quantify GSH and GSSG in juice, ferments, and fnished wines, using GSH derivatisation to preserve the levels of each analyte in existence at the time of sampling.
It was demonstrated that increasing the YAN status of the media minimised the decrease in GSH concentration during fermentation, but the loss of GSH could not be entirely eliminated.Decreases in GSH concentration were still evident with a low initial GSH concentration and high initial YAN concentration, with both of the yeast strains trialled.However, higher YAN concentrations were found to suppress H 2 S formation when higher concentrations of GSH were applied.Tis work raises doubt about the possibility of implementing a level of juice nitrogen that is sufcient to prevent the metabolism by yeast of GSH added prior to fermentation.12  Te second set of samples was also derivatised with NEM and then stored at −20 °C for nine days before analysis (red bars).One-way ANOVA was used to determine whether the freshly prepared and analysed samples were signifcantly diferent from samples there were prepared, stored at −20 °C, and analysed nine days later.Statements ascribing the level of statistical evidence in this report are as follows: P ≥ 0.10 "no evidence" (ns); P ≤ 0.10 "weak evidence" ( * ); P ≤ 0.05 "evidence" ( * * ), P ≤ 0.01 "strong evidence" ( * * * ), and P ≤ 0.001 "very strong evidence" ( * * * * ).
Figure S5.Evaluating the efect of pH on derivatised GSH (a, c) and on GSSG (b, d) quantifcation at pH 3, 5, and 7. Points are the means of duplicate samples with error bars showing standard deviation.Te bar graphs in (c) and (d) represent four repeat analyses of the derivatised GSH (1mM) (c) and GSSG (1 mM) (d) samples over 15 hours to evaluate their stability at each pH level.Te R 2 and P-value describing the comparison of ft for the reanalysed calibration curves are shown for each dataset in each plot.Te dotted lines represent the 95% confdence bands that indicate how well the data defne the best-ft curve.
Analyses.Signifcance tests (repeated measures ANOVA with Tukey mean comparison or Dunnet's mean comparison, nonlinear regression analysis, linearity, analysis of covariance, curve ft, derivatisation at diferent pH levels) were conducted using Prism statistics software (v6.04 GraphPad Software Inc., La Jolla, CA) at a signifcance level of α � 0.05.Values are represented as means and standard deviation (SD).Where mean diferences are presented, they are accompanied by 95% confdence intervals given in square brackets [lower CI, upper CI].Where appropriate, multiplicity adjusted P-values are reported.
hours, (d) model wine analysed two times 5 days apart, and (e) Chardonnay wine analysed three times every 12 hours.Te R2 and P-value describing the comparison of ft for the reanalysed calibration curves are shown for each dataset in each plot.Te dotted lines represent the 95% confdence bands that indicate how well the data defne the best-ft curve.FigureS3.Evaluating the stability of derivatised GSH (a) and GSSG (b) prepared in model wine and derivatised GSH (c) and GSSG (d) prepared in Chardonnay juice.Te samples were analysed immediately after preparation and again after 10 days of storage at −20 °C.Te R2 and P-value describing the comparison of ft for the reanalysed calibration curves are shown for each dataset in each plot.Te dotted lines represent the 95% confdence bands which indicate how well the data defne the best-ft curve.FigureS4.Te stability of derivatised GSH (a) and GSSG (b) measurements was evaluated in ferments prepared in chemically treated with three levels of GSH of 100, 250, and 500 mg/L (0.33, 0.81, 1.63 mM), and a control sample without added GSH.One set of samples was taken and derivatised with NEM and immediately analysed (blue bars).

Table S1 .
Figures of merit for GS-NEM andGSSG for chemically defned media, juice, wine, and model wine.TableS2.Two-way ANOVA analysis of the efect of initial YAN and GSH on ferment duration by yeast strain AWRI1688.Post hoc test employed Tukey's multiple comparison test evaluating simple efects between GSH treatments (within-row efects) at each YAN concentration (within YAN).Contrast specifc P-values are adjusted for multiple comparisons.TableS3.Two-way ANOVA analysis of the efect of initial YAN and GSH on ferment duration by yeast strain AWRI2681.Post hoc test employed Tukey's multiple comparison test evaluating simple efects between GSH treatments (within-row efects) at each YAN concentration (within YAN).Contrast specifc P-values are adjusted for multiple comparisons.TableS4.One-way ANOVA comparison of initial with fnal GSH concentrations.ANOVA analyses were grouped by strain and initial 10 Australian Journal of Grape and Wine Research GSH concentration.Post hoc test employed Dunnett's test to compare the fnal concentration of GSH with the initial concentration.Six tests were conducted to evaluate GSH loss at diferent nitrogen concentrations.Table S5.One-way ANOVA comparison of total strain-specifc H 2 S concentrations produced in low YAN conditions (155 mgN/L) with diferent initial GSH concentrations.ANOVA analyses were grouped by strain and initial GSH concentration.Post hoc test employed Tukey's multiple comparison test with a single pooled variance.Contrast specifc P-values are adjusted for multiple comparisons.Two one-way ANOVA analyses were conducted to evaluate H 2 S production by each of the two yeast strains.Table S6.Two-way ANOVA analysis of the efect of initial YAN and GSH on end of ferment H 2 S concentrations by yeast strains AWRI1688 and AWRI2861.ANOVA analyses were conducted separately for each strain.Post hoc test employed Dunnett's multiple comparison test with individual variances computed for each comparison.Contrast specifc P-values are adjusted for multiple comparisons.Table S7.Two-way ANOVA analysis of the efect of initial YAN and GSH on end of ferment MeSAc concentrations by yeast strains AWRI1688 and AWRI2861.ANOVA analyses were conducted separately for each strain.Post hoc test employed Dunnett's multiple comparison test with individual variances computed for each comparison.Contrast specifc P-values are adjusted for multiple comparisons.(Supplementary Materials)