Using a genome-scale metabolic model for Saccharomyces cerevisiae for understanding and modifying strain performance

Two key metabolic activities of yeast relevant to wine fermentations are nutrient utilization efficiency and wine aroma development. For nutrient utilization efficiency (NUE), variability in yeast cell metabolism results from modulation of cellular processes that include changes in membrane composition along with a range of other metabolic pathways that are not fully understood. This variability often affects the completeness of a fermentation (characterized as “dry,” “sluggish,” or “stuck”). Moreover, variability in yeast species or strains used in wine production results in different concentrations of aroma compounds, which can lead to distinct sensory characteristics. Controlling factors affecting nutrient utilization efficiency and wine aroma profile and mouthfeel characteristics related to yeast requires a detailed understanding of cellular metabolism. To develop such understanding, studies often use large-scale data approaches (e.g. genomics and metabolomics), along with multivariate statistics, to identify key metabolic fluxes or metabolites whose presence favors a specific fermentation outcome. Although these studies are useful in exploring variation between yeasts, they are often not comprehensive enough, especially considering that they are labor intensive and costly. An alternative method is to use genome-scale metabolic models combined with dynamic FBA (flux balance analysis) to predict the flux distribution of all the metabolites within the cell over the course of an entire fermentation. As a part of this grant, our goal is to show that this computational approach can be used to predict experimental wine fermentation data, to understand differences between commercial strains, and to suggest genetic modification strategies towards increasing strain performance and control aroma characteristics. To date, we have been able to simulate anaerobic, nitrogen-limited yeast fermentations with the latest genome-scale yeast model. Predictions for nutrient utilization and production of metabolites such as ethanol, glycerol, and organic acids are quite good. Biomass is somewhat underpredicted using conditions that we would expect. However, predicted biomass increases if we vary amino acid utilization and oxygen utilization at the beginning of fermentation. We have also found that the biomass composition and amino acid composition of proteins are important parameters in predicting maximum biomass concentration. Experimentally, however,
we have found that composition changes between strains and over time. We are now generating a more complete set of this data. In past work, no measurements of this biomass composition were conducted—researchers just assumed composition from old data sets—thus limiting the utility of their predictions. We have also made progress on curating the model for aromatic compounds derived from yeast metabolism and have begun our experiments to quantify the aromatic compounds as a function of time and yeast strain.

Development of a prediction tool for phenolic extraction in red wines as a function of winemaking practices and fermentor design

Red wine fermentations are performed in the presence of grape skins and seeds to ensure extraction of color and other phenolics. The presence of these solids results in two distinct phases in the fermentor, as the solids float to the top to form a “cap.” Modeling of red wine fermentations is, therefore, complex and must consider spatial heterogeneity to predict fermentation kinetics and phenolic extraction. We have developed a reactor-engineering model for red wine fermentations that includes the fundamentals of fermentation kinetics, heat transfer, diffusion, compressible fluid flow, and extraction of phenolics (anthocyanins, skin tannins, and seed tannins).

COMSOL was used to solve all components of the model simultaneously utilizing a Finite Elements Analysis (FEA) approach. Prediction of phenolic concentration gradients and temperature gradients from this model were validated against measurements in 2000 L pilot fermentations. Model prediction and experimental data showed excellent agreement for anthocyanin and tannin concentrations and distributions over the course of fermentation.

After validation, this model was applied to examine how fermentor design (e.g. scale and aspect ratio) and operational decisions (temperature set point, pump over frequency) would affect phenolic extraction rates, relative concentrations of skin to seed tannins, and distribution of phenolics throughout the fermentor in the absence of cap management. These results were a follow up to 2018-2019’s work, where the model was used to explore fermentation dynamics and temperature control in red wine cylindrical fermentors and white wine concrete egg fermentors. Example findings include optimization of skin tannin extraction via cap management, with 1x/day pump overs being found superior to both no cap management and 8x/day pump overs, a finding made possible via the combined spatial fermentation-extraction model.

Our results have opened up two exciting avenues of further investigation. The first is applying our reactor engineering models to isothermal fermentation process acceleration, where wine fermentation process cycle time could be greatly decreased by the judicious application of yeast nutrients throughout the fermentation, maximizing yeast biomass. This would greatly improve productivity in existing wine fermentors and lower the capital cost of new winery equipment. The second is the application of COMSOL extraction models to external grape pomace extraction columns, allowing for the fine-tuning of phenolic profiles in the end wine, potentially in a much more rapid fashion than in-tank extraction.

In the nineteen months since this grant began, we have been highly productive having published six primary research papers, along with a review of wine fermentation process modeling. We have also published three papers in cooperative works stemming from this grant. We have presented this work at various extension venues, as well as technical conferences throughout the nation.

Improvement of Wine Quality: Tannin and Polymeric Pigment Chemistry


As we were breaking new ground from several perspectives, the project took some unexpected turns. At this point, we have completed several goals. First, we have created a table of expected wine pigments based on known reactions between anthocyanins and proanthocyanidins, goals 2a and 2b. These results were published in two papers, the first describing how to enumerate all possible proanthocyanidins that could be distinguished by mass spectrometry, up to a degree of polymerization (DP) of 10 (Cave and Waterhouse 2019), and the second, taking that list and “reacting” those tannins with the pigment to create a list of all possible wine pigments. This was a list of over 1 million structures. (Cave et al. 2019)

Having that list, we then compared the masses of those proposed structures with mass spectral data from two sources, QToF and ICR. In the case of the QToF, we could match about 10% of the several hundred observable compounds (Cave et al. 2020). However, the ICR is much more sensitive and precise, and here we could observe about 18,000 signals, and of those could match over 20% of the signals to entries in our table (Cave et al. 2020). Unfortunately, many signals could be represented by more than one structure, so additional study is needed to determine which of the options is present. These results addressed goals 1a, 1b and 2c. The data we now have will require further mass spectral analysis to discriminate between a number of redundant structures.

Development of a Tribology Method to Assess Mouth-feel Perceptions of Red Wines

We have successfully extracted and chemically characterized red wine with tannins, as well as the interactions between red wine tannins and salivary proteins by chemical methods. The effect of tannin-protein interactions on friction forces has been introduced. A tribology method has been developed to show an effect of red wine tannin on mouth lubrication. The formation of turbidity had the strongest relationship to the sensory effect of astringency.
This research has successfully chemically characterized tannins and the consequences of their interaction with saliva or proteins. The tribology experiments carried out on the two red wines with different lubricant (saliva or mucin) has also given promising results, and received much attention in the press. The research is continuing with Watrelot and Kuhl.

Characterization of Aroma Volatiles and their Glyosidic Precursors in Grapes and Wines

Summary: The complex aroma of wine is derived from many sources, with grape-derived components being responsible for the varietal character. The ability to monitor grape aroma compounds would allow for better understanding of how vineyard practices and winemaking processes influence the final volatile composition of the wine. Previously we developed a procedure using GC-MS combined with solid-phase microextraction (SPME) for profiling the free volatile compounds in grapes and wines. We also developed a method for monitoring the ‘aroma potential’ of grapes and wines without the need for initial isolation of the glycoside precursor fraction. However, this method still depends on indirect measurement of the glycosides and acid or enzymatic hydrolysis is needed to release the volatile aglycone which can result in artifact formation. In the current project we validated a novel method using UHPLC-qTOF MS/MS for direct analysis of intact aroma glycosides in grapes with minimal artifactual changes in composition. Using this method we tentatively identified 27 monoterpene glycosides including two monoterpene trisaccharide glycosides, tentatively identified for the first time in any plant. We measured the terpene glycosides in six cultivars at three maturity time points and demonstrated differential profiles depending on cultivar and maturity. We also modified the method so that it can be used to monitor monoterpene glycosides in wines and during winemaking. We have analyzed the glycoside content during fermentation for wines made in fall 2016 and 2017 with different varieties (Chardonnay, Merlot, Cabernet Sauvignon) and winemaking/processing methods. Monoterpenyl glycoside profiles differed between the grapes and the first alcoholic fermentation samples. In red wines, malonylated monoterpenol glucosides and monoterpenol hexose-pentoses decreased after the completion of alcoholic fermentation. We also measured the volatile composition of the wines during fermentation and we have started to relate changes in terpene volatiles to changes in the glycoside profiles. This work sheds important insight into possible biochemical changes in glycosylation during grape berry maturation. In addition, this research will allow us to better understand the effects of viticultural and winemaking practices on grape and wine components that affect flavor.

Rapid Determination of Molecular and “Truly Free” Sulfur Dioxide by Headspace Gas Chromatography

We completed development of an analytical procedure using headspace gas chromatography (HS-GC) coupled with sulfur chemiluminescence detection (SCD) which can rapidly and precisely quantify molecular and free sulfur dioxide in wine. The method requires minimal sample preparation and involves no chemical reagents (with the exception of a trace internal standard). At room temperature the method can successfully detect levels of molecular sulfur dioxide at concentrations as low as 0.03 mg/L. The total chromatographic time for the method is 8 minutes and, provided that information on the alcohol concentration and pH is readily available, the molecular and free sulfur dioxide concentrations for the sample can be rapidly calculated using simple formulae. The HS-GC method offers a high degree of precision, with a reported coefficient of variation of 3.72%. Comparisons with standard A/O and Ripper results on a large set of wine samples showed large discrepancies for those wines with high anthocyanin levels, suggesting that SO2 bound to anthocyanins is released during those procedures, inflating the amount of free SO2 that is actually available to protect the wine. The characteristics of rapid analysis, good sensitivity, and high precision, demonstrate that the method could be applicable in a production environment, albeit a large scale operation where a Gas Chromatograph could be utilized and maintained.

Oxidation of Wine: Central Role of Iron for Polyphenol Oxidation

We were able to find a quick and simple method to discriminate between the two species of iron in wine, using only a spectrometer and very simple reagents, so the method is accessible in winemaking. We applied the method to measure iron in about 10 wines, and evaluated how the iron species in those wines changed in response to oxygen exposure. The observed responses suggested that the level of iron species in response to oxygen exposure could be used to evaluate a wine’s capacity or reactivity to oxidation. Separately, we also showed the products of reacting caffeic acid quinone with a number of flavonoids, including catechin. The observations help explain the basis of browning in white wines. And looking at quinone reacting with wines, we were able to find specific oxidation products.

Rapid Analysis of Wine Phenolics by Laser-induced Fluorescence

Under the current grant, we conducted an initial evaluation of wine fluorescence properties with a time-resolved (lifetime) fluorescence spectroscopy (TRFS) device that allows for rapid, in situ measurements of fluorescence intensity, spectra and lifetime upon UV laser light excitation and visible autofluorescence light detection. Below, we provide a synopsis of the main results along with an extended report.

First, we measured the excitation-emission matrices of model wine solutions (wine analogs with individual components typically found in wines) that may contribute to the overall fluorescence of finished wine products, together with grape seed extracts and proteins. This analysis will guide the configuration of the TRFS device, which spectral distribution can be adjusted for optimal performance in each application. Moving forward with wine applications, narrower optical filters along the main fluorescence peaks found in the EEMs of wines and their respective components have the potential to increase the detection sensitivity. Only a small subset of wine components were tested in this work, namely caffeic acid, gallic acid, rutin, catechin, and malvidin-3-glucose. Expanding the library of potential contributors to the fluorescence of the final wine product is expected to further guide the optimization procedure of the final device, which could be different depending on the specific application, i.e. it might be of interest to tune the spectral bandwidths to detect a particular contaminant instead of intrinsic wine properties.

Second, we tested the performance of the current lab configuration of the TRFS device to detect spectra (intensity ratio) and fluorescence lifetime of the wine models as well as a variety of wines. Consistently with the EEMs measured in the first place, we found a red shift of red wine with respect to most of the tested wine models, except for caffeic acid, which spectral properties closely resemble those of the finished product. Fluorescence lifetime of all tested wine models was shorter in spectral band 1 than that of red wine. However, for the rest of the spectral bands, fluorescence lifetime of all models except for caffeic acid was longer, where detectable. For caffeic acid, lifetime was always found shorter than for red wine. Commercial wine bottles were then tested. Tannin levels and fluorescence properties were measured to find that both intensity ratio and lifetime in spectral band 4 (570 – 650 nm) better correlate with tannin levels than fluorescence parameters in any other spectral range. This further confirmed some preliminary data that we had acquired previous to this award. Interestingly, the selection of wines for this analysis had a narrow range of tannin levels. Combining the results from the two experiments extends the range of tannins, and initial evaluations indicate a trend: as tannin concentration increases, fluorescence intensity increases, and fluorescence lifetime decreases. With a considerable increase of tested wines and further statistical analysis, these fluorescence parameters have the potential to be used as a proxy for tannin concentration, which would be a faster and economic assay to run compared to current methods. We also applied a multivariate analysis of fluorescence parameters to explore the potential of TRFS to identify or discriminate between different wine varieties. The presented analysis is a very simplified model, but already shows discrimination power. Applying more advanced computational methods and expanding the database with different wines could result in a classifier capable of identifying different wine types in a rapid and inexpensive manner. 2

Third, we evaluated how fluorescence parameters are affected by oxygen levels in wine. This could have potential implications in determining wine quality. For example, after opening a wine bottle, fluorescence parameters could establish when the wine gets spoiled. Our modest first trial showed that wine oxygenation changes some of the fluorescence parameters, but not all. Further measurements and analysis are required to understand these changes and establish an experimental model.

In summary, the studies enabled by this award yielded very promising results and our group plans to continue working on this space, which provides a new and exciting area of research for our time-resolved fluorescence spectroscopy technology that is complementary to our current biomedical applications.


Anthocyanin-Cell Wall Interactions Effect on Tannin Extraction

The objectives of this proposal have been to do the following:

1) Perform extractions on white grape skins at 5 time points (13,17,20,22, and 25 °Brix) throughout maturity with varying levels of added anthocyanins.

• Extracted tannin will be analyzed for concentration, average molecular size, subunit composition, activity, and fraction extracted.
• Measure the levels of pectin methylation in the white grape skins

2) Conduct red-styled fermentations on Sauvignon blanc and Cabernet Sauvignon at 300 pounds per fermentation, with and without an anthocyanin addition. Measurements are the same as Objective 1.

The overall purpose of this proposal is to determine anthocyanins role on the extraction of procyanidin material due to interactions with cell wall material of grape skins. Activity to date has been the isolation and purification of color along with the sampling and extraction of skins with varying concentration of anthocyanins added. Furthermore, 300lb fermentations were conducted, in triplicate, under standard red wine making conditions with both Sauvignon Blanc and Cabernet Sauvignon with a 1.4% (v/v) addition of color concentrate.

Investigating Fruitiness Perception in Red and White Wines

This report details activities that occurred from February 2018 – January 2019. The final date of this project is August 2019 and the next 6 months will include completing the last of the sensory panels and combining all data analysis. A final report will be submitted in January 2020. We are still slightly behind on the timeline due to issues detailed in last year’s report and we also had to renew our IRB (human ethics approval) in June 2018. Once the renewal is submitted it is illegal to run sensory tests on the project until approval is given, which was obtained in September 2018. In January 2019 we completed the last of the Pinot noir sensory panels, although we have not yet done data analysis on the January 2019 panel. Also as stated previously, we have not been able to complete any predictive modeling, some initial reviewer comments said that this objective might have been too ambitious in the timeline and after the 1st year we have to agree and have since removed this objective. We plan on working on predictive modeling in the future but this would be after the current grant is completed.

To date we have investigated 80 different compound combinations and their impact to fruit aroma in Pinot noir wine. We have also completed a panel that shows the influences of phenolic content on fruity aromas in Pinot noir and one panel that shows the impact of ethanol content on fruity aroma in Pinot noir. We have 2 potential marker compounds for red fruit aroma in Pinot noir and 4 red fruit solution sets using fsQCA that show the cause of red fruit aroma in Pinot noir. We have also found 5 solution sets for dark fruit aroma in Pinot noir using fsQCA.

We have also investigated 49 compound combinations for fruitiness in white wine. We are still working on using fsQCA to analyze this data. Preliminary results suggest a combination of low thiols and high esters are responsible for tropical fruit aromas, low to no esters are needed for citrus aromas, and esters and terpenes cause pear, peach and apricot aromas.

We will be running the last 4 Pinot gris sensory panels from February2019-June 2019 and completing the final data analysis. We are in the process of writing the first paper for publication and have done 2 presentations at domestic conferences on the analytical data analysis. Spring /Summer 2019 we will be presenting at 4 different international conferences in Europe and have plans for at least 3 more peer-reviewed publications.