Rapid Evaluation of Smoke exposure in Grapes and Wine by Raman Spectroscopy – A Concept Proposal

Summary – Final Report:

Raman Spectroscopy of Smoke Compounds

We have evaluated ten different smoke compounds in synthetic wine using a 785 nm Wavelength Raman setup using defined parameters to gain spectra.  Individual compounds and combinations were tested to provide us with initial results showing the various Raman fingerprints of the smoke compounds and help determine any potential interferences between them.

Development of a Spectrum Collection Method using Gold Nanoparticle strips

We developed a method for data collection to improve the reproducibility of obtaining the Raman spectrum, using gold nano-strips and a 3D-printed 3×3 grid.  The nano-strips were cut into nine identical pieces, each placed in a well on a 3×3 grid.  Then, our sample was dropped onto the nano-strips for spectrum collection.  Our scan method involves obtaining five separate scans for a set period to determine if signal enhancement improves as the nano-strip becomes more saturated.  We are creating an auto-sampling system for the 785nm Raman to streamline data collection as we finalize the enhancement protocols.

Surface Enhanced Raman Spectroscopy using Gold Nanoparticle Strips

Using the commercially produced gold nano-strips, we obtained scans of smoke compounds in synthetic wine that show strong Raman enhancement properties.  Initial scans of smoke compounds were obtained to have benchmark results, and then the nano-strips were used to determine enhancement effectiveness with lower concentrations.  Two significant results were obtained; a unique fingerprint for guaiacol using the nano-strips was found in the form of a very strong peak at 576 cm-1, and we have been able to detect guaiacol at concentrations in synthetic wine as low as 100ppb (parts-per-billion) using the nano-strips.

Removal of Anthocyanins and Tannins through different Extraction Methods

We have trialed many different extraction methods testing the removal of anthocyanins, tannins, and any other compounds present that interfere with the obtained spectra of smoke compounds in wine.  Various parameters were tested to determine the best method for removal while recovering the smoke compounds present to figure out how to obtain clean scans of smoke compounds in a wine matrix.  Scans of two different types of wines spiked with smoke compounds prior to extraction were obtained to see how well these extraction methods recovered the compounds when collecting Raman spectra.  Un-spiked red wine was also treated through the extraction process and then spiked with smoke compounds to see if there was any interference with scans taken post-extraction using an actual wine matrix rather than a synthetic wine matrix.

Summary of Major Research Accomplishments and Results by Objective

In the 2023-2024 funding year, we have created a new data collection method for observing Raman enhancements using the gold nanoparticle strips, which has produced promising results and allowed us to see phenolic concentrations at very low levels.

We have collected Raman spectra of 10 phenolic compounds of interest (Guaiacol, 4- methylguaiacol, 4-ethylguaiacol, 4-ethylphenol, 4-vinylphenol, whiskey lactone, phenol, o-cresol, m-cresol, and p-cresol) using both of our data collection methods for the 785nm Raman spectrophotometer, the multi-well plate with liquid silver nanoparticles and the 3×3 grid with gold nanoparticle strips.  The gold nanoparticle strips produced data with strong repeatability and noticeable Raman enhancements.

Using these data collection methods, we have established benchmarks for the Raman fingerprints of each of the ten compounds we are testing with the 785nm Raman setup.  This data has been used for signal confirmation of these compounds when conducting scans at low concentrations.

We have successfully produced spectra of some phenolic compounds at low concentrations in a synthetic wine matrix, specifically guaiacol at a parts-per-billion level.  The collected data has been reproduced successfully for method confirmation.

Various solid-phase and stir-bar sterile extraction methods have been tested and refined for sample treatment, attempting to remove anthocyanins/tannins and other compounds that provide spectra interference.  Color removal has been successfully achieved, and now the focus is on increasing signal or concentration enhancement and testing different solvents for the most effective elution process.

Using our previously developed machine learning algorithm, we collected spectra of each compound with the 785nm to remove the Rayleigh scattering and background fluorescence present.  After backgrounds were removed, the intensities of each peak were captured through a regression algorithm.  After the base models were built using average spectra, each sample was measured using juice or wine base models to report the intensity of each peak.  Intensities of extracted peaks were subsequently analyzed using a correlation matrix and principal component analysis.  The above metrics were fed into Orange Data Mining for data processing and ML development.

Wine Compound Interactions That Cause Perceived Smoke Taint in Red Wines.

Objective 1:  Evaluate and select Pinot noir and Cabernet Sauvignon wines that contain various levels of smoke taint.

As detailed in the previous years report more than 120 wines were collected that have various levels of smoke taint. The majority of wines (~100) were Cabernet Sauvignon and Pinot noir, but a few other varietals were also included such as Petit Verdot, Merlot and Syrah. Sensory analysis for these wines were completed in August 2023. We spent 9 months, running weekly sensory panels. This was longer than anticipated but we managed to collect a significant number of smoke impacted wines and did not want to miss the opportunity for larger data set, which can result in stronger causation in the last objective. At the start of each term panelists went through training and they also had retraining after 5 weeks, half way through the 10 week school term. The sensory protocol that was used can be found in Fryer and Tomasino (2022).

Objective 2:  Measure the volatile and nonvolatile chemical composition of the wines from Objective 1.

Measurement of volatile compounds are in progress. We are about half way done with this analysis. The protocol for volatile analysis was adapted from Zhang et al. 2023. We are about 3 months behind on the chemical analysis due to an installation delay with the liquid nitrogen tank used to provide gas to the equeipment and several compounds (utilized as chemical standards) that were back ordered. As of December 2023 analysis was able to recommence. Now that the sample prep for measuring smoke compounds (free volatiles, bound compounds and thiophenols) have been completed the student has started prepping for nonvolatile analysis at the start of January 2023. Free smoke phenols are being measured according to Liu et al. (2020) and bound smoke glycosides according to Caffrey et al. (2019). We are in the process of developing the method to measure the smoke thiophenols. The timeline in the 3rd year grant has been updated to reflect this change in timeline.

Objective 3:  Determine the chemical interactions causing smoke taint perception in Pinot noir and Cabernet Sauvignon wines using fsQCA – this will commence once chemical testing is completed.

Objective 4:  Validate results for fsQGCA using sensory testing – this will commence once chemical testing is completed.

Achieving tropical fruit aroma in white wines: from winemaking to consumer acceptance.

Summary – Final Report

The overall goal of this project is to produce Chardonnay wines with increased tropical fruit aroma perception.  In a previous study developed by Dr. Elizabeth Tomasino’s research group, we found that wines with higher concentrations of fermentation esters and volatile thiols imparted more intense tropical fruit aroma nuances.  Therefore, in this project specific winemaking processes (skin contact, β-lyase addition, and two fermentation gradient temperature regimes) were performed with the intent to either increase or decrease these aromas in the wine.  The first two processes (skin contact and β-lyase addition) are known for increasing volatile thiol concentrations in wine.  The latter (fermentation temperature) is expected to increase thiol concentrations and preserve fermentation esters.

In both micro ferments and larger research scale ferments the Skin contact and combination of skin contact and fermentation gradient resulted in wines with higher concentrations of esters and thiols.  Descriptive analysis (DA) showed that SC was significantly different from both FG and SCFG.  SC presented the most tropical fruit aromas, SCFG presented more stone fruit, and FG presented more honey and lemon/lime.  Liking, using a 7-point Likert scale, showed there was not a significant difference between treatments.  Emotional response, using a 5-point Rate-All-That-Apply (RATA) scale, showed significant differences for ‘calm’ and ‘disgusted’ emotions.  Although using Multiple factor analysis to investigate all the data it was shown that liking was associated with positive emotions which were also associated with tropical fruit aromas.

Together this research shows that skin contact and a fermentation temperature gradient have a positive effect on tropical aroma in Chardonnay wines.  Understanding the link between wine processing and quality and consumer desires is important as wine is subjected to different environmental conditions each year, and consistency can be difficult to achieve.  These conditions of winemaking using skin contact and fermentation temperature gradient would be recommendation for winemakers looking for this tropical fruit aroma sensory signature in Chardonnay wines.

Highly Selective Enzymatic Degradation of Undesirable Phenols in Brettanomyces and Smoke Taint Contaminated Wines.

This is the first-year progress report on a multi-year project divided into two-phases with the first (current) phase focusing on discovery and design of enzymes with exclusive degradation of 4-ethylguaiacol (4EG) and 4-ethylphenol (4EP), the main sensory-active components in wine contaminated with Brettanomyces and responsible for a portion of the ‘taint’ phenols found in smoke-tainted wine (Licker 1998, Mirabelli-Montan 2021).  A second phase of this project—beyond the scope of the current proposal—will use the base of both engineered laccases and the specific knowledge of their active site modifications to create additional enzymes capable of fully, and exclusively, degrading the remaining smoke taint marker compounds including guaiacol, 4-methylguaiacol, o-, p-, m-cresol, and syringol.

Summary of Major Research Accomplishments and Results by Objective

Objective 1 – Library screening:  Library screening resulted in the discovery of 13 laccases with almost 100% 4EG degradation in a buffered pH 7 environment, with three also showing 90% 4EP degradation. When the environment is made more acidic, at pH 4, only laccases #33 and #36 degrade 80-90% 4EG with minimal 4EP activity. When testing in model wine with 4g/L tartaric acid, 13% ethanol, at pH 3.5, only laccases #15 and #16 completely degrade 4EG with small amounts of 4EP activity; with laccases #11 and #34 also degrading 60% 4EG. When the phenols are isolated, then #11, #15, and #36 degrade 80-90% 4EG with #36 degrading almost 30% 4EP after one week. A sequence similarity network that clustered laccases by 70% homology showed multiple clusters had activity on 4EG and 4EP at pH 7. Activity in both pH environments were limited to one cluster. High activity on 4EG in model wine was slightly spread out over three clusters, with additional low to mild activity on 4EP in two of them.

Objective 2 – Computational analysis & design:  Library screening resulted in two pairs of homologous laccases (#15 & #16; #33 & #36) representing two different paths for design towards an exclusively 4EG and 4EP degrading enzyme. However, structural analysis of the active site with the docked phenol revealed a wide-open site on all active laccases. These open active sites greatly increase the risk of off-target degradation of beneficial flavor aromatic compounds. This slowed down the next design steps with first needing to map out the active site landscape for residues required for desired activity by building and testing point mutations to alanine. Current plans include testing these point mutants for altered activities. Then, computational design with newly developed open-sourced AI-tools will be utilized to close the active site with increased specificity and exclusivity on 4EG and 4EP.

Rapid evaluation of smoke exposure in grapes and wine by Raman spectroscopy and machine learning.

Raman Spectroscopy of Smoke Compounds

We have evaluated 10 different smoke compounds in synthetic wine using a 785 nm Wavelength Raman setup using defined parameters to gain spectra. Individual compounds, as well as different combinations, were tested to provide us with initial results showing the various Raman fingerprints of the smoke compounds and help determine any potential interferences between them.

Development of a Spectrum Collection Method using Gold Nanoparticle strips

We developed a method for data collection to improve the reproducibility of obtaining Raman spectrum, using the gold nano-strips and a 3D-Printed 3×3 grid. The nano-strips were cut into 9 identical pieces and then each piece was placed in a well of the 3×3 grid, then our sample was dropped onto the nano-strips for spectrum collection.

Our scan method involves obtaining 5 separate scans over the course of a set period to determine if signal enhancement improves as the nano-strip becomes more saturated. We are in the process of having an auto-sampling system created for the 785nm Raman to streamline our data collection as we are finalizing the enhancement protocols.

Surface Enhanced Raman Spectroscopy using Gold Nanoparticle Strips

Using the commercially produced gold nano-strips, we have obtained scans of smoke compounds in synthetic wine that show very strong Raman enhancement properties. Initial scans of smoke compounds were obtained to have benchmark results, and then the nano-strips were used to determine the effectiveness of enhancement with lower concentrations. Two significant results were obtained; a unique fingerprint for Guaiacol using the nano-strips was found in the form of a very strong peak at 576 cm-1, and we have been able to detect guaiacol at concentrations in synthetic wine as low as 100ppb (parts-per-billion) using the nano-strips

Removal of Anthocyanins and Tannins through different Extraction Methods

We have trialed many different extraction methods testing the removal of anthocyanins, tannins, and any other compounds present that interfere with obtained spectra of smoke compounds in wine. Various parameters were tested to determine the best method for removal while also recovering the smoke compounds present, to figure out how to obtain clean scans of smoke compounds in a wine matrix. Scans were obtained of two different types of wines spiked with smoke compounds prior to extraction, to see how well these extraction methods recovered the compounds when collecting Raman spectra. Un-spiked red wine was also treated through the extraction process and then spiked with smoke compounds, to see if there was any interference with scans taken post-extraction using an actual wine matrix rather than a synthetic wine matrix.

Impact of malolactic fermentation timing and the use of non-Saccharomyces yeast during cold soaking on Pinot noir wine color and sensory properties. (2024-2704

The malolactic fermentation (MLF) is a key process in the production of red wines and some white wines.  While it is commonly conducted after the completion of the alcoholic fermentation (sequential), it can also be induced at the same time where Oenococcus oeni is inoculated shortly after the beginning of alcoholic fermentation (AF).  While a concurrent MLF is typically completed in a shorter time period than a sequential MLF, it is often avoided due to concerns over the production of excess acetic acid, loss of color, and competition with the fermentative yeast Saccharomyces cerevisiae.  Finally, the impact of concurrent MLF on the organoleptic qualities of red wines is relatively unknown and may present an obstacle for the adoption of this practice.  This project addressed some of these concerns and determined the impact of MLF timing on Pinot noir wine chemical and sensory properties.  In addition, the use of non- Saccharomyces yeast during cold soaking and the potential impact on a concurrent MLF was determined.

The timing of MLF inoculation had a significant impact on color with wines produced using a concurrent MLF having lower color density and polymeric pigments.  Interestingly, lower color and polymeric pigment content was only present in concurrent wines that had undergone a cold soak.  While MLF timing affected color, the addition of Torulaspora delbrueckii during cold soaking did not.  For the sequential MLF treatments, there was no difference in color or polymeric pigment content in wines that underwent a cold soak compared to those that did not.  After nine months of aging, the color and polymeric pigment differences present at the end of MLF were no longer present.  Of note was that wines produced with a concurrent MLF increased in color while all other wines saw a reduction in color as is typical during aging.  In addition, all wines increased significantly in their polymeric pigment content.  While all treatments increased in polymeric pigments, the largest increases in polymeric pigment occurred for the three treatments that underwent concurrent MLF.

While measurable color differences in the wines declined during aging, sensory analysis did reveal color differences between some treatments.  Wine produced with a concurrent MLF was noted to be more violet in color than other wines while wine produced with cold soaking + T.delbrueckii + concurrent MLF had higher red intensity.  The more red colored wine was also liked significantly more than the other wines when evaluated in clear glasses.  However, there were no differences in liking when black glasses were used emphasizing the effect of color on liking.  Sensory evaluation also revealed significant differences between the wines in aroma and mouthfeel.  Wines produced without cold soaking had significantly different aroma than those produced with a cold soak, and non-cold soak wines were also significantly different from each other due to MLF timing.  In contrast, timing of MLF did not impact the aroma of the cold soaked wines, whether an addition of T. delbrueckii had been made or not.  Non-cold soaked wines made with a sequential MLF were described as having cedar, fruity, and herbal aromas while those made with a concurrent MLF were described as having spicy, dried fruit, and woody aromas.

Cold soaked wines made with a sequential MLF had dark berry fruit and leather aromas.  For mouthfeel, the major differences were between wines that had T. delbrueckii added and those that did not.  Cold soaked wines with an addition of T. delbrueckii produced wines with desirable mouthfeel characteristics such as smooth, soft, and balanced.

Wines produced in 2023 included an additional variable, the use of Lactiplantibacillus plantarum to conduct MLF.  As was seen with the 2022 wines, the use of T. delbrueckii and/or conducting concurrent MLF had minimal impact on the speed of the alcoholic fermentation.  MLFs conducted by L. plantarum proceeded rapidly due to the higher inoculation rate but those conducted by O. oeni also completed MLF in 14 days or less.  Early analysis of color and phenolic properties shows different color trends between wines produced with concurrent MLFs induced by O. oeni or L. plantarum.  Polymeric pigments  trended higher  in concurrent MLFs with L. plantarum compared to O. oeni, except for concurrent MLF wines with an addition of T. delbrueckii (no cold soak).  Further analysis of these wines will determine which variables are significant as well as any significant interactions between the treatments.

Baseline Levels of Smoke-related Volatiles and their Glycosidic Precursors in California Grapes and Wines

Summary

The increasing incidence of wildfires in grape growing regions of California and the West Coast has highlighted the need for enhanced understanding of the levels of volatile phenols and their non-volatile glycoside precursors that contribute to smoke taint off-flavors in grapes and wines. Within the scope of this project, we conducted measurements of ten volatile phenols in non- smoke exposed grapes aiming to elucidate the baseline levels of these compounds in both red and white grape varieties. Specifically, free and total levels of guaiacol, 4-methylguaiacol (creosol), phenol, 4-ethylguaiacol, o-, m-, p-cresol, 4-ethylphenol, 4-methylsyringol, and syringol were assessed in grapes sourced from various regions across California. Concurrently, air quality data for these regions is presented. The obtained data reveal variability in concentrations across grape varieties, regions, and years. This variability underscores the necessity for extensive databases to comprehensively evaluate the impacts of smoke exposure on the composition of both free and total volatile phenols. We have submitted a proposal to continue this work in 2024-25.

Tropical Fruit Aroma in Wine

Summary The overall goal of this project is to produce Chardonnay wines with increased tropical fruit aroma perception. In a previous study developed by Dr. Elizabeth Tomasino’s research group, we found that wines with higher concentrations of fermentation esters and volatile thiols imparted more intense tropical fruit aroma nuances. Therefore, in this projectspecific winemaking processes (skin contact, β-lyase addition, and two fermentation gradient temperature regimes) were performed with the intent to either increase or decrease these aromas in the wine. The first two processes (skin contact and β-lyase addition) are known for increasing volatile thiol concentrations in wine. The latter (fermentation temperature) is expected to increase thiol concentrations and preserve fermentation esters. The accomplishments for the project for the 2022 year include sensory analysis of wines that were scaled up, including consumer liking and emotional response. Basic wine chemical analysis and ester analysis was performed on all wines. Samples have been prepped for thiol analysis and PI has worked towards synthesizing thiol precursors, as purchasing standards has been challenging. While we cannot yet link the thiol and ester concentrations to the aromas noted in the wines in this project, wine consumers have positive emotional responses to the control, skin contact +fermentation gradient and fermentation gradient wines. The wines with only skin contact were associated with very negative emotions. Additionally, the three wines with positive aromas were described with different tropical fruit descriptors.

Smoke Taint Sensory Interactions

Evaluate and select Pinot noir and Cabernet Sauvignon wines that contain various levels of smoke taint During the first few months of this project I recruited a student to perform this project. We also managed to collect 120 wines, Pinot noir and Cabernet Sauvignon, from industry collaborators that have had some level of smoke impact. Wines were primarily from the 2020 vintage and many were from winemaking tirals, of which we have the trial information. We have started doing sensory analysis of the wines to determine the level of smoke impact. Sensory procedure we are using can be found in (Fryer and Tomasino, 2022). Panels are happening weekly and all wines are being evaluated in duplicate. We are evaluating 10 wines per week, and have sensory panels scheduled through March 2023. After each sensory panel we are pulling 50 mL samples for future chemical analysis. All samples are being stored at -80°C. Storage at the temperature is very important to ensure that no changes happen with the wines during storage. Wines that have been stored in December have started to go through sample prep for bound smoke glycoside analysis, as both bound and free analysis of wines are scheduled to start in March 2023. After the smoke analysis is done we will move onto other compounds.

Raman for Smoke Exposure

During the 2022-2023 funding cycle, we have finished the volatile phenol analysis of 82 wine samples having various degrees of smoke exposure and about 60 pooled samples from different wineries. Those samples will be used for Raman spectra collection once the nanoparticle and Raman signal enhancement protocol is finalized. We have collected grape samples throughout the 2022 season from eight vineyards in California and Oregon and an enology lab. Out of the 214 independent samples, 180 are pinot noir. The samples were collected between September 9 and October 24. About half of the samples were collected from Salem, OR, where the wildfire created a lot of worry during harvest. The lab offered us samples and enology test results in an anonymous way. More grape samples will be collected in 2023. All the grape samples will be analyzed in the 2023-2024 funding cycle simultaneously. Those samples will be used for Raman spectra collection once the nanoparticle and Raman enhancement protocols are finalized. Different sizes of silver nanoparticles have been synthesized, and the Raman enhancements were observed. However, the enhancement depends on the size, shape, and concentration of the nanoparticles. More nanoparticles with different sizes will be synthesized and evaluated in the 2023-2024 funding cycle in conjunction with a linker agent such as benzenethiol. An automated sampling system for the Raman Spectrometer was developed by Dr. Feng Ye’s team from Spectra Scientific company to replace manual spectra collection. The system is based on an alumina 3D printed 117 well plate to eliminate Raman interference from plastics. Because alumina is not stable at acidic pH, we will obtain another 3D printing plate and coat it with gold so it can be used for wine samples. 18 We have established benchmark results for the Raman fingerprints of wine samples, especially using a combination of steady-state electronic and vibrational spectroscopies in a table-top optical setup. We have implemented FSRS with a tunable Raman pump and probe pulses in a femtosecond laser amplifier system for assessing the smoke-exposed wine. We have combined FSRS with AgNPs to optimize conditions for SE-FSRS and achieve higher sensitivity than the resonance Raman enhancement alone. We are investigating nanoparticle receptors for smoke compounds to enhance the Raman signal. Raman spectrum of 82 smoke-exposed wines and 214 smoke-exposed grape samples (grape juice) were collected on the 1064nm FT-Raman system. We have developed an algorithm for machine learning. First, we applied a background subtraction algorithm to remove the Rayleigh scattering and background fluorescence. After backgrounds were removed, the intensities of each peak were captured through a regression algorithm. After the base models were built using average spectra, each sample was measured using juice or wine base models to report the intensity of each peak. Intensities of extracted peaks were subsequently analyzed using a correlation matrix and principal component analysis. The above metrics were fed into Orange Data Mining for data processing and ML development. The model showed an excellent correlation of Raman to titratable acidity. The undergoing research is to use nanoparticles for surface enhancement Raman Spectroscopy and smoke compound receptor to enhance the signal so that the GC-MS data can be correlated with the smoke compounds.