Abstract
Background and goals When vineyards are exposed to wildfire smoke, resulting wines may exhibit unpleasant smoky, ashy attributes (i.e., smoke taint). Smoke exposure leads to elevated levels of volatile phenols and their glycoconjugates in grapes. However, predicting the risk of smoke taint in wine is challenging because some of these markers occur naturally, and few studies outside of Australia have investigated their varietal and/or regional variation. The goal of this study was therefore to explore variation in volatile phenol glycoconjugates (VPG) in California grapes and wine.
Methods and key findings A method for measuring VPGs using Orbitrap mass spectrometry was developed and used to identify the VPGs most indicative of smoke exposure in grapes and wine in CA. Guaiacol, 4-methylguaiacol, cresol and phenol rutinosides, syringol and 4-methylsyringol gentiobiosides, cresol pentose glucoside, and phenol glucoside were consistently found at elevated concentrations (>20 µg/L) in smoke-affected wine. These concentrations were several-fold higher than those of wines from vintages minimally affected by smoke. Satellite imaging data were used to compare the location and density of smoke plumes during the 2018 and 2020 wildfire seasons. Measurement of the above subset of VPGs in grapes sampled during 2018 and 2020 showed they were positively correlated with total glycoconjugates (i.e., the sum of cresol rutinoside, guaiacol rutinoside, 4-methylguaiacol rutinoside, phenol rutinoside, syringol gentiobioside, and 4-methylsyringol gentiobioside concentrations), despite varied levels of total glycoconjugates.
Conclusions and significance These research findings highlight the need for baseline data across different varieties and regions to enable more reliable screening of grapes potentially affected by smoke, and for sampling protocols that account for temporal and spatial variation in smoke exposure.
Introduction
Drier growing seasons and warmer temperatures arising from climate change have intensified the duration, frequency, and scale of wildfires around the world (Bowman et al. 2020). In California, seven of the 10 largest wildfires since 1932 occurred in 2017, 2018, and 2020 (as reported on the CAL FIRE website, www.fire.ca.gov); the 2019/2020 Australian bushfire season was also one of the most devastating on record (Abram et al. 2021). When fires occur, the thermal degradation of lignin in plant material produces volatile phenols (VP) such as guaiacol and 4-methylguaiacol (Maga 1988). VPs can also be found in wines due to their natural presence in grapes (Ristic et al. 2016, Coulter et al. 2022), extraction from oak barrels during wine maturation (Pollnitz et al. 2004), and/or contamination from grapevine exposure to wildfire smoke (Kennison et al. 2007). The latter instigates risk of smoke taint, in which the sensory profiles of wines made from smoke-exposed grapes are characterized by unpleasant ‘smoky,’ ‘burnt,’ ‘drying,’ ‘cold ash,’ and ‘medicinal’ attributes (Kennison et al. 2007, Ristic et al. 2011, Parker et al. 2012).
The quantitation of VPs using gas chromatography-mass spectrometry (GC-MS) is well-established (Pollnitz et al. 2004, Hayasaka et al. 2010a, 2013), but using this metric alone to predict the risk of smoke taint in wine following processing and fermentation of potentially smoke-affected grapes has been impeded by several factors. First, the diversity of natural vegetation that fuels wildfires alters VP profiles in smoke-exposed grapes, adding complexity to establishing thresholds for smoke taint risk (Kelly et al. 2012, Noestheden et al. 2018a). In addition, the rapid uptake and glycosylation of VPs in grapes following smoke exposure hinders accurate risk assessment and increases the likelihood that smoke taint will be underestimated (Szeto et al. 2020).
This complexity arises because VPs are metabolized by grapevines into a wide variety of non-volatile glycoconjugates (Supplemental Figure 1), including glucosides, gentiobiosides, diglycosides (with terminal pentose units), and rutinosides (Hayasaka et al. 2010b), with more recent work also indicating the existence of trisaccharides (Caffrey et al. 2019). Volatile phenol glycoconjugates (VPG) are quantitated by either direct analysis using high-performance liquid chromatography-tandem mass spectrometry (HPLC-MS/MS) (Hayasaka et al. 2010b, 2013, Dungey et al. 2011), or by indirect analysis involving acid hydrolysis of VPGs followed by GC-MS (Noestheden et al. 2017). In situations in which both VP and VPG concentrations are very high, the latter serves to validate the former. However, interpretation remains challenging when concentrations of VPs and their glycoconjugates are incongruent or in the low-to-medium range. This is because of their natural, varietal-dependent abundance (Ristic et al. 2016, van der Hulst et al. 2019, Coulter et al. 2022), delayed accumulation following smoke exposure (Szeto et al. 2020), ability to be hydrolyzed during fermentation (Kennison et al. 2008, Caffrey et al. 2019, Whitmore et al. 2021), and highly subjective, individualized perception thresholds (Parker et al. 2020) due to in-mouth hydrolysis by enzymes during wine tasting (Mayr et al. 2014).
It has become standard practice to include both VP and VPG analysis in smoke taint risk assessment (Favell et al. 2022), even though the time- and resource-intensive process is burdensome during the demanding harvest and winemaking period of vintage. With longer, more intense fire seasons, the widespread concern about smoke taint requires innovative approaches to screen for grapevine smoke exposure (Fudge et al. 2012, Jiang et al. 2021) and ameliorate smoke taint in wine (Fudge et al. 2011, Modesti et al. 2021). However, it also requires a more contextualized interpretation of information acquired from current practice. Thus, the aims of this work were to develop a method for routine VPG analysis using high-resolution tandem mass spectrometry (HR-MS/MS), and to use it for determining VPGs that are specifically elevated due to smoke exposure in grapevines, consistently indicative of smoke exposure across distinct wildfire seasons (e.g., burn conditions, fuel type, age of smoke), and sensitive to low levels of smoke exposure.
Materials and Methods
Reagents
Liquid chromatography-mass spectrometry (LC-MS)-grade solvents were purchased from Sigma-Aldrich. Milli-Q water was sourced from a Purelab Flex 2 system (ELGA LabWater NA). Deuterium-labeled VPG standards (d3-guaicaol glucoside [d3-GuG] and d3-guaiacol gentiobioside [d3-GuGB]) as well as their unlabeled equivalents (guaiacol glucoside and guaiacol gentiobioside) were purchased from Toronto Research Chemicals. β-Glucosidase from almonds (lyophilized, powder, ≥4 U/mg) was sourced from Sigma-Aldrich.
Quantitation of VPGs in grapes and wine by liquid chromatography-high resolution mass spectrometry (LC-HRMS) analysis
LC-HRMS method
A method to quantify VPGs in grapes and wine was adapted to the LC-HRMS interface according to previously published stable isotope dilution analysis methods (Dungey et al. 2011, Hayasaka et al. 2013), with VPG concentrations reported as guaiacol gentiobioside equivalents (for rutinosides, gentiobiosides, and pentose glucosides) and guaiacol glucoside equivalents (for glucosides). Samples were analyzed with a Vanquish ultra-HPLC system coupled to a Q Exactive Orbitrap mass spectrometer (Thermo Fisher Scientific) configured with an Agilent ZORBAX RRHT Eclipse XDB-C18 column (2.1 mm × 100 mm, 1.8-µm particle size) and a heated electrospray ionization source. Sample injection volume was 10 µL.
Column temperature was maintained at 40°C throughout the run. A binary gradient was used, with mobile phases constituted of 0.1% v/v glacial acetic acid in LC-MS-grade water (solvent A) and 0.1% v/v glacial acetic acid in LC-MS-grade acetonitrile (solvent B). The flow rate was set to 0.400 mL/min, with a linear gradient that began at 5% B for 4 min, after which it was increased to 35% B (over 12 min), increased to 100% B (over 1 min) and held for 3 min, and lastly, decreased to 5.0% B (over 0.5 min) and held for 3.5 min.
The Orbitrap was operated in negative ion mode with a capillary temperature of 350°C and an auxiliary gas heater temperature of 425°C. The sheath gas flow rate was 50 arbitrary units (au), the sweep gas flow was 3 au, and the auxiliary gas flow rate was 13 au. The spray voltage was set to −3 kV, and the S-lens radio frequency level was set at 55 au. Spectra were acquired from 100 to 1000 m/z in parallel reaction monitoring (PRM) mode at 5 ppm mass tolerance with 70,000 scan resolution, 1 microscan, 50 ms maximum injection time, and automatic gain control of 1e5. Collision energy was set between 10 and 15 eV. Instrumental control and data acquisition were performed using XCalibur software (ver. 4.1.31.9) (Thermo Fisher Scientific). Included transitions corresponded to glucosides, gentiobiosides, pentose glucosides, and rutinosides monitored in previous work (Hayasaka et al. 2013, Noestheden et al. 2018b) (see Supplemental Table 1).
Sample preparation
Grapes were homogenized, after which 10 g of each sample were spiked with 250 µg/kg of d3-GuG and d3-GuGB as internal standards (ISTD) and prepared for solid phase extraction (SPE). Homogenate samples were centrifuged for 15 min at 7500 rpm using a Beckman Spinchron 15 (Beckman Coulter), adjusted to pH 13 (with 0.5 mL of 10 N NaOH), and processed through a 0.45-µm PTFE syringe filter (Whatman). pH adjustments were made to lower the isobaric interference of phenolic compounds (Noestheden et al. 2018b). To prepare wine samples for SPE, 5-mL aliquots were spiked with 200 µg/L of d3-GuG and d3-GuGB as ISTD and adjusted to pH 13 (with 0.5 mL of 10 N NaOH).
SPE is a time- and resource-intensive procedure; to reduce sample preparation time, automation was trialed with Oasis HLB 96-well SPE plates (30 mg, 30 µm; Waters), hereafter referred to as ‘Oasis HLB plates,’ prepared with a Hamilton Microlab STAR robotic liquid handler (Hamilton Robotics). Wells were pre-conditioned with 0.5 mL of methanol followed by 0.5 mL of water, after which 1 mL of each sample (in triplicate) was loaded onto separate wells and rinsed with 1 mL of water, followed by 0.5 mL of dichloromethane. Plates were dried under vacuum and eluted with 0.5 mL of methanol. Eluates from the three wells corresponding to each sample were combined, transferred into 1.5-mL microcentrifuge tubes, and evaporated to dryness under vacuum in an Eppendorf Vacufuge at room temperature. Lastly, samples were reconstituted in 0.4 mL of water and transferred into a syringeless 0.45-µm filter vial (Agilent Technologies) for LC-HRMS analysis.
As a comparison to high-throughput Oasis HLB plates, wine samples were also prepared with conventional Li-Chrolut EN cartridges (100 mg, 40 to 120 µm) configured to a 20-position manifold (Waters). Cartridges were pre-conditioned with 2 mL of methanol followed by 2 mL of water, after which 5 mL of each sample was loaded onto individual cartridges and rinsed with 2 mL of water. Samples were air dried for 25 min and eluted with 1 mL of methanol. Eluates were dried, reconstituted, and transferred through filter vials for LC-HRMS analysis as outlined above for samples prepared with Oasis HLB plates.
Wine matrix effects
To investigate the impact of matrix effects on calibration linearity and signal intensity in wine samples, model wine and “bag-in-box” dry red wine were prepared with differing concentrations of ethanol. Model wine was prepared with 0 to 20% ethanol (at 5% increments) in water with 7 g/L of tartaric acid, adjusted to pH 3.4 (with 10 N NaOH). The dry red wine (pH 3.4) was analyzed without alcohol adjustment (13.5% ABV), and as its dealcoholized equivalent (0% ABV) following lyophilization. To remove alcohol from the “bag-in-box” red wine, 40-mL aliquots were poured into 120-mL plastic tubes (Sarstedt Pty. Ltd.), weighed, and frozen overnight at −80°C. Caps were replaced with a layer of tightly wrapped, perforated Parafilm, before samples were transferred into purpose-built flasks attached to a freeze dryer (Kinetics FTS FD-3-85A-MP Flexi Dry Microprocessor Freeze Dry Lyophilizer) and kept under 97 mT vacuum and −88°C conditions until a powder was obtained. Samples were then reconstituted with water to their original recorded mass.
Method validation
Frozen, mature Chardonnay grapes were used for method validation in grapes, and model (0% ABV) and dealcoholized dry red wine (0% ABV) were used for method validation in wine. With the exception of model wine samples (which were not pH adjusted due to the simplicity of their matrix), grape and wine samples received an addition of ISTD and an adjustment to pH 13 (as above). Model wine and grape homogenate samples were prepared with Oasis HLB plates, whereas dealcoholized dry red wines were prepared with both Oasis HLB and LiChrolut EN cartridges. A series of standard additions of GuG and GuGB led to eight calibration levels in grapes (at 0, 10, 20, 30, 50, 100, 200, and 500 µg/kg) and to seven calibration levels in model and red wines (at 0, 5, 10, 25, 50, 100, and 250 µg/L). Method precision (reported as percentage relative standard deviation [%RSD] values) and accuracy (reported as percentage recovery values) were examined with 10 replicate samples spiked at 30 µg/kg each of GuG and GuGB (grapes), four replicate samples spiked at 10 and 25 µg/L each of GuG and GuGB (0% ABV model wine), or 5 and 25 µg/L each of GuG and GuGB (dealcoholized dry red wine).
Samples
To identify VPGs that were elevated specifically due to smoke exposure in grapevines, Cabernet Sauvignon wine samples (n = 43) were sourced from commercial wineries across several regions and vintages in CA. Samples from 2011 and 2012 were commercial wines with no known presence of smoke taint and were classified as ‘minimally affected’ by smoke exposure. Samples from 2015, 2018, and 2020 were research wines made from growing regions proximal to wildfires and thus, were classified as ‘affected’ by smoke exposure. This distinction was supported with historical wildfire data and an adapted classification scheme (Crews et al. 2022). Additional details about the sources of wine can be found in Supplemental Table 2. Wine samples were used to identify indicative glycoconjugates instead of grape samples, due to the availability of wine sourced from years with low wildfire activity.
The glycoconjugates most indicative of smoke exposure in wine were quantified in grapes to investigate their consistency across distinct wildfire seasons and their sensitivity to low levels of smoke exposure. Cabernet Sauvignon grapes (n = 108) were sourced from commercial vineyards (at maturity), comprising 61 samples in 2018 (78% from Lake County) and 47 samples in 2020 (74% from Napa Valley). Samples in 2018 and in 2020 were collected between 30 Aug and 11 Oct and between 22 Sept and 8 Oct, respectively. Based on their total glycoconjugate concentrations, these grapes were classified as having had low-to-extreme levels of smoke exposure. In 2020, mature Chardonnay (n = 52), Cabernet Sauvignon (n = 40), and Pinot noir (n = 20) grapes were also sourced from commercial vineyards in the Central Valley (n = 112, 83% from Lodi or Delta). These grapes were classified as having had low-to-moderate levels of smoke exposure, based on their total glycoconjugate concentrations.
Spatial data
Fire perimeters
Fire perimeters from the 2018 and 2020 wildfire seasons in CA were sourced from the CAL FIRE Fire and Resource Assessment Program (FRAP) Geospatial Information System (GIS) database (https://frap.fire.ca.gov/mapping/gis-data). Several statewide agencies contribute to the database, and for a fire to be recorded, it must have burned an area of at least 4, 12, and 120 ha for timber, brush, and grass fires, respectively.
Hectares burned
For each wildfire season from 2008 to 2021, the total hectares burned, number of megafires (i.e., fires with >40,000 ha burned), and size of the largest wildfire were identified using the CAL FIRE historical wildfire activity statistics (http://www.fire.ca.gov/our-impact/statistics), hereafter referred to as CAL FIRE Redbooks.
Smoke plumes
Shapefiles of smoke plumes from the 2018 and 2020 CA wildfire seasons were sourced from the Hazard Mapping System (HMS) Fire and Smoke Analysis, a product offered by the National Oceanic and Atmospheric Administration (NOAA)/National Environmental Satellite Data and Information Service (https://www.ospo.noaa.gov/Products/land/hms.html). Smoke plumes were identified by classifying imagery acquired by two NOAA-NASA (National Aeronautics and Space Administration) satellites from the Geostationary Operational Environmental Satellites (GOES) fleet, GOES-16 and GOES-17. The geographic domain of the HMS is focused on North America, stretching from 14.6°N to 72°N and from 50°W to 170°W. In addition to the location and outline of each smoke plume, aerosol optical depth data can be used to estimate the density of each plume as light (up to 10 μg/m3), medium (10 to 21 μg/m3), and heavy (21 to 32 μg/m3). Daily smoke plumes were generated using a sequence of satellite images collected over a period of one to three hours of daylight.
Polygons of smoke plumes with light, medium, and heavy density were examined over ~80 days in 2018 (24 July to 30 Sept) and 2020 (26 July to 11 Oct). This range of dates was relevant to grape sampling dates within periods of significant wildfire activity. The boundaries of Lake County, Napa County, and Sonoma County (in 2018) and those of Napa County, Sonoma County, and the Lodi/Delta region (in 2020) were used as a base layer. The boundaries for Napa, Sonoma, and Lake County were sourced from the California Open Data Portal, which is procured by the California Department of Technology (https://data.ca.gov/dataset/ca-geographic-boundaries). The boundary for the Lodi/Delta American Viticultural Area (AVA) was sourced from the American Viticultural Areas Digitizing Project Team (https://github.com/UCDavisLibrary/ava). The polygons of smoke plumes were visualized over these perimeters and given a value of 0 if they did not intersect with at least 30% of a region, and a value of 1 if they intersected with at least 30% of a region. This visual analysis was performed for each day, region, and smoke density level.
Vegetation types
A raster layer showing the spatial distribution of CA Wildlife Habitat Relationship (WHR) classes was sourced from the CAL FIRE Hub (https://hub-calfire-forestry.hub.arcgis.com). CA WHR classes are grouped into 13 major land cover types, which include agriculture, barren (or other), conifer forest, conifer woodland, desert shrub, desert woodland, hardwood forest, hardwood woodland, herbaceous, shrub, urban, water, and wetland. In this raster layer, desert shrub and desert woodland were combined into a single ‘desert’ class, and wetland areas were not included. The land area within the fire perimeters identified by the CAL FIRE FRAP GIS database in 2018 and 2020 was mostly accounted for by conifer forests, hardwood forests, hardwood woodlands, herbaceous, and shrub land cover. Zonal histograms were used to calculate the number of pixels corresponding to unique types of land cover within each fire perimeter. The raw raster data had a distance resolution of 50 m and were exported as a GeoTiff image with a larger pixel size of 100 m × 100 m to improve computational efficiency.
Particulate matter (PM)
Daily maximum values of PM from 2018 (Lake County, Napa County, and Sonoma County) and 2020 (Napa County, Lodi/Delta AVA [represented by data from San Joaquin County], and Sonoma County) were sourced from the Air Quality System database (https://www.epa.gov./outdoor-air-quality-data) maintained by the US Environmental Protection Agency. PM data were included if they were tracked using the federal reference method and defined by a diameter of ≤2.5 µm (PM2.5). Most sensors in the network utilized the beta attenuation monitoring to measure PM2.5 concentrations, except for the data from Lake County, which were collected via gravimetric methods.
Data analysis
Spatial data were analyzed and visualized in QGIS (ver. 3.22.8) and projected using the North American Datum of 1983 (NAD83)/California Albers (EPSG:3310) as a coordinate reference system. For visualization purposes, trendlines were fitted using the locally estimated scatterplot smoothing (LOESS) function in ‘ggplot2’ (i.e., geom_smooth()). Chemical data were analyzed in RStudio (RStudio Inc.) ver. 4.1.1 and visualized using the ‘ggplot2’ package.
Results and Discussion
SPE is a common sample preparation technique employed prior to analysis that serves to purify and concentrate analytes of interest. However, the increased prevalence of samples submitted to commercial laboratories for smoke taint analysis—and the time-sensitive decisions resting on the results—would benefit from faster sample processing times than can be achieved with vacuum extraction manifolds. Standard practice involves analysis of VPGs following sample extraction with C18 cartridges and/or filtration through a 0.45-µm filter (Hayasaka et al. 2010a). However, a recent study demonstrated improved recovery of glycoconjugates with the inclusion of an alkaline wash step (0.1 N NaOH, pH 13) to limit interference presented by polyphenolic flavonoids and their corresponding glycoconjugates (Noestheden et al. 2018b). The inclusion of this step requires the replacement of the C18 sorbent, which would not retain the VPGs under such an extreme pH (the typical working pH range for C18 is 2 to 9), with a polymeric equivalent (Noestheden et al. 2018b). Oasis HLB 96-well plates were selected in the present study because they offer the dual benefits of being stable under extremely high pH conditions and adaptable to automation. Their performance was compared to LiChrolut EN, a choice product in conventional SPE workflows with high selectivity for phenolic compounds. The sorbents of both Oasis HLB and LiChrolut cartridges are lined with copolymer substrates.
Red wine and model wine samples were spiked with GuG and GuGB (0 to 250 µg/L) and d3-GuG and d3-GuGB as ISTD (200 µg/L) and analyzed by LC-HRMS in PRM mode. Model wine was prepared with 0% ABV, and red wine was dealcoholized to assess cartridge performance in the absence of competing matrix effects. As shown in Supplemental Table 3, standard curves demonstrated high linearity over the calibration range, with high coefficient of determination (R2) values across both SPE cartridges and wine matrices (R2 range = 0.9995 to 0.9999). Precision values in dealcoholized red wine for GuGB at spiking levels 5 and 25 µg/L were 7% and 4%, respectively (for LiChrolut EN), and 3% and 2%, respectively (for Oasis HLB plates). Precision values in dealcoholized red wine for GuG at spiking levels of 5 and 25 µg/L were 2% and 1%, respectively (for LiChrolut EN), and 2% and 4%, respectively (for Oasis HLB). Accuracy levels for both compounds at spiking levels of 5 and 25 µg/L were 94 to 104% (for LiChrolut EN), and 91 to 101%, respectively (for Oasis HLB plates).
Due to the time-intensive process of dealcoholization, its effect on calibration linearity and peak area intensity was quantified in model wines (containing 0, 5, 10, 15, and 20% ABV) and dry red wine (0 and 13.5% ABV) prepared with Oasis HLB plates. In model wine, peak areas corresponding to GuG and GuGB decreased with increasing %ABV, indicating that fewer glycoconjugates partitioned to the SPE sorbent as matrix polarity decreased. However, as expected, this change was mirrored by the internal standards, thus conserving the analyte/ISTD ratio (Supplemental Table 4). Calibration linearity of GuGB quantitation in model wine was high, with R2 values ≥0.9992 across the range of model wine samples. On the contrary, calibration linearity of GuG quantitation in model wine demonstrated higher sensitivity to matrix effects, with R2 values decreasing from 0.9994 in 0% ABV to 0.9942 and 0.9988 in 15% ABV and 20% ABV matrices, respectively (Supplemental Table 5).
A comparison of red wine without alcohol adjustment (13.5% ABV) and its dealcoholized equivalent demonstrated a very similar trend, with lower absolute GuG and GuGB peak areas in the presence of alcohol (however, this was accounted for with an ISTD). Regardless of dealcoholization of the dry red wine, high calibration linearity was maintained for both GuG and GuGB, with R2 values from 0.9994 to 0.9998. This work highlights the effect of %ABV on the absolute peak area of VPGs. While an ISTD can account for these differences, the absence of deuterated standards corresponding to the full suite of VPGs elevated by grapevine smoke exposure prompted incorporation of a dealcoholization step into the sample preparation procedure.
Following method validation in wines, which demonstrated that the performances of Oasis HLB plates and LiChrolut EN cartridges were comparable, validation in grapes was carried out using Oasis HLB plates only. White grape homogenate was spiked with GuG and GuGB (0 to 500 µg/kg) and d3-GuG and d3-GuGB (250 µg/kg) as ISTD. GuG and GuGB, respectively, demonstrated high linearity (R2 values = 0.9994 and 0.9995), with good precision (6% and 4% RSD) and accuracy (108% and 118% at 30 µg/kg spike level). Further details about method validation in wine and grape matrices can be found in Supplemental Table 3.
Previous studies have shown the effects of timing, duration, and density of smoke exposure on the abundance of VPs and their glycoconjugates in grapes and wine (Kennison et al. 2009, 2011, Szeto et al. 2020). However, the potential effects of these factors on the compositional profile of grape VPGs have yet to be explored in great depth (Jiang et al. 2021). The Australian Wine Research Institute’s commercial services laboratory (Affinity Labs) currently measures six VPGs (alongside free VPs) to inform smoke taint analysis: cresol rutinoside, guaiacol rutinoside, 4-methylguaiacol rutinoside, phenol rutinoside, syringol gentiobioside, and 4-methylsyringol gentiobioside. These glycoconjugates are a subset selected from over 20 VPGs that have been reported in the literature (Hayasaka et al. 2013, Noestheden et al. 2018b, Caffrey et al. 2019, van der Hulst et al. 2019). As yet, few studies have assessed whether this subset of glycoconjugates is relevant to grapevines exposed to wildfire smoke in regions outside of Australia (Crews et al. 2022); thus, Cabernet Sauvignon wines made from grapes affected by different levels of smoke exposure in CA were examined.
In general, a higher intensity of smoke exposure in grapevines increases the risk that smoke taint will be perceived in subsequent wine; however, this relationship is not linear (Kennison et al. 2009, Parker et al. 2012, Szeto et al. 2020). Several methods can be used to classify samples affected by different levels of smoke exposure. They are not grounded in wine sensory studies and remain coarse indications of smoke exposure only. The first option is to use varietal-specific baseline levels established in grape and wine samples collected from several regions and vintages in Australia (Coulter et al. 2022). This system provides a conservative classification that attests the presence or absence of smoke exposure. It is critical for decisions involving samples with lower levels of smoke exposure, but the use of ‘total’ glycoconjugate values is an alternative option for samples with glycoconjugates at levels above baseline. Total glycoconjugate values correspond to the sum of the six glycoconjugates routinely measured by Affinity Labs in their smoke taint analysis. In a recent study, samples were classified into seven levels of smoke exposure based on quantitation of total glycoconjugates, comprising baseline (<6 µg/L), light (6 to 30 µg/L), modest (31 to 100 µg/L), significant (101 to 200 µg/L), elevated (201 to 300 µg/L), substantial (301 to 400 µg/L), and severe (>400 µg/L) (Crews et al. 2022). Herein, wines were classified as ‘minimally affected’ and ‘affected’ based on the severity of relevant wildfire seasons and the latter classification scheme.
In each wildfire season, total hectares burned was used to approximate the prevalence of wildfires and the associated risk that wines were made from smoke-affected grapes. For seasons with a comparable number of total hectares burned, the specific location and duration of major fires (>2000 ha) relative to the appellation of wine samples were examined in greater detail. Between 2011 and 2021, the second lowest value for total hectares burned (92,495 ha) occurred in 2011, whereas the areas burned in 2018 and 2020 were among the top five wildfire seasons, with ~770,000 and 1.74 million total ha burned, respectively (CAL FIRE Redbooks). With such disparate values in the range of total hectares burned over the last 20 years, wines from 2011 and 2020 were considered ‘minimally affected’ and ‘affected,’ respectively.
The total hectares burned in the 2012 and 2015 wildfire seasons were comparable (335,575 and 356,487 total ha burned, respectively) (CAL FIRE Redbooks). However, closer examination showed that the 2012 commercial wines (primarily from Napa Valley, Sonoma County, and the Central Coast) and the 2015 research wines (from Lake County) were likely made from grapes with different amounts of smoke exposure. The CAL FIRE Redbooks consider fires to be ‘large’ if the total area burned exceeds ~120 ha. During 2012, most of the large fires that occurred across CA burned <2000 ha and were contained in under four days (CAL FIRE Redbooks). Napa Valley and Sonoma reported a combined 145 total ha burned, and Monterey and San Luis Obispo Counties on the Central Coast each recorded <1200 total ha burned (CAL FIRE Redbooks). On the other hand, Lake County alone recorded 69,544 total ha burned in 2015, due in part to the Rocky and Jerusalem Fires (28,101 and 10,165 ha burned, respectively), and it took around two weeks to contain each of these fires (CAL FIRE Redbooks). Wine from 2012 was therefore classified as ‘minimally affected,’ whereas wine from 2015 was classified as ‘affected.’
Total glycoconjugate measurements in wine supported the distinction between samples from 2011 and 2012 vintages and those from 2015, 2018, and 2020 vintages (Supplemental Table 6). The total glycoconjugate levels detected in the 2011 and 2012 wines ranged from 3 to 60 μg/L (with a mean value of 23 μg/L). Cresol rutinoside, guaiacol rutinoside, and phenol rutinoside were typically the most abundant glycoconjugates, with mean values between 5 and 8 μg/L. The total glycoconjugate values suggest that wines from 2011 and 2012 were made from grapes with low-to-modest amounts of smoke exposure (i.e., based on the classification system proposed by Crews et al. 2022); values for half of these samples (including all of the 2011 wines) fell within the sum of the 99th percentile values reported for 32 Australian Cabernet Sauvignon wines made from non-smoke exposed grapes (Coulter et al. 2022). Previous research found guaiacol glycoconjugates were stable during bottle aging (Ristic et al. 2017). As such, the low VPG levels observed in older wines were considered to be real rather than an artifact of hydrolysis, especially given that these were commercial wines, i.e., wines with no known presence of smoke taint. For wines from 2015, 2018, and 2020, total glycoconjugate levels indicated that they were likely made from grapes with significant (101 to 200 µg/L total) to severe (>400 µg/L total) levels of smoke exposure, as concentrations ranged from 155 to 871 μg/L (with a mean value of 501 μg/L). Cresol rutinoside, phenol rutinoside, and syringol gentiobioside were the most abundant glycoconjugates in the affected wines, with mean values of 162, 107, and 123 μg/L, respectively. Interestingly, guaiacol rutinoside only had a mean value of 40 μg/L in affected wines. Relative to affected wine from 2020, the most notable differences in affected wine from 2015 and 2018 were the lower levels of guaiacol rutinoside (<10 μg/L, whereas the affected wine from 2020 had a mean value of 44 μg/L).
To identify additional markers that could differentiate minimally affected and affected wine, concentrations of all 24 glycoconjugates were subjected to principal component analysis (PCA) (Figure 1). The first two components captured 78.9% of total variance, with 64.8% attributable to PC 1. Samples with the lowest and highest total glycoconjugates were positioned on opposite ends of PC 1, indicating that separation along this axis is strongly influenced by differences in smoke exposure. However, many individual glycoconjugates also contributed to the variance captured by PC 1, including phenol, guaiacol, and cresol pentose glucosides; phenol, syringol, 4-methylguaiacol, and guaiacol glucosides; and phenol gentiobioside. The relevance of these additional glycoconjugates for differentiating samples with different levels of smoke exposure is uncertain. Therefore, glycoconjugate levels were compared between minimally affected and affected wine (Figure 2). The most indicative markers were identified as those with low concentrations in minimally affected wine and elevated concentrations in affected wine. To compare groups, absolute and relative differences in means were referenced. Among the six established markers of smoke taint, 4-methylsyringol gentiobioside had the lowest mean concentration in affected wine (20 μg/L), and guaiacol rutinoside had the lowest relative difference between affected and minimally affected groups (8.4). Using these numbers as a reference, glycoconjugates were not further examined if they had a mean concentration <20 μg/L in the affected group and/or a relative difference between groups of less than eight.
Principal component analysis scores (A) and loadings (B) plots of volatile phenol glycoconjugate concentrations measured in Cabernet Sauvignon wine samples sourced from vintages with low (2011, 2012) and high (2015, 2018, 2020) wildfire activity in California. MSyG, 4-methylsyringol glucoside; MGuGB, 4-methylguaiacol gentiobioside; MSyPG, 4-methylsyringol pentose glucoside; MGuPG, 4-methylguaiacol pentose glucoside; CrGB, cresol gentiobioside; CrG, cresol glucoside; SyPG, syringol pentose glucoside; GuGB, guaiacol gentiobioside; MSyR, 4-methylsyringol rutinoside; GuG, guaiacol glucoside; MGuG, 4-methylguaiacol glucoside; SyR, syringol rutinoside; SyG, syringol glucoside; GuPG, guaiacol pentose glucoside; PhPG, phenol pentose glucoside; CrPG, cresol pentose glucoside; PhGB, phenol gentiobioside; PhG, phenol glucoside; CrR, cresol rutinoside; GuR, guaiacol rutinoside; MGuR, 4-methylguaiacol rutinoside; PhR, phenol rutinoside; SyGB, syringol gentiobioside; MSyGB, 4-methylsyringol gentiobioside.
Bar chart of mean volatile phenol glycoconjugate concentrations (μg/L) in Cabernet Sauvignon wines made from minimally smoke-affected grapes (from 2011 and 2012; white bars) and smoke-affected grapes (from 2015, 2018, and 2020; gray bars) in California. The mean concentrations of established markers of smoke taint in affected wine are shaded in black (i.e., CrR, GuR, MGuR, MSyGB, PhR, and SyGB). Values are reported as guaiacol gentiobioside equivalents (for rutinosides, gentiobiosides, and pentose glucosides) and guaiacol glucoside equivalents (for glucosides). Error bars represent standard error of the mean. Minimally affected values are plotted as negative for visualization purposes only—all concentration values are positive. CrGB, cresol gentiobioside; CrPG, cresol pentose glucoside; CrR, cresol rutinoside; GuG, guaiacol glucoside; GuGB, guaiacol gentiobioside; GuPG, guaiacol pentose glucoside; GuR, guaiacol rutinoside; MGuG, 4-methylguaiacol glucoside; MGuGB, 4-methylguaiacol gentiobioside; MGuPG, 4-methylguaiacol pentose glucoside; MGuR, 4-methylguaiacol rutinoside; MSyG, 4-methylsyringol glucoside; MSyGB, 4-methylsyringol gentiobioside; MSyPG, 4-methylsyringol pentose glucoside; MSyR, 4-methylsyringol rutinoside; PhG, phenol glucoside; PhGB, phenol gentiobioside; PhPG, phenol pentose glucoside; PhR, phenol rutinoside; SyGB, syringol gentiobioside; SyPG, syringol pentose glucoside; SyR, syringol rutinoside; CrG, cresol glucoside; SyG, syringol glucoside.
Several glycoconjugates had mean concentrations <10 μg/L in affected wine, including cresol, phenol, guaiacol, and 4-methylguaiacol gentiobioside; 4-methylsyringol pentose glucoside; and 4-methylsyringol rutinoside. The remaining glycoconjugates had elevated mean concentrations in the minimally affected wine, ranging from 16 μg/L (4-methylsyringol glucoside) to 25 to 50 μg/L (guaiacol glucoside, guaiacol pentose glucoside, 4-methylguaiacol pentose glucoside, phenol pentose glucoside, and syringol pentose glucoside), and up to >90 μg/L (syringol glucoside, cresol glucoside, and syringol rutinoside). Because the glycoconjugate means in minimally affected wines were elevated, the relative difference to means in affected wines was less than eight. For example, while 4-methylguaiacol glucoside had a mean concentration of 4 μg/L in minimally affected wines, it was only elevated to a mean of 25 μg/L in affected wines; thus, the relative difference between means was less than eight.
The established markers of smoke taint (i.e., cresol, phenol, guaiacol, and 4-methylguaiacol rutinoside, and syringol and 4-methylsyringol gentiobioside) all exceeded these criteria. Additional glycoconjugates that met these criteria included cresol pentose glucoside and phenol glucoside. The potential for pentose glucosides to be included as additional markers of smoke taint has been supported by several other studies (Caffrey et al. 2019, van der Hulst et al. 2019, Szeto et al. 2020, Crews et al. 2022). In contrast, little evidence is available for the inclusion of phenol glucoside as an additional marker of smoke taint. Previous studies conducted in Australia have reported levels of phenol glucoside <5 μg/kg in smoke-affected grape and juice samples (Culbert et al. 2021, Jiang et al. 2021); thus, the potential for phenol glucoside to be used as a marker of smoke taint may be specific to CA samples.
The rate of extraction of VPGs from mature grapes to wine is high (Caffrey et al. 2019, Szeto et al. 2020). Moreover, glycoconjugates have been shown to be stable in wine, even after several years of bottle aging (Ristic et al. 2017). Thus, the markers identified as most important to wine were considered likely to also be representative of smoke exposure in Cabernet Sauvignon grapes grown in vineyards in CA. The occurrence of these markers was examined further in grapes collected from different wildfire seasons.
Previous studies have suggested that the combustion of different fuel sources may lead to differences in exogenous VP loads (Kelly et al. 2014, Noestheden et al. 2018b); however, the extension of this effect to the profile of VPGs in grapes has not been thoroughly evaluated. The results presented above indicate that phenol glucoside may be an example of a regiospecific marker of smoke taint. In the current study, mature grapes were sampled from 2018 and 2020, two of the largest wildfire seasons in the modern history of CA (since 1932) (CAL FIRE Redbooks).
In 2018, grapes were primarily collected from Lake County, and as shown in Figure 3, the wildfire closest to vineyards in Lake County was the Mendocino Complex, which formed through the convergence of the Ranch and River Fires. This was the largest wildfire in 2018, and from its 27 July start date, it spread 185,800 ha over Colusa, Glenn, Lake, and Mendocino Counties and burned 38% of the total area in Lake County before it was contained two months later on 27 Sept. The area burned by the Ranch Fire (166,003 ha) comprised 36% conifer forest, 29% shrub, and 16% hardwood forest land cover, while the smaller River Fire (19,797 ha) was fueled by 71% shrub and 11% hardwood woodland land cover (Figure 3).
Perimeters of major California wildfires near key grapegrowing regions in 2018 (A) and 2020 (B), and the most abundant vegetation types found within the perimeter of each fire (C). Key regions include Sonoma County (SON), Lake County (LAK), Napa County (NAP), and the Lodi AVA (LODI/DELTA).
In 2020, grapes were primarily sourced from Napa County, and several megafires were relevant to this region, including the August Complex, LNU Lightning Complex, and the SCU Lightning Complex (Figure 3). All three complexes began within one day of each other on 16 or 17 Aug. The LNU and SCU Lightning Complexes were contained on 22 and 15 Sept, respectively, while the August Complex was not contained until one month later on 15 Oct. The closest wildfire to vineyards in Napa County was the LNU Lightning Complex, which spread 146,990 ha and burned 41% of the total area in Napa County and 2% of the total area in Lake County. The August Complex burned 417,898 ha and affected 5% of the total area in Lake County. Although the perimeter of the August Complex did not intersect with that of Napa County, the probability of smoke exposure from this fire affecting vineyards in Napa County cannot be discounted given its unprecedented magnitude. Similar reasoning justified the inclusion of details about the SCU Complex, a megafire south of Napa County that burned 160,508 ha, including key regions in the Central Valley such as San Joaquin and Stanislaus Counties. The land cover defining each of the megafires was distinct, with predominant fuel types in each megafire being conifer forest (August Complex), hardwood woodland (SCU Lightning Complex), and shrub (LNU Complex) (Figure 3).
In addition to differences in the scale and fuel load characterizing the 2018 and 2020 wildfire seasons, satellite imagery and PM (PM2.5) concentrations suggest the presence of differences in the duration and density of smoke exposure. In 2018, a steep, four-week accumulation of days with light, medium, and heavy smoke exposure occurred in Lake County that spanned from late July until the end of August (Figure 4), after which the medium and heavy smoke exposure plateaued. Smoke with lighter density continued to linger until the end of September, supported by trends in PM2.5 concentrations (Figure 4). The observed plateau of medium and heavy smoke plumes may reflect the containment of the Ranch Fire on 17 Aug (CAL FIRE Redbooks). In contrast, a gradual accumulation of light, medium, and heavy smoke exposure occurred in 2020 and persisted from the middle of August, when the three megafires ignited, until the end of the season. The rate at which days of heavier smoke exposure accumulated showed a slight reduction from mid-September, which coincided with containment of the LNU Complex (on 22 Sept). Nevertheless, the cumulative number of days during which smoke plumes (across density levels) were observed continued to rise. The PM2.5 concentrations also reflect the persistence of smoke in the air during this period (Figure 4). Thus, 2018 and 2020 were two distinct wildfire seasons, from the perspectives of scale, fuel load, and burn conditions.
Line graphs depicting daily maximum concentrations of particulate matter (PM2.5; μg/m3) (A, B) and the cumulative number of days on which smoke plumes covered at least 30% of a key grapegrowing area in California on a given day (C, D) in 2018 and 2020. Plume density variation is shown as light (– • –), medium (– – –), and heavy (—).
As shown in Figure 5, the distribution of total glycoconjugates measured in grape samples collected in 2018 were largely concentrated within the ‘significant’ (101 to 200 µg/L total) and ‘elevated’ (201 to 300 µg/L total) categories of smoke exposure; outside this trend, total glycoconjugates were <100 μg/L in five samples and >300 μg/L in a single sample. In contrast, grape samples collected in 2020 were seemingly exposed to a wider distribution of smoke, with 35% classified as ‘modest’ (31 to 100 µg/L total glycoconjugates), 12.5% as ‘significant’ (101 to 200 µg/L total glycoconjugates), 15% as ‘elevated’ (201 to 300 µg/L total glycoconjugates), 10% as ‘substantial’ (301 to 400 µg/L total glycoconjugates), and 20% as ‘severe’ (>400 µg/L total glycoconjugates).
Bar graph showing the number of Cabernet Sauvignon grape samples sourced from California that were affected by different levels of smoke exposure during 2018 and 2020, as indicated by total glycoconjugate concentrations (μg/kg), i.e., the sum of cresol rutinoside, phenol rutinoside, guaiacol rutinoside, 4-methylguaiacol rutinoside, syringol gentiobioside, and 4-methylsyringol gentiobioside, reported as guaiacol gentiobioside equivalents. Smoke exposure was categorized as light (6 to 30 μg/L), modest (31 to 100 μg/L), significant (101 to 200 μg/L), elevated (201 to 300 μg/L), substantial (301 to 400 μg/L), or severe (>400 μg/L) (according to the classification system proposed by Crews et al. 2022).
The increased variation in smoke exposure observed in 2020 samples relative to those from 2018 was unexpected, given the apparent prevalence and persistence of smoke exposure that characterized this season. Samples from 2018 were primarily collected in Lake County from late August to mid-October, a period following the observed plateau in the cumulative number of days with medium and heavy density smoke plumes. By contrast, samples from 2020 were primarily collected in Napa County over a period seemingly defined by persistent smoke exposure. Although the uptake of smoke-derived VPs is evident immediately after grapevine smoke exposure (Szeto et al. 2020), the varied glycoconjugate levels observed for samples collected in 2020 may reflect the timing of sampling and the delayed accumulation of glycoconjugates, which can take several weeks (Dungey et al. 2011, van der Hulst 2019, Szeto et al. 2020). The variation in glycoconjugate profiles of samples collected in 2020 may also be attributable to the intermittent nature of smoke exposure during real wildfire scenarios, i.e., smoke density and the duration of smoke exposure vary as factors such as topography, and speed and direction of the wind, change. It should also be noted that while satellite imagery affords valuable insight into the location and density of smoke plumes, plume elevation cannot be determined, and thus, the presence of smoke at ground level (i.e., among grapevines) is uncertain.
The potential effects of acute versus chronic smoke exposure (ostensibly, 2018 versus 2020, respectively) on the uptake and glycosylation of VPs in grapevines over time is not well understood. Previous studies have demonstrated that exposure to higher density smoke results in increased uptake of VPs in grapes as well as enhanced sensory perception of smoke taint in corresponding wine (Szeto et al. 2020, Wilkinson et al. 2021). This highlights the importance of representative sampling protocols that account for both temporal and spatial variation.
In the current study, the subset of eight glycoconjugates all demonstrated positive correlations with the intensity of smoke exposure, measured as total grape glycoconjugates; however, maximum levels varied from ~40 μg/L for 4-methylguaiacol rutinoside and guaiacol rutinoside to >90 μg/L for cresol rutinoside, phenol rutinoside, and syringol gentiobioside. The observed trendlines suggest that cresol rutinoside, phenol rutinoside, and syringol gentiobioside consistently accounted for high proportions of the total glycoconjugates measured, regardless of vintage, whereas 4-methylsyringol gentiobioside consistently accounted for a lower share (with a maximum concentration of 60 μg/L), even in grapes affected by ‘severe’ levels of smoke exposure (as indicated by total glycoconjugates >400 µg/L) (Figure 6). Differences were observed between vintages for cresol pentose glucoside and 4-methylguaiacol rutinoside concentrations, with greater accumulation of these glycoconjugates evident in 2018. The opposite was observed for phenol glucoside and guaiacol rutinoside, which accounted for higher proportions of total glycoconjugates in 2020 than in 2018. These differences might be attributable to the burn conditions or the types of vegetation that fueled the wildfires that burned during each vintage. The flattest curve corresponded to guaiacol rutinoside in 2018, with concentrations that ranged from 1 to 22 μg/L for grape samples classified as having ‘significant’ (101 to 200 µg/L total glycoconjugates) and ‘elevated’ (201 to 300 µg/L total glycoconjugates) levels of smoke exposure.
Trendlines showing correlations between total glycoconjugates (μg/kg) and concentrations of individual glycoconjugates (μg/kg) in Cabernet Sauvignon grape samples sourced from California. Trendlines were fitted using the locally estimated scatterplot smoothing method, and the shaded area reflects the 95% confidence interval. Closed circles (●) fitted with a solid line (—) describe samples from 2018. Open circles (○) with a dashed line (- - -) describe samples from 2020. Total glycoconjugates refer to the sum of cresol rutinoside (CrR), guaiacol rutinoside (GuR), 4-methylguaiacol rutinoside (MGuR), phenol rutinoside (PhR), syringol gentiobioside (SyGB), and 4-methylsyringol gentiobioside (MSyGB). All values reported as guaiacol gentiobioside equivalents. CrPG, cresol pentose glucoside; PhG, phenol glucoside.
The unpredictable nature of wildfires makes it challenging to acquire information needed to inform decisionmaking in vineyards potentially affected by smoke exposure. The best indicators of smoke exposure are VPGs that demonstrate strong correlations to total grape glycoconjugates, independent of burn conditions or fuel type, herein identified as cresol rutinoside, phenol rutinoside, syringol gentiobioside, and 4-methylsyringol gentiobioside. However, glycoconjugates that demonstrate variation between vintages or that occur at lower concentrations do not necessarily need to be removed from smoke taint diagnostics. Prior to their removal, further research is necessary to understand to what extent different profiles of VPGs may contribute to the sensory perception of smoke taint in wine (Parker et al. 2023).
In addition to being consistently indicative of smoke exposure across distinct vintages, it is also critical for key markers to be sensitive to low levels of smoke exposure. As shown in Table 1, total glycoconjugate levels indicate that Cabernet Sauvignon, Pinot noir, and Chardonnay grapes from the Lodi/Delta AVA were affected by low levels of smoke exposure, being ‘light’ (6 to 30 µg/L total glycoconjugates) in Cabernet Sauvignon and Chardonnay and ‘moderate’ (31 to 100 µg/L total glycoconjugates) in Pinot noir. This assessment is supported by a recent study, in which 21, 15, and 9 μg/kg corresponded to the 99th percentile of total glycoconjugate values in non-smoke-exposed Cabernet Sauvignon, Pinot noir, and Chardonnay grapes, respectively (Coulter et al. 2022). Cresol rutinoside and phenol rutinoside were present at the highest levels regardless of variety, ranging from 7 to 13 µg/kg in the red varieties and from 3 to 4 µg/kg in Chardonnay (Table 1). This may indicate that cresol rutinoside and phenol rutinoside are the glycoconjugates most sensitive to small amounts of smoke exposure. The concentrations of the remaining glycoconjugates were <2 µg/kg in Chardonnay and <9 µg/kg in Cabernet Sauvignon and Pinot noir. Concentrations of additional glycoconjugates are shown in Supplemental Table 7. Decisions about harvest and processing of grapes affected by low levels of smoke exposure are challenging without robust baselines. Expanding knowledge about naturally occurring glycoconjugate concentrations across a range of varieties and regions is critical to create a comprehensive, accessible resource for growers and wineries to use to make informed decisions about price adjustments and/or rejection criteria that fairly and accurately reflect grapevine smoke exposure.
Concentration of key volatile phenol glycoconjugates (μg/kg) in Cabernet Sauvignon (n = 40), Chardonnay (n = 52), and Pinot noir (n = 20) grapes collected at maturity in 2020 from the Lodi/Delta regions of California. CrR, cresol rutinoside; GuR, guaiacol rutinoside; MGuR, 4-methylguaiacol rutinoside; PhR, phenol rutinoside; SyGB, syringol gentiobioside; MSyGB, 4-methylsyringol gentiobioside.
The moderate levels of smoke exposure observed in grapes collected during 2020 were unexpected given the unprecedented scale of wildfire activity. In fact, the Lodi/Delta AVA and Napa County appeared to experience comparable levels of smoke exposure (Figure 4). Nonetheless, the samples from the Lodi/Delta AVA were characterized by ‘light’ (6 to 30 µg/L total glycoconjugates) to ‘modest’ (31 to 100 µg/L total glycoconjugates) smoke exposure, whereas fruit from Napa County was affected by ‘light’ (6 to 30 µg/L total glycoconjugates) to ‘severe’ (>400 µg/L total glycoconjugates) smoke exposure. It is again worth noting that the polygons generated for each smoke plume are two-dimensional, static representations of three-dimensional, dynamic phenomena, and they may not accurately depict the density and/or duration of smoke exposure at ground level. This limitation may be overcome by using sensors that monitor PM. Recent studies have deployed PM sensors to monitor temporal changes in the density and location of smoke plumes (Jiang et al. 2021, Wilkinson et al. 2021). The combined use of satellite imagery, PM sensors, and GIS software could lead to customized sampling protocols for affected vineyards and significantly enhance our ability to contextualize wildfire events and the risk of smoke taint.
Conclusions
VPGs have been found in grapevine leaves and fruit, both as natural constituents and following grapevine exposure to smoke, with smoke density, the duration and timing of smoke exposure, and fuel source affecting levels of free and glycosylated VPs, and thus, the risk of perceptible smoke taint in finished wine. An LC-HRMS workflow was developed, validated, and utilized to identify the VPGs most indicative of smoke exposure in grapes and wine from CA wine regions. The six glycoconjugates identified in previous studies from Australia (cresol, phenol, guaiacol, and 4-methylguaiacol rutinoside, and syringol and 4-methylsyringol gentiobioside) differentiated the minimally affected wines and affected wines herein, along with two additional compounds, cresol pentose glucoside and phenol glucoside. Compositional analysis of grapes from the 2018 and 2020 wildfire seasons revealed that syringol gentiobioside and cresol and phenol rutinoside accounted for the highest shares of total glycoconjugates, whereas the contribution of 4-methylsyringol gentiobioside was consistently low. The remaining markers showed greater seasonal variation. It remains unclear which aspects of fire conditions are most critical to the compositional outcomes of grapes affected by wildfire smoke, and to what extent these differences are relevant to wine sensory outcomes. This presents an opportunity for future studies to integrate spatial data into smoke taint risk assessment. Where available, spatial data can approximate the fuel composition, size and location of the fire perimeter, and duration and density of smoke exposure in proximal vineyards. These data could also be used to optimize sampling protocols and better understand the influence of different burn conditions on the sensory perception of smoke taint in wine.
Supplemental Data
The following supplemental materials are available for this article in the Supplemental tab above:
Supplemental Table 1 Retention times, mass spectral data, and collision energies for volatile phenol glycoconjugates (VPG).
Supplemental Table 2 Source, vintage, and regional details of Cabernet Sauvignon wines collected to determine the volatile phenol glycoconjugates (VPGs) most indicative of smoke exposure in wine. Samples were considered smoke-affected (or minimally affected) based on the prevalence of wildfires in each region/vintage and levels of VPGs.
Supplemental Table 3 Figures of merit for the quantitation of guaiacol glucoside (GuG) and guaiacol gentiobioside (GuGB) in dealcoholized model wine (μg/L), red wine (μg/L), and white grape (μg/kg) matrices via liquid chromatography-high resolution mass spectrometry. SPE, solid phase extraction; RSD, relative standard deviation.
Supplemental Table 4 Raw peak areas of standard additions of guaiacol gentiobioside (GuGB) and guaiacol glucoside (GuG) (from 0 to 250 μg/L) and their deuterated equivalents, d3-guaiacol gentiobioside (d3-GuGB) and d3-guaiacol glucoside (d3-GuG), as measured in a range of model wine matrices (0 to 20% ABV), as determined by liquid chromatography-high resolution mass spectrometry. Ratios reflect peak areas of either GuG or GuGB, relative to its respective deuterated equivalent.
Supplemental Table 5 Effects of alcohol by volume (%ABV) on the linearity of calibration (R2) and recovery (%) of guaiacol gentiobioside (GuGB) and guaiacol glucoside (GuG) in model and red wine matrices at different concentrations (μg/L) as determined by liquid chromatography-high resolution mass spectrometry.
Supplemental Table 6 Concentrations of key volatile phenol glycoconjugates (μg/L) in Cabernet Sauvignon wines from 2011, 2012, 2015, 2018, and 2020 across regions in California. Samples are classified based on total glycoconjugate concentrations (µg/L), i.e., the sum of cresol rutinoside (CrR), guaiacol rutinoside (GuR), 4-methylguaiacol rutinoside (MGuR), phenol rutinoside (PhR), 4-methylsyringol gentiobioside (MSyGB), and syringol gentiobioside (SyGB), reported as guaiacol gentiobioside equivalents. Smoke exposure was categorized as light (6 to 30 µg/L), modest (31 to 100 µg/L), significant (101 to 200 µg/L), elevated (201 to 300 µg/L), substantial (301 to 400 µg/L), or severe (>400 µg/L).
Supplemental Table 7 Concentration of additional volatile phenol glycoconjugates (μg/kg) in mature Cabernet Sauvignon (n = 40), Chardonnay (n = 52), and Pinot noir (n = 20) grapes from the Lodi/Delta AVA in California that were affected by light-to-moderate levels of smoke exposure during vintage in 2020. Values are mean concentrations reported as guaiacol gentiobioside equivalents (for volatile phenol rutinosides, gentiobiosides, and pentose glucosides) and guaiacol glucoside equivalents (for volatile glucosides). GuG, guaiacol glucoside; GuGB, guaiacol gentiobioside; GuPG, guaiacol pentose glucoside; MGuG, 4-methylguaiacol glucoside; MGuGB, 4-methylguaiacol gentiobioside; MGuPG, 4-methylguaiacol pentose glucoside; CrG, cresol glucoside; CrGB, cresol gentiobioside; CrPG, cresol pentose glucoside; PhG, phenol glucoside; PhGB, phenol gentiobioside; PhPG, phenol pentose glucoside; SyG, syringol glucoside; SyPG, syringol pentose glucoside; SyR, syringol rutinoside; MSyG, 4-methylsyringol glucoside; MSyPG, 4-methylsyringol pentose glucoside; MSyR, 4-methylsyringol rutinoside.
Supplemental Figure 1 Chemical structures of guaiacol glycoconjugates.
Footnotes
This research was funded by the Australian Research Council Training Centre for Innovative Wine Production (www.ARCwinecentre.org.au; project number IC170100008), funded by the Australian Government with additional support from Wine Australia (www.wineaustralia.com; project number PPA002459), Waite Research Institute, E&J Gallo winery, and industry partners. The authors would like to acknowledge the Viticulture and Chemistry groups within E. & J. Gallo Winery’s Department of Winegrowing Research for their help with collecting and processing samples. The authors declare no conflicts of interest.
Szeto C, Feng H, Sui Q, Blair B, Mayfield S, Pan B and Wilkinson K. 2024. Exploring variation in grape and wine volatile phenol glycoconjugates to improve evaluation of smoke taint risk. Am J Enol Vitic 75:0750013. DOI: 10.5344/ajev.2024.23060
By downloading and/or receiving this article, you agree to the Disclaimer of Warranties and Liability. If you do not agree to the Disclaimers, do not download and/or accept this article.
- Received July 2023.
- Accepted February 2024.
- Published online May 2024
This is an open access article distributed under the CC BY 4.0 license.
References
Vol 75 Issue 1
Jump to section
Related Articles
Cited By...
More from this TOC section
Similar Articles
