Abstract
Background and goals Wine results from the complex interplay between berry chemistry, microbial metabolism, and other abiotic factors. The unpredictability of the final product of these interactions often necessitates fermentations to assess fruit quality, yeast metabolism, etc. Pilot-scale (20 L) fermentation-based experiments with adequate replication are time intensive and costly, although there is also a hesitancy to utilize microvinification (~50 mL). This is because of concerns that small fermentations are highly variable, prone to excessive oxygen exposure, or may lead to other issues, despite previous research demonstrating otherwise. To evaluate these questions, this study investigates previously uncharacterized elements of microvinification, including O2 ingress and the effects of temperature, cap management, and extraction time on wine phenolic and aroma profiles.
Methods and key findings The O2 pickup, phenolic extraction, and volatile release of fermentation in 50-mL microvinifications were quantified and compared to 20-L pilot-scale fermentations. The microvinifications produced consistent results, with no evidence of excess O2 pickup or variability. Phenolic extraction and volatile production varied in some vinifications according to the cap integration method, temperature, and maceration time. Treatment effects similar to those of larger-scale fermentations in previous studies were frequently noted.
Conclusions and significance Microvinifications offer convenience in predicting postfermentation chemistry and yield consistent results with no oxidation problems. Wines similar in phenolic and aroma chemistry to pilot-scale fermentations can be produced on the 50-mL scale provided that representative samples with enough replications are analyzed. Thus, microvinifications can be a valuable tool in improving resolution and replicability in research studies and predicting berry-derived wine chemistry in commercial production.
Introduction
During red wine maceration and fermentation, nonvolatile (e.g., sugars, phenolics, acids, and glycosidically-bound precursors) and free aroma compounds are extracted and modified through microbial-mediated and abiotic chemical reactions. Specific fermentation conditions affect the compounds extracted from the pulp, seeds, and skins, while fermentation, storage, and aging influence metabolite creation from fruit-derived precursors. The production, extraction, or release of esters (Massera et al. 2021), terpenes, norisoprenoids (Hjelmeland and Ebeler 2015), and phenolic compounds (Ichikawa et al. 2012, Alencar et al. 2018), among others, are all affected by the concentrations of specific precursors in the fruit (Gunata et al. 1988, Maicas and Mateo 2005). The release of glycosidically-bound precursors and varietal thiols and the production of esters can all be affected by the fermentation conditions, including the yeast strain, nutrition, temperature, cap management techniques, and basic chemical factors (e.g., pH, sugar, and organic acid levels). Researchers and industry professionals often estimate the flavor potential of grapes by using pilot-scale winemaking (10 to 100 L) (Sampaio et al. 2007). This small-lot fermentation approach is also useful to predict fruit-related faults, including smoke taint (Ristic et al. 2015). However, the observed and modeled kinetics at these small volumes are often significantly different from those of fermentations scaled up to 1000 L or more (Boulton 1979, Sampaio et al. 2007, Sparrow and Smart 2015).
Small-scale fermentations (less than 10-L volumes) are common in experimental winemaking, in part because of the need to replicate fermentations for statistical assessments and to keep field treatment replications separate. Research winemaking is often conducted at scales of at least 20 L, due in part to the belief that they mimic commercial fermentation to a greater extent than smaller fermentation lots in terms of oxidation, sample variability, and kinetics (Romani et al. 2020, Oberholster et al. 2022). However, there are conflicting data on the effect of the fermentation scale on oxygen ingress, with some findings indicating increased dissolved O2 levels with decreases in the volume and increases in the tank surface-to-volume ratio (Hornsey 2007, Sampaio et al. 2007, Schmid et al. 2007). Previous work has successfully conducted fermentations at volumes smaller than 20 L; 1.5-L fermentations were used in one study to measure aroma profiles, including varietal thiols and descriptive sensory analysis of multiple white wines, after reactions with elemental sulfur (Lyu et al. 2021).
Concerns about the applicability of pilot-scale fermentations compared to commercial or large-scale (greater than 50-L volumes) versions persist, despite findings that indicate similar wine chemical parameters. In a recent study evaluating 10, 25, 50, and 100-L fermentations of Cabernet Sauvignon and Tempranillo, factors such as pH, alcohol level, and titratable acidity (TA) remained unchanged, with the lowest variations seen in the 25- and 100-L cases (Sánchez-Ortiz et al. 2021). In another study, no differences in anthocyanins were observed between 80- and 500-kg fermentation lots (Schmid and Jiranek 2011), and results similar to the phenolic content and sensory attributes of larger-scale fermentation lots have been found in 1- to 10-L fermentations (Cadière et al. 2012) and in 200-g to 10-kg must weights (Sparrow et al. 2016). Additionally, it has been reported that the variability in small-scale fermentations is minimal, with low variations in chemical and sensory analyses between 20-, 50-, and 300-kg batches (Schmid et al. 2007).
Size, shape, and temperature considerations can affect various elements of fermentation and the final chemistry of a wine, often with the factors interacting (Lerno et al. 2018). Typical mean winemaking temperatures range from 8 to 15°C for white wine and from 25 to 30°C, or more, for red wine (Sablayrolles 2009, Margalit 2012), with higher cap temperatures observed in red wine fermentations. Fermentation releases heat, which can be either dissipated or retained to raise the must temperature. Higher temperatures in red wine fermentations improve phenolic compound extraction, which is essential for astringency and bitterness (Sacchi et al. 2005). Heat dissipates faster in batches under 200 L, resulting in cooler pilot-scale temperatures, and can be regulated by temperature-controlled chambers, tank jackets, or water baths. For red wines (Pinot noir, Grenache, and Cabernet Sauvignon), a temperature gradient of 10 to 14°C between the pomace cap (up to 33°C) and the vessel bottom has been observed (Schmid et al. 2009). Phenolic extraction varies with the fermentation time and cap management; anthocyanins are extracted early, whereas larger phenolics require higher ethanol levels (Lerno et al. 2017). Furthermore, the fermentation size and shape significantly affect extraction; taller fermenters may require more intense cap management for comparable anthocyanin and color extraction (Miller et al. 2020).
Fermentations at scales of 20 L or more are limited by the need for composite samples from multiple vines or fewer treatments, thus restricting experimental design feasibility. This has led to the reconsideration of small-scale fermentation lots in viticulture research (Sampaio et al. 2007, Sparrow and Smart 2015). Smaller fermentations enable per-vine or per-cluster trials, capturing system variability, which is notably useful in high-throughput phenotyping and “omics” research requiring large sample sizes and high resolution (Alañón et al. 2015, Sirén et al. 2019). For example, marker-assisted grape breeding (using DNA markers to select for desirable plant qualities) involves thousands of F1 hybrids annually, and fermenting at the 20-L scale requires multiple mature vines, rendering large-scale (greater than 20-L volumes) experiments impractical (Karaagac et al. 2012). Alternative methods to vinification exist, such as solvent extraction and glycolysis, and although convenient, these may overestimate compound extraction and fail to replicate the complexity of fermentation (Downey and Hanlin 2016).
In this study, variables such as the temperature, maceration time, and cap integration method for on-skin microvinification (i.e., 50-mL red winemaking) were investigated. Microvinifications were explored in terms of O2 ingress and parameter characterization of fresh versus frozen fruit, and in comparison to pilot-scale fermentations (Figure 1).
Schematic illustrating the study objectives and research questions for characterizing microvinification.
Materials and Methods
Chemicals and reagents
Materials used for winemaking included ICV GRE yeast and Go-Ferm (Lallemand), potassium metabisulfite (KMBS) and L-tartaric acid (Presque Isle Wine Cellars), and AimTab Reducing Substances Tablets (Fisher Scientific). Chemicals used for the Adams-Harbertson assay included glacial acetic acid and bovine serum albumin (both from Thermo Scientific); triethanolamine (Avantor); sodium dodecyl sulfite, sodium hydroxide, potassium bitartrate, and catechin hydrate (all from Sigma-Aldrich); hydrochloric acid (Honeywell); and L-malic acid and ferric chloride (both from Alfa Aesar). Solvents and reagents used for chromatography sample preparation included methanol (all from Sigma-Aldrich) and Everclear grain alcohol (Luxco). Standards used for chromatography quantification and semiquantification included linalool and diethyl succinate (both from Alfa Aesar), methyl caprate (Acros Organics), hexyl acetate and 4-ethyl-2-methoxyphenol (both from Thermo Scientific). 18 standards were used for compound confirmation and calibration (standards listed in Supplemental Table 1), along with 2-octanol (Sigma Aldrich) as an internal standard. All standards purchased were >97% purity.
Grapes used for vinification trials
Interspecific red hybrid grapes were acquired from three vineyards at commercial ripeness for the given year (basic fruit chemistry shown in Supplemental Table 2). The cultivar Chambourcin (~2 kg) was acquired in 2018 from the University of Missouri Southwest Research Center experimental vineyard (Mt. Vernon, MO). Cv. Chambourcin was also acquired in 2021 and 2022 from a commercial vineyard in State College, PA (40°47′N; 77°51′W), and cv. Noiret was acquired in 2021 and 2022 from The Pennsylvania State University Russell E. Larson Agricultural Research Center at Rock Springs, PA (40°71′N; 77°97′W). Approximately 228 kg of each cultivar (3 kg of each was stored at −80°C for microvinification samples) was acquired in 2021 and 2022. Experiments conducted using these samples are described (Table 1).
Experiments conducted, including grapes used and relevant variables.
Pilot-scale (20 L) fermentations
Upon arrival at the research winery, grapes were sprayed with KMBS at a final rate of ~25 mg/kg. Grapes were crushed, destemmed, and homogenized into one large lot using a motorized crusher destemmer (Enoltalia). Following crushing and destemming, must (20 L) was aliquoted into 22.7-L food-grade buckets fitted with airlocks, in triplicate. KMBS was added to the must at a rate of 20 mg/L. The must was inoculated with ICV GRE at 0.264 g/L, which was rehydrated with Go-Ferm yeast nutrient according to manufacturer (Lallemand) directions. The fermentations were kept at room temperature (external temperature of ~25°C) and punched down twice a day. Musts were monitored for residual sugars with AimTab Reducing Substances Tablets. After 14 days, residual sugars were <0.1%, and musts were hand-pressed through cheesecloth into 11.3-L glass carboys for settling of sediment. After an additional 14 days, the wines were filtered through 0.5-μm number 3 filter pads (Buon Vino Mfg.) using a Flojet pump (Buon Vino Mfg.) and racked into clean glass carboys. Malolactic fermentation was not conducted for winemaking. Wines were then bottled in 750-mL clear glass bottles and capped with Stelvin capsule (Saranex lined) 30 × 60 caps (Amcor Flexibles). Wines were stored at 17.8°C for 30 days until used for analysis.
Microvinification (50 mL)
Acquired 2021 grapes (2 kg) were randomly sampled from the pilot-scale grapes for assessment of heterogeneity, then destemmed by hand and stored at −80°C until processing. Frozen grapes were thawed at 4°C for 12 hr. The entire grape sample was crushed by hand in a single-gallon plastic bag until the skin of every berry was broken, similar to that of a commercial crusher. To ensure repeatable ratios of solids-to-juice, the total grape sample was hand-pressed from the juice, and the wet solids (skins and seeds)-to-must volume ratio was found to be 0.57 g of solids per 1 mL of must for 2018 Chambourcin. This ratio was used for all experiments in the current study and is similar to other recorded solids-to-juice ratios (Roby and Matthews 2004). Crushed solids (22.8 g) were allocated into 50-mL centrifuge tubes. Juice was then decanted into the tubes to a final volume of 40 mL, with space left to allow the cap to rise during the fermentation. Must was inoculated as previously described for pilot-scale fermentations. The tubes were placed into a water bath at 35°C (or as specified below for temperature experiments). Microvinifications were conducted in quintuplicate. Fermentations were punched down and degassed twice daily by inverting the tubes and releasing the lids. The musts were fermented until residual sugar concentration was <0.1% (typically by 7 days), as measured by AimTab Reducing Substances Tablets. Samples were pressed (after 5, 7, or 9 days, depending on the treatment) by centrifuging the 50-mL tubes for 10 min at 4816 × g and 4°C in a Sorvall Legend XTR Centrifuge (Thermo Scientific) and vacuum filtered through 110-mm filter paper (Whatman plc) into 15-mL centrifuge tubes to decrease the amount of headspace. Pressed wines were stored at −80°C until analyzed.
For the fermentation condition characterization trials, fermentations were conducted in water baths at 30, 35, or 40°C, or in open air at room temperature (22.5°C). Temperatures were chosen based on typical red wine cap temperatures. Fermentations were punched down and degassed twice a day by inversion (inverted 10 times), stir rod punch-down (glass stir rod), or none (only degassed), depending on treatment. The length of the maceration was 5, 7, or 9 days, depending on the treatment. For the short maceration treatment (5 days), the samples were fermented to a residual sugar of <0.1% after pressing under the normal conditions. The microvinifications had shorter maceration time (to reduce the residual sugar to <0.1%) because of the different kinetics of the fermentation volumes and to avoid O2 ingress post-CO2 production. Treatments were not conducted in a full factorial design because of the number of variables. Statistical analysis of each variable (temperature, cap management technique, and time) was performed separately.
Fresh versus frozen microvinifications
In 2022, samples (the same cultivars and vineyards) were acquired for a fresh versus frozen microvinification trial. Postsampling, each cultivar was separated into two plastic bags with ~1.5 kg each (total of ~3.0 kg/cultivar). One bag was used immediately on sampling day for microvinification following the above procedures. The other bag was placed into a −80°C freezer for 30 days prior to microvinification. Both sets of microvinifications were fermented in quintuplicate.
Fermentation dissolved O2
During punch-down assessment for the 2018 cv. Chambourcin trial, the dissolved O2 was monitored throughout the fermentation and measured using a NOMASense O2 P6000 analyzer (Nomacorc). Within the cap management treatments (i.e., inversion and punch-down), Oxygen Sensor Spots SP-PSt3-NAU-D5-CAF (PreSens) were attached inside the centrifuge tubes (at the 15-mL line, in three tubes of each treatment) according to manufacturer instructions. O2 was measured at the start of fermentation and twice daily after the first day.
Adams-Harbertson assay
A version of the Adams-Harbertson assay using a Clariostar plate reader spectrophotometer (BMG Labtech) was used to measure phenolic compounds in all microvinification samples (2018, 2021, and 2022 vintages), following methods developed by Heredia et al. (2006). The Adams-Harbertson assay was used to measure iron reactive tannins (IRT), iron reactive phenolics (IRP), anthocyanins, large polymeric pigments (LPP), and small polymeric pigments (SPP). The wavelength used for absorbance measurements was 510 nm for IRT and IRP, and 520 nm for anthocyanins, LPP, and SPP.
Headspace solid-phase microextraction gas chromatography-mass spectrometry (HS SPME GC-MS/MS)
Untargeted and semiquantitative measures were conducted using a method developed by Awale et al. (2021) on 2021 samples. The HS SPME GC-MS/MS system uses a Combi-PAL autosampler on an Agilent 7890A gas chromatograph with an Agilent 7000C triple-quad detector (Agilent Technologies). A DVB/C-WR/PDMS fiber (Restek) was used for sampling and extraction. Fibers were conditioned before sampling according to manufacturer recommendations. For extraction, 5-mL samples were placed in 20-mL vials with 2.5 g of salt and internal standard and incubated at 40°C for 5 min. The fiber was exposed at 40°C for 30 min in the headspace of the samples while being agitated at 500 rpm. The fiber was desorbed in the inlet at 260°C for 16 min in splitless mode. A Stabilwax-MS column (30 m × 0.25 mm i.d., 0.25 μm film thickness) and helium gas (flow rate: 2 mL/min) were used for analysis. The GC oven temperature followed the following program: 35°C for 5 min, then increased at a rate of 6°C/min to 240°C, which was held for 10 min. The source temperature, MS1, and MS2 quadrupoles were set to 230, 150, and 150°C, respectively. The electron ionization (EI) source operated at 70 eV. Helium quench gas was set at 2.25 mL/min, and nitrogen (99.999% purity, Praxair) was set at a 1.5 mL/min flow rate for collision gas. EI spectra were collected in scan mode from 35 to 350 amu (atomic mass unit), with 7.5 scans/sec. Qualitative analyses were established using the MassHunter Workstation software ver. B.07.00 (Agilent Technologies). Qualitative data processing was conducted using XCMS online (Tautenhahn et al. 2012) for feature detection, alignment, retention time correction, and normalization. Quantitative analyses were completed using MassHunter Quantitative Analysis software ver. B.07.01 (Agilent Technologies). Frozen wine samples were thawed at 4°C for 12 hr and diluted by 50% in model wine solution (12.5% ethanol, 5 g/L tartaric acid, adjusted to pH 3.5 with 5M sodium hydroxide) for ethanol concentration control and to not overwhelm the GC-MS column or detector with analytes. Aliquots containing 5 mL of diluted sample were added into 20-mL amber SPME vials (Restek). All samples were run in analytical duplicate.
For quantification, 2-octanol was used as the internal standard at 0.05 ng/mL in each vial. A multiple reaction monitoring transition of 97 to 55 m/z ratio was used to better isolate the internal standard. Full scan mode was used for quantification of other compounds (35 to 350 m/z). Dilutions were made from 1 to 500 ng/mL, with calibration curves used depending on the linear range. A table of the chosen m/z ratios and linear ranges for each compound is included (Supplemental Table 1).
Statistical analysis
An analysis of variance (ANOVA) (α = 0.05) was conducted on all phenolic and volatile concentrations when there were more than two treatments, followed by a Tukey honestly significant difference test (α = 0.05) to determine significance between groups. A t-test was conducted on phenolic concentrations (α = 0.05) when there were only two treatments. An unsupervised principal component analysis (PCA) was conducted on volatile metabolomic features found using untargeted GC-MS on all microvinifications and pilot-scale fermentations. Statistical analysis and graphics were computed using the software RStudio (ver. 1.3.1073), including the R packages ‘agricolae’ (de Mendiburu 2020), ‘dplyr’ (Wickham et al. 2022), and ‘ggplot2’ (Wickham 2016).
Results and Discussion
Concerns remain about the effectiveness of small-scale fermentations as tools to characterize fruit chemistry and predict wine qualities, mainly because of issues of consistency and oxidation. This study aimed to address these concerns and optimize fermentation parameters for 50-mL on-skin microfermentation lots. We monitored O2 pickup under different cap management techniques and examined the effects of temperature, fermentation time, and cap management on phenolic content and overall volatile compound composition (experimental outline provided in Table 1).
Adjusting fermentation parameters produced results similar to those observed in larger-scale fermentations. For example, higher temperatures led to greater tannin extraction and lower anthocyanin extraction. Similarly, extended maceration times resulted in increased tannin levels. Comparisons with pilot-scale 20-L fermentations, conducted using a standard protocol, showed comparable outcomes in phenolic extraction.
Microvinification dissolved O2
Reducing fermentation volume may introduce excessive O2 because of the increased ratio of the surface area of the container to the volume of the must. To assess this risk during the microvinification of 50-mL samples, the dissolved O2 was monitored under two cap integration methods: inversion and punching down. At the start of fermentation, both treatments were near saturation levels for aqueous solutions at 35°C, which is expected in fermentations of any volume (Cheynier et al. 1993). While the inversion treatment initially exhibited significantly higher dissolved O2 levels than the punch-down method, with mean values of 5806 and 4890 μg/L, respectively (data not shown in the interest of figure clarity), most of the O2 was consumed on day 1 for both treatments. The concentration dropped to 2.0 and 2.3 μg/L for the inversion and punch-down methods, respectively, after day 1 (Figure 2). On day 2, the concentration of O2 increased in the punch-down approach; subsequently, however, the dissolved O2 remained low in this treatment for the remainder of the fermentation. This is likely because this method required exposing the sample to the atmosphere (through crushing) prior to sufficient CO2 production. The concentration of O2 in experimental samples was considerably lower than reported in other fermentations, including those on the commercial scale (5000 L), which typically have concentrations of at least 50 μg/L (Moenne et al. 2014). With the exception of elevated O2 levels in the punch-down treatment for one early measurement, a maximum of ~35 μg/L of dissolved O2 was observed on the final day. Despite the upward trend observed later in fermentation, the levels remained well below 260 μg/L, where characteristics indicating oxidation (i.e., phenolic oxidation occuring) in red wine are present (Singleton 1987). However, it is likely that once the protection of active CO2 production subsides, smaller fermentation lots take on more O2. Because of this, samples were pressed prior to pomace cap fall (when the solid grape matter is not suspended from CO2 creation and falls to the bottom of the vessel). Ultimately, this experiment demonstrates that even with the inclusion of punching down as a cap management practice, anaerobic conditions can be maintained in microvinifications and are comparable to those of larger volumes (20 L or greater).
Dissolved oxygen concentration of must in microvinifications using two cap integration methods (inversion and active punch-down) for 2018 cv. Chambourcin samples (n = 3). Mean values for inversion and punch-down at the start of fermentation were 5806 and 4890 μg/L, respectively; these values are not significantly different but are excluded from the plot to enhance readability. Data are displayed as the mean ± standard deviation.
Comparison of phenolics between microvinification and pilot-scale fermentation
Pilot-scale fermentations conducted in 2021 at ambient temperature were used to compare microvinification to a common pilot-scale vinification protocol. Because of the differences in fermentation kinetics of the pilot- and microvinification-scale fermentations, the fermentation times until the residual sugar reached <0.1% were ~7 and 14 days, respectively. This difference could potentially introduce changes in phenolic and volatile composition, which are not a result of the volume. Because researchers have identified negligible heat retention during fermentation on the pilot scale, the temperature was not monitored (Sparrow and Smart 2015). While differences in the phenolics were observed within the microvinification treatments and groups of treatments (grouped by the independent variable), overall, the microvinifications were similar to the pilot-scale fermentations (Figure 3). This appears to be a positive outcome, demonstrating how the microvinification replicates (within treatments) had a smaller variance (with a larger number of samples) in comparison to pilot scale. The distribution of each treatment in 2021 (cultivars shown separately) demonstrates that microvinifications can be utilized as a proxy for 20-L fermentations regarding phenolics. Because these data encompass the entire set of microvinifications within the cultivar and year and contain each of the treatments, outliers are expected to be observed. Overall, the microvinifications were observed to be quite similar to the pilot-scale fermentations in terms of phenolics. The phenolic compound groups in the microvinifications were not significantly different from those of the pilot-scale fermentations.
Overall concentrations of iron reactive tannins (IRT) and iron reactive phenolics (IRP), both measured in mg/L catechin equivalents (CE), and anthocyanins (expressed as malvidin-3-glucoside equivalents [M3GE]) for microvinifications and pilot-scale fermentations for 2021 cv. Noiret and 2021 cv. Chambourcin, measured using the Adams-Harbertson assay. The “microvinification” box contains data from the fermentation parameter groups (i.e., temperature, time, and cap management) and all included treatments. Microvinification, n = 50 (across all treatments); pilot, n = 3.
The pilot-scale fermentations were assigned as the control and compared to the various microvinification variables (i.e., temperature, maceration, and cap integration) using Dunnett’s test. Chambourcin anthocyanin concentrations in the room temperature treatment were significantly higher than those in the pilot-scale fermentations (Figure 4). This was unexpected because both were kept at the same ambient external temperature; however, as previously mentioned, small-scale fermentations may not attain the same temperature increase as that of larger-scale during the fermentation process, especially locally within the fermentation lot (Schmid et al. 2007). This potential variation in the fermentation temperature may have slightly elevated the internal temperature of the pilot-scale wines, making their actual temperature more comparable to the microvinifications, with temperature regulation in a water bath. Moreover, controlling the fermentation temperature in winemaking research is considerably more feasible with microvinification in a water bath compared to fermentations at scales of 20 L or more. In Noiret, IRTs were higher in the room temperature treatment than in the pilot-scale treatment, and IRPs were significantly higher with 5-day maceration than the pilot-scale fermentation. There were no other significant differences in the phenolic components of the wines between the microvinification treatments and their respective pilot-scale versions, supporting the idea that controlled-temperature microvinification can be analogous to pilot-scale winemaking in terms of phenolic components.
Concentrations of iron reactive tannins (IRT) and iron reactive phenolics (IRP), both measured in mg/L catechin equivalents (CE), and anthocyanins (expressed as malvidin-3-glucoside equivalents [M3GE]) in 2021 cv. Noiret wine, determined using the Adams-Harbertson assay. Dotted lines indicate different comparison groups and treatments (temperature: room temperature [Room], 30°C [30], 35°C [35], and 40°C [40]; time: 5 days [5D], 7 days [7D], and 9 days [9D]; cap integration: inversion [Inv], no punch-down [None], and punch-down [Punch]; and pilot-scale fermentation [Pilot]). Treatments sharing letters (within groups) are not significantly different according to Tukey’s honestly significant difference test (α = 0.05). An asterisk (*) indicates significant differences with the pilot-scale fermentation using Dunnett’s test (α = 0.05). Microvinification, n = 5; pilot, n = 3. Pilot-scale fermentation values are included for reference.
For both Chambourcin and Noiret microvinifications, sampling was a critical concern because grapes from different locations of the vineyard (or even different vines) could alter the chemistry of the resulting wine. Appropriate vineyard sampling techniques are necessary to obtain representative grape data and are key for a reduction in sample sizes (Meyers et al. 2011). To lessen the differences across the vineyard, microvinification samples were randomly selected from the grapes used in the pilot scale, thus ensuring a more representative sample. For commercial operations, sampling across vineyard plots and conducting individual microvinifications can be valuable to preemptively assess different locations, vine health, and potential characteristics of the resulting wine.
Comparison of volatile compounds between microvinification and pilot-scale fermentation
Similar to the comparison of phenolics, fermentations conducted in 2021 were used to compare microvinification to a common pilot-scale protocol for volatile aroma compounds. Unlike the similarities observed in their phenolic compositions, the two fermentation scales showed significant variations in the volatile compounds (Figure 5). The distribution of each treatment in 2021 (cultivars shown separately) demonstrates that these microvinifications were not optimized for comparison to 20-L fermentations in regard to volatile aroma compounds. Linalool in Chambourcin was the only compound that was not significantly different in concentration between the fermentation volumes. Even though the majority of the aroma compounds are significantly different based on the fermentation scale, aroma differences might not be present. For example, the odor threshold for ethyl caprate is 244 ng/mL in a 9% aqueous ethanol solution (Niu et al. 2019). For Noiret, microvinification (170.0 ng/mL) was observed to be below the limit of odor detection and pilot scale (294.4 ng/mL) was above, while Chambourcin microvinification (33.8 ng/mL) and pilot scale (143.1 ng/mL) were both below the limit of odor detection.
Overall concentrations of linalool, ethyl caprate, and β-damascenone (ng/mL) in wine for 2021 cv. Noiret and 2021 cv. Chambourcin, measured using solid-phase microextraction gas chromatography-mass spectrometry. The “microvinification” box contains data from the fermentation parameter groups (i.e., temperature, time, and cap management) and all included treatments. An asterisk (*) indicates significant differences between the two scales using a t-test (α = 0.05). Microvinification, n = 50 (across all treatments); pilot, n = 3.
Similar to the cumulative microvinifications in comparison to the pilot-scale fermentations, the individual treatments were also different from the latter (Figure 6). Most of the microvinification treatments were significantly different for β-damascenone, ethyl caprate, and linalool in Noiret, using the Dunnett’s test (α = 0.05). Regarding ethyl laurate, isoamyl acetate, and 1-nonanol, the treatments that were significantly different were those that were more extreme (e.g., nontemperature-controlled room temperature microvinifications). For example, the room temperature treatment was significantly different with regard to all compounds included in Figure 6. This demonstrates that the fermentations cannot retain heat at the 50-mL scale, similar to data modeled elsewhere (Colombié et al. 2007, Miller et al. 2019), because temperature drastically affects the extraction and retention of volatile compounds (Massera et al. 2021). Furthermore, other extreme fermentation conditions, including a maceration time of 5 days, led to significantly different results for the majority of compounds in Figure 6. Maceration time has been previously studied in relation to volatile composition for cv. Karaoğlan, with free aroma concentration reaching a maximum at 5 days and decreasing by 10 days of maceration (Yilmaztekin et al. 2015). Conversely, in the same study, glycosidically-bound aroma compounds increased over time, demonstrating that more glycosides can be extracted with longer maceration times. Because malolactic fermentation is important for commercial wineries and alters the aroma composition of wines, further studies should be conducted to characterize differences in pilot-scale fermentations and microvinification that arise from malolactic fermentation.
Concentrations of β-damascenone, ethyl caprate, ethyl laurate, isoamyl acetate, linalool, and 1-nonanol in wine for 2021 cv. Noiret, measured using solid-phase microextraction gas chromatography-mass spectrometry. Dotted lines indicate different comparison groups and treatments (temperature: room temperature [Room], 30°C [30], 35°C [35], and 40°C [40]; time: 5 days [5D], 7 days [7D], and 9 days [9D]; cap integration: inversion [Inv], no punch-down [None], and punch-down [Punch]; and pilot-scale fermentation [Pilot]). Treatments sharing letters (within groups) imply no significant differences using Tukey’s honestly significant difference test (α = 0.05). An asterisk (*) indicates significant differences with the pilot-scale fermentation using Dunnett’s test (α = 0.05). Microvinification, n = 5; pilot, n = 3. Pilot-scale fermentation values are included for reference.
Although many volatile compounds measured in this study were significantly different between microvinification and pilot-scale fermentations, some, such as linalool for Chambourcin, were not. Using this knowledge, specific volatile compounds could be optimized for microvinification according to the study objectives, for a more accurate measurement.
Effect of microvinification parameters on phenolic compounds
Phenolic concentration varied among the treatments and independent variables (temperature, maceration time, and cap integration); the majority of the microvinification treatments caused a significant effect on phenolic concentration (Table 2). Regarding Noiret, only the anthocyanins between the independent variables (temperature, maceration time, and cap management) were not significantly different (α = 0.5). For the 2021 Chambourcin, anthocyanins and LPP were significant, but the independent variables were not. This demonstrates that the different groups of independent variables were quite similar. For the 2018 Chambourcin, only LPP was not significant for both treatments and sets. This level of difference was expected because alterations in fermentation parameters (i.e., temperature, time, and cap management) at any scale led to differences in wine phenolics (Lerno et al. 2017, 2018).
Analysis of variance between all treatments and treatment groups within cultivar and vintage. IRT, iron reactive tannins; IRP, iron reactive phenolics; LPP, large polymeric pigments; SPP, small polymeric pigments.
Temperature
Temperature plays a critical role in the final quality of a wine; as the scale of vinification decreases, atmospheric temperature overwhelms any temperature increase caused by fermentation, which is one of the limitations of small-scale vinification. In red winemaking, heat from the fermentation process is utilized, with higher temperatures allowing for the increased extraction of most grape polyphenols (Sacchi et al. 2005). Engineering-based models have demonstrated that as the surface area-to-volume ratio of the vessel increases, heat is more rapidly lost (Colombié et al. 2007, Miller et al. 2019, Miller and Block 2020). As such, it is common for smaller fermentations to be externally heated to maintain typical commercial-scale fermentation temperatures. Although the exact temperatures observed in commercial fermentation vary, the temperature in a 500,000-L red wine fermentation set-up has been modeled to be between 24 and 36°C between the cap and the bottom of the vessel, whereas the cap may exceed 37°C in the center (Miller et al. 2019). To mimic the temperature distribution present in commercial fermentation, a range of 30 to 40°C was selected for the treatments, with the additional inclusion of an ambient temperature control.
After the treatments, anthocyanin concentrations generally decreased as temperature increased (Figure 4 contains only Noiret data; other comparisons are shown in Supplemental Table 3 and Supplemental Figure 1). However, a few exceptions emerged; for example, the room temperature treatment had among the highest concentrations of anthocyanins for all cultivars and vintages, although significant differences were not observed in all temperature trials. Additionally, the 2021 Chambourcin displayed a lower concentration of anthocyanins at 30°C than at 35°C. The trend of higher fermentation temperatures causing a decrease in anthocyanin concentrations has been observed in other studies; higher temperatures facilitate a faster rate of creation of polymeric pigments, thus lowering the concentration of anthocyanins (Somers and Evans 1986). In the Noiret samples, SPP significantly increased from 1.3 to 2.9 au between room temperature and 40°C (Supplemental Table 3). A difference in IRT was only observed in Noiret wines, with the room temperature treatment leading to an increase over that of all other temperatures. All other evaluated temperature treatments yielded similar IRT and IRP values (Figure 4). This direct relationship between the temperature and tannin concentration in Noiret follows what is expected in extraction during fermentation (Sacchi et al. 2005).
The consistency for Noiret wines, as determined by the coefficient of variation (CV), was generally comparable to that of the pilot-scale fermentation lots for all temperatures other than 40°C. In Chambourcin, the consistency for anthocyanin extraction was acceptable; however, tannin-related metrics such as IRT and LLP had CVs above 50% for many treatments, including pilot-scale, likely because the values were approaching the assay’s limit of detection.
Maceration time
The effect of maceration time was evaluated for fermentations held at 35°C for 5, 7, and 9 days, which all used inversion cap management. A period of 7 days was chosen as the default maceration time because of coinciding with the pomace cap fall, the point at which insufficient CO2 is being produced to protect the fermentation lot from O2. The inversion punch-down approach was chosen as the cap management method because it displayed greater consistency for IRT in the 2018 trial. If using higher volumes (i.e., 1-L fermentations), more complex punch-down methods might be necessary, as previously noted (Sparrow and Smart 2015). With an increased maceration time, the concentration of anthocyanins decreased or remained unchanged (Figure 4 and Supplemental Figure 1). The concentration of anthocyanins significantly decreased in Noiret and the 2021 Chambourcin on day 9; this is consistent with previous observations of larger fermentations, wherein longer maceration times allowed for the conversion of more anthocyanins into polymeric pigments (Somers and Evans 1986). In Noiret, the concentration of SPP increased between 5 and 9 days from 1.94 to 2.37 au (Supplemental Table 3), strengthening previous observations that polymerization occurs with an extended maceration time. The concentration of IRT in the 2018 Chambourcin significantly increased between 5 and 9 days (Supplemental Figure 1), which aligns with expectations that a longer extraction time allows for a higher extraction of tannins from the berry skin and seeds (Alencar et al. 2018). In the 2021 Chambourcin, tannins in all treatments were below the limit of detection for the assay used. No differences were observed in the Noiret tannin concentrations based on the maceration time. IRPs behaved similarly to tannins, significantly increasing between 5 and 9 days in the 2018 Chambourcin, whereas no differences were observed in the 2021 Chambourcin and Noiret (Figure 4).
Although maceration time was investigated for only one temperature in this study, there is potential for an interaction between the temperature and maceration time. Cooler fermentation temperatures have been shown to reduce fermentation rate, leading to slower ethanol creation to facilitate tannin extraction (Massera et al. 2021). As such, further investigations are needed to optimize maceration time at lower temperatures.
Cap integration
Reducing the scale of red wine fermentation presents a challenge, as it leads to limited practical methods for mimicking commercial cap integration. Cap integration at larger volumes is critical to facilitate fermentation and skin extraction and minimize acetobacter growth (Bartowsky and Henschke 2008, Guerrini et al. 2017). However, further investigation was needed to assess the benefits of cap integration, if any, for microvinification. Inversion cap integration and manual punch-down were investigated for all years; a “no-punch-down” treatment was added in 2021 to investigate whether punch-downs were indeed necessary, given that they are a laborious element of microvinification. In the 2021 Noiret samples, anthocyanins and IRP were significantly higher for the inversion treatment compared to both the manual and no-punch-down treatments (Figure 4). In the 2018 Chambourcin, the punch-down treatment significantly increased the IRT compared to inversion, although at a very small and practically unimportant level. No differences from the punch-down method were observed in the 2021 Chambourcin. Generally, the cap integration method had little to no effect on polyphenol extraction in microvinifications, indicating that effort for punch-downs was not necessary. This may be because the shallow cap formed during microvinification, typically only ~1 cm deep, is constantly immersed by the fermenting must. This is unlike the case of a larger fermentation, where one can expect the top of the cap to dry out if not resubmerged.
From a practical standpoint, manual punch-downs were the most labor intensive because the cap needed to be removed and replaced and the punch-down tool needed to be cleaned and sanitized between replications. In terms of labor intensity, the inversion treatment is similar to the no-punch-down method because in both cases, the excess CO2 was allowed to escape twice a day, requiring each cap to be briefly loosened.
Fresh versus frozen microvinification
Because many advanced analytical techniques for grapes and wine require specialized equipment and cannot be done immediately after harvest, samples are typically frozen for later testing. Although freezing grapes prior to actual winemaking is uncommon, it is standard practice to freeze grape samples for analytical analyses (Flora 1976, Schmid et al. 2007, Schmid and Jiranek 2011, Pedrosa-López et al. 2022). Because of the large number of variables explored, the use of frozen fruit was required; this also offered the opportunity to investigate the effect, if any, of freezing fruit prior to microvinification. Freezing grapes prior to fermentation has been believed to affect the extraction of various metabolites, potentially increasing bitterness and astringency while reducing the “fresh flavor” of the wine (Flora 1976). However, more recent studies have shown that freezing fruit prior to fermentation does not significantly affect the metabolomic composition and color of Merlot grapes (Schmid and Jiranek 2011) or the chemical and sensory properties of Cabernet Sauvignon and Cabernet franc (Schmid et al. 2007). Furthermore, comparing fresh versus frozen fruit provided an opportunity to investigate the effect of the inevitable use of frozen fruit for microvinifications as a manner of sample preparation.
We found that the cultivars (Noiret and Chambourcin) were not significantly different in terms of IRT or IRP (Figure 7), which agrees with other observations of freezing grapes prior to maceration (Schmid et al. 2007, Schmid and Jiranek 2011). The freezing treatment resulted in a lower anthocyanin concentration than the fresh treatment at 918.3 and 1148.1 mg/L malvidin-3-glucoside (M3G) equivalents, respectively, for Noiret, and 379.6 and 572.5 mg/L M3G equivalents, respectively, for Chambourcin. Others have observed that pigments either increase, caused by the fragmentation of plant cell structures (Threlfall et al. 2006), or remain unchanged (Schmid et al. 2007) when using frozen grapes. The differences in our findings may be the result of rapidly freezing small volumes of grapes at −80°C, which did not allow for the same degree of ice crystal damage caused in other studies. While the volatile analysis of fresh versus frozen grapes did not fall within the scope of our study, an increase in the concentration of some volatile compounds (terpenes, esters, acids, and β-damascenone) has previously been seen in Muscat wines when freezing fruit prior to fermentation (Pedrosa-López et al. 2022).
Concentrations of iron reactive tannins (IRT) and iron reactive phenolics (IRP), both measured in mg/L catechin equivalents (CE), and anthocyanins (malvidin-3-glucoside equivalents [M3GE]) in wine for 2022 cv. Noiret and Chambourcin, measured using the Adams-Harbertson assay in the fresh versus frozen trials. An asterisk (*) indicates significant differences between treatments (within cultivars) using a t-test (α = 0.05). n = 5 for all treatments.
Effect of microvinification on volatile composition
An untargeted screening was carried out to determine which volatile compounds might be affected by microvinification conditions. Feature extraction of the GC-MS data was conducted using XCMS online (Tautenhahn et al. 2012), where each feature corresponded to a specific m/z value, retention time, and ion intensity. This analysis resulted in 1441, 1333, and 1312 metabolite features, with 1400, 1134, and 1109 features significantly affected by the fermentation parameters (ANOVA [p = 0.05]), in the 2018 Chambourcin, 2021 Chambourcin, and 2021 Noiret microvinifications, respectively. Because many features are ion fragments from a single compound, the actual number of unique metabolites is likely lower. Additionally, analyzing all fermentations (across vintage, cultivar, and scale) in XCMS online identified 64 significant features.
PCA confirmed that microvinification can yield volatile profiles across vintages and cultivars that are comparable to pilot-scale fermentations (Figure 8). PC1 and PC2 explained 62.8% and 11.9% of the variance, respectively, separating samples by fermentation size, vintage, and cultivar. Notably, the 2021 pilot-scale Chambourcin and Noiret overlapped only with their corresponding microvinifications, demonstrating strong similarity. Because microvinification resembles pilot-scale fermentation more closely than proxy laboratory methods (acid or enzymatic hydrolysis), it is a valuable tool for studying wine aroma (Kennison et al. 2008).
Principal component analysis (PCA) score plots of untargeted volatile metabolite features by solid-phase microextraction gas chromatography-mass spectrometry for the 2018 and 2021 cv. Chambourcin (n = 40 and 50, respectively) and 2021 Noiret (n = 50) microvinifications, and the 2021 cv. Chambourcin (n = 3) and 2021 cv. Noiret (n = 3) pilot-scale fermentations. PCA was performed using logarithm-transformed data.
Additional PCAs (Figure 9) were performed for 2018 Chambourcin, 2021 Chambourcin, and 2021 Noiret. PC1 and PC2 explained 45.4% and 16.1% (2018 Chambourcin), 37.7% and 14.9% (2021 Chambourcin), and 30.4% and 14.1% (2021 Noiret), respectively, of the total variance. Pilot-scale and room temperature treatments overlapped in 2021 Noiret and shared the same PC2 position in 2021 Chambourcin, showing similarities at ~25°C. By contrast, the maceration and punch-down treatments were conducted at 35°C, shifting the volatile profiles in line with modifications of fermentation parameters (Salinas et al. 2005, Massera et al. 2021).
Principal component analysis (PCA) score plots of untargeted volatile metabolite features by solid-phase microextraction gas chromatography-mass spectrometry, separated by microvinification treatments in cultivars and vintages: 2021 cv. Chambourcin, 2021 cv. Noiret, and 2018 cv. Chambourcin. Both 2021 cultivars include treatments not included in the 2018 cultivar, designated by an asterisk (*). PCA was performed using logarithm-transformed data. Microvinification, n = 5; pilot, n = 3.
The compounds contributing most to the PCA separations were tentatively identified using the National Institute of Standards and Technology MS Search software. The top 50% of significant features driving the PCA models were selected for quantification, but to ensure confident tentative identification, the quantification was further restricted to those significant features with sufficiently clean MS spectra. Using the untargeted data for 2018 Chambourcin, 18 unique compounds were identified and quantified from the top significant features. The data of this vintage were used solely to identify significantly different compounds for simplicity in the analysis. Nonetheless, many of the retention times and spectra overlapped with compounds driving the PCAs of other wines. Additionally, because of the complex aromatic composition differences among cultivars, especially between hybrids and Vitis vinifera, each cultivar could be affected to varying degrees. Therefore, additional cultivars should be studied to demonstrate these distinct compositional differences.
Volatile compound concentration
Volatile compound concentrations were generally less affected by the microvinification parameters within cultivars than were those of the phenolic compounds for the 2021 Noiret (Figure 6) and 2018 Chambourcin (Supplemental Table 4). In the temperature trials, the linalool concentration was significantly lower in the room temperature treatment than in the 40°C treatment for both Chambourcin and Noiret. This difference in concentration has been previously observed where lower maceration temperatures resulted in lower terpene concentrations (Salinas et al. 2005, Massera et al. 2021), and may be due to increased volatilization or degradation at higher temperatures. Conversely, the ethyl caprate level was significantly higher in both cultivars in the room temperature treatment; because of the role of fermentation temperature in ester formation and retention, this is to be expected (Massera et al. 2021). Previous studies have shown that the highest concentrations of ethyl butanoate and ethyl hexanoate in wines were produced at a fermentation temperature of 20°C, whereas ethyl octanoate peaked at 24°C, compared to higher fermentation temperatures (Rollero et al. 2015). Higher alcohols are not temperature dependent and are instead affected by yeast strain-derived changes (Massera et al. 2021). This was demonstrated for Noiret, with the 1-nonanol concentration showing no significant differences across temperatures.
In the maceration trials, the ethyl caprate concentration was significantly lower after 9 days of maceration, compared to 5 days for both cultivars. This supports the abovementioned point, where free volatiles decrease over time during maceration (Yilmaztekin et al. 2015). For Noiret, many of the compounds (including ethyl laurate and isoamyl acetate) also followed this trend. Conversely, in both cultivars, the linalool concentration increased over time, which was significant only for Chambourcin. This is supported by the literature, where compounds residing predominantly in the skins as glycosides are able to increase in concentration over longer extraction times because of the breaking of glycosidic bonds (Yilmaztekin et al. 2015). A fewer number of the compounds measured in the microvinification trials followed this trend.
Cap management influences the level of β-damascenone, with lower concentrations observed in the punch-down treatment compared to inversion for Chambourcin. On the other hand, the opposite was noted for Noiret, with lower levels of β-damascenone arising from the inversion method of cap management. The ethyl caprate concentration in Chambourcin was significantly higher in the no-punch-down treatment. This may be due to the volatility of esters, allowing their release in the punch-down and inversion methods (Saerens et al. 2010).
Conclusion
Microvinification can serve as a viable tool to assess both the phenolic and volatile profiles extracted and produced during fermentation, functioning as an alternative to conventional solvent- or enzyme-based methods. Although some volatile compounds showed differences from 20-L pilot-scale fermentations, phenolic extraction was highly comparable, with microvinifications replicating the tannin and anthocyanin profiles of pilot-scale wines. When fermentation parameters (e.g., temperature, maceration time, and cap integration) are controlled, microvinification yields dissolved O2 levels and replicability similar to larger volumes. Moreover, microvinification enables researchers to capture the inherent variability of the vineyard using a few grape clusters, rather than pooling fruit from multiple sites or blocks to meet the volume requirements for fermentations of 50 L or more. This makes microvinification especially valuable in exploring viticultural factors that may be masked by large-scale homogenization. Overall, these findings demonstrate that microvinification is not inherently less valid than 20-L pilot-scale fermentations for predicting wine chemistry and can be employed to produce reliable, informative insights into both phenolic and volatile wine composition, provided that grape sampling adequately represents the vineyard block.
Supplemental Data
The following supplemental materials are available for this article in the Supplemental tab above:
Supplemental Table 1 Compounds used for solid-phase microextraction gas chromatography-mass spectrometry, including respective ions and linear ranges.
Supplemental Table 2 Basic chemistry of fruit. TSS, total soluble solids; TA, titratable acidity.
Supplemental Table 3 Phenolic composition of microvinifications. Data are displayed as mean ± standard deviation. IRT, iron reactive tannins; IRP, iron reactive phenolics; LPP, large polymeric pigments; SPP, small polymeric pigments.
Supplemental Table 4 Solid-phase microextraction gas chromatography-mass spectrometry concentrations. Data are displayed as mean ± standard deviation.
Supplemental Figure 1 Concentrations of iron reactive tannins (IRT) and iron reactive phenolics (IRP), both measured in mg/L catechin equivalents (CE), and anthocyanins (malvidin-3-glucoside equivalents [M3GE]) in wine for 2018 and 2021 cv. Chambourcin, measured using the Adams-Harbertson assay. Dotted lines indicate different comparison groups and treatments (temperature: room temperature [Room], 30°C [30], 35°C [35], and 40°C [40]; time: 5 days [5D], 7 days [7D], and 9 days [9D]; cap integration: inversion [Inv], no punch-down [None], and punch-down [Punch]; and pilot-scale fermentation [Pilot]). Treatments sharing letters (within groups) imply no significant difference using Tukey’s honestly significant difference test (α = 0.05). An asterisk (*) indicates significant differences with the pilot-scale fermentation using Dunnett’s test (α = 0.05). Microvinification, n = 5; pilot, n = 3. Pilot-scale fermentation values are included for reference.
Data Availability Statement
All data underlying this study are included in the manuscript and its supplemental information.
Footnotes
Warren ER, Fredrickson A and Kwasniewski MT. 2025. Assessment of on-skin microvinification in the evaluation of berry-derived wine flavor components. Am J Enol Vitic 76:0760013. DOI: 10.5344/ajev.2025.24058
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 November 2024.
- Accepted April 2025.
- Published online June 2025
This is an open access article distributed under the CC BY 4.0 license.
References
Vol 76 Issue 1
Jump to section
Related Articles
Cited By...
More from this TOC section
Similar Articles
