Metschnikowia pulcherrima as a Tool for Sulphite Reduction and Enhanced Volatile Retention in Noble Rot Wine Fermentation

Abstract

The use of non-Saccharomyces species is gaining momentum in modern winemaking as part of broader efforts to reduce chemical inputs and adapt to climate-driven challenges. In this study, Furmint grapes were harvested at two distinct ripeness levels: an early harvest with healthy berries and a late harvest that included botrytized fruit. Two oenological protocols were compared: a conventional sulphur dioxide-based protocol and an alternative bioprotection-oriented approach that minimized SO₂ additions. Bioprotection was carried out using Metschnikowia pulcherrima, followed by sequential inoculation with Torulaspora delbrueckii and Saccharomyces cerevisiae. Grape-derived tannins (from skin and seed) were also added to inhibit oxidative enzymes such as laccase. Fermentation was monitored using standard analytical techniques, with volatile aroma profiles characterized by HS-SPME-GC-MS. Results showed that harvest timing and botrytization strongly influenced the chemical composition of the wines. Moreover, the treatment protocol had a marked effect on the final sensory profile. Wines produced with the bioprotection-oriented protocol displayed enhanced aromatic complexity, particularly through higher concentrations of esters and higher alcohols. Overall, the alternative protocol involving M. pulcherrima-based bioprotection resulted in wines with more pronounced floral and fruity notes, supporting its potential as a viable strategy for producing expressive wines under evolving climatic conditions.

1. Introduction

Aroma and sensory quality are critical parameters in wine production, influencing both consumer acceptance and the perceived typicity of a wine. With growing consumer interest in authenticity and sensory expression, the control of aroma formation and retention has become central in oenological research and practice [1]. Wine aroma arises from a complex interplay of grape-derived precursors, fermentation-derived compounds, and post-fermentation modifications, with volatile esters, higher alcohols, thiols, and terpenes being key contributors to aromatic complexity [2].

Furmint (Vitis vinifera L.), the flagship white grape of the Tokaj Wine Region in northeastern Hungary, offers high acidity, good sugar accumulation, and significant aromatic potential. It is the dominant cultivar in Tokaji wine production, occupying over 60% of vineyard area within the PDO (Protected Designation of Origin) zone [3]. The Tokaj region, recognized as a UNESCO World Heritage Site since 2002, is globally renowned for its botrytized sweet wines (e.g., Tokaji Aszú), produced from grapes affected by Botrytis cinerea under specific microclimatic conditions [4]. Factors such as the presence of fog, extended autumn ripening, and volcanic soils facilitate noble rot development, which profoundly alters the chemical composition of grapes through enzymatic and oxidative transformations [5].

The Furmint grape variety is moderately vigorous, upright in growth, and shows moderate drought tolerance but is sensitive to downy mildew and grey rot in humid climates [6]. Its must contains a range of aroma-active compounds, including terpenic alcohols, six-carbon aldehydes and alcohols, caproic acid, benzyl alcohol, and α-butyrolactone [7,8,9]. Terpenes—present in both free and glycosylated forms—contribute significantly to aromatic complexity, especially when released by enzymatically active yeasts. Additional aroma contributors include esters, nor-isoprenoid ketones (e.g., β-damascenone), and sulphur- or nitrogen-containing compounds such as mercaptans, pyrazines, and pyridines [10,11,12]. Esters, particularly those responsible for fruity and floral notes, are the most abundant volatile compounds in young Furmint wines. Although they contribute to the aromatic profile, they are not grape-derived but are primarily formed during fermentation. Their formation is influenced by yeast metabolism, fermentation temperature, must clarity, grape maturity, and SO₂ addition [13,14,15]. Polyphenols also affect both aroma and mouthfeel. Although volatile phenols are present at low levels in grapes, they can be produced during fermentation by microbial conversion of hydroxycinnamic acids—especially by yeasts such as Brettanomyces, and also by lactic acid bacteria—yielding compounds like 4-vinylguaiacol and 4-ethylguaiacol. These compounds may contribute spicy or smoky notes at low concentrations, but they are generally regarded as spoilage markers when present in higher amounts, particularly 4-ethylphenol and 4-ethylguaiacol [16,17].

Fermentation microbiology plays a central role in wine aroma development. Yeasts, especially S. cerevisiae and non-Saccharomyces species, interact with microbes like B. cinerea to shape volatile compound production, including esters, higher alcohols, and terpenes. Winemakers use microbial management to guide these processes and enhance regional expression [18,19,20]. Several non-Saccharomyces yeasts such as Torulaspora delbrueckii, Lachancea thermotolerans, and Metschnikowia pulcherrima offer bioprotective benefits [21,22]. They suppress spoilage organisms through oxygen competition and metal chelation, while boosting aroma-active compounds (esters, thiols), glycerol, and organic acids [22]. These traits improve microbial stability, sensory quality, and mouthfeel. M. pulcherrima is widely used for bioprotection, showing low alcohol production, support for ester formation and glycerol production with S. cerevisiae, and inhibition of spoilage yeasts and bacteria [23]. It produces pulcherrimin, a red antimicrobial pigment that binds iron, limiting pathogen growth [24].

During late harvest, grape dehydration and B. cinerea activity alter chemical and aromatic profiles [2]. Compounds like furfural, benzaldehyde, and fungal alcohol (1-octen-3-ol) emerge, imparting almond or mushroom-like aromas [25]. Additional esters form, while Botrytis degrades amino acids and terpenes, reducing C3–C5 alcohols and terpene levels [26].

Non-saccharomyces yeasts are increasingly used to reduce alcohol, increase aroma complexity, and meet consumer preferences. S. bacillaris, T. delbrueckii, and P. kluyveri contribute to fruity, floral aromas and increased glycerol while minimizing volatile acidity [27,28,29,30,31,32,33]. L. thermotolerans enhances acidity and stone fruit notes. Some strains also boost polysaccharides, improving mouthfeel [30,31]. S. bacillaris stands out for its glycerol production and clean flavour profile. It is common in botrytized grapes and was first isolated in Tokaj, where it contributes to aroma and texture in Aszú wines [5,34]. Since these yeasts ferment only to low alcohol levels, co-inoculation with S. cerevisiae is required to complete fermentation. This strategy enables complex, full-bodied wines enriched in specific aroma compounds like linalool and succinic acid [35,36]. Enzymatic browning affects both colour and aroma in food products. Polyphenol oxidase (PPO) and catechol oxidase catalyze the oxidation of phenolic compounds into quinones, which polymerize into melanin pigments. Oxygen is essential in this process. Browning extent depends on PPO activity, and control efforts target PPO inhibition [37,38].

Laccase from B. cinerea oxidizes phenolics, promoting browning, off flavours, and SO₂ depletion. Its stability at wine pH makes it persistent during vinification [9,39]. Control methods include minimizing oxygen, removing infected skins, and rapid clarification [40]. While SO₂ remains the main inhibitor, M. pulcherrima shows promise as a biocontrol agent due to pulcherrimin production [41,42,43].

Bustamante et al. [43] demonstrated that M. pulcherrima can prevent oxidative browning in white wines, but only when grapes are healthy; its efficacy diminishes in the presence of laccase, which is often active during noble rot. To address this, oenological tannins were applied as supplementary protection. These tannins, increasingly used in sustainable winemaking, contribute to aroma complexity and structure while reducing the need for sulphur. Chemically, they are classified as hydrolysable (e.g., gallotannins, ellagitannins) or condensed tannins, each with distinct structures and reactivities [44]. Their functions include protein removal, metal chelation, antioxidant protection, and enzymatic oxidation inhibition [45]. Vignault et al. [46] confirmed that wine tannins can suppress laccase activity and help preserve wine colour during oxidation-prone vinification.

These complex interactions raise important questions about microbial contributions to aroma and phenolic stability in noble-rotted Tokaji wines. Based on the preceding context, this research investigates the bioprotective potential of the wild yeast strain M. pulcherrima in Tokaji winemaking, with a particular emphasis on the Furmint grape and botrytized fruit. While its efficacy has been explored in other cultivars, its application to Tokaji grapes remains underexplored [23,47]. The study hypothesizes that M. pulcherrima can inhibit spoilage organisms and mitigate enzymatic browning under botrytized conditions, making it a viable alternative to sulphur dioxide [48,49,50]. However, sulphur reduction alone may be insufficient. We therefore propose a combined strategy involving M. pulcherrima and condensed oenological tannins to enhance phenolic stability. Tannins are expected to bind oxygen and stabilize oxidation-sensitive compounds, thereby supporting the bioprotective performance of M. pulcherrima. Additionally, we examine the yeast’s effect on volatile retention, with the expectation that it enhances key botrytis-related aroma markers (e.g., 2-phenylethanol and its esters) while preserving fruity esters in the absence of SO₂. The study ultimately aims to advance sustainable microbial management and aroma optimization strategies for Tokaji wines.

2. Materials and Methods

2.1. Microvinification and Sampling

Furmint grapes from the Tokaj wine region were used for the study, harvested on 11 September 2024 (with totally healthy state—“early” harvest) and on 21 September 2024 (with 60–80% botrytised grapes—“late” harvest) both manually and through conventional cultivation. The vinification processing took place at Tokajbor-Bene Winery, Bodrogkeresztúr. After destemming, pressing was carried out without crushing, then, after settling, the volume was divided into 50 litre tanks and the treatments presented in Figure 1. were applied in parallel (series a and b) at early and late harvest times. “E” samples represented the “early” harvest: initial juice composition was 222 g/L of sugar, with a total acidity of 5.80 g/L of titrable acid and a pH value of 3.4. “L” samples represented the “late” harvest: initial juice composition was 302 g/L of sugar, with a total acidity of 6.35 g/L of titrable acid and a pH value of 3.43. These values were measured by Fourier transform infrared spectroscopy with LyzaWine 5000^© (Anton Paar, Graz, Austria) analyser. In the case of early harvesting, we aimed for the berries to be fully ripe, healthy, and free of botrytis. In late harvest, we waited until the berries were overripe, with 60–80% botrytis infected, and showing signs of bunching. For bioprotection, M. pulcherrima strain was used (Oenoferm^®MProtect, (Erbslöh Geisenheim GmbH, Geisenheim, Germany)) at a 10 g/hL dosage; for fermentation, T. delbrueckii&S. cerevisiae yeast (Oenoferm^®Wild&Pure) at a 40 g/hL dosage, and liquid nutrient feed with DAP and thiamine (Vitamon^®Liquid, (Erbslöh Geisenheim GmbH, Geisenheim, Germany)) at a 30 mg/hL dosage.

The dosage of the M. pulcherrima product exceeded the manufacturer’s general recommendation of 2–7 g/hL. However, this higher concentration was intentionally chosen in the context of the experimental setting to ensure a more pronounced and measurable effect under the specific conditions of our study. This decision was also based on preliminary trials and literature reports suggesting that, under certain oenological conditions (e.g., must composition, microbial load), higher doses may be required to achieve the desired technological impact.

The yeast strains were provided in dried form and were rehydrated before inoculation according to the manufacturer’s instructions. The sulphur treatments carried out with potassium-bisulphate (K₂S₂O₅). E_M_S, E_M_S_T, L_M_S, L_M_S_T was used at a 5 g/hL quantity. The oenological tannin called Tannivin^®Grape (Erbslöh Geisenheim GmbH, Geisenheim, Germany) was applied; this is a condensed grape tannin variety produced from French grape with the aim to increase the reductive potential of wines. The quantity was 15 g/hL in the E_M_T, E_M_S_T, L_M_T, L_M_S_T sample. The control groups, both the early and late harvests, did not receive a M. pulcherrima bioprotection treatment at harvest time, but were pressed after berrying and then treated with 15 g/hl SO₂ and 40 g/hL Oenoferm^®Wild&Pure with 30 mg/L Vitamon^®Liquid. The optimum fermentation range for the yeast was 16–18 °C; this was controlled by thermometer and tempering by room-cooling, with water jacket-cooling for overheated tanks. The fermentation was followed by a LyzaWine 5000^© (Anton Paar) analyser every other day.

2.2. Microbiological Analysis

The enumeration and differentiation of yeast populations in the must samples were conducted at the SueConsulting microbiology and food hygiene laboratory in Bodrogkeresztúr, Hungary. Samples for microbiological analysis were taken at four time points during the fermentation process: on day 0 (at inoculation), and subsequently on days 4, 8, and 14.

Two different solid media were utilized for the cultivation of yeasts. The primary differentiation medium was WL Nutrient Agar, prepared in-house with the following composition per 1000 mL of distilled water: tryptone (5 g), yeast extract (4 g), glucose (50 g), KH₂PO₄ (0.55 g), KCl (0.425 g), CaCl₂ (0.125 g), MgSO₄ (0.125 g), FeCl₃ (0.0025 g), MnSO₄ (0.0025 g), bromocresol green (0.022 g), and agar (20 g). For complementary analysis, Liofilchem Contact Slide 2 (Ref. 525272-53527, Bentley Labor LLC. ((Edison, NJ, USA) was also used, which contains Rose Bengal CAF Agar on Slide 1 and PCA+TTC Agar on Slide 2.

For yeast enumeration, must samples were serially diluted and filtered through sterile 0.45 µm membrane filters. The filters were then placed onto the surface of the WL Nutrient Agar plates. All plates were incubated under aerobic conditions at 26 °C for 3 days.

Following incubation, yeast colonies were differentiated and counted based on their distinct morphological characteristics on WL Nutrient Agar. However, since all samples were applied via membrane filtration, it was not possible to assess the colour change around the colonies (typically used as an indicator of Saccharomyces activity) due to the membrane covering the medium surface. Consequently, colony differentiation was based only on colour, shape, and texture.

Colonies presumed to be M. pulcherrima were identified by their characteristic pink-to-red pigmentation resulting from pulcherrimin production. However, we acknowledge that not all M. pulcherrima strains exhibit strong pigmentation under all conditions, and colony colour alone may lead to under- or overestimation. Therefore, species identification based on morphology was intended only for monitoring general population trends rather than exact quantification.

While alcoholic fermentation was carried out using a commercial starter culture containing both T. delbrueckii and S. cerevisiae, the microbiological monitoring was designed primarily to assess the behaviour of M. pulcherrima and non-Saccharomyces yeasts in the early fermentation stages. As such, no targeted enumeration of T. delbrueckii or S. cerevisiae was performed. Their presence and dominance were inferred from the known dynamics of inoculated fermentations and the literature. This is noted as a methodological limitation, and we recommend that future studies incorporate species-specific molecular techniques (e.g., qPCR) for more precise tracking of starter strains.

In recognition of these limitations, we have added clarifications in Section 4 and recommend that future studies incorporate molecular techniques (e.g., PCR-based identification) to achieve more accurate and species-specific resolution in yeast population analyses.

2.3. Analytical Characteristics, GC-MS Measurements, Sensory Panel

The following parameters were determined by classical analytical methods [51] at the end of the fermentation. Alcohol content: OIV-MA-AS312-01A method; total acidity: OIV-MA-AS313-01 method; pH: OIV-MA-AS313-15 method; volatile acid: OIV-MA-AS313-02 method.

LYZA 5000 Wine FT-IR (Anton Paar GmbH, Graz, Austria) was used every two days: reducing sugar, glycerol, gluconic acid, lactic acid, malic acid, and tartaric acid.

Thermo Fisher Gallery discrete analyser (Thermo Fisher Scientific Inc., Waltham, MA, USA) [52] was used at the end of the fermentation. Calcium content: UV-VIS 984361 method; potassium content: UV-VIS 984307 method; total polyphenols: UV-VIS 984346 method.

The results are the average values of three parallel measurements. The analysis of the flavouring compounds was performed by GC-MS HS-SPME, i.e., gas chromatography–mass spectrometry, vapour space analysis with solid-phase microextraction.

The sampling procedure used is capable of trapping volatile and semi-volatile components in the air. For the extraction of volatile compounds, 65 µm PDMS/DVB fibre was used, which was conditioned according to the manufacturer’s instructions before measurements (65 µm PDMS/DVB: 250 °C for 0.5 h). This sampling fibre is the most suitable for the capture of volatile compounds [53]. Samples were stored at 23 °C. The SPME sampling time was 15 min; desorption in the gas chromatograph injector was carried out at 240 °C for 1 min. Samples were sampled several times. The values that were used are the average values of two parallel measurements.

Measurements were performed using a Shimadzu GCMS-QP2010 Ultra AOC-5000 Plus gas chromatograph (Shimadzu Corporation, Nakagyo-ku, Kyoto, Japan) (coupled with a sample feeder. Helium 6.0 (99.9999% purity) was used as a carrier gas. The helium flow rate was 0.95 mL/min. Separation was performed using a ZB-WAXplus capillary column (Phenomenex Inc., Torrance, CA, USA) (30.0 m × 250 µm × 0.25 µm) with polyethylene glycol as the composition. The so-called transfer line temperature connecting the gas chromatograph and the mass spectrometer was set at 240 °C (the same as the final temperature of the column). The m/z (mass per unit charge) range tested was between 50 and 400 m/z. Shimadzu GCMSsolution software (version 2.53) was used to control the parameters of the gas chromatography–mass spectrometry system, to search for components, to analyze the mass spectra, and to further evaluate the data and to perform a full comparison of the chromatograms. The gas chromatography procedure was used to identify compounds by spectral library matching. The m/z peaks specific to each component were integrated, thus minimizing the uncertainty due to coelution and peak retention time slip. The identification of the chromatographic peaks, i.e., the resulting components, was performed using the NIST Mass Spectral Search Program (NIST/EPA/NIH Mass Spectral Library, version 2.3, NIST 17) and Wiley FFNSC 2 Mass Spectral Library (Wiley, 2011 edition).

The detected aroma compounds have been grouped according to chemical structure, biological origin, and organoleptic characteristics. The NIST Chemistry WebBook (https://webbook.nist.gov/chemistry/, accessed on 4 December 2024) was used for the chemical structure grouping, the MetaCyc/BioCyc database (https://metacyc.org, accessed on 2 February 2025) was applied as a source for the biological origin, and the Good Scents Company Database C (https://www.thegoodscentscompany.com, accessed on 5 February 2025) was used for the sensory characteristics.

Wine sensory analysis was carried out by a panel composed of six panellists (non-professional ones, “trained assessors”) in the tasting room of University of Eger according to OIV standard, OIV-MA-AS312-01B (Sensory analysis of wine). The sensory booths were compliant with ISO standards, ISO 8589:2022 Sensory analysis—General guidance for the design of test rooms. International Organization for Standardization: Geneva, Switzerland, 2022. Samples of 30 mL were served to the judges in ISO 3591 tasting glasses (ISO 3591:1977—Sensory analysis—Apparatus—Wine-tasting glass. International Organization for Standardization: Geneva, Switzerland, 1977) at 10 °C, with de-carbonated water and neutral bread as flavour-neutralisers. Before the judging started, the judges attended a training session in which they were trained to measure each sensory attribute using standard samples and shown how to use the scale for each attribute. The samples were tasted blindly and randomly according to codes. The sensory evaluation was carried out on a five-point scale, where 1 indicated the absence of the sensation and 5 represented extreme intensity. The visual evaluation focused on colour (from 1—undesirable to 5—highly likeable). The olfactory evaluation included smell purity (from 1—strong off-odours to 5—clean) and aroma purity (from 1—strong off-notes to 5—clean). The taste evaluation considered spiciness (from 1—very little to 5—very rich), roundness (from 1—low to 5—high), and harmony (from 1—disharmonious to 5—elegant).The scores for each property were evaluated with the aid of a statistical programme.

2.4. Statistical Analyses

All statistical analyses were conducted using the R software environment (version 4.4.1) [54]. The specific R packages used for individual analyses are detailed and cited below. Permutational Multivariate Analysis of Variance (PERMANOVA) was performed to test for statistically significant differences in the overall chemical profiles among the experimental groups. The analysis was carried out using the adonis2 function from the R package vegan [55]. The Bray–Curtis dissimilarity index was calculated on relative abundance data for GC-MS volatile profiles if treated compositionally. Statistical significance was assessed using 999 permutations. Partial Least Squares Discriminant Analysis (PLS-DA), a supervised multivariate method, was employed to identify the specific chemical variables that were most influential in discriminating between the predefined treatment groups. PLS-DA models were constructed using the R package mixOmics (version 6.26.0) [56]. Prior Probabilistic Quotient Normalization (PQN) was performed to the dataset. Hierarchical Clustering Analysis (HCA) was applied to the chemical compounds identified by GC-MS to explore natural groupings based on similarities in their concentration profiles across the different wine samples. The clustering algorithm was performed using the hclust function, which is part of the stats package included in the base R distribution. The resulting dendrogram and heatmap were generated using the heatmap function from the R package heatmap [57]. The distance metric used was 1-cor and the linkage method was ward.D2.

Sensory evaluation scores obtained from the trained panel for each were averaged across the five panellists for each wine sample. These mean scores represent the consensus sensory profile for each wine. To visualize these multivariate sensory profiles, spider charts were generated using the radarchart function from the R package fmsb [58].

To investigate the relationship between the objective chemical measurements and the subjective sensory perceptions, a Mantel test was conducted based on Euclidean distances. The analysis was performed using the mantel function available in the vegan R package. The correlation between these two matrices was assessed using the Pearson correlation coefficient. The statistical significance of the observed correlation was determined through a 999 period permutation test.

3. Results

3.1. Yeast Population Dynamics During Fermentation

The evolution of the non-Saccharomyces, and specifically the M. pulcherrima, populations was monitored over a 14-day period. The results, presented in Figure 2, illustrate the distinct growth dynamics of these yeast groups in the early stages of fermentation.

The general non-Saccharomyces yeast population, initially present at approximately 1.9 × 10⁵ CFU/mL, showed a slight increase across all treatments, reaching a peak density of around 2.2 × 10⁵ CFU/mL by day 4 (Figure 2A). Following this initial growth, the population experienced a rapid decline, with viable counts dropping significantly by day 8 and becoming negligible by day 14.

The population dynamics of Metschnikowia showed a clear distinction between the control and the inoculated treatments (Figure 2B). In the treatments inoculated with M. pulcherrima (E_M, E_M_S, E_M_T, E_M_S_T), the yeast established itself rapidly, starting from an initial density of approximately 6.0 × 10⁵ CFU/mL and reaching a peak concentration of 8.0–9.0 × 10⁵ CFU/mL by day 4. In contrast, the control treatment (E_S), which was not inoculated with M. pulcherrima, exhibited a much smaller native population that peaked at only 2.0 × 10⁵ CFU/mL on day 4. In all treatments, the Metschnikowia population declined sharply after day 4, becoming undetectable by day 8 or 14.

In the late harvest treatments (L_S, L_M, L_M_S, L_M_T, L_M_S_T), a markedly different trend was observed. The initial non-Saccharomyces population was substantially higher (approximately 6.0–6.5 × 10⁵ CFU/mL), reaching peak densities of over 6.5 × 10⁵ CFU/mL by day 4, particularly in the L_S and L_M_T treatments (Figure 2A). Despite the higher starting populations, these treatments also showed a rapid decline in viable yeast counts, with almost no detectable cells by day 14.

Similarly, in the case of Metschnikowia (Figure 2B), the late harvest treatments mirrored the trends seen in the E-series, with even slightly higher initial and peak concentrations (approximately 9.0 × 10⁵ CFU/mL). This suggests that the late harvest treatment conditions were conducive to more robust initial growth, though the decline phase followed a similar timeline to that of the early harvest treatments, with populations dropping significantly after day 4.

In both experiments, the Metschnikowia populations increased during the first 4 days but declined sharply thereafter, becoming undetectable by day 8 (Panel A) or day 14 (Panel B). The more prolonged persistence in Panel B may reflect a slower accumulation of inhibitory factors or a less competitive microbial environment compared to Panel A. In both cases, the progressive increase in ethanol concentration likely exceeded the tolerance limits of Metschnikowia, which is generally more sensitive to ethanol stress than Saccharomyces species. Additionally, killer toxin production by the co-inoculated T. delbrueckii-S.cerevisiae strains may have contributed to the rapid loss of viability, as non-Saccharomyces yeasts are often susceptible to these compounds. Finally, nutrient depletion—particularly of assimilable nitrogen—could have further reduced Metschnikowia competitiveness, especially under the more rapid fermentation kinetics apparent in Panel A. The combined action of these factors likely explains the population dynamics observed in both panels.

3.2. Influence of Treatments on Wine Aroma Profiles

GC-MS analysis resulted in the detection of 85 aroma compounds. Using different databases, these compounds were grouped according to their chemical structure, organoleptic characteristics, and microbiological origin (Table S1), and statistical comparisons were made. The results of the PERMANOVA statistical analysis carried out are summarized in

Table 1. Statistical analysis of the effect of different treatments on different aroma compounds, grouped by chemical character, organoleptic property and biological origin—* p < 0.05, ** p < 0.01, *** p < 0.001.

Multivariate Model:	Nested	Sulphur	Tannin	M. pulcherrima	Harvest
All metabolites	0.6540 ***	0.1113 **	0.0509	0.2250 ***	0.4747 ***
Alcohols	0.7146 ***	0.1866 ***	0.0691	0.3180 ***	0.4456 ***
Esthers	0.6117 ***	0.1124 **	0.0575	0.2429 ***	0.4393 ***
Terpenes	0.7143 ***	0.0273	0.0362	0.0528	0.6499 ***
Vegetal	0.7649 ***	0.1022 *	0.0242	0.1663 **	0.6327 ***
Fruity	0.6552 ***	0.1245 **	0.0638	0.2630 ***	0.4568 ***
Tertiary/ageing-related	0.5025 ***	0.0530	0.0468	0.0907 *	0.4118 ***
Off odour	0.4590 ***	0.0328	0.0162	0.0839 *	0.3607 ***
Yeast (Saccharomyces spp.)	0.6850 ***	0.1111 **	0.0589	0.2327 ***	0.4762 ***
Botrytis	0.5106 ***	0.1035 *	0.0514	0.1672 **	0.4224 ***

, where the effect of each treatment (sulphurization, tannin dosage, bioprotection, harvest date) was analyzed according to chemical characteristics (total compounds; most abundant groups of compounds detected—alcohols, esters, terpenes); organoleptic characteristics—vegetal, fruity, tertiary/ageing-related (this category includes compounds associated with wine maturation) and defective odours); and biological origin—wine yeast (Saccharomyces spp. indicates isolates from the Saccharomyces genus without species-level identification, Botrytis). The nested component is the reference effect of the model structure, the pooled effect of the control group.

Nested values are high for all traits (chemical, sensory, biological), so the model explains these traits very well. It can be concluded that the harvest time is the strongest factor affecting all the characteristics. In particular, this is dominant for terpenes, grape aroma carriers and compounds produced by wine yeast. The treatment with M. pulcherrima has a strong effect on all the compounds formed, particularly alcohols and esters for the chemical characteristics, fruity character for the organoleptic parameters, and the compounds produced by Saccharomyces and non-Saccharomyces yeasts for the aroma origin, and also plays an influential role in the activity of Botrytis. No significant effect of the tannin treatment was found for any of the characteristics.

The effect of sulphurization is significant but lags behind the effect of harvest date and bioprotection with M. pulcherrima. The effect is selective, strongly influencing microorganisms, with significant effects on the formation of aroma compounds of all three biological origins, and a strong influence on alcohol and ester formation, as well as on the formation of fruity aromas. PLS-DA Biplot statistical modelling was performed to select the 30 most abundantly detectable aroma components, the results of which are shown in Figure 3.

Component 1 (42.9%) accounts for 42.9% of the variance in the data, and Component 2 (24.2%), together, explains 67.1% of the total variance, so the chart provides good coverage. The separation of the groups visually shows that different treatments result in different aroma profiles without much overlap. The early- and late-harvest samples are completely distinct, further supporting the role of harvest date in the composition of aroma components. Isobutyrate <ethyl>, Phenylacetate <ethyl>, Pentanoic acid <3-methyl>, Hexanoate <ethyl>, Octanoic acid, Linalool and Damascenone are the typical compounds found in the early-harvest samples, giving the wines a fresher, more floral, citrus character. For the late harvest groups, the dominant compounds are Nerol, Dodecanoate <ethyl>, Isobutyl alcohol, Butyrate <ethyl>, Decanoate <ethyl>, so they are richer in esters and higher aromatic alcohols, which give a sweeter, rounder, fruitier, more floral character. Looking at the effect of sulphurization, it is not always the same, because it depends heavily on the time of harvest; it purifies and stabilizes, but the aroma composition is determined by the time of harvest. Isobutyrate, phenylacetate, and hexanoic acid compounds are predominant in early harvesting, characteristic parameters produced by the wine yeast used, which impart fresh, floral, citrus notes to the wines. In the case of late harvests, Nerol, decanoate, dodecanoate, and isobutyl alcohol are the dominant compounds, sweet, fruity notes predominate, a more mature character becomes dominant.

Groups treated with M. pulcherrima show a higher variability of aromas. Typical compounds include Phenylacetate <ethyl>, Butyrate <ethyl>, Nerol, Isobutyl alcohol, and Dodeconate <ethyl>, which are the main characteristic components produced by M. pulcherrima, providing a richer, waxier, and fruitier character. Without bioprotection (E_S; L_S), the harvest date determines the aroma profile, which is not altered by sulphurization. Early harvest is characterized by the floral, herbal Isobutyrate, Hexanoic acid, and Pentanoic compounds, while late harvest is dominated by the fruity Dodecanoate, Butyrate, and Decanoate components.

In the tannin-treated samples, tannin does not dominate the aromatic profile, but it does structure and balance the character of the wine. In early harvests, tannin enhances the floral, fresh notes, while in late harvests it enhances the fruity aromas, especially when used with M. pulcherrima and sulphite. The sample with the richest aroma composition is the Late Harvest_M. pulcherrima+sulphurization+tannin sample (L_M_S_T).

3.3. Characterization and Clustering of Aroma Compounds

Hierarchical Clustering Analysis (HCA) was performed on the quantified aroma compounds to identify groups of compounds exhibiting similar concentration patterns across the different wine samples. The results are visualized as a heatmap in Figure 3, where rows represent the wine samples and columns represent individual compounds. The colour scale within the heatmap indicates the relative concentration of each compound (columns) in each sample (rows), after PQN standardization. The dendrogram associated with the columns illustrates the hierarchical relationships between compounds based on the similarity of their concentration profiles across the samples. Visual inspection of the heatmap (Figure 4), particularly the column dendrogram, reveals distinct clusters of compounds that group together based on their shared variance across the different wine samples. To understand the functional significance of these groupings, the identified compound clusters (from the column dendrogram) were examined in relation to the annotations displayed above the columns, which are described as follows. Chemical Types: by observing the predominant chemical classes (e.g., Esters, Alcohols, Terpenes, Other) within each compound cluster, we can infer the relationships between chemical structure and co-regulation or co-production. For instance, a cluster might be significantly enriched in fermentation-derived esters, suggesting coordinated yeast metabolism influenced by specific treatments visible in the corresponding rows. Biological Source: compound clusters are related to the annotated biological source (e.g., General Yeast, B. cinerea, Other/Grape) helps to attribute groups of co-varying compounds to specific microbial activities or grape metabolism. A cluster strongly associated with B. cinerea might contain compounds whose concentrations are similarly high or low, primarily in the late harvest (“L”) samples. Sensory Effect: compound clusters are correlated with their dominant sensory descriptors (e.g., Fruity, Vegetal, Tertiary/Ageing-Related, Off Odour), providing insights into how groups of covariant compounds collectively contribute to the perceived aroma profile. A cluster characterized by ‘Fruity’ notes might comprise various esters and terpenes whose combined presence is consistently elevated in certain sample groups.

Overall, the hierarchical clustering presented in Figure 4 effectively groups aroma compounds based on their similar behaviour across samples. Relating these compound clusters to their chemical, biological, and sensory annotations helps elucidate the underlying biochemical and microbial processes driving the differences observed in the wine aroma profiles due to harvest time and oenological treatments. The arrangement of the rows clearly demonstrates that harvest time is the primary factor driving the overall aroma profile differences. A distinct separation is observed between the early harvest (“E”) samples clustered in the lower half and the late harvest (“L”) samples clustered in the upper half. Within the early harvest group, the control samples (E_S), which received only standard SO₂ addition post-pressing, form a distinct subgroup. Compared to the other “E” treatments (E_M, E_M_S, E_M_T, E_M_S_T), these E_S samples exhibit notably higher relative concentrations for many compounds clustered on the right side of the heatmap, while showing varied patterns for compounds on the left. This highlights the unique profile of the control “E” wines compared to those undergoing bioprotection, tannin, or alternative SO₂ regimes. Sub-groupings within the late harvest (“L”) samples based on treatment are less visually pronounced from the row clustering alone. The column dendrogram reveals clusters of compounds that exhibit similar concentration patterns across the samples. A general trend is visible: compounds clustering towards the right side of the heatmap tend to have higher relative concentrations in early harvest (“E”) samples (especially E_S), while those clustering towards the left side generally show higher relative concentrations in late harvest (“L”) samples.

However, a key observation is the lack of a strong correlation between these compound clusters and their functional annotations (Chemical Type, Biological Source, Sensory Effect), as shown by the coloured bars above the columns. Most compound clusters identified by the dendrogram are heterogeneous, containing mixtures of different chemical types, attributed biological origins, and sensory effects. This indicates that the factors driving co-variance in this dataset are complex and cut across these broad classification categories. While harvest time clearly differentiates the overall profiles, the specific groups of co-varying compounds reflect the intricate interactions between grape maturity, Botrytis influence, and the metabolic responses induced by the different yeast and oenological treatments, which are not fully captured by simple alignment with the annotation categories alone.

3.4. Sensory Perception of Wine Treatments

The results of the organoleptic evaluation are illustrated in Figure 5 for early harvest (healthy raw material) and late harvest (botrytised raw material).

Differences between treatments can be observed for both harvest dates; these differences are not drastic, but are consistently observed. Early harvest provides a more vibrant colour and fresher flavour, while late harvest provides a more clean, harmonious character. In the control samples, the wine has a more subdued appearance, purity of aroma and flavour, roundness, and harmony at both harvest times. Combined treatments (in particular, Metschnikowia+Sulphite+Tannin) improve the organoleptic qualities, making the individual characteristics more pronounced and resulting in more balanced wines. The Metschnikowia treatment certainly has a positive impact on the organoleptic characteristics of the wines, and this was perceived by the judges. The positive effect of bioprotection is also noticeable in the colour, but the role of sulphurization is not strong. The results show that sulphur dioxide weakens the effect of non-Saccharomyces yeast, but when supplemented with tannin (especially in late harvests) the favourable organoleptic parameters were more pronounced and well perceived.

3.5. Correlation Between Chemical Composition and Sensory Perception

To quantitatively assess the relationship between the measured volatile chemical profiles and the perceived sensory attributes, a Mantel test was conducted. This test compared the distance matrix derived from the GC-MS aroma compound data (Euclidean distance on scaled data) with the distance matrix derived from the averaged sensory panel scores (Euclidean distance). The results of the Mantel test are summarized in

Table 2. Results of Mantel tests comparing the distance matrix derived from sensory panel data with distance matrices derived from GC-MS volatile compound data (overall and by functional category). Values represent the Pearson correlation coefficient (Mantel-r) and the p-value obtained from 999 permutations. ** p < 0.01.

Compound Category	Correlation	p-Value
All	0.2623	0.0080 **
Alcohol	0.3348	0.0020 **
Esther	0.2606	0.0040 **
Terpene	0.0484	0.3220
Vegetal	0.2924	0.0070 **
Fruity	0.3190	0.0020 **
Off Odour	0.0782	0.2460
Tertiary/Ageing-Related	0.0184	0.4220
Yeast (Saccharomyces spp.)	0.2856	0.0050 **
Botrytis	0.1098	0.1740

. A statistically significant positive correlation was found between the overall chemical composition (all 85 compounds) and the overall sensory perception (Mantel r = 0.2623, p = 0.008). This indicates that variations in the global volatile aroma profile are moderately associated with the differences perceived by the sensory panel across the wine samples. Further analysis examining specific compound categories revealed that certain groups contributed more strongly to this overall correlation (Table 2). Significant positive correlations were observed between sensory perception and the profiles of Alcohols (r = 0.3348, p = 0.002), Esters (r = 0.2606, p = 0.004), compounds associated with Vegetal notes (r = 0.2924, p = 0.007), compounds associated with Fruity notes (r = 0.3190, p = 0.002), and compounds attributed to Yeast (r = 0.2856, p = 0.005). Conversely, no significant correlation was found between the sensory profiles and the profiles of Terpenes, compounds associated with Off Odours or Tertiary/Ageing-Related notes, or those specifically attributed to Botrytis within this dataset.

3.6. Monitoring of Calcium, Potassium, and Total Polyphenol Content

During the analysis, the potassium content of the samples varied widely, ranging approximately between 700 mg/L and 1200 mg/L (

Table 3. Results of calcium (Ca), potassium (K), and total polyphenols (TP) content measurements.

ID	Ca [mg/L]	K [mg/L]	TP [g/L]
E_S_a	73.3	901.9	0.20
E_M_a	66.7	739.1	0.43
E_M_S_a	67.2	741.4	0.54
E_M_T_a	66.2	734.9	0.6
E_M_S_T_a	66.8	796.1	0.66
E_S_b	72.8	959.3	0.44
E_M_b	65.6	751.6	0.48
E_M_S_b	65.8	767.3	0.58
E_M_T_b	64.8	795.9	0.61
E_M_S_T_b	65.3	824.8	0.64
L_S_a	88.0	1209.8	0.52
L_M_a	89.0	1201.4	0.62
L_M_S_a	77.0	1099.2	0.61
L_M_T_a	77.0	1069.6	0.66
L_M_S_T_a	78.0	1066.3	0.69
L_S_b	76.0	1093.9	0.69
L_M_b	77.0	1040.9	0.81
L_M_S_b	77.0	1076.8	0.86
L_M_T_b	77.0	1086.6	0.94
L_M_S_T_b	78.0	1067.3	1.00

). The samples labelled “E” (early harvest) generally showed lower potassium levels, typically between 700 and 950 mg/L. In contrast, the “L” samples (late harvest) exhibited consistently higher potassium concentrations, generally ranging from 1000 to 1200 mg/L. The “L” group shows significantly higher potassium levels compared to the “E” group. Lower standard deviation in the “L” group suggests greater consistency across samples. Visual consistency within the “L” group implies a uniform effect of the antioxidant treatment on mineral retention.

In both cases, the samples without M. pulcherrima had the highest values. The only sulphur treatment may have contributed to maintaining or preserving higher potassium levels.

The “a” and “b” designations within sample identifiers showed only minor differences, indicating that the measurement variation between replicates of the same treatment is minimal.

According to the boxplot (Figure 6), wider variability and a lower median potassium level can be found in the “E” group, and a narrower range and higher median can be found in the “L” group.

Ca is higher in the “L” group, and exceeds 80 mg/L value in the samples not treated with bioprotection (L_S_a and L_S_b), which is a risky level for a later Ca precipitation.

The “E” group showed lower polyphenol levels, ranging from 0.19 to 0.65 g/L, the “L” samples, had consistently higher total polyphenol content, up to 1.00 g/L.

The higher total polyphenol content observed in wines made from botrytised grapes can be attributed to several biochemical and technological factors. B. cinerea produces enzymes (e.g., pectinases, cellulases) that degrade grape skin and cell wall structures, facilitating the release of polyphenolic compounds into the must, and the dehydration of the berries caused by noble rot leads to the concentration of solid constituents, including polyphenols. Furthermore, the oxidative environment created by Botrytis may promote polyphenol polymerization, resulting in stable but still detectable forms of these compounds. In the “E” and “L” groups, the application of M. pulcherrima, sulphur, and tannins may prevent the degradation of polyphenols, contributing further to their retention.

4. Discussion

4.1. The Population Dynamics of Yeast During the Initial 14 Days of Fermentation

The microbiological analysis confirmed the successful implantation and early proliferation of M. pulcherrima in the inoculated treatments, reaching over 8.0 × 10⁵ CFU/mL within the first four days. This rapid growth indicates favourable adaptation to the must environment and aligns with previous findings on early-phase non-Saccharomyces activity [41,42,43]. In parallel, the general non-Saccharomyces yeast population, which does not include the inoculated strain, remained relatively stable, increasing only slightly from an initial ~1.9 × 10⁵ CFU/mL to ~2.2 × 10⁵ CFU/mL by day 4 before it also began to decline. As observed in earlier studies, the decline in M. pulcherrima and other non-Saccharomyces yeasts after day 4 is likely due to increasing ethanol levels and the onset of dominance by S. cerevisiae. The higher initial population of M. pulcherrima in inoculated samples compared to controls (E_S) suggests that active inoculation is necessary to establish an effective early presence. This early establishment may contribute to a more controlled fermentation process, especially in musts from botrytized grapes.

The higher initial population of M. pulcherrima in inoculated samples compared to controls (E_S) suggests that active inoculation is necessary to establish an effective early presence. This early establishment may contribute to the more controlled fermentation process, especially in musts from botrytized grapes.

It should be noted, however, that the identification of M. pulcherrima was based on visual colony morphology on WL agar, specifically its typical pink-to-red pigmentation. While this method is commonly used for indicative differentiation, not all M. pulcherrima strains consistently produce visible pigmentation under all conditions. As such, the reported CFU/mL values may be affected by under- or overestimation. Furthermore, because colony morphology alone cannot fully confirm species identity, the data should be interpreted as trend-indicative rather than quantitatively definitive. Future studies should consider the inclusion of molecular methods (e.g., PCR-based identification) to improve accuracy in yeast population assessments.

Although alcoholic fermentation in this study was carried out using a commercial starter culture containing both T. delbrueckii and S. cerevisiae, these species were not specifically enumerated during the microbiological analysis. The focus of yeast monitoring was on early non-Saccharomyces dynamics, particularly the implantation of M. pulcherrima. The presence and activity of T. delbrueckii and S. cerevisiae were inferred from fermentation progression and precedent in the literature. We acknowledge this as a methodological limitation, and future studies should incorporate targeted, species-specific molecular techniques (e.g., qPCR) to enable precise monitoring of starter yeast populations throughout the entire fermentation process.

4.2. Treatments Significantly Modulate Wine Aroma and Sensory Profiles

Harvest time was the most influential factor affecting both chemical aroma profiles and sensory characteristics, as confirmed by PERMANOVA analysis and the clear separation of early- and late-harvest samples in the PLS-DA model. Early-harvest wines were associated with fresher and citrus-like aromas, while late-harvest wines showed higher concentrations of esters and higher alcohols, contributing to rounder and fruitier sensory profiles.

The application of M. pulcherrima also led to notable shifts in volatile compound levels, particularly in esters and higher alcohols (Figure 3), which corresponded with increased fruity attributes in sensory evaluation (Figure 5). These findings are consistent with previous studies [3,9,15] reporting enhanced aroma diversity with non-Saccharomyces yeast use. Panellists perceived the M. pulcherrima treatment as contributing to more complex and balanced aroma profiles, particularly in late-harvest wines.

Sulphur dioxide addition had a statistically significant, although secondary effect, primarily through microbial stabilization. Its impact on aroma composition and sensory expression was dependent on harvest time. Tannin addition alone showed no strong effect but appeared to support aromatic and structural balance when combined with M. pulcherrima and SO₂ in botrytized wines.

Together, these results confirm that both harvest timing and microbial management—particularly the use of M. pulcherrima—have a measurable impact on wine aroma and sensory expression, while tannin and sulphur contribute in context-dependent ways.

4.3. Interpretation of Chemical Profile Modifications in Context

The observed effects of harvest timing, bioprotection with M. pulcherrima, and sulphur dioxide addition are generally consistent with previous findings on aroma development in terpene-rich and botrytized cultivars [59,60,61,62,63,64,65,66,67,68,69]. In our study, early-harvest wines displayed greater concentrations of compounds linked to floral and citrus notes (e.g., linalool, damascenone), while late-harvest samples showed higher levels of esters and higher alcohols such as nerol and ethyl dodecanoate—patterns in line with known ripening and Botrytis effects on aroma precursors.

M. pulcherrima inoculation significantly impacted the volatile profile, supporting earlier research that highlights its role in modulating early microbial dynamics and enhancing ester and alcohol formation [18,42,70]. Although specific enzymatic activity was not directly measured here, the observed increase in compounds like nerol and hexyl acetate suggests enhanced release or production mechanisms, as also reported in the literature. In contrast, sulphur dioxide alone appeared to suppress this aromatic complexity, reflected in the clustering of L_S and E_S samples with reduced volatile diversity and increased off-flavour markers.

The combined use of M. pulcherrima, sulphur dioxide, and oenological tannins showed synergistic effects, particularly in late-harvest wines. These samples were associated with richer and more structured sensory profiles, likely due to enhanced ester and terpene levels. While tannins alone had limited effect, their integration into multifactorial treatments appears to support a stable aroma and structural balance. These observations are consistent with prior reports on the tannin-mediated modulation of oxidative pathways and aroma retention [46,71,72].

Taken together, the results underline the importance of harvest timing, microbial strategy, and antioxidant management in shaping wine aroma. The experimental design allowed for the isolated and combined effects of these variables to be assessed, highlighting the value of integrated oenological approaches.

4.4. Compound Co-Variation: Links Between Chemical, Biological, and Sensory Categories

The hierarchical clustering analysis (Figure 4) provides additional insights into how volatile compounds responded to the various experimental conditions. Consistent with earlier results, harvest time was the primary factor driving the clustering, with specific groups of compounds showing higher abundance in either early (“E”)- or late (“L”)-harvest samples. Compounds enriched in “E” samples tended to group together (right side of the heatmap), while those associated with “L” clustered on the opposite side.

A key observation is that these data-driven clusters did not align strongly with predefined functional categories such as chemical type, biological origin, or sensory effect. While PERMANOVA analysis (Table 1) showed statistically significant treatment effects across these categories, the clustering pattern revealed that compounds frequently co-varied across treatments despite belonging to different annotation classes. Similar inconsistencies between functional annotations and empirical groupings have been noted in other complex fermentation systems [2,9,73].

This outcome reflects the multifactorial nature of aroma formation in wine, where interacting factors—such as grape maturity, Botrytis presence, yeast metabolism, and treatment combinations—can influence a broad spectrum of metabolic pathways simultaneously. As shown in previous studies [23,74], microbial treatments like M. pulcherrima inoculation or sulphur addition alter fermentation dynamics in ways that affect diverse classes of aroma compounds, not just a specific subset.

The lack of tight clustering within functional categories contrasts with findings in simpler biological systems, where compounds within a given pathway tend to be co-regulated [75]. In winemaking, however, the aroma profile is shaped by the combined effects of grape biochemistry, microbial ecology, and processing interventions [76]. Thus, while functional annotations remain useful for interpretation, the actual clustering based on compound covariance provides a more direct understanding of how specific enological variables influence aroma development in practice.

These findings underscore the importance of multivariate and data-driven approaches when analyzing complex aroma systems, as they more accurately reflect the integrated response of the volatile profile to oenological decision-making.

4.5. Sensory Impacts and Relationship with Chemical Drivers

Sustainability considerations and the challenges posed by climate-related shifts in grape composition—such as higher sugar content and increased microbial pressure—have prompted interest in non-Saccharomyces yeasts as alternative tools in winemaking. In our study, sensory analysis confirmed that M. pulcherrima treatments produced wines with enhanced aromatic complexity, particularly in late harvest, botrytized samples. These wines were consistently described as more floral and fruity, with increased aromatic purity compared to controls. The positive effects of M. pulcherrima were also evident in early-harvest wines, where samples exhibited fresher profiles and improved varietal expression.

These findings are consistent with prior studies [48,77,78], which have reported the strain’s potential to enhance ester formation and inhibit spoilage organisms. In our work, the combination of M. pulcherrima, sulphur dioxide, and oenological tannins yielded the most balanced sensory profiles, indicating a synergistic effect. However, sulphur dioxide alone appeared to suppress aroma diversity, suggesting that its use should be carefully timed and dosed when bioprotective yeasts are applied.

In botrytized wines, the observed improvements in floral and citrus notes—often associated with thiols and terpenes such as 3MH, linalool, and nerol [79,80,81,82,83,84]—point to M. pulcherrima’s role in stabilizing or enhancing key aroma compounds, despite the oxidative and enzymatic stress typically caused by B. cinerea [26,73]. Higher levels of phenylethyl alcohol and its acetate were also detected, reinforcing the contribution of this yeast to aromatic finesse in late-harvest matrices.

From a practical perspective, the application of M. pulcherrima supports both sensory optimization and sustainability goals by reducing dependence on SO₂ and minimizing corrective interventions. These findings confirm that bioprotection using non-Saccharomyces yeasts can serve not only as a microbial control strategy, but also as a tool to tailor wine style under variable harvest conditions.

5. Conclusions

This study evaluated how harvest timing (healthy vs. botrytized Furmint grapes) and specific oenological treatment strategies—namely, bioprotection using M. pulcherrima, sulphur dioxide addition, and oenological tannin supplementation—affected the volatile composition and sensory characteristics of the resulting wines. The combined effect of these treatments significantly influenced both chemical and sensory outcomes, particularly in relation to fruity and floral aroma expression.

The application of M. pulcherrima as part of a bioprotection strategy contributed to shifts in volatile profiles, including increased levels of esters and higher alcohol contents that are typically associated with enhanced aromatic complexity. These effects were especially pronounced in late-harvest, botrytized wines. While this approach did not allow for mechanistic conclusions about microbial competition, it demonstrated clear sensory benefits, including elevated fruity and floral notes and improved aromatic freshness.

A strong correlation was observed between chemical markers—particularly esters, alcohols, and compounds with a fruity or vegetal character—and the wines’ sensory profiles. However, the partial mismatch between compound clusters and predefined aroma categories reflects the complexity of aroma formation, highlighting the multifactorial nature of wine sensory expression.

From a practical standpoint, the use of M. pulcherrima in conjunction with oenological tannins and reduced SO₂ represents a promising tool for shaping wine style while aligning with sustainability objectives. In the case of botrytized musts—often low in native tannins—the addition of exogenous tannins was especially valuable, offering oxidative protection and contributing to structural balance and aroma retention.

In summary, the effectiveness of each treatment protocol varied according to grape condition. For wines made from healthy grapes, bioprotection with M. pulcherrima enhanced aromatic expression without introducing sensory faults, supporting its use in fresh, fruit-forward Furmint styles. In botrytized wines, the combined use of M. pulcherrima, SO₂, and oenological tannins resulted in greater aromatic stability and structural integrity. These findings underscore the importance of tailoring vinification strategies to grape condition in order to optimize sensory outcomes in Furmint winemaking.