The Furmint grape must was equalized then cleaned with vacuum drum filter. The inoculation concentration was 5 × 10⁶ cells/ml of must. The fermentations were carried out in steel tanks filled with 100 L must in 2010 and 2011. In 2012 the postharvest and the grape must cleaning process was the same, but the fermentation carried out in 50 L glass carboys. Samples were taken by time course given in Figure 1 to observe the dynamics of the fermentation.

The alcohol and total sugar concentration was measured with a Bruker Alpha FTIR spectrometer (Bruker Optic GmbH, Germany) and the results were processed with the Bruker OPUS software.

Fisher’s test was applied to evaluate significance of differences between variants at p = 0.05 level. The average values of ethanol and sugar concentrations determined in samples taken by time course were used to construct two data matrices

vintage per vintage, while other characteristics of strains of three species (S. cerevisiae, S. uvarum and St. bacillaris) and their 1 + 1 + 1 mixtures fermenting Furmint juice were put into the separated one.

Data matrices comprising time dependent percentage values were subsequently analyzed by percent of ethanol versus log time regression applying second order polynomial functions to elucidate character of dynamic changes in ethanol production during the fermentation following models described by Sváb [3]. The kinetic parameters (lag phase, half time of alcohol production and end point) were correlated by linear regression. Box plots were used to demonstrate differences between performance of three species (S. cerevisiae, S. uvarum and St. bacillaris) and their 1 + 1 + 1 mixtures as well as the wild zymoflora during the vinification process.

Statistical functions of Microsoft Office Excel 2003 (Microsoft, Redmondton, USA) and Statistica 5 program (StatSoft, Tusla, USA) were used for multivariate analysis of data. Graphical presentations of the results of data analysis were edited uniformly in MS Office PowerPoint 2003.

The levels of ethanol (E) and total sugars (TS) could be determined with high accuracy in samples taken during the vinification (F_E = 0.26 < F_TS = 1.56 < F_0.1 = 1.71). The alterations in ethanol production between fermentors processed with consortium of wild strains (11.8 - 12.8 v/v%) did not surpassed those inoculated with mixed strains (11.6 - 12.9 v/v%) or with pure cultures (11.6 - 13.0 v/v%). In general, the effect of vintage (V) on final ethanol concentration was not prominent (F_E,V = 0.22, p = 0.136), thus the difference in 2010 and 2012 was only 0.14% and 0.12%. However, the inoculations in vintage 2011 augmented the ethanol concentrations in new wines at 1.33 ± 0.14 v/v% (Table 1). Similar trend was observed in sugar concentration of new wines, in vintages 2010 and 2011 the differences between wild and inoculated batches were low (max. 0.08 and 0.14 g/L, respectively), while the residual sugar content in vintage 2011 surpassed the wild one at 23.5 ± 3.9 g/L in inoculated fermentors (Table 1). This suggests that the vintage effect manifested itself in the early stages of fermentation by affecting the intensity of sugar consumption (F_TS,V = 20.2, p < 0.01), and not only the amount of sugar remaining was increased but also the amount of alcohol produced, as less sugar was used in other processes.

The dynamics of ethanol production was demonstrated on selected samples of vintage 2010 (Figure 1). The lag phase of ethanol production in spontaneously fermented grape juice was significantly longer (72 h) than in fermentors inoculated with either the strain [10-848] or the mixture of three species (54 and 46 hours, respectively), the time requested to surmount toxic concentration to yeasts [4] [5] was also longer (8 versus 4 - 5 days). The maximum rate of EtOH production of the pure culture of S. cerevisiae (0.92 mM/h) was lower than the mixtures (the spontaneous fermentation also has been carried out with various cohabiting yeasts). In general, contrary to the near uniform composition of new wines vintage per vintage, the dynamics (D) of ethanol production showed high, strain dependent alterations (F_S = 48.58) although the character of the process was similar strain by strain (F_S,D = 1.11), i.e., after strain dependent length of lag phase started a rapid evolution of ethanol concentration, that followed to over 9 % v/v of ethanol in the medium with decreasing intensity up to the ethanol concentration in new wine characteristic to the strain concerned. Although, the vintage effect per se was also prominent (F_S,V = 6.84) in this respect (Figure 2(A)), the dynamics of ethanol production observed in these experiments—carried out in vineries—was similar to that found in highly controlled laboratory conditions [1] [2].

The rate of sugar utilization during lag phase of ethanol production showed high and strain dependent alterations (Figure 3), Nevertheless, the half time of these two processes correlated well (Figure 2(B)). The fermentors separated into two groups by rapidity of the start of ethanol production. The vintage effect in this respect seemingly was insignificant, although the spontaneously fermented batches ranked into group of slow fermenting ones in the first half of vinification (group U). The cause of this separation is unclear, as the starting composition of grape juices in subsequent vintages was different (205, 216 and 222 g/L total sugars, respectively), still groups L and U have heterogeneous components. One may conclude that the composition of must per se is not necessarily connected

to the dynamics of vinification. Examining the process in more details, strain dependent alterations were revealed in the intensity of sugar utilization during lag phase of ethanol production (LS), which can not be related to vintage effect (Figure 3). The time requested to reach the characteristic level of final ethanol concentration showed significant strain dependent differences (c.v. 15% ® 51%) that were influenced by vintage effect as well (Figure 3), having been the vinification process more slow in 2010 (11 - 17 days) then either in 2011 or 2012 variants (6 - 10 and 6 - 16 days, respectively). The maximum rates of both sugar utilization (MIS) and ethanol production (MIE) varied within large limits (3.59 ® 15 and 0.64 ® 2.59 mM/h, respectively) but simultaneously (Figure 4), thus their ratios (MIS/MIE) varied in narrow limits (1:1.82 ® 1:1.89 mM/h; c.v. 2.57%).

The vinification process per se run off by strain dependent manner under strong vintage effect (Figure 3) that was reflected in differences of maximum rate of ethanol production as well as in the maximum rate of sugar utilization as

it was demonstrated using normalized values of these two parameters (Figure 4(A)). The great variations caused by vintage effect and differences in species dependent responses were demonstrated on Figure 5. The wild consortium and Stramerella strains proved to be less sensitive to vintage effect than the pure cultures of Saccharomyces strains, and the mixtures of three species. This was reflected both in dynamics of ethanol production (Figure 5(A)) and in the rates of sugar utilization which varied in significantly higher extent than the ethanol production (Figure 5(B)).

The dynamics of ethanol production observed in these experiments carried out in vineries was similar to that found in highly controlled laboratory conditions [1] with large set of yeast strains [6]. Studying various mathematical models for description of the dynamics of ethanol production we found the second-order polynomial function being most suitable to predict the result of vinification [2]. This model, proposed by several authors [7] [8] as well, permits to weigh the role of constant, primary (linear), and secondary (quadratic) effects as well as to analyze their relationships in strain-dependent manner (Figure 1 and Table 3). The coefficients of polynomial function describing the dynamics of ethanol formation during vinification process can be conceptualized as vectors, i.e., sums of various factors influencing the ethanol-producing capacity of yeast cells in the vinification process connecting them to both extracellular and intracellular factors regulating the performance of proper strains [1]. Thus, the actual ethanol

concentration (Y) is a product of working cell factories and might be extrapolated applying polynomial function Y i = A + [ b 1 × X i ] + [ b 2 × X i 2 ] , where [X] is the actual time (i) counted of the start of fermentation, while the [A] is a time-independent constant, which might be related to a group of properties of yeast strains that take part in ethanol production in a time-independent manners as well as not related to responses of cells to the changing environment in the fermentation tank. The influence of both [b₁] and [b₂] manifests in time-dependent mode, and can be considered to be vectors of primary and secondary factor groups, respectively, and these factors most probably take part in the regulation of the responses of cells to changes in environmental conditions. This concept was discussed in details on the example of microvinification experiments [2], where strict trends (p < 0.01) were elucidated in the manifestation of the simultaneous regulatory effects of these groups of factors. The [b₁] group of factors counteracted to both constant [A] and [b₂] groups. Moreover, the influence of factors [b₁] and [b₂] counteracting synchronously in time dependent manner was about two times stronger than the constant ones, meanwhile, the strength of [b₁] group surpasses that of the [b₂] one about five times. The yeast strains fitted precisely to trend lines (p < 0.01) independently on their taxonomic position or other known properties [1]. Similar relationships could be elucidated analyzing results of the experiments carried out in vineries (V) throughout the subsequent vintages where the conditions were not controlled as strictly as in laboratory (L). Plotting time dependent coefficients (b₁ and b₂) versus time independent one (A) the fermentors fitted strictly to regression line (Figure 1). Comparing the two sets of data, i.e. the coefficients of the polynomial functions obtained of the analysis of results of laboratory and cellar fermentations, a high degree of compliance was demonstrated (t_V,L < 2) between the slopes of regression lines (Table 2). However, the intercepts of regression lines are significantly of different size. On our view the above means, that dissimilarities between

^a= Coefficients of the polynomial functions describing the dynamics of ethanol productind were imported of Kállai et al. 2019 [2]. The two sets overlapped partially, as 17 strains involved into laboratory experiments were old isolates [1] [2] [9]. CA_1-n = Con­stant; PL_1-n = Primary (linear, b₁); SSQ_1-n = secondary (squared, b₂) coefficients of the polynomial functions (Y = A + b₁X + b₂X², where Y is the actual concentration of substances measured in juice at X hours after inoculation) describing dynamics of ethanol production in fermentors in laboratory or vineries, respectively.

two sets relate mainly to quantitative (time independent) characteristics of the fermentation, while similarities between sets can be related to qualitative characteristic, which are connected to those strain (inoculant) dependent factors that influence the dynamics of fermentation in time dependent manner. Most probably, the interaction of the latter two groups of factors in regulation of the fermentation was more tolerant to changes in circumstances than those affecting quantitatively the process.

The weight of these suspected factor groups changed during the vinification process (Table 3). The constant and secondary factors (CA and SSQ [b₂]) counteracted primary one (PL [b₁]) during the whole process, although the power of strain dependent differences only in first half of the fermentation was prominent (Chi-sqr_exp. > Chi-sqr_0.05 = 28.8). Seemingly, the importance of metabolic processes not related to ethanol production became sidelined, and the significance of strain-dependent differences has been reduced in medium with ethanol concentration over 5 - 7 %v/v. One can conclude that; the negative correlation between maximum intensity of ethanol production and linear component of polynomial function (PL, b₁) indicates that some of the strain dependent factors counteract with ethanol production in time dependent manner during mineralization of organic matters in grape juice. The lack of connection between the content of aroma compounds in new wines and the coefficients of polynomial functions describing the ethanol production suggests also that the strain dependent factor groups regulating the dynamics of ethanol production have low

^a= See Table 1 and Figure 1. ^b= Coefficients (C, PL, SSQ) of the functions (D_1-19 = {[CA]_1-19 + [PL]_1-19 + [SSQ]_1-19}), where D_1-19 = dependent variable; CA_1-19 = Con­stant; PL_1-19 = Primary (linear); SSQ_1-19 = secondary (squared) coefficients of the polynomial functions (Y = A + b₁X + b₂X^2, where Y is the actual concentration of substances measured in juice at X hours after inoculation) describing dynamics of ethanol production in the 19 fermentors, respectively. ^c= parameters of the multiple linear regression function: D = f(X₁, X₂, X₃), where D is a dependent variable of the first column, and X₁, X₂, X₃ are the coefficients as given above; ^d= importance of strain dependent properties influencing the dynamics of ethanol production; Chi^2_0.05 = 7.81, Chi^2_0.1 = 6.2, R^2_0.05 = 0.7714; ^e= intervals of vinification process (see Figure 1 and Figure 3); ^f= strain dependent lag phase; ^g= strain dependent time requested to reach the 50% of the final EtOH concentration produced by proper strains as measured of the start of fermentation; ^h= time requested to reach the final EtOH concentration produced by proper strain; ⁱ= maximum rate of alcohol production; ^j= maximum rate of sugar utilization; ^k= portion of utlized sugars (g) during the given ntervals of vinification process; ^l= composition of new wines fermented (see Table 1).

power in determination of bouquet and flavor of new wines.

The Tokaj rural viticulture has a centuries-old tradition and is one of the most important wine-growing areas in Hungary. The extreme and unpredictable weather associated with climate change in the last decades has significantly affected the date of harvest compared to the usual times and made it difficult to predict it accurately. These anomalies are new challenges that request appropriate developments in vine cultivation and wine making. Moreover, improvements are also needed to meet the changing market requirements.

The vinification of grape juice is conducted by yeasts of the genus Saccharomyces, most frequently by S. cerevisiae. This species which genome has been sequenced is also a premier research model system because of its genetic tractability, and an extensive array of molecular technologies having been developed for the genetic manipulation of it [9]. In recent decades, it has become accepted to replace or supplement spontaneous fermentation with traditional autochthonous yeasts with commercially available strains with selected traits to correct processing errors and/or improve the quality of the final product [9] [10].

In order to meet increased demand for meet wines with a more complex taste and aroma the research of starter cultures has focused on non-Saccharomyces species as well [10]. Nevertheless, in spite of the availability of several molecular tools, genetically engineered yeast strains are not yet in use in the commercial wine industry. Especially more data requested on their interactions when applied via co- and sequential inoculation, as we have little knowledge of how the different species and their mixture affect the process of fermentation, its dynamics. Therefore, we consider it is necessary to thoroughly analyze the dynamics of fermentation, because knowing this will allow the process to be controlled and thus also facilitate the planned scheduling of grape processing and winemaking. If we know the analytical parameters of our raw material and we know the fermentation ability of the starter culture we want to apply well, we can predict the duration of fermentation as accurately as possible. Furthermore, we can both the optimal time and the mode of intervention to produce the wine with the desired qualitative indicators.

The production of wine from grapes is unique in many respects, and it is one of the world’s oldest biotechnological processes. This process is traditionally not conducted under sterile conditions, and the resident winery microbiota contributes to wine quality and style characteristics of local wines. This resident microbiota which seems to be stable as was shown for Tokaj region and in cellar reconstructs for centuries [11] impacts wine quality and regionality. Although, any modified organism has the potential to become an enduring resident of the winery flora, further in-depth studies are needed to find out exactly how much of the desired properties are ingested during fermentation.

The correlations revealed in our comparative study of the coefficients of the equations describing the dynamics of alcohol production may shed light on the role of differences in the traits of strains. The role of these strain dependent differences is seemingly significant after inoculation, but decreases as vinification progresses, so it is important to determine what biochemical processes these three mathematically distinct groups of factors describe. The functions of several yeast genes have already been elucidated [12]. With these results, we can equip the yeasts with advantageous and valuable properties for industrial use.

The molecular diagnostics opened new possibilities to analyze the autochthonous yeast consortia fermeting the grape juice, and newer fast and inexpensive test methods will hopefully allow their widespread use among little growers in the near future. Thus, the rapidly evolving methods of “synthetic biology” will also be widely applicable.

Further experiments are needed to develop a mathematical model that better describes the vinification process [13] [14] in parallel with the development of continuous monitoring tools, which allows to predict the dynamics of fermentation more and more accurately, calculated with the effect of more sophisticated winemaking methods, including the different inoculation methods, the interaction of different yeast species and their mixtures, the supply of nutrients, and the regulation of fermentation cycles at different temperatures affecting the whole fermentation, including the time of its duration.