Introduction

ajor expected effects of climate change are an increase in temperature and changes in rainfall patterns and incoming radiation. As a consequence, vine phenology and grape composition at harvest should be dramatically modified. The main consequences on grapes are more sugar and less organic acids aromas and phenolic compounds of wines will also be affected.

Excessive alcohol in wines exerts a number of negative effects that raise microbiological, technical, sensory and economic challenges. First, because ethanol is a chemical stress factor for yeast that is often the underlying cause of sluggish or stuck fermentation. xcessive ethanol levels also impair the sensory quality of the wine by increasing the perception of hotness and by altering the perception of wine aroma complexity (Goldner et al., 2009). Today, wines with moderate ethanol levels are often preferred, in accordance with health prevention policies.

Consequently, reducing alcohol levels in wine has inspired considerable efforts in the past decade and remains a major challenge for the coming years. Different approaches to reduce alcohol levels in wines have been proposed at all stages of the winemaking process.

Some viticultur strategies are promising. Thanks to classical breeding research programs, different new grape varieties were created by successively crossing Muscadinia rotndifolia and 4 different Vitis vinifera cultivars. Such varieties reach full maturity with lower sugar content (i.e. 180 g/l) (Aguera et al., 2010). Two of these new varieties now tested in the experimental vineyard of Inra Pech-Rouge at a 1ha scale. Another of these new grape varieties is their natural resistance against powdery and mildew, thus preserving environment from pesticide use.

In a shorter-term perspective, several physical techniques for dealcoholization have been developed (Schmidtke et al., 2012). Another prospect is the development of yeasts with reduced alcohol yield, which is both very promising and challenging (Tilloy et al., 2015).

High pHs are more and more common, in particular in regions located at low latitudes and acidification may become inevitable for the process (effectiveness of SO2, avoidance of oxidation, etc) and with regard to organoleptic properties. Addition of tartaric acid is often used with inconsistent results. Membrane-based technologies have been developed to decrease the pH of wines by removing potassium content (Lutin et al., 2010) and more generally to control wine pH. Besides these new technologies, adapting winemaking practices may be an effective way to counteract the effects of changes in grape composition. This is particularly the case for fermentation management and oxidation control.

Microbiological strategies to reduce ethanol content of wines

Reducing ethanol yield of Saccharomyces cerevisiae

The yeast S. cerevisiae has been the subject of intensive research for metabolic engineering. Strong fundamental knowledge of its genetics, physiology, systems biology and genomics has facilitated its use as a metabolic engineering platform (Nevoigt, 2008; Borodina and Nielsen, 2014). These approaches have been widely used to optimize various traits of interest in the field of food and fermented beverages. In the recent years reducing ethanol yield has been one of the major targets.

The of conversion of sugars to ethanol presents only minor variations between strains of S. cerevisiae. Hence, to decrease the ethanol production in this yeast, it is necessary to reroute the flux of carbons to other pathways and to the production of other secondary metabolites. However, there are several constraints: it is essential to maintain the redox balance, to avoid the production of compounds that could affect the organoleptic quality of wines and to preserve yeast performance.

To achieve this goal, numerous metabolic engineering strategies have been implemented, such as the expression of a NADH oxidase a lactate dehydrogenase overexpression of GPD1 (reviewed in Tilloy et al., 2015). Among those various strategies, rerouting carbons towards glycerol has emerged as the best option to reduce ethanol yield (Michnick et al., 1997; Remize et al., 1999; Cambon et al., 2006; Varela et al., 2012).

In S. cerevisiae, glycerol plays major roles in redox homeostasis and osmotic stress resistance as it is the main compatible solute in yeast (Blomberg and Adler, 1989). Glycerol is usually found in wines at concentrations ranging from 5 to 9 g/l.

As a proof of concept, wine yeast strains overproducing glycerol and 2,3-butanediol, a polyol with no sensor impact n wines, with a lower ethanol yield and without accumulation of unwanted byproducts have been successfully constructed by metabolic engineering strategies (Ehsani et al., 2009). These strains have the potential to decrease alcohol levels in wines by up to 3% (vol/vol). In recent years, due to the poor public acceptance of GMstrategies, alternative strategies such as evolutionary engineering have been favored.

Evolutionary engineering has proven to be successful to reshape yeast metabolism (Figure 1). The concept of adaptive evolution is that microorganisms tend to evolve their intrinsic characteristics to adapt to new conditions. During this process of evolution, random genetic mutations occur, and if a selection pressure is applied, strains having one or several beneficial mutations in the selective medium will dominate in the culture medium and can thus be selected. This approach is based on the extended cultivation of a strain in controlled selective medium to select for natural genetic variants having beneficial mutations under the conditions used. Since the emergence of mutations is a rare event, several hundred generations are usually necessary before observing an evolution, which can last several months.

Adaptive evolution of a commercial wine yeast strain (Lalvin EC1118®) has been carried out to divert yeast metabolism towards

Figure 1. Evolutionary strategy by serial batch cultivation.

Prolonged cultivation is used to enable selection for successive mutations that confer a selective advantage (as higher biomass, growth rate or survival). Evolved phenotypes are analyzed by genome sequencing and a combination of omics to identify the underlying mutations. The mutation can be reconstructed in the ancestral background to validate their impact on the evolved phenotype (Tilloy et al., 2015).

increased glycerol and lower ethanol roduction using hyperosmotic stress as selective condition. Serial transfers were performed during 300 or 450 generations using either sorbitol or KCl as osmostress or salt stress agent.

The approach with KCl stress succeeded generat strains with redirection of carbon flux towards glycerol. After 200 generations under KCl stress, a first adaptation was detected and the observed increase in glycerol production was maintained after 250 and 300 generations.

The evolved strains showed a gain fitness conferred by a better viability under salt stress and carbon starvation conditions, which was correlated to a greater glycerol production.

Detailed characterization of the KCl-evolved mutants during wine fermentation showed that the evolved strains have substantial changes in central carbon metabolism. Carbons are rerouted towards glycerol, succinate and 2-3-butanediol at the expense of ethanol and without the accumulation of undesirable compounds such as acetaldehyde, acetate and acetoin.

The concentration of the most abundant byproducts after 30 days of fermentation was determined (fermentation was conducted on a synthetic medium as described by Bely et al., ). Carbon and redox balances were close to 100% for all strains. All evolved strains produced glycerol at concentrations 48 to 67% higher than that produced by Lalvin EC1118®, and the ethanol content in the synthetic wines was reduced by 0.45 to 0.80% (vol/vol). The evolved strains also produced larger amounts of succinate and 2,3-butanediol. In contrast, unlike previously described with engineered strains, no significant changes in the production of acetate and acetoin by the evolved strains were found, and the production of acetaldehyde remained in the range of the level found in wines. The evolved strains exhibited an overall decrease fermentation performance in comparison to the ancestral strain but were nevertheless able to complete the fermentation. The final cell populations were the same between the ancestral strain and these two evolved strains.

To understand the metabolic changes underlying these phenotypic differences, the transcriptome and the endo-metabolome of the evolved mutants and of the ancestral strain were compared. The mechanisms underlying the overproduction of glycerol in the evolved mutants remain intriguing. Indeed yeast facing severe KCl stress could have evolved the structural or regulatory genes that are related to glycerol metabolism. However, no significant changes in the transcription levels of genes involved in glycerol metabolism or in the HOG pathway and its target genes that would explain the increase in glycerol production were found. This suggests that glycerol overproduction is not mediated via increased protein synthesis of the glycerol pathway enzymes. Mutations in the evolved strains affecting post-transcriptional mechanisms could also modulate the reaction rates of glycerol production. The increase glycerol production could also be due to a global metabolic effect resulting from metabolic flux changes in central carbon metabolism.

The exact nature of the factors that caused the metabolic rearrangements in the evolved strains during the evolution experiment is difficult to identify. Whole genome sequencing revealed many changes between the ancestral and the evolved strains. Among these, a loss of heterozygosity (LOH) was observed in the evolved strains, particularly those selected in the late phase of the evolutionary process. Furthermore, in addition to the study of SNPs, QTL approaches were implemented to identify the molecular basis of this new phenotype. Bulk segregant analysis (BSA) was designed confronting a pool of segregants from an evolved strain to a pool of segregants derived from the parental strain. Candidate genes were identified and are currently under functional validation to confirm their implication in the phenotype.

To further increase the ability of the evolved strain to produce glycerol and to decrease its ability to produce ethanol during alcoholic fermentation, this strain was further sub to conventional breeding. To this end, about 150 haploid yeast spores from the evolved strain K300.1(b) were generated and haploid strains of opposite mating type having the highest capacity to produce glycerol were selected. Mating between two spores producing 20.9 and 16.2 g/l glycerol during fermentation on MS medium generated a first generation hybrid. After sporulation and spore mating, a second generation hybrid, H2, which produced 16.8 g/l glycerol, was obtained.

To validate the results obtained at laboratory scale, we compared the behaviors and metabolic properties of K300.1(b), H2 and Lalvin EC1118® during pilot-scale fermentation using a Syrah grape must, at 28°C. To avoid stuck fermentation, oxygen and nitrogen were added during fermentation. H2 had a fermentation rate slightly lower than those of Lalvin EC1118® and K300.1(b), but all strains were able to complete the fermentation, despite the high sugar concentration (255 g/l). K300.1(b) and H2 produced more glycerol (14.1 g/l and 17.9 g/l versus 10.8 g/l) and slightly more succinate than Lalvin EC1118®. The ethanol contents of the wines produced by K300.1(b) and H2 were reduced by 0.6% (vol/vol) and 1.3% (vol/vol). The production of acetic acid by the evolved and hybrid strains was greatly reduced compared to that of Lalvin EC1118®.

The selected final strain was assessed at pilot scale on different grape musts during the 2013 and 2014 harvests. The wines obtained contained between 0.6 and 1% vol/vol less alcohol and very little volatile acidity. By contrast, their total acidity was consistently higher, which has a strong interest since the increase wine alcohol content is most often associated with high pH (Tilloy et al., 2015).

In summary, the results obtained in grape must at pilot scale confirm the metabolic shift of the evolved strain and show a greater metabolic reprograming of the hybrid derived from the evolved strain.

This study demonstrates that a combination of adaptive evolution and breeding strategies is a valuable alternative to rational engineering for the generation of non-, low ethanol-producing yeasts.

Reducing ethanol yield by using hybrids and nonSaccharomyces

Interestingly, other species belonging to the Saccharomyces group (S. uvarum and S. kudriavzevii) differ in the metabolism for glycerol production and transport, NADH balance and acetic acid production (Masneuf-Pomarède et al., 2010; López-Malo et al., 2013; Pérez-Torrado et al., 2016). These species can be used as an original genetic material to get non- strains showing a partial reduction ethanol content. However their lower ethanol tolerance was clearly a discriminative trait, distinguishing these species from S. cerevisiae. As industrial winemaking may reach final ethanol content higher than 15° TAV the low ethanol-producing strains must also have a strong ethanol tolerance. The benefits of hybridization have been described in a large range of plant and animal species in an agronomical context. One of the most striking consequence of hybridization the phenotypic superiority of hybrids over two progenitors (i.e. heterosis) (Lippman and Zamir, 2007). Hybridization may also confer an improved phenotypic stability over environmental change (i.e. homeostasis) (da Silva et al., 2015).

Hybrids

In the Saccharomyces sensu stricto clade, hybrids can be easily obtained both at the intra or interspecific level allowing yeast improvement. Recently we obtained a collection of S. uvarum and S. cerevisiae parental strains and all their 55 possible hybrids. The parental strains came from different beverage industries (wine, distillery and cider) as well as from nature isolates.

Fermentations were carried out at 2 temperatures (18°C and 26°C) in Sauvignon banc grape must containing 188 g/l sugar. The phenotypic distribution of 35 traits, i.e. fermentation kinetics, yeast population, aromatic profiles and fermentation products, w measured for the entire data set in triplicates (da Silva et al., 2015). The ethanol/sugar yield (g ethanol/g of sugar consumed) was computed for the strains that achieved alcoholic fermentation leaving less than 1.6 g/l of residual sugar. Within strains of the same group, this parameter showed very low variability (CV <1.2%) underlining the strong robustness of this trait. Nevertheless, at 18°C we observed a significant reduction of ethanol production for the S. uvarum group with an average of 0.30% (vol/vol) reduction compared to the S. cerevisiae group (parental and intraspecific hybrid strains). The interspecific group showed an intermediate ethanol/sugar yield with some individuals producing 0.30% (vol/vol) ethanol than S. cerevisiae group mean. Interestingly, these interspecific hybrids have higher fermentation performances than the S. uvarum group due to the positive contribution of S. cerevisiae genome. Within those hybrids, we selected one background of particular interest EU23.

This hybrid was further tested in synthetic medium with higher sugar content (230 g/l) and compared to 8 strains of commercialized starters. At 18°C, EU23 showed a 0.34% (vol/vol) ethanol decrease related to glycerol overproduction (9.0 g/l compared to 6.5 g/l for commercial strain group). As previously observed in Sauvignon blanc, the temperature slightly impacts the ethanol/sugar yield and only a difference of 0.15% (vol/vol) ethanol was observed in the same media at 28°C. In different red grape musts (Merlot and Cabernet-Sauvignon) this level of discrepancy (ranging between 0.15 and 0.32) was also observed at laboratory and small pilot scale (30 l).


Sequential cultures C. zemplinina/S. cerevisiae

Pure culture

Strain 401

Seq 401

Strain 629

Strain 261

Strain 153

Strain 278

S. cerevisiae

24h

48 h

48h

48h

48h

48h

FX10

Ethanol (%)

13.16 ± 0.07*

13.52 ± 0.18*

13.46 ± 0.05*

13.41 ± 0.08*

13.51 ± 0.04*

13.01 ±0.04*

13.91 ± 0.00

Res. sugar (g/l)

0.87 ± 0.14

0.90 ± 0.20

0.90 ± 0.10

1.33 ± 0.75

0.83 ± 0.06

0.95 ± 0.07

0.95 ± 0.07

Yield (g/g)

0.43 ± 0.00*

0.45 ± 0.00*

0.44 ± 0.00*

0.44 ± 0.00*

0.45 ± 0.00*

0.43 ± 0.00*

0.46 ± 0.00

Vol. acidity (g/l)

0.45 ± 0.01*

0.84 ± 0.15*

0.83 ± 0.07*

0.76 ±0.04*

0.81 ± 0.07*

1.01 ± 0.06*

0.31 ±0.04

Glycerol (g/l)

13.03 ± 0.87*

14.72 ± 0.80*

15.36 ± 0.95*

15.76 ± 0.62*

15.21 ± 0.46*

15.76 ±0.01*

7.30 ± 0.48

Table 1. Sequential cultures C. zemplinina/S.cerevisiae.

Musts were inoculated with 107 viable cells/ml, followed by S.cerevisiae (2.106 viable cells/ml) after 24 or 48 hours of fermentation. Means of triplicate fermentations ± SD. *: Significantly different from pure S. cerevisiae culture.

These results confirmed the interest of interspecific hybrids for lowering alcohol production. However the reduction level obtained is not yet sufficient to guaranty a relevant result for an industrial application. Adaptive evolution experiments are now carried out by applying different osmotic stresses to this interspecific hybrid.

NonSaccharomyces

Several non-Saccharomyces yeast, such as Torulaspora delbrueckii, Schizosaccharomyces pombe Kluyveromyces spp., Issatchenkia spp., Zygosaccharomyces baili, etc, may also contribute to wine fermentation.

One species, Candida zemplinina, was evaluated and compared to S. cerevisiae. Fortyeight C.zemplinina isolates mainly from Bordeaux must and also from Hungary and Italy were used. Three S.cerevisiae were also tested for their fermentation performance. Fermentations were carried out in pasteurized Merlot grape must containing 240 g/l sugar at 4°C.

In pure cultures, C.zemplinina yeast resulted in stuck fermentations, confirming the low fermentation capacity of this species reported in the literature. C.zemplinina group possessed a low alcohol/sugar yield, 12% less compared to S.cerevisiae strains, which can be partially explained by the overproduction of glycerol.

We investigated the possibility of using C.zemplinina, as partner of S. cerevisiae, in mixed fermentations by inoculating 107 viable cells/ml C. zemplinina with 2.106 viable cells/ml S. cerevisiae. In addition, sequential fermentations were inoculated with C. zemplinina (107 viable cells/ml), followed by S.cerevisiae (2.106 viable cells/ml) after 24 or 48 hours of fermentation.

All mixed cultures were able to complete fermentation. In our conditions, no significant differences in ethanol production were observed between S.cerevisiae FX10 Zymaflore culture and co-cultures (simultaneous inoculations) (data no shown). In contrast, sequential fermentations reductions of ethanol production from 0.39 to 0.90% (vol/vol) (Table 1). Note that for the sequential culture with C.zemplinina 401 a higher reduction is observed when S.cerevisiae s added earlier during fermentation. The best sequential multistarter was C.zemplinina 401 with S.cerevisiae added after 24 hours of fermentation.

, a sensory evaluation was performed. Mixed culture wines were judged to be significantly different from those made by single S.cerevisiae. The latter scored better, while those made by the mixed cultures showed high sulfur off-flavor.

Further investigations are required to clarify the molecular mechanism underlying the reduction of ethanol production in C.zemplinina species. Its use in sequential culture appears promising regarding low-ethanol production, but an important effort has to be done to propose C.zemplinina strains with neutral impact on the organoleptic perception of wine.

Technological strategies

Microbial strategies still additional studies to be fully effective when several new techniques and winemaking practices are – or will soon be – available to help the winemakers to correct the physico-chemical and sensory balances of wines.

Membrane-based technologies

Nonporous membrane technologies, due to their potential control by on-line sensors and their narrow specificity, provide powerful tools for the tuning of alcool content and pH.

Reducing wine alcohol content

The removal of ethanol from wine after fermentation can be achieved by using various technologies like membrane filtration, distillation under vacuum or atmospheric pressure, spinning cone columns, adsorption (on resins, silica gels or zeolite), freeze concentration, evaporation and extraction using organic solvent or supercritical solvent. Among all these methodologies, semi-permeable membranes by which alcohol can be separated from wine applied several years (Aguera et al., 2010). Among membrane filtration techniques, reverse osmosis (RO) is the most u technique as it works at low temperatures with few negative effects on wine sensory attributes. Since water is removed along with alcohol, it should be added back to the treated wine or added to wine before RO application. Since addition of water to wine is prohibited in many countries, the resulting permeate should be coupled to steam distillation or membrane contactor technologies to separate water from ethanol (Bes et al., 2010). However, aroma loss and change in flavor components with amount of ethanol removed (Aguera et al., 2010). Therefore the sensory properties and acceptability of ethanolremoved wines may differ based on the final ethanol concentration and consumer preferences.

Adjusting wine pH

Tartrate stabilization by electrodialys is the original "control and monitoring" electro-process, since it combines a mathematical model to calculate and predict the conductivity reduction needed on any wine to guarantee tartrate stability at -4°C/24.8°F for 6 days. Based on the same device, but using different membrane stacks, on-line acidification or de-acidification by bipolar electrodialysis system under enological conditions removes either potassium or organic acids from grape must or wine (Escudier and Le Gratiet, 2012). Depending on membrane type and process selected, it is possible to achieve either (1) the combined extraction of anions and cations to exactly reach the desired level of tartrate stability, (2) the exclusive extraction of cations to accurately adjust wine pH by a desired pH reduction, or (3) the exclusive extraction of anions to reduce wine acidity and increase the pH.

To decrease potassium content and consequently pH, an electro-membrane process with bipolar membranes has been successfully tested in wine (Lutin et al., 2010). The influence of these processes on taste perception (sourness, bitterness, astringency) was investigated. Decreasing wine pH by 0.4 points using an electro-membran technique significantly reduced bitterness and enhanced acidity of red wines, while no effect on astringency was observed (Samson et al., 2009; Caillé et al., 2011). Tannin composition was not modified by the electro-membran treatment. 3.5 pH wines presented lower levels of monomeric anthocyanins but showed higher color intensity than 3.9 pH wines. These changes reflect a higher rate of conversion of monomeric anthocyanins to derived pigments at the lower pH value. Whether these reactions are related to changes in taste properties (especially reduced bitterness of the lower pH wine) remains to be investigated (Müller et al., 2007).

Bipolar membrane electrodialysis as well as use of anionic resins reduce the mineral content of musts, in particular NH4+ (Bouissou et al., 2014). Therefore, acidification of musts that exhibit a low content of assimilable nitrogen may require nitrogen supplementation to avoid slow fermentations.

These new processes recently obtained OIV and are now authorized in Europe (reg. UE N°53/2011).

Control of key winemaking operations

Fermentation

One of the concerns enologist during winemaking is to ensure steady and complete fermentation so all the sugars in the must are converted to alcohol. When sugar concentrations are up to 250 g/l, fermentation is challenging and necessitates an optimal management of yeast nutrients.

Combining addition of ammoniacal nitrogen and oxygen is very efficient because of the positive effects of oxygen on yeast survival and of nitrogen on yeast activity (Sablayrolles et al., 1996). However this addition has to be correctly timed, i.e. at the start of the stationary phase, when about 5% ethanol has been produced (Blateyron and Sablayrolles, 2001).

Casalta et al. (2013, 2016) highlighted the interaction between grape solids (the main source of lipids) and assimilable nitrogen content, pointing out the importance of taking into account the balance between assimilable nitrogen and lipids for controlling fermentations during white winemaking. Thus, the strategy to control the alcoholic fermentation should not only take into account the addition of oxygen and nitrogen but also the lipid status of the must.

Soubeyrand et al. (2005) showed the of adding protective nutritional elements during the rehydration phase. The yeast protector releases specific micronutrients and micro-protectors that move to the active yeast enhancing its effectiveness during alcoholic fermentation (Salmon and Ortiz-Julien, 2007). The micronutrients (vitamins and minerals) allow the yeast cells to reactivate their internal metabolism. The micro-protectors (specific sterols and polyunsaturated fatty acids [PUFA]) gradually integrate into the yeast cell membrane (Luparia et al., 2004), strengthening it and facilitating exchanges with the exterior, thereby preventing the loss of cell material.

Oxidation

the break of grape berry compartmentation during the various technological operations (de-stemming, pressing, crushing, etc), the dissolution of oxygen in the must leads to several oxidation reactions which modify, to varying degrees, the initial chemical composition of the must. This cellular disintegration the substrates oxidation – the native phenolic compounds of the grape –, oxygen and the enzymatic polyphenoloxidase activity (PPO) of the grape. The color of the must then to brown tones with frequent change transparency (Cheynier et al., 2010) and aroma (Hoffman et al., 1996).

In the current context of decreased SO2 effectivness (reduction of the doses combined with pH increase of the musts), the behavior of the grape PPO activity during the winemaking process was recently studied (Frissant et al., 2012; Sire et al., 2016). This work led to new information on protection against oxidation during pressing, i.e. ) the inerting of a pressing system such as developed by certain manufacturers can be advantageously replaced by a reduction in the temperature of the grape harvest and its preservation during pressing and ) the quality increase of the end juices of pressing can also be obtained by specific cooling of the grapes with a very strong sensory gain.

The use of inactivated yeasts enriched in glutathione during alcoholic fermentation was also studied in order to protect rosé and white wines from oxidation during wine aging (Aguera et al., 2012). Aging of the corresponding wines in accelerated oxidative conditions or in standardized conditions clearly evidenced a protection of such treated wines against oxidation, from both a sensory and chromatic point of view. This effect was particularly emphasized for the conservation of odorant varietal thiols.

several years, it has been demonstrated that yeast cells, even in a non-viable physiological state, could exhibit a superfluous oxygen consumption capability, related to a mild oxidation of cellular lipids (Salmon, 2006). Such a potential of oxygen consumption by non-viable yeast cells was therefore applied for the protection of wines against oxidation during storage and/or aging. Yeast derivatives were therefore applied at enological doses on white wines stored under different conditions (with and without SO2 protection). The obtained data suggested that such specific selected inactivated yeast could represent an innovative solution for the protection of wines against oxidation during storage and aging when compared to SO2 addition in the protection against oxidative evolution of stored wines (Sieckowski et al., 2016a, 2016b).

Conclusion

Many different strategies can be used to adapt winemaking to consequences of climate change on grape composition. Some of them are already functional while others are still under investigation. The objective is not only to correct ‘defaults’ such as high ethanol concentrations or low acidity, but also to look for optimal strategies integrating viticultural and winemaking approaches to optimize wine quality.