A systems biology perspective of wine fermentations


  • Francisco Pizarro,

    1. Department of Chemical and Bioprocess Engineering, College of Engineering, Pontificia Universidad Católica de Chile, Casilla 306, Correo 22, Santiago, Chile
    Search for more papers by this author
  • Felipe A Vargas,

    1. Department of Chemical and Bioprocess Engineering, College of Engineering, Pontificia Universidad Católica de Chile, Casilla 306, Correo 22, Santiago, Chile
    Search for more papers by this author
  • Eduardo Agosin

    Corresponding author
    1. Department of Chemical and Bioprocess Engineering, College of Engineering, Pontificia Universidad Católica de Chile, Casilla 306, Correo 22, Santiago, Chile
    • Department of Chemical and Bioprocess Engineering, College of Engineering, Pontificia Universidad Católica de Chile, Casilla 306, Correo 22, Santiago, Chile.
    Search for more papers by this author


The yeast Saccharomyces cerevisiae is an important industrial microorganism. Nowadays, it is being used as a cell factory for the production of pharmaceuticals such as insulin, although this yeast has long been utilized in the bakery to raise dough, and in the production of alcoholic beverages, fermenting the sugars derived from rice, wheat, barley, corn and grape juice. S. cerevisiae has also been extensively used as a model eukaryotic system. In the last decade, genomic techniques have revealed important features of its molecular biology. For example, DNA array technologies are routinely used for determining gene expression levels in cells under different physiological conditions or environmental stimuli. Laboratory strains of S. cerevisiae are different from wine strains. For instance, laboratory yeasts are unable to completely transform all the sugar in the grape must into ethanol under winemaking conditions. In fact, standard culture conditions are usually very different from winemaking conditions, where multiple stresses occur simultaneously and sequentially throughout the fermentation. The response of wine yeasts to these stimuli differs in some aspects from laboratory strains, as suggested by the increasing number of studies in functional genomics being conducted on wine strains. In this paper we review the most recent applications of post-genomic techniques to understand yeast physiology in the wine industry. We also report recent advances in wine yeast strain improvement and propose a reference framework for integration of genomic information, bioinformatic tools and molecular biology techniques for cellular and metabolic engineering. Finally, we discuss the current state and future perspectives for using ‘modern’ biotechnology in the wine industry. Copyright © 2007 John Wiley & Sons, Ltd.


Wine and winemaking
Yeast physiology in wine fermentations
Improvement of wine yeasts
Framework for the integration of genomic information, bioinformatic tools and molecular biology for the rational design of improved wine yeast strains
Future perspectives


The yeast Saccharomyces cerevisiae is an important industrial microorganism. It has long been utilized in bakery to expand or raise dough, and in the production of alcoholic beverages, fermenting the sugars of rice, wheat, barley, corn and grape juice. More recently, yeasts are being used as cell factories for the production of important pharmaceuticals, such as insulin57 and polyketides69. Applications of S. cerevisiae in so-called ‘white biotechnology’ (production of low-cost, high-volume products) are rapidly gaining momentum, e.g. in the production of L-lactic acid for the plastics industry100.

S. cerevisiae is also one of the most widely used model eukaryotic systems. More than 10 years have passed since the sequencing of its genome44. Since then, many functional analysis projects have been dedicated to the investigation of its molecular biology. It was originally believed that, after obtaining the genome sequence, one would have access to comprehensive information on the inner workings of microorganisms. However, the complexity arising from the genome sequencing projects required new comprehensive post-genomic strategies. These included advanced studies on regulatory mechanisms, as well as the application of new high-throughput technologies at the different levels of cellular complexity (genome, transcriptome, proteome and metabolome), and an efficient analysis of the resulting data that requires application of new bioinformatic methods in an integrative or systems biology perspective26.

At present, serial analysis of gene expression (SAGE),115 oligonucleotide microarrays, cDNA microarrays62 and Affymetrix GeneChips120 are techniques routinely used for determining gene expression levels and gene expression ratios in cells under different physiological conditions or environmental stimuli42, 104, 108. Although technically more complex, proteomic studies are steadily picking up the pace, complementing transcriptome studies. These allow the identification of post-transcriptional regulation sites45, 54, 58, 59, 119 and the determination of protein–protein interaction maps43, 46, 55, 111. Metabolome studies have allowed classification of yeast mutants, using metabolic foot- and fingerprinting5, 52, 117. These studies have revealed important features of the molecular biology of S. cerevisiae.

The growth conditions in wine fermentations are very different from laboratory conditions. Therefore, industrial strains present several distinctive features that allow them to adapt to the prevailing conditions. In the following sections, we review the most recent applications of post-genomic techniques to understand yeast physiology in the wine industry and the evolutionary differences that confer desirable traits to industrial strains in general. We then review recent advances in wine yeast strain improvement and propose a reference framework for integration of genomic information, bioinformatic tools, and molecular biology techniques for cellular and metabolic engineering. Finally, we discuss the current state and future prospects for using ‘modern’ biotechnology in the wine industry.

Wine and winemaking

Wine is an alcoholic beverage derived from grape juice (also known as grape must), highly popular throughout the world. With a history of more than 8000 years, wine production is one of the world's oldest biotechnological processes88, 110. From a physicochemical point of view, wine can be defined as a non-ideal, multicomponent liquid solution containing water, ethanol, glycerol and organic acids as major constituents, as well as other minor components, such as flavours, aromas and phenolic compounds. These are either originally present in the grapes or synthesized during winemaking.

S. cerevisiae plays a fundamental role in the anaerobic transformation of grape must into wine. The fermenting medium could be inoculated with commercial, selected yeast strains or it could be left to ferment by the native flora present in the winery or on the grape. In both cases, S. cerevisiae governs the fermentation18.

In the case of natural or spontaneous fermentations, a sequential succession of microorganisms occurs spontaneously as the must conditions change with time41. This includes the interaction of fungi, yeasts, bacteria (acetic and lactic bacteria) and viruses (mycoviruses and bacteriophages)15, 38, 88. In this context, each microbial succession contributes to the change in must composition, with metabolites of some species becoming substrates for others. The interactions of these microbial communities generate synergistic effects on the formation of aromatic compounds that cannot be reproduced by mixing musts from pure-culture fermentations. This has been recently demonstrated by metabolic footprinting of inoculated, natural and mixed-population fermentations52.

Nowadays, however, most wine fermentations begin by inoculating grape must with a specific and pure yeast culture. Culture selection depends on the grape cultivar, must composition, general conditions of fermentation and on the final product required. From a biotechnological point of view, the use of active dry yeast has a high impact on the microbiology of the fermentation process. Unlike the natural process, the addition of pure cultures induces a clear predominance of the inoculated strain over the endogenous strains15, 82.

Grape must composition presents culture conditions that are far from optimal for most microorganisms. Upon inoculation, yeast cells must adapt to the low pH (2.9–3.8) and high osmolarity of the new environment before fermentation begins (sugar concentration up to 300 g/l), as well as to the high SO2 content (40–100 mg/l). After the onset of fermentation, the biological activity of the yeast causes various stress conditions throughout the fermentation process, which include rapid nutrient limitation and starvation (nitrogen, which is available in quantities of just 50–600 mg/l at the start of the fermentation, is soon depleted), temperature variations and ethanol toxicity (final ethanol concentration is 11–14% v/v). The adaptation and response mechanisms to several of these stresses are still largely unknown7, 12, but the recent application of functional genomic approaches under relevant industrial conditions has helped to identify genes that are regulated coordinately, the transcription factors that control metabolism and some relevant technological properties8, 47, 97, 98, 114.

The fermentation process consists of three phases: lag, exponential and stationary. During the lag phase, the inoculated yeast must adapt its metabolism to the high initial glucose/fructose content—usually in a 1 : 1 w/w ratio—in order to grow and transform these sugars into ethanol. In the exponential phase, biomass increases exponentially. One-third of the ethanol and the main fraction of glycerol present in wine are obtained in this phase. In the stationary phase, the remaining two-thirds of the ethanol and the wine aromatic compounds which determine the quality of the final product are produced101. As we will see below, all of these stages impose important stress conditions upon the yeast cell (Table 1).

Table 1. Genes expressed in S. cerevisiae in response to different stress conditions at different stages of the wine fermentation
Growth phaseStress conditionGeneDescriptionReference
LagGrowthRPR, NOP, RPS22A, RPL1A, RPS6BGenes involved in ribosome assembly, protein synthesis, RNA metabolism and nucleotide biosynthesis75,98
 HBT1, UBC8, UBP13Genes related to ubiquitin or ubiquitin-like processes75
 HXT1,5, PYK1Genes involved in glucose metabolism75
 High osmolarityGPD1Glyceraldehyde 3-phosphate dehydrogenase, proposed as an osmotic stress marker56,84,85
 PDC6Isoform of pyruvate decarboxylase; its expression appears to be specific to sugar stress34
 GON7Gene referred to as hydrophilins, involved in the response of osmotic stress75
 AcidityPDR12Gene specifically induced by acid, proposed as a marker of the acid response97
LogSulphurSUL1 -2Genes involved in sulphur transport27
 SSU1Encodes for a plasma membrane protein involved in sulphite resistance47,78
 AnaerobiosisPAUFamily of genes described as seripauperines, related to the TIP/TIR gene stress family. Their function and the conditions that trigger their expression are unknown. Suggested to be expressed in anaerobic conditions93,97
 EthanolFBA1Gene encoding aldolase. With strong expression, FBA1 promoter might be used to overexpress genes at greater levels114
Early stationary 
 Nitrogen deprivationPUT 1,2Encoding proline carriers; these genes are required for the utilization of proline97,114
 DAL1-7,80-82Genes required for degradation of allantoin, a metabolite involved in purine metabolism for nitrogen recycling. Suggested markers for nitrogen limitation20,86,97
 DUR1,2Genes required for the utilization of urea97
Late stationary GAP1, CAN1, DUR3, BGL2, MEP2, PTR2Genes encoding nitrogen permeases27,86,97
 GDH2, GLN1, GAD1, UGA1Genes involved in management of the glutamate pool97
 ESRGroups of genes have been referred to environmental stress response, involving most of the genes shown to respond to nitrogen depletion or to encode proteases or proteins involved in nitrogen utilization and autophagy97
 CERGroups of genes referred to the common environmental response97
 General stressMSN2/MSN4Genes of transcription factors involved in the general stress response, including classical stress-responsive heat shock and responses to ethanol, osmotic and oxidative stresses8,56,83,86,95,97,114
 Oxidative stressAAD, TRX2Genes encoding for an aryl-alcohol dehydrogenases and an antioxidant protein, respectively. TRX2 had been proposed as an oxidative stress marker56,85, 97
 Glucose deprivationHXK1Gene encoding hexokinase 1. This protein exhibits higher affinity for fructose than the HXK2p isoform, favouring fructose utilization97
 Proton homeostasisHSP30, PMP2These genes encode a negative and a positive regulator of the plasma membrane H + -ATPase, respectively114

Wine fermentation ends when dryness is achie- ved, i.e. when the concentration of residual sugars is below 2–4 g/l. This process usually takes 7–10 days for red wines and 15–20 days for white wines, although it could last for several weeks if problematic fermentations arise. The low sugar concentration, in addition to the wine's storage under anoxic conditions, prevents oxidation reactions that could impair the quality of the final product. Moreover, an almost complete removal of sugar prevents the subsequent growth of acetic and lactic acid bacteria, which could metabolize the residual sugar, generating an increase in acidity as well as formation of off-flavours3.

Yeast physiology in wine fermentations

Alcoholic fermentation is a dynamic and complex process during which the grape-must undergoes continuous transformations, which are due to large changes in environmental cultural factors as well as the biological activity of the fermenting yeast. Therefore, wine yeasts have developed mechanisms allowing them to sense environmental changes in order to maintain the integrity of the cell and its metabolic activity throughout the whole winemaking process12.

Laboratory and industrial wine yeasts

Comprehensive studies carried out with yeast laboratory strains have characterized the gene response and adaptation to several stress conditions, such as nutrient limitation20, anaerobic conditions60, 109, ethanol stress4, osmotic stress121 and low temperature51, 86. Yeast also encounters some of these stress conditions throughout the wine fermentation, but caution must be exercised when extrapolating these results to wine yeasts. Indeed, laboratory yeast strains are unable to achieve wine dryness, i.e. complete transformation of sugar into ethanol. The wine yeasts have gone through an extensive selection process and their response differs in some aspects from laboratory strains91. Furthermore, as discussed above, standard culture conditions are usually very different from winemaking conditions, where multiple stresses occur simultaneously and sequentially.

From the different studies on wine yeasts using cDNA microarrays, various approaches have been adopted with respect to the growth conditions and the experimental design of the assay. Some experiments employed standard laboratory culture media for growing both laboratory and wine strains27, 47, 86, while other authors simulated enological conditions by growing the wine yeast on a fully defined culture medium, synthetic must, which mimics grape must in its composition8, 56, 75, 78, 83, 85, 97, 98, 114. Culture conditions differ greatly between experimental set-ups (Table 2).

Table 2. Differences in medium and culture conditions between laboratory-, wine- and industrial- strains of S. cerevisiae
 Laboratory yeast under standard laboratory conditionsWine yeast under laboratory conditionsWine yeast under real conditionsTypical industrial production 
  • *

    Batch fermentations

Medium ConditionsMediumMinimal or rich mediaSynthetic mustReal MustRich medium
 Limiting nutrientCarbon limitedNitrogen limitedNitrogen limitedCarbon limited
 Osmotic PressureLowHigh at start batch ferm.High at start batch ferm.Low
 Oxygen contentMainly aerobicAnaerobicLargely AnaerobicAerobic
 Presence of toxic metalsNoNoPotentially copperNo
 Presence of xenobioticsNoNoyesNo
 Microbial ecologySterileSterileNon-sterileSterile
 SO2 contentNoNo40–100 mg/LNo
Operation ConditionsDuration∼1 day*∼7 Days*White wine: 15–30 days Red wine: 6–10 daysVariable 1–10 days
 Volume (L)0–20–21,000–100,000100–100,000
 Culture systemBatch/ContinuousBatch/ContinuousBatchFed-Batch
 Source of InoculumFrom liquid or solid YPD pre-cultureFrom liquid or solid YPD pre-cultureActive dry yeastActive dry yeast
 AgitationShaker/mixingMixingNatural convection/pump overMixing

Industrial strains of S. cerevisiae differ from most laboratory-bred strains, which are either haploid or diploid. Industrial wine yeast strains are predominantly diploid or aneuploid, and occasionally polyploid10, 11 and show a high leve1 of chromosome length polymorphism17, 92. Furthermore, several commercial traits (e.g. fermentation capacity, metabolite production, stress tolerance) are under multigenic control and their molecular origin is largely unknown. Microarrays have been used for comparative genotyping of standard laboratory, mutant and industrial yeast strains30, 47, 53. For example, a major group of shared genomic patterns that are found in wine strains, but which are not present in laboratory strains, consists of genes, especially transporters and permeases, involved in drug-resistance pathways33. This is probably a result of the adaptative response to the many stresses encountered during winemaking. This technique could be particularly useful to identify single gene amplifications or deletions that are responsible for positive traits found in some native yeasts that come from a similar background as the strain that is used to build the arrays. However, it could yield confusing results with highly divergent strains and will not identify point mutations in coding and regulatory sequences.

Laboratory and wine yeast strains exhibit divergent evolutionary differences in response to environmental changes33. Laboratory yeasts have been isolated and propagated as a pure culture for several decades (70 years for S288C)33, specially selected by their ability to grow under conditions of carbon limitation, exhibiting a more efficient use of the available carbon source, and higher specific sugar uptake rates86. On the other hand, wine yeasts have been selected by their successful adaptation to the severe environmental conditions of winemaking. The latter include higher biomass yields on nitrogen and special adaptation to nitrogen-limiting conditions, which are characterized by an increase in transcription of genes involved in both the anabolic and catabolic phases of nitrogen metabolism86, 88, 97.

Differential gene expression during wine fermentation

In recent publications, two independent groups have described the genome-wide response of active dry yeast to rehydration and inoculation75, 98. Upon rehydration, active dry yeast has to adapt to the change from a carbon-limited, oxidative medium used for yeast production31 to an oxygen-limited, sugar-rich medium with a high osmotic pressure. This change might be accompanied by a transient stress response to the high medium osmolarity, yet both studies have shown no induction of a stress response. The response is best characterized by an activation of the synthetic machinery of the cell75, which is congruent with the sudden availability of nitrogen and carbon sources, and an induction of glycolytic genes and carbon catabolite repression98, as a result of the high concentrations of glucose.

The lag phase is followed by the exponential growth phase, which is characterized by an induction of PAU genes related to cell wall biogenesis and associated with increased anaerobic conditions93, 97. This growth phase is also characterized by a transition from a nitrogen catabolite (NCR)-repressed state to a corresponding nitrogen catabolite-derepressed state as the nitrogen source becomes exhausted16.

The metabolic adaptation of a wine yeast strain to high sugar concentration (osmotic stress) was studied using microarray technology34. Glycolytic genes (HXT1, HXT5, ENO1, PDC6, GLK1, GPM2) are upregulated in response to high sugar concentrations. These results are in agreement with a recent study comparing laboratory and wine yeasts in continuous culture, where HXT1, HXT6/7, ENO1 and PDC5 were upregulated in the wine yeast86. Overall, these results suggest that natural selection in wine yeast strains leads towards higher glycolytic fluxes.

As the culture progresses into the stationary phase, a general stress response is triggered, characterized by an induction of the common environmental response (CER), the environmental stress response (ESR) and few heat-shock genes97. This was also observed by serial analysis of gene expression (SAGE)114 and partly results from nitrogen depletion, but mainly from the response to the increasing ethanol concentration, as shown by the high expression of genes encoding proteins involved in proton homeostasis, HSP30 and PMP2114. This indicates that yeasts are subjected to extreme stress conditions during this phase. In fact, trehalose accumulation was highly correlated with ethanol concentration in these cultures113. An induction of ergosterol biosynthetic genes, the proline oxidase gene and other genes related to electron transport chain was also observed97, 114. These were unexpected results, since they involve pathways that require molecular oxygen as substrate, and they were found even under the strict anaerobic conditions imposed in the work of Varela and co-workers113. Post-transcriptional mechanisms are probably responsible for the regulation of the corresponding proteins, which would enable the cell to rapidly respond when oxygen is again available114.

Both SAGE and microarrays yield similar results112, but because SAGE does not require previous knowledge of gene sequence, it is of great interest for the discovery of novel genes. In fact, in a SAGE study conducted on wine yeasts, 10% of the transcripts matched to unannotated regions within the yeast genome, and 22% of the tags expressed in the late stationary phase did not match any known region within the laboratory yeast genome114. Recent bioinformatic tools have been developed for the unambiguous tag-to-gene assignment in SAGE experiments68. This method reduces ambiguity in tag assignment through a scoring system for assigning match probability. Using this software, five tags matching to intergenic regions show a statistically significant coordinated overexpression during late stationary phase. The regions containing these tags have no homologues in any species studied so far, and could represent the best opportunity to discover novel genes not present in the standard laboratory strain114.

The effect of different amounts of assimilable nitrogen was studied under winemaking conditions8. Nitrogen deficiency triggers a downregulation of biosynthetic genes, with a concomitant reduction in nitrogen-rich macromolecules113. Nitrogen limitation has also been studied in laboratory20 and wine yeasts86 under carefully controlled conditions, using continuous culture. The transcription of many genes is related to the growth rate of the microorganism. In steady-state continuous culture chemostats, the growth rate is a constant function of the feed rate and can be fixed at suboptimal values. This experimental design is very advantageous, as the physiological state of the cell can be controlled and therefore the transcriptional effect of growth rate can be neglected from the assay, thus effectively allowing the researcher to define the boundary conditions of the experiment and isolating the transcriptional effect of the variable in study48. In this respect, DAL genes (involved in allantoin metabolism) have been found to be markers of nitrogen limitation in steady-state chemostats20. Furthermore, we have found differences in the expression levels of these genes between laboratory and industrial strains in nitrogen-limited culture that consumed equal amounts of ammonia. The only differences were found in the intracellular free amino acid content of these cells, suggesting that DAL expression could be used as a marker of the state of nitrogen-stress within the cell86. These results reveal that continuous cultures can be used for the identification of molecular markers of a given environmental condition that can be used to monitor the state of the culture. This might open new avenues for the operation of bioprocesses whereby the operating conditions can be designed to obtain the desired genetic profile for optimum product formation. In the wine field, for example, understanding the effects of microaerophilic conditions through carefully conducted experiments with known dissolved O2 concentrations, at the p.p.m. level, could lead to the rational design of equipment to provide the necessary dose to stimulate fermentation, preventing the oxidation of valuable aromas.

Network component analysis (NCA), which is a methodology for reconstructing regulatory signals from transcriptomic analysis such as DNA microarrays and SAGE, can be used to determine which transcription factors are active at specific points in the fermentation64. The application of this technique to the available transcriptomic data on wine fermentations could help further our understanding of the regulatory response to any specific condition. Furthermore, from transcriptomic studies one can determine gene sets that are expressed only in particular culture conditions (e.g. only in late stationary phase, at low temperature or under nitrogen limitation). These gene subsets are particularly useful to obtain regulatory sequences that can be used to manipulate the expression of natural or introduced gene products that are transcribed only under these conditions.

Improvement of wine yeasts

One of the main uses of genomic techniques is in the metabolic and cellular engineering of bacterial and eukaryotic cells for strain improvement. The need for new wine yeast strains derives from both producer- and consumer-oriented requirements. Winemakers require cost-competitive production of wine with minimum resource inputs. These include improved fermentation performance, in the form of general resilience, ethanol and stress tolerance, efficient nitrogen assimilation and resistance to antimicrobial compounds, among others. Improved processing efficiency can be accomplished by improving protein and polysaccharide clarification and controlled cell sedimentation, flocculation and flotation. Great commercial losses can be avoided by improving the biological control of spoilage microorganisms through the expression of antimicrobial enzymes and peptides or through the metabolic production of sulphur dioxide89. Consumer-orientated tendencies include the requirement of new healthful wines, i.e. wines that enhance the benefits and decrease the risks associated with moderate consumption88. This includes wines with higher antioxidant (resveratrol) content and lower concentrations of toxic substances, such as ethyl carbamate or biogenic amines, and also wines with lower ethanol content. Increasing the organoleptic properties of wine includes enhanced production of desirable volatile esters, optimization of phenolic content and reduced production of higher alcohols, sulphites and sulphides18, 28, 63, 116. In this way, both producers and consumers are pressing for new specialized and improved wine yeast strains88.

The genetic improvement of industrial strains has traditionally relied on classical genetic techniques—such as variant selection, mutagenesis, hybridization (mating, spore-cell mating), rare mating, cytoduction and spheroplast fusion—followed by selection for broad traits, such as fermentation capacity, ethanol tolerance and absence of off-flavours (e.g. hydrogen sulphide). These methods are especially advantageous to improve and combine traits under polygenic control and do not give rise to products that are included in the definition of genetically modified organisms (GMOs). Therefore, strains improved with these methods are not treated with the same level of public suspicion as are wine yeasts that have been transformed with foreign DNA and are not subject to the same strict regulations that pertain to GMOs89. Despite considerable work, a major limitation of these techniques is their difficulty for adding or removing features from a strain without altering its performance32.

Other techniques for strain optimization include directed evolution in continuous culture, e.g. chemostats, Brown and Oliver Interactive Continuous Selection (BOICS)23, 36, 61. An example of this methodology is the optimization of glucose consumption in yeast36. In this work, the performance of a parental yeast strain was compared against three evolved strains that had grown for 250–500 generations under glucose limitation. After this time, each of the evolved strains had increased biomass yield on glucose (∼four-fold more cells, three-fold greater biomass) and each left ∼three-fold less glucose unused. Moreover, all three evolved strains produced at least one order of magnitude less ethanol than the parental strain. In this experimental set-up, microarrays helped to identify the changes in the transcriptome of the evolved cells. The patterns of gene expression in the evolved strains were remarkably similar to one another. Genes with altered expression in the three evolved strains included genes involved in glycolysis, the tricarboxylic acid cycle, oxidative phosphorylation and metabolite transport36. Although this experiment was conducted under carbon limitation and is particularly interesting for the industrial production of yeasts, the same experimental set-up may be used to obtain improved wine yeast strains that are, for example, more efficient in nitrogen assimilation, by growing wine strains under nitrogen limitation.

In the last decades, gene technology has broadened the possibilities to add new traits and improved characteristics into the target cell. This has been the topic of excellent reviews18, 32, 88, 89 and therefore will only be briefly discussed here. Some of these advances include glycerol overproduction,70, 71, 94 lower ethanol yields50, 67, 72, increased release of phenolic compounds103, malo-ethanolic118 and malo-lactic yeast strains6, 21, decreased ethyl carbamate production28, production of the grape antioxidant resveratrol13, 14 and increased ester production65, among others.

Driven by a consumer tendency favouring wines with lower alcohol content, the wine industry is leaning towards the production of these wines. However, in order to favour high flavour intensity, winemakers are harvesting fully matured, sometimes overripe grapes (with very high sugar concentrations) that result in wines with excessive alcohol levels. Actual procedures for alcohol removal include technologies such as spinning cone column and reverse osmosis, both of which have a detrimental effect on wine quality, are expensive and difficult to operate. A biological alternative includes using yeast strains that give lower ethanol yields. This can be accomplished by diverting carbon flow from ethanol toward other metabolites. This has been carried out very elegantly by engineering the redox metabolism of S. cerevisiae with the expression of a water-forming NADH oxidase from Lactococcus lactis49. Expression of this gene alone results in a wild-type phenotype under anaerobic conditions, but allows a 15% reduction in the ethanol yield under microaerobic conditions. Unfortunately, the transformation also causes a sharp decrease in the fermentation rate, an increase in the acetaldehyde concentration and a stuck fermentation if the initial sugar concentration is over 100 g/l. However, when culture conditions were modified to precisely control oxygen additions, the expression of the NADH oxidase gene allowed a reduction of 7% in the ethanol yield, without affecting fermentation performance50. This was accomplished by adding the oxygen in the stationary phase, but could also be done by engineering the heterologous expression of this gene to be exclusively expressed in the stationary phase.

Glycerol is the most abundant by-product of alcoholic fermentation, after ethanol and carbon dioxide. This compound may contribute to the mouth feel and perceived sweetness of wine. Therefore, many attempts have been made to increase the glycerol yield during fermentation. Another interest for rerouting the carbon flux towards glycerol is the decrease in ethanol yield. This constitutes a microbial alternative (or complement) to the current commercial procedures for ethanol removal from beverages (see above). The most direct approach for overproducing glycerol was to overproduce GPDH. Overexpression of GPD1 in wine and brewer's yeasts resulted in a two- to three-fold increase in glycerol production and a lower ethanol yield70–73, 94. These strains exhibited increased production of undesired by-products, mainly acetate and acetoin, and other products such as 2,3-butanediol and succinate, as a result of an altered NADH metabolism and a rearrangement of carbon fluxes. This was partially solved by knocking out the ALD6 gene in these strains. With this strategy, high amounts of glycerol were obtained without increasing acetate formation, whilst increasing carbon flux towards the formation of succinate, acetoin and 2,3-butanediol25. We postulate that the use of a more rational approach, including the use of metabolic models, could have predicted the increase in acetate prior to the actual gene modification87. As we will see in the following section, the use of bioinformatic tools, associated with genome-wide analysis in metabolic engineering, could aid in the identification of the most appropriate candidate gene knockouts for the redirection of carbon flux towards products that do not have a detrimental effect on wine quality.

Framework for the integration of genomic information, bioinformatic tools and molecular biology for the rational design of improved wine yeast strains

Metabolic engineering is defined as the directed improvement of the cellular properties achieved from the interplay of theoretical analysis, relying on biochemical information, and the application of genetic engineering9, 106. What distinguishes metabolic engineering from classical applied molecular biology is the use of a directed, rational approach. This implies that it is necessary to have a solid knowledge of the system being studied and requires a careful analysis of the cellular system for the construction of the recombinant strain74, 76.

Until now, most of the successes in genetic engineering have been based on qualitative or intuitive design principles. However, even though there are several success stories in genetic engineering, there are also many attempts that have failed, due to the lack of rational strategies based on robust predictive analysis tools81. A logical framework structure to proceed in a metabolic engineering approach for the improvement of wine yeasts is summarized in Figure 1. The desired traits are determined from an analysis of either producer or consumer preferences. Then, in a metabolic engineering approach, the candidate gene(s) subject to modification are identified, aided by bioinformatic tools (see below), in the design section. Afterwards, the appropriate modification of the target genes is performed in the synthesis section. Finally, a complete systems analysis is carried out on the new strain. Usually one does not reach the desired phenotype in one round and therefore several iterations are necessary.

Figure 1.

Framework for the integration of genomic information, bioinformatic tools and molecular biology for the rational design of improved wine yeast strains

It is difficult to predict the effects of introducing genetic modifications in the design of a metabolic engineering strategy because microbial metabolism is often subjected to tight regulation and is constrained by mass, energy and redox conservation laws on a large number of intracellular metabolites. Moreover, as metabolic pathways and related regulatory processes form complex molecular and functional interaction networks54, 80, it is only through analysis of the metabolism in an integrative systems approach107 that one may evaluate the effect of specific genetic modifications81. Here, metabolic models that predict cellular physiology based on biochemical information are of great assistance87, 99, especially if they are built on a genome scale29, 90. Based on a genome-scale model of yeast40, Patil and co-workers81 have developed a bioinformatic tool named OptGene that uses an evolutionary programming algorithm to identify sets of gene deletions that maximize the yield—or the productivity (yield and growth rate)—of a given metabolite. This approach suggests strategies that are non-intuitive genetic modifications that span several different pathways. Complementing this approach, the addition of heterologous genes (pathways) may also be evaluated24.

The metabolic engineering approach presented in Figure 1 is mostly directed towards the genetic manipulation of organisms. Despite scientific accomplishments in strain development through gene technology, commercial strain development for the wine industry cannot involve genetic engineering techniques at this time (see review102). However, if different approaches for modification of the yeast strains are used, such as directed evolution, mutation/selection, hybridization or any of the methods mentioned above, the available tools for genome-wide analysis can be used to determine whether the acquired traits fit the desired phenotype and could be used for screening of mutants that carry natural deletions in the genes of interest.

After obtaining the modified strains, a careful analysis should be performed on the new strain, involving a systems approach that uses transcriptome, metabolome and/or flux analysis (preferably in combination) to evaluate the physiological impact of the performed genetic modification. This system allows better evaluation of possible side-effects arising from the genetic modification (e.g. formation of an off-flavour). These data are also of great value in the design stage of further modifications (iterative process).

Of course, one is not restricted to enter the cycle in the design section. Alternatively, one may compare strains with different productivities for a given trait, and analyse the genome-wide differences that lead to the different phenotypes. Then, in a reverse metabolic engineering approach22, first comes the analysis, followed by design and development of the strategy to modify the candidate genes in the target cell. This is particularly useful for the wine industry, where autochthonous S. cerevisiae strains have been shown to have considerable differences regarding volatile compounds, for example, compared to commercial wine yeast strains79, 96, 105.

Non-Saccharomyces strains have been shown to produce significant benefits for wine, such as secretion of several enzymes, including pectinases (which increase juice extraction, improve clarification and facilitate wine filtration), β-glycosidases (which may hydrolyse non-volatile glycosidic aromatic precursors from the grape) and proteases (which improve clarification), among others, that contribute to reveal the varietal aroma and improve the winemaking process35, 37, 39, 77. Commercial exploitation of these strains has not been successful, due to their limited fermentation properties (e.g. poor ethanol tolerance, low fermentation rate). However, according to the GOLD database (http://www.genomesonline.org)66, multiple genome sequencing projects are under way for relevant microorganisms related to wine fermentations, such as Saccharomyces bayanus, Pichia pastoris and the spoilage yeast Zygosaccharomyces rouxii, which will allow the development of genome-wide tools for the analysis of these species.

Future perspectives

The shortest path to commercial implementation of genetically modified (GM) yeasts will probably lie in strains developed through self-cloning techniques that are based on the use of host-derived material28. Recently, a saké yeast strain that was improved through self-cloning was approved by the Japanese Government and does not need to be treated as a GMO2, 102. This technology has been used to develop commercial wine yeast strains that produce significantly lower amounts of the carcinogen ethyl carbamate, by placing the DUR1,2 gene under the control of the PGK1 promoter28. This strain received Generally Regarded as Safe (GRAS) status from the FDA in 2006. The potential health benefits associated with the use of this strain for wine production may help improve consumer perception of GM products.

There is vast potential benefit to the wine consumer and producer alike in the application of gene technology. Experts agree, however, that the benefits will be realized only if the application is judicious, systematic and achieved with high regard for the unique nature of the product. The first GM wine products should unequivocally demonstrate organoleptic, health and economic advantages, primarily for the consumer, before general public acceptance18, 89.

A promising research field links the components of the wine metabolome (grape- and yeast-derived) with the sensory impact they have on wine through gas chromatography–mass spectrometry (GC–MS) coupled with GC–olfactometry1, 19. This technique could determine how different styles of winemaking affect the wine metabolome and quantify the sensory impact on wine. Coupled to hedonic preference mapping, this will allow further development of custom-made strains to consumer preferences. Arguably, this should not lead towards more uniform wines, but rather to a plethora of wines destined to occupy small market niches suited for particular preferences of a particular consumer sector. Thus, gene technology is harnessed in this way to expand the diversity of high-quality wines.

Finally, we must not forget that wine is a product of science and art. In this article, we have addressed the scientific issues in the field of yeast development for the wine industry, but we must emphasize that these are (potential) tools for the winemaker. These are new brushes and exotic colours that the artist may use to develop this wonderful beverage called wine.