As a non-recombinant means of strain improvement, adaptive evolution is a technique with great potential. In this first report of the use of adaptive evolution in the improvement of a commercial wine yeast strain, a sequential batch fermentation system was used to adaptively evolve the wine strain L-2056 and its haploid derivative, C9. Mutants were isolated under the selective pressures of a winelike fermentation after approximately 350 generations (from L-2056) and 250 generations (from C9) and were demonstrated to have altered production of metabolites, including ethanol, glycerol, and succinic and acetic acid. Additionally, the evolved isolate of the commercial wine yeast was able to more rapidly catabolize all available sugars under these conditions. These results endorse the potential of adaptive evolution as a tool for the non-recombinant modification and optimization of industrial yeast strains.
In the production of fermented beverages such as wine, beer, and cider, yeast facilitate the biochemical conversion of sugars to ethanol and carbon dioxide and produce a number of sensorily important metabolites (e.g., higher alcohols, organic acids, and esters), consequently influencing product quality (Romano et al. 1998, Lambrechts and Pretorius 2000). Selected strains of Saccharomyces cerevisiae are widely used in these industries because of their generally greater fermentation reliability and their preferred sensory contribution to the product. Nevertheless, their application does not guarantee fermentation will proceed without difficulties. For example, a common problem affecting winemaking, often despite inoculation with selected strains, is the occurrence of stuck or sluggish fermentations, which result from a attenuation of yeast sugar catabolic capacity and/or restricted biomass formation (Salmon 1996, Bisson and Butzke 2000). The problem can arise for many reasons (Alexandre and Charpentier 1998), but common causes are nutrient deficiencies or the inability of the yeast to function effectively in the hostile environment prevailing in grape juice: low pH values (~3.5), high osmolarity (sugar content, up to 300 g L−1 for dry table wines), temperature extremes during processing (0 to 35°C), low oxygen availability or oxidative stress through aerobic handling, among others (Boulton 1980, Bisson 1999). Incompletely fermented wines not only fall outside specifications for some wine styles but also are at increased risk of microbial spoilage and oxidation, and hence loss of commercial value. As a consequence, there is a strong demand for strains that are better suited to these conditions and that can make a positive sensory contribution to the wine.
Optimized strains can be developed by several methods including hybridization or genetic engineering. However, hybridization strategies may be difficult to apply since wine yeasts are generally homothallic and frequently exhibit poor sporulation and spore viability (Snow 1983, Walker et al. 2003). Additionally, screening for hybrids can be complicated since industrial strains do not necessarily have auxotrophic markers and dominant markers (arising through spontaneous mutation) must be specifically isolated and checked for retention of original fermentative properties (Pérez et al. 2000, Ambrona et al. 2005). As an alternative, recombinant techniques are precise but the use of this technology to optimize strains intended for fermented beverage production is not favored by some consumer groups (Pretorius and Bauer 2002). Adaptive evolution is an approach that has yet to be widely used in the development of optimized industrial yeast.
Adaptive evolution defines a set of mutations that occur in response to a specific challenge and that are advantageous to the cell under these conditions (Foster 1999). By this technique an organism is subjected to serial or continuous cultivation for many generations under conditions to which it is not optimally adapted (Brown et al. 1998, Ferea et al. 1999). In the presence of such an environmental stress, an increase in fitter variants occurs because of natural selection. The mechanism of adaptive evolution is still debated, but three models for this process are postulated: mutations are targeted toward genes in which a mutation directly relieves the stress (directed mutation model), genome wide mutation rates increase such that both adaptive and nonadaptive mutations are stimulated (hypermutation model), or mutations arise infrequently but DNA replications acting on multiple DNA copies making normal mutation appear enhanced or gene specific (cryptic growth model) (Hersh et al. 2004).
As a strain-optimization technique, adaptive evolution has the advantage of not necessarily requiring previous genetic modification of the strain of choice or application of a complicated method to identify desirable derivatives or knowledge of the genes involved in the attribute. Furthermore, the increased ploidy (that is, aneuploid and polyploid) of wine strains could be well suited to adaptation through this technique (Bakalinsky and Snow 1990, Querol et al. 2003). Consequently, adaptive evolution may offer a versatile non-recombinant means for the development of optimized wine yeast.
The aim of this study was to investigate the merits of the use of adaptive evolution to derive novel wine yeast strains, especially those suited to the stresses of fermentation. Given the complex nature of industrial grape juice/must fermentations, we used a system that would emphasize a specific subset of stresses prevalent under such conditions. These stresses were chosen to reflect typical practices in New World winemaking regions: deliberate inoculation of each batch of grape juice with an aerobically propagated starter culture. Accordingly, apart from the high sugar and low pH environment of the growth medium, we used a sequential batch fermentation system with each batch subjected to an aerobic propagation step followed by anaerobic fermentation. It was hypothesized that this multistressed approach would result in adaptation of the yeast culture to one or more of these parameters, yielding derivatives that were better able to grow under the imposed stress or yield a different profile of metabolites.
Materials and Methods
Saccharomyces cerevisiae strains L-2056 (Lallemand, Adelaide, Australia, selected from vineyards of the Cotes du Rhone) and C9 (L-2056 ho MATa), a stable haploid derivate thereof, were used. The construction of C9 by disruption of HO using a removable KanMX-based disruption construct is described elsewhere (Walker et al. 2003).
Sequential batch fermentation.
A starter culture was grown aerobically for 24 hr at 30°C in a 250-mL shake flask containing 50 mL of YPD medium (1% w/v yeast extract, 2% w/v bactopeptone, 2% w/v d-glucose) and shaken at 160 rpm on a rotary platform. Adaptation of strains to high-stress conditions was achieved by growth in 1 L of a chemically defined grape juice medium (CDGJM) in sequential batch fermentations in bench-top fermentors (2 L, B Braun Biotech, Melsungen, Germany). The formulation of CDGJM used here differed from that originally reported (Henschke and Jiranek 1993) in that it contained glucose and fructose (each at 100 g L−1) and ammonium sulfate (580 mg free amino nitrogen L−1) as the sole nitrogen source (to eliminate mutants with amino acid auxotrophies, which would be impractical in industry) (Table 1⇓). The sterols and unsaturated fatty acids, otherwise known as oxygen substitutes or survival factors (Larue et al. 1980), were deliberately omitted from the medium. Fermentations were initiated by introduction of the yeast starter culture to 5 x 106 cells mL−1 and the inoculated medium aerated using sterile compressed air (0.45 μm, 2 VVM) for 4 hr with agitation at 350 rpm. Aeration sought to replicate typical industry practices for generation of an inoculum by aerobic cultivation, during which a theoretical preaccumulation of sterols and unsaturated fatty acids could occur. Such an extensive aeration was considered to lead to an oxidative stress. Culture vessels were then fitted with a fermentation lock to allow them to become self-anaerobic and were maintained at 30°C with agitation at 150 rpm. Fermentations were complete when less than 2.5 g L−1 of residual sugars remained, at which point the fermentor was drained. Sufficient culture remained in the fermentor to inoculate a fresh batch of media to ~1 x 105 cells mL−1. The refilled fermentor was sparged with compressed air as before, the fermentation lock refitted, and the culture was again monitored for fermentation completion before the entire cycle (draining, refilling, sparging, fermentation) was repeated.
Analysis of culture isolates.
Given each batch fermentation was inoculated to ~1 x 105 cells mL−1 and resulted in ~2 x 108 cells mL−1, there were of the order of 11 generations per fermentation cycle. Culture samples were collected at 50-generation (culture doubling) interims, checked for purity by plating, and designated FM1 to FM5 (for C9) and FM10 to FM16 (for L-2056). The fermentation performance of cultures from each sample was compared to the parental strains in separate shake-flask cultures. Upon identification of novel attributes, after some 250 and 350 generations for strains C9 and L-2056, respectively, the fermentation performance of both the mixed cultures and the 10 single-colony clones of each evolved culture was evaluated. Clones were chosen randomly through colony isolation after streak-plating and incubation on YPD. Batch fermentations where conducted by inoculating yeast from plates into 10 mL YPD and grown overnight (to ~2 x 108 cells mL−1) before inoculation at 1 x 105 cells mL−1 into 100 mL CDGJM. Triplicate fermentations of parental strains and mixed culture isolates were conducted. Single-colony isolates were conducted as single fermentations. Fermentations were performed as described previously (Walker et al. 2003) in 250-mL conical flasks fitted with a fermentation lock. Near-anaerobic conditions of industrial fermentations were mimicked by flushing the headspace of each flask after inoculation with filtered (0.45 μm), high-purity nitrogen (~20 VVM, 5 min). Flasks were incubated at 30°C with shaking at 160 rpm.
Culture development was monitored during logarithmic phase by determining the optical density at 600 nm (OD600). Culture dry weight was determined by collecting biomass from 5-mL culture aliquots on a predried and preweighed 0.22-μm membrane filter, followed by washing, drying, and reweighing. Commercially available kits were used for enzymatic quantitation of glucose and fructose (Boehringer Mannheim, Mannheim, Germany). Succinic acid, acetic acid, acetaldehyde, glycerol, and ethanol were determined by HPLC analysis as described previously (Walker et al. 2003).
Chemostat selection and analysis of candidate variants.
In order to isolate yeast strains that displayed phenotypes of greater fermentation performance under stressful winelike conditions, S. cerevisiae strains L-2056 and C9 were subjected to adaptive evolution through sequential batch fermentation. Experimental design was based on the notion that the condition of excess oxygen before an extended anaerobic fermentation of a medium devoid of sterols and unsaturated fatty acids, high in sugar, and low in pH would provide a selective pressure(s). Low inoculum density further exacerbated the influence of these stresses by encouraging greater cell division and therefore a reduction in the cellular concentrations of sterols and unsaturated fatty acids at the end of fermentation. It was hypothesized that evolved strains dominating the culture after extended exposure to these stresses would be better able to complete fermentation during exposure to one or more of these parameters.
Each chemostat was sampled every 50 generations, and a number of fermentation parameters were analyzed. Comparative small-scale batch fermentations of mixed cultures were conducted to allow monitoring of fermentation duration, glucose, and fructose utilization and the formation of biomass, acetic acid, succinic acid, glycerol, and ethanol. Fermentation duration by each 50-generational sample was seen to match the parent strains until a point where an adaptive mutation(s) was proposed to have occurred: after 250 generations under selective pressure for C9 and after 350 generations for L-2056. The mixed culture (FM16-M; M denotes the mixed, unpurified culture) derived from L-2056 displayed superior fermentative abilities (Figure 1⇓) as well as altered metabolite yields (data not shown). While that was initially also the case for cultures derived from C9 (FM5), subsequent analysis revealed that significant differences were limited to parameters other than fermentation kinetics (Table 2⇓). Thus, FM5-M produced 8.6% more biomass at the end of fermentation than the parent strain. Also, ethanol yield was decreased, with a concurrent increase in glycerol, acetic acid, and acetaldehyde.
For FM16-M, the differences between the evolved cultures and the parent were typically greater in magnitude (Table 3⇓). FM16-M was able to catabolize all available sugar after 96.5 hr, which is ~69% of the time required by L-2056 from which FM16-M was derived. No deviation in biomass formation was apparent between FM16-M and the parent strain; however, the production of most metabolites by FM16-M did deviate. For instance, there was a marked increase (12%) in glycerol production in the mixed evolved culture FM16-M and less succinic acid produced compared with L-2056 (Table 3⇓). Unlike the FM5-M and C9 comparison, there was little deviation between FM16-M and L-2056 in the production of acetic acid. Culture growth, estimated by absorbance during exponential growth and as a function of cell numbers at the end of fermentation, did not differ markedly between evolved cultures and their respective parental strain (data not shown).
To further elucidate the phenotypic differences between the two mixed-culture evolved cultures and their respective parents, 10 random single-colony representatives from the FM5-M and FM16-M populations were selected to undergo preliminary fermentation analysis (metabolite and fermentation data are shown in Tables 2⇑ and 3⇑, respectively). In both instances, one of the 10 analyzed clones failed to complete fermentation within the time frame of the experiment; therefore, the data for this strain was omitted from mean data calculations. For the most part, the mean for the clones from each culture modeled the trends of the corresponding mixed culture. However, there was occasional large variation in the production of metabolites and the measured fermentation dynamics for the single clones. For example, the mean succinic acid yields by the clonal cultures were seen to increase by ~10% rather than remaining largely unchanged in the mixed culture derived from FM5. Interclonal variation was also highlighted by the measures of dry cell weight; however, again the mean resembled that of the parent, C9. Metabolite differences between C9 and FM5, which are of greatest industrial importance, are highlighted in Figures 2⇓ and 3⇓. Most notably, FM5-M clearly produces more glycerol and less ethanol, with almost all FM5-C strains mimicking this result. Preliminary results for FM5-C3 suggest that this strain may demonstrate the most favorable metabolite production with regard to enological significance; that is, an increase in glycerol production and a reduction in ethanol production. Further characterization of this strain will better define the scale of these highlighted changes.
Similar patterns were seen in the FM16 single isolates, with the mean values for metabolite yields by these isolates broadly agreeing with those established for the mixed culture. Succinic acid values did vary widely, however (Table 3⇑). Fermentation duration was dramatically reduced compared to the parent strain, L-2056 (Figure 4⇓), although the variation between clonal cultures was wide (77.5 to 119 hr). Glycerol production by the same evolved isolates also consistently increased over the parent strain (Figure 5⇓), and the mean of the single isolates was similar to that of the mixed culture. As with the analysis of the clonal derivatives from C9, examples can be found among the set of derivative of L-2056 which indicate marked changes in metabolite yields and fermentation duration. Because of the potential sensory and processing significance of these changes, such examples (FM16-C7) warrant further investigation under a broader set of conditions.
The sluggish catabolism of sugars displayed by L-2056, as correlated by the protracted fermentation compared to the evolved cultures (Figure 1⇑), is considered highly undesirable under industrial enological conditions. When sugar catabolism subsides toward the end of fermentation, it has a high likelihood of becoming stuck, which is problematic because residual sugar of even 5 to 10 g L−1 can dramatically change the style of a wine, especially when dry styles are the intended product (Iland and Gago 2002). In addition, restarting a fermentation in order to bring a wine within specification can be extremely difficult where arrest has occurred in the later stages when ethanol concentrations are high.
Numerous stresses are typically imposed on yeast under enological conditions, and such stresses may have contributed to the unintentional adaptive evolution of contemporary strains from less well-suited progenitors (Querol et al. 2003). It was therefore hypothesized that deliberate application of a selective pressure(s) through a laboratory-based adaptive evolution experiment could be used to yield yeast more adept at completing fermentation under winelike conditions. We applied an environment reminiscent of winemaking that could be considered highly stressful: oxidative stress from prefermentation propagation and aerobic handling of grape must/juice, extended fermentation induced by high concentrations of glucose and fructose, anaerobic growth in the absence of oxygen substitutes (sterols and unsaturated fatty acids), low pH, and a low rate of inoculation. The work was performed with a wine strain (L-2056) and, to facilitate later genetic characterization, a haploid derivate (C9) (Walker et al. 2003).
The measured fermentation duration for generational isolates up to and including FM4 (200 generations) and FM15 (300 generations) were similar to the parental strain in each case. However, adaptation to the selective pressure occurred in the haploid C9 culture after approximately 250 generations and the commercial L-2056 culture after approximately 350 generations. The evolved cultures were able to ferment more rapidly and/or produce altered amounts of metabolites with sensorial or quality importance for wine. Consequently, it is proposed that a defining mutation or mutations occurred after the points where FM4 and FM15 had been collected.
While the rapid fermentation seen by FM16-M and clonal isolates may be considered undesirable under optimal conditions, under suboptimal conditions such as during the nutrient limitations that commonly occur during grape juice fermentations (Henschke and Jiranek 1993, Alexandre and Charpentier 1998, Bisson 1999), this property may help overcome protracted fermentation. For the wine industry, fermentation reliability is a key yeast attribute (Pretorius 2000, Pretorius and Bauer 2002); thus a strain with such an improvement would be highly prized by winemakers. Perhaps of equal importance to rapid fermentation is completion of fermentation, again a property displayed by the majority of the clonal isolates from FM16-M. Together the attributes of reduced fermentation duration and complete catabolism of sugars can go a long way to achieving a final product without off-flavors and poor palate structure, thereby enhancing (or preserving) the commercial value of the wine.
Without large-scale gene or genome analysis techniques, the existence and precise nature, number, or origin of the mutations proposed to have occurred in this study cannot be defined. Nevertheless some comments about the adaptive evolution process are warranted. First, yeast in this experiment spent most time in a resting state. Thus of the three mechanisms by which mutations may arise (directed mutation model, hypermutation model, or the cryptic growth model) (Hersh et al. 2004), any mutations that occurred were likely to have arisen from a replication independent process (Heidenreich et al. 2003). It is also noteworthy that adaptive evolution of the haploid strain occurred some 100 generations sooner than it did for the commercial strain (presumed to be a diploid) (Walker et al. 2003). Ploidy may be important in this outcome (Orr and Otto 1994, Zeyl et al. 2003), although this disparity may not be the same in other circumstances. By definition, a new allele requires a molecular mutation event followed by a fixation process, whereby the new allele replaces all others in the population (Zeyl 2004). While higher ploidy reduces the waiting period by providing more alleles from which the next adaptive mutation can arise (Paquin and Adams 1983), nondominant mutations may take longer to fix because they confer smaller selective advantages on heterozygotes (Zeyl et al. 2003). Thus in a larger population, as in this experiment, fixing time rather than the waiting time takes precedence, and haploids should evolve more rapidly (Orr and Otto 1994). Further experiments with strains of different ploidy would help explain the basis for the different evolution rates.
Given that such an extended incubation was required to yield a culture with enhanced enological properties, a clearly predominant, single clone might have been expected (Sonderegger and Sauer 2003). That was not the case, however. Characterization of clonal isolates from the mixed cultures, FM5-M and FM16-M, revealed clear heterogeneity. The single-clone fermentations derived from both FM5-M and FM16-M generally yielded a mean response that was similar to that of the mixed-culture populations (Tables 2⇑ and 3⇑). One advantage of considering individual clones is that the performance of some appeared better than that of the mixed cultures. More detailed characterization of these and additional isolates might uncover even better clones.
Numerous groups have isolated evolved strains using continuous culture with defined working volumes and specified dilution rates (Ferea et al. 1999, Sonderegger and Sauer 2003). Here, use of a sequential batch cultivation system ensured that any evolved strains selected would be suited to fermentation under enological conditions and seemed to eliminate selection of alterations that are disadvantageous to general fermentation performance. This system, although useful in this application, extended the duration of the selection process considerably. In a similar work on the adaptation of a haploid recombinant strain to anaerobic xylose utilization, Sonderegger and Sauer (2003) isolated an evolved strain after 460 generations over 266 days, whereas isolation of FM5-M after some 250 generations required 198 days. The difference in total duration between our batch-fed continuous fermentation and a continuous culture is mainly due to the prevalence of stationary phase during high sugar fermentations.
The increased biomass production (dry weight) demonstrated by evolved cultures at the end of fermentation might have arisen through a divergence of carbon during glycolysis. With FM5-M, increased biomass yield could be partly related to a reduced production of ethanol and/or a response to the toxicity of these metabolites (Aranda and del Olmo 2003). In a winemaking context, reduced ethanol yield can be a very positive outcome. Ethanol increases the perception of “heat” on the back palate and masks many flavor and aroma compounds of wine (Iland and Gago 2002). Thus the avoidance of higher ethanol concentrations, particularly those resulting from the use of overripened fruit, can be most beneficial. So-called low-ethanol yeasts are attracting interest (Pretorius 2000, Pretorius and Bauer 2002), particularly in New World wine-producing countries where flavor-ripe grapes are readily achieved and the resulting flavor-intense wines are valued by consumers. A reduced ethanol yield would presumably also benefit the yeast itself by reducing the toxic effects of this metabolite during fermentation (Ansanay-Galeote et al. 2001). The superior performance of FM5 may in part be a result of the lower ethanol concentration of these fermentations.
Glycerol yields were increased by approximately 10% in both mixed evolved cultures. Recombinant strains that overproduce glycerol have previously been shown to also yield elevated amounts of acetic acid because of an oxidation of acetaldehyde by aldehyde dehydrogenases (Eglinton et al. 2002) as a mechanism for maintaining redox balance. While increases in acetic acid production were seen by the strains investigated here, the extent to which such a compensatory process might explain these yields is uncertain. Similarly, the basis for the glycerol response of the evolved strains investigated here is yet to be determined, but mechanism aside, increased diversion of carbon to glycerol is likely to be viewed favorably by winemakers, as glycerol has the potential to enhance the mouthfeel of wine (Nieuwoudt et al. 2002).
Succinate production was increased in the haploid-evolved strains; however, a reduction was generally observed in the diploid. There may be several reasons for this deviation. Diversion of more carbon through the tri-carboxylic acid pathway is one possibility. During fermentation, the tricarboxylic acid pathway operates in a branched fashion, and it is thought that most succinic acid is formed by the reductive pathway (Camarasa et al. 2003). Alternate routes rely on glutamate being present and therefore are probably unimportant for the ammonium-grown cultures investigated (Albers et al. 1998, Coleman et al. 2001, Tate and Cooper 2003). However, the genes involved in these pathways will be targeted in further work to help determine the extent of their involvement in the modified succinic acid yield of FM5-M, FM16-M, and clonal isolates.
Although metabolic engineering has become a standard practice for strain improvement in other industries, issues concerning consumer acceptance currently impede such use of strains in the wine industry. Adaptive evolution offers a non-recombinant technique to isolate commercial wine yeast strains that are highly tailored to the stressful conditions of a typical wine fermentation. To our knowledge, this report is the first to detail application of this technique to a commercial wine yeast strain. Our findings indicate the feasibility of this technology in this field even if some selection strategies, such as the sequential batch approach used here, require extended culture times.
↵3 (present address) School of Pharmacy and Medical Sciences, Division of Health Sciences, University of South Australia, Adelaide, SA 5000, Australia.
Acknowledgments: This project was supported by Australia’s grapegrowers and winemakers through their investment body the Grape and Wine Research and Development Corporation, with matching funds from the Australian Government (projects UA 01/04 and UA 05/01). C. McBryde holds a University of Adelaide Postgraduate Research Scholarship.
- Received May 2006.
- Revision received June 2006.
- Copyright © 2006 by the American Society for Enology and Viticulture