Evolutionary engineering of Saccharomyces cerevisiae for improved industrially important properties


  • Z. Petek Çakar,

    Corresponding author
    1. Istanbul Technical University, Dr Orhan Öcalgiray Molecular Biology, Biotechnology and Genetics Research Center, ITU-MOBGAM, Istanbul, Turkey
    • Department of Molecular Biology and Genetics, Faculty of Science and Letters, Istanbul Technical University, Istanbul, Turkey
    Search for more papers by this author
  • Burcu Turanlı-Yıldız,

    1. Department of Molecular Biology and Genetics, Faculty of Science and Letters, Istanbul Technical University, Istanbul, Turkey
    2. Istanbul Technical University, Dr Orhan Öcalgiray Molecular Biology, Biotechnology and Genetics Research Center, ITU-MOBGAM, Istanbul, Turkey
    Search for more papers by this author
  • Ceren Alkım,

    1. Department of Molecular Biology and Genetics, Faculty of Science and Letters, Istanbul Technical University, Istanbul, Turkey
    2. Istanbul Technical University, Dr Orhan Öcalgiray Molecular Biology, Biotechnology and Genetics Research Center, ITU-MOBGAM, Istanbul, Turkey
    Search for more papers by this author
  • Ülkü Yılmaz

    1. Department of Molecular Biology and Genetics, Faculty of Science and Letters, Istanbul Technical University, Istanbul, Turkey
    2. Istanbul Technical University, Dr Orhan Öcalgiray Molecular Biology, Biotechnology and Genetics Research Center, ITU-MOBGAM, Istanbul, Turkey
    Search for more papers by this author

Correspondence: Z. Petek Çakar, Department of Molecular Biology and Genetics, Faculty of Science and Letters, Istanbul Technical University, 34469 Maslak, Istanbul, Turkey. Tel.: +90 212 285 72 63; fax: +90 212 285 63 86; e-mail: cakarp@itu.edu.tr


This article reviews evolutionary engineering of Saccharomyces cerevisiae. Following a brief introduction to the ‘rational’ metabolic engineering approach and its limitations such as extensive genetic and metabolic information requirement on the organism of interest, complexity of cellular physiological responses, and difficulties of cloning in industrial strains, evolutionary engineering is discussed as an alternative, inverse metabolic engineering strategy. Major evolutionary engineering applications with S. cerevisiae are then discussed in two general categories: (1) evolutionary engineering of substrate utilization and product formation and (2) evolutionary engineering of stress resistance. Recent developments in functional genomics methods allow rapid identification of the molecular basis of the desired phenotypes obtained by evolutionary engineering. To conclude, when used alone or in combination with rational metabolic engineering and/or computational methods to study and analyze processes of adaptive evolution, evolutionary engineering is a powerful strategy for improvement in industrially important, complex properties of S. cerevisiae.


‘Rational’ metabolic engineering and its limitations

Since its first definition in 1991 by Jay Bailey as ‘the improvement of cellular activities by manipulation of enzymatic, transport, and regulatory functions of the cell with the use of recombinant DNA technology’ (Bailey, 1991), metabolic engineering has been an increasingly important interdisciplinary field of biotechnology owing to new developments in high-throughput analytical technologies in functional genomics and bioinformatics.

The classical or so-called rational metabolic engineering involves proposal of a defined genetic manipulation that is expected to provide a benefit via the perturbation of the known biochemical network, based on the knowledge of the metabolic system of interest (Bailey et al., 1996). However, trying to reengineer a complex cellular machine without a detailed plan and extensive knowledge of how it really works is the major problem in rational metabolic engineering. Consequently, despite some successful examples of this rational approach in bacterial amino acid and fine chemical processes, many other ‘rational’ metabolic engineering studies resulted in metabolic consequences different from those expected upon the genetic changes introduced (Bailey, 1991; Bailey et al., 1996). Such failures are resulting from the following limitations of ‘rational’ metabolic engineering: (1) requirement of extensive biochemical and genetic information on the metabolism or metabolic pathway(s) of interest, regulatory factors, enzymes involved and their kinetics, flux-limiting steps, etc.; (2) complexity of cellular physiological responses, such as activation of an alternative metabolic pathway by the cells as a response to the inhibition of an existing pathway by metabolic engineering; thus, a high number of ‘stimulus–response’ experiments are necessary to estimate the results of a perturbation realistically; and (3) difficulties of cloning in industrial strains, mainly resulting from their genetic complexity such as polyploidy, or regulatory issues such as the use of genetically modified organisms (GMO) in food industry, or simply the lack of necessary vectors and methods for their genetic manipulation (Bailey, 1991; Bailey et al., 1996).

Inverse metabolic engineering and its applications

To overcome limitations of ‘rational’ metabolic engineering, a ‘bottom-up’ approach called ‘inverse metabolic engineering’ was defined by Bailey and coworkers (Bailey et al., 1996) as an alternative strategy. In this approach, the starting point is the identification, isolation, or calculation of a desired phenotype. The next step is the determination of the genetic basis and/or the particular environmental factors that confer that phenotype. This step is the most challenging step of inverse metabolic engineering. However, with rapid developments in high-throughput analytical technologies in functional genomics (Bro & Nielsen, 2004) as well as proteomics, it has become possible to analyze the whole genome or proteome of an organism in detail in a relatively short period of time. The last step of inverse metabolic engineering strategy is the transfer of the identified genetic basis to a chosen, suitable industrial organism to obtain the desired phenotype in that organism. As inverse metabolic engineering already starts with the last step of the ‘rational’ approach as its first step, which is the identification of the desired phenotype, this strategy is also called as the ‘bottom-up’ approach (Bailey et al., 1996). The major advantage of inverse metabolic engineering is that, unlike ‘rational’ metabolic engineering, it does not initially require extensive biochemical, genetic, or regulatory information on the organism of interest to obtain the desirable phenotype.

Inverse metabolic engineering studies include a variety of successful examples such as improvement in folate production in lactic acid bacteria (Sybesma et al., 2003), the use of phosphagen kinase systems to improve cellular energy metabolism (Sauer & Schlattner, 2004), improvement in galactose uptake and fermentation of Saccharomyces cerevisiae (Bro et al., 2005; Lee et al., 2011), and increasing alcohol tolerance in S. cerevisiae (Hong et al., 2010). However, when the desirable phenotypes are more complex and involve many genes or more than one operon, evolutionary strategies for the selection of desirable phenotypes might be advantageous.

An inverse metabolic engineering strategy: evolutionary engineering

Continuous evolution procedures based on the application of a selection procedure to obtain a desired phenotype are generally called ‘evolutionary engineering’ (Butler et al., 1996; Sauer, 2001). Evolutionary engineering is one of the approaches that can be employed to generate diversity in performance, which is the starting point for inverse metabolic engineering studies. A classical empirical method for microbial strain development is physical or chemical mutagenesis followed by direct selection on solid culture medium that provides the selective pressure for improved mutants, such as high salt and metal concentrations, high or low temperatures. The major drawback of this simple method, however, is the fact that repeated cycles of mutation and selection on solid plates usually result in highly specialized but crippled mutant strains (Çakar et al., 2005). In other words, the mutants obtained may have one highly improved property under selective conditions, but they may have significantly lower growth rates than the wild type.

Contrary to repeated cycles of mutation and selection of colonies on solid culture plates, evolutionary engineering involves a more systematic approach: repeated batch cultivations can be performed in the presence of a selective pressure, or alternatively, prolonged chemostat cultivations can be performed under selective conditions. Those cultivations can be performed with a wild-type or, to increase genetic diversity, a chemically/physically mutagenized strain. Spontaneous or induced mutagenesis of the initial monoclonal population results in the formation of fitter variants in the initial monoclonal population as a consequence of the selection pressure applied throughout the cultivations. These fitter variants can survive and grow better than the original cells under the selection conditions. Thus, in chemostat or serial batch cultures, the ratio of the number of fitter cells to the total number of cells in the culture will periodically increase, and the fitter cells will dominate the culture (Dykhuizen & Hartl, 1983; Sauer, 2001).

Successful evolutionary engineering examples exist with a variety of industrially important microorganisms, such as Escherichia coli, where physiological properties such as stress resistance (Weikert et al., 1997), recombinant protein production (Weikert et al., 2000), metabolic activity, (Sonderegger et al., 2005) and lactic acid production (Fong et al., 2005) were improved, Acetobacter aceti with improved acetate resistance (Steiner & Sauer, 2003), Bacillus subtilis with improved plasmid stability (Fleming et al., 1988), and Actinobacillus succinogenes (Roca et al., 2010) and Mannheimia succiniciproducens (Lee et al., 2010) with enhanced succinate production. However, many other evolutionary engineering studies are with the baker's yeast S. cerevisiae to improve a variety of physiological and production properties of this microorganism. Considering the industrial importance of S. cerevisiae in bakers and brewers industry, in wine and ethanol production, and in biological research as a well-known eukaryotic model organism for studying complex processes, it may not be surprising at the first sight to encounter several evolutionary engineering studies to improve S. cerevisiae. However, it is important to note that improving industrial yeast strains by evolutionary engineering may be more difficult than improving laboratory strains. The main reason for this is the ploidy issue. Industrial yeast strains that are usually polyploid are less likely to accumulate relevant recessive mutations by laboratory evolution. Nevertheless, successful studies on laboratory evolution using polyploid industrial strains exist (see for example Teunissen et al., 2002). The major reason and – at the same time – advantage of using evolutionary engineering approach to improve the properties of yeast is the fact that introduction of a foreign/recombinant gene into the microorganism of interest is not necessary for evolving an existing/natural characteristics of that microorganism. This approach simply mimics the nature by random mutation of the microorganisms’ own genes, followed by selection under suitable conditions to favor the desired phenotype. Thus, S. cerevisiae and other microorganisms to be used in food and beverage industry will not be considered as GMOs, when improved by evolutionary engineering, and will most likely have higher public acceptance as also mentioned previously (Çakar et al., 2005; Kutyna et al., 2010).

Evolutionary engineering studies with S. cerevisiae

Major applications of evolutionary engineering in S. cerevisiae are summarized in Table 1 and will be discussed in this section in two categories: (1) evolutionary engineering of substrate utilization and product formation and (2) evolutionary engineering of stress resistance.

Table 1. Major applications of evolutionary engineering in Saccharomyces cerevisiae
Evolutionary engineering applicationSelected references
Ethanol fermentation using xyloseSonderegger & Sauer (2003), Kuyper et al. (2005), Van Maris et al. (2007), Liu & Hu (2010)
Ethanol fermentation using arabinoseWisselink et al. (2007)
Ethanol fermentation using mixed substrates (xylose and arabinose)Wisselink et al. (2009), Sanchez et al. (2010)
Ethanol fermentation using lactoseGuimaraes et al. (2008ab)
Four-carbon dicarboxylic acid productionZelle et al. (2010)
Improved resistance to inhibitory compounds of lignocellulosic biomassTomas-Pejo et al. (2010), Wright et al. (2011)
Improved freeze toleranceTeunissen et al. (2002)
Improved ethanol toleranceStanley et al. (2010ab)
Improved multistress resistanceÇakar et al. (2005)
Improved cobalt resistanceÇakar et al. (2009)

Evolutionary engineering of substrate utilization and product formation in S. cerevisiae

A major area of evolutionary engineering research with yeast is bioethanol production based on lignocellulosic biomass. Lignocellulosic substrates can be obtained from industrial, municipal, and agricultural wastes and forestry residues. Thus, they are cheap and highly abundant substrates suitable for significantly increasing ethanol production (Almeida et al., 2007). Lignocellulosic biomass is usually hydrolyzed by acid pretreatment, resulting in hydrolysis of hemicellulose that consists of pentose sugars such as xylose and arabinose. Thus, microorganisms that can grow on pentose sugars and convert them to ethanol at high yields are highly desirable.

Saccharomyces cerevisiae is the usually preferred microorganism for ethanol production as it can produce high amounts of ethanol, is more robust than bacteria based on its higher ethanol tolerance (Almeida et al., 2007), and has Generally Regarded as Safe (GRAS) status. However, it cannot utilize pentose sugars such as xylose and arabinose. On the contrary, other yeasts such as Pichia stipitis (now called Scheffersomyces stipitis) can utilize pentose sugars such as xylose. Although recent studies have been focusing on ethanol production by S. stipitis strains (Ferreira et al., 2011) that can be potentially used for industrial xylose fermentation, many other studies have been focusing on the use of recombinant S. cerevisiae as a preferred ethanol producer. Thus, using rational metabolic engineering approach, the genes responsible for the xylose utilization pathway in P. stipitis, xylose reductase (XYL1) and xylitol dehydrogenase (XYL2), were transferred to S. cerevisiae, and a recombinant strain able to grow aerobically on xylose was obtained (Kötter et al., 1990). However, for ethanol fermentation, anaerobic growth on xylose is necessary. Thus, an early study of yeast evolutionary engineering was conducted for anaerobic growth of that recombinant xylose-utilizing strain on xylose (Sonderegger & Sauer, 2003). Selection in a chemostat for 460 generations by starting with aerobic growth on xylose and slowly adapting the culture to microaerobic and then to anaerobic conditions yielded mutant clones with improved anaerobic xylose utilization and consequently up to 19% higher ethanol production (Sonderegger & Sauer, 2003). Similarly, in a rational ‘minimal’ metabolic engineering study where a heterologous xylose isomerase from an anaerobic fungus Piromyces spp. was functionally expressed in S. cerevisiae for efficient anaerobic xylose fermentation, the recombinant strain was further improved by evolutionary engineering via serial batch cultivation conditions on xylose, first aerobically and then under severe oxygen limitation. Thus, it was suggested that activities and/or regulatory properties of native S. cerevisiae gene products can be improved by evolutionary engineering (Kuyper et al., 2004). In a follow-up study, Kuyper and coworkers applied evolutionary engineering approach for mixed-substrate utilization of a xylose-fermenting recombinant S. cerevisiae strain. In that study, the challenge was to improve xylose consumption in the recombinant S. cerevisiae strain in glucose–xylose mixtures, which was achieved by prolonged cultivation in xylose-limited, automated sequencing batch reactors on glucose–xylose mixtures. Additionally, the xylose uptake kinetics of the recombinant strain was also improved by prolonged cultivation in an anaerobic, xylose-limited chemostat (Kuyper et al., 2005). Industrial pilot-scale experiment resulted in high ethanol yields from the d-xylose substrate initially present in plant biomass hydrolysates as reported by the same research group in 2007 (van Maris et al., 2007).

Despite the general information that S. cerevisiae cannot grow by utilizing xylose as the sole carbon source, Attfield & Bell (2006) applied natural selection and breeding to native S. cerevisiae strains and obtained strains that can grow slowly on xylose as the sole carbon source, with doubling times < 6 h. This completely nonrecombinant strategy is a unique example for the development of xylose-utilizing S. cerevisiae strains without the use of recombinant DNA technology and was suggested to find better acceptance for food and fermentation applications by developing nongenetically modified strains (Attfield & Bell, 2006).

In a more recent study, combined approaches of genetic engineering, chemical mutagenesis by ethyl methane sulfonate (EMS), and evolutionary adaptation were employed to obtain a recombinant S. cerevisiae strain with high-efficiency xylose fermentation. The chemically mutagenized, recombinant S. cerevisiae strain was exposed to serial batch cultivations for growth in xylose medium, first by aerobic and then under oxygen-limiting conditions and at increasing xylose concentrations for selection. Evolutionary adaptation yielded an improved mutant producing 11% more ethanol in oxygen-limited fermentation than in aerobic fermentation conditions and consuming xylose efficiently (Liu & Hu, 2010).

Similar to the work described on xylose, there are also some studies on evolutionary engineering of arabinose fermentation in S. cerevisiae. For this purpose, bacterial l-arabinose utilization pathway, including B. subtilis AraA and E. coli AraB and AraD, was overexpressed in S. cerevisiae, with simultaneous overexpression of the l-arabinose-transporting yeast galactose permease. Sequential transfer of the resulting recombinant strain in l-arabinose media revealed an l-arabinose-utilizing yeast strain that produces ethanol under oxygen-limited conditions. Molecular analysis results of the finally obtained yeast strain showed that low L-ribulokinase activity was crucial for efficient l-arabinose utilization, as well as high l-arabinose uptake rates and enhanced transaldolase activity (Becker & Boles, 2003). In a later study on arabinose utilization, the major aim was to achieve efficient alcoholic fermentation of l-arabinose under anaerobic conditions (Wisselink et al., 2007) unlike oxygen-limited conditions, which had been described previously. Firstly, structural genes for l-arabinose utilization pathway of Lactobacillus plantarum were expressed in S. cerevisiae, along with the overexpression of the S. cerevisiae genes coding for the enzymes of the nonoxidative pentose phosphate pathway. Secondly, the resulting recombinant microorganism was exposed to extensive evolutionary engineering under both aerobic and anaerobic serial batch cultivation conditions, to improve the rate of arabinose consumption. Finally, an improved strain was obtained that had high rates of ethanol production, arabinose consumption, and a high ethanol yield (0.43 g g−1) during anaerobic growth on l-arabinose as the sole carbon source. Additionally, ethanol was produced efficiently from glucose and arabinose mixtures, which is an industrially desired property (Wisselink et al., 2007). In another study, codon optimization technique was used to improve l-arabinose fermentation in a recombinant S. cerevisiae strain encoding genes from bacterial l-arabinose utilization pathway. The codon usage of bacterial l-arabinose pathway genes was adapted to that of the highly expressed genes of glycolytic enzymes in S. cerevisiae, which resulted in improved l-arabinose conversion rates. However, for further improvement in the strain to be used more efficiently in industrial applications, evolutionary engineering was recommended (Wiedemann & Boles, 2008).

As lignocellulosic biomass consists of different types of sugars, rapid and efficient conversion of all sugars derived from this biomass is necessary for the efficiency of fermentation processes. For this purpose, evolutionary engineering approach was employed for rapid utilization of glucose, xylose, and arabinose by repeated batch cultivation with repeated cycles of growth in three different media (glucose, xylose, and arabinose; xylose and arabinose; and arabinose only) having different compositions. The evolved strain had improved specific rates of xylose and arabinose consumption (Wisselink et al., 2009). Similarly, an industrial recombinant S. cerevisiae strain was improved by evolutionary engineering for simultaneous conversion of xylose and arabinose to ethanol by prolonged aerobic, continuous cultivation first in xylose and then in xylose and arabinose medium, where xylose and arabinose transport and consumption rates as well as anaerobic ethanol production were improved (Sanchez et al., 2010).

As the challenging step of inverse metabolic engineering is the identification of the molecular basis of the desired phenotype obtained, current ‘omics’ technologies are crucial for such complex and detailed analysis. To this end, metabolome, transcriptome, and metabolic flux analyses of a recombinant S. cerevisiae strain evolved for l-arabinose fermentation were performed, and the results were compared with those of the original, wild-type strain. It was found that the expression level of the GAL-regulon was increased, implying that galactose transporter is crucial for arabinose growth. Additionally, transketolase and transaldolase enzymes were found to be possibly involved in flux-controlling steps of arabinose fermentation, emphasizing once more the role of pentose phosphate pathway in efficient pentose sugar utilization (Wisselink et al., 2010).

When compared to other yeasts such as P. stipitis that can naturally utilize xylose for growth, it was previously shown that S. cerevisiae uses the nonoxidative branch of the pentose phosphate pathway much less than P. stipitis, under batch growth conditions in glucose medium (Fiaux et al., 2003). As efficient utilization of pentose phosphate pathway is desired for many industrial applications including ethanol production from lignocellulosic biomass, an evolutionary engineering strategy was designed recently to increase the carbon flux to pentose phosphate pathway in an industrial S. cerevisiae strain. This strain was exposed to long-term batch cultivation on gluconate, which is a substrate metabolized by pentose phosphate pathway. C-13 metabolic flux analysis on glucose results showed that the evolved strain had higher flux into the pentose phosphate pathway and showed different physiological properties such as increased aroma compound production, higher rates of fermentation, and lower acetate production during wine fermentation (Cadiere et al., 2011).

Lactose fermentation by S. cerevisiae is another biotechnologically important area of evolutionary engineering, particularly regarding cheese whey fermentation. In a recent study, a recombinant S. cerevisiae strain carrying lactose utilization genes of Kluyveromyces lactis coding for lactose permease and β-galactosidase was improved by evolutionary engineering for higher lactose fermentation efficiency. A batch selection strategy based on serial transfer and dilution in lactose medium was employed, which yielded an evolved strain with twofold faster lactose consumption, 30% higher ethanol production, and efficient fermentation of concentrated cheese whey (Guimaraes et al., 2008a). The molecular basis of the evolved phenotype was also determined by plasmid copy number analysis and determination of LAC gene expression levels by quantitative real-time RT-PCR (Guimaraes et al., 2008a) and by comparative transcriptomic analysis in a follow-up study with original and evolved strains, using cDNA microarrays (Guimaraes et al., 2008b).

Another recent application area of evolutionary engineering is the biotechnological production of four-carbon dicarboxylic acids, such as succinate, fumarate, and aspartate as these compounds can be better alternatives to oil-derived chemical intermediates. In a recent study, evolutionary engineering enabled pyruvate-carboxylase-negative S. cerevisiae mutants constitutively overexpressing phosphoenolpyruvate carboxykinase from S. cerevisiae or A. succinogenes, a succinate-producing prokaryote, to grow on glucose as the sole carbon source. The evolved mutants were analyzed in detail to identify their molecular basis and to engineer low-cost and efficient anaerobic four-carbon dicarboxylic acid production in S. cerevisiae (Zelle et al., 2010). Evolutionary engineering of prokaryotic succinate production including A. succinogenes (Roca et al., 2010) and M. succiniciproducens (Lee et al., 2010) has also been reported recently.

Evolutionary engineering of stress resistance in S. cerevisiae

An important area of evolutionary engineering research is the improvement in S. cerevisiae's resistance to a variety of industrial stress conditions. Tolerance to freezing, for example, is desirable for S. cerevisiae, particularly for frozen dough production in the bakery industry. During frozen dough preparation, because of the initiation of fermentation, the freeze resistance of yeast decreases rapidly, which causes a significant decrease in its dough-rising capacity (Teunissen et al., 2002). To overcome this problem, polyploid industrial yeast strains were subjected to UV mutagenesis and screened for survival in dough exposed to 200 freeze–thaw cycles. A resulting mutant strain had higher freeze tolerance and gassing power and performed also well under industrial conditions (Teunissen et al., 2002).

High tolerance to acids is also an industrially desirable property for S. cerevisiae. Various studies exist in the literature regarding improvement in yeast acid tolerance. Early reports involve work on isolation of acetic acid-tolerant S. cerevisiae mutants in a turbidostat (Aarnio et al., 1991). In recent years, improvement in acetic and formic acid tolerance of S. cerevisiae became increasingly important because of bioethanol production from lignocellulosic plant biomass hydrolysates, particularly at low pH. Both acetic acid and formic acid are released during solubilization and hydrolysis of lignocelluloses and have inhibitory effects on growth, ethanol fermentation, and productivity (Hasunuma et al., 2011). In a recent study, two different evolutionary engineering strategies were employed to increase acetic acid tolerance of xylose-fermenting S. cerevisiae. The first strategy was based on sequential anaerobic, batch cultivation at pH 4 at increasing concentrations of acetic acid. The second strategy involved prolonged anaerobic continuous cultivation with controlling pH. In the latter strategy, ammonium assimilation causes acidification, which results in selective conditions for acetic acid tolerance. The mutant cultures obtained from both selection strategies had improved acetic acid tolerance (up to 6 g l−1) after about 400 generations. However, upon storage at −80 °C and cultivation in the absence of acetic acid, the mutants could not grow on xylose at pH 4 in the presence of 5 g l−1 acetic acid. Further analysis and characterization in chemostat cultures with linear acetic acid gradients showed that, in chemostat selection strategy, the acetic acid tolerance phenotype was acetate inducible (Wright et al., 2011).

The presence of inhibitory compounds such as acetic acid is an important issue in ethanol production from lignocellulosic compounds. The inhibitory compounds include furan derivatives, weak acids such as acetic acid, formic acid, and levulinic acid, and phenolics (Almeida et al., 2007). These compounds usually inhibit yeast growth and ethanol production. To overcome such inhibitory effects, a variety of methods have been used, including fermentation control strategies such as increasing cell density to enhance volumetric conversion rates of inhibitors (Palmqvist & Hahn-Hagerdal, 2000) or operating in fed-batch mode to keep the inhibitor concentration low (Taherzadeh et al., 2001; Rudolf et al., 2005). Alternatively, recombinant S. cerevisiae strains were constructed with improved inhibitor tolerance. Evolutionary engineering of S. cerevisiae for improved hydroxymethylfurfural and furfural tolerance was also employed in batch cultures by serial transfer of cultures to media containing the inhibitors at increasing concentrations. The evolved strains were shown to be more tolerant to the inhibitors tested and metabolized glucose faster than the control strain (Liu et al., 2005). The importance of and improvement in robustness of xylose-fermenting S. cerevisiae strains to be used in industrial application with respect to the tolerance to inhibitory compounds in lignocellulosic hydrolysates were also emphasized (van Maris et al., 2007). Heer & Sauer (2008) performed a chemical analysis of four different lignocellulosic hydrolysates and identified furfural as an important compound that causes hydrolysate toxicity for S. cerevisiae. Additionally, they applied targeted evolution to a half industrial S. cerevisiae strain for 300 generations in the presence of furfural as the only inhibitor compound and also in a medium supplemented with hydrolysate. The resulting strains had significantly higher bioconversion properties than the parental strain with an ability to grow at hydrolysate concentrations that are lethal for the parental strain. The higher viability of the resistant strains during the prolonged lag phase induced by furfural was stated as the basis for their improved resistance (Heer & Sauer, 2008). A recent study with xylose-fermenting S. cerevisiae employed evolutionary engineering for improved ethanol production at high substrate loading, in different fed-batch SSF processes. The evolved strain was more tolerant to inhibitors with 65% and 20% improvement in xylose consumption and final ethanol concentrations, respectively (Tomas-Pejo et al., 2010).

High tolerance to ethanol is also desirable for yeast strains to be used in industrial ethanol fermentations. Thus, many reports exist in the literature on improvement in yeast ethanol tolerance by evolutionary approaches. An early example of continuous selection experiments in a chemostat to obtain ethanol-tolerant yeast mutants dates back to 1982 (Brown & Oliver, 1982). Since then, different studies have been reported to improve ethanol tolerance and investigate the complex genetic mechanism of this phenotype. In a recent study, for example, mutagenized and nonmutagenized S. cerevisiae populations were subjected to adaptive evolution in the presence of ethanol stress, and ethanol-tolerant mutants were obtained. Those mutants had higher growth rates than the wild type when grown in sublethal ethanol concentrations. They also survived better than the wild type in lethal ethanol concentrations. In the presence of ethanol, the mutants had higher glucose utilization rate and produced more glycerol than the wild type (Stanley et al., 2010a). Transcriptomic analysis of the mutants and the wild type showed that improved ethanol tolerance is related to increased mitochondrial and NADH oxidation activities, which also stimulates glycolysis (Stanley et al., 2010b). To improve glucose/ethanol tolerance of S. cerevisiae, particularly for very high gravity (VHG) fermentations in the ethanol industry, Alper et al. (2006) employed a genetic-driven selection approach called ‘global transcription machinery engineering (gTME)’. This approach is based on the alteration of key proteins that regulate the global transcriptome. Thus, diversity is obtained at the transcriptional level. The transcription factor Spt15p was mutated, and upon selection in the presence of high glucose and ethanol, dominant mutations conferring high glucose and ethanol tolerance were identified (Alper et al., 2006). Santos & Stephanopoulos (2008) also reviewed such combinatorial engineering methods based on recombinant DNA technology to improve cellular phenotypes. These methods involve synthetic promoter libraries, tunable intergenic regions, artificial transcription factor engineering, global transcription machinery engineering, ribosome engineering, genome shuffling, etc. Among these, a successful example of artificial transcription factor libraries in S. cerevisiae was described: zinc-finger motifs attached to effector domains were used to create these libraries. Upon expression in S. cerevisiae, yeast strains with improved tolerance to a variety of stress conditions were obtained (Park et al., 2003). Following technological advances, these combinatorial approaches are expected to become more common (Santos & Stephanopoulos, 2008).

In our research group, we have been working on applying evolutionary engineering strategies to improve resistance of S. cerevisiae to different stresses. Our ultimate aims are first to obtain generally more ‘robust’ strains and second to understand the complex genetic basis of resistance to one particular stress type, multiple-stress resistance, and the common genetic factors between resistance to different stress types, which can also be a good model for analyzing stress-related events in complex eukaryotes.

As S. cerevisiae cells are exposed to many different stress types during industrial bioprocesses, such as oxidative, freeze–thaw, high temperature, and ethanol, to have robust strains that are multistress resistant is highly desirable. To this end, we employed several batch and chemostat selection strategies in the presence of oxidative, freeze–thaw, high temperature, and ethanol stresses. The starting strain was a well-known laboratory strain (CEN.PK 113-7D) that was chemically mutagenized by EMS treatment prior to selection, to increase the genetic diversity of the initial population for selection. It is important to note that, unlike evolutionary engineering of xylose, arabinose, or lactose conversion to ethanol, no recombinant strain is necessary for evolutionary engineering of stress resistance. Utilization of an unusual substrate by S. cerevisiae requires transfer of genes of the related metabolic pathway from another organism that can utilize that substrate naturally. However, resistance to stress is a cellular property that can be improved simply by evolutionary engineering without transfer of foreign genes into S. cerevisiae, making the approach more suitable for food and beverage industry as a non-GMO application. Results of the parallel selection experiments for multistress-resistant S. cerevisiae showed that batch selection for freezing–thawing stress resistance was the best strategy to select multistress-resistant yeasts that are resistant not only to freezing–thawing stress but also to ethanol, oxidative, and high-temperature stresses (Çakar et al., 2005). This finding was also supported by Wei and coworkers, where they improved the multistress tolerance of an ethanologenic S. cerevisiae strain by freeze–thaw treatment (Wei et al., 2007). Additionally, the final resistant mutant populations were found to be heterogeneous with highly varying stress resistance levels, indicating the multigenic and complex nature of stress resistance (Çakar et al., 2005).

Heterogeneity of yeast populations is an interesting issue described by various research groups. Attfield et al. (2001) studied stress gene expression and stress resistance among individual S. cerevisiae cells using flow cytometry and fluorescence techniques. They observed heterogeneity in the strength of stress response in the clonal population tested and stated that this heterogeneity was mainly physiologically based, resulting mainly from asynchronous growth in batch cultures (Attfield et al., 2001). Phenotypic heterogeneity was also discussed as a potentially beneficial property to confer populations the advantage of persisting during perturbations. For this purpose, survival of stress-sensitive S. cerevisiae mutants was tested and compared with that of the wild type. It was shown that the mutants had increased heterogeneity, which confers benefits under high-stress conditions by increasing occasional cell survival (Bishop et al., 2007). Similarly, the high heterogeneity of our final evolved populations may also provide a survival advantage at high-stress levels. Other recent reports focus on comparative genomic analysis of wild-type yeast strains to identify the major type of genome variability that confers genetic diversity to natural yeast populations (Carreto et al., 2008), the development of single-cell analysis methods to detect cell-to-cell variability upon stress exposure (Tibayrenc et al., 2011), and the contribution of positive and negative epistasis to adaptation and reproductive isolation of S. cerevisiae in low-glucose and high-salt environments (Parreiras et al., 2011). The high level of heterogeneity of the final evolving populations in our studies might also have resulted from the negative epistasis of different mutations, which individually confer an increased fitness in those evolving populations.

Apart from commonly encountered stress types in yeast bioprocesses, our group also focuses on metal stress and evolutionary engineering of metal-resistant S. cerevisiae. Particularly, metals with widespread industrial use and less-known biological function and/or toxicity mechanism are preferred. One such example is the evolutionary engineering of cobalt-resistant S. cerevisiae (Çakar et al., 2009). In that work, selection was performed under batch serial cultivation conditions with cobalt stress applied continuously or as a pulse, at constant or gradually increasing levels. The best selection strategy to yield the most cobalt-resistant mutant was determined as batch selection under gradually increasing, continuously applied cobalt stress. Upon 25 repeated batch cultivations, the final population of that selection resisted up to 8 mmole l−1 CoCl2. Similar to our previous study on multistress resistance (Çakar et al., 2005), cobalt resistance of the final mutant population was again highly heterogeneous, ranging between threefold and 3700-fold of the wild-type resistance level. The intracellular cobalt contents of the mutants were determined by flame atomic absorption spectrophotometry and showed that the cobalt hyper-resistant mutants had 2–4 times lower intracellular cobalt contents than the wild type, suggesting a resistance mechanism that prevents cobalt uptake. Additionally, they were found to have significant cross-resistance to other metals such as nickel, iron, zinc, and manganese. The mutants evolved under pulse stress exposure, however, were cross-resistant to heat shock and hydrogen peroxide stress, implying that the resistance mechanisms to continuously applied and pulse ethanol stresses are complex and may involve different molecular factors (Çakar et al., 2009). Comparative genetic and genomic analyses of the cobalt hyper-resistant mutant and the wild type are in progress to identify the related molecular resistance mechanism to cobalt, along with evolutionary engineering studies with other possibly related metals such as nickel (Küçükgöze et al., 2011) and iron.

The physiological characterization of the evolved yeast strains after evolutionary engineering should be performed carefully. This is because of the observation that evolutionary engineering to improve a particular cellular property might lead to trade-off situations between different characteristics. Wenger et al. (2011) recently reported this issue with a particular focus on ‘hunger artists’, describing yeasts that show trade-off situations under carbon sufficiency when adapted to carbon limitation before. They showed that S. cerevisiae evolved under aerobic, glucose-limited conditions had only very few trade-offs, i.e. reduction in other traits, in other carbon-limited conditions. However, they performed less well in aerobic, carbon-rich conditions. Thus, the cost of their fitness at resource-limiting conditions was paid when the same resources were nonlimiting. Whole-genome sequencing of the mutants revealed that the mutations were in genes responsible for glucose sensing, signaling, and transport, which explains their poor performance when glucose was abundant (Wenger et al., 2011). A similar observation was made before with maltose-limited S. cerevisiae cultures. Prolonged chemostat cultivations of S. cerevisiae under maltose limitation revealed mutants with high maltose affinity. Additionally, the mutants had maltose hypersensitivity under excess maltose conditions. Exposure to excess maltose resulted in unlimited uptake in maltose-adapted cells because of the accumulative properties of S. cerevisiae maltose–proton symport system (Jansen et al., 2004).

Another example of trade-off situations in evolutionary engineering studies involves impaired glucose growth of S. cerevisiae mutants evolved for anaerobic growth on xylose. These mutants could simultaneously utilize both xylose and glucose under batch cultivation conditions but had impaired growth on glucose (Sonderegger & Sauer, 2003).

Trade-off situations also occurred when evolving S. cerevisiae for a particular stress resistance. When S. cerevisiae was evolved for cobalt resistance under continuously applied cobalt stress conditions, the resulting mutants had sensitivity to some other stress types, such as ethanol and hydrogen peroxide (Çakar et al., 2009).

As trade-off situations are not rare in evolutionary engineering studies and the ‘cost of adaptation’ is an important issue in evolutionary biology (Wenger et al., 2011), the evolved mutants should be characterized in detail, based on their physiology and genetics, to determine whether they have any important traits with reduced function, as the ‘cost’ of their evolved trait(s).

Emerging technologies and new, high-throughput analytical instruments in biological research are expected to change evolutionary engineering studies and experiments tremendously. Recently, microfluidics has become a hot topic in biological research. It is an analytical system that allows processing and manipulation of fluids in small amounts (Vinuselvi et al., 2011). The advantages of this system are low cost and labor, high resolution and precision, and high-throughput continuous and batch processing of multiple samples. Microfluidic bioreactors are microfluidic devices that can be used for evolutionary adaptation during long-term cultivation, microbial growth, and cell quantification (Vinuselvi et al., 2011). Examples include a microfluidic chemostat for bacterial and yeast cell cultivations (Groisman et al., 2005); the ‘Envirostat’, a microfluidic bioreactor that allows constant environmental conditions (Kortmann et al., 2009); and the bacteria–algae–yeast (BAY) microbioreactor without valves, mixers, and pumps, powered by digital microfluidics, where a fluorescent viability assay and the use of a fluorescent reporter gene allowed measurements of cell growth and density (Au et al., 2011). A diffusion-based microreactor system for continuous cultivation was also described (Edlich et al., 2010). It has the advantage of integrating online measurement technique for dissolved oxygen (DO) and optical density (OD). The oxygen supply was monitored online by integrated DO sensors, on the basis of a fluorescent dye complex (Edlich et al., 2010). An emerging microscopic technique based on coherent anti-Stokes Raman scattering allowed quantitative glucose imaging at the single-cell level, allowing monitoring glucose fluxes in microfluidic reactors (Akeson et al., 2010). Such developments in microfluidics will most likely change yeast cultivations and evolutionary engineering selection experiments in batch and chemostat conditions tremendously.


As an inverse metabolic engineering approach, evolutionary engineering can be applied easily to improve diverse and complex microbial properties, such as utilization of different, industrially suitable substrates, formation of valuable products, and improvement in stress resistance. Additionally, the recent developments in functional genomics methodologies allow analysis and identification of the genetic basis of the desired phenotype. Depending on the cellular property to be modified, combined approaches such as rational metabolic engineering for the transfer of a nonexisting pathway, followed by extensive evolutionary engineering of the resulting recombinant strain to improve process efficiency, are also used. Additionally, computational algorithms can be used to study and analyze adaptive evolution processes (Hua et al., 2006; Sorribas et al., 2010). To conclude, evolutionary engineering is a powerful approach with widespread use in improving industrially important and genetically complex properties of S. cerevisiae and other microorganisms. It is expected that emerging high-throughput technologies in biological research, such as microfluidics, will change evolutionary engineering experiments by decreasing their costs and increasing their precision and allowing high-throughput continuous and batch processing of multiple samples. Additionally, development in high-throughput screening and analytical technologies will further increase the utilization frequencies of evolutionary engineering and genetic-driven selection strategies, which are both based on screening of a large number of samples for a desirable trait.


Our research presented in this mini review was supported by Turkish State Planning Organization (DPT), Turkish Scientific and Technological Research Council (TÜBİTAK) (project no: 105T314, 107T284, 109T638, PI: Z.P.Ç.), Istanbul Technical University (ITU) Research Funds (project no: 30108, PI: Z.P.Ç.), ITU-Institute of Science and Technology-Research Funds (project no: 33237 to Z.P.Ç. and B.T.-Y., project no: 34200 to Z.P.Ç. and C.A., project no: 33567 to Z.P.Ç. and Ü.Y.), and FEMS Research Fellowship (2009-2, to B.T.-Y.). We wish to thank Can Holyavkin, Nazlı Kocaefe, and Arman Akşit for scientific contribution and technical assistance and other former and present ITU students U.Ö.Ş. Şeker, T. Sezgin, M. Şen, G. Küçükgöze, B.G. Balaban, Ş.H. Tekarslan, Ç. Karabulut, O. Ercan, B. Yılmaz, A. Altıntaş, S. Mutlu, M. M. Hız, F.Ş. Küçük, N.S. Alikişioğlu, G. A. Akçeoğlu, A.K. Korkmaz, and S. Erbil for their contributions to the work done in our group. We are also grateful to Uwe Sauer, Jean-Marie François, Laurent Benbadis, Süleyman Akman, Candan Tamerler, Mehmet Sarıkaya, and Bülent Balta for their valuable scientific contributions to our work presented in this review.