Craig MacLean, NERC Center for Population Biology, Imperial College London, Silwood Park Campus, Ascot SL5 7PY, UK. Tel.: +44 (0)20 7594 2475; fax: +44 (0)13 4487 3173; e-mail: firstname.lastname@example.org
Traits that do not contribute to fitness are expected to be lost during the course of evolution, either as a result of selection or drift. The Leloir pathway of galactose metabolism (GAL) is an extensively studied metabolic pathway that degenerated on at least three independent occasions during the evolutionary diversification of yeasts, suggesting that the pathway is costly to maintain in environments that lack galactose. Here I test this hypothesis by competing GAL pathway deletion mutants of Saccharomyces cerevisiae against an isogenic strain with an intact GAL pathway under conditions where expression of the pathway is normally induced, repressed, or uninduced. These experiments do not support the hypothesis that pleiotropy drives GAL pathway degeneration, because mutations that knock out individual GAL genes do not tend to increase fitness in the absence of galactose. At a molecular level, this result can be explained by the fact that yeast uses inexpensive regulatory proteins to tightly regulate the expression of structural genes that are costly to express. I argue that these results have general relevance for our understanding of the fitness consequences of gene disruption in yeast.
Galactose metabolism in yeast is controlled by the GAL pathway, which is made up of structural genes that convert galactose into glucose-1-phosphate and regulatory genes that modulate the expression of the pathway in response to galactose availability (Lohr et al., 1995; Ideker et al., 2001; Dickinson & Schweizer, 2004). The classical GAL pathway is as follows: a hexose transporter (Gal2p) transports galactose into the cytoplasm where it is converted to glucose-1-phosphate by the sequential action of galactokinase (Gal1p), uridylyltransferase (Gal7p) and epimerase (Gal10p). Gal4p is a DNA-binding protein that strongly activates the expression of the structural genes, provided that it is not bound to Gal80p, the pathway repressor, which shuttles between the cytoplasm and the nucleus. The presence of galactose activates Galp3p, the signal transducer, which then binds to Gal80p in the cytoplasm releasing Gal4p from inhibition. A schematic of the GAL pathway is shown in Fig. 1. The expression of the GAL structural genes is also modulated by Gal80p independent mechanisms of catabolite repression that actively repress the expression of GAL structural genes when glucose is present in the environment (Johnston et al., 1994; Gancedo, 1998).
The ability to metabolize galactose was lost on at least three separate occasions during the evolutionary diversification of yeasts, and in each case this was caused by the degeneration of the GAL pathway (Hittinger et al., 2004). The well-known molecular biology of GAL (Lohr et al., 1995) combined with the availability of genome-wide data on changes in gene expression and protein abundance in response to systematic environmental and genetic perturbations of GAL (Ideker et al., 2001) make the pathway an attractive system with which to link the fitness consequences of gene loss with fundamental aspects of biochemistry. Here I test the hypothesis that pleiotropy promotes the degeneration of GAL by measuring the competitive fitness of GAL pathway single gene deletion mutants against an isogenic GAL+ strain in galactose-limited cultures, where expression of the pathway is induced, glucose-limited cultures, where the expression of the pathway is repressed, and glycerol-limited cultures, where the pathway is neither induced or repressed.
Materials and methods
Strains used in this study were derived from the haploid strain BY4741 (ATCC#201388, Mat a his3Δ1 leu2Δ0 met15Δ0 ura3Δ0). Mutants gal1Δ, gal2Δ, gal3Δ, gal4Δ, gal7Δ, gal10Δ, gal80Δ and ura4Δ were constructed by replacing the relevant ORF with the KanMX module and a unique oligonucleotide barcode as part of the Saccharomyces Genome Deletion Project (Giaever et al., 2002).
Prior to competition, each strain was grown up overnight at 30 °C in 2×YPAD with continuous shaking. One millilitre of overnight culture of each galΔ mutant and the ura4Δ mutant were then mixed together to form a ‘common pool’ containing all seven GAL deletion mutants and the ura4Δ tester. One hundred microlitres of the common pool was inoculated into flasks containing 20 mL of supplemented minimal medium (Yeast Nitrogen Base, 1.7 g L−1, Ammonium Sulphate 5 g L−1, uracil 20 mg L−1, histidine 20 mg L−1, leucine 100 mg L−1, methionine 20 mg L−1) containing either 2% Glucose, 2% Galactose, or 3% Glycerol. Each competition was repeated three times (glucose, galactose) or four times (glycerol). Genomic DNA was extracted from the remainder of common pool immediately after inoculation into competition flasks using a Wizard kit (Promega, Southampton, UK). Competition flasks were incubated at 30 °C with constant shaking (125 rpm). Cells were harvested after 2 days of incubation (glucose and galactose competitions) or 5 days of incubation (glycerol competitions) and genomic DNA was immediately extracted. Preliminary experiments established that the ura4 mutation is selectively neutral in all three media used in our experiments, implying that this strain is an appropriate tester strain with an intact GAL pathway.
Quantitative PCR fitness assay
Quantitative PCR was used to assay the change in the relative frequency of gal deletion mutants and ura4Δ during competition. In each deletion mutant strain the relevant GAL gene was replaced by a deletion module consisting of a KanMX module flanked by unique 5′ and 3′ 20-mer ‘barcode’ sequences unique to each deletion strain (Giaever et al., 2002). The strategy to quantify the abundance of each strain in the mix was to amplify DNA a using a common reverse primer in the KanMX module (KanB) and a series of unique forward primers consisting of the unique 5′ barcode of each strain (Table 1). The change in the abundance of each strain in each competition was calculated by determining the concentration of the DNA sequence unique to each strain before and after competition by comparison with a standard curve of DNA concentration vs. CT for each strain. A schematic of this protocol is shown in Fig. 2. Fitness was estimated as log w, the change in the log ratio of galΔ/ura4Δ, such that a value of 0 represents equal fitness. This protocol worked for every gal deletion mutant except gal10Δ, which gave an irregular amplification profile.
Table 1. Sequences of oligonucleotides used in this study.
Sequence (5′ to 3′)
ATA GAA GTA TAA GTG CGC CGC G
TTG CCA GAG AGG AAA CGC C
GCC GAC GTT AAA TAC ACT CAC G
CCT ATA CTG AAA CTG ACG CA
CCG ACT AGC AGA TAG CCG C
ATA GTG CCA GAC CGC CGC
GCC ACC ATA AAT GAG TCC ACG
TAC GGT AGT CGA TAC TCT GTC G
CTG CAG CGA GGA GCC GTA AT
gal10Δ fitness assay
The competition experiment was repeated as per the above protocol, except that only gal10Δ and BY4741 were competed against each other. Changes in the density of each strain were measured by plating out the common pool and the competition cultures on YPD, which allows the growth of both gal10Δ and BY4741, or on YPD supplemented with geneticin, which selects for the gal10Δ strain. Fitness was estimated as log w, the change in the log ratio of gal10Δ/BY4741, such that a value of 0 represents equal fitness.
Fitness consequences of GAL disruption in the presence of galactose
The entry of galactose into the cell relieves the repression of gal4p by gal80p, resulting a massive (>1000-fold) increase in the expression of GAL structural genes (Lohr et al., 1995). As a proof-of-principle experiment, all GAL pathway single gene deletion mutants were competed against an isogenic GAL+ tester strain in galactose-limited batch cultures (Fig. 3a).
The purpose of this experiment is twofold. First, we have a clear expectation that the deletion of the GAL structural genes and genes that are required for activating the expression of the GAL pathway (GAL3 and GAL4) will result in a fitness loss in a galactose-limited environment. Secondly, data from this assay can be compared with previously published growth rates of pure cultures of gal deletion mutants in galactose-limited medium to ask if assaying the fitness of mutants in pure and mixed cultures give equivalent results.
Consistent with our expectations, disruption of GAL genes tends to reduce fitness when galactose is present as sole carbon source (mean log w = −0.57, SEM = .29, t = 2, P = 0.048), although there is considerable fitness variation among mutants (F6,14 = 235, P < .001). Competitive fitness data from my assays are in excellent agreement with pure-culture doubling rates in galactose (Ideker et al., 2001): all galΔ mutants, except gal80Δ, have increased doubling times and the rank order correlation between these two measures of fitness is very strong (Spearman's r = −.79, F1,5 = 8.06, P =.036). The only exception to this trend is the gal1Δ mutant, which has a lower competitive fitness than what would be expected based on its doubling time alone. If this mutant is excluded, the rank order correlation between these two measures of fitness is perfect (Spearman's r = −1). This result continues to hold when the gal10Δ mutant, whose fitness was assayed in a slightly different manner, is excluded from the above analysis (Spearman's r = −1).
Fitness consequences of disrupting a repressed GAL pathway
The presence of glucose in Saccharomyces cerevisiae cells triggers multiple independent mechanisms of repression that prevent expression of GAL structural genes, even if galactose is also available (Johnston et al., 1994; Gancedo, 1998). One potential adaptive explanation for GAL degeneration is that loss of the pathway may increase fitness under conditions where GAL expression is actively repressed. Contrary to this prediction, the average fitness of galΔ deletion mutants in glucose-limited cultures does not differ significantly from zero (Fig. 3b; mean log w =−0.109, SEM = 0.077, t6 = 1.4, P = 0.11), although there is variance in fitness among mutants (F6,14 =14.9, P < 0.001).
Fitness consequence of GAL pathway disruption under noninducing conditions
In the absence of a specific inducer, the GAL regulatory genes are constitutively expressed at low levels, resulting in cells with an inactive GAL pathway that are poised for rapid induction of the pathway if galactose becomes available (Lohr et al., 1995). One possible explanation for the degeneration of GAL is that the pathway is costly to maintain in an uninduced state. To test this hypothesis I assayed the fitness of galΔ deletion mutants in glycerol-limited cultures. On average, GAL pathway deletion mutations do not increase fitness in glycerol-limited cultures (Fig. 3c: mean log w = −0.28, SEM = 0.2 t6 =1.48, P = 0.18), although there is some variation in fitness among mutants (F6,18 = 5.07, P = 0.0033).
The degeneration of unused traits is an important feature of evolution that can be addressed using both population and molecular genetics. Previous microbial experiments have provided evidence that unused metabolic pathways can degenerate as a result of selection due to antagonistic pleiotropy (Cooper & Lenski, 2000; MacLean et al., 2004) or as a result of neutral drift (MacLean & Bell, 2002; Collins & Bell, 2004), but these experiments have not investigated the genetic causes of trait degeneration. The well-documented transcriptomic and proteomic changes that occur in response to GAL pathway disruption combined with multiple independent instances of GAL degeneration in natural populations make this pathway an ideal system to link the population genetic and molecular genetic causes of trait degeneration. My results clearly show that there is no general tendency forGAL pathway deletion mutations to increase fitness in the absence of galactose (Fig. 3b,c), suggesting that antagonistic pleiotropy was not the cause of GAL pathway degeneration during the evolutionary diversification of yeasts.
Fitness of GAL pathway mutants in the presence of galactose
As a proof-of-principle experiment I assayed the fitness of GAL pathway deletion mutants in galactose-limited cultures. The results of this assay are in excellent agreement with previous pure-culture fitness estimates of these mutants under this set of conditions, which validates the mixed-culture approach to assaying fitness used in these experiments.
Although the mean effect of GAL mutations is to decrease fitness in the presence of galactose, there is considerable variation in fitness among mutants (Fig. 3a). To what extent can this variation be explained in terms of the well-characterized biology of this pathway? The simplest genetic manipulation of GAL is to prevent induction of the pathway by disrupting GAL3 or GAL4. As we would expect, gal3Δ and gal4Δ deletions result in equivalent fitness losses in galactose-limited medium (Fig. 3a). Disruption of GAL structural genes often reduces fitness more than regulatory mutations that prevent pathway expression. This can be explained by considering the effects of structural gene mutations on pathway expression and function. GAL structural genes are expressed at low levels in response to induction by galactose in gal7Δ and gal10Δ strains (Ideker et al., 2001). This imposes a large cost because the expression of a nonfunctional GAL pathway that lacks either uridylyltransferase (gal7p) or epimerase (gal10p) results in the intracellular accumulation of toxic galactose-1-phosphate (Lai & Elsas, 2000). The gal1 mutation results in a large cost because this mutation does not reduce the expression of other GAL structural genes (Ideker et al., 2001), so that the cell makes a large investment in producing a nonfunctional GAL pathway. The comparatively mild effects of the gal2Δ mutation are not surprising given that deleting GAL2 results in an increase in the expression of functionally redundant hexose transporters that can transport galactose into the cell (Ideker et al., 2001).
Fitness of GAL pathway mutants in the absence of galactose
To test the hypothesis that carrying an intact GAL pathway is inherently costly, I assayed the fitness of GAL pathway deletion mutants in glucose and glycerol-limited cultures. My justification for the use of these assay environments is as follows. Glucose is the preferred carbon source of yeast: S. cerevisiae has a number of adaptations for ensuring rapid utilization of available glucose and for repressing the metabolism of other substrates in the presence of glucose (Gancedo, 1998; Dickinson & Schweizer, 2004). Glycerol, on the other hand, is a poor carbon substrate for yeast and the expression of genes for catabolizing alternative substrates is not actively repressed in the presence of glycerol. This makes glycerol-limited cultures an ideal environment for testing the cost of carrying an unused GAL pathway. Deleting GAL pathway genes does not tend to increase fitness in either of these environments (Fig. 3b,c), demonstrating that carrying a repressed or un-induced GAL pathway does not incur a fitness cost. Although it might be argued that testing the fitness of deletion mutants in only two environments is not a sufficient test for pleiotropy, the GAL pathway can only exist in two different states in the absence of galactose, uninduced and repressed, and both of these states can be generated by competition in glucose and glycerol limited cultures.
Can the absence of a fitness cost associated with carrying an unused GAL pathway be explained in terms of molecular biology? The expression of the GAL pathway is so efficiently regulated in cells with an intact GAL pathway that GAL pathway structural proteins are essentially undetectable in cells grown in the absence of a GAL pathway inducer or in the presence of a GAL pathway repressor (Lohr et al., 1995; Ideker et al., 2001), as occurs in glycerol-limited cultures (Lohr et al., 1995). The benefit of such tight regulation is clear: deleting GAL80 results in constitutive expression of the GAL structural genes in glycerol-limited cultures (Lohr et al., 1995) and this is associated with a large fitness cost (Fig. 3c, log w = −1.05, SE = 0.18, t3 = 6, P = 0.009). Moreover, the direct cost of such regulation is minimal: GAL4 and GAL80 are constitutively expressed at low levels in the presence of glucose (Lohr et al., 1995), but deleting these genes does not result in a significant increase in fitness under these conditions (gal4Δ: log w = −0.19, SE = 0.034, t25.8, P = .03; gal80Δ: log w = .21, SE = 0.07 t2 = 2.97, P = 0.098). This is the appropriate measure of the costs of regulating GAL expression in the absence of galactose because the expression of GAL genes is so strongly repressed in the presence of glucose that any fitness benefit of deleting regulatory genes under these conditions can be attributed to the cost of expressing the regulatory genes themselves.
I argue that by tightly regulating the expression of a pathway that is costly to express with cheap regulatory proteins, yeast can maintain an unused GAL pathway without incurring any direct cost. The tight pattern of regulation of gene expression that is exemplified by GAL is typical of other yeast metabolic pathways (Dickinson & Schweizer, 2004) and our results may help to explain why so few yeast gene deletions increase fitness under laboratory conditions (Sliwa & Korona, 2005). Bacteria possess analogous mechanisms for regulating the expression of catabolic operons, and our results may help to explain why many studies have failed to find evidence for the loss of unused metabolic pathways as a result of natural selection (Korona, 1996; Travisano, 1997; Riley et al., 2001; MacLean & Bell, 2002) More generally, it is possible that tight regulation of expression is a general property of extant metabolic pathways precisely because only tightly regulated pathways that can be carried without cost can escape degeneration over evolutionary time scales.
The author would like to thank C. Brandon for technical assistance, G. Bell for insightful discussion, and anonymous reviewers for their helpful comments. Funding was provided by grants from NERC (UK) to the Centre for Population Biology.