Evolution of mutational robustness in the yeast genome: a link to essential genes and meiotic recombination hotspots

Abstract

Deleterious mutations inevitably emerge in any evolutionary process and are speculated to decisively influence the structure of the genome. Meiosis, which is thought to play a major role in handling mutations on the population level, recombines chromosomes via non-randomly distributed hot spots for meiotic recombination. In many genomes, various types of genetic elements are distributed in patterns that are currently not well understood. In particular, important (essential) genes are arranged in clusters, which often cannot be explained by a functional relationship of the involved genes. Here we show by computer simulation that essential gene (EG) clustering provides a fitness benefit in handling deleterious mutations in sexual populations with variable levels of inbreeding and outbreeding. We find that recessive lethal mutations enforce a selective pressure towards clustered genome architectures. Our simulations correctly predict (i) the evolution of non-random distributions of meiotic crossovers, (ii) the genome-wide anti-correlation of meiotic crossovers and EG clustering, (iii) the evolution of EG enrichment in pericentromeric regions and (iv) the associated absence of meiotic crossovers (cold centromeres). Our results furthermore predict optimal crossover rates for yeast chromosomes, which match the experimentally determined rates. Using a Saccharomyces cerevisiae conditional mutator strain, we show that haploid lethal phenotypes result predominantly from mutation of single loci and generally do not impair mating, which leads to an accumulation of mutational load following meiosis and mating. We hypothesize that purging of deleterious mutations in essential genes constitutes an important factor driving meiotic crossover. Therefore, the increased robustness of populations to deleterious mutations, which arises from clustered genome architectures, may provide a significant selective force shaping crossover distribution. Our analysis reveals a new aspect of the evolution of genome architectures that complements insights about molecular constraints, such as the interference of pericentromeric crossovers with chromosome segregation.

Figure 1. Conceptual overview of the methods and main results of this study.

We developed a computer simulation (S. digitalis) to test the hypothesis of a feedback between sexual processes (meiosis and mating) and genome architecture that would facilitate purging of lethal mutations in populations of unicellular organisms (such as baker's yeast, S. cerevisiae). The implementation of the simulation is based on population genetic concepts and on our results from mutation-accumulation experiments with an Msh2 mutator strain (see main text). Experiments performed with digitally implemented populations of yeast genomes revealed a competitive advantage of yeast-alike chromosome architectures and enabled the evolution of genomes with yeast-alike fitness. These investigations provide reasoning for several hallmarks of yeast genome evolution and for various parameters that describe the population genetics of yeasts.

Figure 2. Yeast life cycles and mutation homeostasis.

(A) The key elements of natural yeast life cycles. Life cycles in natural yeast populations are regulated by changes in the environment, which either permit mitotic growth or induce the formation of spores via meiosis. All S. cerevisiae isolates from nature were found to be diploid, indicating a return to a diploid lifestyle immediately following germination. This may occur by intra-tetrad spore-mating (inbreeding/automixis), or by mating of spores from different tetrads (outbreeding/amphimixis). Optionally, mating type switching followed by mother-daughter mating may generate diploid cell lines (see corresponding sections in Results and Discussion). Predominant diploid life cycles can also be found for many other species throughout the yeasts . (B) Loss of individuals from a population occurs either at random or due to a decrease in fitness caused by mutations. The simulation implements populations that are cycling through vegetative and sexual stages of the life cycle (mitosis, meiosis and mating). The genomes are constantly exposed to mutations. A homozygotisation of recessive lethal mutations (mutations that inactivate essential genes) leads to the death of an individual. Alternatively, individuals may be removed at random from the population due to limitations in the nutritional supply (starvation). In a dynamic equilibrium, in which the average population growth equals zero, the influx of new recessive lethal mutations is equal to the outflux associated with the death or removal of individuals. For more details on the fitness impact of mutations, see main text and Supplementary Figure 1A and 1B in Text S1.

Figure 3. Digital genomes and the life cycle of digital populations.

(A) Schematic illustration of the genetic building blocks of the digital genomes (top): essential genes (wild type, white squares; and mutated, crossed squares); non-essential genes (black squares); intergenic elements (circles) are either crossing over proficient (recombination hotspots, white circles) or silent (crossed circles). Unique identifiers represent essential genes. In the simulation, each individual possesses two copies of one single chromosome. Such a diploid genome is considered viable if it contains at least one functional copy of each individual essential gene. The simulation framework implements entire populations of individuals with diploid genomes, which are arranged in a population matrix (bottom). Individuals with different chromosome architectures belong to different species. (B) The life cycle of digital populations. See main text for details. (C) Crossover interference is implemented using an Erlang probability density function with a shape factor k = 4 (F. Stahl, personal communication, and references [46],[92],[93]). The function is based on a genetic distance definition (hotspot density). Rescaling of the Erlang function along the distance axis allows adjusting the crossover frequency per chromosome. (D) Breeding can occur between haploid gametes from the same meiosis (intratetrad mating/inbreeding) or between gametes from different meioses (outbreeding). Mating types are optional. See main text for details.

Figure 4. Competitive fitness advantage of yeast chromosomes in the presence of lethal mutations.

(A) Mutational robustness R _max of digital populations with random chromosomes or with S. cerevisiae chromosome IX, respectively, at different inbreeding fractions. 3,000 experiments were performed per chromosome configuration. A different random architecture composed of the same total number of meiotic recombination hotspots and essential genes was generated for each reference experiment. A histogram summarizing these results is provided as Supplementary Figure 2A in Text S1. The digitized chromosome IX architecture is based on essential gene data obtained from www.yeastgenome.org as well as on data on the genome-wide distribution of meiotic recombination double strand break sites obtained from . The representation in the inset shows meiotic recombination hotspot (HS) and essential gene (EG) densities of chromosome IX in a sliding window analysis (see Text S1). The population size cap was set to 200 individuals. (B) Results of survival competition experiments for S. cerevisiae chromosome IX versus randomly generated chromosome architectures (n = 10 experiments per grid point), for the entire inbreeding/outbreeding domain and for high mutation rates up to R_max. The level of dominance is color-encoded (bright red: chromosome IX wins in all experiments, bright blue: random architecture wins in all experiments). The population size cap was set to 2,000 individuals. (C) Results of survival competition experiments for the architectures described in (B), for a wide range of mutation rates (in logarithmic steps, R between 10⁻⁴ and 1) and for different population size caps (150, 1,500 and 15,000). The competitive advantage (“comp. adv.”) indicates the percentage of chromosome IX wins over random architectures. This percentage is formed as an average over the entire inbreeding/outbreeding domain (bars indicate SD). (D) Results of survival competition experiments of the sixteen yeast chromosomes versus randomly generated chromosome architectures (n = 170 experiments per bar; population size cap: 10,000), averaged over the entire inbreeding/outbreeding domain. Mutation rates are color-encoded; green: 10⁻², blue: 10⁻¹, red: 1. A competitive advantage of 100% indicates that the yeast chromosome architecture always outperformed the random architectures. A value of 50% indicates that the yeast chromosome exhibited a performance identical to that of random architectures. For simulation details and statistical information see Text S1, section “S. digitalis Simulation Settings.”

Figure 5. Effect of essential gene clustering on mutational robustness.

(A) Mutational robustness R _max as a function of the inbreeding ratio (n = 3 per inbreeding ratio; error bars indicate SD), for a clustered model chromosome (seven EG clusters separated by regions containing meiotic recombination hotspots) and for random chromosomes. Each experiment was performed with a different randomly generated chromosome containing the same number of elements (400 NEGs, 100 EGs, 166 recombination hotspots). The inset shows the average increase of R_max and its SD over the inbreeding/outbreeding domain in response to the introduction of a MAT locus. The red and blue dots on the R_max-axis indicate the average over all inbreeding fractions (left: without MAT, right: with MAT). In clustered model chromosomes, the MAT locus was placed in the first EG cluster; in random chromosomes, the MAT locus was placed in the centre of the chromosome. The population size cap was set to 200 individuals. (B) (average of the entire inbreeding/outbreeding domain, bars indicate SD) as a function of the level of EG clustering (ranging from perfect clustering (1) with all EGs joined in one single continuous cluster, to a maximally unclustered architecture (100) with each pair of EGs separated by at least one meiotic recombination hotspot). The population size cap is color-encoded (ranging from 50 to 1,600 individuals). Simulation time was conservatively assigned; a further increase in the maximum number of generations did not change the simulation outcome significantly. (C) Results of survival competition experiments for model chromosomes with one MAT-linked cluster and five clusters in the chromosomal arm region (CF1+5) versus model chromosomes with one MAT-linked cluster and a random EG distribution in the arm region (CF1+R), for the entire inbreeding/outbreeding domain and high mutation rates up to R_max (n = 10 experiments per grid point). The competitive advantage is color-encoded (bright red: CF1+5 won all competitions, bright blue: CF1+R won all competitions). The population size cap was set to 1,000 individuals. For simulation details and statistical information see Text S1, section “S. digitalis Simulation Settings.”

Figure 6. Fitness advantage of achiasmate meiosis.

(A) Results of survival competition experiments for random chromosomes versus achiasmate chromosomes (super-cluster architectures) with all EGs in one cluster not linked to (top panel) or linked to the MAT locus (lower panel). The matrix shows the color-encoded level of dominance for the entire inbreeding/outbreeding domain and for high mutation rates up to R_max. The population size cap was set to 2,000 individuals. (B) Results of survival competition experiments for achiasmate chromosomes versus random chromosomes, for a wide range of mutation rates (in logarithmic steps, R between 10⁻⁴ and 1) and for different population sizes (150, 1,500, and 15,000). The MAT locus was placed on the same chromosome. The competitive advantage indicates the percentage of achiasmate chromosome wins over random architectures. This percentage is formed as an average for all inbreeding/outbreeding ratios (bars indicate SD across the inbreeding/outbreeding domain). (C) As for (B), but with the MAT locus positioned next to the centromere on a different chromosome in the genome. (D) Autosomal chromosomes are linked to the MAT locus via their centromeres in the case of intratetrad mating. The cartoon illustrates a simple scenario with one autosomal chromosome and the sex chromosome. The autosomal chromosome contains a heterozygous mutation (1). The mutation may be linked to the centromere; this is always the case for achiasmate meiosis. Also the MAT locus may be linked to the centromere; this is the case for S. cerevisiae and for achiasmate meiosis. If both loci are linked to the corresponding centromere, intratetrad mating (from (4) to (5)) will involve chromatids that are each derived from one of the homologous chromosomes that are segregated in meiosis I (from (2) to (3)). This will reconstitute the heterozygous situation of all sites in the genome that fulfill this criterion. In population genetic terminology, this type of mating is called intratetrad mating with first division restitution. It has been demonstrated to occur in many species, including some fern, flies and fungi and may have constituted a driving force for the evolution of sex chromosomes. For simulation details and statistical information see Text S1, section “S. digitalis Simulation Settings.”

Figure 7. Maintenance and de novo evolution of essential gene clustering.

(A) Maintenance of EG clustering in inbreeding populations. Initially, EGs and meiotic recombination hotspots were arranged in two separate clusters. Chromosomes were allowed to rearrange by positional swapping of genes and meiotic recombination hotspots during an evolution period of 150,000 generations. The level of essential gene clustering was determined at the end of each experiment using the clustering score analysis (as described in Text S1). Color scheme: green indicates significant clustering after 150,000 generations. A high clustering score indicates a high level of essential gene clustering in regions of low meiotic recombination hotspot density. (B) Time-course of the clustering value in four different regimes in (A). The evolution of populations 1 and 2 is visualized side by side in Video S1. (C) Schematic illustration of a small model genome with five essential genes, which we designed to study the mechanisms involved in the evolution of clustering (see (D)). The clustering scores (red) of all possible architectures (green) are shown below the scheme. Symbols: light box = EG; black box = NEG; circle = meiotic recombination hotspot; circle with cross = meiotic recombination coldspot. (D) Distributions of clustering scores in inbreeding populations evolved from the initial genome shown in (C). The simulation was performed in the presence of a MAT locus (green line) as well as without a MAT locus (red line) (n = 1,000 experiments for each configuration). The distribution of clustering scores for random architectures is shown in blue and serves as a reference (n = 2·10⁶). In the presence of a MAT locus, clustering evolved over a broad range of rearrangement rates (Supplementary Figure 4 in Text S1). For simulation details and statistical information see Text S1, section “S. digitalis Simulation Settings.”

Figure 8. Evolution of yeast-like essential gene clustering.

(A) Evolution of chromosome architectures with a yeast chromosome IX-like size and genetic content. Beginning with completely unclustered chromosomes, populations with and without MAT were allowed to evolve at different mutation rates R as indicated (n _red = 830, n _green = 826, n _blue = 839 experiments). Clustering was scored by measuring both the size of the largest essential gene cluster and the average size of the remaining clusters. For genomes containing a MAT, the largest cluster was always observed to be linked to the MAT. The scatter plot shows the scores obtained after an evolution period of 100,000 generations. Score distributions for the different populations are spanned along the axes (including reference distributions for randomly generated populations of the same size). The “clustering score” ν quantifies the level of essential gene clustering. It is defined as the sum of squared sizes of all essential gene clusters (maximum-sized groups of essential genes not disrupted by a hotspot), normalized by the total number of essential genes and meiotic recombination hotspots. Random architectures with a chromosome IX genetic content score ν = 107±17 (SD), whereas the chromosome IX architecture itself scores ν = 224. The percentages indicate the fraction of experiments yielding genomes with a clustering value ν at least 2σ above the average score of randomly generated genomes (ν≥141). The insets show meiotic recombination hotspot and essential gene densities obtained by sliding window analyses of four selected genomes (indexed i to iv). (B) Survival competition of the evolved genomes shown in (A) versus chromosome IX and random genome architectures. The matrices show the average statistical results for all evolved genomes of the green and red groups in (A). (C) The bars indicate the total number of wins of the evolved architectures and of chromosome IX respectively in the competition experiments shown in (B). For simulation details and statistical information see Text S1, section “S. digitalis Simulation Settings.”

Figure 9. The effect of mating type switching on population fitness.

Results of survival competition experiments for S. cerevisiae chromosome IX versus randomly generated chromosome architectures (n = 10 (top row) or 100 (bottom row) experiments per grid point), in the presence (10%) and absence of mating type switching. Different breeding conditions and mutation rates (between 10⁻³ and R_max) were considered. The matrixes in the top row show the outcome of a high-resolution analysis of the high-R regime, for which the most pronounced differences were observed. The level of dominance is color-encoded (bright red: chromosome IX wins in all experiments, bright blue: random architecture wins in all experiments). The population size cap was set to 2,000 individuals. For simulation details and statistical information see Text S1, section “S. digitalis Simulation Settings.”

Figure 10. Natural crossing over rates for yeast chromosomes are optimal.

(A) as a function of the number of crossovers per meiosis (logarithmic scale) for populations of random chromosomes with 750 genes (20% EGs). Bars indicate the SD of R _max for the entire inbreeding/outbreeding domain. Inset: Number of crossovers per meiosis required to reach 95% and 99.5% respectively of the maximum mutational robustness R _max, for chromosomes of different lengths (250, 500, 750, 1,000, and 1,500 genes) and random configurations. (B) Results of survival competition experiments between two chromosome IX populations subjected to different crossing over rates (n = 5 experiments per grid point), for mutation rates R up to R_max and for the entire inbreeding/outbreeding domain. The color-code indicates the winning chromosome architecture. (C) Results of survival competition experiments between yeast chromosome IX and random chromosomes at different crossing over rates (n = 10 experiments per grid point). Same experimental conditions as in (B). Supplementary Figure 8 in Text S1 shows plots for average fitness advantages of the results in (B) and (C). “minimum” = one crossover per chromosome and meiosis; “maximum” = all 58 hotspots are active in each meiosis. For simulation details and statistical information see Text S1, section “S. digitalis Simulation Settings.”

Figure 11. Mutation accumulation and genetic analysis of lethality and mating.

(A) Outline of the mutation accumulation experiment. Diploid GalS-MSH2/GalS-MSH2 cells were grown in the presence of glucose to repress MSH2 expression (MSH2 ^off) for three growth periods of 24 hours (1/1000 dilution between the cultures). After each growth period, an aliquot was removed and the GalS promoter was induced by spreading the cells on galactose-containing plates for 24 hours in order to prevent a further accumulation of mutations. Upon sporulation, the products of meiosis were analyzed for viability using tetrad analysis (results are given in the main text) and sorting of single spores or dyads by FACS. The ability of mating was analyzed by testing the mating types of sorted dyads in order to determine whether the formed colonies were haploid (one spore unviable, no mating) or diploid (mating). (B) Msh2 levels in GalS-MSH2 and control cells were grown first on galactose (1), then on glucose for 24 hours (2) and finally on galactose again for 6 hours (3); detection via Western blotting and antibodies specific to Msh2 (Santa Cruz, Msh2 (yC-15): sc-26230). (C) Viability of spores and dyads. For each time point 1,536 spores and 384 dyads were FACS-sorted in array patterns onto YP-Gal plates (140 mm diameter). Spore viability was determined by counting the number of formed colonies. Mating efficiency of spores in dyads was determined by counting the fraction of the colonies that consisted of diploid cells. In wild type cells and at 0 h in the GalS-MSH2 strain, this fraction was 53%. The probability of having two viable spores of opposite mating type in a dyad was calculated from the time-dependent spore viability and from the fraction of diploid formation at 0 h. These data allowed us to calculate the probability of diploid formation (mating), if unviable spores were unable to mate (dashed line).