B R I E F C O M M U N I C AT I O N doi:10.1111/evo.12228
RAPID EVOLUTION OF CHEATING MITOCHONDRIAL GENOMES IN SMALL YEAST POPULATIONS Jean-Nicolas Jasmin1,2,3 and Clifford Zeyl1 1
Department of Biology, Wake Forest University, Winston-Salem, North Carolina 27109
2
Current address: CEFE-UMR5175, Montpellier 34293, France 3
E-mail: [email protected]
Received February 13, 2013 Accepted July 30, 2013 Data Archived: Dryad doi:10.5061/dryad.2018h Outcrossed sex exposes genes to competition with their homologues, allowing alleles that transmit more often than their competitors to spread despite organismal fitness costs. Mitochondrial populations in species with biparental inheritance are thought to be especially susceptible to such cheaters because they lack strict transmission rules like meiosis or maternal inheritance. Yet the interaction between mutation and natural selection in the evolution of cheating mitochondrial genomes has not been tested experimentally. Using yeast experimental populations, we show that although cheaters were rare in a large sample of spontaneous respiratory-deficient mitochondrial mutations (petites), cheaters evolve under experimentally enforced outcrossing even when mutation supply and selection are restricted by repeatedly bottlenecking populations. KEY WORDS:
Cost of sex, experimental evolution, genetic conflicts, mitochondrial genetics, selfish genetic elements, symbiosis.
Genes with Mendelian inheritance are potentially in evolutionary conflict with those in the mitochondrial genome, which is transmitted in the cytoplasm. Mitochondrial genomes are inherited in three distinct ways, each giving rise to distinct genetic conflicts (Barr et al. 2005). First, in species with maternal inheritance, mitochondrial variants biasing the individual sex ratio toward females can spread (e.g., cytoplasmic male sterility in plants) despite autosomal genes being selected to restore the Fisherian sex ratio (McCauley and Olson 2008). Second, when clonal lineages of unicells or cell lineages in multicellular organisms grow mitotically, mitochondrial mutants transmitting more often than their competitors can spread, even if as a result they lower the fitness of their carriers (Taylor et al. 2002). Finally, cheating mitochondrial genomes can invade outcrossing populations with biparental inheritance, generating a cheating load analogous to the virulence of horizontally transmitted parasites (Hastings 1999; Chinnery et al. 2000; Travisano and Velicer 2004). The detailed study of mitochondrial genetics began with petite mitochondrial mutations (ρ−) of the yeast Saccharomyces cerevisiae (Ephrussi et al. 1955; Evans and Clark-Walker 1985; C
269
Williamson 2002), which arise spontaneously at very high rates (around 1% of cell divisions; Baruffini et al. 2007) and inflict two types of fitness cost: (1) cells fixed (homoplasmic) for ρ− grow ∼30% more slowly than wild-type cells in standard rich media, in part because they cannot respire the ethanol generated during fermentation (Brown et al. 1984; Zeyl and de Visser 2001); and (2) diploid ρ− are sterile because they cannot sporulate. In Saccharomyces yeast, both parents contribute to the mitochondrial genomes inherited by zygotes, allowing some ρ− mitochondrial genomes to cheat by out-replicating wild-type (ρ+) mitochondrial genomes within cells that are heteroplasmic (mixed) for ρ− and ρ+ genomes, leading to their overrepresentation in daughter cells (Ephrussi and Grandchamp 1965; Uchida and Suda 1978; Hurst 1994; Hastings 1999; MacAlpine et al. 2001; Taylor et al. 2002; Barr et al. 2005; Bernardi 2005). Spontaneous ρ− mutants that outcompete ρ+ in heteroplasmic diploids during mitosis can make up >10% of all ρ− mutants (Petersen et al. 2002). Besides its importance for the mutational load, such cheating is also the basis of models of the evolutionary transition from biparental mitochondrial inheritance to the uniparental inheritance typical
C 2013 The Society for the Study of Evolution. 2013 The Author(s). Evolution Evolution 68-1: 269–275
B R I E F C O M M U N I C AT I O N
of plants and animals. These models assume that mitochondrial cheaters can evolve from wild types with unbiased transmission (Hastings 1999; Randerson and Hurst 1999), but this idea lacks empirical support. Here we assessed the availability of mutations for cheating by manipulating the mutation supply and the intensity of selection on three independent ρ− ancestors that were not cheaters. To do this, we experimentally enforced outcrossing upon the naturally inbreeding yeast S. cerevisiae (Tsai et al. 2008). We found that transmission rates of ∼85% evolved at all but the smallest effective population sizes for two out of three ancestral ρ−. In the third ρ−, transmission rates never evolved beyond 55%, indicating that mutations for strong cheating are less accessible to some ρ− genomes than others. Moreover, in contrast to previous studies that measured cheating during mitosis alone, we did not observe mutants with transmission rates >95%. Overall, our results suggest that, provided that their fitness costs are not too large (Hastings 1999), cheating mitochondrial mutants would rapidly invade outcrossing yeast populations, which could favor the evolution of uniparental mitochondrial inheritance (Randerson and Hurst 1999).
Methods SPONTANEOUS PETITES AND TRANSMISSION RATE ASSAYS
To obtain spontaneous ρ− mutants, ρ+ clones of genotypes MATa and MATα (strain YPS670 ho::KANMX4) were grown in liquid yeast extract-peptone-dextrose (YPD) and mixed for mating on solid YPD (2% dextrose, 2% bacto-peptone, 1% yeast extract, plus 2% agar for solid medium). Cells were allowed to mate for 1.5– 2 h at 30◦ C and transferred to sporulation plates (2% potassium acetate, 0.05% zinc acetate, 2% agar) and incubated for 2 days at 30◦ C. Spores were isolated and streaked on yeast extract-peptoneglycerol-dextrose (YPGD) plates (2% glycerol, 0.1% dextrose, 2% bacto-peptone, 1% yeast extract, 2% agar), on which petite colonies are smaller than grande colonies because petites cannot metabolize glycerol. Small colonies were restreaked on YPGD plates to confirm that they could not respire. A single petite colony was stored frozen in 15% glycerol for each of 124 replicates. To assay the transmission rate of these 124 petites, each mutant was crossed to the ancestral grande clone of opposite mating type, following the same procedure as for mutant isolation. After spore suspensions were streaked on YPGD plates, petite colonies were differentiated from grandes using soft agar containing tetrazolium chloride (0.07% tetrazolium chloride, 1% dextrose, 1.4% agar, poured at 50◦ C; Ogur et al. 1959). Grande colonies turn red whereas petite colonies remain white.
270
EVOLUTION JANUARY 2014
Nearly all prior studies reporting transmission rates of ρ− genotypes (Ephrussi et al. 1965; MacAlpine et al. 2001; Petersen et al. 2002; Bernardi 2005) defined transmission rates by the ability of ρ− to fix in zygotic clones, which measures the intracellular fitness of ρ− during diploid mitosis. This is not relevant to the evolutionary genetics of ρ− because diploids homoplasmic for ρ− cannot sporulate and are “dead ends” for ρ−. The transmission rate of ρ− in the sexual cycle, as used here, is the parameter relevant for theoretical models of mitochondrial cheating (Hastings 1999; Randerson and Hurst 1999). SPORE ISOLATION
Sporulated populations were harvested from sporulation plates and resuspended in a sterile predigestion solution (0.8 M potassium chloride, 0.1 M sodium sulfite). After 0.5–1 h at 30◦ C, sterile digestion solution was added (final concentrations: 0.1 mg/mL zymolyase 100T, 2.5 μL/mL β-mercaptoethanol, 50 mmol/L sorbitol, 50–100 μL glass beads) and the cell suspensions were incubated for 3 to 5 h at 30◦ C while vortexing at every ∼30 min. Next, we added Triton-X and vortexed for 1 min. These steps were aimed at killing vegetative cells, digesting asci, and separating spore tetrads to enforce outcrossing. The cell suspensions were washed twice in water. Finally, the cells were resuspended in 1 mL water and mineral oil and vortexed for 1–2 min, after which the mineral oil was transferred to a new tube. Spores are lipophilic and segregate to the oil phase, whereas surviving vegetative cells are hydrophilic and remain in the water phase. The mineral oil was then washed with water and streaked onto YPGD plates. Several controls indicated that less than 1% of the colonies formed on YPGD arose from vegetative cells. SELECTING FOR CHEATERS
Selection lines were established by three ancestral ρ− sampled randomly from the library of 124 spontaneous ρ−, each propagated independently throughout the selection cycle (see Fig. 1 for selection procedure) at five effective population sizes (the number N of passaged colonies: 1, 3, 10, 32, or 100), for a total of 15 lines. We verified that each of the N putative petite sampled at each selection cycle was respiratory-deficient by replating each colony on YPGD. The colonies arising from a single putative petite colony always had similar size on YPGD, supporting the assumption that the cells within a colony arising from an isolated spore were homoplasmic for ρ−. After 20 selection cycles, the non-Mendelian basis of evolved petites was confirmed by dissecting tetrads derived from three consecutive crosses of the ρ+ ancestor to a single ρ− clone isolated from each selection line. The transmission rate of a single ρ− clone per line, isolated after these three consecutive crosses, was estimated along with those of the three ancestral ρ− clones using the same assay as for the spontaneous ρ− mutants (strains were randomized for the assay), except that the mating
B R I E F C O M M U N I C AT I O N
mix and mate sporulaƟon
ρ+ +
ρ− isolate and plate spores
a
α pick and culture N peƟtes
Procedure for the experimental selection of cheating ρ− mitochondrial genomes that outcompete wild-type ρ+ genomes
Figure 1.
in heteroplasmic cells during meiosis. Starting from the top left, ρ− strains (comprising both mating types) and ancestral ρ+ (50% MATα, 50% MATa) are mixed in equal proportions for mating on a Petri dish. After 1.5 h, the mating population is transferred to sporulation medium for 2 days. Spores are harvested, isolated, and plated on rich media (YPGD). Petite colonies result from ρ+ × ρ− crosses. N ρ− colonies (where N is 1, 3, 10, 32, or 100 ρ−) are cultured independently to initiate the next selection cycle, and checked to ensure respiratory inability (not shown on diagram). These N ρ− clones are then mixed in equal proportions with ancestral ρ+ of both mating types (which are regrown from frozen stocks at every selection cycle). There were three independent ancestral ρ− per sample size (15 selection lines in all).
phase lasted from 90 to 100 minutes. We analyzed the variation in the petite counts of evolved genotypes assuming it follows a binomial distribution, in a generalized model accounting for a quadratic relationship between transmission rates and effective population size. The model included the ancestral genotype effect (fixed), the effective population size effect (fixed), and the interaction between ancestral genotypes and the quadratic and linear effects of the effective population size (Proc GENMOD procedure in SAS 9; SAS Institute Inc., Cary, NC). The significance of this interaction was tested by model selection based on likelihood ratio tests. We also obtained Bayesian estimates of the binomially distributed transmission rates of spontaneous ρ−, of Mendelian mip1 mtDNA-free genotypes, and of evolved ρ−, based on two replicate assays of each one.
Results We first studied the transmission rates of spontaneous petite mutants to obtain a point of comparison with the evolved ρ− genomes and to assess the genetic variation in transmission rates available to selection. Transmission rates were estimated by mass-mating each mutant to the ρ+ ancestor and sporulating the resulting
heteroplasmic diploids immediately to limit selection for respiratory ability. As a control for the transmission rate assays, we sporulated a strain heterozygous for a deletion of the mip1 gene (genotype mip1/MIP1), coding for the mitochondrial DNA polymerase. This control tests for unintended selection against petites that might arise between spore isolation and colony phenotyping. For example, spores that are more likely to yield a petite colony may also have lower viabilities. In the absence of such selection we expect the mip1/MIP1 heterozygote to yield petites (mip1) and grandes (MIP1) colonies in a 1:1 ratio. We observed a 4.7% bias against mip1 petites (95% confidence interval, 43.3–47.2% based on a binomial distribution, see Supporting Information for statistics of transmission rates). Therefore, when testing for greater than Mendelian transmission rates (i.e., cheating), we compared transmission rates to the estimated 45.3% rate accounting for selection against petites. However, selection against mitochondrial petites is likely to be weaker than that against mip1 petites if mitochondrial mutations are recessive, such that the fitness of nuclear petites (here, mip1 haploids) is similar to genotypes homoplasmic for ρ−, but is lower than that of heteroplasmic cells. Thus, selection against petites is probably overestimated by the mip1 control, save perhaps for the ρ− genotypes with transmission rates well above 50%. Out of 124 independent spontaneous mutants, eight had higher transmission rates than the Mendelian control, four (3.2%) of these significantly so (binomial 2.5–97.5 percentiles did not overlap with those of mip1, see Supporting Information; Fig. 2A), and three of those (2.4%) had transmission rates also significantly greater than 50%. In general, spontaneous ρ− fixed in only one out of four haploid lineages arising from meiotic products (the median transmission rate is 25.5%). However, these estimates need to be taken with caution due to a gradient of selection against petites resulting from the order in which spontaneous mutants were transferred to sporulation medium (Fig. S1). Harvesting mating mixtures from plates required more time than mixing and spreading them, such that the first mutant was transferred to sporulation medium after 1.5 h of mating, whereas mutant number 124 spent an extra 30 min on the mating plate, where there is strong selection against petites. We found that, on average, the first mutants had a 10.4% higher transmission rate than the last mutants (a logistic regression of transmission rate on mutant number yielded a highly significant coefficient of −8.38 × 10−4 ; Fig. S1). We randomly picked three independent spontaneous ρ− to establish experimental lines selecting for cheating ρ−. We named those lines X (ancestral transmission rates of ∼17%), Y (∼18%), and Z (∼42%). After 20 cycles of experimental selection (Fig. 1), transmission rates had increased significantly in all 15 lines except line Z at N = 1, which actually declined significantly (Fig. 2B). We defined significant changes in transmission rates, and the
EVOLUTION JANUARY 2014
271
B R I E F C O M M U N I C AT I O N
Figure 2. Evolution of cheating mitochondrial genomes from spontaneous petites. A, distribution of transmission rates of 124 independent spontaneous ρ− strains obtained from spore suspensions after crossing two ρ+. Three of these spontaneous ρ− genotypes were
selected for 20 cycles (see Fig. 1) at various effective population size N. B, transmission rates of each of these three ρ− before (anc) and after experimental selection at effective population size N (ancestor X, triangles and dashed line; Y, empty circles and dotted line; Z, filled circles and solid line). The horizontal dotted line shows the transmission rate of the mip1 Mendelian control. All quadratic and linear coefficients of each curve are significantly different from zero (Wald χ2 > 119, P < 0.0001).
evolution of cheating, by comparing the 2.5th–97.5th percentiles of transmission rate estimates for evolved lines with those of their ancestors, and of the Mendelian mip1 control, respectively (Supporting Information). Cheaters evolved in all five lines from the Y ancestor, in all lines but N = 1 from ancestor Z, and in the N = 3 and 10 lines of ancestor X. Note that variation in the length of time spent on mating plates could have affected our transmission rate estimates by at most ∼1.6% (8.38 × 10−4 × 19 genotypes in each assay). Transmission rate evolution was influenced by both effective population size (N) and ancestry, as well as the interaction between those factors. The response to selection was, predictably, smallest at N = 1, and transmission rates plateaued rapidly with increasing N, as indicated by the significantly negative coefficients for the quadratic terms in the generalized model (quadratic coefficients ranged from, mean ± SE, −1.1 ± 0.05 for Z to −0.51 ± 0.05 for Y; Table 1 and Fig. 2B). For two ancestors, the quadratic relationships (Fig. 2B and Table 1) suggest that transmission rates were lower at the largest N values (32 and 100) than at N = 10, but our experimental design lacks the evolutionary replication needed to distinguish statistically such a decline from an asymptotic relationship between N and evolved transmission rates. Lines founded by different ancestors differed in their response to effective popTable 1.
ulation size, suggesting that some ρ− respond less to selection than others (a model including the interaction term gave a better fit than one without, df = 4, χ2 = 190, P < 0.0001; see also Table 1). The difference among ancestors is especially striking for X versus Y, which had nearly identical ancestral transmission rates (the intercept of the generalized linear model is −0.91 ± 0.06 for X and −0.04 ± 0.06 for Y; Fig. 2B). When N ≥ 10, transmission rates increased up to 86% and 85% for the Y and Z ancestors, respectively, whereas ancestor X failed to evolve transmission rates higher than 55% at any population size.
Discussion Saccharomyces is highly inbred (Tsai et al. 2008), which ensures that the mitochondrial genomes that a zygote inherits from its parents are closely related, minimizing genetic conflicts (Hurst 1994). We disrupted this system of inbreeding and biparental inheritance by enforcing outbreeding onto experimental populations. Outbreeding lowers relatedness among mitochondrial genomes when their transmission is biparental, thus generating cytoplasmic competition that selects for cheating genomes (Hurst 1994; Frank 2003). To assess the potential for cheaters to evolve, we manipulated the supply of mutation and the strength of
Generalized linear model of the effect of ancestral ρ− genotype and the effective population size on evolved transmission
rates.
Effect
df
χ2
P
Ancestral ρ− Effective population size (N) N2 N × ancestral ρ− N2 × ancestral ρ−
2 1 1 2 2
379 1446 857 51 65
RAPID EVOLUTION OF CHEATING MITOCHONDRIAL GENOMES IN SMALL YEAST POPULATIONS Jean-Nicolas Jasmin1,2,3 and Clifford Zeyl1 1
Department of Biology, Wake Forest University, Winston-Salem, North Carolina 27109
2
Current address: CEFE-UMR5175, Montpellier 34293, France 3
E-mail: [email protected]
Received February 13, 2013 Accepted July 30, 2013 Data Archived: Dryad doi:10.5061/dryad.2018h Outcrossed sex exposes genes to competition with their homologues, allowing alleles that transmit more often than their competitors to spread despite organismal fitness costs. Mitochondrial populations in species with biparental inheritance are thought to be especially susceptible to such cheaters because they lack strict transmission rules like meiosis or maternal inheritance. Yet the interaction between mutation and natural selection in the evolution of cheating mitochondrial genomes has not been tested experimentally. Using yeast experimental populations, we show that although cheaters were rare in a large sample of spontaneous respiratory-deficient mitochondrial mutations (petites), cheaters evolve under experimentally enforced outcrossing even when mutation supply and selection are restricted by repeatedly bottlenecking populations. KEY WORDS:
Cost of sex, experimental evolution, genetic conflicts, mitochondrial genetics, selfish genetic elements, symbiosis.
Genes with Mendelian inheritance are potentially in evolutionary conflict with those in the mitochondrial genome, which is transmitted in the cytoplasm. Mitochondrial genomes are inherited in three distinct ways, each giving rise to distinct genetic conflicts (Barr et al. 2005). First, in species with maternal inheritance, mitochondrial variants biasing the individual sex ratio toward females can spread (e.g., cytoplasmic male sterility in plants) despite autosomal genes being selected to restore the Fisherian sex ratio (McCauley and Olson 2008). Second, when clonal lineages of unicells or cell lineages in multicellular organisms grow mitotically, mitochondrial mutants transmitting more often than their competitors can spread, even if as a result they lower the fitness of their carriers (Taylor et al. 2002). Finally, cheating mitochondrial genomes can invade outcrossing populations with biparental inheritance, generating a cheating load analogous to the virulence of horizontally transmitted parasites (Hastings 1999; Chinnery et al. 2000; Travisano and Velicer 2004). The detailed study of mitochondrial genetics began with petite mitochondrial mutations (ρ−) of the yeast Saccharomyces cerevisiae (Ephrussi et al. 1955; Evans and Clark-Walker 1985; C
269
Williamson 2002), which arise spontaneously at very high rates (around 1% of cell divisions; Baruffini et al. 2007) and inflict two types of fitness cost: (1) cells fixed (homoplasmic) for ρ− grow ∼30% more slowly than wild-type cells in standard rich media, in part because they cannot respire the ethanol generated during fermentation (Brown et al. 1984; Zeyl and de Visser 2001); and (2) diploid ρ− are sterile because they cannot sporulate. In Saccharomyces yeast, both parents contribute to the mitochondrial genomes inherited by zygotes, allowing some ρ− mitochondrial genomes to cheat by out-replicating wild-type (ρ+) mitochondrial genomes within cells that are heteroplasmic (mixed) for ρ− and ρ+ genomes, leading to their overrepresentation in daughter cells (Ephrussi and Grandchamp 1965; Uchida and Suda 1978; Hurst 1994; Hastings 1999; MacAlpine et al. 2001; Taylor et al. 2002; Barr et al. 2005; Bernardi 2005). Spontaneous ρ− mutants that outcompete ρ+ in heteroplasmic diploids during mitosis can make up >10% of all ρ− mutants (Petersen et al. 2002). Besides its importance for the mutational load, such cheating is also the basis of models of the evolutionary transition from biparental mitochondrial inheritance to the uniparental inheritance typical
C 2013 The Society for the Study of Evolution. 2013 The Author(s). Evolution Evolution 68-1: 269–275
B R I E F C O M M U N I C AT I O N
of plants and animals. These models assume that mitochondrial cheaters can evolve from wild types with unbiased transmission (Hastings 1999; Randerson and Hurst 1999), but this idea lacks empirical support. Here we assessed the availability of mutations for cheating by manipulating the mutation supply and the intensity of selection on three independent ρ− ancestors that were not cheaters. To do this, we experimentally enforced outcrossing upon the naturally inbreeding yeast S. cerevisiae (Tsai et al. 2008). We found that transmission rates of ∼85% evolved at all but the smallest effective population sizes for two out of three ancestral ρ−. In the third ρ−, transmission rates never evolved beyond 55%, indicating that mutations for strong cheating are less accessible to some ρ− genomes than others. Moreover, in contrast to previous studies that measured cheating during mitosis alone, we did not observe mutants with transmission rates >95%. Overall, our results suggest that, provided that their fitness costs are not too large (Hastings 1999), cheating mitochondrial mutants would rapidly invade outcrossing yeast populations, which could favor the evolution of uniparental mitochondrial inheritance (Randerson and Hurst 1999).
Methods SPONTANEOUS PETITES AND TRANSMISSION RATE ASSAYS
To obtain spontaneous ρ− mutants, ρ+ clones of genotypes MATa and MATα (strain YPS670 ho::KANMX4) were grown in liquid yeast extract-peptone-dextrose (YPD) and mixed for mating on solid YPD (2% dextrose, 2% bacto-peptone, 1% yeast extract, plus 2% agar for solid medium). Cells were allowed to mate for 1.5– 2 h at 30◦ C and transferred to sporulation plates (2% potassium acetate, 0.05% zinc acetate, 2% agar) and incubated for 2 days at 30◦ C. Spores were isolated and streaked on yeast extract-peptoneglycerol-dextrose (YPGD) plates (2% glycerol, 0.1% dextrose, 2% bacto-peptone, 1% yeast extract, 2% agar), on which petite colonies are smaller than grande colonies because petites cannot metabolize glycerol. Small colonies were restreaked on YPGD plates to confirm that they could not respire. A single petite colony was stored frozen in 15% glycerol for each of 124 replicates. To assay the transmission rate of these 124 petites, each mutant was crossed to the ancestral grande clone of opposite mating type, following the same procedure as for mutant isolation. After spore suspensions were streaked on YPGD plates, petite colonies were differentiated from grandes using soft agar containing tetrazolium chloride (0.07% tetrazolium chloride, 1% dextrose, 1.4% agar, poured at 50◦ C; Ogur et al. 1959). Grande colonies turn red whereas petite colonies remain white.
270
EVOLUTION JANUARY 2014
Nearly all prior studies reporting transmission rates of ρ− genotypes (Ephrussi et al. 1965; MacAlpine et al. 2001; Petersen et al. 2002; Bernardi 2005) defined transmission rates by the ability of ρ− to fix in zygotic clones, which measures the intracellular fitness of ρ− during diploid mitosis. This is not relevant to the evolutionary genetics of ρ− because diploids homoplasmic for ρ− cannot sporulate and are “dead ends” for ρ−. The transmission rate of ρ− in the sexual cycle, as used here, is the parameter relevant for theoretical models of mitochondrial cheating (Hastings 1999; Randerson and Hurst 1999). SPORE ISOLATION
Sporulated populations were harvested from sporulation plates and resuspended in a sterile predigestion solution (0.8 M potassium chloride, 0.1 M sodium sulfite). After 0.5–1 h at 30◦ C, sterile digestion solution was added (final concentrations: 0.1 mg/mL zymolyase 100T, 2.5 μL/mL β-mercaptoethanol, 50 mmol/L sorbitol, 50–100 μL glass beads) and the cell suspensions were incubated for 3 to 5 h at 30◦ C while vortexing at every ∼30 min. Next, we added Triton-X and vortexed for 1 min. These steps were aimed at killing vegetative cells, digesting asci, and separating spore tetrads to enforce outcrossing. The cell suspensions were washed twice in water. Finally, the cells were resuspended in 1 mL water and mineral oil and vortexed for 1–2 min, after which the mineral oil was transferred to a new tube. Spores are lipophilic and segregate to the oil phase, whereas surviving vegetative cells are hydrophilic and remain in the water phase. The mineral oil was then washed with water and streaked onto YPGD plates. Several controls indicated that less than 1% of the colonies formed on YPGD arose from vegetative cells. SELECTING FOR CHEATERS
Selection lines were established by three ancestral ρ− sampled randomly from the library of 124 spontaneous ρ−, each propagated independently throughout the selection cycle (see Fig. 1 for selection procedure) at five effective population sizes (the number N of passaged colonies: 1, 3, 10, 32, or 100), for a total of 15 lines. We verified that each of the N putative petite sampled at each selection cycle was respiratory-deficient by replating each colony on YPGD. The colonies arising from a single putative petite colony always had similar size on YPGD, supporting the assumption that the cells within a colony arising from an isolated spore were homoplasmic for ρ−. After 20 selection cycles, the non-Mendelian basis of evolved petites was confirmed by dissecting tetrads derived from three consecutive crosses of the ρ+ ancestor to a single ρ− clone isolated from each selection line. The transmission rate of a single ρ− clone per line, isolated after these three consecutive crosses, was estimated along with those of the three ancestral ρ− clones using the same assay as for the spontaneous ρ− mutants (strains were randomized for the assay), except that the mating
B R I E F C O M M U N I C AT I O N
mix and mate sporulaƟon
ρ+ +
ρ− isolate and plate spores
a
α pick and culture N peƟtes
Procedure for the experimental selection of cheating ρ− mitochondrial genomes that outcompete wild-type ρ+ genomes
Figure 1.
in heteroplasmic cells during meiosis. Starting from the top left, ρ− strains (comprising both mating types) and ancestral ρ+ (50% MATα, 50% MATa) are mixed in equal proportions for mating on a Petri dish. After 1.5 h, the mating population is transferred to sporulation medium for 2 days. Spores are harvested, isolated, and plated on rich media (YPGD). Petite colonies result from ρ+ × ρ− crosses. N ρ− colonies (where N is 1, 3, 10, 32, or 100 ρ−) are cultured independently to initiate the next selection cycle, and checked to ensure respiratory inability (not shown on diagram). These N ρ− clones are then mixed in equal proportions with ancestral ρ+ of both mating types (which are regrown from frozen stocks at every selection cycle). There were three independent ancestral ρ− per sample size (15 selection lines in all).
phase lasted from 90 to 100 minutes. We analyzed the variation in the petite counts of evolved genotypes assuming it follows a binomial distribution, in a generalized model accounting for a quadratic relationship between transmission rates and effective population size. The model included the ancestral genotype effect (fixed), the effective population size effect (fixed), and the interaction between ancestral genotypes and the quadratic and linear effects of the effective population size (Proc GENMOD procedure in SAS 9; SAS Institute Inc., Cary, NC). The significance of this interaction was tested by model selection based on likelihood ratio tests. We also obtained Bayesian estimates of the binomially distributed transmission rates of spontaneous ρ−, of Mendelian mip1 mtDNA-free genotypes, and of evolved ρ−, based on two replicate assays of each one.
Results We first studied the transmission rates of spontaneous petite mutants to obtain a point of comparison with the evolved ρ− genomes and to assess the genetic variation in transmission rates available to selection. Transmission rates were estimated by mass-mating each mutant to the ρ+ ancestor and sporulating the resulting
heteroplasmic diploids immediately to limit selection for respiratory ability. As a control for the transmission rate assays, we sporulated a strain heterozygous for a deletion of the mip1 gene (genotype mip1/MIP1), coding for the mitochondrial DNA polymerase. This control tests for unintended selection against petites that might arise between spore isolation and colony phenotyping. For example, spores that are more likely to yield a petite colony may also have lower viabilities. In the absence of such selection we expect the mip1/MIP1 heterozygote to yield petites (mip1) and grandes (MIP1) colonies in a 1:1 ratio. We observed a 4.7% bias against mip1 petites (95% confidence interval, 43.3–47.2% based on a binomial distribution, see Supporting Information for statistics of transmission rates). Therefore, when testing for greater than Mendelian transmission rates (i.e., cheating), we compared transmission rates to the estimated 45.3% rate accounting for selection against petites. However, selection against mitochondrial petites is likely to be weaker than that against mip1 petites if mitochondrial mutations are recessive, such that the fitness of nuclear petites (here, mip1 haploids) is similar to genotypes homoplasmic for ρ−, but is lower than that of heteroplasmic cells. Thus, selection against petites is probably overestimated by the mip1 control, save perhaps for the ρ− genotypes with transmission rates well above 50%. Out of 124 independent spontaneous mutants, eight had higher transmission rates than the Mendelian control, four (3.2%) of these significantly so (binomial 2.5–97.5 percentiles did not overlap with those of mip1, see Supporting Information; Fig. 2A), and three of those (2.4%) had transmission rates also significantly greater than 50%. In general, spontaneous ρ− fixed in only one out of four haploid lineages arising from meiotic products (the median transmission rate is 25.5%). However, these estimates need to be taken with caution due to a gradient of selection against petites resulting from the order in which spontaneous mutants were transferred to sporulation medium (Fig. S1). Harvesting mating mixtures from plates required more time than mixing and spreading them, such that the first mutant was transferred to sporulation medium after 1.5 h of mating, whereas mutant number 124 spent an extra 30 min on the mating plate, where there is strong selection against petites. We found that, on average, the first mutants had a 10.4% higher transmission rate than the last mutants (a logistic regression of transmission rate on mutant number yielded a highly significant coefficient of −8.38 × 10−4 ; Fig. S1). We randomly picked three independent spontaneous ρ− to establish experimental lines selecting for cheating ρ−. We named those lines X (ancestral transmission rates of ∼17%), Y (∼18%), and Z (∼42%). After 20 cycles of experimental selection (Fig. 1), transmission rates had increased significantly in all 15 lines except line Z at N = 1, which actually declined significantly (Fig. 2B). We defined significant changes in transmission rates, and the
EVOLUTION JANUARY 2014
271
B R I E F C O M M U N I C AT I O N
Figure 2. Evolution of cheating mitochondrial genomes from spontaneous petites. A, distribution of transmission rates of 124 independent spontaneous ρ− strains obtained from spore suspensions after crossing two ρ+. Three of these spontaneous ρ− genotypes were
selected for 20 cycles (see Fig. 1) at various effective population size N. B, transmission rates of each of these three ρ− before (anc) and after experimental selection at effective population size N (ancestor X, triangles and dashed line; Y, empty circles and dotted line; Z, filled circles and solid line). The horizontal dotted line shows the transmission rate of the mip1 Mendelian control. All quadratic and linear coefficients of each curve are significantly different from zero (Wald χ2 > 119, P < 0.0001).
evolution of cheating, by comparing the 2.5th–97.5th percentiles of transmission rate estimates for evolved lines with those of their ancestors, and of the Mendelian mip1 control, respectively (Supporting Information). Cheaters evolved in all five lines from the Y ancestor, in all lines but N = 1 from ancestor Z, and in the N = 3 and 10 lines of ancestor X. Note that variation in the length of time spent on mating plates could have affected our transmission rate estimates by at most ∼1.6% (8.38 × 10−4 × 19 genotypes in each assay). Transmission rate evolution was influenced by both effective population size (N) and ancestry, as well as the interaction between those factors. The response to selection was, predictably, smallest at N = 1, and transmission rates plateaued rapidly with increasing N, as indicated by the significantly negative coefficients for the quadratic terms in the generalized model (quadratic coefficients ranged from, mean ± SE, −1.1 ± 0.05 for Z to −0.51 ± 0.05 for Y; Table 1 and Fig. 2B). For two ancestors, the quadratic relationships (Fig. 2B and Table 1) suggest that transmission rates were lower at the largest N values (32 and 100) than at N = 10, but our experimental design lacks the evolutionary replication needed to distinguish statistically such a decline from an asymptotic relationship between N and evolved transmission rates. Lines founded by different ancestors differed in their response to effective popTable 1.
ulation size, suggesting that some ρ− respond less to selection than others (a model including the interaction term gave a better fit than one without, df = 4, χ2 = 190, P < 0.0001; see also Table 1). The difference among ancestors is especially striking for X versus Y, which had nearly identical ancestral transmission rates (the intercept of the generalized linear model is −0.91 ± 0.06 for X and −0.04 ± 0.06 for Y; Fig. 2B). When N ≥ 10, transmission rates increased up to 86% and 85% for the Y and Z ancestors, respectively, whereas ancestor X failed to evolve transmission rates higher than 55% at any population size.
Discussion Saccharomyces is highly inbred (Tsai et al. 2008), which ensures that the mitochondrial genomes that a zygote inherits from its parents are closely related, minimizing genetic conflicts (Hurst 1994). We disrupted this system of inbreeding and biparental inheritance by enforcing outbreeding onto experimental populations. Outbreeding lowers relatedness among mitochondrial genomes when their transmission is biparental, thus generating cytoplasmic competition that selects for cheating genomes (Hurst 1994; Frank 2003). To assess the potential for cheaters to evolve, we manipulated the supply of mutation and the strength of
Generalized linear model of the effect of ancestral ρ− genotype and the effective population size on evolved transmission
rates.
Effect
df
χ2
P
Ancestral ρ− Effective population size (N) N2 N × ancestral ρ− N2 × ancestral ρ−
2 1 1 2 2
379 1446 857 51 65
