The environmental change experienced by many contemporary populations of organisms poses a serious risk to their survival. From the theory of evolutionary rescue, we predict that the combination of sex and genetic diversity should increase the probability of survival by increasing variation and thereby the probability of generating a type that can tolerate the stressful environment. We tested this prediction by comparing experimental populations of Chlamydomonas reinhardtii that differ in sexuality and in the initial amount of genetic diversity. The lines were serially propagated in an environment where the level of stress caused by salt increased over time from fresh water to the limits of marine conditions. In the long term, the combination of high diversity and obligate sexuality was most effective in supporting evolutionary rescue. Most of the adaptation to high-salt environments in the obligate sexual-high diversity lines had occurred by midway through the experiment, indicating that positive genetic correlations of adaptation to lethal stress with adaptation to sublethal stress greatly increased the probability of evolutionary rescue. The evolutionary rescue events observed in this study provide evidence that major shifts in ways of life can arise within short time frames through the action of natural selection in sexual populations.

A population may adapt through natural selection to conditions lethal to its ancestors. This is the process of “evolutionary rescue.” Adaptation will occur whenever there is variation in relative fitness; evolutionary rescue requires in addition that there is variance in absolute fitness such that the fittest genotype at least achieves positive growth when a stress is applied. Populations are likely to survive if the time for a resistant type to spread through the population and restore population growth is less than the time for population density to decrease to critically low levels (Gomulkiewicz and Holt 1995). The probability of being rescued by natural selection depends on factors such as the rate of environmental change, the strength of selection, population size, and the amount of genetic variation (Lynch et al. 1991; Burger and Lynch 1995; Bell and Collins 2008; Orr and Unckless 2008). In recent years, empirical investigations have provided support for the predictions that larger population sizes (Bell and Gonzalez 2009; Willi and Hoffmann 2009), more genetic variation (Agashe et al. 2011), and lower rates of environmental deterioration (Perron et al. 2008; Bell and Gonzalez 2011) increase the probability of survival. Although most natural populations of eukaryotes are sexual and genetically diverse, few experiments have investigated the importance of standing genetic variation and recombination on the probability of survival in a deteriorating environment.

If the starting population is isogenic, adaptation will occur solely through the fixation of novel mutations (Bell and Collins 2008; Orr and Unckless 2008). The rate of adaptation will increase with population size as the mutation supply rate increases although the relationship might flatten out at high population sizes because competition among clones bearing different mutations in asexual populations may delay the fixation of the best (Fisher 1930; Muller 1932) through “clonal interference” (Gerrish and Lenski 1998). In practice, the rate of adaptation in asexual populations increases at a decreasing rate with population size (Samani and Bell 2010).

If the base population is genetically diverse, adaptation will occur through the sorting of genetic variation that has been retained through processes such as frequency-dependent selection, mutation, and gene flow. Adaptation from standing genetic variation may occur rapidly because the alleles are already present in the population at high frequencies (reviewed by Barrett and Schluter 2008). In very large populations of microbes, the mutation supply rate may be so high that mutations form a pool of variability that is constantly available to selection, and the distinction between adaptation from standing genetic variation and from mutations becomes less well defined.

Adaptation from standing genetic variation can occur through the fixation of alleles that were previously deleterious or neutral but are beneficial in the new environment. Neutral alleles may drift in the population, whereas deleterious alleles may be maintained at mutation–selection equilibrium. The probability of being rescued by previously deleterious alleles is dependent on the frequency of the allele in the population. This frequency is a factor of the population size and the deleterious selection coefficient sd in particular. Assuming no back selection, the probability P that a previously deleterious allele will rescue a population is equal to P= 1 –[(sbr+sd)/sd)]−2Noμ where sb is the beneficial selection coefficient (Orr and Unckless 2008). The result is similar for the probability that a population will be rescued by a previously neutral allele with the added effect of the rate of back mutation. Survival is more likely to occur through the fixation of previously neutral alleles than from previously deleterious mutations because they are usually at higher frequencies in the population.

Finally, if the population is sexual, recombination will indirectly increase the rate of adaptation in a homogeneous but changing environment either by generating novel combinations of beneficial alleles (Weismann 1889; Fisher 1930; Muller 1932), or by freeing beneficial alleles found in inferior genetic backgrounds (Hill and Robertson 1966; Peck 1994). There are a number of empirical studies that support the hypothesis that recombination increases the rate of adaptation in a novel environment (Colegrave 2002; Kaltz and Bell 2002; Goddard et al. 2005; Morran et al. 2009) or during coevolutionary interactions (Morran et al. 2011). More specifically, empirical studies have demonstrated that the effects of recombination occur on two different time scales. Over the short term, it will decrease the mean fitness of offspring because of recombination load, but in the longer term the genetic variance generated by sexual episodes will favor the sexual populations over asexual populations when they are facing a novel environment because it will increase their ability to respond to selection (Greig et al. 1998; Colegrave et al. 2002; Kaltz and Bell 2002; Becks and Agrawal 2011).

Here, we report a test of the theory that sexuality combined with genetic variation increases the probability of evolutionary rescue by increasing the rate of adaptation and thereby enabling populations to survive higher levels of stress. We exposed experimental populations of the unicellular alga Chlamydomonas reinhardtii to an environment where the salt (NaCl) concentration increased over time from freshwater to marine conditions. Salt imposes osmotic and oxidative stresses by disrupting the homeostasis of ions (Na+, Cl, K+, and Ca2+), and inhibits growth by decreasing the rates of photosynthesis (Husic and Tolbert 1986). Reynoso and de Gamboa (1982) used optical density readings coupled with hemocytometer measurements to make the observation that in C. reinhardtii, growth is suppressed by 50% at a salinity of 0.085 M (4.97 gL−1), and completely suppressed by salinities higher than 0.26 M (15.12 gL−1). Similarly, Moser and Bell (2011) found that optical density was reduced by 50% at NaCl concentrations of about 5 gL−1, and almost completely suppressed at 8 gL−1. We exposed populations to salt concentrations that were initially low but increased steadily over time until they exceeded the limit of tolerance of their ancestors. These populations differed in the initial amount of genetic diversity and in the expression of sexuality. We found that the combination of high initial diversity and obligate sexuality increased the likelihood of evolutionary rescue in populations exposed to a deteriorating environment by increasing their ability to respond to selection.



We constructed base populations by isolating one spore from each of 40 different lines that have been cultured in the dark, using sodium acetate as a carbon source, for about three years. These lines and the lines from which they were derived had not been exposed to high-salt environments during more than 10 years of laboratory culture. The mating type (plus or minus) of each spore was determined by crossing it with testers, and the same number of spores from each of the two mating types was used to build the library. Populations were cultured in their ancestral environment, Bold's medium (Harris 2009) supplemented with 1.2 gL−1 sodium acetate, before the start of every assay.


We propagated each spore in the ancestral medium without added salt for two growth cycles before transferring 5% of each culture to a series of different salt concentrations between 0 and 7 gL−1 NaCl on a 96-well plate. The maximum concentration tested was set at 7 gL−1 NaCl because Moser and Bell (2011) showed that it causes almost complete suppression of growth. Growth was monitored throughout the second cycle in the assay environments by measuring optical density at 660 nm with a Bio-Tek synergy HT plate reader (BioTek Instruments, Highland Park, Winooski, VT). In all fitness assays, we measured optical density nine to 10 times during a growth cycle and we used the maximum optical density reached during the growth cycle as the measure of fitness. Maximum growth reached during the growth cycle was used instead of final growth because maximum growth is independent of the initial inoculum size and unlike final growth, is independent of the final state (stationary or death phase) of each culture. Additionally, measurements were always made during the second cycle in the assay environments because carry-over effects from growth in the ancestral environment were often observed during the first cycle.


Six kinds of experimental lines were assembled based on two treatments: diversity and mode of reproduction. Low diversity and high diversity lines were assembled by randomly sampling either two or 12 spores, respectively, without replacement, from the 40 base populations. Asexual lines were assembled by sampling spores from only one mating type. We assembled an equal number of asexual lines with spores of only mating type + as of asexual lines with spores of only mating type –. Sexual lines were assembled by sampling an equal number of spores from each mating type. During the sexual cycle, obligate sexual lines were exposed to chloroform that kills the vegetative cells but not the zygotes (Ebersold and Levine 1959), insuring that only the daughter cells resulting from meiosis were transferred to the next growth cycle. In the facultative sexual lines, chloroforming was not performed, meaning that daughter cells from both the vegetative and sexual cycles were possibly transferred to the next growth cycle. Each treatment combination was represented by 16 randomly assembled replicate lines, for a total of 96 experimental lines.

Experimental lines were cultured in 24-well plates with 2 mL of Bold's liquid medium and serially transferred using a 5% inoculum at the end of each cycle (every three to four days). Sterile breathable sealing films (AeraSeal, Fisher, Nepean, Ontario, Canada) were used to cover the plates to avoid uneven evaporation across the plate and cross-contamination from condensation that usually happens with the use of the plastic lids. The plates were placed in a growth chamber, with temperature set at 25°C, constantly shaken at 200 rpm (3-mm rotation diameter) and illuminated at 100 μE.


Each replicate line was exposed to two different environments: a constant environment without supplemented salt (control lines) and a deteriorating environment in which the salt concentration increased by 1 gL−1 every two growth cycles (selection lines). The experiment was terminated when all lines had gone extinct, which occurred at 33 gL−1 NaCl, after 66 growth cycles, or roughly 300 generations.

At each salt concentration up to and including 7 gL−1 we imposed two cycles of asexual reproduction and one cycle of sexual reproduction. We initiated the sexual cycle by transferring the cultures to nitrogen-free medium and incubating them in the light for about 24 h to allow gametogenesis and mating to occur. The asexual lines were not expected to mate during the sexual cycle because they contained only one mating type.

The plates holding asexual or facultative sexual populations were then wrapped in aluminum foil and stored in the dark for five days to allow the zygotes to mature. We then added 1 mL of Bold's medium to each well to stimulate zygote germination and incubated the plates in the light for 48 h (Wegener et al. 1989). Obligate sexual lines were spread on agar plates after mating, stored in the dark for five days, and then exposed to chloroform vapor to kill unmated cells before zygote germination was initiated by incubating in the light (Levine and Ebersold 1960). Sexual cycles were not induced after growth cycle 14 (7 gL−1 NaCl) because the populations had become too sparse to induce a mating reaction. Ten obligate sexual selection lines (three from the high diversity treatment and seven from the low diversity treatment) were lost because no zygotes were produced during the sexual cycle and consequently were discontinued.

The lines were axenic when established but became contaminated with bacteria and fungi, which had spread to all lines by cycle 33. Because most lines became contaminated well before the first extinctions were recorded, and all lines were contaminated well before the different treatments combinations had diverged in terms of their extinction dynamics, it appears that interactions possibly occurring among C. reinhardtii and the contaminants did not significantly affect the outcome of selection. The presence of fungi and bacteria may have affected the optical density readings, but had no effect on the estimation of cell density from colony counts.


Each line was visually inspected under a dissecting microscope before each transfer. The absence of living cells was recorded as an extinction event and was corroborated with the absence of growth in subsequent assays.


The control lines corresponding to the selection lines that had survived up to cycle 40 (20 gL−1 NaCl) were transferred to a series of salt concentrations, from 0 to 30 gL−1 NaCl, with 2 gL−1 increments. We recorded the proportion of control lines that were surviving (determined by colony counts on agar) after two cycles of growth in the assay environment, in parallel with the selection experiment, where the selection lines were exposed to each salt concentration for two growth cycles. This enabled us to compare the selection lines, in which selection for resistance to salt had occurred, to the control lines, in which selection had not occurred.


The fitness of the experimental lines was assayed at intervals of four cycles for most of the experiment, beginning at 4 gL−1 NaCl, and ending at 22 gL−1 NaCl, using reciprocal transplants. Before the start of each assay, the lines were transferred to the ancestral environment (without salt supplementation) for two cycles to acclimate the selection and control lines to the same environment. The lines were then inoculated into their home environment (salt or no salt) and into the alternative environment for two growth cycles, recording optical density during the second cycle.


At 24 gL−1 NaCl, the overall degree of adaptation of the surviving lines to salt-supplemented environments was determined by assaying the lines in a gradient of salt concentrations, from 0 to 36 gL−1 NaCl, with 4 gL−1 increments. Cell density was calculated by counting the number of colonies formed in a layer of soft agarose (0.9 gL−1) poured on top of agar; agarose was used because many of the selection lines lost the ability to grow on hard agar.


We tested if adaptation to high salt had occurred early (4 gL−1 NaCl), midway (8 gL−1 and 9 gL−1 NaCl), or late (24 gL−1 NaCl) in the experiment by reviving the stored populations of three obligate sexual-high diversity lines that had survived up to at least 24 gL−1 NaCl and assaying them in a gradient of salt concentrations, from 0 to 36 gL−1 NaCl.



A concentration of 7 gL−1 NaCl reduced growth in spores from the ancestral populations to 48% of the maximum optical density (mean OD = 0.59) in the first assay (mean OD = 0.28, SD = 0.060, n= 58) and to 58% of the maximum optical density (mean OD = 0.54) in the second assay (mean OD = 0.31, SD = 0.091, n= 58). These results are consistent with those of Moser and Bell (2011).


The among-spore variance component of growth decreased with salt concentration, but the coefficient of variation remained about the same (with values of 0.2–0.25) for concentrations between 1 and 7 gL−1. Hence, there was substantial genetic variation for salt resistance among the ancestral spores.


The history of the lines can be divided into four periods: an initial period when all survived, a first mass-extinction period, a stationary period, and a second mass-extinction period (Fig. 1A). No extinctions occurred during the initial lag period, which lasted 17 growth cycles, from 1 to 9 gL−1 NaCl. The first three extinctions occurred simultaneously during cycle 18 and were followed by a period of mass extinction lasting eight growth cycles (from 10 to 12 gL−1 NaCl) during which 42% of the selection lines became extinct. This corresponds to a probability of extinction of 0.054 per cycle. After this mass-extinction period, the probability of extinction decreased from 0.054 to about 0.013 per cycle for the next 27 cycles. A second period of mass extinction lasting nine growth cycles (from 26 to 30 gL−1 NaCl) then saw the extinction of 35% of the initial number of selection lines, at a rate of 0.10 per cycle. Only two selection lines survived this mass-extinction period: one was a facultative sexual-high diversity line, which went extinct when transferred to 32 gL−1 NaCl at cycle 63, and the second was an obligate sexual-high diversity line, which went extinct when it was transferred to 33 gL−1 NaCl at cycle 65 (Fig. 1B).

Figure 1.

Extinction dynamics. (A) The overall proportion of surviving lines over time. (B) The proportion of surviving lines for each treatment combination. Lines that went extinct for reasons other than salt stress were not included in the calculation of the proportion.


The regression of growth on salt concentration, over the linear part of the range between 4 and 24 gL−1 was significantly shallower for the selection lines than for the control lines (ANCOVA, F1,179= 16.03, P < 0.001 for assay × selection environment interaction; Fig. 2A). The selection lines had consistently higher growth over the whole of this range, and their advantage increased with salt concentration.

Figure 2.

Linear regressions showing overall response to salt stress when assayed late in the experiment of (A) the control lines (empty) and the selection lines (filled); (B) each treatment combinations (asexual: triangles; facultative sexual: squares; obligate sexual: circles; high diversity: empty; low diversity: filled), using only the selection lines. The treatment combinations are ordered in the legend so as to match their rank at 24 gL−1 NaCl.


The regressions of the selection lines were analyzed separately to identify differences among treatments (Fig. 2B). There is significant variation among the slopes of the different treatment combinations (ANCOVA, F2,91= 5.02, P= 0.008), reflecting variation among degrees of sexuality (ANCOVA, F2,91= 5.32, P= 0.006) and among levels of diversity (ANCOVA, F1,91= 8.53, P= 0.004). More specifically, the decline in cell density is significantly more gradual in obligate sexuals (mean slope =– 0.30, SD = 0.066) than in asexuals (mean slope =– 0.47, SD = 0.059; Tukey's HSD, P adj = 0.001), and more gradual, although not quite significantly, than in facultative sexuals (mean slope =– 0.38, SD = 0.17; Tukey's HSD, P adj = 0.07). The decline in cell density is also more gradual in facultative sexuals than in asexuals, although not quite significantly (Tukey's HSD, P adj = 0.08). In terms of diversity, the decline in cell density is more gradual in high diversity lines (mean slope =– 0.32, SD = 0.11) than in low diversity lines (mean slope =– 0.44, SD = 0.12; Tukey's HSD, P adj = 0.002). Overall, the two treatment combinations showing the most gradual decline in cell density with increasing salt concentration are the facultative sexual-high diversity lines and the obligate sexual-high diversity lines. The two treatment combinations showing the steepest decline in cell density over salt concentration are the facultative sexual-low diversity lines and the asexual-low diversity lines. The treatment combinations with a more gradual decline in cell density over salt concentration have a higher degree of adaptation in high salt concentrations and a lower degree of adaptation in lower salt concentrations compared to the treatment combinations with a steeper decline of adaptation.


One treatment combination stands out from the others in terms of its extinction dynamics. The obligate sexual-high diversity treatment combination had a higher proportion of surviving lines than the other treatment combinations, and alone experienced no extinctions during the stationary phase (Fig. 1B). At the end of the stationary period (cycle 51), immediately before the second mass-extinction period, the combination of obligate sexuality and initial high diversity led to a statistically significant lower risk of extinction compared to the expected risk, as calculated using all the other treatment combinations (exact binomial test: proportion of obligate sexual-high diversity lines surviving = 0.62; expected = 0.32; P= 0.03). None of the other treatment combinations had a significantly different risk of extinction than expected (asexual-high diversity, asexual-low diversity, facultative sexual-high diversity, facultative sexual-low diversity: estimate = 0.31 for all these treatment combinations, expected = 0.38, P= 0.8; obligate sexual-low diversity: estimate = 0.37, expected = 0.36, P= 1). The half-life of the obligate sexual-high diversity lines was 53 cycles, compared with 28.5–33 cycles for the other treatment combinations.

In the assay of the surviving lines, sexual and high-diversity lines were disproportionately represented among those surviving extreme stress. Five lines survived at 28 gL−1 NaCl: one asexual-low diversity, two facultative sexual-high diversity, one obligate sexual-low diversity, and one obligate sexual-high diversity. Only two selection lines survived the second mass-extinction period: one was a facultative sexual-high diversity line, which became extinct when transferred to 32 gL−1 NaCl at cycle 63, and the second was an obligate sexual-high diversity line, which went extinct when it was transferred to 33 gL−1 NaCl at cycle 65.


At 4 gL−1 NaCl, the asexual lines had a greater direct response to selection than both the facultative sexual lines and the obligate sexual lines (F2,83= 5.05, P= 0.008; Fig. 3A). Furthermore, the low diversity lines had a greater direct response to selection than the high diversity lines (F1,83= 8.37, P= 0.005). Neither of the treatments had a significant effect on growth of the selection lines at 6 gL−1 NaCl. The reverse pattern was observed at 8 gL−1 NaCl: the obligate sexual lines had a greater direct response to selection than both asexual and facultative sexual lines (F2,80= 3.56, P= 0.03), and the high diversity lines had a greater direct response to selection than the low diversity lines, although this was not formally significant (F1,80= 2.85, P= 0.1; Fig. 3B).

Figure 3.

Effects of mode of reproduction and diversity on the direct response to selection, as measured during the reciprocal transplants at (A) 4 gL−1 and (B) 8 gL−1 NaCl. The main horizontal line in each panel indicates the mean direct response to selection of all treatment combinations.


Obligate sexual-high diversity populations from early in the experiment (4 gL−1 NaCl) have a significantly lower degree of adaptation across the salt gradient than populations from late in the experiment (t=–3.63, P < 0.001; Fig. 4). None of the lines from early in the experiment survived concentrations higher than 20 gL−1 NaCl. The degree of resistance midway through the experiment is lower than late in the experiment, although the difference in not significant (8 vs. 24 gL−1 NaCl: t=– 0.76, P= 0.4. 9 vs. 24 gL−1 NaCl: t=–1.03, P= 0.3).

Figure 4.

Linear regressions showing overall response to salt stress of obligate sexual-high diversity lines when assayed early (4 gL−1), midway (8 gL−1, 9 gL−1), and late (24 gL−1) in the experiment. 4 gL−1: circles; 8 gL−1: pluses; 9 gL−1: triangles; 24 gL−1: crosses.


The extinction schedule of the selection lines can be compared with the expected outcome without selection for resistance to salt, from the assay of control lines (Fig. 5). The fraction of control lines surviving declines continuously with increasing salt concentration, dropping below 50% at 16 gL−1 and below 10% at 24 gL−1 NaCl. In contrast, all the corresponding selection lines had survived to 16 gL−1 and 80% survive at 24 gL−1 NaCl.

Figure 5.

Proportion of lines surviving high salt concentrations when evolved in the presence of selection for resistance to salt (filled circles) and in the absence of selection for resistance to salt (empty circles).



The goal of this study was to determine if natural selection could enable populations to survive in an environment that its ancestors could not tolerate. Even though we took all the lines to extinction such that no single line survived the entire spectrum of salt concentrations, we detected two clear events of mass extinction. We take these events as an indication of genetic hurdles requiring genetic adaptation for survival. Hence, we refer to the survival of lines past a mass-extinction event as an event of evolutionary rescue. We detected the occurrence of two events of evolutionary rescue, following each of the two mass-extinction events. The first event of evolutionary rescue is most likely attributable to sorting of the genetic variation initially contained in each line and to selection on variation generated mainly by recombination. The second event of evolutionary rescue, which led to the survival of only two lines, may have been attributable to selection on variation that was generated by mutation, enabling the lines to survive past stress levels that were lethal to the ancestral population. The salt concentrations experienced by the experimental lines in this experiment span the whole range from fresh water conditions to marine conditions. The evolutionary rescue events that we observed provide evidence that major shifts in ways of life can arise within short time frames, through the action of natural selection.

A number of studies have investigated how many and which genes might be involved in resistance to NaCl in Chlamydomonas. Based on the mutation frequency observed using insertional mutagenesis, Prieto et al. (1996) estimated that 10–12 loci in the genome of Chlamydomonas are involved in NaCl and osmotic stress resistance. The genes they found were most likely linked to general osmotic homeostasis and reduction of ion toxicity by regulation of the transport of ions through the cell membrane. Using functional expression screening, many more genes from a halotolerant species of Chlamydomonas have been identified in regards to NaCl and oxidative stress tolerance. Miyasaka et al. (2000) identified 35 homologues to genes having already been identified, either with or without a link to salt or oxidative stress resistance, and 55 unknown genes. Among the genes that were identified based on NaCl-resistance screening only, there is glutathione peroxidase, Bbc1 (breast basic conserved), alternative oxidase (AOX in mitochondria), and a number of ribosomal and photosystem proteins. Other proteins have since been identified, in particular one that is homologous to a late embryogenesis abundant (LEA) protein, which is highly hydrophilic and protects against desiccation during osmotic stress (Tanaka et al. 2004). The implication of many genes in the regulation of NaCl stress suggests that resistance in this experiment may have been attributable to the fixation of beneficial alleles at more than one locus.


Evolutionary rescue in a deteriorating environment is facilitated by positive genetic correlations of survival at lethal stress with survival at sublethal stress. If this were not the case, specific adaptation to every level of deterioration would need to occur as the environment was changing, and populations would become extinct if the environment deteriorates faster than beneficial alleles can spread (Samani and Bell 2010). Genetic correlations in C. reinhardtii between environments differing in the concentration of NaCl have been measured previously (Moser and Bell 2011). In this experiment, most of the adaptation to salt concentrations that were lethal to the ancestral population had occurred in the obligate sexual-high diversity lines by the time the experiment reached about 8 gL−1 NaCl, which corresponds to the time sexual cycles were stopped. This suggests that the combination of alleles—generated by recombination—that enabled the lines to survive the first mass-extinction period also enabled the lines to survive more stressful conditions.


Initially, low diversity enhanced adaptation, most likely because a lower degree of clonal interference allowed superior types to spread more rapidly. In the high diversity lines, competition between resistant genotypes could have interfered with the progress of any particular beneficial allele toward fixation, whereas competition would have been much weaker in the low diversity lines. Subsequently, it was high diversity that enhanced adaptation and lowered the risk of extinction. Here, the greater range of variation enhanced the probability of evolutionary rescue because it increased the probability of there being a resistant type that could spread through the population and restore growth. As salt concentration increased and population size fell, the effects of clonal interference may also have been reduced (Perfeito et al. 2007), enhancing the advantage of high diversity.


Lack of sexuality initially enhanced adaptation, perhaps because genotypes that were beneficial at zero supplemental salt were also beneficial at low salt. Recombination has been shown to reduce the fitness of sexual lines by breaking down associations between loci (recombination load) in Chlamydomonas (Colegrave et al. 2002; Kaltz and Bell 2002), in rotifers (Becks and Agrawal 2011), and in yeast (Greig et al. 1998). The lower degree of adaptation initially in the obligate sexuals is also possibly attributable to a reduced effective population size because chloroforming, by killing all vegetative cells, greatly reduced the inoculum size. Subsequently, it is sexuality that enhanced adaptation and lowered the risk of extinction. This may be attributable to the generation of a broader range of selectable variation by recombination in repeated sexual cycles early in the experiment. The effect of sexuality was lesser in facultative sexual lines than in obligate sexual lines early in the experiment. This may have been attributable to the dilution of meiospores by the vegetative growth of unmated spores before germination of the zygotes in the facultative sexual lines. The cost of the time needed for the zygotes to germinate would have lowered the net advantage of sexuality in the facultative sexual lines, but not in the obligate sexual lines where germinating zygotes were not competing with vegetative cells.


The combination of obligate sexuality and high diversity was most effective in enhancing survival as the environment deteriorated. This is because sex is only effective in the presence of genetic variation (Greig et al. 1998). When genetic variation is low, recombination does not generate much greater variation among the offspring than there was among the parents, and therefore does not greatly increase the probability of generating types that are resistant to the stress. The level of replication in our experiment (16 replicates of each treatment combination) was too low for us to estimate the strength of this effect accurately. Nevertheless, the empirical evidence presented here is in agreement with the proposition that the combination of obligate sexuality with high diversity greatly reduces the risk of extinction during environmental deterioration by increasing the rate of response to selection and the degree of adaptation to the stress.

To the best of our knowledge, this is the first experiment to demonstrate the extinction dynamics of populations exposed to environmental conditions that span an entire ecological spectrum, in this case from freshwater conditions to marine conditions. Although most Chlamydomonas spp. inhabit freshwater, some have been isolated from marine environments and shown to be tolerant to high salt concentration (Hellebust 1985; Miyasaka et al. 1998; 2000) suggesting that the ecological transition has occurred in the past. A similar transition might occur as the result of the salinification of ground water, marine intrusions, and extreme high tides, or even salt runoff from roads in recent times. We recorded the occurrence of two episodes of evolutionary rescue associated with two clear episodes of mass extinction. The dynamics observed in this experiment are likely to be specific to the organism and the stress that we used, but they suggest that there might sometimes be more than one genetic hurdle preventing a population from adopting a new way of life and that surmounting these hurdles is facilitated by a high level of sexually generated variation. This is consistent with the general observation that secondarily asexual lineages of plants and animals have on average a shorter life span than sexual lineages. It also supports the idea that the longer life span of sexual lineages is attributable to their ability to respond to selection more effectively and thereby to adapt to a deteriorating environment more rapidly.

Associate Editor: S. West


We thank P. Homme and G. A. Meyer for laboratory assistance, and two anonymous reviewers for comments. This work was funded by the Natural Sciences and Engineering Research Council of Canada (JL and GB).