Chlamydomonas (Chlorophyta) can grow as a heterotroph on medium supplemented with acetate in the dark. A long-term experiment to investigate adaptation to dark conditions was set up with hundreds of replicate lines. Growth was initially slow, and most lines became extinct when transferred every few weeks. Some lines survived through the expansion of lineages derived from cells with extreme phenotypes and exhibited a U-shaped curve of collapse and recovery. Two short-term experiments were set up to evaluate the effect of sex on the frequency of ‘evolutionary rescue’ by deriving replicate lines from ancestral populations with contrasting sexual histories that had been cultured in the light for hundreds of generations. When transferred to dark conditions of growth, lines derived from obligately sexual populations survived more often than lines derived from facultatively sexual or asexual populations. This reflected the higher initial frequency of cells able to grow in the dark, due to greater genetic diversity supported by sexual fusion and recombination. The greater probability of evolutionary rescue suggests a general reason for the prevalence of sexual species.
A population whose environment deteriorates will tend to adapt through natural selection as it decreases in abundance. It may nevertheless become extinct, especially if the deterioration is rapid and severe. In some circumstances, the population will persist, after an initial period of decline, because it has become adapted to conditions lethal to its ancestor. This is the phenomenon of ‘evolutionary rescue’. This paper describes how experimental populations confronted with a sudden extreme stress either adapted or became extinct. The populations consist of a few thousands of cells of a unicellular green alga, Chlamydomonas, sampled from a larger population maintained for many years in the laboratory under standard conditions of photoautotrophic growth on mineral medium. The stress is transferred to complete darkness on medium containing sodium acetate, which supports a low level of heterotrophic growth. Regular transfer to fresh medium dilutes the growing populations, in proportion to the frequency of transfer, so that they die out if their rate of growth is less than the rate of dilution. All populations will tend to adapt to life in the dark, provided that there is some genetic variation in the rate of growth in heterotrophic conditions. This is not sufficient to ensure survival, however, unless the range of variation encompasses the types whose rate of growth exceeds that necessary for replacement. Evolutionary rescue requires not only that there should be some variation in relative fitness but also that the range of variation in absolute fitness should include types with positive rates of increase for the combination of stress and dilution imposed on the population.
Whether or not a population survives a sudden stress depends primarily on the range of variation it currently expresses . A sample from the population will subsume a larger or smaller part of this range, in proportion to its size. When exposed to the stress, its composition will become biased in favour of more resistant types. In some cases, these samples will include types whose growth rate exceeds dilution, and will thereby found long-lasting populations; in other cases, no type will be able to replace itself, and the population will dwindle to extinction. The successful populations will expand as rapidly growing lineages spread, until their growth rate is equal to the dilution rate. If they continue to evolve, through the spread of more efficient or competitive lineages, their overall growth may temporarily exceed the rate of dilution, although it is unlikely that growth will exceed dilution by very much or for very long.
In the longer term, mutation and recombination may assemble genotypes that can persist in a gradually deteriorating environment [2–6], but such processes will not have sufficient time to act when the population is abruptly challenged by a potentially lethal stress. The fate of a population will then depend on its history. In particular, it will depend on its sexual history, because this will govern current levels of variation. Asexual populations are likely to be genetically uniform because they will be dominated by the single most successful clone. Sexual populations will tend to be more variable because recombination obstructs the fixation of multilocus genotypes. The effect of sex will be less when mating and recombination are less frequent relative to vegetative reproduction, or when mating occurs between close relatives. Evolutionary rescue should therefore be most effective in outbred sexual populations and least effective in asexual populations, with inbred or facultatively sexual populations being intermediate. The purpose of the experiment reported here is to test whether the sexual history of a population affects the likelihood of evolutionary rescue in the expected way.
2. Material and methods
(a) Base populations
Ancestral populations were established from a compound cross among wild-types and subsequently propagated on solid Bold's agar medium (see  for recipes and general procedures) for several thousand vegetative generations over a period of 12 years. These populations and procedures are described in more detail by Bell . Each population was transferred in one of four ways.
(1) Obligate large sexual populations (three replicate lines) were founded from a mixture of both mating types and subsequently allowed to grow and mate on the agar surface in each cycle, after which they were re-spread, incubated in the dark to allow zygotes to mature, exposed to chloroform vapour to kill unmated cells and then incubated in the light to stimulate zygote germination and subsequent vegetative growth. At least 200 zygotes were transferred in each cycle.
(2) Single zygote ‘bottlenecked’ obligate sexual populations (two replicate lines) were transferred in the same way, except that a single zygote alone was chosen at random to propagate the line. Chlamydomonas reinhardtii is a vegetatively haploid species whose sexual system is heterothallic, with two non-switching mating types; consequently, the diploid zygote is heterozygous for mating type and the F1 is a segregating generation, producing haploid zoospores of both genders that will also vary with respect to other segregating loci.
(3) Facultatively sexual populations (two replicate lines) were established from a mixture of mating types but subsequently transferred without enforcing a sexual cycle. From visual surveys it is thought that mating is very infrequent in these lines, if it occurs at all, because zygotes have not been observed for several years.
(4) Asexual populations (two replicate lines) were established from spores of the same mating type and never enter the sexual cycle.
(b) Dark lines
Dark lines were established by transferring samples of these populations to agarose medium supplemented with 1.2 g l−1 sodium acetate and incubated in 96-well microwell plates in the dark. Two series of experiments are reported here.
(1) The long-term experiment was set up with 960 replicate lines from each of the three outbred sexual ancestors. They were initially transferred at monthly or longer intervals with 25 µl of inoculum in a culture volume of 200 µl. The initial history of these lines is described here; their subsequent adaptation to dark growth over greater than 50 growth cycles will be analysed in a subsequent report.
(2) The short-term experiment was set up from all four sexual treatments and transferred by an inoculum of 10 µl every two weeks. Twenty replicates per line were propagated for the obligate outbred sexual treatment and 10 replicates per line for the other three treatments, comprising 120 selection lines in all. These lines were followed for 20 growth cycles.
Both experiments were scored by recording optical density (OD) at 665 nm immediately after inoculation and immediately before transfer in each growth cycle: the difference between the two is used as an estimate of growth in OD units. Extinction of lines was confirmed by microscopy.
(c) Initial state
To evaluate the initial state of the base populations with respect to dark growth, samples from the two outbred, two bottlenecked and two asexual populations were suspended in liquid medium, thin-spread on acetate plates (two replicates each) and incubated in the dark for three weeks. Most cells died; the small minority of putative viable cells formed small pale colonies. To confirm viability, the plates were incubated in the light for 4 days and the number of green colonies scored. At the same time, a corresponding set of plates was spread from the same suspensions and incubated in the light to give an estimate of total cell density.
(d) Statistical analysis
The frequency of rescue was analysed by logistic regression using the R statistical package. Raw data are available as electronic supplementary material.
(a) Long-term experiment
In the long-term experiment, most lines become extinct within the first 10 growth cycles. The average growth of these lines falls exponentially towards zero. The growth of the surviving lines at first falls at about the same rate, but then recovers in a U-shaped curve until the rate of growth necessary for long-term persistence has been attained (figure 1). This rate of growth was maintained for the next 50 cycles.
(b) Short-term experiment
In the short-term experiment, the frequency of surviving lines varied among treatments (χ2 = 13.7, d.f. = 3, p < 0.001). Survival was greatest for the obligate outbred sexual lines (74%); then for the obligate bottlenecked sexual lines (65%); then for the facultatively sexual lines (25%); with the lowest rate of survival (10%) among the asexual lines (figure 2). Logistic regression comparing the asexual lines with the others showed that they differed significantly from the obligate outbred lines (z = 4.25, p = 0.00002) and from the obligate bottlenecked lines (z = 3.20, p = 0.0014) but not from the facultative lines (z = 1.21, p = 0.2). Other comparisons divide the treatments into two groups, such that all comparisons between groups are significant and all comparisons within groups are non-significant: one group comprises the obligate sexual lines, with the higher rate of survival, and the other comprises the facultative and asexual lines, which have the lower rate of survival.
(c) Initial state
The frequency of viable cells in samples from the base populations, summed over replicates, was 3 in 113 (2.6%) and 1 in 188 (0.5%) for the two asexual populations; 9 in 258 (3.5%) and 7 in 173 (4.0%) for the two bottlenecked populations; and 37 in 1036 (3.6%) and 27 in 402 (6.7%) for the two outbred populations. Hence, most cells were incapable of growth in the dark. The variation among treatments is not significant for the arcsin square-root transformation (F = 2.7; d.f. = 2,3; p > 0.05), but pooling colony number for asexual and sexual lines under light and dark conditions in a 2 × 2 contingency table does suggest a significant difference (χ2 = 5.7, d.f. = 1, 0.025 > p > 0.01). Moreover, the frequency of cells able to grow in the dark in asexual, bottlenecked and outbred treatments is correlated with the frequency of rescue among lines (r = 0.815, d.f. = 4, p = 0.05).
The initial phase of the long-term experiment illustrates the process of evolutionary rescue. The ancestral populations contain a range of phenotypes that enable any replicate line to adapt to the dark. Adaptation is only permanent, however, if this range includes types whose absolute fitness is large enough to permit indefinite growth. Many populations adapt but nevertheless become extinct. This aspect of evolution has seldom been emphasized, although its roots extend back to Haldane's concept of the cost of natural selection . The U-shaped pattern of collapse and recovery was predicted theoretically by Gomulkiewicz & Holt  and documented for a yeast–salt system by Bell & Gonzalez . The replicate lines in the long-term experiment illustrate both the exponential collapse of doomed populations and the U-shaped recovery of rescued populations.
The ordering of lines with respect to evolutionary rescue in the short-term experiment was: obligate outbred > obligate bottlenecked > facultative > asexual. This suggests that outbred populations are the most, and asexual populations the least likely, to survive an abruptly imposed stress. One interpretation is that the sexual populations, for unknown reasons, had a uniformly greater initial capacity for dark growth than the asexual populations. In this case, most cells from all populations would grow, but those from the sexual populations would grow more. An alternative interpretation is that the sexual populations were initially more variable with respect to dark growth. This variability is unlikely to have arisen during the course of the short-term experiment itself, because all populations were propagated vegetatively during the experiment. With no attempt to enforce mating, there was little if any variation generated by recombination during the course of this experiment, although some low level of successful mating cannot be ruled out. Some variation may have been generated by mutation, but the effective population size is small (Ne ≈ 4000) and few beneficial mutations are expected; in any case, the number of mutations would be the same, on average, for each line. Most of the variability in dark growth is therefore likely to reflect variability already present in the ancestral populations. In this case, most cells would be inviable in the dark, but the extreme variants capable of growth would be more frequent in the sexual populations. The outcome of the screens of the ancestral populations supports the second interpretation. The difference in response between treatments is therefore likely to be attributable to the quantity of standing genetic variation for dark growth, greatest in the outbred lines and least in the asexual lines.
Alleles that enhance dark growth are likely to be neutral or slightly deleterious, since the ancestral populations have been maintained for thousands of generations in the light. They will accumulate through drift, or will be maintained at mutation–selection equilibrium. From time to time, a beneficial mutation conferring greater fitness in the light may arise and spread. In an asexual population, any such mutation is necessarily linked to all the alleles at other loci in the genome where it originally appears. Hence, its spread will eliminate genetic variation, including variation for dark growth, through the fixation of a single allele at every locus (‘periodic selection’: [11,12]). Such genome-wide selective sweeps will not occur as readily in sexual populations because recombination transfers the initial mutation onto different genetic backgrounds. Recombination is less effective in bottlenecked populations, because they are less diverse and because sexual partners are genetically correlated, and in facultative populations, where recombination is less frequent, if it occurs at all. This argument is not necessarily decisive, because a beneficial allele that sweeps through an asexual population may carry with it a number of mutations that are slightly deleterious in the light but may be beneficial in the dark . This depends on the frequency of sweeps and the genetic correlation between light and dark growth. The suppression of genome-wide selective sweeps by sexual fusion and recombination, however, is the simplest explanation for the greater variability in the outbred populations with respect to dark growth.
This variability confers greater evolutionary potential when conditions change and underlies the high frequency of rescue in the outbred populations. It also offers a very simple explanation for the maintenance of sexuality: if populations repeatedly experience severe stress, sexual populations are more likely to be rescued whereas asexual populations are more likely to become extinct. Several experiments have shown that sexual populations tend to adapt more rapidly than comparable asexual populations to novel and stressful conditions of growth [14,15], provided that they are sufficiently large  and genetically variable . This might create a long-term advantage for sex by favouring the competitive replacement of asexual by sexual populations. The experiments reported here extend this interpretation to the situation where absolute fitness drops below replacement level and asexual populations are less likely to recover. This is consistent with the paucity of ancient obligately asexual clades in natural communities [18,19].
The experiments were transferred and monitored with great care by Kathy Tallon. The study was funded by the Natural Science and Engineering Research Council of Canada.
One contribution to a Special Feature ‘Experimental evolution’ organized by Paul Sniegowski, Thomas Bataillon and Paul Joyce.
- Received August 30, 2012.
- Accepted October 3, 2012.
- © 2012 The Author(s) Published by the Royal Society. All rights reserved.