Sex has at least a two-fold cost. Many different approaches have been taken to explain why sexual genetic systems persist in spite of this. Reviewed here are the DNA damage-repair hypothesis; simultaneous fixation of several favourable mutations at different loci; purging of deleterious mutations and Muller's ratchet or mutational meltdown; ecological hypotheses about heterogeneity in space and time, sibling competition, and diversification as a habitat becomes saturated with species; and the parasite-host coevolution hypothesis. None of these explain the empirical evidence fully. I conclude that several of the proposed selection pressures may contribute to the evolution and maintenance of sex.
Natural selection is by now a well-known phenomenon. It operates where the following three conditions are met by populations of organisms: That there is variance between at least some individuals in the population, the peculiarities of each individual being passed on to at least some of its offspring, and finally, these peculiarities must influence the number of offspring of an individual. If we generalise these preconditions and define them as variability between information units, continuity of their properties through generations and differential reproduction, we arrive at a wider range of systems that are governed by that same selective mechanism. Any such system is a complex adaptive system, and quite a range have come to light (Gell-Mann 1994).
Among them is the genetic system that has produced, among many others, the human species. One integral and remarkable aspect of that system is the fact that it is sexual. This remains one of the great mysteries of evolutionary biology.
The purpose of this essay is to outline the adaptive significance of sexual and asexual reproductive systems. Because the debate on this topic persists, I can only discuss hypotheses that have been suggested as explanations of the predominance of sexual genetic systems in higher eukaryotes, inevitably also discussing the advantages and domains of asexual systems. When I speak of sexual reproduction in this essay, I mean anisogamous amphimixis. Further, I will mostly be thinking of dioecious and gonochoristic species. Asexual reproduction will refer to apomixis.
The reason why scholars have found it hard to explain the prevalence of sex is its cost. John Maynard Smith (1971) first pointed out that there is a 50% loss of reproductive output in sexual populations relative to asexual ones. As only half the population are producing eggs, only half of which will hatch as females and in their turn produce eggs, any sexual population can only reproduce at half the rate of an asexual equal. Maynard Smith concludes that the advantage of sexuality must be enormous to outweigh this high cost. To justify sexual reproduction, any daughter produced in this way should have to be twice as fit as one produced by a parthenogenetic mother. This argument assumes that the number of eggs laid and their chance of survival are independent of whether the female is parthenogenetic (e.g. where the male does not care for his offspring). The assumption probably holds true for most taxa. There are also likely to be further costs in some species, such as those that arise from the necessity to find a mate (Stearns & Hoekstra 2000). Sessile organisms lose a certain proportion of their gametes because they do not get fertilised due to the dilution in their dispersion medium. However, heterogonic species, those that can reproduce both sexually and asexually, are no rarity, and G. C. Williams (1975) has suggested that both modes of reproduction must be similarly beneficial to be maintained in the lifecycle, and each may be appropriate in particular circumstances. There is an abundance of hypotheses as to what this advantage of sexual reproduction might be, and it is suggested that it must be an advantage to the individual, as individuals rather than groups of individuals are the unit of selection. Group selection, of course, does work in theory, but it is likely to be much weaker than selection on individuals, and there is a general consensus among biologists that no gene can invade a population if advantageous only to the group, not the individual (Gell-Mann 1994). However, a sexual population may be resistant to invasion of parthenogenetic individuals on group selectionist grounds if the rate at which asexual females evolve is very slow (Ridley 1996).
The fact that sexual reproduction necessitates a diploid life stage has led some authors to suggest that a property of the diploid stage may be a reason for sex to evolve. They suggest that repair of lesions and breaks (summarised as damage) in chromosomes may be enhanced by the presence of a second chromosome of the same kind (Bernstein et al. 1981, 1988, Bernstein & Bernstein 1991). An advantage of this hypothesis is that it allows sex to be immediately beneficial to the individual. But these benefits are conveyed by any genetic system involving diploid or polyploid individuals, including parthenogenetic individuals. Along another line of investigation, Cleveland pointed out in 1947 that possibly the alternation of meiosis and a subsequent doubling of chromosome number could have evolved before syngamy, after observing a regular endomitosis-meiosis cycle in the primitive flagellate Barbulanympha. Further to this, another flagellate, Pyrsonympha, exhibits several cycles of endomitosis, followed by a series of reductional divisions (Hollande & Carruette-Valentin 1970). Margulis & Sagan (1986) picked up this idea and argued that each ploidy phase in the life cycle might be adapted to different external conditions, and indeed haploids and diploids do differ in ways that may represent adaptations. This evidence further suggests that perhaps syngamy evolved from a lifecycle that already consisted of alternating haploid and diploid phases (Maynard Smith & Szathmáry 1995), discounting the possibility of syngamy (and hence sex) evolving in concert with meiosis, and thus the DNA repair hypothesis of Bernstein et al.
To find the reason for sex, we must investigate the effects of meiosis and syngamy, which are its defining characteristics, but absent from parthenogenesis. The recombination characteristic of meiosis generates and multiplies genetic variation in the population by crossing over. This has diverse effects.
R. A. Fisher (1930) saw that evolution can proceed faster in sexual populations if only the rate at which favourable mutations appear in the population is great enough. If we imagine that the finite range of possible beneficial mutations form a hierarchy determined by the magnitude of their fitness advantage, we perceive that if they all occur at about the same time in an asexual population, only one of them can become fixed (the one giving the greatest fitness advantage) while all others are purged, then another (whichever one of the next number of mutations has the highest fitness) and so on. In sexual populations, those mutants with very good mutations will interbreed with those possessing slightly inferior, but still beneficial, ones, so that good alleles accumulate in the population more quickly. However, this argument applies only to populations in which the rate at which favourable mutations arise is high enough so it is likely that once one of the mutations has occurred, the other will do so in an individual without the first mutation before the first mutation has become fixed in the population (Crow & Kimura 1965).
At the same time, deleterious mutations are more easily purged from sexual populations, especially if double mutants are more deficient in fitness than predicted from multiplying the fitness values of single mutants, that is, if the deleterious effects interact. Fewer genetic deaths will be required, and with a sufficiently high mutation rate, this gives a short-term advantage to genes for sex and recombination (Kondrashov 1982, 1993, Feldman et al. 1980). Asexual populations may furthermore accumulate deleterious mutations by drift. Muller's ratchet moves forward metaphorically as the classes of individuals with the lowest number of mutations at each point in time in an asexual population consecutively become eliminated by genetic drift (Muller 1964). This process, tending towards extinction of the asexual population, is termed mutational meltdown (Lynch et al. 1993). The standard deviation of the number of deleterious mutations is also greater in sexually reproducing populations than in asexual ones. This makes it possible to eliminate them more easily in sexual populations because more individuals will be more seriously impaired by expressing several deleterious mutations.
The flip-side of this coin is that recombination can also unlink adaptive combinations of genes, thus decreasing fitness. If AB and ab phenotypes are most adaptive and both occur in a population, recombination will produce individuals with the maladapted Ab and aB phenotypes. In this situation, newly evolved parthenogenetic individuals of genotypes AB and ab have a huge selective advantage, and the asexual system will become fixed in the population (Futuyma 1997).
Brooks (1988) has reviewed the abundance of evidence for recombination frequencies under genetic control. The effects of this genetic control can be fairly localised, affecting particular loci. Likewise, artificial selection can produce parthenogenetic taxa. This is more evidence supporting the idea that there must be short-term benefits to recombination and sexual reproduction as the evolution of lower recombination and parthenogenesis appear to be selected against.
Some hypotheses on the benefit of genetic variation consider environments heterogeneous either in time or space, or both. The recombining of different alleles may be beneficial when a polygenic character is under stabilising selection and the optimum of this selection varies with environmental fluctuations (Maynard Smith 1980). This applies similarly to spatially heterogeneous habitats (Ghiselin 1974, Williams 1975, Bell 1982). If the distribution of niches occupied by individuals of a species should not follow a Normal distribution, though, asexual reproduction will be favoured, since then each niche can be occupied by as many copies of a particular genotype as the niche can hold. In this situation, parthenogenesis can spread through the population and extinguish sexual individuals (Case & Taper 1986). G. C. Williams refined one interesting hypothesis from this debate, namely that sexual females have a greater chance of producing at least some offspring that are going to be very well suited to the habitat they come to live in, because they are all different. An asexual female's offspring can only succeed in the habitat their mother was adapted to (Williams 1975). Others have pointed out that as members of a clone, such as asexual siblings, are adapted to using the resources in the same way, they are more likely to compete and interfere with each other, whereas sexual siblings may use resources differently even in a fairly homogeneous habitat (Bell 1982, Price & Waser 1982). Ghiselin extends this to argue that species must diversify as their habitat becomes saturated, and he equates sex with diversification (Trivers 1983).
W. D. Hamilton (1980, 1993, but see Jaenike 1978) is the most prominent proponent of a hypothesis according to which parasite-host coevolution changes both the parasite and host environments at such a high rate that sex becomes advantageous in the short term. This is because recombination generates new gene combinations or new genes altogether in the host that the parasite is badly adapted to. Segregation and recombination can also regenerate genes that have been purged due to negative frequency-dependent selection, or if recessive, they may be unmasked by recombination. Similarly, the parasites whose offspring have a high degree of variation will be selected for as some of them will be suited to the new environment provided by the new host genotypes. In some species, pairs of genes are identifiable between host and parasite which should be as compatible as possible for the sake of the parasite's fitness, but as different as possible from the host's point of view. In a gene for gene situation, where one parasite protein is adapted to interact with one host protein, such mechanisms have been experimentally shown (Seger and Hamilton 1988).
On the other hand, sexual populations are also vulnerable to genetic parasitism, such as meiotic drive. This gives a tendency for sexual populations to become extinct more often (Stearns & Hoekstra 2000).
The best way to confirm all of these thoughts is to turn to nature and see whether the distributions of sexual and asexual species agree with our predictions, and whether the two tend to differ in any significant ways. Firstly, sex is common in both animals and plants, whereas most taxa that reproduce parthenogenetically are of recent evolutionary origin, and polyphyletically distributed through animal and plant kingdoms, suggesting they are successful in the short term but tend to become extinct rather sooner. Bell (1982) says that asexual species are more often found in freshwater than marine environments, in small bodies of water rather than large, and at high latitudes rather than the tropics. He generalises that asexual reproduction is common in novel and disturbed habitats. This is support for Ghiselin's hypothesis, showing that asexual reproduction is indeed more useful in environments of abundant opportunity, where only the rate of reproduction counts for survival.
The parasite-host coevolution hypothesis is supported by empirical studies showing that sexual species have higher parasite pressure than asexual species in taxa containing both (Ridley 1996). On the other hand, these results could be interpreted simply as sexual species being more susceptible to parasites. Clarification is needed from parasite species which use both a sexual and an asexual host, preferably in the same habitat and at similar densities. Other predictions of the hypothesis include that parasites be found in most sexual species and that cycles of allele frequencies occur in nature, and neither of these can be confirmed yet.
Interestingly, the debate about the cause of sex is far from resolved, and I agree with West, Lively and Read (1999) in anticipating that what we are looking for, amidst all the plausible hypotheses, is perhaps a pluralistic explanation since none of the hypotheses stand out as being able to account for a two-fold average fitness advantage of sexual females over asexual ones.
As a conclusion, properties of the genetic system within which evolution occurs are themselves under selection for their effects. Mutation rates, recombination rates and which kind of sexual or asexual scheme is being followed all can change given the right conditions, and the concept of natural selection, though possibly given a different name, becomes recognisable as a universal function linking all kinds of complex adaptive systems. Furthermore, whilst the origin, and cause of evolution, of sex are not known, the sister phenomenon to natural selection, genetic drift, is acknowledged with a possible role in this causation.
Bernstein, H., Hopf, F. A. & Michod, R. E. 1988. Is meiotic recombination an adaptation for repairing DNA, producing genetic variation, or both? In: Michod, R. E. & Levin, B. R. (eds.), The evolution of sex: An examination of current ideas. Sinauer Associates, Sunderland, Massachusetts.
Brooks, L. D. 1988. The evolution of recombination rates. In: Michod, R. E. & Levin, B. R. (eds.), The evolution of sex: An examination of current ideas. Sinauer Associates, Sunderland, Massachusetts.
Hollande, A. & Carruette-Valentin, J. 1970. Appariement chromosomique et complexes symptomatiques dans le noyeaux en cour de depolyploidisation chez Pyrsonympha flagellata: le cycle evolutif des Pyrsonymphines symbiontes de Reticulotermes lucifugus. Compte Rendus de l'Académie des Sciences, Paris D270, 2550-3.
Note added 22 November 2003:
A colleague has been so kind as to raise the issue of the role of hermaphrodites in our hunt for the benefit of sex. I am adding this note with the caveat that it is not supposed to bring this essay up to date - much has happened since I wrote it. Rather, I want to point out that hermaphrodites bear the same two-fold cost that all sexual organisms pay - unless they are in an environment where they experience reduced competition with other males (local mating competition, LMC; e.g. a male who has exclusive access to his sisters (but no other females), as in some fig wasps), they should invest half of their "sexual energy" in male functions. Which emphasises the special role of self-fertilising organisms, since these experience zero local mating competition, and as long as their gamete production cannot be biased (e.g. by segregation distorters), they will show minimal investment in male reproductive functions.