Sex differences in recombination

One of the stronger empirical generalizations to emerge from the study of genetic systems is that achiasmate meiosis, which has evolved 25-30 times, is always restricted to the heterogametic sex in dioecious species. usually the male. Here we collate data on quantitative sex differences in chiasma frequency from 54 species (4 hermaphroditic flatworms, 18 dioecious insects and vertebrates and 32 hermaphroditic plants) to test whether similar trends hold. Though significant sex differences have been observed within many species, only the Liliaceae show a significant sexual dimorphism in chiasma frequency across species, with more crossing over in embryo mother cells than in pollen mother cells; chiasma frequencies are unrelated to sex and gamety in all other higher taxa studied. Further, the magnitude of sexual dimorphism. independent of sign, does not differ among the three main ecological groups (dioecious animals, plants, and hermaphroditic animals), contrary to what would be expected if it reflected sex-specific selection on recombination. These results indicate that the strong trends for achiasmate meiosis do not apply to quantitative sex differences in recombination, and contradict theories of sex-specific costs and benefits. An alternative hypothesis suggests that sex differences may be more-or-less neutral, selection determining only the mean rate of recombination. While male and female chiasma frequencies are more similar than would be expected under complete neutrality. a less absolute form of the hypothesis is more difficult to falsify. In female mice the sex bivalent has more chiasmata for its length than the autosomes, perhaps compensating for the absence of recombination in males. Finally, we observe that chiasma frequencies in males and females are positively correlated across species, validating the use of only one sex in comparative studies of recombination. 259 260 Burt et al. Introduction Observations of sex differences in the amount of recombination at meiosis are common, even among autosomal genes, and date back to the early days of genetics (Morgan, 19 12, 1914; Haldane, 1920). These differences can be usefully divided into three types, according to their cytogenetics. First, both sexes may have normal chiasmate meiosis, but with quantitative differences in the number or position of cross-overs (e.g. mice). Second, one sex may have an achiasmate meiosis, with no crossing-over of homologous chromosomes at all (e.g. male fruit flies). Finally, there may be neither independent segregation of nonhomologous chromosomes nor crossing-over in one sex (always the male), as in haplodiploid and parahaplodiploid species (e.g. bees, scale insects). Here, we will be mainly concerned with the evolution of quantitative sex differences in recombination. Haldane ( 1922) gave the first general treatment of the problem, advancing the empirical claim that recombination tends to be reduced in the heterogametic sex. Huxley ( 1928) similarly suggested that whenever a marked sex difference in recombination occurred, it was always the heterogametic sex that had the lower value. Both authors proposed the same explanation: if gender is determined by two or more loci on the sex chromosomes, then selection against intersexes will favour reduced recombination between these chromosomes in the heterogametic sex, and as a pleiotropic effect the recombination of autosomal chromosomes may also be reduced. These views have been questioned on occasion, both because there are some exceptions to the empirical generalization (e.g. Dunn and Bennett, 1967; Callan and Perry, 1977) and because the proposed explanation cannot account for observed sex differences in hermaphrodites (e.g. Ved Brat, 1966). However, there was no alternative theoretical perspective until Trivers (1988) recently revived the subject, with slightly different empirical claims and a provocative new explanation. According to Trivers, recombination tends to be lower in males than females, as well as lower in the heterogametic sex than the homogametic sex, though he acknowledges that there are many exceptions to these rules. Trivers suggests that both reduced recombination and heterogamety are consequences of selection being more intense in one sex (usually the male) than the other. He argues that reproducing individuals of the sex experiencing more intense selection will, on average, have better combinations of genes than those of the other sex, and so the cost of breaking up those combinations should be higher. Bernstein et al. (1988) counter with an alternative explanation, that rates of recombination tend to be higher in females because oogenesis is associated with higher metabolic rates, and thus more DNA damage, than spermatogenesis; however, they admit to being puzzled by the association with gamety. Most other theories of recombination can be adapted to predicting sex-specific optimal recombination rates. For example, many theorists believe that the main function of recombination is to reduce linkage disequilibrium (e.g. Felsenstein, 1988; Maynard Smith, 1988; Kondrashov, 1988). As two potentially important sources of linkage disequilibrium are selection and drift, one might expect that the Sex differences in recombination 261 sex experiencing the more intense selection, or otherwise having the higher variance in reproductive success, should have more recombination. This prediction is exactly opposite to that made by Trivers. Other predictions follow from the various proposed diversity theories of recombination (Williams, 1975; Bell, 1982; Tooby, 1982). Alternatively, sex differences in recombination may be more-or-less invisible to natural selection, the latter determining only the mean value. Simulations by Nei ( 1969) indicate that sex differences in recombination may have very little effect on population mean fitness. Information currently available on achiasmate meiosis in no way contradicts Haldane’s, Huxley’s, and Trivers’ empirical claims: we know of 25-30 independent origins of achiasmate meiosis among dioecious animals (A. Burt, unpublished; see Serrano, 1981; Bell 1982; Nokkala and Nokkala, 1986 and references therein) and every time it has evolved in the heterogametic sex, which all but twice is the male (exceptions are Copepoda and Lepidoptera/Trichoptera). Here, we bring together the available data on quantitative sex differences in chiasma frequencies, to further test the strength of the proposed trends and, if possible, to test the various explanations. To this end we also examine the magnitude of sexual dimorphism in chiasma frequency, independent of sign, and look for evidence of compensation between the sexes. One further motivation for this study is to estimate the correlation between male and female chiasma frequencies across species, thus determining whether the value for one sex is a good indicator of what is happening in the other sex and in the species as a whole. This estimate is important because male meiosis is usually more easily studied than female meiosis, and so comparative surveys of chiasma frequencies tend to only use data for males (e.g. Burt and Bell, 1987; Sharp and Hayman. 1988). There are about 20 times more chiasma frequencies for males in the literature as for females. Data and Analysis Rates of recombination can be measured both by counting chiasmata through the microscope and by crossing marked individuals to construct a linkage map. Counts of chiasmata are available for many more species than are extensive linkage maps, and here we will restrict ourselves to the former. As with any comparative analysis using data from the literature, the quality of estimates varies for example, in techniques and sample sizes. Actually counting chiasmata in some species is quite straight-forward and in others quite difficult; female mammals are notoriously difficult. For plants, often only metaphase figures are available, whereas counts at the earlier diplotene stage are usually considered more accurate. Perhaps more importantly, the methods used are often different for the two sexes, so that observed differences between males and females may be due to differences of technique rather than real. This problem is particularly acute when the data for the two sexes come from different studies (3 of 6 amphibians and 2 of 4 mammals in our data set). 262 Burt et al. Perhaps the best measure of recombination to be got from a meiotic spread is the proportion of the genome which recombined. This value can be calculated by measuring the distances between the ends of chromosomes and the nearest chiasma and between neighboring chiasmata, and expressing these as a proportion of the total genome length. For n bivalents and C chiasmata, there will be n + C such distances, di. The proportion of the genome which recombines is then equal to the proportion of pairs of loci which are on different segments: P = 1 Ed;. This value will be a function of the number and size distribution of chromosomes and the number and position of cross overs. Corrections could be made for obviously noncoding fractions of the genome simply by not including them in the calculations. Unfortunately, this proportion has yet to be reported for any species. Instead, we shall use simple counts of the number of chiasmata, noting that for any given distribution of cross overs along the genome, our measure P increases monotonically with the number of chiasmata. Ignoring possible sex differences in the position of cross overs will lead to some inaccuracy: for example, Fletcher & Hewitt (1980) observe that males of Chr)~sochraon dispar have slightly more chiasmata per bivalent than females, but that they are terminalized to such an extent that the effective amount of recombination is greater in females. However, quantitative information on the position of chiasmata is available for very few species. One possible check on the data is to compare sex differences in chiasma frequency and linkage map lengths. Unfortunately, we know of map length data for only three species in our data set, all mammals: Sminthopsis crassicauduta (Bennett et al., 1986) mice (Dunn and Bennett, 1967), and humans (Donis-Keller et al., 1987). For S. crassicauduta and mice the sex differences in chiasma frequency and map lengths are in the same direction, but not for humans: the cytogenetic data suggest that males have more chiasmata than females (51 vs 43; Lange et al., 1975; Jagiello et al., 1976) but the genetic data indicates they have shorter map lengths (2017 vs 3857 CM; Donis-Keller et al., 1987). Apparently, the female chiasma frequencies are greatly underestimated. This corroborates Chandley’s (1988 : 20) statement that, due to technical difficulties, “accurate counts of chiasmata for the human female still remain to be established.” As the problems of getting sufficient appropriate material (oocytes at time of ovulation) are much greater for human females than for other species, this discrepancy is unlikely to be representative of the rest of the data. Indeed, among other organisms for which both chiasma frequencies and extensive genetic maps exist, there is a strong correlation between the two (r = 0.85, n = IO; A. Burt, unpublished). Here, we have excluded humans from further analysis. Having decided to use counts of chiasmata at meiosis, there still remains a number of possible indices of recombination. Burt and Bell (1987) defined the excess chiasma frequency as the number of chiasmata per bivalent in excess of one, summed across bivalents. This measure was considered to most accurately reflect selection for recombination, independently of the various constraints on changes in chromosome number and the mechanical role of chiasmata in proper segregation. However, it does not make much biological sense for polyploid and achiasmate species, both of which are represented in our data set. Therefore we use here the Sex difTerences in recombination 263 number of chiasmata per autosomal bivalent. Interpretations are also made easier by this choice, since in our data set the chiasma frequency per bivalent is independent of chromosome number (r = -0.105, n = 54, p > 0.4), while excess chiasma frequency is positively correlated with chromosome number (r = 0.317, n = 54, p < 0.02). In any case, choice of index does not affect the conclusions drawn. Data is available for 54 species of animals and higher plants (Appendix), approximately 0.002% of all known animals and higher plants. Unfortunately, the data set is taxonomically unrepresentative: there are 8 species of acridid grasshoppers, but no other arthropods; 4 Triturus newts, but no fish, reptiles, or birds; 22 species in the Liliaceae, but only two dicots. This nonrandomness means that we cannot put much weight on overall trends and must instead look within lower taxa: since we cannot make definitive statements about all animals and higher plants, we shall try to say something about acridid grasshoppers, Triturus, and the Liliaceae. Results Correlations Across all chiasmate species there is a positive correlation between male and female chiasmata per bivalent (r = 0.75, n = 54, p < 0.001; Fig. 1). this result seems to be fairly robust, as the sign of the correlation is positive in 9 of 11 independent taxa (Table 1). The exceptions are amphibians and Oedipodinae, a subfamily of grasshoppers, though neither are significantly negative. Sexual dimorphism Across all species females seem to have more chiasmata than males (paired t-test, t = 2.49, n = 54, p < 0.02). Closer examination of the data shows that this trend holds for Lilium (all 8 species, p 0.008) and probably Liliaceae genera (all 4 genera have more species with more chiasmata in the female than the male, p = 0.0625). However, there is no evidence that the trend applies to other plant taxa or any animal taxon (Table I). As many species individually show significant sex differences in chiasma frequency (Appendix), this result indicates that there is a large sex x species interaction effect. All dioecious species in the data set are male heterogametic (except one species with unknown sex chromosome system), so the absence of a consistent sex difference also indicates that there is no consistent difference between homoand heterogametic sexes. Ranges The magnitude of sexual dimorphism, independent of sign, is given by Imale-femalel. This is also the range, a measure of dispersion. The idea of sex-specific