Cis- and trans-acting genetic factors contribute to heterogeneity in the rate of crossing over between the Drosophila simulans clade species

Authors


Correspondence: Maria Victoria Cattani, Center for Genomics and Systems Biology, New York University, New York, NY 10003, USA.

Tel.: +(212) 992 7541; fax: (212) 995 4015; e-mail: vmc7@nyu.edu

Abstract

In the genus Drosophila, variation in recombination rates has been found within and between species. Genetic variation for both cis- and trans-acting factors has been shown to affect recombination rates within species, but little is known about the genetic factors that affect differences between species. Here, we estimate rates of crossing over for seven segments that tile across the euchromatic length of the X chromosome in the genetic backgrounds of three closely related Drosophila species. We first generated a set of Drosophila mauritiana lines each having two semidominant visible markers on the X chromosome and then introgressed these doubly marked segments into the genetic backgrounds of its sibling species, Drosophila simulans and Drosophila sechellia. Using these 21 lines (seven segments, three genetic backgrounds), we tested whether recombination rates within the doubly marked intervals differed depending on genetic background. We find significant heterogeneity among intervals and among species backgrounds. Our results suggest that a combination of both cis- and trans-acting factors have evolved among the three D. simulans clade species and interact to affect recombination rate.

Introduction

Meiotic recombination ensures the segregation of homologous chromosomes (Hawley, 1988). A failure to produce a crossover between homologous chromosomes can result in nondisjunction, with the consequent aneuploidy being potentially catastrophic for zygotic development (Hassold & Hunt, 2001; Manheim et al., 2002). The number and distribution of crossovers is regulated, with crossover assurance mechanisms ensuring at least one crossover event per chromosome arm and crossover interference mechanisms limiting the number of crossovers per chromosome arm (reviewed in Lindsley & Sandler, 1977). In addition, the mean rate of crossing over per physical length of chromosome varies among genomic regions (Lindsley & Sandler, 1977; True et al., 1996; Myers et al., 2005; Nachman, 2002; Petes, 2001), among individuals (Coop et al., 2008) and among species (True et al., 1996; Jensen-Seaman et al., 2004; Ptak et al., 2005; Winckler et al., 2005; Dumont et al., 2011; Dumont & Payseur, 2011).

Within Drosophila genomes, there is substantial variation in recombination rate at both broad and fine scales. At broad scales, recombination rates can vary dramatically and tend to be strongly suppressed near centromeres and telomeres (Lindsley & Sandler, 1977). In Drosophila melanogaster, the autosomes and the X chromosome differ, with stronger centromeric suppression on the autosomes and stronger telomeric suppression on the X (Lindsley & Sandler, 1977). At finer scales, recombination rates can vary up to approximately 3.5-fold in a 1.2-Mbp interval on the D. melanogaster X chromosome (Singh et al., 2009) and approximately 40-fold in a 2-Mbp interval on the Drosophila pseudoobscura X (Cirulli et al., 2007). Across eukaryotes, fine-scale variation in recombination rates is correlated with particular sequence motifs, transposons, repetitive DNAs, GC content, and nucleotide diversity (Begun & Aquadro, 1992; Marais, 2003; Myers et al., 2005; Nachman, 2002; Petes, 2001; Singh et al., 2009). Some of these correlates directly contribute to recombination rate variation (e.g. hotspot DNA sequence motifs in mammals), whereas others are evolutionary consequences of recombination rate variation (e.g. nucleotide diversity). For still other correlates, the basis of the relationship is unclear (e.g. repetitive DNAs).

Within species, recombination rates can differ among strains of D. melanogaster (Lawrence, 1958; Green, 1959; Kidwell, 1972; Brooks & Marks, 1986), where genetic map lengths can vary up to approximately 14% for the major autosomes (Brooks & Marks, 1986). Artificial selection experiments designed to alter genetic map distances in Bombyx mori (Turner, 1979; Ebinuma & Yoshitake, 1981) and in D. melanogaster (Chinnici, 1971; Kidwell, 1972; Charlesworth & Charlesworth, 1985) provide further direct evidence for genetic variation affecting recombination rates. Indeed, the rate of recombination often evolves as a correlated response to strong selection on other quantitative traits (reviewed in Otto & Barton, 2001).

Between distantly related Drosophila species, large differences in total map length exist. The genetic map of Drosophila virilis is, for instance, approximately twice as long as that of D. pseudoobscura (Gubenko & Evgen'Ev, 1984; Anderson, 1993), which is in turn twice as long as that of D. melanogaster (Lindsley & Zimm, 1992). Considerable differences also exist between D. melanogaster and its close relatives of the Drosophila simulans species complex, D. simulans, Drosophila mauritiana and Drosophila sechellia. The total genetic map lengths of D. mauritiana and D. simulans are approximately 1.8 and approximately 1.3 times longer than the standard total genetic map of D. melanogaster, respectively (Sturtevant, 1929; True et al., 1996). Most of the D. melanogaster–D. simulans difference is attributable to the autosomes, as the genetic map length of the X chromosome of the two species is virtually identical. The genetic map of the D. mauritiana X chromosome, however, is approximately 1.8 times longer than both (True et al., 1996). Little information exists on the D. sechellia genetic map length, but for at least two markers on the third chromosome, there is no significant difference between D. mauritiana and D. sechellia (Coyne & Charlesworth, 1997). In addition to such differences in average genetic map length, closely related species have evolved differences in the broad-scale chromosomal distribution of recombination events: centromeric and telomeric suppression is strong in D. melanogaster but weak in D. mauritiana (True et al., 1996).

Chromosomal, individual and interspecific variation in recombination rates can in principle be attributed to cis- and trans-acting genetic factors. Trans-acting factors have the potential to affect recombination rates genome-wide, whereas cis-acting factors should primarily affect the local rate of recombination in the chromosomal region in which they reside. Trans-acting factors might include proteins involved in chromatin remodelling or crossover formation (e.g. Prdm9 in mammals; Parvanov et al., 2009). Cis-acting factors, on the other hand, might include chromosomal rearrangements, centromeric heterochromatin or particular sequence motifs (e.g. recombination hotspots; Petes, 2001; Myers et al., 2008). Previous genetic analyses show that a combination of cis- and trans-acting factors contribute to variation in local recombination rates within species (Brooks & Marks, 1986; Timmermans et al., 1997; Yandeau-Nelson et al., 2006; Baudat & de Massy, 2007; Parvanov et al., 2009; Baudat et al., 2010; Grey et al., 2011), but little is known about the genetic control of recombination rate differences between species.

Here, we explore the genetic control of crossover differences between closely related Drosophila species. In particular, we estimate crossover frequencies for seven contiguous segments spanning most of the euchromatic portion of the D. mauritiana X chromosome when placed into three different genetic backgrounds: its own and those of its two sibling species, D. simulans and D. sechellia. We test whether crossover frequencies within each interval differ depending on species genetic background. If species differences in the rate of crossing over are determined solely by global trans-acting factors, we might expect all seven segments to show similar shifts in crossover rates depending on background. If, however, differences in the rate of crossing over are also determined by cis-acting factors, we might expect heterogeneity among intervals across the three different genetic backgrounds. Our results suggest that both cis- and trans-acting factors that affect the local rate of crossing over on the X chromosome have evolved among the three D. simulans clade species.

Materials and methods

Construction of lines

Starting with nine D. mauritiana white (w) lines, each bearing a unique semidominant P[w+] marker inserted at known cytological and DNA sequence positions on the X chromosome (described in True et al., 1996 and Araripe et al., 2006; kindly provided by Y. Tao, Emory University), we generated seven lines having two adjacent P[w+] markers. These seven lines were then made homozygous and maintained as stocks. The seven doubly P[w+]-marked intervals tile across most of the X chromosome euchromatin (Fig. 1). For five doubly P[w+]-marked segments, we could distinguish four genotypes: the parental genotype with two P[w+] inserts (hereafter ‘2P’), the parental genotype with no P[w+] inserts (hereafter ‘0P’) and the two recombinant genotypes with a single P[w+] insert (hereafter ‘1P’). For two doubly marked segments, 2P-1 and 2P-7, we could only distinguish three genotypes: the parental 2P, the parental 0P, but the left (distal) and right (proximal) 1P recombinant classes were indistinguishable (Fig. 2). Eye colour in these lines is sensitive to copy number (0P vs. 1P vs. 2P), to genotype (heterozygous vs. homozygous) and to position of the P[w+] insert, providing a range of phenotypes from pale yellow eye colour in heterozygotes to dark orange in homozygotes (see also True et al., 1996 and Tao et al., 2001 for other instances in which these markers were reliably used). The particular P[w+] elements used were selected from a larger set based on our ability to reliably distinguish 0P[w+], 1P[w+] and 2P[w+] individuals.

Figure 1.

2P intervals spanning the Drosophila mauritiana X chromosome. Inverted black triangles represent the insertion sites of P[w+] elements; names above correspond to the different P[w+] elements used (True et al., 1996).

Figure 2.

Scoring crossover frequencies. 2P females were crossed to Drosophila mauritiana w, Drosophila simulans w or Drosophila sechellia w males. For each cross, we counted the 0P, 1P, and 2P daughters. Grey bars represent X chromosome; grey, short bar with a hook represents Y chromosome. Inverted black triangles represent P[w+] markers.

We introgressed all seven doubly marked 2P segments from D. mauritiana into mutant white-eyed D. simulans wXD1 and D. sechellia w genetic backgrounds (Fig. 2). Briefly, we crossed sibling species (D. simulans wXD1 and D. sechellia w) females to D. mauritiana w 2P males. Virgin F1 hybrid females heterozygous for the 2P segment were then backcrossed to sibling species males. We repeated this backcrossing procedure for 15 generations, selecting virgin heterozygous 2P females each generation. This procedure gradually places the D. mauritiana 2P segment into a largely sibling species' genetic background.

Estimation of crossover frequencies

We estimated the rate of crossing over between adjacent P[w+] inserts in heterozygous 2P females as the frequency of heterozygous recombinant 1P daughters (we excluded sons because hemizygous recombinant 1P and parental 2P genotypes could not always reliably be distinguished). We estimated crossover frequencies at two generations of the introgression procedure, generation 11 and again at generation 15. At both generations, we collected approximately 10 virgin females from the D. simulans wXD1 and D. sechellia w introgression lines and from the original D. mauritiana w 2P lines and aged them for 2 days. We then crossed them to approximately 15 corresponding sibling species males (D. simulans wXD1, D. sechellia w or D. mauritiana w) that were aged for at least 2 days. We set three replicate crosses per species per interval per generation (three replicates × three species × seven intervals × two generations). Crosses were carried out in standard cornmeal-agarose Drosophila medium and kept in an incubator at 25 °C with a 12-h light cycle. After 5 days for D. simulans and D. mauritiana, and seven for D. sechellia (which has lower fecundity), the parents were removed, and the vials hydrated with a solution of 0.5% propionic acid. For each cross, we counted the number of 0P, 1P and 2P daughters (Fig. 2). Offspring were counted within 8 h of eclosion, when differences in eye colour are most extreme. As the left (distal) and right (proximal) recombinant 1P genotypes could not always be distinguished from one another, we pooled them. For each species and interval, at least 632 female progeny were scored (mean = 1330; range = 632–1951) (Table 1). Because it is possible that differential viability of females with and without P[w+] introgressions might affect our estimates of crossover frequencies, we tested for significant deviations from the expected 1 : 1 ratio of 2P vs. 0P daughters in each of the 21 lines (seven intervals × three genetic backgrounds) using Yates' Chi-square test. Crossover frequencies were not converted to genetic map distances using mapping functions, as the observed crossover frequencies for all intervals were comparable and relatively small (≤0.174), making the expected frequency of double crossovers negligible.

Table 1. Crossover frequency differs among lines and intervals, and it is not correlated with interval size, GC content or density of repetitive satellites
IntervalSpeciesProgeny scoredCrossover freq. ± SECytological intervalInterval size (Mbp)Crossover per Mbp ± SE% GCRepetitive DNA per Mbp
2P-1Drosophila mauritiana16130.158 ± 0.0201EF-4B3.730.042 ± 0.02046.38365.95
Drosophila simulans19050.041 ± 0.0030.011 ± 0.003
Drosophila sechellia8320.055 ± 0.0150.015 ± 0.015
2P-2D. mauritiana12680.057 ± 0.0074B-6EF2.360.024 ± 0.00745.12441.95
D. simulans18660.058 ± 0.0040.024 ± 0.004
D. sechellia11160.043 ± 0.0030.018 ± 0.003
2P-3D. mauritiana15580.060 ± 0.0076BC-9A2.990.020 ± 0.00744.22356.52
D. simulans14540.053 ± 0.0300.017 ± 0.030
D. sechellia12990.079 ± 0.0110.026 ± 0.011
2P-4D. mauritiana16180.077 ± 0.0079A-10EF2.230.035 ± 0.00744.69356.05
D. simulans10090.057 ± 0.0080.025 ± 0.008
D. sechellia9900.114 ± 0.0110.051 ± 0.011
2P-5D. mauritiana16810.143 ± 0.01110EF-12F2.910.049 ± 0.01143.53395.53
D. simulans16420.091 ± 0.0030.031 ± 0.003
D. sechellia6370.036 ± 0.0070.012 ± 0.007
2P-6D. mauritiana19510.174 ± 0.00912F-15A3.680.047 ± 0.00944.87351.90
D. simulans12200.107 ± 0.0130.029 ± 0.013
D. sechellia6560.098 ± 0.0120.027 ± 0.012
2P-7D. mauritiana14730.092 ± 0.01015A-18DE1.380.067 ± 0.01048.12422.46
D. simulans15110.069 ± 0.0100.050 ± 0.010
D. sechellia6320.071 ± 0.0090.051 ± 0.009

Statistical analyses

For statistical analyses, crossover frequencies were angularly transformed (arcsine of the square root of the proportion of recombinant females) and analysed by analysis of variance (anova). The two different generations at which crossover frequencies were estimated (generations 11 and 15) were treated as random blocks in the anova model, with species and interval set as fixed factors within each block. Significant main effects were followed by Tukey post hoc tests. We confirmed that the data conformed to anova assumptions. We also analysed the data with two additional models. First, we analysed the data using a two-way anova with species and interval as fixed factors and pooled generation data (= 6 crosses for each combination of species and interval). As the results are quantitatively similar to our blocked model, we present only the former analysis in the 'Results' section. Second, we also performed the blocked anova analysis for crossover frequencies that were standardized by the length (Mbp) of each interval. Even after multiple transformations, however, these data failed to conform to the homogeneity of variance assumption. Nevertheless, the results do not qualitatively differ from those reported from the blocked model. To test whether crossover frequencies varied consistently in the three species' backgrounds, we estimated Spearman rank correlations among crossover frequencies between each species pair.

To test for possible correlates of recombination rate, we calculated the GC content of each 2P interval using the newly generated genome sequence data from the D. mauritiana reference strain, which is the same strain used in the present experiments (Garrigan et al., 2012). We used the Tandem Repeats Finder software (Benson, 1999) to determine the density of repetitive satellite DNAs in each of the 2P intervals (i.e. number of satellite loci divided by the length of the interval). For each interval, we scored the density of repetitive DNAs of different period sizes: microsatellites (period ≤ 5), minisatellites (period ≤ 20) and repetitive DNA loci of any period size. We then tested for correlations between standardized crossover rates, GC content and the densities of different kinds of repetitive satellites. As the results do not differ for repetitive DNAs of different period sizes, we report the correlation for the density of repetitive DNAs of all sizes. To further test whether any of the candidate cis-acting factors is related to crossover frequency, we performed multiple regression analyses for each of the three species with interval size, GC content and density of repetitive DNAs as explanatory variables and crossover frequency as the dependent variable. All statistical analyses were performed using the statistica 7.0 software (StatSoft, Tulsa, OK, USA).

Results

We estimated the frequency of crossover events within seven intervals on the D. mauritiana X chromosome, each marked by two visible semidominant P[w+] inserts (hereafter, 2P), in D. mauritiana w, D. simulans wXD1 and D. sechellia w genetic backgrounds (Fig. 1). To investigate the contribution of cis- and trans-acting factors to differences in crossover frequencies among species, we estimated crossover frequencies for each combination of species background and 2P interval twice, at generations 11 and 15 of the introgression procedure, scoring the numbers of parental (0P and 2P) and recombinant (1P) daughters produced by heterozygous 2P mothers (Fig. 2; Table 2). We first tested whether differential viability of females with or without P[w+] introgressions might affect estimates of crossover frequencies, but found that heterozygous 2P daughters do not have reduced viability relative to 0P daughters in 20 of the 21 crosses (Table 2). Only interval 2P-2 in a D. simulans background deviated significantly from a 1 : 1 ratio (= 0.016), with a small dearth of heterozygous 2P females. Table 3 shows that crossover frequencies vary significantly among intervals (F6,84 = 36.17, = 0.0002) and depend on the species genetic background (F2,84 = 42.96, = 0.0227) suggesting that both cis- and trans-acting factors contribute to the differences in crossover frequencies. The generation at which the crossover frequencies were estimated had no effect (F1,84 = 0.31, = 0.6285). Tukey post hoc tests show that crossover frequencies are, on average, approximately 1.4-fold higher in D. mauritiana than in D. simulans or D. sechellia.

Table 2. Viability differences between 0P and 2P daughters cannot explain observed differences in crossover frequencies
Species backgroundIntervalFemalesPa
0P1P2P
  1. a

    Yates' chi-square test.

Drosophila mauritiana2P-13391273410.975
D. mauritiana2P-2305362890.538
D. mauritiana2P-3382473560.357
D. mauritiana2P-4357603640.823
D. mauritiana2P-53321143540.423
D. mauritiana2P-63721553660.854
D. mauritiana2P-7316633030.629
Drosophila simulans2P-1522434760.154
D. simulans2P-2492564180.016
D. simulans2P-3360393360.383
D. simulans2P-4246292370.715
D. simulans2P-5412813980.648
D. simulans2P-6285703000.563
D. simulans2P-7368553710.944
Drosophila sechellia2P-1227252050.312
D. sechellia2P-2303262710.196
D. sechellia2P-3335553070.286
D. sechellia2P-4282692520.209
D. sechellia2P-5168131760.706
D. sechellia2P-6189401780.602
D. sechellia2P-7206312010.843
Table 3. Results of anova testing for differences in crossover frequency among intervals and species backgrounds (Drosophila mauritiana, Drosophila simulans and Drosophila sechellia) for the two generations assayed (generations 11 and 15)
Source of variationd.f.MSFP% variance
Generation10.00010.310.630
Species20.018542.960.022717
Generation × Species20.00040.76 0
Interval60.010836.170.000223
Generation × Interval60.00030.53 0
Species × Interval120.00679.940.000239
Generation × Species × Interval120.00071.19 1
Residuals840.0006   

To test whether trans-acting factors alone could explain differences in crossover frequencies, we asked whether crossover frequencies varied consistently in each species. Within D. mauritiana, post hoc comparisons reveal three groups of intervals (< 0.001): one with a single interval with the significantly highest normalized crossover frequency of 0.067 (2P-7), a group with intermediate crossover frequencies ranging from 0.042 to 0.047 (2P-1, 2P-5 and 2P-6) and a last group with crossover frequencies ranging from 0.020 to 0.035 (2P-2, 2P-3, and 2P-4) (Fig. 3). Similarly, within D. simulans, there are also three groups of intervals (< 0.01): one with a single interval with the significantly highest crossover frequency of 0.050 (2P-7), a group with intermediate crossover frequencies ranging from 0.024 to 0.031 (2P-2, 2P-4, 2P-5 and 2P-6) and a last group with crossover frequencies ranging from 0.011 to 0.017 (2P-1 and 2P-3) (Fig. 3). Within D. sechellia, there are also three clusters (< 0.05): one group with the highest crossover frequency of 0.051 (2P-4 and 2P-7), one with intermediate crossover frequencies ranging from 0.026 to 0.027 (2P-3 and 2P-6) and a last group with crossover frequencies ranging from 0.012 to 0.018 (2P-1, 2P-2 and 2P-5) (Fig. 3). There is no significant Spearman rank correlation between crossover frequency and the length of the interval (Mbp) for any species (= 0.58, = 0.17 for D. mauritiana;= 0.25, = 0.59 for D. simulans;= 0.08, = 0.96 for D. sechellia), although with just seven intervals these analyses have limited power.

Figure 3.

Mean values ± standard error for crossover frequency per Mbp for Drosophila mauritiana (black bars), Drosophila simulans (grey bars) and Drosophila sechellia (white bars) in seven intervals along the X chromosome.

Crossover frequencies are, on average, significantly higher in D. mauritiana than in the other two species, consistent with the action of trans-acting species differences. However, Table 3 reveals a significant interaction between species background and particular 2P interval, suggesting that trans-acting factors alone could not explain the differences in crossover frequencies. Indeed, 39% of the variance is attributable to the interaction between species and interval, greater than any other source. Consistent with the large species × interval interaction effect, there is no correlation among crossover frequencies in different genetic backgrounds: intervals with high crossover frequencies in one species do not always have high crossover frequencies in the other two (rsim vs. sech = −0.071, = 0.857; rsim vs. mau = 0.357, = 0.379; rsech vs. mau = 0.071, = 0.857). We find that the normalized crossover frequency varies among species background for four of the seven intervals (Fig. 3). For interval 2P-1 and interval 2P-6, the crossover frequencies in D. mauritiana are significantly higher than those of D. simulans and D. sechellia (< 0.001 for 2P-1; < 0.05 for 2P-6). As the 2P-1 interval is near the X chromosome telomere, this observation shows that telomeric suppression is weak or absent in a D. mauritiana genetic background but can be induced when the D. mauritiana distal segment is introgressed into a D. simulans or D. sechellia genomic background. Surprisingly, for interval 2P-4, the crossover frequency in D. sechellia (0.051) is significantly higher than those of D. mauritiana (0.035) and D. simulans (0.025) (< 0.05). For interval 2P-5, there is a significant difference in crossover frequency among all three species: D. mauritiana has the highest crossover frequency (0.049), followed by D. simulans (0.031) and then D. sechellia (0.012; < 0.001 for all three comparisons). There are no significant differences in crossover frequency among species backgrounds for intervals 2P-2, 2P-3 or 2P-7.

To try to identify candidate causative cis-acting factors that might contribute to the differences in crossover frequencies, we also tested whether standardized crossover rates (i.e. crossover frequency per Mbp) are correlated with the GC content or the density of repetitive satellite DNA loci within each interval for each species background. We found no significant correlation between standardized crossover rate and GC content in any of the three species backgrounds (< ∣0.26∣, > 0.29). Similarly, we found no significant correlation between standardized crossover rate and the density of repetitive DNA loci in any of the species background (< 0.42; > 0.45). To further test whether any of the candidate cis-acting factors is related to crossover frequency, we performed a multiple regression analysis for each of the species treating interval size, GC content and density of repetitive DNAs as explanatory variables and crossover frequency as the dependent variable. In all three cases, the P values of the model were ≥ 0.196, suggesting that none of the explanatory variables have a significantly effect on crossover frequency (math formula = 0.543, = 0.445; math formula = 0.084, = 0.959; math formula = 0.748, = 0.196). As above, however, these analyses have limited power as they involve just seven data points.

Discussion

The analyses here suggest that the genetic control of crossover frequencies has diverged between the D. simulans clade species, with both cis- and trans-acting factors appearing to contribute to the differences in the genetic control of recombination rate on the X chromosome. Crossover frequencies observed in D. mauritiana were, on average, approximately 1.4-fold higher than those of its sister species, D. simulans and D. sechellia. Our results are thus consistent with those of True et al. (1996), where the total genetic map length for the D. mauritiana X chromosome was approximately 1.8-fold higher than that of D. simulans. This main effect – a higher average rate of crossing over in a D. mauritiana background – is, most simply, consistent with a species difference in trans-acting factors. While it is formally possible that many dominant cis-acting changes have accumulated over the length of the D. mauritiana X chromosome that on balance increase the rate of crossing over, whole-genome sequence data do not support a simple D. mauritiana -specific accumulation of changes on the X (Garrigan et al., 2012). In other work, we have identified a large-effect trans-acting gene responsible for much of the species difference between D. mauritiana and D. melanogaster (M.V. Cattani, S.B. Kingan & D.C. Presgraves, unpublished); it remains to be determined whether this gene also contributes to the species differences observed among the D. simulans clade species. It is however clear that trans-acting factors have evolved to effect species differences in the rate of crossing over in Drosophila.

The evolution of species differences in trans-acting factors alone is not, however, sufficient to explain the observed patterns. For interval 2P-4, for instance, the crossover frequency is significantly higher in D. sechellia than in D. mauritiana and D. simulans, but, for interval 2P-5, all three species differ significantly, with D. sechellia having the lowest crossover frequency of the three. In principle, the latter observation could be explained if a previously unidentified inversion in D. sechellia hindered recombination. However, the inversion would have to be small enough to go undetected in traditional polytene chromosome analyses but large enough to suppress crossover frequencies over 3 Mbp. These observations imply that cis-acting differences must have evolved in the different species to account for such species-by-interval interactions in crossover frequency (Table 3). Indeed, as our analyses show that most of the variance in crossover frequency comes from the interaction between species and chromosomal interval, we infer that both cis- and trans-acting factors have evolved in this relatively young species clade. As an alternative to the evolution of cis-acting modifiers of crossing over, it is possible that trans-acting factors have evolved to target different sets of cis-acting sequences in the three species, with differences in target number explaining the observed differences in crossover frequencies for a particular interval among species' backgrounds.

On a broad chromosomal scale, our results show that crossover frequencies normalized by interval length can differ by at least approximately 2.8-fold on the D. mauritiana X chromosome (e.g. interval 2P-2 vs. 2P-7; Table 1). As in True et al. (1996), however, we find that D. mauritiana lacks the pronounced telomeric suppression of crossing over at the tip of the X chromosome as seen in D. melanogaster and in D. simulans (Lindsley & Sandler, 1977; True et al., 1996): in D. mauritiana's own genetic background crossover frequency at the tip of the X (0.042) is similar to the average crossover frequency for the entire X chromosome (0.040). However, telomeric suppression becomes evident when the tip of the D. mauritiana X is introgressed into D. simulans or D. sechellia: crossover frequencies for the distal-most interval 2P-1 are reduced threefold in both species' genetic backgrounds (Table 1). These findings allow us to make inferences about the evolutionary history and genetic control of telomeric suppression. First, the presence of telomeric suppression in the outgroup species, D. melanogaster, and the two sister species D. simulans and, based on the present findings, D. sechellia suggests that the absence of telomeric suppression evolved recently in D. mauritiana, sometime within the last approximately 240 000 years. It has been suggested that telomeric suppression could be a mechanism to ensure X chromosome segregation, as exchanges that happen too near the telomere might cause failure of centromeres to properly orient (Koehler et al., 1996). This model, however, seems unable to account for the absence of telomeric suppression in D. mauritiana, whose karyotype is not different from that of its sister species. Second, the absence of telomeric suppression cannot be explained by dominant cis-acting changes D. mauritiana, as these would cause release from telomeric suppression even when heterozygous in the genetic background of its sister species. Instead, telomeric suppression must result either from dominant cis-acting factors in D. simulans and D. sechellia that make them refractory to crossovers or from trans-acting factors in D. simulans and D. sechellia that prevent crossovers near the X chromosome telomere.

The conclusions drawn here come with several caveats. First, viability differences between females with and without the introgression might bias our estimates of crossover frequencies. However, Table 2 shows that inviability effects are unlikely, as we find no significant deviation from a 1 : 1 ratio of 2P vs. 0P daughters in 20 of the 21 crosses (the one small deviation disappears if we correct for multiple tests). Second, it is possible that the sequence homozygosity of the D. mauritiana stocks may quantitatively bias estimates of crossover frequencies. However, studies in D. melanogaster and in D. mauritiana suggest that there is little or no effect of stock isogenicity on crossover frequencies (Rutherford & Carpenter, 1988; True et al., 1996). Third, as with any analysis of phenotypic differences between species, it is possible – indeed likely – that some fraction of the quantitative differences in crossover frequencies observed here reflect the contributions of variation segregating within species. More work will be needed to partition genetic variation for crossover rates into that segregating within vs. that fixed between species, a nontrivial task requiring the estimation of crossing over for many lines within multiple species. Once the genes and sequences responsible for species differences in genetic map lengths have been identified, we can begin to ask about the population genetic and molecular basis of the evolution of recombination rates.

Acknowledgments

This work was supported by an Ernst Caspari Fellowship from the University of Rochester to MVC and by funds to DCP from the University of Rochester, the Alfred P. Sloan Foundation and the David and Lucile Packard Foundation. We thank Colin Meiklejohn and two anonymous reviewers for helpful comments on this manuscript. The authors declare no conflicts of interest.

Data depostited at Dryad: doi:10.5061/dryad.7p52d

Ancillary