DISCUSSION

Exceptions to the chromosomal homologies and rate of gene transposition: The ancestral karyotype of the genus Drosophila consisted of five acrocentric chromosomes and a dot (Muller's elements A–F). Our results are in good agreement with an extensive conserved synteny of Muller's elements A and D during the evolution of the D. melanogaster and D. repleta lineages. Therefore, and with the exception of 2 out of 145 clones, the established chromosomal homologies between chromosome X of D. melanogaster and D. repleta and between chromosomal arm 3L of D. melanogaster and chromosome 4 of D. repleta (Figure 1) are firmly corroborated. Our results also indicate that no exchange of information occurred via pericentric inversions after the centric fusion between Muller's elements D and E that gave rise to the metacentric chromosome 3 in the D. melanogaster lineage. Mammals also show an extensive conserved synteny of chromosome X, even though translocations have often rearranged the genome of mammalian species (Ohno 1967; Landeret al. 2001). However, the conservation of the X chromosome in mammals and in the Drosophila genus likely results from different mechanisms. In the case of mammals, there is a need to keep the level of expression of the X-linked genes adjusted to one single copy due to the dosage compensation mechanism (Hartl and Lozovskaya 1994; Graves 1996). In Drosophila, without discarding adjustments on gene dosage, the main reason is probably the reported lack of interchromosomal rearrangements, which holds for both the X chromosome and the autosomes (Powell 1997).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Comparison of the molecular organization of Muller's element A (chromosome X) between D. melanogaster and D. repleta. Connecting lines match the cytological position of orthologous markers. Shaded rectangles show conserved segments with two or more consecutive markers. The estimated size of each conserved segment is given on the leftmost column. The asterisk (*) indicates those clones yielding more than one hybridization signal. The names of those clones in parentheses are incorrect (these clones have probably been mislabeled during the clone distribution process).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

Comparison of the molecular organization of Muller's element D between D. melanogaster (chromosomal arm 3L) and D. repleta (chromosome 4). For meaning of the symbols, see legend of Figure 2.

Because of this absence of interchromosomal rearrangements in Drosophila, the two exceptional clones that fail to obey the synteny conservation likely indicate transposition events.

1. The gene Lsp1alpha is located on the X chromosome (element A) of D. melanogaster but maps to chromosome 2 (element E) of D. repleta. This gene is also localized in element E in eight different species belonging to the Sophophora and Drosophila subgenera (Brock and Roberts 1983) and it has been suggested that it recently transposed onto the X chromosome in the species belonging to the melanogaster subgroup (Smithet al. 1981). The D. buzzatii Lsp1 genes are being cloned and sequenced in our laboratory to test this hypothesis (J. González, F. Casals and A. Ruiz, unpublished results).

2. Cosmid clone 174F6 maps to the euchromatin-heterochromatin boundary of the D. melanogaster X chromosome (20A-C). However, it hybridized near the centromere (polytene band G5d) of chromosome 4 in D. repleta. This cosmid clone contains the suppressor of forked [su(f)] gene (Madueñoet al. 1995). Analysis of a 33-kb chromosomal walk around the su(f) locus in D. melanogaster revealed that most of this interval consists of repetitive sequences. In fact the su(f) gene is flanked by a 1.5-kb direct repeat sequence (Tudoret al. 1996). Sequences homologous to the 1.5-kb repeats are found in the euchromatin-heterochromatin boundary of chromosome arms 2L, 2R, and 3L of D. melanogaster. Therefore, ectopic exchange events involving some homology between donor and target site, which are a possible mechanism of gene transposition in the Drosophila genome (Yi and Charlesworth 2000), could explain the case of cosmid clone 174F6.

A crude estimate of the rate of gene transposition in the Drosophila genus can be produced by combining the results of the present work with results previously obtained for chromosomal elements B (Gonzálezet al. 2000) and E (Ranzet al. 2001). In these latter works no cases of gene transposition were detected. Therefore we have observed two possible transposition events out of a total of 328 clones hybridized to the D. repleta chromosomes (Table 1). This amounts to a rate of 4.9 × 10⁻⁵ transpositions/gene/myr, which is quite low. This rate, however, does not include tandemly repeated genes such as histone or rRNA genes, which often show transposition (Alonso and Berendes 1975; Fitchet al. 1990). It also does not include intrachromosomal transpositions. Given the differentiation undergone by the banding pattern and morphology of the polytene chromosomes of so distantly related species as D. repleta and D. melanogaster, the only transpositions that we can safely detect with our mapping procedure are those taking place between different chromosomal elements. Transpositions within the same chromosomal element are probably overlooked, although they do exist. For instance, the seven in absentia (sina) gene is nested within an intron of the Rh4 opsin gene in chromosomal arm 3L of D. melanogaster (Montellet al. 1987). However, in D. virilis (Neufeldet al. 1991) and D. repleta (this work) the two genes are located at distant sites of the homologous element (chromosome 3 in D. virilis and chromosome 4 in D. repleta). To explain the different structural arrangement of these two genes between D. melanogaster and D. virilis, Neufeld et al. (1991) proposed a retrotransposition event of the Rh4 gene in the lineage leading to D. virilis. Our results support their interpretation because two P1 phages (DS00383 and DS00052) from the sina/Rh4 region in D. melanogaster map near sina but far from Rh4 in D. repleta. In addition, if a single transposition event was involved, our result indicates that it took place after the separation of the D. melanogaster and D. virilis ancestral lineages but before the divergence between the D. virilis and D. repleta lineages (~30 myr ago). Further work with phylogenetically closer species, whose polytene chromosomes are more easily compared, is required to obtain more accurate estimates of the rate of gene transposition.

Comparative mapping and rates of fixation of paracentric inversions: Our fairly dense physical maps of D. repleta chromosomes X and 4 allow a detailed comparison of their gene arrangements with those of the homologous D. melanogaster elements X and 3L (Figures 2 and 3). In both elements, a considerable reshuffling of gene order extends from telomere to centromere. The rank order correlation is in both cases nonsignificant (for chromosome X, Spearman's R = 0.062, P > 0.05, six ties; for chromosome 4, Spearman's R = 0.184, P > 0.05, two ties), indicating that gene order is effectively randomized. This profound rearrangement found for Muller's elements A and D can be attributed mainly to the fixation of paracentric inversions if we consider that transposition rates are low (see above) and that paracentric inversions are the prevailing chromosomal rearrangement in Drosophila both as intraspecific polymorphisms and as interspecific fixed differences (Hartl and Lozovskaya 1994; Powell 1997; Ranzet al. 2001). Using the maximum-likelihood method devised by Ranz et al. (1997) we have estimated the number of inversions fixed between D. repleta and D. melanogaster since their most recent common ancestor (Figure 1). An estimate (±SD) of 118 ± 17 for Muller's element A and 56 ± 8 for Muller's element D were obtained (Table 3). The coefficient of variation (CV) of both estimates is reasonably low (14%) and comparable to the most detailed comparative maps carried out in Drosophila (Ranzet al. 2001). Obviously, the accuracy with which the rates of chromosomal evolution are estimated increases with the number of markers used in comparative mapping (Schoen 2000). However, computer simulations made with the method of Ranz et al. (1997) show that the CV is not a lineal function of the number of markers but follows a negative exponential function (D. Schoen, personal communication). This implies that a decrease of the CV below the actual values would require a disproportionate increase in the number of markers.

The evolution rates for chromosomal elements A and D can be compared with those for elements B and E, which have been previously estimated using the same pair of species (Gonzálezet al. 2000; Ranzet al. 2001). Because the different chromosomal elements vary in size, to make the data comparable, we have calculated the density of breakpoints per megabase by dividing the number of breakpoints by the size of each element in megabases in D. melanogaster (Table 3). Breakpoint density varies up to four times among chromosomal elements and the differences are statistically significant. The breakpoint density for element A (X chromosome) is the highest (10.83) but there are differences between the three analyzed autosomes as well. Element E exhibits the highest density (8.14) whereas element B shows the lowest (2.57) and element D is intermediate (4.63). The weighted average for the whole genome (the euchromatic portion of the six chromosomal elements A–F) is 6.54 breakpoints/Mb, which allows us to infer that 393 paracentric inversions have become fixed in the whole genome between D. melanogaster and D. repleta (Table 3). Taking 62 myr as the divergence time between the two subgenera, we obtain conservative estimates for the rate of disruptions per megabase per million years. These estimates range from 0.021 for element B up to 0.087 for element A with a weighted average of 0.053 for the whole genome (Table 3).

These results agree fairly well with the scarce reliable estimates previously reported in other Drosophila species. Segarra et al. (1995) compared the X chromosome between D. melanogaster and D. pseudoobscura and estimated that 0.086 disruptions/Mb/myr have occurred since the divergence of these two lineages. Likewise, Vieira et al. (1997a,b) compared the gene order of three different chromosomes among D. virilis, D. montana, and D. novamexicana, three species of the virilis group of subgenus Drosophila. Their rates (taking D. melanogaster as the reference species for chromosome sizes as before) were 0.036–0.056 disruptions/Mb/myr for chromosome X, 0.032 for chromosome 2 (Muller's element E), and 0.009–0.014 for chromosome 3 (Muller's element D). It is remarkable that the ranking order between chromosomal elements in these studies (A > E > D) is the same that we have observed (Table 3), which suggests a genus-wide pattern regardless of the phylogenetic distance of the species compared.

In addition, our results support the previous finding that the rate of genome rearrangement in Drosophila is about two orders of magnitude higher than that in mammals and several times higher than that in the most dynamic plant lineages (Ranzet al. 2001). This conclusion was drawn from the comparative analysis of the Muller's element E between D. melanogaster and D. repleta, which represents the ~23% of the euchromatic fraction of the D. melanogaster genome. Now, with comparative data in D. repleta of >60% of the D. melanogaster genome, the same conclusion still holds. Most current comparative maps of mammals (and also plants) have a relatively poor resolution. Consequently, as the number of orthologous markers mapped increase, it is likely that more rearrangements (e.g., paracentric inversions) will be discovered in some lineages and the evolution rates in these lineages rise accordingly (Sunet al. 1999; Mülleret al. 2000; Puttaguntaet al. 2000; Frönicke and Wienberg 2001). In our view, however, such increase will not equalize the disparate evolution rates that we have observed. The plausible reasons for the faster chromosomal evolution in the genus Drosophila have been discussed elsewhere (Ranzet al. 2001).

What factors can account for the remarkable variation in evolution rate observed between the chromosomal elements of Drosophila? Factors affecting the rate of chromosomal evolution can be classed in two groups: mutational and selective. A higher rate of inversion fixation would be expected if mutation rate were higher (other things being equal). In Drosophila, transposable elements (TEs) have been implicated in the origin of natural inversions, which can originate through ectopic recombination between TE copies located in opposite orientation in different sites of the same chromosome (Montgomeryet al. 1991; Lim and Simmons 1994; Andolfattoet al. 1999; Cáceres et al. 1999a, 2001; Mathiopoulos et al. 1999). Thus, a higher mutation rate could be due to a higher proportion of repetitive DNA or to a higher recombination rate. If we consider the recently sequenced genome of D. melanogaster (Table 4), it would appear that the X chromosome, which in our study showed the highest breakpoint density, does not show a TE density higher than that of the autosomes (Rizzonet al. 2002). On the other hand, chromosome X does possess a microsatellite density that is at least twice that of any of the autosomes (Kattiet al. 2001), a fact that likely reflects the presence of several repeated sequences that are exclusive of the X chromosome euchromatin or more abundant in the X chromosome than in the autosomes (Huijseret al. 1987; Pardueet al. 1987; Waring and Pollack 1987; Lowenhauptet al. 1989; Dibartolomeiset al. 1992; Bachtroget al. 1999). Moreover, the microsatellite density of the autosomes (Table 4) parallels their evolution rates (Table 3). This is an intriguing observation because micro-satellite sequences can generate unstable secondary structures (Mitas 1997; Mooreet al. 1999) that could be involved in the origin of chromosome rearrangements (Pletcheret al. 2000; Puttaguntaet al. 2000). So far, however, no evidence for the implication of microsatellites in the origin of Drosophila inversions has been found. Obviously, more data are needed on the distribution of TEs and repeated sequences among chromosomal elements in other Drosophila species apart from D. melanogaster.

The second factor that might be affecting inversion production is recombination rate. If we assume that ectopic recombination is correlated with regular meiotic recombination (Montgomeryet al. 1991), then meiotic recombination rates in D. melanogaster can be considered in search of a pattern. Chromosome X exhibits a higher average recombination rate (3.35 cM/Mb) than the autosomes (Table 4) in good agreement with its faster chromosomal evolution. However, the average recombination rate of chromosomal arm 2L, which shows the slowest evolution rate, is comparable (or superior) to that of both arms of chromosome 3 (Table 4). Therefore, no consistent effect of recombination on evolution rate is apparent. On the other hand, it is clear that recombination rates vary between Drosophila species (Trueet al. 1996; Cácereset al. 1999b) and the D. melanogaster rates may not have a genus-wide validity.

Another group of factors comprises those selective causes affecting the probability of fixation of inversions. For instance, Charlesworth et al. (1987) showed that the X chromosome should evolve faster than the autosomes due to a higher fixation probability of underdominant and favorable partial or fully recessive rearrangements. Also, when the chromosomal rearrangements have an antagonistic effect in the two sexes, they will invade the population under a wider range of conditions if they are X linked than if they occur in the autosomes (Rice 1984). These predictions might help to explain the fast evolution rate of the X chromosome but would not explain the rate variation between the autosomes. A higher evolution rate could also be due to less functional constraints, as would be expected in regions with low gene density (Landeret al. 2001). Fixed inversions are more likely to have their breakpoints between genes (as found by Cireraet al. 1995 and Cácereset al. 1999a) than within transcriptions units (as in Schneuwlyet al. 1987). In the latter case, an inversion would probably have a strong deleterious effect and would be quickly eliminated by natural selection before fixation. The average density in Drosophila is one gene per 9 kb but there is substantial variation in gene density throughout the genome (Adamset al. 2000). Nevertheless, the average gene density (as inferred from release 2 of the D. melanogaster genome sequence; Adamset al. 2000) seems to be comparable for elements X, 2L, and 3L and only slightly higher for elements 2R and 3R (Table 4). Thus, no systematic correlation is apparent between gene density and evolution rate in Drosophila. However, we have to take into account that there can be local differences in gene density within chromosomal elements. Jabbari and Bernardi (2000) pointed out that the gene concentration in GC-rich regions is sevenfold higher than that in GC-poor regions in the Drosophila genome and our results suggest comparable density differences within the D. repleta X chromosome. Chromosomal arms rich in gene-poor intervals might have more fixed breakpoints than arms with little or no variation in gene density.

Chromosomal inversions may have diverse effects at the genetic and phenotypic level, which will affect their probability of fixation in a complex manner. For instance, in heterokaryotypes, inversions reduce recombination rate in the inverted chromosomal segment (Navarroet al. 1997) but may increase it in the nonhomologous chromosomes (Lucchesi and Suzuki 1968). Accordingly, the fate of an inversion is considered to depend strongly on the epistatic combinations of alleles caught by the inversion at the moment of its appearance (Charlesworth and Charlesworth 1973; Charlesworth 1974) and the species recombination rate (Cácereset al. 1999b). Finally, in some cases (e.g., the sex ratio arrangement of chromosome X), inversions may be associated with meiotic drive alleles and be preferentially transmitted to the offspring (Ashburner 1989). Given these manifold effects of inversions, it seems improbable that a single factor explains the variation in evolution rate among Drosophila chromosomal elements and we must cautiously conclude that several causes, some of them discussed above, contribute to this variation.

Conserved chromosomal segments and functional constraints: The chromosomes of D. repleta can be regarded as a mosaic of relatively small segments homologous to those in D. melanogaster chromosomes. Our estimates of the number of inversions fixed in elements A and D allow us to predict that the average size of such segments is 92 kb for chromosome X and 214 kb for chromosome 4 (Table 3). The comparison of the physical maps of D. melanogaster and D. repleta led us to the identification of 9 and 13 conserved segments with two or more consecutive markers in elements A and D, respectively (see Figures 2 and 3). There were also 39 and 29 singletons, segments that contained only a single marker. In chromosome X the size of the 9 segments with two or more consecutive markers ranged from 4.4 to 576 kb with an average length of 170.7 kb. Likewise, in chromosome 4 the size of the 13 conserved segments ranged from 152 to 939 kb with an average length of 386.9 kb. In both cases, the average length of the observed conserved segments is bigger than the predicted average size (Table 3). This is expected because there is a discovery bias that favors big conserved segments at the initial stages of comparative mapping and also because only conserved segments delimited by two or more markers have been considered to estimate the average length of observed segments (Nadeau and Sankoff 1998b).

Blocks of genes that are conserved during long periods of time may represent gene combinations that interact functionally and are therefore maintained together by natural selection, the so-called “functional constraints” hypothesis (Maieret al. 1993; Randazzoet al. 1993; Wright 1996). However, because all genomes are phylogenetically related, colinear groups of genes may also reflect the fixation of a limited number of genomic rearrangements with random breakpoints since both species diverged, the “random breakage” (RB) hypothesis (Nadeau and Taylor 1984; Nadeau and Sankoff 1998a). Previous comparative mapping results in Drosophila (Gonzálezet al. 2000; Ranzet al. 2001) and other organisms (Nadeau and Sankoff 1998a; Huynenet al. 2001; Landeret al. 2001) have found little evidence for functional constraints. The RB hypothesis can be tested comparing the observed length distribution of conserved segments with that predicted under the RB hypothesis, which will approximate a negative exponential distribution (Figure 4). It can be seen that, in both chromosomal elements, the empirical distribution fits in general inside the theoretical distribution as expected because only a subset of all conserved segments has been detected. Only one segment in each chromosome seems to depart significantly from the expectations (Figure 4). The size and gene content of these segments should be further investigated.

We can look at the functional constraints hypothesis from a different perspective. There are a few examples of gene complexes in D. melanogaster chromosomes X and 3L whose members show a coregulated expression. We can ask whether or not such complexes are conserved in D. repleta. The achaete-scute complex (AS-C) has been studied extensively (reviewed in Modolell and Campuzano 1998). It spans ~90 kb of the X chromosome where only six transcription units are separated by very large stretches of nontranscribed DNA. This DNA contains many cis-regulatory sequences that coregulate the achaete and scute (but not the lethal-of-scute and asense) genes of the complex. The molecular organization of the 210-kb D. melanogaster segment delimited by the markers 125H10-65F1 has been studied in D. repleta. This 210-kb segment contains, among others, the genes of the AS-C (see appendix). When the physical maps of this region are compared, only a segment of ~130 kb, which includes the genes achaete and scute, has been conserved. This gene complex is also conserved in D. virilis (Beamonte 1990). Thus, the molecular organization of this gene complex seems to have been preserved during the 80–124 myr of divergence of these species. Two gene complexes of chromosome 3L also appear to show coregulated expression. The genes araucan (ara) and caupolican (caup), two members of the Iroquois complex, have similar patterns of expression and apparently share cis-regulatory sequences (Gómez-Skarmetaet al. 1996). They are closely linked in D. melanogaster (comprising a genomic segment of ~40 kb) and also in D. repleta (as shown by the coincident hybridization sites of caup, ara, and DS08512). Finally, knirps (kni) and knirps-related (knrl) are two neighboring and functionally equivalent genes mapping to a region of 100 kb (Lundeet al. 1998). They are affected by a cis-acting regulatory sequence (ri) lying immediately upstream of the kni transcription unit. A chromosomal segment of 245 kb around the kni locus (comprising kni, DS01369, and DS00239) is conserved in D. repleta, suggesting that both genes and their regulatory sequences have not been disrupted. In summary, the conservation of these three small gene complexes suggest that natural selection may play a role in some (perhaps exceptional) cases to keep together functionally related Drosophila genes.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

Expected (□) and observed (▪) distribution of the length of conserved segments under the random breakage hypothesis (Nadeau and Taylor 1984; Nadeau and Sankoff 1998b). The empirical distribution fits in general inside the theoretical distribution as expected because only a subset of all conserved segments has been detected (see text for details).

Overall our results are in agreement with a modular organization of the Drosophila genome (Ranzet al. 2001). Thus, the genome of Drosophila can be seen as a mosaic of independent modules that can change their localization within the euchromatin without loss of function. Usually these modules change their localization within the chromosomal arm and only occasionally between chromosomal arms. Each module may consist of a gene plus its regulatory sequences (as proposed by Ranzet al. 2001) or perhaps a small group of nearby genes. If Drosophila euchromatin possess expression domains organized by insulators or boundary elements (Gerasimovaet al. 2000; Mongelard and Corces 2001), this undoubtedly will influence the molecular consequences of inversion breakpoints. We can speculate that breaks taking place within trancriptionally independent domains will have more disturbing consequences than those occurring between domains. The rough agreement between the size of these expression domains (Gerasimovaet al. 2000) and the average size of the conserved segments (Table 3) is certainly intriguing and deserves more work. Recent comparisons of yeast genomes show the prevalence of small inversions in gene order evolution between Saccharomyces and some Candida species (Llorenteet al. 2000; Seoigheet al. 2000). A similar result was found when comparing zebrafish and human genomes (Postlethwaitet al. 2000), supporting the hypothesis that inversions have been a more frequent force in the shaping of vertebrate karyotypes than translocations. Altogether these results suggest a major role for inversions in the genome shuffling process (Huynenet al. 2001).