INTRODUCTION

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION METHODS WEB SITE REFERENCES REFERENCES

Transposable elements (TEs) have been found in organisms as different as bacteria, nematodes, yeast, plants, fishes, and mammals including humans. Evidence is accumulating that they are agents of genome restructuring and, as such, appear to be a major constituent of genomes (Kidwell and Lisch 1997; Shapiro 1999; Tomilin 1999). TEs have a transposition capacity that enables them to invade the genome, leading to insertional mutations and chromosomal rearrangements. Therefore, organisms have developed various mechanisms to limit their number. However, the relative importance of the forces that control the dynamics of TEs in natural populations is still controversial (Biémont et al. 1997; Charlesworth et al. 1997; Kidwell and Evgen'ev 1999; Nuzdhin 1999). It has been proposed that containing the number of TE copies must involve either selection against rearrangements caused by ectopic recombination between TE insertions (Langley et al. 1988; Charlesworth et al. 1994, 1997; Zhang and Peterson 1999; Gray 2000) or selection against TE-induced mutations (Biémont et al. 1997). There are data consistent with both of these hypotheses. For example, in humans, ectopic exchange between Alu sequences seems to be more important in producing deleterious mutations (0.2%-0.3% of diseases) than insertional mutagenicity (0.1%; Roy et al. 1999), whereas ~50% of mutations in Drosophila melanogaster are attributable to TE insertions (Finnegan 1992). If the ectopic exchange in a region is proportional to the meiotic exchange in that region (Langley et al. 1988; Petes and Hill 1988; Montgomery et al. 1991; Goldman and Lichten 1996, 2000), then the number of TE insertions should be negatively correlated with the recombination rate. The same is true for selection against insertional mutations, because selection is weaker in regions of low recombination (Hill and Robertson 1966, Charlesworth et al. 1993; Kliman and Hey 1993). Hence, the density of FL L1 elements has been found to be negatively correlated with recombination rate in humans (Boissinot et al. 2001), suggesting that purifying selection against TEs is occurring. However, the distribution of TE insertion sites over the chromosomes of D. melanogaster shows no evident relationship between the frequency of recombination and TE density (Hoogland and Biémont 1996), although it is well known that TEs accumulate in heterochromatic regions such as the chromocenter and the bases of chromosomes, which are characterized as sites where there is little or low recombination (Charlesworth et al. 1992a,b). The TEs of Drosophila also seem to be more abundant on chromosome 4 and within some inversions, both of which have a low recombination rates (Montgomery et al. 1987; Langley et al. 1988; Eanes et al. 1992; Sniegowski and Charlesworth 1994). In plants, many elements are located in clusters at the paracentric heterochromatin (Brandes et al. 1997), the copia-like elements are concentrated in the centromeric regions (Heslop-Harrison et al. 1997), and the regions flanking the centromeres are densely populated by TEs (Copenhaver et al. 1999). It is becoming increasingly evident, however, that TEs are major constituents of the centromeric regions in Drosophila (Pimpinelli et al. 1995; Zuckerkandl and Hennig 1995; Pardue et al. 1996; Eissenberg and Hilliker 2000), and that this is not merely the result of passive accumulation in such regions caused by the absence of strong forces tending to eliminate them.

A recent study in Caenorhabditis elegans produces even more puzzling results, with a recombination rate that was found positively correlated with the amount of transposons (DNA-based elements), but not with the amount of LTR and non-LTR retrotransposons (RNA-based elements; Duret et al. 2000). The importance of the forces regulating TE distribution may thus vary according to the genome, indicating the need for comparative studies (Kidwell and Evgen'ev 1999). The difference between the findings from studies of C. elegans and Drosophila may also be attributable to the type of data used, population data in the case of Drosophila, data from the genome of a single individual in that of C. elegans. Because we now possess the sequence of the entire D. melanogaster genome, we analyzed the distribution of its TEs in relation to the recombination rate. LTR and non-LTR retrotransposons tend to accumulate in regions of low recombination, but with no clear tendency to parallel the recombination rate along the chromosomes. There is, however, a negative correlation between the recombination rate and the density of transposons.

	RESULTS

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION METHODS WEB SITE REFERENCES REFERENCES

Table 1 shows the reference used for each element and the number of sequences retrieved from the D. melanogaster genome. Among the 54 TE families, 10 were transposons, 28 LTR retrotransposons, and 16 non-LTR retrotransposons. We thus collected 1007 insertions, 185 transposons (DNA-based elements), 572 LTR retrotransposons, and 250 non-LTR retrotransposons (RNA-based elements). No copies of P and HeT-A elements were identified in the genome. Figure 1 reveals that TEs accumulate mainly in pericentromeric regions and in chromosome 4, but not in the telomeric regions, which is consistent with what is usually observed (Charlesworth et al. 1992a,b).

View larger version (14K): [in this window] [in a new window]   Figure 1   Box diagram of TE densities according to specific chromosomic regions, showing the minimal, quartile 1, median, quartile 3, and maximal values of the TE densities found in 0.25-Mb genome fragments.

TE Density According to Recombination Rate

Because the accumulation of TEs in Drosophila heterochromatin is likely to be a consequence of the peculiar properties of the heterochromatic material (Pimpinelli et al. 1995), it is not possible to establish a direct relationship between the accumulation of TEs in heterochromatin and the recombination rate (Garcia Guerreiro and Fontdevila 2001). As the pericentromeric regions and chromosome 4 are heterochromatic regions, the relationship between TE density and recombination rate was studied both with and without these regions to avoid statistical bias. Moreover, as telomeric regions have peculiar evolutionary dynamics (Marais et al. 2001), we also performed the analyses with and without these regions.

Figure 2 shows the relationship between the mean density of TEs and the recombination rate for each TE type. The density of LTR retrotransposons (Fig. 2a) was high for low values of recombination, but was homogeneously distributed afterward, and the density of non-LTR retrotransposons did not seem to follow any clear tendency (Fig. 2b). However, the Spearman rank correlations were significant ( = 0.13, P < 0.01 for LTR retrotransposons,  = 0.15, P < 0.01 for non-LTR retrotransposons), and the ² values, calculated for four classes of recombination each containing 25% of the values, were also significant (² = 113.12, P < 0.0001 for LTR retrotransposons, ² = 52.46,  P < 0.001 for non-LTR retrotransposons), with accumulations of insertions in the lower recombination rate class. We did the calculation again, this time without the pericentromeric and telomeric regions and chromosome 4. The Spearman rank correlations were no longer significant ( = 0.02, P > 0.05 for LTR retrotransposons,  = 0.08, P > 0.05 for non-LTR retrotransposons), although the chi² statistics were still highly significant for LTR retrotransposons (² = 34.42, P < 0.0001) and to a lesser degree, significant for non-LTR retrotransposons (² = 8.94, P = 0.03), as a result of accumulations of insertions in the lowest recombination class value. Therefore, we conclude that the densities of LTR and non-LTR retrotransposons do not parallel the recombination rate along the chromosomes, although these TEs do tend to accumulate in regions with low recombination rates.

View larger version (35K): [in this window] [in a new window]   Figure 2   Density of LTR retrotransposons (a), non-LTR retrotransposons (b) and transposons (c) according to recombination rate of the genome of Drosophila melanogaster.

There was a significant negative correlation between the density of transposons and the recombination rate (Spearman rank correlation coefficient,  = 0.29, P < 0.0001; Fig. 2c), which remained significant after eliminating the pericentromeric, telomeric, and chromosome 4 regions ( = 0.19, P < 0.0001). The chi² tests on transposon density over the four recombination classes defined above were highly significant (² = 197.84, P < 0.0001) with an accumulation of transposons in the lowest recombination rate class when all of the genomic regions were taken into consideration. The ² remained highly significant after eliminating the telomeric and pericentromeric regions and chromosome 4 (² = 71.78, P < 0.0001). We thus conclude that the density of transposons decreases when the recombination rate increases.

We performed a detailed analysis for the LTR retrotransposons on individual chromosomes. Such analysis was not done for transposons or non-LTR retrotransposons because of the lack of data for a reliable statistical analysis. The telomeric, pericentromeric, and chromosome 4 regions were not included in the analysis. No significant linear correlation between the recombination rate and the density of LTR retrotransposons was found for any of the chromosome arms (Spearman rank correlation,  = 0.13, P > 0.05 for 2L;  = 0.12, P > 0.05 for 2R;  = 0.16, P > 0.05 for 3L;  = 0.09, P > 0.05 for 3R), although all values were negative. These negative correlation values were due to an accumulation of LTR retrotransposon copies in regions of low recombination, which was significant for the 2L arm (Fig. 3a: ² = 43.33, P < 0.0001) and the 3R arm (Fig. 3d: ² = 14.56, P < 0.01), but not for the 2R or 3L arms. The X chromosome gave different results. Because of the narrow range of recombination values in the middle portion of this chromosome, the distribution of LTR retrotransposon insertions was analyzed along the entire chromosome by use of nonparametric statistics. The X arm was split into 5000-bp fragments, which were coded 1 if they contained at least one TE insertion and 0 if no insertion was detected. This allowed us to calculate the Variance of Ranks, VR, which detects aggregation in the middle of the sequence of 0 and 1 (low variance value) or at its ends (high variance value), and the Multiple Pool, MP, which detects aggregation at various regions in the sequence (see Aulard et al. 1995, for details of these statistics). These tests (VR = 3.5, P < 0.001; MP = 3.73, P < 0.001) showed that copies of LTR retrotransposons were at least grouped near the X centromere. A one-sample Kolmogorov-Smirnov test, which detects aggregation in a given region, revealed a significant group of LTR retrotransposon copies around 3.33 Mb, outside but not far from the telomeric region (KS = 0.15, P < 0.01). Therefore, there was no tendency for LTR retrotransposons to accumulate in the X telomere despite its low recombination rate.

View larger version (44K): [in this window] [in a new window]   Figure 3   Density of LTR retrotransposons according to the recombination rate in the chromosome arms 2L (a), 2R (b), 3L (c), and 3R (d) in the genome of D. melanogaster.

To make it easier to compare our data with that from the population analysis (Hoogland and Biémont 1996), we calculated the Spearman rank correlation between recombination rates and the mean density found in each recombination rate class of the elements roo/B104, mdg3, mdg1, copia, 412, 297, I, and hobo, for which we had reliable data from both studies. Once again, the pericentromeric, telomeric, and chromosome 4 regions were excluded. The  values were not significant, and fell within the range of values of the population study (roo/B104,  = 0.14, P > 0.05; mdg3,  = 0.25, P > 0.05; mdg1,  = 0.16, P > 0.05; copia,  = 0.25, P > 0.05; 412,  = 0.10, P > 0.05; 297,  = 0.21, P > 0.05; I,  =  0.04, P > 0.05; hobo,  = 0.14, P > 0.05), suggesting that there was no great difference between the two approaches.

Gene Density According to Recombination Rate

No significant correlation was detected between gene density and recombination rate (Spearman rank correlation,  = 0.09, P > 0.05), suggesting that the relationship between TE density and recombination rate was not biased by the amount of genes.

TE Density on Autosomes and on the X Chromosome

We compared the TE density on autosomes and the X chromosome for each class of TEs and for all of the TEs taken globally. When all regions were considered, the chi² statistics were significant for transposons (² = 8.40, P < 0.01) and non-LTR retrotransposons (² = 5.48, P = 0.02), both of which showed a deficit in the number of copies on the X chromosome. ² was not significant for LTR retrotransposons (² = 3.69, P > 0.05) or for all of the TEs pooled (² = 0.92, P > 0.05). Eliminating the pericentromeric, telomeric, and chromosome 4 regions rendered the tests nonsignificant (² = 1.53, P > 0.05 for non-LTR retrotransposons; ² = 1.63, P > 0.05 for transposons; ² = 1.58, P > 0.05 for all TEs pooled) except for LTR retrotransposons (² = 9.41, P = 0.002), which showed an accumulation of copies on the X chromosome.

	DISCUSSION

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION METHODS WEB SITE REFERENCES REFERENCES

As in all previous studies, our study confirms the accumulation of TEs in centromeric and pericentromeric regions, with a low TE density in subtelomeric regions. However, after eliminating these specific regions, the densities of LTR and non-LTR retrotransposons appeared to be high in regions with low recombination rates, but to have no direct, linear relationship with the recombination rate along the chromosome arms. In contrast, a negative correlation was found between transposon density and recombination rate. The accumulation of TEs in regions of reduced recombination rate is compatible both with selection acting against deleterious mutations caused by TE insertions, and with selection acting against chromosomic rearrangements caused by ectopic recombination between TE copies, because, in both cases, selection is expected to be weaker in regions of reduced recombination. The nonlinearity in the relationship between recombination and LTR and non-LTR retrotransposon densities might result from nonlinearity in Hill-Robertson effects and ectopic exchanges with meiotic recombination. However, the observations that the transposon density linearly decreased with recombination rate and that the accumulation of LTR retrotransposons was statistically significant only for a few chromosome arms suggest that selection is not sufficient to explain the distribution of TEs along chromosomes in the D. melanogaster genome.

An absence of linear correlation between retrotransposon density and recombination rate when centromeric and pericentromeric regions were eliminated from the calculation is congruent with previous population studies (Hoogland and Biémont 1996), in which, however, no negative correlation was detected between transposon density and recombination rate. A major difference between the two studies is that the population approach involved only two transposons, P and hobo. The present study relied on 10 transposons, without the P element, which was not detected in the sequenced genome. The significant negative correlation observed between transposons and recombination rate, strongly supports the hypothesis that selection acts against these TEs. But if this is so, then why is the relationship between transposon density and recombination negative in Drosophila and positive in the nematode (Duret et al. 2000)? Differences in meiotic pairing and recombination mechanisms could account for the contrasting relationships between TEs and recombination rate in these two species. But Drosophila and C. elegans use the same recombination-independent mechanisms to align homologs (McKim et al. 1998), and ectopic recombination is several orders of magnitude less frequent than allelic recombination in both organisms (Virgin and Bailey 1998). Yeast, however, initiates homolog colocalization and alignment by homology-dependent DNA-DNA interactions (Kleckner and Weiner 1993), and shows only small, but significant, differences between ectopic and allelic recombination frequencies (Kupiec and Petes 1988; Goldman and Litchen 1996). TEs are not randomly distributed in yeast, but are mainly located in genes transcribed by RNA polymerase III, such as tRNA genes (Kim et al. 1998). It is, however, difficult to explain why the non-uniform homolog pairing in Saccharomyes cerevisiae (Kleckner and Weiner 1993), as opposed to the close homolog alignment in Drosophila and C. elegans, could account for the differences between TE distribution in these two species and in the yeast genome.

Could a difference in breeding system account for the reverse correlations between transposon density and recombination rates in Drosophila and the nematode? If the deleterious effects of TEs are mostly recessive, then selection against TEs should be most effective in populations with high levels of homozygosity (Wright and Schoen 1999). In contrast, because ectopic exchanges occur preferentially between heterozygous TE insertions (Montgomery et al. 1991, Charlesworth and Charlesworth 1995), according to this selective model, selection should be most effective in out-crossing populations. Because the C. elegans breeding system is presumed to be mostly inbreeding (Baird et al. 1992), unlike that of Drosophila, the effects of selection can be expected to differ in these two species. Moreover, because self-fertilization can theoretically be expected to reduce the recombination rate (Morgan 2001), studying the relationship between recombination rate and TE density would not easily detect selection against TE-induced mutations in C. elegans. However, the pres- ent sequenced Drosophila genome is derived from a laboratory strain that is undoubtedly homozygous, whereas the genomes of individuals from natural populations are likely to be highly heterozygous. Therefore, homozygosity may have interfered with the mechanisms controlling the TE copy number in this specific genome during its long history in the laboratory, during which the loss or mobilization of specific TEs cannot be excluded (Biémont et al. 1987).

The contrasting relationships between recombination rate and transposon density in the nematode and Drosophila could be attributable to the very large population size in Drosophila. A large population means that selection against TE insertions are likely to outweigh drift (Charlesworth and Charlesworth 1995), so that selection can interfere with recombination, leading to fewer transposon insertions in regions of high recombination in which selection is strongest. If selection is a weaker force in the nematode, then it should be possible to detect the preference for transposons to become inserted in regions of high recombination rate, as a result of their mechanisms of transposition in this species (Duret et al. 2000). Although it is difficult to compare effective population sizes between two species, the specific reproductive system of the nematode, which is assumed to be mostly inbreeding (Baird et al. 1992), may have an effect similar to that of a smaller population. This means that the interaction between selection, effective population size, and recombination is of great importance in the structuring of genomes.

The presence of high densities of TEs in regions of low recombination and the significant negative correlation between transposons and recombination rate suggest that selection may act against the insertion of TEs. According to this hypothesis, fewer TE insertions can be expected on the X chromosome than on the autosomes. Because the males are hemizygous in Drosophila, deleterious TE insertions on the X should be selected against to a greater extent than insertions on the autosomes (Montgomery et al. 1987; Langley et al. 1988; Charlesworth et al. 1994). We did not detect any reduction in TE density on the X chromosome compared with the autosomes for the TEs of the sequenced genome. This observation and the fact that differences in TE amount between the autosomes and the X chromosomes have been observed for some elements, but not for others, in studies of populations of Drosophila (Montgomery et al. 1987; Biémont 1992; Charlesworth et al. 1992a,b; 1994; Biémont et al. 1997), suggest that selection against the insertional effects of TEs is not the main force controlling TE copy number. This is also consistent with the data on C. elegans, which shows no evidence that there are fewer TE insertions on the X chromosome than the autosomes (Duret et al. 2000). We must, however, consider the possibility that the TEs may have only been mobilized recently within the genome of the Drosophila stock used for sequencing, and that selection had not yet reduced the TE copy number on the X chromosome.

We cannot rule out the possibility that the distribution of TEs is not directly associated with recombination, but depends on other factors that could themselves be associated, possibly fortuitously, with recombination, and so depends on the individual genome considered and the species. This is illustrated by the observations that LTR retrotransposons aggregated in a small region of the X chromosome, but were homogeneously distributed in regions of middle and high recombination rates, and accumulated in regions of low recombination other than pericentromeric regions only in the 2L and 3R chromosome arms.

The distribution of target sites for TE insertions could vary with the DNA base composition (Sharp and Matassi 1994), and this would account partly for the distribution of TEs along the chromosomes. In the C. elegans and Drosophila, the G+C content is positively correlated with the recombination rate, both in noncoding regions and in synonymous positions of codons (Marais et al. 2001). This might lead to a link between the distribution of target sites for TE insertions and the recombination rate. Many TEs seem to be inserted in AT-rich, late-replicating, DNA regions (Le et al. 2000), and their target insertion sites are often a succession of A and T. For example, the human L1 elements show target specificity for TTTTAA, which leads to a linear negative relationship between L1 density and GC richness. This has also been shown for LTR retrotransposons 1731, 17.6 in D. melanogaster, TRIP in sea urchin, Mag of Bombyx mori (Springer et al. 1995), and certain retroviruses (Bernardi et al. 1985), but it has also been shown that TEs are globally AT-rich (Sharp and Matassi 1994; Lerat et al. 2000, 2002), so that insertions of numerous TE copies in a region leads to a low GC value. On the other hand, TEs could accumulate in low gene density regions, as reported for the Arabidopsis genome (The Arabidopsis genome initiative 2000), and could be associated with low GC content (Kumar and Bennetzen 1999; for review, see Lin et al. 1999; Adams et al. 2000; Jabbari and Bernardi 2000), and reduced recombination rate (for review, see Kliman and Hey 1993; Charlesworth 1994; Fullerton et al. 2001; Marais et al. 2001). In the present study, however, we did not detect any relationship between gene density and recombination rate. Moreover, the GC-rich SINE elements are located in GC-rich regions (Korenberg and Rykowski 1988; Boyle et al. 1990; Jurka 1997), and the region around P insertion sites in Drosophila are GC-rich (Liao et al. 2000). In humans, the distributions of young and old copies of Alu elements have been found to be different (Smit 1999), suggesting that Alus integrate randomly but are preferentially fixed in GC-rich DNA as the result of some force of selection. These data do not allow us to conclude that there is any specific or general relationship between base composition and TE distribution.

TEs may insert preferentially in regions in which other sequences are already inserted, making the correlation between TE density and recombination rate merely fortuitous and variable from genome to genome and species to species, especially if such sequences could be recombinogenic, as postulated for the CeRep sequences of C. elegans (Cangiano and La Volpe 1993; Barnes et al. 1995). If the specific sequences in which the TEs are inserted vary according to the TE family, then the association between recombination rate and TE density will vary with the TE considered. For example, microsatellites accumulate preferentially in genome regions in which recombination is infrequent (Charlesworth et al. 1994). However, the density of microsatellites is not influenced by the recombination rate in D. melanogaster (Bachtrog et al. 1999) and, there is no strong evidence for the overall insertion of different TEs in specific sequences, such as micro and mini-satellites. We have evidence, however, that some of the microsatellite sequences may, in fact, result from the TEs themselves (Nadir et al. 1996; Jarne et al. 1998; Toth et al. 2000), although a high density of microsatellites does not always coincide with a high density of TEs, especially in Arabidopsis thaliana (Schlötterer 2000), even though retrotransposons have been found near microsatellites in barley (Kalendar et al. 1999).

These findings suggest that various features (local genomic composition and structure, chromatin conformation, DNA nick repair, number of DNA replications, effective population size, reproductive system, and history of the host) could variously influence and even blur the impact of natural selec tion acting against the TE insertions along the chromosome.

	METHODS

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION METHODS WEB SITE REFERENCES REFERENCES

Sequence Data and Locations of Transposable Elements

The sequences of the D. melanogaster chromosome arms X, 2L, 2R, 3L, 3R, and 4 were retrieved from the unannotated version 1 of the genome (Adams et al. 2000; BDGP 2000). The retrieved data (all tracks of N omitted) thus totaled 114.5 Mb, corresponding to 64% of the whole genome sequence (the actual sequence represented 95% of the total euchromatin). Therefore, all of the TE sequences collected were from the euchromatic part of the genome, because most of the heterochromatin, including the Y chromosome, was not sequenced. Heterochromatin is composed of many transposable elements, mostly in an inactive state and in the form of defective sequences. Hence, the TE sequences studied here represent only a fraction of all the TE sequences in the Drosophila genome. Our analysis is thus pertinent for comparison with data on TE chromosomal locations obtained from population analyses in which the in situ hybridization technique used gives information on the copy number of the TEs that are inserted in the euchromatin, along polytene chromosomes.

A bank of reference sequences for TE families was constituted with sequences retrieved from the FlyBase database (FlyBase 1998; http://flybase.bio.indiana.edu/). When the sequences from this database corresponded to incomplete copies of LTR retrotransposons, we searched for full-length copies with the BLAST program in the high-quality genomic clone sequences from BDGP/EDGP (http://www.fruitfly.org). The list was completed with sequences retrieved from EBI (The European Bioinformatics Institute 2001; http://www.ebi.ac.uk/index.html, sequence set: ftp://ftp.ebi.ac.uk/pub/databases/edgp/sequence_sets) and with the pilgrim sequence (Costas et al. 2001). The distributions of TE insertions in the whole shot-gunned genome sequence were then analyzed by use of the above reference sequences and by the program RepeatMasker (A.F.A. Smit and P. Green, unpubl.; http://repeatmasker.genome.washington.edu/cgi-bin/RM2_req.pl). Because many TE insertions were consensus or mixed sequences, we only worked on localization data of TEs. For the LTR retrotransposons, we considered as insertions only the retrieved sequences with at least one complete or one incomplete typical LTR to avoid wrongly attributing very divergent copies to a given element. Each match was then checked to decide whether different copies of a given LTR retrotransposon were present, or only one copy detected by different matches. For non-LTR retrotransposons and transposons, we considered those retrieved sequences that were >400 bp, so as to eliminate short, deleted, highly divergent sequences. Each match was also checked as above.

The TE density was estimated from the number of TE insertions per base pair in the sequence fragments considered, excluding the number of N. For each class of recombination rate (see below), all genome regions corresponding to a given range of recombination rate were pooled. To analyze the relationship between TE density and recombination rate, the Drosophila genome was cut into nonoverlapping fragments of 0.25 Mb. The sequences corresponding to the telomeric, centromeric, and chromosome 4 regions were defined by use of the Gadfly annotations (http://hedgehog.lbl.gov:7081/annot/). When these regions were to be excluded from the analysis, the genome was cut into nonoverlapping fragments after removing the corresponding sequences.

Estimation of Gene Densities along the Chromosomes

Because DNA sequences on each chromosome arm were not yet fully annotated, we searched for the chromosomal location of the 13376 known and predicted genes of the Drosophila genome by using BLAST and the data on transcribed gene sequences of release 1 (na.gadfly.dros.RELEASE1 at the Berkeley Drosophila Genome Project site: http://www.fruitfly.org). Gene density along the chromosomes was thus determined as described above for the TEs. To analyze the relationship between gene density and recombination rate, the Drosophila genome was cut into nonoverlapping fragments of 0.25 Mb.

Estimation of Recombination Rate

The rate of recombination along the chromosomes was determined by use of a procedure similar to that described by Kliman and Hey (1993). The D. melanogaster genetic map data was taken from FlyBase (FlyBase 1998). We selected the 892 loci that had been located in both the genetic map and the genomic sequence. The recombination rate was estimated for each chromosome arm by taking the derivative of the best fitting polynomial function of the genetic distance versus the nucleotide coordinate in the genomic sequence. Second-degree polynomial curves fitted the data set (r2 > 0.97) well for all chromosome arms. The fitting by the polynomial function is clearly correct for the points in the center of the curve, but deviates for the subtelomeric (chromosomal sections 20, 40-41, 80-81) and pericentromeric (chromosomal sections 1, 21, 60-61) regions, in which the recombination rates are low (Kliman and Hey 1993). To analyze the relationships between recombination rate and TE density, and between recombination rate and gene density, for each genomic fragment of 0.25 Mb (see above), the recombination rate was estimated from the value of the derivative of the polynomial curve at the middle position of the fragment. The 0.0 cM/Mb value was assigned to the fragments of the telomeric and pericentromeric regions and of chromosome 4. We also defined 49 classes of recombination rate from 0.0 to 4.9 cM/Mb at intervals of 0.1 cM/Mb. The telomeric and pericentromeric regions, and the chromosome 4 were assigned to the 0.0-0.1 cM/Mb class of recombination.

	WEB SITE REFERENCES

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION METHODS WEB SITE REFERENCES REFERENCES

http: //flybase.bio.indiana.edu/; the FlyBase database of the Drosophila Genome Projects and community literature. FlyBase 1998.

http: //hedgehog.lbl.gov:7081/annot/; the genome annotation database. GadFly 2000.

http: //repeatmasker.genome.washington.edu/cgi/RM2_req.pl; RepeatMasker Smit, A.F.A. and Green, P. 2000.

http: //www.ebi.ac.uk/index.html; the European Bioinformatics Institute 2001.

http: //www.fruitfly.org/index.html; the Berkeley Drosophila Genome Project 2000.