INTRODUCTION

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION METHODS WEB SITE REFERENCES REFERENCES

The genes of Caenorhabditis elegans appear to have an unusually rapid rate of evolution. The substitution rates of many C. elegans genes are twice those of their orthologs in non-nematode metazoans (Aguinaldo et al. 1997; see Fig. 3 in Mushegian et al. 1998). Even among nematodes, the C. elegans small subunit ribosomal RNA gene evolves faster than its orthologs in most of the major clades (see Fig. 1 in Blaxter et al. 1998). It has been estimated that two-thirds of C. elegans protein-coding genes evolve more rapidly than their Drosophila orthologs (Mushegian et al. 1998). In vertebrates at least, the rate of nucleotide substitution is correlated with that of chromosomal rearrangement (Burt et al. 1999).

Ranz et al. (2001) reported that Drosophila chromosomes rearrange at least 175 times faster than those of other metazoans, and at a rate at least five times greater than the rate of the fastest plant genomes. However, no Caenorhabditis rate data existed to compare with the Drosophila data. Given their fast rate of nucleotide substitution, we guessed that Caenorhabditis genomes might have a fast rate of rearrangement. Here, we have estimated the rate of rearrangement since the divergence of C. elegans from its sister species Caenorhabditis briggsae, using the complete C. elegans genome sequence (The C. elegans Sequencing Consortium 1998) and 13 Mb of sequence from C. briggsae released by the Washington University Genome Sequencing Center (http://genome.wustl.edu/gsc/). Previous studies have shown that C. elegans and C. briggsae have conservation of gene order over stretches of chromosome that can be up to six genes long (Kuwabara and Shah 1994; Thacker et al. 1999).

To calculate the rate, we estimated the number of chromosomal rearrangements since the speciation of C. elegans and C. briggsae. Because both species have six chromosomes (Nigon and Dougherty 1949), we assumed that there have not been any fusions or fissions of whole chromosomes since they diverged. Kececioglu and Ravi (1995) and Hannenhalli (1996) have developed computer algorithms that deduce the historical order and sizes of the reciprocal translocations (whereby two nonhomologous chromosomes exchange chunks of DNA by recombination) and/or inversions that have occurred since the divergence of two multichromosomal genomes. However, the C. elegans genome evolves not only by reciprocal translocations and inversions, but also by transpositions (whereby a chunk of DNA excises from one chromosome and inserts into a nonhomologous chromosome) and duplications (Robertson 2001). We designed a simple algorithm to calculate the number and sizes of such mutations, although not the order in which they occurred. Our method starts by finding all perfectly conserved segments between two species, in which gene content, order, and orientation are conserved. Next, these segments are fused into larger segments that have been splintered by duplications, inversions, or transpositions. When no more segments can be merged, the final fused segments are assumed to have resulted from fissure of chromosomes by reciprocal translocations.

To convert the observed number of rearrangements into a rate, it is necessary to have an accurate estimate of the briggsae-elegans divergence date. Emmons et al. (1979) were the first to estimate this date, using restriction fragment data, venturing that it must be "tens of millions of years" ago. Butler et al. (1981) speculated that the date was 10-100 million years ago (Mya), judging from 5S rRNA sequences, anatomical differences, and protein electrophoretic mobilities. Subsequent estimates based on sequence data were 30-60 Mya (Prasad and Baillie 1989; one gene), 23-32 Mya (Heschl and Baillie 1990; one gene), 54-58 Mya (Lee et al. 1992; two genes), and 40 Mya (Kennedy et al. 1993; seven genes). Nematode fossils are extremely scarce (Poinar 1983). Therefore, to calibrate the molecular clock, these studies either assumed that all organisms have the same silent substitution rate (Prasad and Baillie 1989; Heschl and Baillie 1990) or nonsilent substitution rate (Lee et al. 1992), or that C. elegans has the same silent rate as Drosophila (Kennedy et al. 1993). These are dubious assumptions; for example, Mushegian et al. (1998) showed that about two-thirds of C. elegans genes have a higher rate of nonsilent substitution than their orthologs in Drosophila. To gain a more reliable interval estimate of the briggsae-elegans speciation date, we used phylogenetic analysis of all genes for which orthologous sequences were available from C. elegans, C. briggsae, Drosophila, and human. Only those genes that did not have a significantly different amino acid substitution rate in the four taxa were used to produce date estimates.

The briggsae-elegans sequence data set is the largest available for any pair of congeneric eukaryotes. Such a big sample has a high power for detecting genome-wide trends. For example, the breakpoints of reciprocal translocations and inversions are frequently near repetitive DNA. This has been observed in bacteria (Romero et al. 1999), yeast (Fischer et al. 2000), insects (Cáceres et al. 1999), mammals (Dehal et al. 2001), and plants (Zhang and Peterson 1999), but not yet in nematodes. Rearrangements near transposable elements may happen when the element is transposing (Zhang and Peterson 1999), but most rearrangements are hypothesized to occur by homologous recombination between nontransposing transposable elements, dispersed repeats, or gene family members. We find that translocation and transposition breakpoints are strongly associated with repeats in the C. elegans genome.

	RESULTS

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION METHODS WEB SITE REFERENCES REFERENCES

Detection of Conserved Segments and Their Length Distribution

Using the BLASTX algorithm (Altschul et al. 1997), we predicted the positions of 1784 genes in the 12.9-Mb sample of C. briggsae genomic DNA. The 1784 genes partition the DNA into 756 segments that have been perfectly conserved between the two species. In C. briggsae, the segments range from 1 to 19 genes, or 0.6-154 kb. These segments were merged to recreate 252 longer segments that have been fractured by duplications, inversions, or transpositions since speciation. The 252 segments, which we assume to have resulted from fissure of chromosomes by reciprocal translocations, range from 1 to 109 genes in C. briggsae, or 1.3-1040 kb (average, 53 kb). In C. elegans, the corresponding segments cover 13.7% of the genome, the smallest segment being one gene (0.4 kb), and the largest 167 genes (954 kb; Fig. 1A,B). The segments can be browsed at the web address http://wolfe.gen.tcd.ie/worm/results.html. An example of the representation of a conserved segment on the web site is shown in Figure 2.

View larger version (14K): [in this window] [in a new window]   Figure 1   Distribution of sizes of conserved segments, measured in units of kilobases (A) and genes (B) with respect to Caenorhabditis elegans. These conserved segments were assumed to have resulted from fissure of chromosomes by reciprocal translocations.

View larger version (16K): [in this window] [in a new window]   Figure 2   The Caenorhabditis elegans region surrounding the sex-1 locus (F44A6.2B), and the corresponding region in Caenorhabditis briggsae. The pale blue bar wrapped over two lines represents the region between coordinates 10.18-10.23 Mb of C. elegans chromosome X, and the navy bar represents 3.50-3.57 Mb of C. briggsae contig RWRA. There are five orthologous briggsae:elegans genes (red) in the conserved segment, which are identified by the same number (1-5) in the two species, and are named on the C. elegans map. Inversion (I) and transposition (T) breakpoints are marked with orange arrows, which are shown arbitrarily on the C. briggsae chromosome. A region including three genes (C07A4.1, F09B9.3, and F09B9.2) has been inverted in either C. elegans or C. briggsae since speciation. Furthermore, a region comprising six genes (gray) between C07A4.1 and F44A6.1 in C. elegans has transposed to another part of the C. briggsae genome, or has transposed into this part of the C. elegans genome.

If the nine C. briggsae supercontigs are concatenated, we have one large 13.3-Mb chunk (the 12.9 Mb sample including internal gaps). If we assume that the 251 translocation breakpoints (and supercontig ends) are distributed at random along this chunk, the probability of recovering a segment L Mb by chance is e^251L/13.3 (Ranz et al. 2001). Of the sample of 252 segments detected, after we use the Bonferroni correction for multiple testing, only one is large enough to give a significant result (P = 8 × 10⁷). This is a 1.04-Mb segment containing 109 C. briggsae genes conserved between C. briggsae supercontig FORK and C. elegans chromosome X. Gene Ontology classifications are only given in WormBase (http://www.wormbase.org/) for 19 of the C. elegans orthologs of these 109 C. briggsae genes, and there is no obvious relationship between their functions that might provide a selective explanation for why this large segment has been conserved.

Differences among and along C. elegans Chromosomes

The median length of a conserved segment is significantly greater on the C. elegans X chromosome (40.6 kb) than on autosomes (17.0 kb; Mann-Whitney test: P < 0.01). It is not known which (if any) of the nine C. briggsae supercontigs in the sample originated from its sex chromosome. However, in C. elegans, sex is determined by counting X chromosomes via X signal elements on the X chromosomes (Akerib and Meyer 1994). We found the ortholog of the strongest C. elegans X signal element, the sex-1 gene (Carmi et al. 1998), on C. briggsae supercontig RWRA (Fig. 2). We suggest that RWRA, the largest supercontig (5.0 Mb) in the C. briggsae sample, is part of its sex chromosome. RWRA consists of 95 conserved segments matching C. elegans autosomes and 23 segments matching the C. elegans X chromosome. If RWRA is the C. briggsae sex chromosome, the C. briggsae sex chromosome must have undergone many reciprocal translocations with autosomes since divergence from C. elegans. Conversely, the C. elegans X chromosome consists of conserved segments matching five different C. briggsae supercontigs, which are unlikely to be all derived from the C. briggsae X chromosome.

The 252 conserved segments are scattered over all six C. elegans chromosomes (Fig. 3A), with 211 being on autosomes and 41 on the X chromosome. Taking Barnes et al.'s (1995) division of C. elegans autosomes into arms and centers, we found 102 conserved segments on autosome centers, and 109 on autosome arms (Fig. 3A). The median length of a conserved segment was not significantly different among the centers (20.5 kb), the left arms (17.5 kb), and the right arms (15.1 kb) of autosomes (Kruskal-Wallis test: P = 0.5).

View larger version (19K): [in this window] [in a new window]   Figure 3   (A) Location of conserved segments in the Caenorhabditis elegans genome. Gray bars under the autosomes show the "central clusters" described by Barnes et al. (1995). The segments cover ~15% of chromosome I, 7% of II, 10% of III, 13% of IV, 15% of V, and 20% of X. (B) Matrix plot comparison between the C. elegans genome (vertical axis) and the nine Caenorhabditis briggsae contigs (horizontal axis). Conserved segments are indicated by lines drawn between the positions of the outermost genes in each species.

Estimating the briggsae-elegans Divergence Date

Using the divergence of the nematodes from the arthropods at 800-1000 million years ago (Mya; Blaxter 1998; Brooke 1999) to calibrate the molecular clock, we estimated the briggsae-elegans divergence date from 92 sets of orthologs. Each set comprised a C. briggsae gene, its C. elegans ortholog, one or more orthologs from Drosophila, and one or more human orthologs. When the nematode-arthropod divergence is taken to be 800 Mya, a 95% confidence interval for the median briggsae-elegans speciation date is 49-94 Mya (median, 70 Mya). If the nematode-arthropod divergence is taken to be 1000 Mya, the interval becomes 61-118 Mya (median, 88 Mya; Fig. 4). Our best estimate of the briggsae-elegans speciation date is therefore ~50-120 Mya.

View larger version (24K): [in this window] [in a new window]   Figure 4   Estimates of the briggsae-elegans speciation date from 92 sets of Caenorhabditis briggsae, Caenorhabditis elegans, Drosophila, and human orthologs, calculated by taking the nematode-arthropod divergence date to be 1000 Mya.

Duplications

From phylogenetic trees, we identified 27 C. briggsae genes that have arisen by 14 duplications from 13 ancestral orthologs at the time of speciation. In 10 of these duplicate pairs, one duplicate has transposed, whereas four of the duplicate pairs have remained adjacent. In two of the four adjacent pairs, one of the duplicates has inverted. Of the 10 duplicates that have transposed, two of the duplicates are on different C. briggsae supercontigs. These 10 transpositions and two inversions in C. briggsae are the only rearrangements for which we know the genome in which they occurred.

Rates of Reciprocal Translocation, Inversion, and Transposition

Translocations

Inversions

View larger version (27K): [in this window] [in a new window]   Figure 5   (A) Sizes of inversions in kilobases, with respect to Caenorhabditis elegans. (B) Sizes of inversions, measured in units of genes. (C) Sizes of transpositions in kilobases. (D) Sizes of transpositions, measured in units of genes.

Transpositions

10

25

View larger version (30K): [in this window] [in a new window]   Figure 6   Method of detecting inversions and transpositions. (A) To detect transpositions to or from unsequenced parts of the Caenorhabditis briggsae genome, we looked along C. briggsae contigs for adjacent genes b1 and b2 whose Caenorhabditis elegans orthologs e1 and e2 are on the same chromosome, where between e1 and e2 there are 1-50 C. elegans genes with unknown C. briggsae orthologs. We assumed that the genes between 1 and 2 have transposed in either C. briggsae or C. elegans. (T) Transposition breakpoints. (B) To detect transpositions to or from sequenced parts of the C. briggsae genome, we looked along C. briggsae contigs for three conserved segments in a row, where in C. elegans the first and third segments were close together on the same chromosome, and the middle segment was far away on the same C. elegans chromosome or on a different C. elegans chromosome. We assumed that the middle segment (genes 4-5) had transposed in either C. briggsae or C. elegans. (C) To detect inversions, we looked along C. briggsae contigs for three conserved segments in a row, where in C. elegans the first and third segments were close together on the same chromosome, and the middle segment was far away on the same C. elegans chromosome or on a different C. elegans chromosome, and either the first or third segment, or both, had inverted in either C. briggsae or C. elegans. Here the third segment (genes 6-7) has inverted.

Overall Rate of Rearrangement

Association of Breakpoints with Repetitive DNA

We obtained the distribution of 33 dispersed repeat sequences in the C. elegans genome from WormBase (http://www.sanger.ac.uk/Projects/C_elegans/WORMBASE/GFF_files.shtml; Stein et al. 2001). When we pool all 33 repeats, there is a significant association between dispersed repeats and both translocation and transposition breakpoints in C. elegans (Table 1). However, no association is seen for inversion breakpoints. For two individual dispersed repeats, the association with transposition breakpoints is significant (P < 0.05; Table 2): CeRep20 and CeRep37. However, the significance of the association is marginal for CeRep20 (P = 0.045), whereas the small sample size for CeRep37 makes the test result unreliable.

Translocation breakpoints tend to be next to four different repeats: CeRep13, CeRep15, CeRep19, and CeRep32. C. elegans has compound repeats, listed on the Sanger Institute web site (http://www.sanger.ac.uk/Projects/C_elegans/repeats/). The only one associated with translocation breakpoints is CeRep13-CeRep18-CeRep18-CeRep33-CeRep18-CeRep13 (P = 0.01; Table 2). However, this is simply owing to the association of CeRep13 with breakpoints, because breakpoints are often near CeRep13/CeRep18/CeRep33, but not CeRep13 + CeRep18 + CeRep33. The association of CeRep19 and CeRep32 with translocation breakpoints is marginally significant (P  0.05), but that of CeRep13 and CeRep15 is strong (P  0.005). CeRep13 is a 26-bp sequence that is repeated ~1350 times in the C. elegans genome, whereas CeRep15 is a 63-bp sequence of which there are about 910 copies. Both these repeats seem to be derived from transposable elements. CeRep13 is 96% identical over 24 bp to the 24-bp terminal inverted repeat (TIR) of Cele11, which is thought to be a nonautonomous relative of Tc2, a Tc1/mariner family transposon (Oosumi et al. 1996). CeRep15 is 89% identical over 63 bp to part of the 170-bp TIR of Cele7, also thought to be a nonautonomous DNA transposon (Oosumi et al. 1995). We searched the C. briggsae genomic DNA for CeRep13 and CeRep15 using FASTA (Pearson and Lipman 1988). Homologs of CeRep13 seem to be present in the C. briggsae genome, because it has hits of 91% identity over 22 bp.

It is possible that rearrangement breakpoints could be associated with repeated gene sequences. To investigate this, we used BLASTP (Altschul et al. 1997) with an E-value cutoff of 10¹⁰⁰ to define families of highly similar genes. There are 1252 families, containing 3901 genes. The proportion of translocation breakpoints that have a gene family member on one or both sides (41%) is significantly greater than the proportion of all C. elegans intergenic spacers having a family member on one or both sides (33%; one-sided ² test: P  = 0.0001). A strong association is also seen for transposition breakpoints (one-sided ² test: P = 0.0002), but none for inversion breakpoints.

	DISCUSSION

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION METHODS WEB SITE REFERENCES REFERENCES

The average size of a conserved segment is 53 kb in C. briggsae. This is much larger than the 8.6-kb average found by Kent and Zahler (2000), even though they analyzed a subset of the same C. briggsae sequences (8.1 Mb of 12.9 Mb). There are three reasons for the difference. First, Kent and Zahler did not realize that the order and spacing of clones along C. briggsae chromosomes are known. The average size of the clones in their sample was 36 kb, whereas the average size of the supercontigs in our sample is 1486 kb. They underestimated the average size of a conserved segment because many clones end before the segment ends. Second, because their method allowed up to 50 kb of contiguous nonsyntenous DNA within a conserved segment in C. elegans but only up to 1 kb in C. briggsae, it was biased toward finding shorter conserved regions in C. briggsae than C. elegans. Third, instead of their approach of defining conserved segments by an arbitrary gap size, we strove for a more biologically meaningful approach by searching for the fragments into which chromosomes have been splintered by translocations. We followed Sankoff's (1999) suggestion and regarded inversions and transpositions within translocated segments as noise. For example, Kent and Zahler split the chromosomal region containing the sex-1 locus into nine segments, partitioning the DNA at poorly conserved noncoding stretches or where there have been small inversions and transpositions. In contrast, we found one large conserved segment in the sex-1 region (Fig. 2).

Kent and Zahler (2000) found that 63 of their 100 longest conserved segments were near the middle of C. elegans autosomes and surmised that "chromosome arms appear to be more susceptible to rearrangement." However, Kent and Zahler's result may merely reflect an ascertainment bias in the data set. We found that although the centers account for two-fifths of the length of autosomes (33 of 82 Mb), half of our conserved segments are from autosome centers (102 of 211 segments). We found no significant difference between the lengths of conserved segments on C. elegans autosome arms and centers, which indicates that arms and centers undergo translocations with equal frequency. The rate of chromosomal rearrangement may not be correlated with the nucleotide substitution rate, which does seem to be faster on arms than centers (The C. elegans Sequencing Consortium 1998; Koch et al. 2000).

The difference between the median segment size on X (41 kb) and on autosomes (17 kb) seems far too large to be attributable to lower detection sensitivity in gene-poor regions. Rather, X appears to be better conserved than the autosomes, which must be caused by a lower rate of occurrence or fixation of translocations of X. There may be fewer X translocations than autosomal translocations because of the lower density of gene family members and of some dispersed repeats on C. elegans X chromosome (The C. elegans Sequencing Consortium 1998; Surzycki and Belknap 2000). Alternatively, the rate of fixation of translocations may be different for X chromosomes and autosomes. Ohno (1967) hypothesized that in mammals and other species such as C. elegans that have dosage compensation systems in which X genes in XX organisms are down-regulated, X-autosomal translocations will be more deleterious than autosome-autosome translocations. This is because X genes (for example, dosage-sensitive genes involved in sex determination or sexual dimorphism) that are normally repressed on X through dosage compensation would become derepressed when translocated to autosomes, and autosomal genes that are not normally repressed would be repressed when translocated to X. There is greater selection against deleterious recessive mutations on the sex chromosome than on autosomes (Montgomery et al. 1987), therefore if most translocations are deleterious recessive, for example, because they upset regulation of expression, we would expect X translocations to be fixed less than autosomal translocations. On the other hand, if most translocations are selectively neutral, X may have a lower fixation rate because of a lower susceptibility to hitchhiking effects compared with the centers of autosomes (Barnes et al. 1995). A further possibility is that there is selection against translocations of the (as yet imprecisely mapped) region(s) of the X chromosome from which dosage compensation is initiated, as is seen in mammals (Nesterova et al. 1998).

Translocation and transposition breakpoints are often near repetitive DNA in the C. elegans genome, such as gene family members and dispersed repeats. Ectopic recombination between repeats may cause reciprocal translocations. Further study is needed to find out why transposition breakpoints tend to be near dispersed repeats (Table 1) and gene family members. It is possible that we have sometimes mistaken two translocations that were between sites close to each other on the same pair of chromosomes as a transposition.

When counting rearrangements, we could detect only inversions or transpositions of genes within conserved segments. As a result, some transpositions may have been mistaken for translocations, for example, a three-gene segment that transposed to a position between conserved segments. Furthermore, we may not have detected all inversions and transpositions, for example, if an entire conserved segment was inverted. Another possible source of error is that we assumed that stretches of C. elegans genes whose C. briggsae orthologs have not been found were caused by transpositions to or from an as-yet-unsequenced region of the C. briggsae genome (Fig. 6A), but it could be that the C. briggsae orthologs have been deleted. Our count of rearrangements may also have been affected by problems that are not specific to our method. First, the average size of a C. briggsae supercontig in our sample was 1486 kb, so we may not have detected rearrangements >1.5 Mb. Second, rearrangements that occur twice cannot be detected (Sankoff 1999). Third, it can be impossible to distinguish between three overlapping inversions and a single transposition (Blanchette et al. 1996). Following Nadeau and Taylor (1984), we attributed the few such ambiguous cases to inversions. However, some such inversions may have been in fact transpositions, because we found that transpositions are more common than inversions in Caenorhabditis. Fourth, we could not tell a reciprocal translocation apart from a chromosome fusion followed by a fission unless both of the translocation breakpoints had been found. We assumed ambiguous cases to be reciprocal translocations, not chromosome fusions or fissions, because both species have six chromosomes (Nigon and Dougherty 1949). Thus, we will not have detected if a fusion was followed by a fission in one of the species, or if a chromosome fission occurred in both species since divergence.

We estimate that Caenorhabditis has a rearrangement rate of 0.4-1.0 breakages/Mb per Myr. This is 1400-17,000 times the mammalian rate calculated by Ranz et al. (2001). Moreover, it is 4-20 times faster than the rate in Drosophila, previously reported to be the fastest rate among eukaryotes (0.05-0.09 breakpoints/Mb per Myr; Ranz et al. 2001). Error in the estimated briggsae-elegans divergence date would make our rate estimate inaccurate, but it seems unlikely that we have overestimated the rate of rearrangement. For nematodes to have the same rearrangement rate as Drosophila, the briggsae-elegans divergence date would have to be 560-1020 Mya; however, the nematode order to which Caenorhabditis belongs arose only ~400 Mya (Vanfleteren et al. 1994). Caenorhabditis and Drosophila differ not only in the rate, but also in the type, of rearrangement seen. In Caenorhabditis, translocations and inversions are almost equally frequent. In contrast, in Drosophila translocations are rare compared to inversions (Ranz et al. 2001), but in mammals translocations are roughly four times more common than inversions (Ehrlich et al. 1997).

Ranz et al. (2001) analyzed in situ hybridization data from Drosophila melanogaster chromosome 3R and Drosophila repleta chromosome 2, and used a maximum likelihood method to estimate the number of inversions that have occurred since divergence of the two chromosomes. Their likelihood method was designed to give an unbiased estimate of the number of rearrangements, thus differences between our Caenorhabditis results and their Drosophila results are probably not caused by differences between the methods used. However, some differences between the results are probably due to differences in data quality. For example, it is likely that they have underestimated the rate of small rearrangements in Drosophila for two reasons. First, because the orientation of the Drosophila markers was not known in both species, they could not detect inversions of single markers (for comparison, ~40% of the Caenorhabditis inversions we detected were one gene long; Fig. 5B). Second, their physical map only had one marker per 49 kb in its densest regions, thus the smallest inversion that they could detect was ~100 kb long (for comparison, ~95% of the Caenorhabditis inversions detected were <100 kb long; Fig. 5A). By comparing the D. melanogaster genome to that of the mosquito Anopheles gambiae (soon to be released, Hoffman et al. 2002), it may be possible to estimate the rate of small rearrangements in insects. In contrast, Ranz et al. (2001) will have detected more long inversions than we did, because the average size of a C. briggsae supercontig in our sample was ~1.5 Mb, whereas their markers spanned 28 Mb.

We suggest four reasons why Caenorhabditis chromosomes may have a faster rearrangement rate than those of Drosophila. First, the generation time of Caenorhabditis is 4-5 times shorter (3-4 d compared with ~2 wk). Second, C. elegans and C. briggsae may have a smaller effective population size than Drosophila, because they are largely self-fertilizing but Drosophila is not. Third, C. elegans chromosomes may be more prone to hitchhiking effects than those of Drosophila, because in C. elegans the most gene-rich regions of autosomes have the lowest recombination rates, but the opposite is true for Drosophila (Barnes et al. 1995). In other words, if a selectively neutral rearrangement occurs near a positively selected gene, in C. elegans it is less likely to be separated from the selected gene by meiotic recombination, and so is more likely to undergo a selective sweep with that gene. These three reasons may also lie behind the faster substitution rate in C. elegans compared with Drosophila. However, the amino-acid substitution rate is usually less than two times faster in a C. elegans gene than in its Drosophila ortholog (see Fig. 3 in Mushegian et al. 1998), whereas the rearrangement rate is at least four times faster in C. elegans. Our fourth reason is the only one that may contribute to a higher rearrangement rate in C. elegans but not to a higher substitution rate. It is that in selfing species like C. elegans and C. briggsae, rearrangements that are deleterious when heterozygous are more likely to persist than in an out-crossing species, because homozygous individuals arise sooner (Lande 1979). If this is true, we would expect non-selfing species of Caenorhabditis, such as Caenorhabditis remanei (Baird et al. 1992), to have a lower rate of rearrangement compared with Drosophila than do C. elegans and C. briggsae. We would also expect greater karyotype variability in C. elegans populations than in Drosophila or C. remanei populations. Genomic sequence from non-selfing Caenorhabditis species and data on the karyotype variability in wild C. elegans populations could provide clues as to why there is a rate difference.

	METHODS

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION METHODS WEB SITE REFERENCES REFERENCES

Sources of Sequence Data

Nine supercontig DNA sequences from the C. briggsae sequencing project at the Washington University Genome Sequencing Center (http://genome.wustl.edu/gsc/) generated from a fingerprint map of C. briggsae (M. Marra, J. Schein, and R. Waterston, unpubl.) were downloaded from the WormBase site (http://www.wormbase.org/; Stein et al. 2001) in July 2001. The C. briggsae data consist mainly of genes requested by the Worm Community to be sequenced (Baillie and Rose 2000) and are therefore not a random sample of the genome. The nine supercontigs range from 70 to 5015 kb. Because some of these supercontigs contained large internal gaps, we subdivided supercontigs at any internal gap of >2 kb. The resulting 20 contigs range from 51 to 2288 kb (median, 369 kb) and totaled 12.9 Mb. The 19,957 C. elegans protein sequences from Wormpep54 were downloaded from http://www.sanger.ac.uk/Projects/C_elegans/wormpep/ in July 2001. We discarded 586 Wormpep proteins from the genes of transposable elements and genes similar to transposable-element genes, 31 from genes whose chromosomal coordinates in C. elegans are unknown, and 713 from alternatively spliced genes (retaining the longest splice variant only); 18,627 proteins remained. C. elegans gene coordinates corresponding to ACeDB release WS44 were downloaded in July 2001 from http://www.sanger.ac.uk/Projects/C_elegans/WORMBASE/GFF_files.shtml.

Predicting C. briggsae Genes

The C. briggsae contigs were largely unannotated, so we predicted C. briggsae genes using a spliced alignment approach similar to that of Mironov et al. (1998). This was feasible because protein-coding regions are conserved between the two species, but intergenic regions and introns are not (Kent and Zahler 2000). Regions of the C. briggsae contigs homologous to C. elegans proteins were identified using BLASTX (Altschul et al. 1997) with the BLOSUM62 scoring matrix (Henikoff and Henikoff 1992), using the SEG filter (Wootton and Federhen 1996), and storing all hits with an E-value of 0.1. There were 99,221 BLASTX hits. Because BLASTX does not always accurately distinguish between orthologs and paralogs, we kept any overlapping hits having E-values within a factor of 90 of each other. Nearby BLASTX hits to the same C. elegans protein were assumed to correspond to the exons of a C. briggsae homolog, and were merged so long as they were on the same strand of the C. briggsae contig. To avoid merging hits that were implausibly far apart on a C. briggsae contig, any C. briggsae intron could not be >7700 bp, and the summed length of introns in a C. briggsae gene could not exceed 8150 bp. These numbers (90, 7700, and 8150) were chosen on inspection of the results. To prevent mistaken merging of tandemly repeated genes on a C. briggsae contig, the following rule was used, where "left" and "right" refer to the position on the C. briggsae contig. The left BLASTX hit had to start in the C. elegans protein before the right hit ended in the C. elegans protein, and the left hit had to end in the C. elegans protein before the right hit started in the C. elegans protein, or the hits overlap by <1100 amino acids. After merging BLASTX hits to predict genes, we found a nonoverlapping set of the most significant C. briggsae genes along each supercontig. Lastly, C. briggsae genes that hit <45% of the length of the C. elegans protein, or had BLASTX E-values of 10⁵, were deleted, as they were probably pseudogenes. On the nine supercontigs, we predicted 1934 C. briggsae genes. We will not have detected C. briggsae genes that do not have homologs in C.elegans, if any such C. briggsae genes exist. In a search of the literature we could not find any examples of C. briggsae genes that do not have a C. elegans homolog.

Finding Orthologs

We did not use synteny data to define orthologs, only sequence identity and phylogenetic trees, because we wanted to use orthologs to gauge synteny conservation. The 1934 C. briggsae genes hit 1804 different C. elegans proteins in BLASTX. If a C. briggsae gene hit only one C. elegans protein in BLASTX, and no other C. briggsae genes hit that C. elegans protein, then the C. briggsae and C. elegans genes were taken to be one-to-one orthologs. Based on BLASTX results alone, 1704 one-to-one ortholog pairs were found. Some of these orthologous pairs were detected from BLASTX hits having E-values as high as 10⁶. For the remaining 230 C. briggsae genes, it was necessary to draw 151 different phylogenetic trees to deduce orthology. To find an outgroup for a tree of a C. briggsae gene and its C. elegans hits we used BLASTP (Altschul et al. 1997) with an E-value cutoff of 0.1 to compare the C. elegans hits with Wormpep54 (19,957 proteins) and to SWISS-PROT (July 2001). For each C. elegans protein in the tree, the outgroup was either the top-scoring C. elegans hit for which a C. briggsae ortholog had previously been identified from BLASTX results, or the top-scoring non-elegans hit, whichever had the highest score. The sequences for a tree were aligned using CLUSTALW (Thompson et al. 1994), and a maximum parsimony phylogenetic tree was drawn using protpars (Felsenstein 1993). We bootstrapped the trees using 1000 bootstrap replications in the seqboot algorithm (Felsenstein 1993). Only nodes with bootstraps of 80% were used to deduce orthology.

Our final C. briggsae data set contains 1934 genes: 1744 genes in one-briggsae-to-one-elegans orthology relationships, and 190 other genes in the following relationships:

(1)	13 genes in one-briggsae-to-many-elegans relationships;
(2)	46 genes in many-briggsae-to-one-elegans relationships: 23 for which the C. elegans ortholog is known and 23 for which the C. elegans ortholog is unresolved;
(3)	4 genes in many-briggsae-to-many-elegans orthology relationships;
(4)	13 genes whose C. elegans ortholog has been deleted since speciation or had not been sequenced yet; and
(5)	114 remaining C. briggsae genes whose orthology is unresolved. For 137 of the genes, orthology could not be decided owing to lack of a suitable outgroup or low bootstraps in trees.

The 137 C. briggsae genes of unresolved orthology (mainly histone genes) and the 13 with deleted orthologs were ignored in the subsequent analysis, leaving 1784 C. briggsae genes that hit 1792 different C. elegans genes, of which 1744 were one-to-one orthologs.

Estimating the briggsae-elegans Divergence Date

We downloaded 161,296 human proteins and 35,108 Drosophila proteins from GenBank (http://www.ncbi.nlm.nih.gov/Entrez/; December 2001). To find C. elegans orthologs of these proteins, we compared them with Wormpep using BLASTP (Altschul et al. 1997) with the SEG filter (Wootton and Federhen 1996). If a human protein hit a C. elegans protein with a BLASTP E-value of <10²⁰, and the C. elegans protein with the second strongest hit had an E-value that differed by a factor of 10²⁰ or more, then the C. elegans protein was considered to be the ortholog of the human protein. We found 238 sets of orthologs, each set containing a C. briggsae gene, its C. elegans ortholog, one or more human orthologs, and one or more Drosophila orthologs. For each set, we aligned the proteins using CLUSTALW (Thompson et al. 1994), and made a guide-tree using protdist and neighbor from the PHYLIP package (Felsenstein 1993). We discarded 33 ortholog sets for which the human sequences did not group together and/or the Drosophila sequences did not group together, leaving 205 sets. For each ortholog set, the alignment and guide-tree were used as input for Gu and Zhang's (1997) program GAMMA, which estimated an  parameter for the  distribution used to correct for rate variation among amino acid sites. For 31 trees, GAMMA could not estimate the  parameter. For the remaining 174 trees, we used the two-cluster test (Takezaki et al. 1995) to check for unequal rates between lineages, taking human to be the outgroup to Drosophila and Caenorhabditis (Aguinaldo et al. 1997); 92 trees passed the test at the 5% significance level. For each tree, the branch lengths were reestimated under the assumption of rate constancy, using Takezaki and Nei's (Takezaki et al. 1995) program with the  correction for multiple hits. Although the exact branching order of the chordates, arthropods, and nematodes continues to be hotly debated (Mushegian et al. 1998; Wang et al. 1999), most estimates of the divergence of these three phyla range from 800 to 1000 Mya (Blaxter 1998; Brooke 1999). We calibrated the linearized trees by taking the nematode-arthropod divergence to be 800-1000 Mya.

Finding Conserved Segments and Classifying Breakpoints According to Mutation Type

When two species are compared, any region of their genomes in which gene content and order are conserved is a "conserved segment" (Sankoff 1999). Between two adjacent conserved segments is a "breakpoint" (Sankoff 1999) caused by translocation, inversion, duplication, or transposition. We searched for all perfectly conserved segments on the C. briggsae supercontigs: segments in which gene order and orientation are perfectly conserved with C. elegans. To estimate the size distribution of different types of mutation, the breakpoints within C. briggsae contigs were classified as duplication, translocation, inversion, or transposition breakpoints as described below.

From phylogenetic trees, we identified C. briggsae genes that have arisen by duplication since speciation. If two C. briggsae duplicates that arose from one ortholog were adjacent, we called the breakpoint between them a duplication breakpoint; if one of the duplicates is inverted, it is also an inversion breakpoint. These breakpoints were subsequently ignored, thereby enlarging the original conserved segments. A conserved segment was then taken to be the region between two as-yet-unexplained breakpoints. Transpositions and inversions were detected as shown in Figure 6. The final conserved segments left after all inversions and transpositions had been found were assumed to be segments whose breakpoints were due to translocations. The final conserved segments were manually edited where, for example, two segments were close in the C. elegans genome and probably were the same conserved segment.

Because the lengths of those transpositions involving C. elegans genes whose C. briggsae orthologs have not yet been sequenced can be measured only in units of C. elegans genes (Fig. 6A), the sizes of all inversions and transpositions have been given in terms of the number of C. elegans genes. If a transposition had occurred within an inverted segment, the size of the inversion was taken to include the transposed genes; likewise, if an inversion had occurred within a transposed segment, the size of the transposition was taken to include the inverted genes.

Testing Whether Breakpoints Are Associated with Repeats

The positions of 33 dispersed repeat families in the C. elegans genome were downloaded from http://www.sanger.ac.uk/Projects/C_elegans/WORMBASE/GFF_files.shtml. The arrangement of these dispersed repeats into compound repeats was taken from http://www.sanger.ac.uk/Projects/C_elegans/repeats/. To group C. elegans proteins into families, we compared Wormpep to itself using BLASTP (Altschul et al. 1997) with the SEG filter (Wootton and Federhen 1996). Proteins A, B, and C were assumed to belong to the same family if A hit B with an E-value of <10¹⁰⁰ and B hit C with an E-value of <10¹⁰⁰.

	WEB SITE REFERENCES

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION METHODS WEB SITE REFERENCES REFERENCES

http://evolution.genetics.washington.edu/phylip.html; PHYLIP phylogeny inference package.

http://www.genome.wustl.edu/gsc/; Genome Sequencing Center, Washington University, St. Louis.

http://www.ncbi.nlm.nih.gov/Entrez/; GenBank.

http://www.sanger.ac.uk/Projects/C_elegans/repeats/; List of C. elegans repeat families at the Sanger Institute.

http://www.sanger.ac.uk/Projects/C_elegans/WORMBASE/GFF_files.
shtml; WormBase release files at The Sanger Institute.

http://www.sanger.ac.uk/Projects/C_elegans/wormpep/; C. elegans protein database Wormpep.

http://www.wormbase.org/; WormBase.

http://wolfe.gen.tcd.ie/worm/results.html; Views of worm segments as in Figure 2.