MBE Advance Access originally published online on July 23, 2007
Molecular Biology and Evolution 2007 24(10):2169-2179; doi:10.1093/molbev/msm148
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Research Articles |
A New Method for Assessing the Effect of Replication on DNA Base Composition Asymmetry
uleaUniversité de Lyon; université Lyon 1; CNRS; UMR 5558, Laboratoire de Biométrie et Biologie Evolutive, 43 boulevard du 11 novembre 1918, Villeurbanne F-69622, France; and HELIX, Unité de recherche INRIA
E-mail: necsulea{at}biomserv.univ-lyon1.fr.
 |
Abstract |
|---|
 TOP Abstract IntroductionMaterial and MethodsResults and DiscussionConclusion and PerspectivesSupplementary MaterialAcknowledgementsReferences |
|---|
DNA base composition asymmetry is at the basis of numerous in silico methods for the detection of the origin and terminus of replication in prokaryotes. However, most of these methods are unable to identify the evolutionary mechanisms that cause the base composition asymmetry. In prokaryotic chromosomes, due to the tendency for coorientation between replication and transcription, compositional biases that discriminate the leading strand from the lagging strand can be produced by 2 superposing mechanisms: replication-associated mutation bias and coding sequence–associated bias (such as transcription-related mutational processes or selective pressures on codon usage). We propose here a new method for the analysis of nucleotide composition asymmetry that allows the decoupling of replication-related and coding sequence–related mechanisms. This method is inspired by a recent work (Nikolaou and Almirantis 2005) that proposed an artificial chromosomal rearrangement meant to create a perfect gene orientation bias. We show that the study of nucleotide skews on the artificially rearranged chromosomes is a powerful means to assess the contributions of the 2 types of mechanisms in generating the base composition asymmetry. We applied our method to all completely sequenced prokaryotic chromosomes available. Our results confirm that in most species the replication mechanism has an important effect on base composition asymmetry but also that it has different impacts on GC and AT skews. We also analyzed the variability in AT skew direction encountered in prokaryotes. In disagreement with a recent report (Worning et al. 2006), we find that the polymerase-
subunits encoded in a genome are not sufficient to predict the sign of the AT skew on its leading strand for replication.
Key Words: DNA base composition asymmetry • origin of replication • replication-induced mutation bias • transcription orientation • bacterial genomes
|
Introduction |
|---|
|
TOPAbstractIntroductionMaterial and MethodsResults and DiscussionConclusion and PerspectivesSupplementary MaterialAcknowledgementsReferences |
|---|
With the increasing availability of complete genomic sequences, it has become clear that the DNA replication mechanism is an essential factor in the organization of prokaryotic genomes. As a general rule, prokaryotic chromosomes are characterized by one bidirectional origin of replication and one terminus that define the boundaries of the replichores. The replication process is asymmetric with respect to the 2 DNA strands; the leading strand is synthesized continuously, whereas the lagging strand is replicated in a fragmented manner. Within a replichore, the 2 DNA strands can be highly asymmetric with respect to gene content and nucleotide composition.
The influence of the replication mechanism on the structure of prokaryotic chromosomes was documented before the genomic era, when an analysis of the Escherichia coli chromosome showed that genes coding for ribosomal proteins and rRNAs tend to be localized on the leading strand, so that transcription and replication occur in the same direction (Nomura et al. 1977
); it was later confirmed that the coorientation between transcription and replication can be extended to other classes of protein-coding genes in E. coli (Brewer 1988). An explanation was provided by the "polymerases collision" hypothesis, stating the existence of a selective pressure that minimizes the risks of head-on collisions between the DNA replication complex and the RNA polymerase, likely to be deleterious for the organism (Nomura et al. 1977; Brewer 1988). The analysis of the first complete genomic sequences proved that the extent of the bias in protein-coding gene orientation varies widely among species: in Synechocystis sp. and in E. coli, the proportion of genes encoded on the leading strand reaches only 50% and 54%, respectively, whereas in Bacillus subtilis and Mycoplasma genitalium it reaches 74% (Blattner et al. 1997; McLean et al. 1998). A characteristic that seems to be shared by most bacterial species is that the orientation bias is greater for highly expressed genes (McLean et al. 1998), which seems to be in agreement with the polymerases collision hypothesis because the risk of collision is likely to be dependent of the transcription rate. However, recent studies have suggested that the main determinant of the gene strand bias may be the "essentiality" of genes rather than the level of expression (Rocha and Danchin 2003a, 2003b).
Another genomic structure associated with the replication mechanism is the DNA base composition asymmetry. E. Chargaff experimentally determined the approximate equimolarities [A] 
[T] and [G] [C] for long, single-stranded DNA molecules (Rudner et al. 1968). It was later proved theoretically that these equalities (referred to as the second parity rule, or PR2) should be observed at the equilibrium state, when mutational and selective pressures are symmetric with respect to the 2 DNA strands (Sueoka 1995; Lobry 1995). In prokaryotic genomes, local deviations from PR2 (also called GC skews and AT skews) are widely encountered, and they generally divide the chromosome into 2 regions with opposite signs for the base composition skews (Lobry 1996a). The switch in skew direction generally occurs at the origin and terminus of replication (Lobry 1996a, 1996b; Francino and Ochman 1997; Grigoriev 1998; Rocha et al. 1999); this compositional property is the basis of many in silico methods for the prediction of the origin of replication (Grigoriev 1998; Salzberg et al. 1998; Frank and Lobry 2000; Zhang R and Zhang CT 2002; Worning et al. 2006).
The causes of the base composition asymmetries observed in prokaryotes have been widely discussed in the recent litterature—see Frank and Lobry (1999) and Rocha (2004) for a review. One plausible explanation is that the nucleotide skews are caused by a mutation bias associated with the replication mechanism itself, given the strand asymmetry of this process; the replication mechanism is thus considered a direct cause of base composition asymmetry. Another hypothesis states that the deviations from PR2 are related to the strand asymmetry of the transcription mechanism, in combination with the biased gene orientation encountered in bacterial chromosomes (for most genes, the coding strand is also the leading strand for replication). Similarly to the latter explanation, it has been proposed that selective forces acting on protein-coding sequences (e.g., through biased codon usage) can cause the nucleotide skews, again when associated with gene orientation bias. In the latter hypotheses, the replication mechanism is considered to be only an indirect factor in base composition asymmetry, acting through the bias in gene orientation.
There is extensive evidence supporting the replication mechanism as a direct cause of base composition asymmetries. The analysis of the codon usage patterns, through statistical methods such as correspondence analysis and correspondence discriminant analysis, showed that in some bacterial species genes encoded on the leading and lagging strands can be distinguished by their synonymous codon choice (Perrière et al. 1996; McInerney 1998). Using the analysis of variance (ANOVA) method on GC and AT skews, with gene direction and replication orientation as explanatory variables, Tillier and Collins (2000) proved that the base composition of a gene is significantly influenced by its position on the leading or the lagging strand for replication. A study performed on 43 prokaryotic chromosomes confirmed that deviations from PR2 differ significantly between leading and lagging strands, as expected under the hypothesis of replication-induced mutation bias (Lobry and Sueoka 2002).
However, there is also evidence suggesting that an indirect effect of replication, through gene orientation bias and transcribed/nontranscribed strand asymmetries, would be sufficient to explain the base composition asymmetries. The analysis of the substitution pattern in E. coli coding sequences suggested that the asymmetry is probably generated by a process linked to transcription, and not to the mode of replication, because the substitution patterns were similar on the leading and lagging strands (Francino et al. 1996).
A recent paper (Nikolaou and Almirantis 2005) has brought an interesting contribution to the area, in favor of the latter type of mechanism. At the basis of their findings is the observation that in bacterial species the base composition skew is generally encountered in combination with a strong coding sequence (CDS) orientation bias. In order to prove that a strong gene orientation bias is a sufficient cause for the emergence of nucleotide skews, the authors propose to artificially rearrange the bacterial chromosome in order to produce a perfect gene orientation bias, and they show that in the case of Nostoc sp., this rearrangement is followed by a strong trend in base composition asymmetry. Thus, according to the results of Nikolaou and Almirantis (2005), only an indirect role of replication on base composition asymmetry seems to be supported, whereas the gene orientation bias appears to be the main factor responsible for the existence of the nucleotide skews.
We present here a method for distinguishing between the various sources of base composition asymmetry, based on the artificial genome rearrangement proposed by Nikolaou and Almirantis (2005). We show that the graphical representation of the nucleotide skews on the rearranged chromosomes is a powerful tool for the visual discrimination between the 2 types of mechanisms. We use this method to analyze all the currently available prokaryotic genomic sequences and we examine the distribution of the various patterns exhibited by the rearranged nucleotide skews.
We show that in most of the species the nucleotide skews induced by the artificial genome rearrangements exhibit significant breakpoints at the origin and terminus of replication. Our results are in favor of the hypothesis of a direct effect of the replication mechanism on nucleotide skews, independent of coding sequence orientation. We agree nevertheless that asymmetries between the coding and noncoding strands, caused by mechanisms such as transcription-induced mutation bias or codon choice, are responsible for a large part of the nucleotide skews encountered in prokaryotes.
Our results confirm that the impact of the replication mechanism on nucleotide asymmetries is different for GC and AT skews. Although the effect of replication on the GC skew is generally very strong, we show that in numerous species the AT skew is exclusively caused by coding sequence–related mechanisms.
Finally, we analyzed the variability in AT skew sign among species. A recent work has proposed that the direction of AT skew in prokaryotes is determined by the polymerase- subunit that replicates the leading strand (Worning et al. 2006). Our results show that the direction of the AT skew caused by replication alone is highly variable even within taxonomic groups, such as Firmicutes, that use the same subunit for the replication of the leading strand.
|
Material and Methods |
|---|
|
TOPAbstractIntroductionMaterial and MethodsResults and DiscussionConclusion and PerspectivesSupplementary MaterialAcknowledgementsReferences |
|---|
Sequence Data
The data set consists of 389 complete prokaryotic chromosomes, extracted from the National Center for Biotechnology Information complete genomes database on July 26th, 2006. The annotated sequence files were retrieved from ftp://ftp.ncbi.nih.gov/genomes/Bacteria/and analyzed using functions from the seqinR package (Charif and Lobry 2006
) of the R statistical computing environment (R Development Core Team 2005). The annotations were parsed to extract coding sequence boundaries; protein-coding genes that were annotated as partial or with introns were ignored.
Artificial Genome Rearrangements
The chromosomes were artificially rearranged to obtain a perfect gene orientation bias (also called here CDS skew), as described in Nikolaou and Almirantis (2005). All the genes encoded on the forward strand were moved to the first half of the chromosome, and all the reverse strand encoded genes were relocated on the second half. This rearrangement conserved the relative order of genes within each of the 2 groups—both forward-encoded and reverse-encoded genes are placed on the rearranged chromosome in increasing order of their coordinates on the real chromosome. We computed the base composition skews only on protein-coding sequences; therefore, intergenic sequences were ignored when rearranging the genome, and the artificially generated nucleotides skews were plotted against the gene index, instead of the actual coordinates on the chromosome.
Cumulative Skews
The base composition asymmetry of the rearranged bacterial chromosomes was described through cumulative GC and AT skew diagrams. The cumulative skews for a chromosome were obtained by summing the relative skews for each protein-coding gene, computed on third codon positions: GC skew = (G3 – C3)/(G3 + C3) and AT skew = (A3 – T3)/(A3 + T3).
The skew computations were performed exclusively on the published DNA strand, not on the coding strand of the genes.
The gene orientation bias is measured through the CDS skew. Each gene is assigned a value, either +1 or –1, according to the transcription orientation (+1 for forward-encoded genes, –1 for reverse-encoded genes); the cumulative sum of these values gives the CDS skew.
Breakpoint Detection and Significance
Under the assumption that base composition asymmetries are exclusively caused by a local bias in coding sequences (such as a transcription-related mutation bias, or a selective mechanism such as biased codon usage), in combination with gene orientation bias, the artificial nucleotide skews generated through genome rearrangements should follow the perfect CDS skew (Nikolaou and Almirantis 2005). Thus, within each group of genes (forward encoded and reverse encoded), the cumulated artificial nucleotide skews should vary linearly along the rearranged chromosome, as does the cumulated artificial CDS skew (cf. fig. 1a).
|
If we assume, on the contrary, that the replication process has a nonnegligible effect on base composition asymmetry, the difference between the leading and the lagging strand should be visible as a change of slope in the linear regression of the artificial nucleotide skews against the chromosome position, for both forward-encoded and reverse-encoded genes (cf. fig. 1b). We therefore wanted to verify if the relationship between the cumulated nucleotide skews and the position on the chromosome can be best described by a standard linear model or by a piecewise linear model, with breakpoints corresponding to the origin and terminus of replication.
We analyzed the relationship between the cumulated nucleotide skews and the position on the chromosome, separately for the forward-encoded and reverse-encoded groups of genes. We chose the gene index as a measure of the position on the chromosome, instead of the actual coordinate.
We used the segmented.lm function in the segmented package (Muggeo 2004) in R to search for breakpoints in the linear regression of the artificial cumulated GC and AT skews against the gene index, for each of the 2 groups. This function implements an iterative algorithm to estimate the position of the breakpoints, by trying to maximize the likelihood function and to minimize the residual sum of squares of the regression, and requires starting values for the breakpoints (Muggeo 2003).
As we could not tell beforehand how many breakpoints might the linear regression display, we used segmented.lm to determine 5 potential breakpoints, and we then evaluated their significance as described in the next paragraph. In order to avoid the situation where the estimated optimal set of breakpoints corresponds only to a local extremum of the likelihood and residual sum of squares functions, we performed a grid search by applying the segmented.lm function to a series of 150 sets of initial breakpoint values, drawn from a uniform distribution, and we then selected among the results the one that minimized the residual sum of squares. We then repeated the same procedure, changing the number of breakpoints to 4, 3, 2, and 1.
We then performed an additional procedure meant to refine the position of the breakpoints, by repeating the grid search on a restricted region surrounding each of the previously found breakpoints, of maximum length equal to 200 genes (100 genes in the 5' region and 100 genes in the 3' region of the breakpoint).
In order to test the significance of the breakpoints, we used as a measure for the "strength" of each breakpoint the difference in slopes between the 2 segments it delimits. We then obtained the expected distribution of the slope differences for each chromosome, within each group of genes (forward encoded and reverse encoded), and for each number of expected breakpoints (from 1 to 5), by randomly permuting the gene order, detecting the breakpoints and then computing the slope differences on the permuted data sets; we performed 150 permutations for each group of genes. For each "real" breakpoint, we could thus compute a p-value, indicating whether this breakpoint could have been obtained simply by chance; the p-value of a breakpoint is equal to the proportion of times that a slope difference at least as large (in absolute value) was observed in the permuted data sets. Our p-value allows us to test the following null hypothesis: the breakpoints in rearranged nucleotide skews derive only from the intrinsic variation in GC and AT skews among genes and not from their order on the chromosome.
We chose a threshold of 5% and a top-down approach to determine the number of significant breakpoints as follows: we first computed the p-values for the data sets with 5 breakpoints. If at least one of the 5 breakpoints was considered nonsignificant (p-value higher than 5%), we moved on to the data set with 4 breakpoints, and so on, until all the remaining breakpoints were significant or until no breakpoints remained.
We set the maximum number of breakpoints at 5 because in most of the cases only 1 or 2 breakpoints should be observed, corresponding to the origin or terminus of replication. However, in some particular species, such as Pasteurella multocida or Yersinia pestis, several segmental inversions with respect to the origin of replication can be observed; these inversions create a number of additional breakpoints. For these species, we used the same procedure but setting the maximum number of breakpoints at 7. We used the same procedure for archaeal species of the genera Sulfolobus and Halobacterium, where multiple origins of replication have either been determined experimentally (Lundgren et al. 2004; Robinson et al. 2004) or inferred through computational methods (Zhang R and Zhang CT 2003).
Due to the length of the significance computations, we computed the number of significant breakpoints only for one species (or strain) per genus, which allowed us to restrict the data set to 173 chromosomes. For the other species, we assumed that the number of significant breakpoints is constant within a genus, unless the visual inspection of the nucleotide skews graphs indicated otherwise, and we then determined their location according to the first part of our procedure.
Origins of Replication Dataset
To verify if the breakpoints detected in the nucleotide skews correspond to the origin and the terminus of replication, we needed to know the positions of these 2 points on all completely sequenced chromosomes.
We used Oriloc (Frank and Lobry 2000) to predict the origin and terminus of replication in prokaryotic chromosomes, through the following procedure: we computed the cumulated composite skew given by this program (a linear combination of the GC and AT skews) in third codon position, and we determined its minimum and maximum. We used the gene orientation bias to decide which extremum corresponds to the origin of replication, by assuming that most genes are encoded on the leading strand for replication—in this way, we make no a priori assumption on the direction of the nucleotide skews on the leading strand.
As a complement, we used the predictions for the origin and terminus of replication provided by Worning et al. (2006) for circular chromosomes (http://www.cbs.dtu.dk/services/GenomeAtlas/suppl/origin/). We also used the procedure described by Mackiewicz et al. (2004), which predicts the position of the origin of replication based on the DnaA box distribution and on the position of the dnaA genes. The results are available at http://pbil.univ-lyon1.fr/datasets/Necsulea2007/origins.
Simulated Dataset
In order to verify the specificity of the breakpoint detection method, we built a data set that simulates a chromosome with 1000 genes, for which the rearranged skews display 2 breakpoints, corresponding to the origin and terminus. The skews were drawn from a Gaussian distribution, with the parameters (m, sd) for the leading strand and (m + r, sd) for the lagging strand. The first and third segments delimited by the breakpoints (that correspond here to the leading strand) cover a fraction fr and 0.5 – fr of the chromosome, respectively, whereas the second segment (that corresponds here to the lagging strand) covers half of the chromosome. We allowed the parameters m and r to vary between 0.001 and 10, sd between 0.001 and 1, and fr between 0.01 and 0.3. We repeated the simulation 100 times for each set of parameters, and we permuted the simulated data set 100 times to compute the significance of the breakpoints.
Reproducibility and Supplementary Material
The data set used in this paper is available at the following URL: http://pbil.univ-lyon1.fr/datasets/Necsulea2007/dataset. All the programs used in this work are available upon request; the functions that allow the computation and the graphical representation of the rearranged nucleotide skews are available in the 1.0.7 version of the seqinR package in R.
The results of our analyses for all 389 chromosomes are available at the following URL: http://pbil.univ-lyon1.fr/datasets/Necsulea2007/html/index.html.
|
Results and Discussion |
|---|
|
TOPAbstractIntroductionMaterial and MethodsResults and DiscussionConclusion and PerspectivesSupplementary MaterialAcknowledgementsReferences |
|---|
Analysis of the Nucleotide Skews in Artificially Rearranged Chromosomes
Figure 1 illustrates how the analysis of the nucleotide skews in rearranged chromosomes can be a useful method to distinguish between the 2 main sources of base composition asymmetry: coding sequence–related bias (transcription-induced mutation biases or codon usage bias) and replication-related mutation bias.
In the case of Gloeobacter violaceus, the rearranged GC skew is perfectly correlated with the CDS skew; no significant breakpoints are detected. This means that the transcription orientation of genes is a sufficient predictor of their nucleotide skew; the replication mechanism does not have a significant effect on base composition asymmetry in this species.
In the case of E. coli K12, 2 significant breakpoints are detected for each group of genes (forward encoded and reverse encoded). For forward-encoded genes, the coordinates of the breakpoints (with respect to the original chromosome) are approximately 1.605 and 3.962 Mb, whereas for reverse-encoded genes the coordinates are approximately 1.603 and 3.858 Mb. The positions of the origin and terminus of replication in E. coli have been determined experimentally; the origin was identified near the ilv locus, at 74 min on the genetic map of the chromosome, and the terminus near the trp locus, on a point diametrically opposed to the origin (Bird et al. 1972). In terms of position on the chromosome sequence, these locations correspond to approximately 3.9 and 1.6 Mb; these positions correspond as expected with the switches in nucleotide skew direction (Blattner et al. 1997). Therefore, the breakpoints observed in the rearranged skew correspond to the origin and terminus of replication in this chromosome.
The existence of breakpoints in the rearranged skews and their localization at the origin and terminus of replication is a clear indication that the replication mechanism has a strong direct effect on nucleotide skews. The slope in GC skew is thus significantly different between the leading and the lagging strands; for this particular example, the slope in GC skew is higher in the leading strand than in the lagging strand. In this case, the effect of replication is an enrichment in G over C in the leading strand.
Different Skew Patterns in Prokaryotes
Figure 2 illustrates the different rearranged skew patterns that are encountered in completely sequenced prokaryotic chromosomes.
|
In the first situation, exemplified by the Anabaena variabilis chromosome (fig. 2a), our method does not detect any significant breakpoints, neither for the GC skew nor for the AT skew, or if significant breakpoints are detected, their position is distant from the origin and terminus for replication. Our interpretation for this situation is that the replication mechanism has no direct effect on base composition skews. This situation is encountered in 44 of 389 chromosomes (cf. table 1). In Bacteria, only 33 of 360 chromosomes show no significant effect of replication. The absence of direct replication effects on base composition skews seems to be more frequent in certain bacterial families, such as Mollicutes, where 10 of 16 chromosomes show no effect of replication on nucleotide skews, or Cyanobacteria (7 of 17 chromosomes). In Archaea, this situation is more frequent (11 of 29 chromosomes), but the number of completely sequenced Archaeal chromosomes is still too small to draw a general conclusion.
|
The opposite situation, illustrated here by the case of Ehrlichia canis Jake (fig. 2b), is where both types of skews display significant breakpoints that coincide with the origin and/or terminus of replication. This situation is more widely encountered (246 of 389 chromosomes). We confirm the existence of a general trend in G over C enrichment in the leading strand; only 4 species were found to have an apparent enrichment in C over G in the leading strand—Halobacterium sp. NRC-1, Streptomyces coelicolor, Thermobifida fusca, and Natronomonas pharaonis.
The direction of the AT skew shows more variability; in most cases (including E. canis Jake), the leading strand is enriched in T over A. The opposite pattern is encountered exclusively within Firmicutes and Fusobacteria (cf. The Effect of Replication on AT Skews in Firmicutes for further discussion). The case of E. canis Jake is also interesting because the AT skew displays 2 additional breakpoints, for each of the forward-encoded and reverse-encoded group of genes, which seem to have no connection with the replication origin or terminus. These breakpoints delimit 2 segments situated at 1239–1291 kb and at 1018–1052 kb on the chromosome. The GC skew, however, does not display any significant breakpoints at these positions. This situation, where additional significant breakpoints other than the terminus or origin of replication are detected, is not unfrequent in Bacteria. The existence of these additional breakpoints could in some cases be explained by recent strand inversions (indeed, skew diagrams have been previously used to detect strand inversions and deletions [Grigoriev 1998, 2000]), but it is also possible that they may delimit clusters of genes with unusual codon usage, perhaps, acquired by horizontal transfer or simply annotation errors.
In numerous prokaryotic species, the replication mechanism seems to have a direct effect only on one type of skew, either GC or AT. For example, in Carboxydothermus hydrogenoformans Z-2901 (fig. 2c), significant breakpoints are found only on the GC skew curve. This situation, where the replication mechanisms have a significant impact only on the GC skew, is shared by 79 prokaryotic chromosomes (74 Bacteria and 5 Archaea).
The opposite case, where only the AT skew displays significant breakpoints, is illustrated here by Methanopyrus kandleri (fig. 2d). In Bacteria, this situation is only encountered in 14 chromosomes. There is no clear pattern in the phylogenetic distribution of these chromosomes within Bacteria—species with this type of nucleotide skews have been encountered in numerous classes (Actinobacteria, Cyanobacteria, Deinococcus/Thermus, Mollicutes, and Proteobacteria). In Archaea, this pattern is encountered in 6 chromosomes.
Archaeal Species with Multiple Origins of Replication
Until recently, archaeal species were thought to follow the bacterial mode of replication, with a single origin and terminus. However, experimental analyses of the cell cycle have identified 3 origins of replication in the Sulfolobus acidocaldarius and Sulfolobus solfataricus chromosomes (Robinson et al. 2004; Lundgren et al. 2004). Previous in silico studies had suggested that multiple origins of replication may also be present in Halobacterium sp. NRC-1 (Zhang R and Zhang CT 2005), although only one origin was determined experimentally (Berquist and DasSarma 2003).
We wanted to verify whether our method could detect a significant effect of replication on base composition asymmetry in chromosomes with multiple origins. Figure 2e and f shows the results of our approach for S. acidocaldarius and Halobacterium sp. NRC-1. For these 2 species, we have constrained the number of breakpoints to 5, the minimum number expected for a chromosome with 3 origins of replication, although not all of them have a significant p-value.
For S. acidocaldarius, the 3 origins of replication were identified at the beginning of the published sequence, at 0.63 and 1.2 Mb (Lundgren et al. 2004). The second and the fourth breakpoints identified by our approach correspond (roughly) to the positions of the second and third origin (the positions of these breakpoints are at approximately 0.59 and 1.19 Mb for the GC skew and at approximately 0.60 and 1.21 Mb for the AT skew). It must be noted, however, that even these breakpoints are not all significant according to our approach; for 2 of the breakpoints the p-value is greater than 0.05.
For Halobacterium sp. NRC-1, a previous study had predicted 2 origins of replication, at 0.92 and 1.8 Mb (Zhang R and Zhang CT 2005), the latter of which was also confirmed experimentally (Berquist and DasSarma 2003). With our approach, the second and the fifth breakpoints are placed at approximately 0.9 and 1.8 Mb, for both types of skews. Once again, the breakpoints are not all significant; for the GC skew, only the one placed at 1.8 Mb has a p-value lower than 0.05, whereas for the AT skew none of the breakpoints is significant. We confirm that the replication mechanism seems to favor a C over G enrichment in the leading strand for this species.
These results suggest that, although our approach seems to be able to identify the positions of the multiple origins of replication, the effect of the replication in these species appears to be relatively weak, as compared with other prokaryotic species where our procedure identifies significant breakpoints.
The Effect of Replication on AT Skews in Firmicutes
Although the G over C enrichment in the leading strand seems to be a nearly universal feature of bacterial chromosomes, the sign of the AT skew on the leading strand shows much greater variability. A recent paper (Worning et al. 2006) has suggested that the sign of AT skew is determined by the polymerase- subunit that replicates the leading strand. In B. subtilis, the 2 polymerase- subunits are different, encoded by the polC and the dnaE genes; the former replicates the leading strand and the latter the lagging strand (Dervyn et al. 2001). The genomes of Firmicutes, Fusobacteria, and Aquificales encode homologues to both these genes (Worning et al. 2006). Worning et al. have shown that in genomes where both genes are present, the AT skew is positive on the leading strand, whereas in genomes that contain only dnaE the AT skew is negative. The authors suggested that the explanation for this pattern may be related to the proofreading capacity of the polymerase replicating the leading strand.
The application of our method to all the completely sequenced genomes allowed us to test if the link between the presence or absence of polC and the sign of the AT skew is still valid when the replication-related and coding sequence–related biases are decoupled. Our results indicate that the effect of replication on AT skew is highly variable even within the Firmicutes (cf. fig. 3).
|
In the first Firmicute genome represented in figure 3 (Bacillus anthracis Ames), the AT skew displays one significant breakpoint, corresponding to the terminus of replication, for each group of genes (forward encoded and reverse encoded). The slope of the AT skew is higher for genes encoded on the leading strand of the replication than for lagging strand genes, regardless of the transcription orientation of the genes. Thus, the leading strand contains indeed more A than T. The same situation is encountered for Lactobacillus salivarius UCC 118, as well as in Clostridium species and most of the Bacillus, Staphylococcus, and Lactococcus species (cf. Supplementary Material online). It must be noted, however, that not all Bacillus species show the same pattern; for example, in Bacillus halodurans and B. subtilis the AT skew does not exhibit clear breakpoints, which suggests that the effect of replication on the AT skew is weaker in these species.
The rearranged AT skew for Lactobacillus acidophilus NCFM does not display any significant breakpoints, suggesting that the replication mechanism has no significant effect on this nucleotide skew. This is in striking contrast with the pattern exhibited by the rearranged GC skew, where the difference in slope between the leading and the lagging strands is very strong. This situation is encountered in all other Lactobacillus species analyzed, with the exception of L. salivarius.
An even more surprising situation is the case of Thermoanaerobacter tengcongensis, a member of the Clostridia family. On this chromosome, significant breakpoints corresponding to the terminus of replication are found on the rearranged AT skew. The slope of the AT skew is higher for genes encoded on the lagging strands for replication; thus, the leading strand contains more T than A. The same T over A enrichment in the leading strand is encountered in Geobacillus kaustophilus.
If PolC is effectively used by all Firmicutes for the replication of the leading DNA strand, it becomes reasonable to assume that the effect of the replication mechanism on the AT skew is not determined exclusively by the differential usage of DnaE or PolC.
The inconsistency between our results and those of Worning et al. (2006) might be explained by the differences in methodology. In our approach, just as in Oriloc (Frank and Lobry 2000), the nucleotide skews are computed only on protein-coding sequences, and only for third codon positions, whereas the more complex method developed by Worning et al. (2006) computes oligomeric skews, on the entire chromosome sequence. More importantly, our method is specifically devised to separate the effects of replication and of coding sequence–related mechanisms, whereas the distinction between the possible sources of the base composition asymmetry is not a priori searched by Worning et al. (2006). As suggested by an anonymous reviewer, a possible confounding factor is the relation between the excess of genes encoded on the leading strand and the presence or absence of the polC gene (Rocha 2002).
Performance of the New Method
The method presented here is focused on the impact of the replication mechanism on the base composition asymmetry. Detecting the origin and terminus of replication is not the objective of our approach, although the coordinates of the significant breakpoints in the rearranged nucleotide skews can in most cases confirm their positions.
An important aspect is that here we compute the nucleotide skews only on third codon positions of protein-coding sequences. This choice can be justified by the generally high gene density of prokaryotic genomes; when using a sliding window approach that does not distinguish between coding and noncoding sequences, a large proportion of the nucleotides analyzed may have a functional role (e.g., first and second codon positions), which means that selective effects may superpose with mutational pressures. By taking into account only third codon positions in the skew computations, the effect of selective pressures is likely to be less important, given that these positions are mostly synonymous and, therefore, potentially neutral.
Sensitivity Analysis
Our approach relies heavily on the breakpoint detection algorithm described by Muggeo (2003). This breakpoint estimation method has been shown to be unbiased, but its accuracy is influenced by the breakpoint location and the difference in slopes (Muggeo 2003). In order to verify if this aspect could influence our results, we estimated the sensitivity of our approach using a simulated data set (cf. Material and methods and Supplementary Material online).
The results of the simulations show that the influence of the breakpoint location on the accuracy of its prediction is weak. The position of the breakpoint can be generally estimated with good accuracy, except for situations where segments are smaller than 5% of the total length of the chromosome (cf. Supplementary fig. 3). Our method might therefore fail to detect origins or termini of replication that are very close to the start or the end of the published sequence. For circular chromosomes, it is possible to circumvent the bias by rotating the sequence, but for linear chromosomes, this may indeed be a problem.
We also evaluated the impact of the slope difference (which corresponds here to the effect of replication on nucleotide skews) on the breakpoint estimation. We find that our method is able to accurately detect the breakpoint position even for a very weak effect of replication. For example, when the standard deviation of the simulated skews is set at 0.01, we can detect significant breakpoints within 2% of the chromosome length of the real breakpoints, even for slope differences as small as 10–4 (cf. Supplementary fig. 2). The global slope of the cumulated skews, that corresponds to the coding sequence–related effects, does not appear to influence the accuracy of the prediction; however, the variance of the simulated skews has a strong impact on the position of the estimated breakpoints (cf. Supplementary figs. 1 and 4).
Comparison with Other Methods for the Detection of Replication-Related Effects
The study of DNA base composition asymmetry has been the basis for the development of numerous computational methods for the detection of the origin and terminus of replication. Ranging from simple nucleotide skew analyses (Grigoriev 1998; Frank and Lobry 2000) to more complex approaches based on oligomer frequencies (Salzberg et al. 1998; Zhang R and Zhang CT 2005; Worning et al. 2006), these methods have been used to detect or to confirm the boundaries of the replichores in most of the recently published prokaryotic genomes. Comparatively less attention has been given to the study of the evolutionary mechanisms that may cause the base composition asymmetries, although several statistical methods that assess the relative contributions of the different types of mechanisms have been proposed (Perrière et al. 1996; Tillier and Collins 2000; Lobry and Sueoka 2002).
We performed a comparison of this new approach with the methods described by Tillier and Collins (2000) and Lobry and Sueoka (2002) on the 173 chromosomes for which significance computations were performed. Both these methods are based on a comparison between the means of the nucleotide skews on the leading and lagging strands for replication; the former uses a 2-way ANOVA of the skews, with the replication and transcription orientation as explanatory variables, whereas the latter performs a t-test to compare the skews computed on the transcribed strand of the genes between the leading and the lagging strand for replication. We followed the conventions of Tillier and Collins (2000) and Lobry and Sueoka (2002) and considered that the replication mechanism had a significant effect on the nucleotide skews when the p-value was lower than 0.01. We did not use the same significance threshold for our approach because our randomization test was considerably less powerful than the previously cited parametric tests; we chose instead a threshold of 0.05 for the significance of the breakpoints. In our approach, we consider that replication has an effect on the base composition asymmetry when the rearranged skews show significant breakpoints close to the origin and terminus (see Supplementary Material online for more details). With these conventions, we find that there is a general agreement between the 3 methods, although our approach seems to be more conservative. For the AT skew, our approach disagrees with the methods described by Tillier and Collins (2000) and Lobry and Sueoka (2002) in only 18 and 19 chromosomes, respectively. For the GC skews, the numbers are similar (14 and 13, respectively)—see table 2 and Supplementary table 2. Almost all these incoherences between the 3 approaches correspond to situations where our method does not detect a significant effect of replication. A possible explanation for these inconsistencies may be that the approach presented here lacks sensitivity, as compared with the parametric tests used by Tillier and Collins (2000) and Lobry and Sueoka (2002). However, although these 2 methods are more sensitive, they may also be prone to biases such as uneven distribution of highly expressed genes, with highly biased codon usage, between the leading and lagging strands.
|
|
Conclusion and Perspectives |
|---|
|
TOPAbstractIntroductionMaterial and MethodsResults and DiscussionConclusion and PerspectivesSupplementary MaterialAcknowledgementsReferences |
|---|
We presented here a new computational approach for the analysis of the base composition asymmetry, aimed for the specific detection of replication-associated effects, and we applied it to a large number of complete bacterial and archaeal genomes. We were able to show that in many species the replication mechanism has different impacts on the 2 types of nucleotide skews. Our method has allowed us to address the question of the variability of the direction of the AT skew in prokaryotic species, and we concluded that the differential usage of the PolC/DnaE polymerase
-subunits is not the only factor that influences the sign of the AT skew on the leading strand.
A recent work has demonstrated that the mutational processes that cause base composition asymmetry vary considerably between species, even when the compositional biases are similar (Rocha et al. 2006); another interesting study has shown that differences in skew intensity between closely related species may be caused by the loss of genes involved in replication restart pathways (Klasson and Andersson 2006). We have shown here that the asymmetry patterns are more complex than originally thought; we believe that the methods used by Rocha et al. (2006) and Klasson and Andersson (2006) may also be able to shed more light on the causes of the variability in base composition skews.
|
Supplementary Material |
|---|
|
TOPAbstractIntroductionMaterial and MethodsResults and DiscussionConclusion and PerspectivesSupplementary MaterialAcknowledgementsReferences |
|---|
Supplementary figs 1–4, table 2, and other Supplementary Materials are found at Molecular Biology and Evolution online (http://www.mbe.oxfordjournals.org/).
|
Acknowledgements |
|---|
|
TOPAbstractIntroductionMaterial and MethodsResults and DiscussionConclusion and PerspectivesSupplementary MaterialAcknowledgementsReferences |
|---|
We thank the IN2P3 Computing Center for providing the computer resources. We thank Bastien Boussau as well as 2 anonymous reviewers for critically reading the manuscript and for their helpful suggestions. This work was supported by the Agence Nationale de la Recherche (GIP ANR JC05-49162) and by the Centre Nationale de la Recherche Scientifique. A. N. is supported by a scholarship from the Ministère de l'Education Nationale, de l'Enseignement Supérior et de la Recherche.
|
Footnotes |
|---|
Aoife McLysaght, Associate Editor
|
References |
|---|
|
TOPAbstractIntroductionMaterial and MethodsResults and DiscussionConclusion and PerspectivesSupplementary MaterialAcknowledgementsReferences |
|---|
Berquist BR, DasSarma S. An archaeal chromosomal autonomously replicating sequence element from an extreme halophile, Halobacterium sp. strain NRC-1. J Bacteriol. (2003) 185:5959–5966.
Bird R, Louarn J, Martuscelli J, Caro L. Origin and sequence of chromosome replication in Escherichia coli. J Mol Biol. (1972) 70:549–566.[CrossRef][Web of Science][Medline]
Blattner FR, Plunkett G, Bloch CA. (17 co-authors). The complete genome sequence of Escherichia coli K-12. Science (1997) 277:1453–1462.
Brewer BJ. When polymerases collide: replication and the transcriptional organization of the E. coli chromosome. Cell (1988) 53:679–686.[CrossRef][Web of Science][Medline]
Charif D, Lobry JR. SeqinR 1.0–2: a contributed package to the R project for statistical computing devoted to biological sequences retrieval and analysis. In: Structural approaches to sequence evolution: molecules, networks, populations. Biological and Medical Physics, Biomedical Engineering—Bastolla U, Porto M, Roman HE, Vendruscolo M, eds. (2006) New York: Springer Verlag.
Dervyn E, Suski C, Richard D, Bruand C, Chapuis J, Errington J, Jannière L, Ehrlich SD. Two essential DNA polymerases at the bacterial replication fork. Science (2001) 294:1716–1719.
Francino MP, Chao L, Riley MA, Ochman H. Asymmetries generated by transcription-coupled repair in enterobacterial genes. Science (1996) 272:107–109.[Abstract]
Frank AC, Lobry JR. Asymmetric substitution patterns: a review of possible underlying mutational or selective mechanisms. Gene (1999) 238:65–77.[CrossRef][Web of Science][Medline]
Frank AC, Lobry JR. Oriloc: prediction of replication boundaries in unannotated bacterial chromosomes. Bioinformatics (2000) 16:560–561.
Francino MP, Ochman H. Strand asymmetries in DNA evolution. Trends Genet. (1997) 13:240–245.[CrossRef][Web of Science][Medline]
Grigoriev A. Analyzing genomes with cumulative skew diagrams. Nucleic Acids Res. (1998) 26:2286–2290.
Grigoriev A. Graphical genome comparison: rearrangements and replication origin of Helicobacter pylori. Trends Genet. (2000) 16:376–378.[CrossRef][Web of Science][Medline]
Klasson L, Andersson SGE. Strong asymmetric mutation bias in endosymbiont genomes coincide with loss of genes for replication restart pathways. Mol Biol Evol. (2006) 23:1031–1039.
Lobry JR. Properties of a general model of DNA evolution under no-strand-bias conditions. J Mol Evol. (1995) 40:326–330.[CrossRef][Web of Science][Medline]
Lobry JR. Asymmetric substitution patterns in the two DNA strands of bacteria. Mol Biol Evol. (1996a) 13:660–665.[Abstract]
Lobry JR. Origin of replication of Mycoplasma genitalium. Science (1996b) 272:745–746.[CrossRef][Web of Science][Medline]
Lobry JR, Sueoka N. Asymmetric directional mutation pressures in bacteria. Genome Biol. (2002) 3:research0058.[Medline]
Lundgren M, Andersson A, Chen L, Nilsson P, Bernander R. Three replication origins in Sulfolobus species: synchronous initiation of chromosome replication and asynchronous termination. Proc Natl Acad Sci USA (2004) 101:7046–7051.
Mackiewicz P, Zakrzewska-Czerwinska J, Zawilak A, Dudek MR, Cebrat S. Where does bacterial replication start? Rules for predicting the oriC region. Nucleic Acids Res. (2004) 32(13):3781–3791.
McInerney JO. Replicational and transcriptional selection on codon usage in Borrelia burgdorferi. Proc Natl Acad Sci USA (1998) 95:10698–10703.
McLean MJ, Wolfe KH, Devine KM. Base composition skews, replication orientation, and gene orientation in 12 prokaryote genomes. J Mol Evol. (1998) 47:691–696.[CrossRef][Web of Science][Medline]
Muggeo VMR. Estimating regression models with unknown break-points. Stat Med. (2003) 22:3055–3071.[CrossRef][Web of Science][Medline]
Muggeo VMR. Segmented: segmented relationships in regression models. R package version 0.1–4 (2004).
Nikolaou C, Almirantis Y. A study on the correlation of nucleotide skews and the positioning of the origin of replication: different modes of replication in bacterial species. Nucleic Acids Res. (2005) 33:6816–6822.
Nomura M, Morgan EA, Jaskunas SR. Genetics of bacterial ribosomes. Annu Rev Genet. (1977) 11:297–347.[CrossRef][Web of Science][Medline]
Perrière G, Lobry JR, Thioulouse J. Correspondence discriminant analysis: a mutivariate method for comparing classes of protein and nucleic acid sequences. Comput Appl Biosci. (1996) 12:519–524.
R Development Core Team. R: a language and environment for statistical computing (2005) Vienna: Austria. R Foundation for Statistical Computing. ISBN 3-900051-07-0.
Robinson NP, Dionne I, Lundgren M, Marsh VL, Bernander R, Bell SD. Identification of two origins of replication in the single chromosome of the archaeon Sulfolobus solfataricus. Cell (2004) 116:25–38.[CrossRef][Web of Science][Medline]
Rocha EPC. Is there a role for replication fork asymmetry in the distribution of genes in bacterial genomes? Trends Microbiol. (2002) 10:393–395.[CrossRef][Web of Science][Medline]
Rocha EPC. The replication-related organization of bacterial genomes. Microbiology (2004) 150:1609–1627.
Rocha EPC, Danchin A. Gene essentiality determines chromosome organisation in bacteria. Nucleic Acids Res. (2003a) 31:6570–6577.
Rocha EPC, Danchin A. Essentiality, not expressiveness, drives gene-strand bias in bacteria. Nat Genet. (2003b) 34:377–378.[CrossRef][Web of Science][Medline]
Rocha EPC, Danchin A, Viari A. Universal replication biases in bacteria. Mol Microbiol. (1999) 32:11–16.[CrossRef][Web of Science][Medline]
Rocha EPC, Touchon M, Feil EJ. Similar compositional biases are caused by very different mutational effects. Genome Res. (2006) 16:1537–1547.
Rudner R, Karkas JD, Chargaff E. Separation of B. subtilis DNA into complementary strand, III. Direct analysis. Proc Natl Acad Sci USA (1968) 60:921–922.
Salzberg SL, Salzberg AJ, Kerlavage AR, Tomb JF. Skewed oligomers and the origin of replication. Gene (1998) 217:57–67.[CrossRef][Web of Science][Medline]
Sueoka N. Intrastrand parity rules of DNA base composition and usage biases of synonymous codons. J Mol Evol. (1995) 40:318–325.[CrossRef][Web of Science][Medline]
Tillier ERM, Collins RA. The contributions of replication orientation, gene direction, and signal sequences to base-composition asymmetries in bacterial genomes. J Mol Evol. (2000) 50:249–257.[Web of Science][Medline]
Worning P, Jensen LJ, Hallin PF, Staerfeldt HH, Ussery DW. Origin of replication in circular prokaryotic chromosomes. Environ Microbiol. (2006) 8:353–361.[CrossRef][Medline]
Zhang R, Zhang CT. Single replication origin of the archaeon Methanosarcina mazei revealed by the Z curve methode. Biochem Biophys Res Commun. (2002) 297:396–400.[CrossRef][Web of Science][Medline]
Zhang R, Zhang CT. Multiple replication origins of the archaeon Halobacterium species nrc-1. Biochem Biophys Res Commun. (2003) 302:728–734.[CrossRef][Web of Science][Medline]
Zhang R, Zhang CT. Identification of replication origins in archaeal genomes based on the Z-curve method. Archaea (2005) 1:335–346.[Medline]
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  F.-B. Guo and J.-B. Yuan Codon Usages of Genes on Chromosome, and Surprisingly, Genes in Plasmid are Primarily Affected by Strand-specific Mutational Biases in Lawsonia intracellularis DNA Res, April 1, 2009; 16(2): 91 - 104. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 K. Vetsigian and N. Goldenfeld Genome rhetoric and the emergence of compositional bias PNAS, January 6, 2009; 106(1): 215 - 220. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 N. V. Sernova and M. S. Gelfand Identification of replication origins in prokaryotic genomes Brief Bioinform, September 1, 2008; 9(5): 376 - 391. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||