Separating the effects of mutation and selection in producing DNA skew in bacterial chromosomes

Abstract

Background: Many bacterial chromosomes display nucleotide asymmetry, or skew, between the leading and lagging strands of replication. Mutational differences between these strands result in an overall pattern of skew that is centered about the origin of replication. Such a pattern could also arise from selection coupled with a bias for genes coded on the leading strand. The relative contributions of selection and mutation in producing compositional skew are largely unknown.

Results: We describe a model to quantify the contribution of mutational differences between the leading and lagging strands in producing replication-induced skew. When the origin and terminus of replication are known, the model can be used to estimate the relative accumulation of G over C and of A over T on the leading strand due to replication effects in a chromosome with bidirectional replication arms. The model may also be implemented in a maximum likelihood framework to estimate the locations of origin and terminus. We find that our estimations for the origin and terminus agree very well with the location of genes that are thought to be associated with the replication origin. This indicates that our model provides an accurate, objective method of determining the replication arms and also provides support for the hypothesis that these genes represent an ancestral cluster of origin-associated genes.

Conclusion: The model has several advantages over other methods of analyzing genome skew. First, it quantifies the role of mutation in generating skew so that its effect on composition, for example codon bias, can be assessed. Second, it provides an objective method for locating origin and terminus, one that is based on chromosome-wide accumulation of leading vs lagging strand nucleotide differences. Finally, the model has the potential to be utilized in a maximum likelihood framework in order to analyze the effect of chromosome rearrangements on nucleotide composition.

Figure 1

Chromosome division. The distribution of C_d (chromosome division, Equation 10) based on [S₁, S₂]^ML pairs for the 326 chromosomes indicated in the text. C_d is plotted along the X axis and represents the deviation from equal chromosome division.

Figure 2

Strand asymmetry across bacterial chromosomes. A scatter plot of R_T and R_G for D₄ sites on the leading strands of the 352 bacterial chromosomes. The values represent the average effect for the two leading strands in each of the replication arms, with the bars indicating the 95% uncertainty. The straight line represents R_T = R_G.

Figure 3

Strand asymmetry in Firmicutes. The same plot as in Figure 2 with the Firmicute chromosomes indicated as open points. Open squares indicate those of the class Mollicutes (which includes the Mycoplasma genus) while the open circles indicate all other Firmicutes.

Figure 4

Likelihood surface for the E. coli K12 chromosome. This representation of the likelihood surface of the E. coli K12 chromosome (NC_000913) is based on the M_OBS model (see Methods). The two axes represent relative chromosome length so that every point represents the likelihood analysis on the pair of chromosome locations [S₁, S₂]. The grey scale inset shows the conversion of log likelihood to grey value, with the maximum log likelihood (LL_max) as black and the bars indicating -10 decrements. Since not every pair of chromosome locations was sampled the points are dispersed. The likelihood analysis is symmetrical around the diagonal so the two maxima are identical and represent just one pair of chromosome locations, but interchanging leading and lagging strands. The negative lines represent the location of the annotated ori and ter. The positive lines represent the maximum likelihood (ML) values and Monte Carlo estimated 95% ranges.