The source of laterally transferred genes in bacterial genomes

Abstract

Background: Laterally transferred genes have often been identified on the basis of compositional features that distinguish them from ancestral genes in the genome. These genes are usually A+T-rich, arguing either that there is a bias towards acquiring genes from donor organisms having low G+C contents or that genes acquired from organisms of similar genomic base compositions go undetected in these analyses.

Results: By examining the genome contents of closely related, fully sequenced bacteria, we uncovered genes confined to a single genome and examined the sequence features of these acquired genes. The analysis shows that few transfer events are overlooked by compositional analyses. Most observed lateral gene transfers do not correspond to free exchange of regular genes among bacterial genomes, but more probably represent the constituents of phages or other selfish elements.

Conclusions: Although bacteria tend to acquire large amounts of DNA, the origin of these genes remains obscure. We have shown that contrary to what is often supposed, their composition cannot be explained by a previous genomic context. In contrast, these genes fit the description of recently described genes in lambdoid phages, named 'morons'. Therefore, results from genome content and compositional approaches to detect lateral transfers should not be cited as evidence for genetic exchange between distantly related bacteria.

Figure 1

Principle of the detection of recently acquired and lost genes using parsimony. Genes present in species A and absent from species B and C (+A -B -C) are likely to have been acquired recently if the number of lost genes in the sister species (+A -B +C) is relatively small.

Figure 2

Results of the approach described in Figure 1 in three groups of closely related bacteria. Italic numbers refer to lost genes. A list of the acquired genes is available as an Additional data file.

Figure 3

Gene acquisitions and losses. The method described here (see Figure 1) only identifies losses of genes (in genome A) that have been conserved in the two other lineages considered (genomes B and C; in green). If recent acquisitions (in black) are deleted shortly after their integration in the genome, what we observe is a high number of acquisitions compared with losses. This, rather than an increase in genome size, may explain the results presented in Figure 2.

Figure 4

Intraspecies FCA. (a) E. coli; (b) Salmonella enterica; (c) S. pneumoniae; and (d) H. pylori. Both genes (top) and codons (bottom) are plotted on the two first axes of the FCA. Codons are labeled according to the nature of the base at the third position. The percentages of variability explained by the axes are shown between brackets.

Figure 5

Interspecies FCA for the four groups of species considered. (a) Native genes; (b) LTG; (c) IS; and (d) codons are presented separately in superimposable figures. The first two axes, which represent 22.98% and 7.29%,, respectively, of the total variability, are shown. Ellipses represent 90% of the points of each cloud. The arrows in (b) and (c) represent the displacement of the center of the ellipse relative to that of the native genes. Phages were not included in the present analysis because no sequences were found in the H. pylori and the S. pneumoniae genomes. In (d), green squares represent A+T-rich codons and purple squares G+C-rich codons.

Figure 6

G+C content at the third position of codons in the different classes of gene. IS and phages are absent from certain species because their numbers were insufficient. Bars represent 95% of confidence interval.

Figure 7

Relative neutrality plots for the different classes of gene in E. coli O157:H7. GC12 is plotted as a function of GC3 and the slope of the correlation (bold line) is computed.