Rapid pathway evolution facilitated by horizontal gene transfers across prokaryotic lineages

Abstract

The evolutionary history of biological pathways is of general interest, especially in this post-genomic era, because it may provide clues for understanding how complex systems encoded on genomes have been organized. To explain how pathways can evolve de novo, some noteworthy models have been proposed. However, direct reconstruction of pathway evolutionary history both on a genomic scale and at the depth of the tree of life has suffered from artificial effects in estimating the gene content of ancestral species. Recently, we developed an algorithm that effectively reconstructs gene-content evolution without these artificial effects, and we applied it to this problem. The carefully reconstructed history, which was based on the metabolic pathways of 160 prokaryotic species, confirmed that pathways have grown beyond the random acquisition of individual genes. Pathway acquisition took place quickly, probably eliminating the difficulty in holding genes during the course of the pathway evolution. This rapid evolution was due to massive horizontal gene transfers as gene groups, some of which were possibly operon transfers, which would convey existing pathways but not be able to generate novel pathways. To this end, we analyzed how these pathways originally appeared and found that the original acquisition of pathways occurred more contemporaneously than expected across different phylogenetic clades. As a possible model to explain this observation, we propose that novel pathway evolution may be facilitated by bidirectional horizontal gene transfers in prokaryotic communities. Such a model would complement existing pathway evolution models.

Figure 1. Schematic representation of the gene-content reconstruction algorithm adopted in the present study.

In this model example, there are six genomes and four ortholog groups. The algorithm first estimates the most likely gene gain/loss rates on each branch of the phylogenetic tree. Then, it reconstructs the most likely history of the gene-content evolution by using the estimated gene gain/loss rates, without arbitrary parameters.

Figure 2. Prokaryotic species analyzed and the phylogenetic tree, overlaid with rapid pathway acquisition events.

Species codes and taxonomic classifications are from the KEGG database (for complete species names, see Table S1). Branch lengths reflect evolutionary time estimated by Ciccarelli et al. to some extent, without impairing readability. Symbols on the tree indicate the places where pathways were acquired rapidly (see the main text). Small circles indicate a single appearance, whereas large symbols of the same shape and color indicate the independent acquisition of the same pathway at different places on the phylogenetic tree. Colors indicate the metabolic categories in which the pathway is involved; shapes have no special meaning.

Figure 3. Schematic representation of the procedure used to detect the acquired pathways.

Note that the gene/enzyme connected via a compound that already existed in the ancestral metabolic network is not comprised in the acquired pathway (indicated by an asterisk).

Figure 4. Functional distribution of the acquired pathways.

Number of pathway acquisition events for each functional category is shown. The vertical axis corresponds to the observed numbers, whereas the horizontal axis represents the expected numbers based on the functional distribution of all genes in this study. The functional categories above the dashed 45° line were more frequently observed than expected.

Figure 5. Modes of pathway acquisition.

Each sequential line from the bottom left corner to the top right represents how rapidly a pathway was acquired. A segment corresponds to one branch during the evolutionary period when the pathway was acquired. Its slope represents the proportion of acquired genes divided by the branch length derived from the linearized phylogenetic tree (see Methods). Segments were sorted in descending order of their slopes to visualize how strongly the acquisition was biased. If the gradual acquisition scenario holds true, sequential lines will approach the 45° line, whereas the rapid acquisition scenario will produce lines that are strongly convex upward. To permit comparison, representative lines for the present and the randomized data are shown as gray dashed lines.