Lateral gene transfer as a support for the tree of life

Abstract

Lateral gene transfer (LGT), the acquisition of genes from other species, is a major evolutionary force. However, its success as an adaptive process makes the reconstruction of the history of life an intricate puzzle: If no gene has remained unaffected during the course of life's evolution, how can one rely on molecular markers to reconstruct the relationships among species? Here, we take a completely different look at LGT and its impact for the reconstruction of the history of life. Rather than trying to remove the effect of LGT in phylogenies, and ignoring as a result most of the information of gene histories, we use an explicit phylogenetic model of gene transfer to reconcile gene histories with the tree of species. We studied 16 bacterial and archaeal phyla, representing a dataset of 12,000 gene families distributed in 336 genomes. Our results show that, in most phyla, LGT provides an abundant phylogenetic signal on the pattern of species diversification and that this signal is robust to the choice of gene families under study. We also find that LGT brings an abundant signal on the location of the root of species trees, which has been previously overlooked. Our results quantify the great variety of gene transfer rates among lineages of the tree of life and provide strong support for the "complexity hypothesis," which states that genes whose products participate to macromolecular protein complexes are relatively resistant to transfer.

Fig. 1.

LGTs are informative for phylogeny. An LGT can be viewed as a phylogenetic character, and the sharing of an LGT can be viewed as a shared derived character. An objective criterion to select a species tree can be the number of LGTs that need to be invoked in a collection of gene trees. In this illustration, two candidate species trees and two gene trees are presented. Given these species trees, different LGT events are needed to reconcile gene trees. Gene tree 1 has a score of 2 LGTs given species tree 1 and 1 LGT given species tree 2. A global score on all gene trees can be assigned to each candidate species tree. Here, species tree 1 has a score of 3 LGTs, whereas species tree 2 has a score of 2 and is therefore more parsimonious.

Fig. 2.

LGT scores for different candidate species trees. For each of the 16 phyla analyzed, eight candidate species trees (T1–T8) derived from different methods (Materials and Methods and SI Materials and Methods) were tested [T1, concatenated tree (concat); T2, recoded concatenated tree (recod); T3, consensus of trees from universal family (consense); T4, consensus of jackknife replicates (jackknife); T5, super distance matrix tree (SDM); T6, 16S rRNA (16s); T7, 23S rRNA (23s); and T8, 16S + 23S rRNA (16s + 23s)]. Each hypothesis has an associated LGT score that corresponds to the number of LGTs inferred by Prunier in the families under study (Nb fam). Dark blue cells represent trees that are topologically identical to the most parsimonious tree in terms of LGT. Light blue cells correspond to trees that are not statistically different from the most parsimonious tree. Trees with white cells are significantly different from the best tree. Numbers of transfers per hundred branches per family are shown (LGT rate). Chlamydiales-Verruco, Chlamydiales-Verrucomicrobia; Nb sp, number of species.

Fig. 3.

A prokaryotic tree of life with branch-wise rates of transfers. Within-phyla relationships correspond to the minimum LGT tree, and relationships among phyla are reconstructed from a concatenate of 157 protein alignments after removal of transferred genes (Fig. S1). For each phylum, the root corresponding to the result of the outgroup method is presented, except for Actinobacteria and Gammaproteobacteria for which is shown the minimum LGT root, which was significantly better in terms of LGT score (Table S4; P < 0.05). Branch lengths were estimated on the basis of universal single-copy gene concatenation, and each branch is colored according to the proportion of gene families analyzed where the branch was transferred. Spiro, Spirochaetes; Chlamy, Chlamydiales-Verrucomicrobia; Bacter/Chloro, Bacteroidetes/Chlorobi; Delta, Deltaproteobacteria; Epsilon, Epsilonproteobacteria.

Fig. 4.

LGT rates across different functional categories. The dataset was annotated with the three COG categories: cellular processes and signaling (Cell), information storage and processing (Info), and metabolism (Meta). Average rates of LGT per 100 branches and per family are given for the three categories in corresponding boxes for each phylum and for the pooled dataset (All Phyla). Different colors correspond to classes of statistical equivalence in pairwise tests (Tukey–Kramer test, α = 0.05). Transparent boxes indicate no significant difference between transfer rates of the three categories. Chlamydiales-Verruco, Chlamydiales-Verrucomicrobia.