A spectrum of verticality across genes

Abstract

Lateral gene transfer (LGT) has impacted prokaryotic genome evolution, yet the extent to which LGT compromises vertical evolution across individual genes and individual phyla is unknown, as are the factors that govern LGT frequency across genes. Estimating LGT frequency from tree comparisons is problematic when thousands of genomes are compared, because LGT becomes difficult to distinguish from phylogenetic artefacts. Here we report quantitative estimates for verticality across all genes and genomes, leveraging a well-known property of phylogenetic inference: phylogeny works best at the tips of trees. From terminal (tip) phylum level relationships, we calculate the verticality for 19,050,992 genes from 101,422 clusters in 5,655 prokaryotic genomes and rank them by their verticality. Among functional classes, translation, followed by nucleotide and cofactor biosynthesis, and DNA replication and repair are the most vertical. The most vertically evolving lineages are those rich in ecological specialists such as Acidithiobacilli, Chlamydiae, Chlorobi and Methanococcales. Lineages most affected by LGT are the α-, β-, γ-, and δ- classes of Proteobacteria and the Firmicutes. The 2,587 eukaryotic clusters in our sample having prokaryotic homologues fail to reject eukaryotic monophyly using the likelihood ratio test. The low verticality of α-proteobacterial and cyanobacterial genomes requires only three partners-an archaeal host, a mitochondrial symbiont, and a plastid ancestor-each with mosaic chromosomes, to directly account for the prokaryotic origin of eukaryotic genes. In terms of phylogeny, the 100 most vertically evolving prokaryotic genes are neither representative nor predictive for the remaining 97% of an average genome. In search of factors that govern LGT frequency, we find a simple but natural principle: Verticality correlates strongly with gene distribution density, LGT being least likely for intruding genes that must replace a preexisting homologue in recipient chromosomes. LGT is most likely for novel genetic material, intruding genes that encounter no competing copy.

Fig 1

Comparison of estimated verticality and number of genomes in a protein cluster for a. all clusters (n = 101,422) and b. all conserved clusters (average branch length ≥ 0.1; n = 8,547). Unrooted trees were analyzed if at least two taxonomic groups were present. Verticality was calculated as the sum of monophyletic taxonomic groups in a cluster adjusted by the fraction of a taxonomic group represented in the cluster. The procedure for determining verticality on the basis of an example is shown in S3 Fig. This value correlates with the number of genomes, an approximation of universality, which is even more apparent when clusters of high evolutionary rate were filtered out (a.: p < 10⁻³⁰⁰, Pearson´s R² = 0.726; b.: p < 10⁻³⁰⁰, R² = 0.829). In both plots clusters of special interest were marked: The eukaryotic-prokaryotic clusters (EPCs) are highlighted in red and the clusters that correspond to a gene from the mitochondrial genome of Reclinomonas americana [45] are displayed in blue triangles along the abscissa of the plot and in the graph. For the latter, the gene identifier was noted above each plot. Ribosomal proteins are indicated by the black diamond on the right of each plot and in the graph [6]. Notably, the ribosomal protein clusters show a steep gradient of verticality among conserved clusters with similarly wide distribution.

Fig 2. Comparison of estimated verticality and number of genomes [%] for the 100 most vertical clusters.

Identity and Annotation of clusters can be found in S6 Table. This is a representation of some of the clusters shown in the blue rectangle of Fig 1A.

Fig 3. Relative occurrence of a taxonomic group as the sister group of each clade in the unrooted trees.

For each taxonomic group in a cluster the sister was determined and counted. Multiple occurrences of different groups in the sister were accounted for by their relative occurrence. If the taxonomic group was paraphyletic, each monophyletic subgroup was determined and the sister of these were noted. The values of these subgroups were added up by multiplying the individual values of the sister by the fraction of the subgroup of the whole taxonomic group. To compare, the final values of each taxonomic group were normalized by dividing by the highest count a possible sister has gotten. It is apparent that Gammaproteobacteria are always overrepresented. It is not clear if the observed effects are due to overrepresentation of certain taxa in the data set or due to relative abundance of LGT.

Fig 4. Identification of the prokaryotic sister group to the eukaryotes in 2,575 eukaryotic-prokaryotic unrooted gene trees (EPC).

a. shows the average clade sizes for eukaryotes, the sister group to eukaryotes and the outgroup in the analyzed trees for (right) the 229 trees with only plastid derived lineages and (left) for the 456 EPCs containing all taxa except photosynthetic lineages. b. details the list of bacterial (top) and archaeal (bottom) phyla occurring in the trees with only plant lineages (right) and all other trees (left) that were filtered for conservation (average branch length of the tree ≤ 0.1). Archaeal and bacterial phyla with less than 5 representative species in the dataset were collapsed into ‘other archaea’ and ‘other bacteria’ groups. P_mono refers to the proportion of trees with a branch (split) separating the species of the respective phylum from all the others in the tree; S_non refers to the number of occurrence of the phylum only in the outgroup clade; S_mix refers to the number of occurrences of the phylum as a mixed sister (more than one phylum in the clade); S_pure refers to the number of occurrences of the phylum as pure sister (as the single phylum); S_p,avg shows the average size of the sister clade when the phylum occurs as a pure sister clade. N_trees show the number of occurrences of the phyla across the trees and N_gen indicates the number of species in each taxon included in the complete dataset.

Fig 5

Mapping of EPCs to prokaryotic clusters. The EPCs were separated according to the pure sister group of eukaryotes in the unrooted trees: a. and b. Alphaproteobacteria, c. and d. Cyanobacteria, e. and f. Gammaproteobacteria. The left panel shows EPCs that may include all eukaryotic supergroups but no groups that include only photosynthetic lineages, the right panel shows only EPCs that only include photosynthetic eukaryotes (lineages from SAR, Hacrobia and Archaeplastida). Meaning the latter are indicative of plastid endosymbiosis. Plots for all taxa see S5 Fig.