Evolutionary analysis of a streamlined lineage of surface ocean Roseobacters

Abstract

The vast majority of surface ocean bacteria are uncultivated. Compared with their cultured relatives, they frequently exhibit a streamlined genome, reduced G+C content and distinct gene repertoire. These genomic traits are relevant to environmental adaptation, and have generally been thought to become fixed in marine bacterial populations through selection. Using single-cell genomics, we sequenced four uncultivated cells affiliated with the ecologically relevant Roseobacter clade and used a composition-heterogeneous Bayesian phylogenomic model to resolve these single-cell genomes into a new clade. This lineage has no representatives in culture, yet accounts for ∼35% of Roseobacters in some surface ocean waters. Analyses of multiple genomic traits, including genome size, G+C content and percentage of noncoding DNA, suggest that these single cells are representative of oceanic Roseobacters but divergent from isolates. Population genetic analyses showed that substitution of physicochemically dissimilar amino acids and replacement of G+C-rich to G+C-poor codons are accelerated in the uncultivated clade, processes that are explained equally well by genetic drift as by the more frequently invoked explanation of natural selection. The relative importance of drift vs selection in this clade, and perhaps in other marine bacterial clades with streamlined G+C-poor genomes, remains unresolved until more evidence is accumulated.

Figure 1

Regression model for Roseobacter SAG genome size estimation based on genome statistics for 40 cultured Roseobacter strains. The x axis shows the ratio of the number of conserved single-copy genes universally present in fully sequenced Roseobacter genomes to the number of predicted protein-coding genes in a genome; the y axis is the number of nucleotides sequenced. The data were fit to a polynomial regression model (R²=0.93), and the model was used to estimate the genome sizes of the four SAGs (C03, O19, K06 and J04), which were found to be 2.64, 3.10, 3.50 and 2.65 Mb, respectively. The prediction interval (PI) is also shown.

Figure 2

Genomic characteristics of the three SAG-O19 clade members compared with metagenomic roseobacter sequences and isolates. Distribution of G+C content sampled from Global Ocean Survey (GOS) metagenomic data sets (a) and cultured Roseobacter genomes (b). For panel a, the GOS roseobacter reads (n=5608) were identified using the d_N pipeline software, which assigns metagenomic reads to a microbial clade with high confidence (Luo et al., 2012). For panel b, the cultured roseobacter genomes were randomly sheared to generate an in silico metagenome, and then analyzed as described for panel a. The gray arrow in panel b indicates the average G+C content of the SAGs. The fraction of noncoding DNA (c) and estimated average genome size (d) in Roseobacter GOS reads, SAGs and isolates. The box-and-whiskers plots indicate the median (horizontal line), the boundary of the first (Q1) and third quartile (Q3) (box) and the range from Q1–1.5(Q3–Q1) to Q3+1.5(Q3–Q1) (whiskers); circles indicate outliers. The details for noncoding DNA fraction and genome size estimation can be found in the Materials and methods.

Figure 3

Bayesian phylogenomic tree of the Roseobacter clade using a composition-heterogeneous model in the P4 software package based on a concatenation of 52 single-copy orthologous protein sequences. The scale bar indicates the number of substitutions per site. The value near each internal branch is the posterior probability for that branch. The tree is rooted using four genomes from sister clades in the Alphaproteobacteria (Hyphomonadaceae, Caulobacterales, Rhizobiales); the outgroups are not shown. The SAG clade is highlighted in blue, an example control clade is highlighted in orange and the remaining members used to calculate d_R/d_C ratios are indicated in green. The basal lineage consisting of HTCC2255 and SAG C03 were not included in the analyses. G+C content and genome size are indicated in the two right columns, the latter estimated using a regression model for incomplete genomes (Figure 1). Plasmids, if any, are not included in genome size calculation.

Figure 4

The ratio of radical (d_R) to conservative (d_C) substitution rates based on physicochemical properties (shown here, charge) of amino acids for the SAG-O19 clade and six control clades. Bars indicate one standard deviation of the mean. For the control clades, only strain names are shown; full names can be found in Figure 3.