Phylogenomics reveals the slow-burning fuse of diatom evolution

Abstract

Evolution is often uneven in its pace and outcomes, with long periods of stasis interrupted by abrupt increases in morphological and ecological disparity. With thousands of gene histories, phylogenomics can uncover the genomic signatures of these broad macroevolutionary trends. Diatoms are a species-rich lineage of microeukaryotes that contribute greatly to the global cycling of carbon, oxygen, and silica, which they use to build elaborately structured cell walls. We combined fossil information with newly sequenced transcriptomes from 181 diverse diatom species to reconstruct the pattern, timing, and genomic context of major evolutionary transitions. Diatoms originated 270 Mya, and after >100 My of relative stasis in morphology and ecology, a radiation near the Jurassic-Cretaceous boundary led to the diversity of habitats and cell wall architectures characteristic of modern diatoms. This transition was marked by a genome duplication and high levels of gene tree discordance. However, short generation times increase the probability of coalescence between speciation events, minimizing the impacts of incomplete lineage sorting and implicating sequence saturation and gene tree error as the main sources of discordance. Nevertheless, a rigorous tree-based approach to ortholog selection resulted in strongly supported relationships, including some that were uncertain previously. Three pulses of accelerated speciation were detected, two of which were associated with the evolution of novel traits and ecological transitions. The first 100 My of diatom evolution was a slow-burning fuse that led to a burst of innovations in ecology, morphology, and life history that are hallmarks of contemporary diatom assemblages.

Keywords: concordance; discordance; diversification; incomplete lineage sorting; microbes.

Fig. 1.

High-quality transcriptomes from ecologically and phylogenetically diverse diatoms. Numbers represent total diatom strains, most of which are distinct species (A). Transcriptomes recovered a large percentage of conserved eukaryotic (B) and stramenopile (C) orthologs (BUSCOs). The estimated completeness of newly sequenced transcriptomes compared favorably to reassembled diatom transcriptomes from the MMETSP (23).

Fig. 2.

The diatom tree of life. Relationships are based on maximum likelihood analysis of 240 genes and 171,434 amino acids under the LG+PMSF(C20)+F+G model. Ten numbered clades are highlighted to facilitate discussion. (A) Order-level phylogeny with major groups identified by common name. (B) Time-calibrated phylogeny. Node bars represent CI of divergence time estimates. Geologic periods are indicated as: P (Permian), Tr (Triassic), J (Jurassic), K (Cretaceous), Pg (Paleogene), and Ng (Neogene). Scanning electron microscope images show representative diatom species. Fully labeled phylogram and chronogram figures are available as SI Appendix, Figs. S5 and S6.

Fig. 3.

The rapid radiations of polar centric and raphid pennate diatoms were marked by low gene concordance and low phylogenetic signal. (A) Species phylogeny with branches colored by the amount of data necessary to resolve each branch across simulated datasets. Most branches were consistently resolved with <200 loci, but some were unresolvable with 1,290 loci. (B) Sequence saturation across 1,290 loci, calculated as the slope of a regression between pairwise patristic distances and uncorrected genetic distances. (C) Gene concordance factor for each node in the species tree. Nodes with open circles or diamonds correspond to identically marked nodes in panel (A); gray bars represent 95% CI around age estimates; the red line indicates the fit of a generalized additive model. (D) Average distance (discordance) between the species tree and empirical gene trees ($\overline{\mathrm{x}}=0.48$) far exceed expected levels based on coalescent simulations across a range of plausible values for theta, parameterized with values from diatoms (Fragilariopsis [1] and Phaeodactylum [3]), green algae (Chlamydomonas [2]), and coccolithophores (Gephyrocapsa [4]). (E) Average discordance between the species tree and empirical gene trees ($\overline{\mathrm{x}}=0.48$) far exceeded discordance levels with gene trees simulated across a range of realistic values for effective population size and generation time.

Fig. 4.

Species diversification and turnover are lowest in radial centric diatoms and highest in pennate diatoms. (A) Three pulses of net diversification (speciation minus extinction) occurred in marked lineages. (B) Net turnover (speciation plus extinction) was also highest in Bacillariales and Naviculales but was sensitive to tree topology in Thalassiosirales. The analysis was run on 100 bootstrap trees that captured topological uncertainty within polar centrics and raphid pennates.