Non-random sharing of Plantae genes

Abstract

The power of eukaryote genomics relies strongly on taxon sampling. This point was underlined in a recent analysis of red algal genome evolution in which we tested the Plantae hypothesis that posits the monophyly of red, green (including plants), and glaucophyte algae. The inclusion of novel genome data from two mesophilic red algae enabled us to robustly demonstrate the sisterhood of red and green algae in the tree of life. Perhaps more exciting was the finding that >1800 putative genes in the unicellular red alga Porphyridium cruentum showed evidence of gene-sharing with diverse lineages of eukaryotes and prokaryotes. Here we assessed the correlation between the putative functions of these shared genes and their susceptibility to transfer. It turns out that genes involved in complex interactive networks such as biological regulation and transcription/translation are less susceptible to endosymbiotic or horizontal gene transfer, when compared to genes with metabolic and transporter functions.

This article refers to:

The Plantae hypothesis posits that Rhodophyta (red algae; source of nori, agar and carrageenan), Viridiplantae (green algae and plants) and Glaucophyta (a small, poorly studied algal group that includes Cyanophora paradoxa) share a common ancestral origin.1,2 These taxa have played a pivotal role in the field of endosymbiosis because the Plantae share many traits associated with plastid function and photosynthesis,3–6 suggesting their ancestor captured the plastid (e.g., chloroplast) that has spread throughout the tree of life. Recent phylogenomic analyses of eukaryotes however, recover little7,8 or no9,10 support for Plantae monophyly. This may be explained by the fact that investigators have until now relied primarily on data from the highly reduced and specialized genome of the red algal extremophile Cyanidioschyzon merolae11 to represent this species-rich and ancient lineage comprised primarily of mesophilic taxa. Using phylogenomic analysis of >60,000 novel red algal genes from two mesophilic species, Porphyridium cruentum and Calliarthron tuberculosum, we recently found an emerging signal of red + green (RG) monophyly that was supported by ∼50% of the examined protein phylogenies.12 Interestingly, we also recovered strong evidence of reticulate evolution between the RG lineage and other phyla of eukaryotes and prokaryotes involving 1,808 protein phylogenies (non-parametric bootstrap ≥90%; terminal taxa ≥30, number of distinct phyla ≥3). This surprising result is likely explained by endosymbiotic or horizontal gene transfer (E/HGT) involving the red and/or green algal lineages,13,14 particularly within the context of plastid evolution.15,16 Little is known, however, about the functional bias (if any) of this substantial level of non-lineal transfer between eukaryotes. In comparison, the correlation between gene function and susceptibility to HGT has recently been described in prokaryotes.17

Even though the function of many P. cruentum genes is unknown (62.3% do not have a hit to publicly available gene data12), here we asked the question if functional bias exists among the 1,808 genes implicated in E/HGT, when compared to the total number (8,082) of annotated genes in the dataset. Adopting the approach of Chan et al.17 we postulate that putative functions over-represented in the E/HGT set are associated with genes that are more susceptible to gene sharing/transfer when compared to functions distributed across all annotated genes; i.e., the null model postulates no significant difference between the distribution of annotated functions in the E/HGT set and that across all genes. Figure 1 shows the comparative representation of functional categories between these two groups of genes for three principal groups of “biological process,” “cellular component,” and “molecular function” as assigned based on sequence similarity18 and information extracted from the Gene Ontology database (www.geneontology.org/).

Within the principal group of “biological process” (Fig. 1A), the categories metabolic process, cellular process and localization (p ≤ 0.01) are significantly over-represented in genes showing non-lineal sharing, whereas categories such as biological regulation, cellular component organization and developmental process are under-represented (p ≤ 0.01). This finding suggests that genes involved in anabolic or catabolic pathways could have been more susceptible to non-lineal transfer than the others, and that genes involved in the intricate network of cell development and organization are less likely to have undergone such transfer. Similarly within the principal group of “cellular component” (Fig. 1B), protein functions related to membrane-bounded organelle (e.g., plastid and mitochondrion), vesicle (p ≤ 0.01, respectively), and protein complex (p ≤ 0.05) are more susceptible to E/HGT than others, in contrast to those related to ribosomes, the cytoskeleton, and chromosomes (non-membrane-bounded organelle; p ≤ 0.01) or to structural constituents of a cell (cell part; p ≤ 0.05).

Interestingly, for the principal group of “molecular function” (Fig. 1C), categories related to transporter activities, i.e., transmembrane (p ≤ 0.01) and substrate-specific proteins (p ≤ 0.05), as well as proteins binding to metal cluster and tetrapyrrole (p ≤ 0.01, respectively) are over-represented in E/HGT-implicated genes compared to the overall dataset. These proteins are crucial for the transport of secondary metabolites and photosynthesis. Biomolecules that bind to metal cluster and tetrapyrrole include heme (resulting in hemoproteins; e.g., cytochrome) and chlorophyll, which are key components of the photosynthetic reaction complex. This observation suggests that algal genes encoding photosynthetic functions or transport among different cellular compartments are more likely to have a non-lineal gene history than expected under the null model.

Our results underline the notion that endosymbiosis has a significant impact both on the quantity (extensive E/HGT) and quality (function) of gene transfer between algal lineages. We found functional categories related to the transcription/translation apparatus (e.g., protein binding) and structural constituents of the ribosome (p ≤ 0.01 respectively), among other enzymatic activities, are under-represented in the 1,808 genes compared to the overall dataset. The complexity hypothesis19,20 postulates that genes encoding “informational” proteins that are involved in complex interactive networks implicating multi-protein complexes (e.g., processes related to transcription and translation) are less likely to be horizontally transferred than genes encoding “operational” proteins. A recent study21 suggests that the connectivity of protein-protein interaction networks rather than biological function better explains the susceptibility of a gene to undergo HGT. This aspect remains to be rigorously tested, because the complexity of protein-protein interactions is necessarily related to overall biological function. Our results do not validate nor did we intend to test the complexity hypothesis in this study, but our findings are consistent with this working hypothesis. In conclusion, we demonstrate potential functional biases in non-lineal gene sharing in Plantae lineages that are associated with endosymbiosis, e.g., towards functions related to the establishment of cell organelles. The addition of genome data from understudied groups such as mesophilic red algae promises to further advance understanding of eukaryote evolution in general.

Figures and Tables

Figure 1 Annotated functional categories of proteins encoded by Porphyridium cruentum. The charts are organized based on three principal categories in the Gene Ontology database: (A) biological process, (B) cellular component and (C) molecular function. The blue bars show the proportional representation of the functional categories (in percentage) across the overall dataset (8,082 annotated genes), whereas the red bars show the proportional representation of the same functional categories (in percentage) among the 1,808 genes implicated in E/HGT. The categories are numbered independently for each panel, as shown in the legends. Significance of over- or under-representation is represented by single (p ≤ 0.05) and double asterisks (p ≤ 0.01), as inferred based on the approach of Chan et al.17

Acknowledgments

This research was supported by grants from the National Science Foundation (EF 08-27023, DEB 09-36884). We acknowledge the support of our colleagues in the RedToL project (lead PI, Hwan Su Yoon).

Addendum to: