The evolution of vertebrate tetraspanins: gene loss, retention, and massive positive selection after whole genome duplications

Abstract

Background: The vertebrate tetraspanin family has many features which make it suitable for preserving the imprint of ancient sequence evolution and amenable for phylogenomic analysis. So we believe that an in-depth analysis of the tetraspanin evolution not only provides more complete understanding of tetraspanin biology, but offers new insights into the influence of the two rounds of whole genome duplication (2R-WGD) at the origin of vertebrates.

Results: A detailed phylogeny of vertebrate tetraspanins was constructed by using multiple lines of information, including sequence-based phylogenetics, key structural features, intron configuration and genomic synteny. In particular, a total of 38 modern tetraspanin ortholog lineages in bony vertebrates have been identified and subsequently classified into 17 ancestral lineages existing before 2R-WGD. Based on this phylogeny, we found that the ohnolog retention rate of tetraspanins after 2R-WGD was three times as the average (a rate similar to those of transcription factors and protein kinases). This high rate didn't increase the tetrapanin family size, but changed the family composition, possibly by displacing vertebrate-specific gene lineages with the lineages conserved across deuterostomes. We also found that the period from 2R-WGD to recent time is controlled by gene losses. Meanwhile, positive selection has been detected on 80% of the branches right after 2R-WGDs, which declines significantly on both magnitude and extensity on the following speciation branches. Notably, the loss of mammalian RDS2 is accompanied by strong positive selection on mammalian ROM1, possibly due to gene loss-induced compensatory evolution.

Conclusions: First, different from transcription factors and kinases, high duplicate retention rate after 2R-WGD didn't increase the tetraspanin family size but just reshaped the family composition. Second, the evolution of tetraspanins right after 2R-WGD had been impacted by a massive wave of gene loss and positive selection on coding sequences. Third, the lingering effect of 2R-WGD on tetraspanin gene loss and positive selection might last for 300-400 million years.

Figure 1

The functions and structures of vertebrate tetraspanins. (A) Schematic of some tetraspanin functions. (B) The tetraspanin overall structure. C1~C8 indicate the conserved cysteines; 100% conserved cysteines are labeled; 'CCG' is the so-called tetraspanin signature. (C) The LEL structure of tetraspanins.

Figure 2

The protein phylogenetic (ME) tree of invertebrate deuterostome tetraspanins. Proteins marked with a rectangle are members of the TSPAN4 gene cluster. This tree includes all 39 amphioxus sequences but omitted 9 C. instestinalis sequences and 5 S. purpuratus. Readers interested in the omitted genes are referred to Figure S1-4.

Figure 3

The chordate TSPAN4 gene clusters in amphioxus and humans. Triangles show the gene direction. Picture is not drawn to scale. The relationships between four human paralogous region are shown using a tree pattern. An inferred TSPAN4 cluster in the pre-WGD vertebrate ancestor is also provided.

Figure 4

The protein phylogenetic (ME) trees of all 17 ancestral (pre-WGD) vertebrate tetraspanin lineages. The bony vertebrate ortholog lineages and the ohnolog patterns are highlighted as shown in panel A. Red-filled rectangles are used to mark each ancestral lineage in panel J and K. #1, a distant tetraspanin, only found in ray-finned fish otocephala; #2, found in ray-finned fishes, xenopus and reptiles, probably an independent duplicate of TSPAN4; #3, a tetraspanin pseudogene from mouse; #4, a divergent tetraspanin found in all bony vertebrates, originated by retrotransposition; #5, mammalian ROM1, the true ortholog of bony vertebrate ROM1, but too divergent to cluster with other ROM1; #6, mammal TSPAN16, the true ortholog of teleost TSPAN16, but too divergent to cluster with other TSPAN16; #7, a divergent lineage, its position is not determined; #8 and #9, no synteny shared between TSPAN33 and TSPAN33-like, but since they were separated before the radiation of bony vertebrates, here we treated them as an ohnolog pair; #10, found in reptiles and mammals, as a result of tandem duplication of CD81, becoming too divergent to cluster with CD81; #11, only found in ray-finned fish otocephala, as a result of independent duplication of CD9, becoming too divergent to cluster with CD9; #12, has weak support for its clustering with tetrapod CD37 and sharing no synteny, so its identity is questionable.

Figure 5

(A) Schematic representation of the tree topology. Three types of branches (in thick lines) are selected for branch-site model tests for positive selection. (B) The one-ratio tree of RDS. The branches selected for tests are marked with "#". (C-F) RDS gene trees inferred using branch-specific models, with the mammalian ROM1 branch as the foreground branch. (C) Synonymous substitution (d_S) tree of transmembrane region (TM). (D) Nonsynonymous substitution (d_N) tree of TMs. (E) d_S tree of the non-TM regions. (F) d_N tree of the non-TM region.