BEL/Pao retrotransposons in metazoan genomes

Abstract

Background: Long terminal repeat (LTR) retrotransposons are a widespread kind of transposable element present in eukaryotic genomes. They are a major factor in genome evolution due to their ability to create large scale mutations and genome rearrangements. Compared to other transposable elements, little attention has been paid to elements belonging to the metazoan BEL/Pao subclass of LTR retrotransposons. No comprehensive characterization of these elements is available so far. The aim of this study was to describe all BEL/Pao elements in a set of 62 sequenced metazoan genomes, and to analyze their phylogenetic relationship.

Results: We identified a total of 7,861 BEL/Pao elements in 53 of our 62 study genomes. We identified BEL/Pao elements in 20 genomes where such elements had not been found so far. Our analysis shows that BEL/Pao elements are the second-most abundant class of LTR retrotransposons in the genomes we study, more abundant than Ty1/Copia elements, and second only to Ty3/Gypsy elements. They occur in multiple phyla, including basal metazoan phyla, suggesting that BEL/Pao elements arose early in animal evolution. We confirm findings from previous studies that BEL/Pao elements do not occur in mammals. The elements we found can be grouped into more than 1725 families, 1623 of which are new, previously unknown families. These families fall into seven superfamilies, only five of which have been characterized so far. One new superfamily is a major subdivision of the Pao superfamily which we propose to call Dan, because it is restricted to the genome of the zebrafish Danio rerio. The other new superfamily comprises 83 elements and is restricted to lower aquatic eumetazoans. We propose to call this superfamily Flow. BEL/Pao elements do not show any signs of recent horizontal gene transfer between distantly related species.

Conclusions: In sum, our analysis identifies thousands of new BEL/Pao elements and provides new insights into their distribution, abundance, and evolution.

Figure 1

Overview over analyzed genome sequences and their taxonomic classification. The names of 62 non-mammalian species whose genomes we analyzed are grouped by phylum. 11 additional mammalian genomes we analyzed are summarized as "11 Mammals". Next to each species, the number of BEL/Pao elements we identified is shown. If we were not able to identify any element in one genome, the genome name is shown in red. Genome sequences where BEL/Pao elements had been already identified previously are marked with an asterix (*). For completeness we list seven additional species where no complete genome sequence was available but where BEL/Pao elements had been identified previously (shown in parentheses). A: Abe et al. (2001) [20], C1: Cook et al. (2000) [13], C2: Copeland et al. (2005) [16], J: Jurka and Kohany (2010) [21], L: Llorens et al. (2008) [23], S: Steinemann and Steinemann (1997) [19].

Figure 2

Separation of LTR elements in different classes for each genome. For each genome that contains BEL/Pao elements, we show the abundance of BEL/Pao, Ty1/Copia, Ty3/Gypsy and DIRS elements, based on our classification. The horizontal axis indicates the relative abundance of each element class in each genome, normalized to the interval (0,1). The difference in the length of each bar to the frequency of one reflects unclassified transposable elements. Their frequency is quite large in some genomes. The three most abundant phyla in our data set (arthropods, nematodes, chordates) are labeled.

Figure 3

Markov clustering of D. melanogaster BEL/Pao elements. We clustered all 178 BEL/Pao elements from D. melanogaster into ten families based on their sequence similarity using the MCL algorithm [27] from within BioLayout [46]. We refer to these families as MCL families. A node in the graph represents one element. Edges represent nucleotide sequence similarity between two elements (see Methods for details). Elements clustered into the same family are shown in the same color. The absence of an edge between two elements, indicates that the elements do not share sufficiently high sequence similarity over at least 500 bp (see Methods). We compared the elements of the MCL families to previous annotation of these elements in the D. melanogaster genome and in Repbase Update. Each MCL family is labeled with two names separated by a slash. The left name is from the Drosophila genome annotation, the right name is from Repbase Update. Dashes '-' indicate that a family has not been previously annotated. Note that one family has two names: 3S18 in the genome annotation and BEL in Repbase Update. Elements from family d) were previously annotated as belonging to two different families.

Figure 4

Phylogenetic tree of BEL/Pao families. The phylogenetic tree is based on the concatenated amino acid sequence of the three protein domains protease, reverse transcriptase, and integrase. The tree is based on 893 transposable element families for which we could construct a consensus sequence for all three domains. In addition to data from these families, we used in this analysis all BEL/Pao elements from Repbase Update [25] where we could identify all three domains (total of 92 elements). Some of these Repbase Update elements are from species whose genomes we did not analyze here. Furthermore, we included 16 elements from the Gypsy element Database GyDB [23], in order to associate clades in the tree to previously described subclades of the BEL/Pao element class. A list of these elements is found in additional file 2. Different colors indicate major clades. A) Same phylogenetic tree as in A) but with major clades collapsed into triangles indicating superfamily divergence and statistical support values for the branches. We constructed the tree using PhyML with an approximate likelihood ratio test to estimate the statistical support of the tree topology [31]. This statistical support is indicated by the numbers at the branches which range from 0 (least significant) to 1 (highly significant). All branches leading to the major clades have very high support. C) Same phylogenetic tree as in A) but without divergence triangles, and branches are reorganized to facilitate comparison with trees in D) and E). D) and E) Phylogenetic relationship of BEL/Pao clades as described by Copeland et al. [16] and Llorens et al [23], respectively. High resolution tree and graphic files are available from the authors upon request.