Widespread horizontal transfer of retrotransposons

Results and Discussion

To determine the sequence conservation of BovB across taxa and examine the evidence for HT, we identified BovB sequences in all publicly available genomes and in several low coverage genomic survey (454 shotgun) sequences using RepBase (12) consensus sequences for BovB as BLAST (13) queries (SI Appendix, Tables S2–S4). The BovB sequences available in Repbase (12) include a sequence extracted from the horn-nosed viper (Vipera ammodytes) BovB VA, that contains Chicken Repeat 1 (CR1) elements on both the 3′ and 5′ ends (SI Appendix, Fig. S4). This means that early during its colonization of the squamates, it somehow acquired the CR1 sequences now present at both ends. We used a trimmed version of BovB VA in our sequence similarity searches, noting that use of the untrimmed RepBase BovB VA sequence leads to false discovery in birds and nonsquamate reptiles, as reported in turtles and the tuatara (10) (basal to squamates), which have an abundance of CR1 elements.

Other squamates may have CR1 fragments on their BovB consensus sequences too. However, due to the abundance of CR1 in the squamate genomes and the low coverage reads from which the squamate BovBs were built, all CR1 fragments had to be removed to reliably assemble a BovB consensus from those species. Hence additional sequencing in a greater range of reptiles would be required to determine when CR1 ends were acquired by the squamate BovB lineage. Interestingly the BovB sequences for the python and the copperhead that were extracted from RepBase do not have the CR1-like ends that are present in BovB VA. This could be due to a different repeat building process used by Castoe et al. (14).

Our searches revealed that BovB is highly abundant in cow, sheep, and Afrotheria (basal mammals), with significant portions of these genomes resulting from BovB contribution (Table 1). BovB is thus capable of significantly altering genome structure and therefore function. BovB sequences contribute to more than 1% of anole, opossum, platypus, and wallaby genomes but only exist as relatively few copies in the horse, sea urchin, silkworm, and zebrafish. BovB was not found in the tuatara, turtles, birds, or other mammals. BovB was also not found in mosquitos despite the presence of an RTE element (SI Appendix, Table S3).

Within the horse genome, just 31 regions were extracted by LASTZ (15) when searched for 80% coverage of the BovB query sequence. We checked to ensure that this was not contamination by searching for 5′-truncated BovB sequences in the genome, which we expected to find if the reverse transcription step of the copy-and-paste movement was truncated due to premature termination. We were able to find >100 5′-truncated BovB per chromosome, indicating that BovB has been undergoing transcription, reverse transcription, and insertion, but at a much more limited scale than in ruminants or afrotheria. Finally, the presence of horse-specific SINEs inserted into some of the full-length horse BovBs indicated that BovB has been present in the horse genome for some time (Fig. 1).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

SINEs inserted into BovB in the horse genome. This is a visual representation of 3 of 31 nearly full-length horse BovBs according to RepeatMasker (31) using the RepBase (12) horse repeat library. Mid-gray rectangles indicate masking as RTE-1 EC, which is the RepBase equivalent of the horse BovB consensus sequence we constructed. Square gray boxes represent the presence of horse ERE SINE sequences.

We constructed a consensus from the BovB sequences recovered from each species where possible and conducted phylogenetic analyses using both maximum likelihood (Fig. 2) and Bayesian methods (SI Appendix, Figs. S9 and S10). Both methodologies gave similar tree topologies varying only in the placement of the zebrafish, silkworm, and sea urchin sequences. Excluding these, the consensus sequences were resolved into two major clades of BovB (Fig. 2).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Phylogenetic tree of BovB sequences showing the distribution of BovB across taxa. Maximum likelihood tree built using FastTree (35) from the full-length BovB sequences extracted from full genome sequence and those constructed from low coverage reads. The sequences were aligned with MUSCLE (29) and processed by Gblocks (34) to limit the effect of indels, making an alignment that was 2858 bp long. Local support values are only shown for those nodes with support less than 0.9. Branch colors indicate important BovB clades: marsupials in purple, reptiles/ruminants/ticks in green, monotremes/Afrotheria/horse in orange, and the RTE clade in maroon, used to root the tree. Taxa showing BovB are colored taxonomically, with marsupials in purple, reptiles in green, ruminants in dark blue, arthropods in yellow, Afrotheria in red, monotremes in pink, horse in blue, zebrafish in gray, sea urchin in light blue, and silkworm in orange. The RTEs are in maroon.

The largest clade comprised BovB consensus sequences from the marsupials, ruminants, ticks, and all but one of the squamates examined. Whereas the marsupials robustly grouped together, the resolution within the clade was too low to allow analysis of HT between marsupials. However, as no nonmarsupials are present in this clade, we have concluded that it is likely that BovB was present in the common ancestor of marsupials and potentially no HT has occurred since the divergence of marsupials from other mammals. Analyses with additional taxa will be required to test this hypothesis.

The BovB sequences constructed for the two reptile tick species (Bothriocroton hydrosauri and Amblyomma limbatum) nested within the squamate clade. Although the two tick species were collected from the same host (Tiliqua rugosa), they contributed independent BovB sequences to this analysis, neither of which clustered with the BovB sequences from the host. Both species feed on a diverse range of squamates (16) and the potential exists for contamination from the nucleated red blood cells of the lizard in their gut. For this reason, A. limbatum tick legs were sequenced to remove the potential for contamination. Although contamination is a concern, and can come from various sources, we do not believe it affects our results. The DNA samples for 454 sequencing came from a number of different laboratories and samples were not extracted in one laboratory, or by one person. Pre-PCR and PCR steps were carried out in different laboratories to prevent contamination. Furthermore, the species where we have identified BovB were not amplified/sequenced together but were amplified/sequenced in conjunction with samples where BovB was not detected. If contamination were an issue one would expect the pattern of occurrence to be random, not lineage specific, e.g., all marsupials, reptiles. We describe our controls for false BovB hits in SI Appendix, section 1.7.3. We have also directly tested for contamination by PCR amplifying and sequencing BovB from a subset of critical taxa: horse, both tick species, and the Lord Howe Island Gecko (Christinus guentheri). We were able to validate BovB in these species using freshly extracted DNA from independent specimens that were not sourced from the laboratories where the original samples were obtained, and we describe our methods and report representative results in SI Appendix, section 3.6. Sequences of validation samples have been deposited with GenBank.

It is also important to note that the topology seen in the squamate BovB subtree is not the topology expected from a tree built from gene orthologs or fossil records (SI Appendix, Fig. S8). This indicates that BovB has been moving horizontally among the squamates as well as between them, ruminants, and marsupials.

The second major BovB clade includes monotremes, African mammals, the horse, and one species of gecko. The Lord Howe Island gecko appeared to have two subclasses of BovB during the consensus construction process, but only one subclass was deemed of sufficient quality to use in phylogenetic analysis. To get a suitable quality sequence for phylogenetic analysis of the other subclass, significantly more data would be required to build the other BovB subclass in this gecko. There is no suggestion of a vector at present and more widespread sequencing would be required to find a parasite or virus vector that would facilitate the HT of the BovB within this clade. The BovB from the African mammals displayed the relationship expected when building a tree from orthologous sequences, which implies that BovB was present in the common ancestor of Afrotheria and has not moved horizontally between African mammals since its incorporation in the ancestral afrotherian genome.

We compared the tree constructed from BovB sequences to the tree constructed from protein orthologs in OrthoDB: Database of Orthologous Groups (17), and TimeTree of Life data (18) using the program SPRIT (19), that estimates the number of required subtree prune and regrafts (SPR) to transform one tree into another. It is apparent from the representation of SPRIT output shown in Fig. 3 that nine SPR are required to explain the observed BovB-based topology. Each SPR corresponds to at least one HT event, therefore we conclude that at least nine interspecies HT events have occurred during the evolutionary history of BovB. This is significantly more than previous estimates of one or two (9, 20) and could increase with the inclusion of new taxa and higher quality data that refines the position of taxa on the BovB and protein ortholog trees.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Horizontal transfers. This is a representation of the least number of subtree prune and regrafts (SPR) required to turn the control tree built from protein orthologs (A), into the tree built from the BovB sequences (J) through intermediates B–I. The movement that corresponds to the SPR in the next tree is shown by the arrow and the SPR that made the current tree is shown in red. D, Danio rerio; Eq, Equus caballus; Mo, Monotremata; Ec, Echinops telfairi; Lo, Loxodonta africana; Pr, Procavia capensis; Ig, Iguanidae; Me, Metatheria; Ag, Agamidae; Ru, Ruminantia; Bt, Bothriocroton hydrosauri; Sc, Scincidae; Ge, Gekkonidae; Am, Amblyomma limbatum; Se, Serpentes; Bm, Bombyx mori; and St, Strongylocentrotus purpuratus.

BovB is capable of expanding within a diverse range of species including warm- and cold-blooded animals and shows a large variability in its accumulation of substitutions in different species; showing a low number of substitutions per site in the anole and a very high number in the opossum (Table 2).

The analysis of BovB HT revealed that ticks may have transferred DNA between snakes and lizards and into ruminants and marsupials. Although we cannot identify the exact tick species, it is known that species of Amblyomma and Bothriocroton infest mammals, marsupials, and monotremes, and that Amblyomma sp. are highly important parasites of domestic animals and man in Africa and America (21). Further work is needed to understand why BovB has been so successful at colonizing some genomes, for example the cow and elephant, and so unsuccessful in others, like the horse. In extreme cases such as the cow, almost a quarter of the genome is the result of BovB and derived SINE sequence retrotransposition, with one reported instance of exaptation into a protein-coding gene (22). The timing of HT for BovB is difficult to determine. HT in terrestrial animals could have occurred via a common mechanism/vector before the breakup of Gondwanaland 175–140 Mya (23). Alternatively, it could have occurred much later if migratory birds or insects were transfer partners. In this context it is worth noting that immature stages of Amblyomma sp. are found on wild birds (24). Resolution of these phylogeographic alternatives will have to await the availability of additional genome sequence data.

The frequent horizontal movement of BovB illustrates the significant impact HT has had on animal genomes; expansion of BovB in various lineages has contributed large amounts of sequence (and presumably structural variation) to the genomes of distantly related species. It is tempting to speculate that BovB is not the only retrotransposon to have jumped between species, and further investigation will be required to test this hypothesis. Despite public concern over the transfer of genetic material to create genetically modified organisms, it appears that Mother Nature has been quietly shuffling genomes for some time.