Molecular evolution and tempo of amplification of human LINE-1 retrotransposons since the origin of primates

Abstract

We investigated the evolution of the families of LINE-1 (L1) retrotransposons that have amplified in the human lineage since the origin of primates. We identified two phases in the evolution of L1. From approximately 70 million years ago (Mya) until approximately 40 Mya, three distinct L1 lineages were simultaneously active in the genome of ancestral primates. In contrast, during the last 40 million years (Myr), i.e., during the evolution of anthropoid primates, a single lineage of families has evolved and amplified. We found that novel (i.e., unrelated) regulatory regions (5'UTR) have been frequently recruited during the evolution of L1, whereas the two open-reading frames (ORF1 and ORF2) have remained relatively conserved. We found that L1 families coexisted and formed independently evolving L1 lineages only when they had different 5'UTRs. We propose that L1 families with different 5'UTR can coexist because they don't rely on the same host-encoded factors for their transcription and therefore do not compete with each other. The most prolific L1 families (families L1PA8 to L1PA3) amplified between 40 and 12 Mya. This period of high activity corresponds to an episode of adaptive evolution in a segment of ORF1. The correlation between the high activity of L1 families and adaptive evolution could result from the coevolution of L1 and a host-encoded repressor of L1 activity.

Figure 1.

(A) Structure of a modern human full-length element. A full-length element is 6 Kb long and contains a 5′ untranslated region (5′UTR), two open-reading frames (ORFI and ORFII), and a 3′UTR. The 5′UTR has a regulatory function (Swergold 1990; Minakami et al. 1992). ORFI encodes a protein with nucleic acid-binding properties that can also act as a nucleic acid chaperone (Martin et al. 2000; Martin and Bushman 2001). ORFIp also contains a coiled-coil domain (C-C) that mediates interaction of ORFIp with itself (Martin et al. 2000). ORFII encodes a protein with endonuclease (EN) (Feng et al. 1996) and reverse transcriptase (RT) activity (Mathias et al. 1991). The 3′UTR contains a conserved poly-purine tract (Howell and Usdin 1997). Genomic copies of L1 are typically flanked by an A-rich tail at their 3′ end. (B) Functional motifs in the 5′UTR of a modern L1 element (L1PA1). The first 100 bp (100bp) of the 5′UTR was shown to be critical for transcription (Swergold 1990). The 5′UTR contains a YY1 binding site that plays an important role in transcription initiation (Athanikar et al. 2004), a functional RUNX3 binding site (Yang et al. 2003), two functional SRY-related transcription factor binding sites (SRY-A and SRY-B) (Tchenio et al. 2000), and two cellular factor-binding motifs (B and C) (Minakami et al. 1992). The 5′UTR also contains an antisense promoter (AS) between positions 400 and 600 that can drive transcription of adjacent cellular genes (Speek 2001).

Figure 2.

Phylogeny of L1 consensus sequences. This maximum likelihood tree is based on the consensus sequences of the ORF1 and ORF2 of 27 L1 families. The numbers above the nodes indicate the percentages of time the labeled node was present in 1000 bootstrap replicates of the data. Asterisks indicate branches on which the free-ratio model assigned estimates of ω > 1.

Figure 3.

L1 families of similar ages have unrelated 5′UTR sequences. This dot plot analysis, based on the first 1000 bp of the 5′UTR, shows that the 5′UTR of family L1PA13A is unrelated to the 5′UTR of families L1PA14 (A) and L1PA13B (B). In contrast, families L1PA14 and L1PA13B (C) and families L1PA13A and L1PA15 (D) are relatively similar.

Figure 4.

Phylogeny of L1 genomic sequences. These trees were built using the neighbor-joining method based on Kimura's two-parameter distances. These two trees were built using the same full-length elements but using different regions of L1. Tree A was built using the 3′ end of the elements (2000 bp) and tree B was built using the 5′ end of the elements (300 bp). Only bootstrap values > 80 are shown. The gray boxes indicate that the L1PA8A family is forming a distinct lineage (see text).

Figure 5.

L1 lineages have frequently recruited novel 5′UTR sequences. This scenario was inferred from the phylogenetic tree in Figure 2. An arrow indicates the acquisition of a new 5′UTR. The 5′UTR sequences are drawn proportionally to their size. As the L1PA14 family is nested within the L1PA13A family (data not shown), it is likely that the modern type of 5′UTR was probably recruited independently by the L1PA14 and L1PA13B families.