Previously, we discovered that DINE-1 is highly abundant and appears to have experienced a recent transpositional burst in the lineage leading to D. yakuba 23. Using the D. yakuba DINE-1 consensus sequence, we searched using BLAST for related sequences in 11 other sequenced genomes of Drosophila (Figure 1). This initial screen suggested that all 12 Drosophila species contain hundreds to thousands of copies of DINE-1-related sequences. To infer the structure of DINE-1 in each species, we manually aligned 50 sequences with the highest BLAST scores from each species. Then we aligned together these sequences from all the species. This analysis revealed that DINE-1-related sequences from the ten newly analyzed Drosophila species share a number of structural similarities with DINE-1 from D. yakuba and D. melanogaster (Figure 2; Table 1; also see Additional data file 1).

We previously defined D. yakuba DINE-1 as beginning at its 5' inverted repeat, based on the assumption that this sequence is a TIR 23. We have here placed this repeat one nucleotide internal to D. yakuba DINE-1. This change in based on the recent designation of DINE-1 as a Helitron element 2428. Our designation of the boundaries of DINE-1 differs by one nucleotide from that in 28, based on our analysis presented below of polymorphic insertions in D. yakuba (Figure 3). The precise boundaries of DINE-1 in each species are difficult to determine because of their preference for inserting in T-rich regions. We have annotated the sequences in Figure 2 in order to maximize similarity to D. yakuba DINE-1. Based on this alignment the 5' inverted repeat ranges from being terminal to 2 bp internal; in all species the corresponding 3' inverted repeat is clearly internal. We thus refer to these repeats as subterminal inverted repeats (subTIRs).

The common features of DINE-1 from all species include: 13 bp subTIRs (the exact location of such repeats differs by 1-2 nucleotides among species); a partial inverted repeat next to the 5' subTIR; terminal regions that are relatively well-conserved within species, called blocks A and B; a GTCY-rich microsatellite repeat of variable length within block A; a variable central repeat region, which is responsible for most of the total length variation among elements; the lack of any significant open reading frames; and a propensity to insert between TT dinucleotides (discussed further below).

Our analysis revealed one novel feature not previously described for DINE-1, SGM or mini-me, namely a short hairpin stem-loop structure (with 7-11 nucleotide-long stems) located a few nucleotides downstream of the 3' subTIR (Figure 2). The sequence of the self-complementary stem differs among species, suggesting that compensatory mutations maintain its structure. This stem-loop may function as a terminator during rolling-circle replication (see Discussion).

Several of the features we characterized refine structural features inferred previously from SGM 26 and mini-me 27. TIRs from mini-me were reported to vary from 10-20 bp in length from different species, while our analysis identified 13 bp TIRs in all species. These differences likely reflect the fact that we have analyzed many more sequences. The 13 bp subTIRs from D. melanogaster and D. virilis that we describe contain the 10 and 11 bp sequences reported by Wilder and Hollocher 27. Likewise, the 17 bp TIR reported previously for D. subobscura mini-me contains the 13 bp subTIR reported here for D. pseudoobscura, with one internal base-pair difference. The partial inverted repeat flanking the 5' subTIR we identified is more variable than that reported based on two copies of mini-me, being a partial and/or interrupted repeat of the 5' subTIR in different species.

Wilder and Hollocher 27 also reported that mini-me elements from all species contain a highly conserved 33 bp core region. We find that this conserved core region actually extends over 90 bp, including the (TA)₄ repeat and additional sequences 5' to this repeat. These sequences partially overlap with the LS module described for SGM elements 26. The striking conservation of this core among the 12 species suggests that it is of functional significance for DINE-1 transposition.

We previously proposed that transposition of D. yakuba DINE-1 creates a dinucleotide (TT) TSD upon insertion 23. However, based on analysis of the genome sequence from a single strain this conclusion was not definitive because some copies did not have a TT dinucleotide flanking both sides. This could be due either to accumulated mutations after insertion or because the TT site preference is not absolute. We performed a preliminary test of this hypothesis by comparing the sequences of three polymorphic insertion sites of DINE-1 in multiple strains of D. yakuba 23. Here we extended this analysis with seven additional DINE-1 insertions that are polymorphic among different D. yakuba strains (see Materials and methods).

We found that all of these ten sites have a similar sequence structure (Figure 3). The interpretation of these data, however, depends on where precisely DINE-1 starts. Our previous interpretation of DINE-1 beginning at the 5' subTIR is consistent with insertions causing a TT TSD (Figure 3c). With the re-designation of the 5' end of DINE-1, these two nucleotides are instead part of the element, and DINE-1 would not create a TSD. The lack of a TSD is consistent with the proposal that DINE-1s are Helitrons (see below). Based on our designation of the DINE-1 boundaries, all D. yakuba insertions occur between the dinucleotide TT (Figure 3a). Using the DINE-1 boundaries from 28, the insertion site preference is more variable (Figure 3b). We then examined the sequences flanking the putative DINE-1s identified in the other Drosophila species. The majority (> 80%) of DINE-1s are flanked by TT dinucleotides. The conservation of this site preference in all 12 species, combined with the numerous other similarities described above, strongly suggests that each of these elements is in fact a species-specific DINE-1 and that they likely share a common mechanism of transposition.

Two pieces of evidence demonstrate that DINE-1 is highly homogeneous within 11 of the 12 species, with D. willistoni discussed below as being exceptional. First, we performed BLAST searches using the 90 bp sequence of the core region, which is conserved among all types of DINE-1s. By comparing the TIRs and block A sequences we found that each species contains only one type of DINE-1. Second, we searched for DINE-1s in one genome using the DINE-1 consensus sequences from other genomes as queries, and found only the same sets of sequences.

Among the 11 species (again excluding D. willistoni), there are 5 different subTIRs (Table 1). All five melanogaster subgroup species have the same subTIR sequence. D. ananassae has a unique subTIR, while the closely related species D. pseudoobscura and D. persimilis share the same subTIR, as do D. virilis and D. mojavensis. The DINE-1 subTIR from D. grimshawii shares 12/13 bp with D. virilis and D. mojavensis. The central repeat is the most diverse region of DINE-1 among species. Even species sharing the same subTIRs, such as D. virilis and D. mojavensis, have unrelated central repeat regions.

Analysis of D. willistoni gave uniquely different results. D. willistoni contains three different subtypes of DINE-1s, each with different subTIRs and different central repeat sequences (Figure 2). Phylogenetic evidence presented below suggests that they have at least two independent evolutionary origins.

DINE-1 is highly abundant in all 12 Drosophila species. It is difficult to determine an exact number because each species contains small and fragmented copies that cannot always be unambiguously identified as DINE-1s. We therefore used stringent criteria to identify DINE-1s in order to obtain a reliable comparison among species (Table 1). For example, this search identified 355 copies in D. melanogaster compared to previous analyses that suggested that D. melanogaster has approximately 1,000 copies 22. Using identical search criteria, we found vast differences in the copy number of DINE-1s among species, ranging from 334 in D. grimshawii to 6,297 in D. willistoni.

We identified similar numbers of DINE-1s in the D. melanogaster sister species D. simulans and D. sechellia compared to D. melanogaster. In contrast, more than ten-fold more copies were identified in D. yakuba. This high copy number is due to the large number of closely related copies in D. yakuba, and is consistent with previous work that suggested that DINE-1 has been inactive in D. melanogaster but underwent a recent transpositional burst in D. yakuba 23.

We therefore sought to determine whether other species with high copy number also show evidence of recent transpositional bursts. We used BLAST percent identity scores as an approximate method to estimate divergence among individual copies within species (Table 1; Figure 4). This method accurately recapitulates previous analyses for D. yakuba and D. melanogaster that were based on estimates of per-site divergence 23: DINE-1s from D. melanogaster have a broad peak of identities centered approximately around 90%, while D. yakuba shows a peak from approximately 96-100%, with a long tail of more diverged copies. These differences are highly significant (Mann-Whitney U test, two-tailed, p < 0.001).

D. simulans and D. sechellia show distributions similar to D. melanogaster (Figure 4; p > 0.05) and have average percent identities around 90-91% (Table 1). D. erecta has an average percent identity more similar to D. melanogaster than to D. yakuba; however, its distribution is significantly different from both species (p < 0.001). These data suggest that D. yakuba is the only melanogaster subgroup species that experienced a recent transpositional burst. D. grimshawii also has a distribution with very few copies of high similarity, and a similar copy number to D. melanogaster, suggesting that DINE-1 has not been recently active in this species. In contrast, DINE-1s from D. pseudoobscura, D. persimilis, and D. ananassae have average percent identities > 95% with distributions highly skewed toward young copies, suggesting recent transpositional bursts in these species. The distributions in D. pseudoobscura and D. persimilis are significantly different (p < 0.001; see Discussion). D. ananassae in particular stands out for having many copies identical to the consensus sequence (in the block A region). D. virilis and D. mojavensis also have substantial numbers of young copies but more broad distributions, suggesting the possibility that multiple rounds of transposition, silencing and reactivation may have occurred in these species.

D. willistoni has more than 1,000 copies of each of its three subtypes (Table 1), with subtype 1 and 3 having about twice as many copies as subtype 2. Each subtype has a peak near 100% identity, suggesting recent transpositional activity; however, their distributions are significantly different from each other (p < 0.001). Interestingly, these subtypes also have different phylogenetic patterns (see below).

These very different estimates of DINE-1 divergence within different species, and in particular our evidence for recent transpositional bursts, raises the question of whether DINE-1 may have undergone horizontal transmission into some Drosophila species. To understand the evolutionary dynamics of DINE-1s and their association with their host species, we analyzed the phylogenetic relationship of DINE-1 consensus sequences from the 12 Drosophila species (Additional data file 1) and compared it with the known phylogeny of Drosophila 29. Because of the rapid evolution in the central repeat region, reliable alignment for phylogenetic reconstruction could be obtained only for blocks A and B.

With the exceptions of D. willistoni and D. ananassae, the phylogenetic relationships of the DINE-1 sequences are, in general, consistent with the host species phylogeny (Figure 5). The grouping of DINE-1 in a separate clade containing D. ananassae and two of the three types from D. willistoni is surprising. The fact that this clade is an outgroup suggests that this result is not due to horizontal transfer from other Drosophila into these species. One possibility is that DINE-1 was horizontally transferred from non-Drosophila into these species. Alternatively, DINE-1 sequences resembling ancestral copies may have become reactivated in D. ananassae and D. willistoni. These sequences would be related to ancestral copies that were vertically inherited in the common ancestor of Drosophila.

In situ hybridizations to polytene chromosomes from several species using species-specific DINE-1 probes revealed strong signals in the heterochromatic chromocenter region, with additional hybridization observed along chromosome arms (data not shown). These non-chromocenter sites of hybridization led us ask whether DINE-1 insertions (whole or partial) are found in or near to protein-coding genes. SGM-related sequences were previously found in introns or adjacent to a number of D. melanogaster genes 26. We found hundreds of DINE-1 insertions in predicted introns in all species (Table 2; Additional data file 1). We also found many DINE-1 copies within 1 kb of genes, which could potentially be in either untranslated regions or in regulatory regions. These results suggest that DINE-1 has had a significant impact on gene structure evolution throughout the Drosophilidae. Few DINE-1 insertions were found in predicted coding sequences (CDSs). Strikingly, the largest number (15) was found in D. ananassae, a species that has a very high copy number of highly similar (young) DINE-1 copies (Table 1; Figure 4). This result suggests that recent transpositional activity of DINE-1 in D. ananassae has resulted in mildly deleterious insertions into coding regions that have not yet been removed by selection.

Kapitonov and Jurka 2428 recently proposed that DINE-1 is related to Helitron, a family of DNA-mediated TEs. They reported consensus sequences of autonomous and non-autonomous copies of Helitron in D. yakuba and D. virilis. The non-autonomous consensus sequences are closely similar to our consensus sequences reported here. The consensus autonomous copies have an open reading frame (ORF) encoding the RepHel protein found in many other Helitrons, and sequences at each end similar to what we report here for DINE-1. These include block A at the 5' end and block B at the 3' end. We searched these two species using the RepHel portion of the autonomous consensus sequence as a query to determine whether these species contain potentially active copies. Among the top ten hits in D. yakuba, none have a fully intact RepHel ORF. Three copies have DINE-1 sequences flanking both sides of the RepHel sequences; two of these have over 500 bp of DINE-1 sequence at each end while the third has only approximately 50-60 bp of DINE-1 sequence at each end. Six of the remaining copies have DINE-1 sequences flanking one side of the RepHel sequence, and the last hit has no flanking DINE-1 sequences. Among the top ten hits in D. virilis we again found no copies with a fully intact RepHel ORF. One copy has DINE-1 sequences flanking both sides and five copies have DINE-1 sequences flanking one side. Among the remaining copies, one is in a highly repetitive region and could not be further analyzed, and the remaining four copies have no flanking DINE-1 sequences. We conclude that D. yakuba and D. virilis are unlikely to contain currently active autonomous Helitrons.

DINE-1 has many features characteristic of MITEs - small size, lack of coding potential, high copy number, and frequent association with genes. On the other hand, most MITEs have TIRs, which are presumably sites of transposase binding. A few MITE-like elements have been discovered that have subTIRs rather than TIRs but their corresponding autonomous elements have not been identified 373839. DINE-1 has inverted repeats and their conservation in structure despite ongoing changes in primary sequence argues strongly that they are of functional importance. We have placed the 5' inverted repeat 0-2 nucleotides internal to the end of DINE-1 in different species. Under this annotation, D. yakuba DINE-1 insertions would not cause a TSD (Figure 3). If the true 5' end of DINE-1 instead corresponds to the 5' inverted repeat, then D. yakuba DINE-1 insertion would cause a 2 bp TSD, as seen in other MITEs. The 3' inverted repeat, however, is clearly subterminal, which would be unusual for a MITE element.

Kapitonov and Jurka 2428 have recently proposed that DINE-1 is instead a non-autonomous Helitron element. They noted that DINE-1 has a number of features unusual for Helitrons. One was the absence of a short hairpin or palindrome at the 3' end, which is thought to function as a replication terminator. We have identified here a 3' hairpin structure in all 12 species that may fulfill this function. A number of unusual features remain. Foremost are the termini. Helitrons do not contain TIRs but instead have highly conserved 5' TC or 3' CTRR sequences. In contrast, DINE-1 lacks these short termini sequences but instead contains conserved subTIRs. The presence of relatively long blocks of conserved sequence between non-autonomous DINE-1 and the proposed autonomous copies also contrasts with other species. For example, bats contain several families of very high copy number non-autonomous Helitrons, which differ almost entirely from their autonomous master copies other than at their di- and tetra-nucleotide termini 40.

The most decisive evidence favoring the Helitron hypothesis is the association of DINE-1 elements with non-functional but recognizable partial ORFs of the RepHel protein in D. yakuba and D. virilis, making these copies the candidate autonomous elements responsible for the recent transpositional bursts of non-autonomous DINE-1s in these species. Considering some of the unusual features mentioned above, it will be of great interest to investigate experimentally the mechanism of DINE-1 transposition.

From our previous study 23, we found that D. melanogaster and D. yakuba contain structurally similar types of DINE-1s. The species differed significantly, however, in their distributions of sequence divergence and the chromosomal location of their DINE-1 copies. D. yakuba contains many similar copies, and these apparently younger copies have a higher relative frequency in euchromatic regions compared to older, more diverged copies. We hypothesized that DINE-1s in D. melanogaster and D. yakuba derive from a common ancestor that existed before the divergence of the melanogaster subgroup species. This hypothesis was tested here by our identification of DINE-1 from three other species of the melanogaster subgroup, D. simulans, D. sechellia and D. erecta. All five subgroup species share the same TIRs, core, central repeat unit, and 3' end stem-loop sequences. DINE-1s from the three newly characterized species have similar copy numbers and distributions of sequence divergence, an observation consistent with the hypothesis that DINE-1 was active and then silenced in the common ancestor of the melanogaster subgroup. D. yakuba is the only species showing evidence of a second, recent transpositional burst. We did find that DINE-1s from D. erecta have a different sequence in the region joining the central repeat to block B, suggesting that this is the most rapidly evolving region of DINE-1.

Our analysis reveals that several species outside the melanogaster subgroup have distributions of DINE-1 identity similar to that described above for D. yakuba, suggesting that DINE-1 has undergone multiple, independent transpositional bursts. D. ananassae and D. willistoni show the strongest evidence, with distributions skewed toward 100% (Figure 4). D. virilis has a somewhat broader distribution, with many similar copies suggestive of recent transpositional activity. D. mojavensis shows a broad distribution that is suggestive of multiple rounds of transposition and silencing at different times.

D. pseudoobscura and D. persimilis have distributions with peaks around 98-99% identity. These species diverged less than one million years ago 41, which might suggest that the similarly high identity in both species reflects activity of DINE-1 before or during their speciation. However, the distributions are significantly different, with D. pseudoobscura retaining proportionally more copies of high divergence. One possible explanation is that DINE-1 remained active more recently in D. persimilis. Alternatively, the strength of selection against older copies may differ between the species.

The discovery of multiple and relatively distant species each showing evidence for recently active DINE-1 copies raises the question of whether this element has been transmitted vertically or horizontally. The phylogenetic relationship among different DINE-1s (Figure 5) suggests vertical inheritance, with transpositional bursts resulting from existing copies escaping from host suppression. DINE-1 from D. ananassae and two subtypes from D. willistoni give a pattern discordant from the accepted species phylogeny but this pattern is also not consistent with a simple model of horizontal transfer among Drosophila species. Instead, we suggest that the phylogenetic pattern is likely to reflect reactivation of a related ancestral element in both D. ananassae and D. willistoni. Our analysis is necessarily limited by the relatively short sequences available for analysis. We suggest that further phylogenetic analysis of the autonomous elements from each species will help to further understand the evolution of DINE-1. Nevertheless, the combination of our phylogenetic analysis and the divergence data indicates that the activity of DINE-1 is extremely dynamic. The activation and suppression of the element seems to have evolved rapidly and repeatedly in multiple lineages leading to the 12 species.

Similar dynamics of transposition and suppression are found in LINEs and SINEs of mammalian genomes 424344 and MITEs in plant genomes (see review in 14). Some insertions of MITEs in plants have been shown to affect gene regulation 4546. We have found that DINE-1 insertions are frequently found in the flanking regions and introns of genes, suggesting that some copies may also influence gene regulation (Table 2).

Highly abundant interspersed repetitive sequences can also serve as targets for ectopic recombination. Such recombination may be deleterious by promoting genome instability 4748, but may also catalyze structural evolution of existing genes and contribute to new gene formation 49. DINE-1 is a candidate for causing analogous phenomenon in Drosophila. The testis-expressed gene hydra is one well-characterized example 50. hydra exists only in the melanogaster subgroup, and its exon 1 has undergone multiple independent duplications. Many of these duplicated exon 1s are flanked by DINE-1 insertions, which suggests that DINE-1 may have facilitated some of these duplications by providing homologous target sequences for unequal crossing over. Given its high abundance and evidence for multiple rounds of transpositional activity, DINE-1 has clearly had a significant impact on Drosophila genome evolution, and we suggest that other examples of gene structural evolution associated with DINE-1 will be found among these species.

DINE-1 can also be a valuable system for studying rates and patterns of mutations. One can study de novo mutations in species that have had recent transpositional bursts by comparing the sequence variation among young, recently inserted DINE-1 copies. One can also use DINE-1 to examine substitution patterns between species. Previous comparative analysis of the chromosome distribution of DINE-1 in D. yakuba and D. melanogaster suggests that most new insertions are eliminated from the genome by negative selection 23. Old copies that remain are thus likely to be evolving neutrally. One could therefore identify orthologous insertions between D. melanogaster and D. simulans, whose insertions must predate the divergence of these species, in order to infer the substitution pattern along lineages leading to both species. The ability to perform similar studies in multiple Drosophila species will allow unprecedented power for determining whether and how patterns of mutations vary in different lineages.

Using the D. yakuba DINE-1 consensus sequence as a query, we searched for DINE-1-related sequences in all 12 Drosophila genome databases (from Comparative Assembly Freeze 1 (CAF1) 51) using BLAST with the default setting of the parameters (Figure 1). Note that D. persimilis and D. sechellia were sequenced at only approximately three- to four-fold coverage and, thus, are incomplete. We retrieved the 50 copies of DINE-1 with the lowest E-value in each species, aligned them, and derived a consensus sequence for each species (Additional data file 1).

We then BLAST-searched each genome using part of the consensus sequences of DINE-1 (5' end to end of core sequence) from its own species, using the default settings of the program. All BLAST hits greater than 100 bp were retained. The frequencies of percent identity between the query sequence and all hits were plotted for each species.

Sequences were aligned using ClustalW 52 with the default parameter settings. Alignments were further improved by manual adjustment. Inferred phylogenetic trees of the species consensus of DINE-1 were constructed using the neighbor-joining method with bootstrap resampling (500 replicates) using MEGA 3.0 53.

The UCSC Genome Browser Gateway 54 was used to obtain locations of DINE-1 in the annotated genomes, with the exception of D. willistoni. DNA sequences were retrieved using the Genes and Gene Prediction tracks (track setting: Other RefSeq) and grouped into the following categories: category 1, 1,000 bp upstream of CDS; category 2, CDS; category 3, introns; and category 4, 1,000 bp downstream of CDS. Categories 1 and 4 were then merged into a single class of flanking sequences. We then performed BLAST search to each of these three classes of sequences, using the block A regions of the DINE-1 consensus of each species. Only hits longer than 40 bp with an E-value lower than 10^-5 were included.

From the D. yakuba genome database, seven sites of DINE-1 insertion with sequence similarity > 97% to the D. yakuba DINE-1 consensus were chosen for analysis. PCR primers complementary to the 100 bp flanking sequence of each site were designed (Additional data file 2). A total of ten lines of D. yakuba, including nine Cy lines from a natural population (gift from Dr Peter Andolfatto at UCSD) and the strain Tai18E2, which was used for whole-genome sequencing, were checked for the presence of DINE-1 insertions at each site. Genomic DNA was phenol-chloroform extracted from 20-30 flies per line followed by ethanol precipitation. The program for PCR reaction was: 94°C for 5 minutes, followed by 30 cycles of 94°C (30s), 60°C (30s), and 70°C (1 minute), and extension at 70°C for 7 minutes. For lines not containing the DINE-1 insertion, PCR products were directly sequenced using ABI BigDye (Applied Biosystems, Foster City, CA, USA) technologies.