The number of full-length Bari sequences varied by species: 11 were found in D. melanogaster, two were found in both D. simulans and D. sechellia, seven in D. erecta and four in D. ananassae (see Table S2 in Additional File 2). The only sequence found in D. yakuba was a 215 bp fragment resembling the Bari of D. erecta. The ML reconstruction of evolutionary relationships among the full-length sequences is shown in Figure 1A and Additional file 2 Figure S2. The sequences of D. ananassae and D. erecta are clustered in well-supported monophyletic clades. Also well-supported is the clade grouping the sequences of the melanogaster complex (D. melanogaster, D. simulans and D. sechellia), albeit in a polytomic branch; however, the Bari sequences of D. ananassae cluster more closely to those of the melanogaster complex than to those of D. erecta, which is inconsistent with the species phylogeny (Additional File 1). The K2p distances further contribute to this incongruence, with the sequences of the melanogaster complex and D. ananassae being less divergent from each other (melanogaster complex vs. D. ananassae = 0.171 ± 0.012; melanogaster complex vs. D. erecta = 0.372 ± 0.024; D. ananassae vs. D. erecta = 0.380 ± 0.023; see Table S3 in Additional File 2).

Two processes could be responsible for the phylogenetic incongruence observed in the Bari phylogeny: recent invasion of the melanogaster complex by a Bari sequence from D. ananassae (or vice versa) or the existence of an ancestral polymorphism followed by stochastic inheritance. To distinguish between these two possibilities, we estimated Ks, the rate of synonymous substitutions per synonymous site, which provide a measure of divergence in neutral sites, and the time of divergence of the Bari sequences and of the host genes (i.e., ADH and GAPDH). If the incongruence in the phylogeny is due to differential fixation of Bari variants in the common ancestor and is evolving vertically, then the Ks values of Bari and of the host genes and their time of divergence should be equivalent; however, if the estimates for the transposable element sequences are significantly lower than those for the host genes, then Bari was likely transferred after the species divergence, through horizontal transfer or species hybridization. The sequences of D. sechellia and D. erecta were not utilized in this analysis because they contained large numbers of stop codons and small deletions. The Bari sequences of D. ananassae showed few premature stop codons, all of which were excluded from the alignment, so these sequences were used in the Ks estimate. The average Ks values of the Bari sequences and host genes were as follows: D. melanogaster vs. D. ananassae, Bari Ks = 0.409 ± 0.049 and host genes Ks = 0.428 ± 0.058; D. simulans vs. D. ananassae, Bari Ks = 0.409 ± 0.047 and host genes Ks = 0.422 ± 0.055 (see Tables S4 and S5 in Additional File 2). Using the estimated Ks and 0.011 substitutions per site per million years (My) 25 as the rate of synonymous substitution, the average time of divergence of Bari in these species was estimated at 17.68 My, and that of the host genes was 19.34 My. During this latter period, the lineages that gave rise to D. ananassae and the melanogaster subgroup were still evolving in Asia 10 , suggesting that the Bari sequences diverged from a common sequence in the common ancestor. These estimates suggest that the incongruence resulted from stochastic retention of the same Bari1-like variant by the ancestors of D. ananassae and of the melanogaster complex. The loss of parts of the TIRs followed, yielding long TIRs in D. ananassae and short TIRs in the melanogaster complex 23 .

The ML tree did not allow us to resolve the relationships among the Bari1 sequences within the melanogaster complex (D. melanogaster, D. simulans, D. sechellia) because the sequences are very similar and cluster within an unresolved branch. We therefore reconstructed the sequence relationships using a network tree. This approach can resolve relationships among sequences with low diversity and can thus be used to infer the origin of multiple copies from a unique sequence. In addition, the network reveals relationships between ancestral and derived sequences and introduces median vectors, which represent ancestral, lost or unsampled sequences 26 27 ; these relationships cannot be inferred from the classical phylogenies.

The network shows a second phylogenetic incongruence in the Bari1 sequences, revealing a closer relationship between the copies of D. simulans and D. melanogaster than either has to D. sechellia (Figure 1B). The two full-length sequences of D. simulans are directly related to a unique sequence of D. melanogaster, which is centrally positioned on the network and is the sequence from which all the other sequences of this species diverged. Moreover, all sequences of D. melanogaster are very similar (K2p = 0.0020 ± 0.0006), suggesting a recent origin. The age of these copies was estimated at ~ 40,000 y (Ks = 0.00086 ± 0.00084), and the longest time of divergence between the Bari1 sequences of D. melanogaster and that of D. simulans was estimated at ~ 196,900 y (Ks = 0.0004 ± 0.0004; average time = 32,800 y; shortest time = 0; see Table S4 in Additional File 2). Because the D. melanogaster and D. simulans lineages split from a common ancestor between 2 and 3 Mya 11 12 , the network topology and the divergence time of the Bari1 sequences in both species are inconsistent with the species phylogeny and the estimated divergence time. These data suggest a very recent transfer of Bari1 from D. simulans to D. melanogaster and further suggest that following this transfer, the sequence remained active and dispersed within the D. melanogaster genome, producing the similar copies that are observed today. The short branches and the presence of several similar copies in the D. melanogaster network, all derived from the same sequence, are evidence of transposition burst, a common process after the introduction of a new element into a naïve genome 20 , supporting the hypothesis of the transfer from D. simulans to D. melanogaster.

The similarity of the two D. simulans Bari1 sequences to that of D. melanogaster is high, with the sequences differing at only a few sites. To identify whether the proposed recent transfer is exclusive to the sequenced genomes, we sequenced a region of the transposase gene in different strains of both species and reconstructed their relationships (Figure 1C). The strains analyzed represented natural populations of different geographic origins: Africa, the ancestral site; Asia and Europe, continents that were first colonized by both species (ancient invaders); and Brazil, where colonization occurred relatively recently (recent invaders). The evolutionary reconstruction shows a sequence shared among all strains. This central sequence, which likely corresponds to the central sequence depicted in Figure 2B, could be the sequence transferred between the species. The sharing of this sequence among strains of different geographic origins (Africa, Asia, Europe and Brazil) suggests that transfer of Bari occurred before the global dispersal of D. melanogaster.

As with Bari, we found different numbers of full-length copies (with both LTRs) of the retrotransposon 412 in different species of the melanogaster group. The smallest numbers were found in D. simulans (2) and D. yakuba (2), followed by D. erecta (8), D. sechellia (11), D. ananassae (14) and D. melanogaster (27; see Table S5 in Additional File 3). The ML phylogeny based on the gag region shows monophyletic branches clustering with the sequences of D. ananassae, D. erecta and D. yakuba with high statistical support (Figure 2A and Additional file 3: Figure S3). Like the species, the D. ananassae 412 was the first to diverge. Next, the ancestral element of D. yakuba and D. erecta diverged, also mirroring the species divergence (Additional File 1); however, in the melanogaster complex, the 412 sequences did not form monophyletic groups within each species. Given the very recent divergence of D. melanogaster, D. simulans and D. sechellia (~2 and 3 Mya and 0.5 Mya, respectively 11 12 28 , the sequences of 412 could not have had time to coalesce, potentially explaining the unresolved relationships among them. Alternatively, the 412 sequence could have been exchanged between these species.

To resolve the phylogenetic relationships of 412 within the melanogaster complex, we again used the network approach (Figure 2B). The reconstruction shows long and short branches connecting the sequences of D. sechellia, suggesting the presence of old and young copies in this species. These sequences are connected to two copies of D. simulans through median vectors (ancestral or unsampled sequences). In D. simulans, only two full-length sequences were sampled; these sequences are located in different regions of the network. One is related by a long branch to the old sequences of D. sechellia through median vectors, whereas the other is more closely related to all the sequences of D. melanogaster. As shown in Figure 2B, all copies of D. melanogaster are directly derived from this sequence of D. simulans, and all have short branches, indicating a very recent origin. This relationship suggests a transfer of 412 from D. simulans to D. melanogaster.

To confirm this pattern in strains derived from several natural populations, we sequenced the integrase region of the 412 element in the same strains of D. melanogaster and D. simulans analyzed for Bari (from Africa, Asia, Europe and South America) and reconstructed the network (Figure 2C). A similar relationship was observed in the strains of both species, where most of the D. melanogaster sequences show short branches and are derived from only one D. simulans sequence. Two sequences of D. melanogaster are excluded from this group. They are unresolved and connected by a median vector to the D. simulans sequences, as is apparent from the reticulation in the network. There are two main groups of D. simulans sequences: one with resolved relationships and short branches, and the other presenting reticulations and several long branches. This scenario reflects the ancient and complex evolutionary history of 412 in this species. Note that the sequence transferred to D. melanogaster is shared by all D. simulans strains, indicating that it is an active ancestral sequence in this species. In addition, all D. melanogaster strains, regardless of origin, share the sequence derived from D. simulans, which differs by only one point mutation. The sharing of ancestral sequences among strains from different continents suggests that, as with Bari, transfer occurred before the dispersal out of Africa.

The age of the proposed transfer was estimated using the molecular clock equation (t = Ks/2r). To calculate Ks, the 27 full-length sequences of gag + pol ORFs extracted from the sequenced D. melanogaster genome were compared to the ancestral sequence from D. simulans (Figure 2B, detail). The oldest age estimated is 146,000 y (average = 33,674 y ± 0.0084; lowest = 0.0 y), suggesting that the transfer of 412 from D. simulans to D. melanogaster occurred very recently (see Table S7 in Additional File 3). Indeed, the ages of the insertions in the D. melanogaster genome, as calculated by the divergence between the LTRs of each copy are 94,697 y (the highest) and 0.0 y (the lowest; see Figure S4 in Additional File 3). It is known that at least two factors can introduce biases into this estimation. First, the LTRs evolve, in general, faster than the coding domains of the retrotransposon sequence. Second, because the LTRs of the new copy are synthesized from only one maternal LTR at the moment of reverse transcription, the new LTRs are identical at the point of insertion. This process may conceal the accumulation of divergence between copies. Despite these biases, this approach has been widely used 29 30 and provides useful information about the date of insertion of each copy. The estimated age indicates that transposition would have started soon after the introduction of the copy and continued until very recently, congruent with empirical data and simulations 7 . The analysis shows that all full-length 412 D. melanogaster sequences were inserted into the genome no more than 0.1 Mya, while in D. simulans, the insertion occurred approximately 0.3 Mya.

We performed an in silico search for the DNA transposon Bari in the sequenced genome of species of the melanogaster group, and we sequenced a region of the transposase in different geographic strains of D. melanogaster and D. simulans. The element exhibits structural variations related to its TIRs, which characterize various Bari subfamilies 23 . Both long and short TIRs were observed in the sequences analyzed in our study. The sequences of elements found in D. ananassae and D. erecta, which contain long TIRs, included stop codons and are therefore inactive, whereas the sequences of D. melanogaster, D. simulans and D. sechellia included short TIRs. In D. melanogaster and D. simulans, there were full-length sequences without stop codons, which therefore suggest putatively active copies. In contrast, the full-length sequences in D. sechellia included multiple stop codons.

The element found in D. ananassae, with long TIRs, is more closely related to the element of D. melanogaster, which has short TIRs, than to that of D. erecta, which also has long TIRs. These relationships produce incongruences between the element and species phylogenies. The element found in D. erecta (Bari2 subfamily) is widely distributed in Drosophila, whereas those in D. ananassae (Bari1-like subfamily) and in the melanogaster complex (Bari1 subfamily) are restricted to their respective species; this pattern indicates that the Bari element of D. erecta is older than that of D. ananassae 23 . Therefore, we propose that the ancestor of the melanogaster species group possessed at least two Bari variants that were stochastically inherited by the derived species. The estimated age of the common ancestor of the Bari sequences in the genomes of D. ananassae and D. melanogaster (~ 17 Mya) is similar to the ages of their host genes (~ 19 Mya). During this period, the proto-melanogaster lineage was still diversifying from its sister subgroups in Africa. The only published estimate for the diversification between the melanogaster and ananassae subgroups is ~ 44 Mya 25 , but it could have occurred more recently, as all of the divergence times estimated in that study are more than two times higher than other estimates (e.g., between the melanogaster and montium subgroups, 41.3 Mya 25 and 12.7 Mya 33 and between D. melanogaster and D. simulans, 5.4 Mya 25 , 2.3 Mya 33 and 2–3 Mya 11 ). In conclusion, i) the phylogenetic incongruence arising from clustering the Bari sequences of D. ananassae with those of the melanogaster complex and ii) the older estimated age of their Bari ancestor with respect to the age of diversification of the melanogaster subgroup and the migration of its ancestral lineage to Tropical Africa support the hypothesis of vertical inheritance with stochastic retention of polymorphic sequences of Bari in these species. The ancestral polymorphism hypothesis is also supported by the smaller distances between the elements of the D. ananassae vs. melanogaster complex than between those of the D. erecta vs. melanogaster complex.

As described in the Background section, few reports clearly demonstrate retention of ancestral polymorphisms. One such study examines the DNA transposon mariner, which occurs in the melanogaster subgroup in D. simulans, D. sechellia, D. mauritiana, D. yakuba and D. teissieri, but not in D. melanogaster, D. erecta and D. orena. It is proposed that mariner was present in the ancestral species prior to the radiation of the melanogaster species subgroup and that the element was lost independently in the lineages leading to D. melanogaster and D. orena - D. erecta. In addition, the mariner sequences of D. simulans and D. mauritiana share active copies, a subset of all mariner sequences, that cluster together rather than according to the species phylogeny. This shared polymorphism in populations of D. simulans worldwide and in D. mauritiana indicates retention of ancestral polymorphisms 34 . In both the mariner study and in ours, the rates of evolution of the DNA transposon and of a host gene were compared to test the ancestral-polymorphism hypothesis and to explain incongruence between the phylogenies of the species and the transposable element.

The incongruence between the species phylogeny and the phylogenies of the Bari and 412 elements, along with the ages of the sequences shared between D. melanogaster and D. simulans (less than the age of the species divergence), could arise through either horizontal transfer or introgressive hybridization. An increasing number of reports in the last two decades (mostly published following the rise of large-scale genome sequencing, which allows analysis of most copies from a given genome as well as broader comparative evolutionary analysis) suggest that horizontal transfer of transposable elements occurs frequently in eukaryotes (for a review see 7 ), especially in Drosophila (for a review see 8 9 ). Particularly for D. melanogaster and D. simulans, evidence of horizontal transfer is accumulating in the literature 19 35 36 37 38 39 40 41 42 , including the elements Bari 19 37 and 412 19 37 42 , the focus of this study; however, D. melanogaster and D. simulans can hybridize, even today, in both the laboratory 43 and in nature 44 ; therefore, it is also possible to introduce transposable elements via hybridization.

In order for either horizontal transfer or introgression to occur, species must overlap in time, space and habitat. Moreover, for horizontal transfer, a shared potential vector (e.g., a virus or endobacterium) is required. Currently, the sibling species D. melanogaster and D. simulans are cosmopolitan and are sympatric in many parts of the world; however, our analyses suggest that transfer events occurred before their worldwide expansion but after species divergence. It is thought that D. melanogaster was the first species to diverge from a common ancestor in West Africa and that the ancestor of D. simulans, the proto-simulans lineage, migrated to east Africa and occupied the Pacific islands and then diversified. After the divergence, D. simulans returned to the mainland and expanded, coming into contact with D. melanogaster populations in Africa during the Late Pleistocene (around 120 and 9 thousand years ago) 10 ; we estimate that transfer of both elements occurred during this period. Overlaps in space, time, and most likely niche, may have provided the necessary conditions for both horizontal transfer and introgression. Regardless of the precise mechanism, after the transfers occurred, both species expanded out of Africa, D. melanogaster with the Homo sapiens migration and D. simulans more recently, most likely during the great navigations 12 .

The presence of defective fragments of Bari and 412 elements in the sequenced genomes of both species and the presence of two more divergent sequences of the retrotransposon 412 in populations of D. melanogaster (as shown in the network) indicate that both elements were present in the common ancestor of these species and were inherited by both species. The presence of full-length and putatively active copies in D. melanogaster, which were derived exclusively from sequences transferred from D. simulans, suggests that D. melanogaster either did not inherit active copies of both Bari and 412 from a common ancestor or lost these copies early in its diversification. The defective fragments that are still present today would then be remnants of the copies inherited from its ancestor. After the reintroductions of Bari and 412, the transferred sequences remained active in D. melanogaster, giving rise to the majority of copies that are currently present in this species. We show here that the amplification of these copies occurred in a short period of time and at elevated rates, resulting in a burst of transposition. This process can be deduced from the network by the presence of identical sequences clustered in nodes or by similar sequences connected by short branches; these characteristics are observed in both species, in both the sequenced genomes and in natural populations. Bursts of transposition have been previously reported, in silico, for other elements in the genomes of the melanogaster subgroup, such as the Helitron DINE-1 in D. yakuba and D. ananassae 45 and numerous LTR retrotransposons in the D. melanogaster euchromatin 20 .

The element 412 occurs in two subfamilies in D. simulans. Only one is observed in D. melanogaster and is very similar to one of the subfamilies found in D. simulans 24 30 . Therefore, according to our data, an ancestral sequence of the melanogaster complex was likely inherited by both species, but diversification occurred only in D. simulans; later, one of the two subfamilies was transferred to D. melanogaster, which at the time had only defective copies derived from its ancestral lineage. The subfamilies present in D. simulans were the result of rearrangements, indels and point mutations in the regulatory sequences in the 5' LTR –UTR region; then, these changes gave rise to elements that were able to overcome the host control for transposition and thus able to became invaders 24 30 . This process may explain why the retrotransposon 412 remained active in D. simulans following its divergence and retained its capacity for amplification following its transfer to D. melanogaster, whose control host system, which had coevolved related with the sequences inherited from the ancestral copies, could not recognize this new element. Data from the literature (reviewed in 9 ) suggest that several elements have been independently transferred between the two species over time (e.g., Copia, tirant, Opus, Gypsy 2, Gypsy 5, Gypsy 6, 297, 17.6), but several others may have been transferred simultaneously and very recently (e.g., 412, Blood, Stalker 2, Transpac, Flea), as can be deduced from the very similar ages of the transfers. There is a consensus that multiple elements have recently arrived in D. melanogaster; however, their origins either have not been suggested 20 23 or were not clearly demonstrated 38 46 . Utilization of the network approach allowed us to propose, at least for Bari and 412, that D. simulans was the donor species.

The search for copies of the retrotransposon 412 and of the DNA transposon Bari in the sequenced genomes of species of Drosophila melanogaster group (release 5.18 of D. melanogaster and 1.3 for all other species 16 ) was performed using BLASTn 47 . The deposited sequences (Repbase databank 48 ) described in D. melanogaster were used directly to search in this species. The Bari sequence used [GenBank: X67681] is 1,728 bp long and encodes a 340 aa transposase. The 412 sequence [GenBank: X04132] is 8,039 bp long and encodes a 452 aa gag-like protein and a 1,001 aa pol-like polyprotein (Reverse transcriptase, RNase H and Intregrase, respectively). For the other species, the deposited sequences were used to identify the reference sequence, i.e., the most complete and conserved sequence of each element in each species. These sequences were used to search for complete and incomplete copies in each genome. The complete copies were tested for the presence and integrity of the transposase gene (Bari) and gag and pol (412) using the software ORF Finder 49 . In the search, the hits with the smallest e-values (> 10^-5) and highest RM scores (> 225) were selected. Alignment, reconstruction of the phylogeny by maximum likelihood (ML), calculation of the rate of synonymous substitutions per synonymous site (Ks) and analysis of the Kimura 2-parameters distance (K2p) 50 were performed only for full-length sequences with both TIRs and LTRs, using the package Mega5 51 . The evolutionary relationships between sequences were also reconstructed using the package Network 52 . The ages of the transposable elements were estimated using the molecular clock equation r = k/2 T, where r is the rate of neutral synonymous substitution in the genus Drosophila (r = 0.011/site/million years 25 and k is the rate of divergence in the synonymous sites (Ks). The molecular-clock hypothesis assumes that when genes from different species are compared, the number of nucleotide changes is proportional to the speciation time. We then estimated the divergence time between species of the melanogaster group using the Ks values of the CDS of two host genes (ADH: Alcohol dehydrogenase and GAPDH: Glyceraldehyde 3 phosphate dehydrogenase 1; see Table 5S in Additional File 2) and of the Bari transposase. The ages of the insertions of the retrotransposon 412 were estimated using the date of divergence between both LTRs of each copy by K2p in the molecular clock equation.

The phylogenetic relationships between the Bari sequences of strains of D. melanogaster and D. simulans of different geographic origin (Table 1) were also reconstructed. These strains were classified as ancestors (sampled in Africa) or invaders (ancient, sampled in Asia and Europe; or recent, sampled in Brazil) according to place of origin and literature reports 10 .

Classification with the regard to geographic origin of the strains, name of the collectors or stock numbers and GenBank sequence accession numbers.

Genomic DNA was extracted from 50 individuals 53 . DNA concentration and integrity was analyzed via spectrophotometer (NanoDrop). Amplification (PCR) was performed using specific primers that anneals to nucleotides 412 to 1,133 (total length of 722 bp) in the Bari transposase gene (Bari_F 5’ CGG GCT GGT ATT GTT GCT AGG TTT 3’ and Bari_R 5’ ATC CTA CCC TTA TGG CAT GGA GCA 3’) and to nucleotides 5,622 – to 6,499 (total length of 878 bp) in the 412 integrase gene (412_F 5' TGG SCR AGG TCA WAR GAC AT 3’ and 412_R 5' RCT TTS TAT STT ATA GGG CC 3’), 0.625 unit of Taq polymerase (Invitrogen), 200 ng genomic DNA, 1 mM of MgCl₂, 1 X buffer, 0.08 mM of dNTPs and 0.4 mM of each primer, for a final volume of 25 μL. PCR conditions were as follows: initial denaturation (94°C, 120 s); followed by 30 cycles of denaturation (94°C, 45 s), annealing (69°C for Bari and 59°C for 412, 45 s) and extension (72°C, 60 s). Each PCR product was analyzed by gel electrophoresis on a 1.0% agarose gel, purified (DNA GFX DNA & Gel Band, GE) and cloned (TOPO TA Cloning kit, Invitrogen) according to the manufacturer’s specifications. Approximately 30 (D. melanogaster) and 20 (D. simulans) clones were selected for plasmid extraction by phenol/chloroform protocol and sequenced using the universal primers M13F and M13R. The sequences produced were deposited in the GenBank database (Table 1).