The evolutionary history of mariner elements in stalk-eyed flies reveals the horizontal transfer of transposons from insects into the genome of the cnidarian Hydra vulgaris

Abstract

The stalk-eyed flies (Diopsidae, Diptera) are a family of approximately 100 species of calypterate dipterans, characterised by extended head capsules. Species within the family have previously been shown to possess six subfamilies of mariner transposons, with nucleotide substitution patterns suggesting that at least two subfamilies are currently active. The vertumnana subfamily has been shown to have been involved in a horizontal transfer event involving Diopsidae and a second dipteran family in the Tephritidae. Presented here are cloned and sequenced mariner elements from three further diopsid species, in addition to a bioinformatic analysis of mariner elements identified in transcriptomic and genomic data from the genus Teleopsis. The newly identified mariner elements predominantly fall into previously recognised subfamilies, however the publicly available Teleopsis data also revealed a novel subfamily. Three of the seven identified subfamilies are shown to have undergone horizontal transfer, two of which appear to involve diopsid donor species. One recipient group of a diopsid mariner is the Bactrocera genus of tephritid flies, the transfer of which was previously proposed in an earlier study of diopsid mariner elements. The second horizontal transfer, of the mauritiana subfamily, can be traced from the Teleopsis genus to the cnidarian Hydra vulgaris. The mauritiana elements are shown to be active in the recipient H. vulgaris and transposase expression is observed in all body tissues examined in both species. The increased diversity of diopsid mariner elements points to a minimum of four subfamilies being present in the ancestral genome. Both vertical inheritance and stochastic loss of TEs have subsequently occurred within the diopsid radiation. The TE complement of H. vulgaris contains at least two mariner subfamilies of insect origin. Despite the phylogenetic distance between donor and recipient species, both subfamilies are shown to be active and proliferating within H. vulgaris.

Introduction

Transposable elements (TEs) are almost universal components of eukaryotic genomes [1], that are capable of driving their own replication and movement within their host genome. They are divided into two classes, based upon their transposition mechanism. The Class I elements are retrotransposons, which transpose via an RNA intermediate; in contrast, Class II elements are transposons that move as DNA molecules. Transposition of many DNA transposons is facilitated by a Transposase (Tnpase) enzyme. The Tnpase binds to the inverted terminal repeats (ITRs) that flank the transposon and generates double stranded DNA breaks in order to excise the parental element and then integrate the element into a new genomic location. Daughter elements may be generated during transposition if the double stranded break created to excise the parental element is repaired using a copy of the transposon as a template (reviewed in [2]). DNA transposons can be subdivided through the phylogenetic analysis of their Tnpase sequences. Within Metazoa, one of the most widespread forms of DNA transposon is the Tc1/mariner superfamily, named after elements originally discovered in Caenorhabditis elegans [3] and Drosophila mauritiana [4].

TEs are recognised as a major source of mutation within their hosts’ genomes. Mutations may be deleterious due to a variety of different mechanisms. Insertion mutations result from TEs integrating within or adjacent to host genes and thereby either altering their expression or mRNA sequence [5, 6], whilst recombination between similar TEs in different genomic locations can produce gross chromosomal rearrangements and result in selection against ectopic exchange [7]. The presence of TEs within their hosts’ genomes results in a metabolic burden due to RNA production, protein synthesis and the repair of double stranded DNA breaks, so the transposition process itself can be deleterious to the host organism [8]. It can be seen that for each of these processes individuals harbouring higher TE copy numbers will be at a greater selective disadvantage than those with lower copy numbers. TEs can be considered to be in a state of genomic conflict with their hosts, as their ability to proliferate may be opposed by natural selection through their hosts. Furthermore, active elements are constantly acquiring mutations and it has been proposed that TEs have a natural life cycle within their hosts, with elements entering a naïve genome and proliferating, before deactivating mutations, as well as host repression mechanisms, result in the loss of all active copies [9].

TE families may maintain on-going transposition through their horizontal transfer into a new host population. The new host is unlikely to have defences, such as RIP, RNAi or protein targeting [10–12], against the invading TE, which will only evolve once the host adapts to the new TE family [13]. Horizontal transfer has been shown to be a common feature of TE evolution [14–17] and has been shown to occur between closely related species [18], as well as species from different eukaryotic supergroups [19]. The different mechanisms which underpin horizontal transfer are currently unknown, although it has been speculated that shared parasites, viruses and introgression between closely related taxa may facilitate the transfer of TEs from one species to another [12, 20, 21]. Horizontal transfer events may be identified through incongruencies between phylogenetic trees, where TE phylogenies show strongly supported differences to host species phylogenies. As phylogenies frequently have poor support, due to the rapid rate of TE evolution [22, 23], phylogenetic trees may often be consistent with both horizontal transfer and vertical inheritance. The direction of a horizontal transfer event may however be established if TEs from one taxon are nested, with strong support, within the TEs of the second taxon. In such circumstances the nested taxon is likely to be the recipient that has acquired the TE from the donor in which its TEs are nested.

The Tc1/mariner superfamily has been extensively studied with regard to both horizontal transfer and vertical inheritance [24–26]. Carr [16] showed that 14 species of diopsid stalk-eyed fly possess a minimum of six subfamilies of the mariner elements, with evidence for on-going transposition uncovered in two families. Diopsids form a family of acalyptrate dipterans (Fig 1) that exhibit hypercephaly, with head capsules laterally extended into eyestalks [27]. One subfamily of the Diopsidae, Centrioncinae, is non-hypercephalic, whilst the second subfamily, Diopsinae, is only made up from hypercephalic species [28]. The Diopsinae are further divided into two tribes, the Sphyracephalini and the Diopsini [27].

Fig 1. Representative phylogeny of diopsid species.

Cladogram highlighting the relationships between the diopsid species involved in this study, based upon Kotrba and Balke [28] and Kotrba et al [29].

One of the two identified putatively active mariner subfamilies within the diopsids, the vertumnana subfamily, presented a phylogeny consistent with the vertical inheritance of the transposon through the Diopsini tribe [16]. In addition, a vertumnana mariner tnpase from the dipteran Bactrocera neohumeralis (Tephritidae) was recovered as nested, with strong support, within the diopsid vertumnana sequences, consistent with the horizontal transfer of the vertumnana subfamily from the diopsids into Bactrocera [16]. The geographically close habitat ranges of Bactrocera and the potential donor genus Teleopsis led to the proposal that the horizontal transfer may have occurred in New Guinea. The species composition of Teleopsis is under debate [29–31], however here it is used within its broadest sense to include the putatively nested or synonymous genera Cyrtodiopsis and Megalobops.

Experimental aims

The presented work expands upon the original diopsid mariner survey of Carr [16] through a PCR screen of two additional species from the Diopsinae, Diasemopsis aethiopica and Diopsis apicalis, and the first investigated species from the Centrioncinae in Teloglabrus entabensis. Furthermore whole genome bioinformatic surveys of Teleopsis dalmanni and Sphyracephala brevicornis were performed in order to overcome the limitations of PCR screening to investigate diopsid mariner diversity. Additional bioinformatic analyses were performed using RNA-Seq reads to identify novel mariner sequences in Teleopsis species and determine relative gene expression levels. The diopsid-Bactrocera horizontal transfer, proposed in Carr [16], is re-evaluated and two further horizontal transfers, between insects and the cnidiarian Hydra vulgaris (Hydridae), are analysed in depth.

Methods

DNA extraction and PCR

Whole flies of D. aethiopica, Di. apicalis and Teloglabrus entabensis were provided by Andrew Pomiankowski of University College, London. DNA was extracted from individual specimens using the Proteinase K/NaCl protocol of Carr et al. [32]. PCR was carried out in 50μl volumes (5U Abgene Red Hot DNA Polymerase, 2.5 mM MgCl2, 0.4mM dNTP), using the MAR124F and MAR276R primers of Robertson [24], which amplify multiple mariner subfamilies. The annealing temperature was 52°C, with an extension step of 72 °C of 1 minute per cycle other than the final cycle which utilised a 10 minute extension step. PCR was undertaken over 25 cycles.

PCR products were ligated into the pGEM-T Easy Vector (Promega) and transformed into Subcloning Efficiency DH5α Chemically Competent Cells (Invitrogen). Plasmid DNA was extracted using the Qiagen Spin Miniprep kit and mariner DNA sequenced using T7 and SP6 primers (Macrogen Inc, Seoul, Korea). All sequences have been deposited in GenBank (accession numbers MN719915-MN719940).

Bioinformatic identification of novel mariner sequences

Genomes of T. dalmanni (assemblies NLCU01 and JXPO01), S. brevicornis (JXPL01), H. vulgaris (ACZU01 and ABRM01), H. oligactis (PJUT01) and H. viridissima (PJUU01) were downloaded from NCBI. Each was screened with RepeatMasker [33] specifying a library of Drosophila sequences available in RepBase [34] with the following options: -species drosophila -pa 4 -nolow -no_is -inv -a. Custom mariner RepeatMasker libraries were compiled and genomes re-screened to obtain all hits not collected using Drosophila sequences using the same options as above with the exception of -lib custom_library.

Putative miniature inverted-repeat transposable element (MITE) families were identified in the genomes of T. dalmanni, H. vulgaris and S. brevicornis using MITE Tracker [35]. The resulting families_nr.fasta file was BLASTed using Tdmar, Hvmar and Sbmar nucleotide queries to ascertain the regions where MITEs may have been derived from the full-length mariner sequences.

Full-length mariner elements were identified in the T. dalmanni genome by undertaking BLASTn similarity searches of the whole genome shotgun contigs using the partial, putatively autonomous, tnpase sequences identified through PCR and RepeatMasker screening. Contigs with tnpase hits for individual subfamilies were aligned with MAFFT v7.309 [36] using the L-INS-I strategy and default parameters; this strategy resulted in the mariner elements in the contigs being aligned with each other. Diagnostic TA target site duplications were identified to confirm that the 5’ and 3’ termini had been identified. Illumina RNA-Seq sequencing reads from T. quinqueguttata (SRX1490590, SRX1490591) and T. whitei (SRX485305) were mapped onto the complete tnpase (coding sequence) cds from each T. dalmanni subfamily in order to generate reconstructed species-specific sequences (S1 Dataset). Reads were mapped onto tnpase cds with SMALT v. 0.2.6 (https://www.sanger.ac.uk/science/tools/smalt-0). The number of reads mapped to each tnpase was calculated in Tablet v.1.19.09.03 [37] from the SMALT output SAM files.

A full-length mauritiana subfamily sequence from H. vulgaris was produced using 180bp query sequences from Tdmar2 (NLCU01006112, 92267–90983). Reads were identified using the Trace Archive Nucleotide BLAST on the Hydra magnipapillata–wgs database. A H. vulgaris mosellana partial tnpase was uncovered (accession number ABRM01005801) in the BLASTn screening for the subfamily phylogeny, using Tdmar2 as a query sequence. The ABRM01005801 hit was then used as a query sequences for the H. vulgaris whole genome shotgun contigs, using BLASTn, in order to identify contigs containing mosellana subfamily elements. Full-length sequences were then uncovered by aligning contigs in MAFFT. Contig ABRM01012171 was assembled with an intact mosellana element that possessed a premature stop codon at positions 431–433. Screening of the Trace Archive for the Hydra magnipapillata–wgs database revealed the majority of reads showed guanosine at position 431 rather than the thymine present in ABRM01012171. Replacing the thymine with guanosine results in a glutamate residue in the Tnpase, as opposed to the stop codon represented in contig ABRM01012171 (S1 Dataset).

Phylogenetic analyses

The sequenced diopsid mariner clones were aligned against homologous regions of tnpase from publicly available diopsid sequences, as well as the uncovered tnpase sequences identified in the whole genome contigs of S. brevicornis and T. dalmanni and the sequence read archive (SRA) RNA-Seq files of T. whitei in MAFFT using the L-INS-I strategy and default parameters. The alignment was manually edited by eye in order to minimise indel regions. The resulting alignment was then subjected to maximum likelihood analysis with the raxmlGUI [38], using the thorough bootstrapping methodology and 1,000 bootstrap replicates. The ML tree was generated from 100 starting parsimony trees and created with the GTRCAT model, following the RAxML author’s recommendation. Bayesian inference phylogenies were created with MrBayes 3.2.6 [39] on the Cipres Science Gateway server [40]. The analyses were run with the GTR+I+Γ model and a four category gamma distribution to correct for among site rate variation. The MCMC analyses consisted of 5,000,000 generations with two parallel chain sets run at default temperatures and a sampling frequency of 1000, with a burnin value of 1250.

The diopsid irritans subfamily dataset was constructed only from putatively autonomous sequences generated through PCR and bioinformatic screening. Nucleotide sequences were acquired for the mosellana subfamily phylogeny using BLASTn, with the tnpase cds of Tdmar4 used as a query sequence. The nr/nt database was screened without an organism limitation, whilst the wgs database was limited to screening Metazoa (taxid: 33208). Bactrocera sequences in NCBI were screened for the vertumnana subfamily phylogeny using the T. quinqueguttata sequence Tqmar1.1 (DQ197023) with BLASTn with screened organisms limited to Bactrocera (taxid: 27456) in both the nr/nt and wgs databases. Extracted sequences were aligned to diopsid sequences in MAFFT. Maximum likelihood and Bayesian inference phylogenies were created for the irritans, mauritiana and mosellana subfamilies using the same protocols as for the diopsid mariner sequences.

Subfamily maximum likelihood and Bayesian inference phylogenies, created with individual insertion sequences, were generated using the same protocols as the diopsid mariner phylogeny. For Hvmar1 and Hvmar2 the Trace Archive for the Hydra magnipapillata–wgs database was screened through NCBI using the 5’ ITR and untranslated region (UTR) as query sequences. 5’ITR/UTR query sequences were also used for Tdmar2-4 insertions phylogenies. BLASTn searches were conducted using the SRA database on the file SRX2950777, which contained 9,283,997 reads of male T. dalmanni genomic DNA. For both the T. dalmanni and H. vulgaris TEs, individual insertions were identified using the 5’ flanking DNA in the sequencing read, with reads that contained less than 12bp of flanking DNA discarded. Sequences were additionally discarded if the 5’ and 3’ termini of the ITR/UTR regions were not intact.

All alignment used in the phylogenetic analyses are presented in the Nexus format in the S2 Dataset.

Gene expression analysis of mariner tnpase

Raw Illumina RNA-Seq transcriptome read files were downloaded from NCBI (see S3 Table for accession numbers, as well as the tissue type and number of reads for each file). Reads were mapped onto tnpase cds with SMALT. The number of reads mapped to each tnpase was calculated in Tablet v.1.19.09.03 [37] from the SMALT output SAM files. Normalised gene expression levels were calculated as transcripts per million (TPM) [41].

Results

A revised phylogeny of mariner within Diopsidae

Cloned PCR products amplified from genomic DNA extractions of Diasemopsis aethiopica, Diopsis apicalis and Teloglabrus entabensis were subject to BLASTn similarity searching, which resulted in 26 partial mariner sequences being identified (S1 Table). The PCR screen only identified the vertumnana subfamily in Di. apicalis. One clone did not show any obvious null mutations and may have been amplified from an autonomous element, whilst the other two clones contained premature stop codons in the amplified region. The irritans, mauritiana and mellifera subfamilies were amplified from the genomic DNA of D. aethiopica. Only one sequence, from the irritans subfamily, did not contain premature stop codons. The genome of the centrioncid Te. entabensis harbours a minimum of four subfamilies, with the capitata, irritans, mellifera and vertumnana subfamilies all amplified in the PCR screen. Possible autonomous elements, lacking premature stop codons, were amplified from the irritans, mellifera and vertumnana subfamilies, whilst the single copy amplified from the capitata subfamily contained multiple internal stop codons.

The newly generated sequences were combined with the diopsid mariner sequences from Carr [16] to produce a customised RepeatMasker [33] library, in order to screen the whole genome contigs of T. dalmanni and S. brevicornis. The S. brevicornis screen revealed the presence of two mariner elements with putative, albeit imperfect, ITRs (accession numbers: JXPL01000092 and JXPL01000142, S1 Dataset). Neither copy was assembled as encoding a functional tnpase, indicating that mariner is no longer active in S. brevicornis. The screen of T. dalmanni identified 93 mariner-like sequences within the assembled whole genome shotgun (wgs) contigs. Phylogenetic analyses of the T. dalmanni sequences resulted in four distinct clades (97–100% maximum likelihood bootstrap percentage (mlBP); 1.00 bayesian inference posterior probability (biPP)), which correspond to four different mariner subfamilies (S1 Fig). Confirming the results of the PCR screen in Carr [16], the mauritiana and vertumnana subfamilies were uncovered with RepeatMasker; furthermore the irritans subfamily and a previously unidentified subfamily were also present. The mariner complement of T. dalmanni recovered in the RepeatMasker screen was dominated by elements from the mauritiana subfamily, with 68 of 93 identified insertions belonging to this subfamily. Full length, putatively autonomous, elements were uncovered for three of the subfamilies in the T. dalmanni whole genome contigs (S1 Dataset) through the presence of inverted terminal repeats and diagnostic TA target site duplications generated by mariner elements [42]. No functional copy of the vertumnana subfamily was identified, with only degraded pseudogenes present in the assembled genome.

Further bioinformatic searching for diopsid mariner sequences was performed by mapping SRA sequencing reads onto the mariner clones sequenced here and in Carr [16], as well as the full-length mariner tnpase open reading frames (ORFs) identified in the contigs of T. dalmanni. RNA-Seq SRA files were screened for T. quinqueguttata and T. whitei. The vertumnana, cecropia, mauritiana and mellifera subfamilies, previously identified in T. quinqueguttata [16], were all expressed at relatively low levels (S2 Table), however no additional subfamilies were identified in the T. quinqueguttata RNA-Seq reads. From the RNA-Seq reads of T. whitei complete tnpase ORFs could be reconstructed for the irritans, mauritiana subfamilies as well as the novel subfamily uncovered in T. dalmanni (S1 Dataset).

The combined diopsid dataset for all identified mariner sequences comprised 217 sequences and was analysed using both ML and BI methodologies (Fig 2). The six mariner subfamilies identified in diopsids by Carr [16] were recovered, however the capitata subfamily was not recovered as monophyletic, albeit with only moderate phylogenetic support (55% mlBP, 0.81 biPP). A seventh subfamily was identified, with strong support (100% mlBP, 1.00 biPP), which was made up only from mariner elements from T. dalmanni and T. whitei. The two non-functional S. brevicornis mariner elements were recovered with strong support (79% mlBP, 1.00 biPP) as members of the irritans subfamily.

Fig 2. Maximum likelihood phylogeny of diopsid mariner sequences.

The phylogeny was constructed from 480 aligned nucleotide positions using the GTRCAT model, and estimated nucleotide frequencies. Values for mlBP and biPP are shown above and below the branches respectively. 100% mlBP and 1.00 biPP are both denoted by “*”. Values <70% mlBP and <0.97 biPP are denoted by “-”. The scale bar represents the number of substitutions per site. Individual mariner subfamilies are bracketed and colour-coded.

The presence and identity of subfamilies in the screen undertaken here and Carr [16], as well as the absence of families in the whole genome contigs of S. brevicornis and T. dalmanni are shown in Table 1.

Table 1. The presence of mariner subfamilies within screened diopsid species.

Species	capitata	cecropia	irritans	mauritiana	mellifera	mosellana	vertumnana
Centrioncinae
Te. entabensis	Temar4	-	Temar2	-	Temar3	-	Temar1
Sphyracephalini
C. seyrigi	-	-	-	-	Csmar1-2	-	-
S. beccarii	-	-	-	-	-	-	-
S. brevicornis	Absencea	Absencea	Sbmar1	Absencea	Absencea	Absencea	Absencea
S. europaea	Semar3, Semar4	-	Semar1	Semar2	Semar5	-	-
Diopsini
D. aethiopica	-	-	Damar3	Damar1	Damar2	-	-
D. comoroensis	-	-	-	-	Dcmar2	-	Dcmar1
D. dubia	Ddmar2	Ddmar3	-	-	Ddmar1	-	-
D. meigenii	-	-	-	-	-	-	Dmmar1
D. signata	-	-	-	-	Dsmar2	-	Dsmar1
Di. apicalis	-	-	-	-	-	-	Dioamar1
T. breviscopium	-	-	-	Tbmar2	Tbmar3	-	Tbmar1
T. dalmanni	Absencea	Absencea	Tdmar3	Tdmar2	Absencea	Tdmar4	Tdmar1
T. quadriguttata	-	-	-	-	-	-	Tqdmar1
T. quinqueguttata	-	Tqmar4	-	Tqmar2	Tqmar3	-	Tqmar1
T. rubicunda	-	-	-	-	-	-	Trmar1
T. thaii	-	-	-	-	-	-	Ttmar1
T. whitei	-	-	Twmar3b	Twmar2b	-	Twmar4b	Twmar1

Dash represents the absence of the subfamily in PCR and transcriptome screens.

^a: The subfamily is absent from the assembled whole genome contigs.

^b: Sequences identified in RNA-Seq reads.

Due to low numbers of putatively autonomous copies, individual phylogenies were not created for the diopsid representatives of the capitata, cecropia, mellifera subfamilies and the novel subfamily uncovered in Teleopsis. The capitata and mellifera subfamilies were both identified in the centrioncid Te. entabensis, as well as the Sphyracephalini and Diopsini tribes of Diopsinae (Fig 2). The cecropia subfamily elements are currently limited to two species, D. dubia and T. quinqueguttata, with no additional members identified in either the PCR or bioinformatics screens conducted here.

Four subfamilies were each analysed phylogenetically. The irritans subfamily was identified in six species, of which five harboured potentially autonomous elements. A phylogeny of the putatively autonomous diopsid elements of the irritans subfamily, rooted with sequences from the centrioncid Te. entabensis, is broadly congruent with the host species phylogeny; the only element in an unexpected position is a long-branched sequence from D. aethiopica which clusters with moderate to strong support (65% mlBP; 0.98 biPP) with an irritans tnpase from S. europaea (S2 Fig). The phylogenies of the mauritiana and vertumnana subfamilies, as well as the novel subfamily uncovered in T. dalmani are presented individually below.

Phylogenetic analysis of the mosellana subfamily

The seventh mariner subfamily, identified within the whole genome shotgun contigs of T. dalmanni and RNA-Seq transcriptome reads of T. whitei, was not uncovered in the diopsid PCR screen of Carr [16]. In order to investigate the evolutionary origins of the subfamily in the diopsids BLASTn similarity searching, through NCBI, was undertaken with the nucleotide sequence of a putatively autonomous tnpase from T. dalmanni. Only the top hit uncovered for each species was extracted for phylogenetic analysis. The screen uncovered the presence of 41 tnpase sequences that clustered with the novel Teleopsis sequences (95% mlBP; 1.00 biPP, Fig 3). The earliest published sequence from the subfamily was identified in the genome of the dipteran Sitodiplosis mosellana and accordingly the subfamily is named here as the mosellana subfamily.

Fig 3. Maximum likelihood phylogeny of the mosellana subfamily.

The phylogeny was constructed from 1032 aligned nucleotide positions using the GTRCAT model, and estimated nucleotide frequencies. The mosellana elements are bracketed and rooted with tnpase sequences from the irritans and mellifera subfamilies. Values for mlBP and biPP are shown above and below the branches respectively. Diopsid sequences are shown in purple, other dipteran sequences are shown in light blue. Dark blue sequences are from hymenopteran hosts, pink sequences from lepidopteran hosts, the orange sequence is from an archaeognath host and light green sequences are from copepod hosts. The brown sequence is from a hydrozoan host and the mustard sequence from a fish host. The outgroup sequences are from the irritans subfamily (Semar1 and Temar2.2) and the mellifera subfamily (Temar3.1). 100% mlBP and 1.00 biPP are both denoted by “*”. Values <50% mlBP and <0.70 biPP are denoted by “-”. The scale bar represents the number of substitutions per site.

The Teleopsis sequences fell within two distinct groups in the mosellana subfamily, however there were no strongly supported branches (≥70%mlBP and ≥0.97biPP) separating the two clusters (Fig 3). Of the two groups, one consisted of three non-autonomous T. dalmanni sequences which all harboured premature stop codons. The second group possessed sequences from both T. dalmanni and T. whitei, with both species harbouring putatively functional tnpase sequences. The potentially active Teleopsis elements formed a monophyletic group with tnpase sequences from three Bactrocera species and S. mosellana, however this clade of dipteran elements had no phylogenetic support.

Thirty nine of the mosellana sequences were uncovered from pancrustacean species, with hosts falling within the insect orders Archaeognatha, Coleoptera, Diptera, Hymenoptera and Lepidoptera, as well as the crustacean Copepoda order. The two remaining sequences were identified in the whole genome shotgun contigs of the hydrozoan cnidarian Hydra vulgaris (synonymised with H. magnipapillata, H. littoralis and H. attenuata) and the ray-finned fish Cyprinodon variegatus. The sequence from the C. variegatus contig showed a single nucleotide difference (99.9% nucleotide identity) from a mosellana subfamily element uncovered in the copepod Caligus rogercresseyi. This is consistent with a horizontal transfer between the two species, with C. variegatus being a marine fish and C. rogercresseyi a sea louse parasite of fish. The more biologically plausible direction of transfer would be from C. rogercresseyi to C. variegatus, as the fish mariner sequence is nested within those of the pancrustaceans by three strongly supported branches (all ≥96% mlBP and 1.00 biPP). An alternative explanation is that the sequenced genomic DNA of C. variegatus was contaminated by DNA from C. rogercresseyi. Consistent with this hypothesis only a single copy of the transposon was uncovered in the whole genome contigs of C. varietgatus, present in a contig (JPKM01108314) which only contained 53bp of flanking DNA. Two copies, possessing different flanking DNA, of the mosellana element were identified in C. rogercresseyi contigs (LBBV01023767 and LBBU01012271). Mapping RNA-Seq reads of C. rogercresseyi (54.6 million reads, SRX864101-2 and SRX1481244) to the mosellana subfamily sequence (accession number LBBV01023767) revealed 468 reads that spanned the entire coding region. The subfamily however could not be identified in C. variegatus RNA-Seq reads (124.7 million reads, SRX3140005-6, SRX3140009-11, SRX5103143, SRX5103155 and SRX5103167). The lack of expression in C. variegatus suggests either an unsuccessful horizontal transfer or contamination of genomic DNA during whole genome sequencing.

A full-length consensus sequence, labelled Hvmar1, was generated for the mosellana subfamily element in the genome of H. vulgaris (S1 Dataset). The H. vulgaris tnpase clustered with mariner elements from the dipteran Bactrocera tryoni and the hymenopteran ants Camponotus floridanus and Pseudomyrmex gracilis (100% mlBP, 1.00biPP, Fig 3). Hvmar1 exhibited 83.7% and 87.0% nucleotide identity with the C. floridanus (accession number XM_011264844) and P. gracilis (accession number XM_020442418) mariner elements respectively, despite cnidarians and insects last sharing a common ancestor approximately between 600–700 million years ago [43–45]. Hvmar1 appears to be a genuine component of the H. vulgaris genome, as similarity searching of the NCBI Hydra magnipapillata wgs Trace Archive with BLASTn revealed 148 distinct 5’ termini. Furthermore, mapping RNA-Seq reads to the tnpase also showed Hvmar1 to be expressed in whole polyps, as well as head, foot, tentacle and body tissue (S2 Table).

The high nucleotide identity between Hvmar1 and the hymenopteran mariners, as well as the nested position of Hvmar1 within the insect mosellana sequences is consistent with a horizontal transfer event. The donor species would appear to be an insect, although the donor insect order cannot be confirmed, as Hvmar1 clusters with mariner elements from dipteran and hymenopteran species but is not nested within either group. Similarity searching of whole genome contigs, with BLASTn, using Hvmar1 as a query sequence failed to uncover mosellana sequences in either H. viridis or H. oligactis, as well as the genomes of other cnidarians. The absence of the subfamily within other Hydra species is consistent with H. vulgaris being the recipient species of the horizontal transfer event in Cnidaria.

A reassessment of the diopsid-Bactrocera vertumnana horizontal transfer

Carr [16] proposed a putative horizontal transfer event between Diopsini stalk-eyed flies and the tephritid Bactrocera neohumeralis. Due to the B. neohumeralis sequence (clone Bnmar29, accession number AF348438.1) being nested within the Teleopsis sequences, it was suggested that the direction of transfer was from Teleopsis to Bactrocera, although the actual donor and recipient species could not be confirmed. The identification of vertumnana sequences from two additional diopsid species (Te. entabensis and Di. apicalis), as well as the additional sequencing of whole genomes from Bactrocera species, allowed a review of the phylogenetic relationships between vertumnana elements from diopsid and Bactrocera hosts.

BLASTn similarity searching of Bactrocera wgs contigs and the nr/nt database, using Tqmar1.1 (Accession Number DQ197023), uncovered vertumnana sequences from five species in addition to clone Bnmar29 from B. neohumeralis. The vertumnana subfamily therefore has a greater distribution within Bactrocera than recognised in Carr [16]. Diopsid vertumnana sequences, taken from all twelve species in which the subfamily has been identified, were aligned with seven Bactrocera elements. A single sequence was used for each diopsid species in which the subfamily has been identified with the exception of T. dalmanni; due to internal deletions two sequences were used for T. dalmanni. The resulting alignment was subjected to maximum likelihood and Bayesian inference analyses (Fig 4). The Bactrocera elements formed a strongly supported monophyletic group (96% mlBP, 1.00 biPP), as did the vertumnana sequences from Teleopsis (94% mlBP, 1.00 biPP) and Diasemopsis (88% mlBP, 0.99 biPP). The expanded vertumnana phylogeny therefore rejects the placement of the B. neohumeralis transposon within the grouping of Teleopsis elements recovered in the Carr [16] ML phylogeny. Rooting the phylogeny with the mariner sequenced from the earliest branching diopsid genus, Teloglabrus, recovers the expected relationships between elements from the host diopsid genera under the model of vertical inheritance. The Bactrocera elements are recovered as nested within the Diopsini vertumnana elements (100% mlBP, 1.00 biPP), as the sister-group to the tnpases sequenced from Teleopsis species (99% mlBP, 1.00 biPP). Re-rooting the phylogeny with the Bactrocera sequences fails to recover any of the expected genera relationships based upon the host species phylogeny. All further potential rooting of the phylogeny also fail to recover the expected genera relationships based upon the host species phylogeny. The most biologically plausible rooting for the phylogeny is therefore between the Teloglabrus tnpase and the remaining transposons, with the tephritid vertumnana being recovered as the sister-group to the Teleopsis transposons. The increased dataset presented here therefore provides further evidence for a horizontal transfer event of the vertumnana subfamily from the Diopsini to Bactrocera.

Fig 4. Maximum likelihood phylogeny of the vertumnana subfamily.

The phylogeny was constructed from 510 aligned nucleotide positions using the GTRCAT model and estimated nucleotide frequencies. Support values are shown in the same format as Fig 3.

Phylogenetic evidence for a cross-phyla horizontal transfer event from Teleopsis to Hydra

BLAST screening of the newly sequenced mariner elements from D. aethiopica unexpectedly recovered a high scoring hit (E value: 5e-175) from H. vulgaris (accession number U51187, clone Hydra.vulgaris.6) when Damar1.1 of the mauritiana subfamily was used as a query sequence. The only other mariner sequences uncovered with similarly high BLAST scores were from other diopsid species. A BLASTn screen of the H. vulgaris whole-genome shotgun contigs, using sequence Hydra.vulgaris.6 as a query sequence, confirmed the presence of 218 similar sequences recovered with E values of 0.0 indicating that the original Hydra.vulgaris.6 sequence was a genuine component of the H. vulgaris genome and not a PCR contaminant. Robertson [46] highlighted the relationship of clone Hydra.vulgaris.6 with insect mariner sequences, but did not propose whether a horizontal transfer donor was a cnidarian or an insect.

The finding raised the possibility of the identification of a second horizontal transfer event of a mariner subfamily from diopsids, as well as a second insect to Hydra horizontal transfer. A reciprocal BLASTn of Hydra.vulgaris.6 against diopsid sequences in the nr/nt database resulted in high scoring hits (E value: 0.0, query coverage: 100%, nucleotide identity: >95%) for mariner sequences deposited from T. breviscopium, T. dalmanni and T. quinqueguttata. A BLASTn screen of the whole-genome shotgun contigs of T. dalmanni uncovered a complete, intact mauritiana mariner element (Accession Number NLCU01006112, position 92,267–90,983), designated as Tdmar2 on the basis of identity with the sequenced clones generated by Carr [16]. Tdmar2 possessed 27bp inverted terminal repeats (ITRs) and a putative tnpase ORF of 1,038bp in length (S1 Dataset).

Screening the NCBI Sequencing Trace Archive of the H. vulgaris genome with 180bp query sequences of Tdmar2 uncovered hits across the entire element. The concatenated hits produced a putative mauritiana subfamily element, designated as Hvmar2, which exhibited 31 nucleotide differences from Tdmar2 out of 1,287 sites (97.6% nucleotide identity). The coding regions of Tdmar2 and Hvmar2 showed 19 nucleotide differences (98.2% nucleotide identity), which resulted in 14 amino acid differences between the putative Tnpases.

Similarity screening with BLASTn of the wgs contigs of H. oligactis and H. viridis, as well as all available cnidarian whole-genome contigs in NCBI failed to uncover orthologous mauritiana subfamily sequences. Phylogenetic analyses of Hvmar2 and the diopsid mauritiana elements clustered the Hvmar2 with the Teleopsis elements in a group with robust support (84% mlBP, 0.99 biPP, Fig 5) and furthermore nested Hvmar2 within the Teleopsis sequences (97% mlBP, 1.00 biPP). Rooting the mauritiana sequences with those from the Sphyracephalini S. europaea recovered the expected relationships between the diopsid TEs based upon their host species, consistent with their vertical inheritance since the origin of the Diopsinae. An alternative rooting, between Hvmar2 and the diopsid elements, failed to recover the expected species relationships, indicating that this is not the correct root for the phylogenetic tree. The mauritiana subfamily Hvmar2 subfamily therefore appears to have had a recent origin from within the diopsids and specifically from the Teleopsis genus. T. dalmanni does not appear to be the donor species for Hvmar2, as Tdmar2 is recovered as being more closely related to mauritiana elements from both T. whitei and T. breviscopium in Fig 5. Based upon the phylogeny the donor species would appear to be a closer relative of T. breviscopium, T. dalmanni and T. whitei than T. quinqueguttata, as the T. quinqueguttata elements are recovered with strong support at the base of the Teleopsis mauritiana elements. The basal position of the T. quinqueguttata elements mirrors the basal position of T. quinqueguttata in the host species genus [28, 29], further highlighting the reliability of the phylogenetic signal in the mariner sequences.

Fig 5. Maximum likelihood phylogeny of the mauritiana subfamily.

The phylogeny was constructed from 492 aligned nucleotide positions using the GTRCAT model and estimated nucleotide frequencies. Support values are shown in the same format as Fig 3.

Mauritiana tnpase is expressed in the genomes of Teleopsis species and H. vulgaris

The phylogenetic analyses of tnpase from the mauritiana subfamily indicated that an active mariner element had undergone horizontal transfer from an unknown Teleopsis species into the cnidarian H. vulgaris. In order to investigate this proposed transfer the activity of mariner was analysed in both Teleopsis, as the putative donor genus, and H. vulgaris, from the putative recipient genus.

RNA-Seq reads were mapped onto the mariner subfamily tnpase sequences of T. dalmanni, T. quinqueguttata and T. whitei (S2 Table). Within T. dalmanni it was possible to investigate gene expression within the male and female germ cells, as well as whole larvae and adult head cells. Consistent with the lack of identified autonomous vertumana elements in T. dalmanni, the subfamily exhibited the lowest mapping coverage (<1.5 TPM, ≤0.001% of RNA-Seq reads in all five tissue types), albeit based upon a shorter CDS (S2 Table). The irritans, mauritiana and mellifera subfamilies were all expressed in both ovary and testes cells. Expression was also observed in male and female heads, raising the possibility that transposition occurs in both somatic and germline cells. Across the examined five tissue types in T. dalmanni, tnpase TPM and the absolute number of mapped reads was higher for the mosellana subfamily (Tdmar4) than the mauritiana subfamily (Tdmar2). T. whitei testes expression patterns were similar to those of T. dalmanni, with the vertumnana TPM being an order of magnitude lower than values of the irritans, mauritiana and mosellana subfamilies. As in T. dalmanni testes, the mosellana subfamily TPM was more than sevenfold than higher than the value of the mauritiana subfamily. In contrast to T. dalmanni and T. whitei, the vertumnana subfamily showed the highest TPM in T. quinqueguttata testes. However only 39 out of 84,490,068 sequencing reads mapped to the cecropia, mauritiana and mellifera subfamilies, indicating that the other subfamilies may either be non-functional or silenced in T. quinqueguttata testes.

Unlike the diopsids, H. vulgaris lacks a sequestered germline [47], as gametes develop from interstitial cells which also produce neurons, cnidocytes and secretory cells [48, 49]. As with Hvmar1 of the mosellana subfamily, Hvmar2 was shown to be expressed in all cell types (S2 Table), consistent with its on-going transposition. Mapping coverage for the mosellana subfamily was fourfold to eightfold higher compared to the mauritiana subfamily for all examined tissues, mirroring the expression patterns observed in T. dalmanni and T. whitei.

Phylogenetic analyses of individual mariner insertions in the genomes of H. vulgaris and T. dalmanni

The expression of tnpase is required for the transposition of mariner in both species. However, the observed expression does not confirm that the mauritiana elements are transposing, since the expression of tnpase mRNA could reflect the presence of a repressor isoform, such as the KP protein that supresses transposition of the P element in Drosophila melanogaster [8, 50]. The mariner elements in both H. vulgaris and T. dalmanni were therefore subjected to phylogenetic analysis in order to uncover evidence of on-going transposition. Upon transposition daughter elements should possess identical sequences to their parental elements in phylogenetic trees, despite being present at different genomic locations. Nucleotide differences will accumulate over time, as the parental and daughter elements begin to diverge following transposition.

A phylogeny of the Hvmar1 5’ ITR/UTR region revealed that insertions were mainly present on short terminal branches (S3 Fig), with 62 identical copies at different genomic locations. The Hvmar2 phylogeny of 103 insertions was also generated from 5’ ITR/UTR sequences. The phylogeny was similar to that of Hvmar1, with a large number of short branched sequences, as well as 26 identical paralogous sequences. At the base of the phylogeny was a weakly supported grouping of long branched sequences, which appear to be older elements that are no longer transposing. The long branch sequences contained unique indels, consistent with their greater antiquity in the H. vulgaris genome, whilst the presence of identical paralogous sequences indicate that there is on-going transposition of both Hvmar1 and Hvmar2 in H. vulgaris.

The 5’ termini sequences identified in the sequencing reads of T. dalmanni also uncovered identical paralogous copies for the subfamilies Tdmar2-4 (S4 Fig). Within the 86 sequences of Tdmar2, four insertions at different genomic locations showed identical sequences; for Tdmar3 there were 9 identical paralogous insertions across 134 sequences. In contrast to the RepeatMasker tnpase screen of the whole genome contigs, which only uncovered 11 Tdmar4 sequences, the BLAST screen of the 5’ ITR/UTR of sequencing reads identified 2,422 distinct insertions. Of those, 1,790 copies were identical to other copies in different genomic locations, indicating a high level of recent transposition of the mosellana subfamily in T. dalmanni.

Mariner-derived MITEs are abundant in the genome of T. dalmanni

Due to the discrepancy between the results of the T. dalmanni RepeatMasker tnpase screen, where Tdmar2 appeared to have the highest copy number, and the ITR/UTR wgs Trace Archive BLAST screen, which was dominated by Tdmar4 insertions, the genomes of T. dalmanni, S. brevicornis and H. vulgaris were screened for MITEs. MITEs are non-autonomous transposons that possess ITR sequences, but lack internal coding tnpase sequences [51]. MITE families are derived from autonomous transposons and are mobilised via the Tnpase enzymes of autonomous element copies [52]. The T. dalmanni screen uncovered the presence of 342 putative MITE families, of which 55 appeared to have originated from mariner elements (Table 2). Within the mariner-derived MITEs 25 families had an origin within Tdmar4 of the mosellana subfamily and Tdmar4-derived MITEs contributed 2017 out of 2489 MITE insertions in the whole genome contigs (Table 3). Consistent with the lack of autonomous Tdmar1 copies, the vertumnana subfamily possessed the fewest MITEs and only contributed 73 MITE insertions, split across eight families.

Table 2. MITE sequences BLASTed against full-length sequences Hvmar1-2, Sbmar1 and Tdmar1-4.

Species	Total number of MITE families	Number potentially mariner -derived
Diopsidae
S. brevicornis	36	0
T. dalmanni	342	55
Hydridae
H. vulgaris	126	0

Table 3. Characterisation of MITE families for each mariner subfamily in T. dalmanni.

Subfamily	No. of MITE families	Total copy number
Tdmar1	8	73
Tdmar2	5	98
Tdmar3	17	301
Tdmar4	25	2017

MITEs sharing identity with multiple subfamilies were assigned to that with the most significant e-value.

The S. brevicornis genome possessed 36 MITE families, none of which appeared to have a mariner-derived origin (Table 2), further highlighting the paucity of mariner-like elements within the species. In contrast to S. brevicornis the genome of H. vulgaris was rich in MITE elements. A total of 126 MITE families were identified, however none of the families were found to have an origin from mariner elements (Table 2).

Discussion

Multiple mariner subfamilies were present in the ancestral diopsid

The original diopsid mariner study of Carr [16] identified six subfamilies of mariner transposons within 14 species of diopsid, however, with the exception of the vertumnana subfamily, no attempt was made to determine the evolutionary origins of the subfamilies. The sequencing of thirteen mariner clones from the centrioncid Te. entabensis, allows greater insight into the subfamilies present in the last common ancestor (LCA) of diopsids.

Through the use of subfamily phylogenies, as well as the distribution of subfamilies across species, it is possible to estimate the latest evolutionary points of origin of the subfamilies within the diopsids (S5 Fig). The capitata, irritans, mellifera and vertumnana subfamilies are all present in the genome of the centrioncid Te. entabensis as well as multiple Diopsinae species, consistent with their presence in the diopsid LCA. Subfamily phylogenies of both irritans and vertumnana also indicate their vertical inheritance since the origin of the Diopsidae.

The mauritiana subfamily was amplified in species across the Diopsinae, but not from the centrioncid Te. entabensis. Only 13 clones were sequenced from the genomic DNA of Te. entabensis, so it is possible that the absence of the mauritiana subfamily was due to the limited sample size. The distribution of mauritiana elements, as well as the phylogenetic relationships presented in Fig 5 indicate that the subfamily was present in the genome of the Diopsinae LCA, but an earlier origin in the ancestor of all diopsids cannot be excluded on the basis of the current dataset.

The cecropia subfamily has only been identified within Diopsini species (Table 1), but its absence in both the Sphyracephalini and Centrioncinae is currently mainly posited on small-scale PCR screens which may not uncover low copy number, or non-functional degenerate, subfamilies. Finally, only two Teleopsis species have been shown to harbour mosellana elements, consistent with a late origin of the subfamily in the diopsids. The subfamily has not been amplified through PCR in any diopsid species or the earlier screen of H. vulgaris genomic DNA [46]. The WVPHEL amino acid motif on which the Robertson [24] forward degenerate mariner primer was designed is not present in the reconstructed mosellana Tnpase sequences in T. dalmanni, T. whitei and H. vulgaris (S1 Dataset). Conceptual translations of the mosellana tnpase sequences used to generate the subfamily phylogeny in Fig 3 also lack the WVPHEL motif. The homologous region of the Tnpase could be translated in 33 species and 19 encoded the amino acids LVPKEL. The Robertson primers appear to lack specificity to mosellana elements in a broad range of species; therefore the absence of amplified PCR product may be the result of failed primer binding rather than the absence of mosellana elements in a species’ genome. The loss of the highly conserved WVPHEL motif will require the design of alternative degenerate primers to amplify mosellana elements from genomic DNA. The lack of the mosellana subfamily from the transcriptome of T. quinqueguttata is consistent with an origin of the subfamily in Teleopsis after the lineage leading to T dalmanni and T. whitei split from the T. quinqueguttata lineage. It remains however possible that the mosellana subfamily has greater antiquity in Teleopsis, and perhaps other diopsid taxa, and has undergone stochastic loss in T. quinqueguttata.

The mariner subfamilies present in the T. dalmanni genome have persisted for sufficient time in order to generate non-autonomous MITE families. The MITE complement is dominated by transposons generated from Tdmar4 of the mosellana subfamily, however all four subfamilies have produced multiple MITE families. The available data suggest that the ancestral diopsid possessed a diverse complement of mariner elements, with a minimum of four subfamilies residing in the genome. Due to the limitations of small-scale PCR screens and limited whole genome availability across the Diopsidae, the origins of the cecropia, mauritiana and mosellana subfamilies are unresolved and they may be either ancestral or more recent acquisitions into diopsid genomes. Stochastic loss of mariner subfamilies has occurred within the diopsids, as can been seen by the absence of any active copies in the sequenced genome of S. brevicornis, as well as the absence of mariner elements in PCR screen of the genomic DNA of S. beccarii reported in Carr [16]. S brevicornis is a closer relative of S. europaea, a species which possess multiple mariner subfamilies, than S. beccarrii (Fig 1), indicating the absence of mariner elements in S. beccarrii and S. brevicornis is due to independent losses.

Within T. dalmanni the irritans, mauritiana and mosellana subfamilies are all expressed in both the male and female germline. This finding is consistent with mariner transposition occurring in both sexes, unlike the sex-restricted transposition of some TE families, such as copia and Doc, observed in D. melanogaster [53].

The vertumnana subfamily diopsid-Bactrocera horizontal transfer event

The phylogeny presented here recovers a nested position of the Bactrocera elements within those of the Diopsini in the vertumana subfamily (Fig 4). Carr [16] proposed a horizontal transfer event within New Guinea from either a Teleopsis species, or related diopsid, to Bactrocera, based primarily upon the phylogeny of the mariner elements, but also the then recognised distribution of Teleopsis and Bactrocera species. More recently, Feijen and Feijen [54] stated that the direction of horizontal transfer should be re-evaluated, as they considered Teleopsis species to be absent from New Guinea. The enlarged vertumnana phylogeny presented here, with additional diopsid and Bactrocera mariner sequences, provides an ideal opportunity to reassess the inheritance of the subfamily. The increased diversity of Bactrocera mariner elements from the vertumnana subfamily highlights that the proposed horizontal transfer event did not occur into B. neohumeralis, but within an ancestor of at least five Bactrocera species. The monophyly of the Bactrocera elements, which are nested within the paraphyletic Diopsini vertumnana elements with strong support, points toward the diopsids being the donor group and Bactrocera being the recipients. Feijen and Feijen’s [54] alternative argument failed to take into account the required horizontal transfer of a vertumnana mariner from the Australasian B. neohumeralis into African Diasemopsis species if the hypothesis of Bactrocera being the donor group was correct. The revised phylogeny presented here provides additional evidence for the Diopsidae being the donor to Bactrocera. The putatively recipient Bactrocera species are present across South Asia [55] and not confined to Australasia, thereby expunging the argument that the direction of transfer could not have been from diopsids to Bactrocera due to the absence of Teleopsis from New Guinea. The presence of the vertumana subfamily in additional African diopsid genera, in Diopsis and Teloglabrus, would require a further two independent, intercontinental horizontal transfer events, under the Bactrocera to Diopsidae horizontal transfer route. The phylogeny presented here requires a single horizontal transfer event from Diopsini to Bactrocera within South East Asia, as was the case in the original Carr [16] phylogenetic tree. However the alternative route from Bactrocera to Diopsidae requires a minimum of four independent horizontal transfer events, into the genera Diasemopsis, Diopsis, Teleopsis and Teloglabrus.

Horizontal transfer events of mariner from insects to H. vulgaris

The genome of H. vulgaris is rich in TEs, with approximately 57% of the genome being made up from over 500 TE families [49]. The H. vulgaris genome is considerably larger than that of its distant congener H. viridis [56] and the difference has been speculated to be the result of bursts of transposition by multiple TE families [49, 56, 57]. DNA transposons contribute to 21% of the H. vulgaris genome, with mariner elements alone making up 4% of the sequenced genome [49]. PCR screens of the genomes of H. vulgaris, including the North American subspecies/sister-species H. littoralis, have identified mariner elements from the capitata, cecropia, irritans and mauritiana subfamilies [46].

Two of the mariner subfamilies identified here in diopsid species are also present in the genome of H. vulgaris, these being the mauritiana and mosellana subfamilies. Orthologous mariner elements appear to be absent from the sequenced genomes of both H. oligactis and H. viridis, consistent with their invasion of Hydra occurring after the H. vulgaris lineage split from other Hydra approximately 21–28 million years ago [58]. The presence of the subfamilies in the ancestors of either all three Hydra species or only H. vulgaris and H. oligactis is a less parsimonious explanation, which would require multiple stochastic loss events in addition to the gains of the two subfamilies through horizontal transfer. The mosellana subfamily, designated Hvmar1, was not identified in a previously published PCR screen in either H. vulgaris or H. littoralis genomic DNA [46], however this may be due to the lack of subfamily primer specificity due to the loss of the WVPHEL amino acid motif. The phylogeny of the mosellana subfamily presented here indicates a horizontal transfer event from an unknown insect donor into H. vulgaris (Fig 3). The lack of a clear donor species, or even donor insect order, has resulted in Hvmar1 being placed on a relatively long branch within the subfamily phylogeny; it is therefore unclear as to whether the horizontal transfer was an ancient or more recent event. The transfer of Hvmar1 appears to have been a successful one. The tnpase is expressed across multiple cell types and, based upon the ITR/UTR phylogeny which showed 62 identical paralogous copies, Hvmar1 is currently transposing within the H. vulgaris genome.

In contrast to the mosellana subfamily, the mauritiana subfamily, designated here as Hvmar2, was amplified by Robertson with clone Hydra.vulgaris.6 [46]. The relationship of Hydra.vulgaris.6 to other mariner elements was not robustly resolved in the Robertson phylogeny, but it was nested within a clade of insect mariner elements. No diopsid mariner elements were included in the phylogeny, with Hydra.vulgaris.6 clustering with mauritiana elements from D. mauritiana and the hymenopteran Myrmecia occidentalis. The mauritiana phylogeny presented here robustly nests Hvmar2 within the diopsid elements and highlights a putative horizontal transfer between an unknown Teleopsis species and H. vulgaris. The Tnpases of the mauritiana subfamily were shown by Carr [16] to be evolving under purifying selection on their amino acid sequences, indicating the elements are active and therefore potentially viable donors. The mauritiana horizontal transfer into H. vulgaris has also been successful, with Hvmar2 tnpase expression observed across body tissues and multiple identical paralogous insertions identified in the whole genome sequencing reads.

The mechanism, or mechanisms, that have facilitated the horizontal transfer events from insects into H. vulgaris are difficult to envisage. Aquatic cnidarians and terrestrial insects sharing mutual parasites or viruses appears to be unlikely, given their approximately 600 million year divergence time and different habitats. Terrestrial insect larvae or imagos which fall into the water column may be preyed upon by Hydra, which are known to feed upon dipteran larvae and can engulf prey items in excess of 30mm in length [59]. As Teleopsis species, as well as members of other diopsid genera, often live over bodies of water [29, 30, 60] opportunistic predation may potentially allow diopsid mariner DNA to be taken up by Hydra cells resulting in horizontal transfer.

The lack of mariner-derived MITEs is consistent with both Hvmar1 and Hvmar2 being recent invaders in the H. vulgaris genome and contrasts with the proliferation of MITE families in T. dalmanni. The absence of orthologous families of Hvmar1 and Hvmar2 in the whole genome contigs of both H. viridis and H. oligactis suggests that the horizontal transfer events from insect donors occurred within the H. vulgaris species complex. The lack of available sequence data means that the approximate age of the mosellana transfer cannot be gauged, however the very low nucleotide divergence (~2.5%) between Tdmar2 and Hvmar2 suggest that the mauritiana transfer occurred very recently in the evolutionary history of Hydra. A donor Teleopsis species has not been identified, with the Hvmar2 tnpase showing 98.2% and 98.5% nucleotide identity to the tnpase sequences of Tdmar2 and Twmar2 respectively. Teleopsis species are restricted to eastern Asia, with many species present in south east Asia [30, 54] therefore a broader screen of Teleopsis species will be required in order to determine if the genus harbours the donor species. The subspecies, or strain, of H. vulgaris which has been shown to harbour Hvmar2 in its genome, H. magnipapillata, was isolated from Japan and closely related populations are present in south east Asia [58], overlapping with Teleopsis species and indicating a possible east Asian location for the mauritiana subfamily horizontal transfer event. The H. vulgaris AEP strain is a North American laboratory-produced line, generated through a cross of strains from California and Pennsylvania [58]. The presence of Hvmar2 in the transcriptome reads from the AEP strain highlights the transcontinental movement of this mauritiana element in the H. vulgaris global population.

Conclusions

Sequencing of mariner elements from the basal centrioncid Te. entabensis points to a minimum of four subfamilies being present in the ancestral diopsid. A total of seven subfamilies have now been identified within Diopsidae genomes. The identification of the mosellana subfamily in whole genome sequence data highlights the inherent dangers of relying upon degenerate primers in PCR screening for mariner elements, as the widely used primers designed by Robertson [37] do not amplify this subfamily. Two diopsid mariner subfamilies appear to have undergone horizontal transfer to species outside of the family. One of the putative recipient species, H. vulgaris, has also acquired a mariner element from a second, unidentified, insect donor. Despite the great evolutionary distance between insects and cnidarians, both transferred mariner elements have successfully proliferated in Hydra contributing to the diverse TE complement of this species.

Supporting information

S1 Dataset. Annotated sequences of the mariner subfamilies characterized in S. brevicornis, Teleopsis species and Hydra vulgaris.

The full-length sequences for each identified subfamily are presented, along with putative open-reading frames, untranslated regions and flanking repeats. Conceptual translations of encoded proteins are provided.

(TXT)

S2 Dataset. Alignments used in the phylogenetic analyses.

Alignments are provided in the Newick format. Columns present within square brackets were excluded from the phylogenetic analyses.

(TXT)

S1 Fig. Maximum likelihood phylogeny of the T. dalmanni mariner sequences uncovered in whole genome shotgun contigs using the customized mariner RepeatMasker library.

The phylogeny was constructed from 1099 aligned nucleotide positions using the GTRCAT model, and estimated nucleotide frequencies. Values for mlBP and biPP are shown above and below the branches respectively. 100% mlBP and 1.00 biPP are both denoted by “*”. Values <50% mlBP and <0.70 biPP are denoted by “-”. The scale bar represents the number of substitutions per site. Individual mariner subfamilies are bracketed and colour-coded.

(PDF)

S2 Fig. Maximum likelihood phylogeny of diopsid irritans tnpase sequences.

The phylogeny was constructed from 500 aligned nucleotide positions using the GTRCAT model, and estimated nucleotide frequencies. The phylogeny layout is the same as in S1 Fig.

(PDF)

S3 Fig. Maximum likelihood phylogenies of mariner 5’ ITR/UTR sequences within the H. vulgaris genome.

Phylogenies were generated using the GTRCAT model with empirical base frequencies. A) Hvmar1 created from 246 aligned nucleotide positions, B) Hvmar2 created from 291 aligned nucleotide positions. OTU labels are the 5’ flanking DNA of the ITR. The phylogeny layouts are otherwise the same as in S1 Fig.

(PDF)

S4 Fig. Maximum likelihood phylogenies of mariner 5’ ITR/UTR sequences within the T. dalmanni genome.

Phylogenies were generated using the GTRCAT model with empirical base frequencies. A) Tdmar2 created from 187 aligned nucleotide positions, B) Tdmar3 created from 204 aligned nucleotide positions, C) Tdmar4 created from 197 aligned nucleotide positions. OTU labels are the 5’ flanking DNA of the ITR for A and B. C is presented as a radial tree and both the support values and OTU labels are omitted due to the large number of sequences. The phylogeny layouts are otherwise the same as in S1 Fig.

(PDF)

S5 Fig. Representative diopsid phylogeny showing the latest possible points of origins of mariner subfamilies.

Circles represent the putative origin points of the subfamilies. The tree layout is the same as Fig 1.

(PDF)

S1 Table. mariner sequences generated in this study.

(DOCX)

S2 Table. Number of RNA-Seq reads mapped to mariner tnpase sequences.

(DOCX)

S3 Table. SRA RNA-Seq files used in gene expression analyses.

(DOCX)

Data Availability

All newly generated sequences have been deposited in GenBank and the accession numbers stated in the manuscript.(accession numbers MN719915-MN719940). Newly assembled transposon sequences are provided in the Supporting Information file S1 Dataset. Alignments used in the phylogenetics analyses are provided in the Supporting Information file S2 Dataset.

Funding Statement

The author(s) received no specific funding for this work.