The initial search of the M. truncatula genome aimed at the identification of putative autonomous elements, i.e. those carrying an ORF showing homology to the predicted MtMaster TPase (transposase) protein sequence 14 and flanked with terminal inverted repeats of at least 14 bp, containing the G(N)₅GTT motif, and followed by a 3 bp-long TSD (TAA or TTA). This resulted to 44 sequences showing significant homology (E-value < 10^-20) to the TPase, after eliminating the redundancy coming from overlapping BACs. We obtained precisely the same hits using the whole TPase sequence and the DDE region, likely because of the very rigorous E-value threshold imposed during the search. Of the identified sequences, 22 were flanked by TIRs and TSDs characteristic for PIF/Harbinger-like elements and these were assumed to represent complete transposable elements. They ranged in length from 2,180 to 25,288 bp. In 11 of these elements, another coding region, similar to the MtMaster orf1 with E-value ranging from 10^-4 to 10^-99, could be found. The relative order of both ORFs was variable – five elements had orf1 upstream and six downstream the TPase (Table 1).

A phylogenetic analysis of the DDE domain region of the TPase revealed that the M. truncatula PIF/Harbinger-like elements could be divided into five lineages. Nine elements, including the previously described MtMaster, were grouped into lineage M, together with carrot DcMaster 14. In seven of these elements the orf1 preceded the TPase as expected, while for the remaining two the orf1 was absent, most likely because of an internal deletion. Eight elements formed a new lineage designated as A6. Typically for the group A, the orf1 was located downstream the TPase in the five elements carrying both coding regions. Another new lineage, designated as E, was formed by two elements. In none of them could the orf1 be identified. Two other elements were included into lineage A5, together with maize ZmPIF 12 and one was placed into lineage D (Figure 1). However, in the latter case the orf1 was located downstream to TPase, contrary to previously described elements from that lineage 15.

In addition to the TPase-containing elements described above, using a strategy outlined in the Methods section, we identified additional 67 elements lacking any coding capacity and thus considered as non-autonomous. List of all identified elements and their coordinates are given in the Additional File 1. The grouping of the identified transposable elements was based on the full element sequence similarity or 5' and 3' terminal sequence similarity using two approaches: hierarchical clustering and multidimensional scaling (Additional File 2). This strategy allowed us to define families and subfamilies of PIF/Harbinger-like transposable elements in M. trunactula (Table 2), where families essentially reflected the lineages previously identified on the basis of the TPase phylogeny, and subfamilies grouped elements carrying homologous TIRs (Table 3) and showing a degree of overall DNA sequence similarity. For each but two subfamilies, one or two putatively autonomous core elements could be identified. The exception was a low copy number family MtPH-E, for which none of the elements contained a region homologous to the orf1.

The largest family, MtPH-A6, contained 54 elements, while family MtPH-D was represented only by a single element. The second most abundant family, containing 27 elements, was MtPH-M (Master), of which 18 was grouped into subfamily 3.

MtPH-A6 consisted of four subfamilies represented by putative autonomous elements sharing similar ORF organization, i.e. a TPase containing two introns, followed by orf1. MtPH-A6 TPases formed a well supported clade, containing four subclades with high bootstrap values, representing the corresponding subfamilies (Figure 1).

Subfamily MtPH-A6-1 contained nine elements ranging in length from 802 to 8,707 bp, the longest element carrying a nested insertion of the 7,555 bp long RAM12 gypsy-like retrotransposon.

Subfamily MtPH-A6-2 grouped six elements, 898 to 4,218 bp long, all being simple internal deletion derivatives of the core element MtPH-A6-2-Ia.

Sixteen elements belonged to subfamily MtPH-A6-3, ranging in length from 553 to 4,845 bp, except for a much larger, 23,892 bp long element MtPH-A6-3-IIIa, initially identified as being flanked by 15 bp TIRs unrelated to those of the MtPH-A6-3 subfamily. However, the element contained a 4.9 kb region 74% identical to the two elements mentioned above, but lacking the first 8 bases in the 5' TIR (Figure 2). Hence, the true boundaries of the elements could not be initially identified using our mining strategy. An interesting feature of that subfamily was the presence of a perfect microsatellite site in the first intron of the TPase. The three elements containing the region coding for the TPase, MtPH-A6-3-Ia, MtPH-A6-3-IIa, and MtPH-A6-3-IIIa had, respectively, 27, 8, and 21 repeats of the (TA)_n core motif (Figure 2).

MtPH-A6-4 subfamily members ranged in length from 431 to 25,288 bp. Among the 23 members of that subfamily, 18 were characterized by the presence of imperfect 60 bp long tandem repeats, variable in number, while in the remaining five elements the core repeat was entirely absent. Each repeat itself contained a triplicated AAACNNCTTATT motif. These elements contained from 2 to 35 repeats that in extreme cases covered almost the entire region between the TIRs (Figure 3A and 3B). In some elements, tandem repeats were present only in one subterminal region, while for the others they were present in both subterminal regions in opposite orientation. The 60 bp tandem repeats were identified in 27 other sites in the M. truncatula genome, initially not identified as occupied by MtPH-A6-4 elements. However, BLAST search with the terminal 214 bp + 3 bp TSD of the MtPH-A6-4-Ia and MtPH-A6-4-IIa elements indicated that in all instances at least one of the regions flanking the repeats showed residual homology to the TE terminus (E value < 1e-08). The presence of tandem repeats facilitated internal rearrangements resulting in inversions of the internal region (Figure 3C). Two nested insertions were identified in the longest element MtPH-A6-4-IIa, which showed three blocks of significant homology to the MtPH-A6-4-Ia core element, interrupted by an unidentified element of 2,191 bp carrying 15 bp TIRs and flanked by a 5 bp long TSD and a gypsy-like retrotransposon (Figure 3D).

MtPH-M family included three subfamilies with short (14 bp), similar TIRs and orf1 followed by TPase. Subfamily MtPH-M-1 contained only four elements, ranging in length from 812 to 5,140 bp. Two of them, MtPH-M-1-Ia (previously described as MtMaster 10) and MtPH-M-1-IIa (showing 90% overall sequence identity to MtMaster) had both ORFs, and the remaining two were internally deleted derivatives.

Five elements were grouped into subfamily MtPH-M-2, three of them carrying both orf1 and TPase. The region containing element MtPH-M-2-IIa occurred to be a composite structure of two related TEs. The initially identified sequence flanked by TIRs and TSDs spanned over 21,696 bp. The 5,303 bp element MtPH-M-2-IIa occupied the 5' region of that sequence, however the downstream sequence also contained blocks of homology to the core element MtPH-M-2-Ia, and a nested insertion of a Gypsy-like retrotransposon (Figure 4). It indicates that an ancient copy of a TE related to those belonging to the subfamily MtPH-M-2 became a target for subsequent nested insertions. Other elements from that family ranged in length from 2,240 to 7,816 bp.

Subfamily MtPH-M-3 was the largest within the family and contained 18 elements, of which two carried both ORFs. Their length varied from 442 to 4,048 bp, and interestingly, two 442 bp-long elements were 100% identical. As their length resembled that of miniature inverted repeat elements (MITEs), but unlike MITEs, their number in the M. truncatula genome was low, it would be tempting to speculate that these copies might become founders of a new MITE family. A slightly more advanced stage of proliferation of MITE-like elements could be observed with a group of 10 short (776–905 bp) elements from the same family. A more detailed comparison of the element sequences provided a further insight into the evolution of MtPH-M-3 subfamily. Internal deletions were accompanied by differentiation and rearrangement of variant sequences (blocks A, B, and C in Figure 5, Additional File 3) in the subterminal regions. Two lineages could be traced that originated from the core element MtPH-M-3-Ia, that included respectivley 5 and 11 elements. The element MtPH-M-3-VI showed apparently a mosaic structure, as it contained the 3' subterminal region from lineage I, while the major portion of the element contained sequence variants chracteristic for the lineage II (Figure 5).

Family MtPH-A5 was represented by four elements ranging in length from 1,182 to 6,770 bp. The two putative autonomous elements were 72% similar over the entire sequence, but within the coding region the nucleotide sequence similarity reached 95%. Two shorter elements were deletion derivatives of full-length elements. Interestingly, a recently reported MITRAV family of miniature elements of barrel medic 22 showed a high nucleotide sequence similarity of their termini to the MtPH-A5 elements, spanning over ca. 40 bp on both ends of the element.

Family MtPH-E consisted of three elements, none of which carried both ORFs. The elements ranged from 1,508 to 3,957 bp. The two largest elements were very similar, differing by one indel, while the similarity of the shortest element to the other two was restricted only to the 180 bp of the 5' terminus and 70 bp of the 3' terminus.

Family MtPH-D was represented by a single element of 3,160 bp, carrying both ORFs. However, their orientation was opposite to that of typical PIF/Harbinger-like elements representing the D lineage 15. Its localization in the D lineage was not strongly supported by bootstrap analysis (Figure 1). The fact that no internally truncated elements were identified could suggest that the element might be capable of perfect excision, not triggering the process of abortive gap repair.

In order to find evidence for a possible mobility of identified elements we implemented a strategy proposed by Le et al. 8, i.e. we searched for regions, called RESites (Related to Empty Sites), paralogous to sequences flanking the insertion sites, but lacking the transposable element. We identified 11 RESites, of which five represented insertion sites of non-autonomous elements belonging to the MtPH-A6-4 subfamily, while two and one of them were related to non-autonomous elements of the MtPH-A6-3 and MtPH-A6-2 subfamilies, respectively. The remaining three RESites represented insertion sites of the putative autonomous (core) elements belonging to family MtPH-E and subfamilies MtPH-M-2, and MtPH-M-3 (Figure 6).

We identified several M. truncatula ESTs showing high similarity to putative expression products (orf1 and TPase) of the mined autonomous elements (Additional File 4). However, ESTs directly corresponding to the putative expression products, both to the orf1 (CX532696, 641 bp, 94% identity) and the TPase (AW686181, 304 bp, 99% identity), could be detected only in case of elements representing the MtPH-M-1 subfamily (Additional File 5). Interestingly, A number of ESTs similar to non-coding terminal regions of the TEs could also be identified (data not presented).

The PCR assay of MtPH insertion polymorphism was performed on eight M. truncatula populations selected to represent genetic diversity of the species, as proposed by Ronfort 31. Fifty-six insertion sites identified in the reference genome of cv. Jemalong A17 were checked for presence of the TE. Thirty-seven primer pairs yielded products of the expected size for the reference sample, while 11 generated complex profiles, likely indicating that insertions were present in repetitive regions. The remaining eight primer pairs produced ambiguous results. Of the 37 successful amplifications, 20 occurred to be polymorphic. Usually, the size the shorter amplicon corresponded to the predicted size of the product amplified from the unoccupied site. However, amplicons slightly differing from the expected size were also observed, indicating a possible imperfect excision event (Figure 7).

The experiment was performed on the M. truncatula genomic DNA sequence database consisting of 1540 BACs, updated Aug 2005 32. As the size of the whole M. truncatula genome ranges from 500 to 600 Mbp 17 and the average non-overlapping coverage by each BAC was ca. 100 Kb 32, we estimated that the input sequence data amounted 26–30% of whole genome.

The predicted protein sequence of DDE domain and the whole TPase sequence of the previously identified MtMaster element 14 was used as the initial query for a TBLASTN search against the BAC sequence database, using the E-value threshold of 1e-20. The output file was then processed to eliminate redundancy coming from overlapping BACs, and significant hits were extracted, along with up to 30 kb flanking sequences. The extracted sequences were scanned for the presence TIRs and TSDs, using a newly developed tool named TIRfinder, identifying TIRs and TSDs and returning a file with a list of found elements fulfilling user-defined requirements. To provide fast computation on whole genome, the algorithm uses very efficient data structures, such as suffix trees. TIRfinder is an open source software accessible online 33. The program was written in Java and can be run on Windows or Linux.

We allowed up to four mismatches inside 14 bp of the TIRs and no mismatch in TSDs. Another condition was the presence of the conserved G(N)₅GTT motif at the 5' end of the TIR. In silico prediction of the presence of coding regions was performed for all identified sequences using FGENESH 34.

To identify internally deleted copies of elements related to those found previously, 217 bp-long (3 bp TSD + 14 bp TIR + 200 bp subterminal sequence) terminal regions were extracted from all putative autonomous elements. These sequences were used to scan the M. truncatula genomic DNA sequence database (BLASTN, E-value threshold – 1e-10), and regions showing homology to any of the terminal regions were identified. The output was automatically filtered to find sequences of length ranging from 400 to 30,000 bp, flanked with TIRs showing homology to the same autonomous element on both ends. All newly found sequences have been checked whether they contained a region coding for the TPase. All TEs were scanned using Censor 35, to identify the presence of nested elements.

Multiple alignment of 48 transposase sequences of PIF/Harbinger-like transposable elements was obtained using T-Coffee 36. Bootstrap analysis was performed with PHYLIP using seqboot, neighbor, protdist and consense programs 37. The sequence similarity of 89 TEs was analyzed by the hierarchical clustering method and visualized with help of multidimensional scaling. For both tasks we used the R statistical environment 38. As a measure of dissimilarity between sequences we used the E-value of BLAST. Hierarchical cluster analysis of a set of dissimilarities was done by hclust (complete linkage) method 39. Multidimensional scaling 40 visualization is primarily dependent on the analogy of similarity and proximity (and hence of dissimilarity and distance). It re-scales a set of dissimilarity data into distances and produces the low-dimensional configuration that generated them. The visualization for our data was obtained with isoMDS R procedure.

Sequences were visually compared, aligned, edited, and analysed using BioEdit and the included accessory applications 41. Pairwise sequence comparisons were performed using 'blast 2 sequences' 42 and Yass 4344. Dot-plots were generated using Nucleic Acid Dot Plots 45 with a window size of 25 nucleotides and a mismatch limit of 5 positions. Tandem repeats identification was performed using 'mreps' software 4647.

In order to find RESites (Related to Empty Sites) in the M. truncatula genome we performed a computer-based search, essentially as described by Le et al. 8. Briefly, we extracted 1 Kb sequence flanking both sides of each of the mined elements, combined them into one sequence of 2 Kb, and used it as a query for a BLASTN search on the whole BAC sequence database. Hits spanning on both sides of the insertion were considered as those representing RESites.

EST search was performed using nucleotide sequences of the putative autonomous elements, using a BLAST tool run against the M. truncatula EST database 48.

PCR assay was performed on plants representing cv. Jemalong A17 and seven populations from the core M. truncatula collection (CC8, as described by Ronfort et al. 31). Primer pairs were anchored in the regions flanking the mined elements. They were designed using Primer3 49 to obtain amplification of ca. 600 bp long fragment for the putative empty site. Two cycling protocols were employed. For TEs of length not exceeding 2 Kb a standard PCR was performed. The reaction was set up in the volume of 20 μl and contained 0.25 mM each dNTP, 2 mM MgCl₂, 10 pmol of each primer, 1 unit of TAQ polymerase (Fermentas) and 2 μl of the PCR buffer supplied by the manufacturer. The thermal profile of the reaction was as followed: 94°C for 2 min., 35 cycles of: 94°C for 30 s, 53°C for 30 s, and 68°C for 90 s, and completed with 68°C for 5 min. For larger elements we used long PCR protocol. Amplification was performed in the volume of 20 μl containing 0.25 mM each dNTP, 10 pmol of each primer, 0,5 unit of long PCR enzyme mix (Fermentas) and 2 μl of the Long PCR buffer supplemented with MgCl₂ (Fermentas), using the following thermal profile: 94°C for 2 min., 35 cycles of: 94°C for 15 s, 53°C for 30 s, and 68°C for 7 min., and completed with 68°C for 10 min. All reactions were carried out in the Mastercycler or Mastercycler Gradient (Eppendorf). Amplification products were separated on 1% agarose gels and visualized with ethidium bromide under UV.