LTR retrotransposon landscape in Medicago truncatula: more rapid removal than in rice

Abstract

Background: Long terminal repeat retrotransposons (LTR elements) are ubiquitous Eukaryotic TEs that transpose through RNA intermediates. Accounting for significant proportion of many plant genomes, LTR elements have been well established as one of the major forces underlying the evolution of plant genome size, structure and function. The accessibility of more than 40% of genomic sequences of the model legume Medicago truncatula (Mt) has made the comprehensive study of its LTR elements possible.

Results: We use a newly developed tool LTR_FINDER to identify LTR retrotransposons in the Mt genome and detect 526 full-length elements as well as a great number of copies related to them. These elements constitute about 9.6% of currently available genomic sequences. They are classified into 85 families of which 64 are reported for the first time. The majority of the LTR retrotransposons belong to either Copia or Gypsy superfamily and the others are categorized as TRIMs or LARDs by their length. We find that the copy-number of Copia-like families is 3 times more than that of Gypsy-like ones but the latter contribute more to the genome. The analysis of PBS and protein-coding domain structure of the LTR families reveals that they tend to use only 4-5 types of tRNAs and many families have quite conservative ORFs besides known TE domains. For several important families, we describe in detail their abundance, conservation, insertion time and structure. We investigate the amplification-deletion pattern of the elements and find that the detectable full-length elements are relatively young and most of them were inserted within the last 0.52 MY. We also estimate that more than ten million bp of the Mt genomic sequences have been removed by the deletion of LTR elements and the removal of the full-length structures in Mt has been more rapid than in rice.

Conclusion: This report is the first comprehensive description and analysis of LTR retrotransposons in the Mt genome. Many important novel LTR families were discovered and their evolution is elucidated. Our results may outline the LTR retrotransposon landscape of the model legume.

Figure 1

RT phylogenetic tree of 74 LTR families. A Bel-Pao type RT (BEL-1-I_NVp from Repbase) is used as outgroup. The 74 families are grouped into Copia or Gypsy superfamily. In the tree, each family is described by its name and a superfamily label. The superfamily label is given according to the order of domains in the POL. RT similarity and domain organization give consistent categorization. Mtr3, Mtr5 and Mtr62 lack other domains except RT, so they are categorized directly though RT similarity and are marked by the lowercase initials of the superfamilies. The 14 clades, to which the 74 families belong, are shown in the figure. The placement of Mtr3 and Mtr39 is unresolved and they are marked by grey dots [see Additional file 4].

Figure 2

Structure of LTR families. Each sub-figure gives the structure of a family. the X-axis displays coordinates of nucleotides and the Y-axis displays average similarities among the full-length members of that family (calculated using the PLOTCON program in the EMBOSS package [46]). Grey stripes show the positions of the LTR and the domains detected. We display ORFs that are >500 bp in length. The arrow under a ORF label represents the length of that ORF. In Mtr67, the ORFs are found in both chains and their orientation is indicated by the arrows above the ORF labels. Sudden collapse of similarity (e.g. 1.8–2 Kb of Mtr8 and 2–3.5 Kb of Mtr10) is caused by the insertion or deletion events in one or two family members.

Figure 3

Half-life of full-length LTR retrotransposons in Mt and rice. 526 Mt and 705 rice full-length elements are analyzed. Each bin represents 0.1 MY. Vertical bars under the histogram represent insertion events. a) The distribution of the insertion date of Mt elements. Fitting of this distribution to a exponential curve indicates that the insertions in the recent 0.1 MY have been significantly active. b) Fitting the dates to the exponential curve. The logarithm of the dates fits the straight line y = 0.52 - 2.71x well. Therefore the rate of the exponential curve is α = -2.71, which corresponds to a half-life of 0.26 MY. c) and d) display the fitting in rice, which gives a half-life of 0.4 MY.

Figure 4

Homologous matches of an candidate. The lines represent matches generated by whole-genome homology search of a reference candidate. Some matches are made of several pieces (segments on the same horizontal line). All the matches are categorized into Part I and Part II. Part I (Rset) consists of the matches that cover both the LTR region and the internal domain. They are reliable copies of the reference candidate. Part II is further classified into pseudo-copies and "copies in part II". Pseudo-copies are the matches that correspond to unrelated sequences. Unrelated sequences (dark grey regions) are the subregions that have significantly high matches (grey stripes) or that match some LTR elements well (not showing here). At last, "copies in part II" and Rset are combined to obtain the total copies of the candidate.