Treasures in the attic: rolling circle transposons discovered in eukaryotic genomes.

S the advent of methodologies to analyze the content of whole genomes (e.g., renaturation kinetics and Cot analysis), it has been known that a large fraction of eukaryotic genomes is highly repetitive (1, 2). Recent computer-assisted analysis of several sequenced eukaryotic genomes, including Caenorhabditis elegans, Drosophila melanogaster, Arabidopsis thaliana, and humans, has demonstrated that most repetitive DNA is composed of or derived from transposable elements (TEs). In the human genome, for example, TEs are the single most abundant component, accounting for over 40% of the total DNA (3). Although this amount of TEs is viewed as a hindrance to those engaged in the determination and assembly of DNA sequence, the availability of both complete and partial eukaryotic genome sequences is providing TE biologists with a bonanza of raw material that is being used to understand how genomes evolve. Before the report in PNAS by Kapitonov and Jurka (4), all eukaryotic TEs were thought to use one of two mechanisms for transposition. Class 1, or retrotransposons, transpose via an RNA intermediate in reactions catalyzed by element-encoded proteins, including reverse transcriptase. In contrast, the transposon itself is the intermediate for class 2 elements where an element-encoded transposase catalyzes reactions, resulting in TE excision from one site and reinsertion elsewhere in the genome (the so-called cut-and-paste mechanism). In addition to these two mechanisms, some prokaryotic TEs (called IS or insertion sequences), move by another mechanism called rolling circle (RC) transposition (5, 6). This process is similar to the RC replication of some plasmids, single-stranded (ss) bacteriophage, and plant geminiviruses. In a recent issue of PNAS, Kapitonov and Jurka (4) report that RC transposons also occur in eukaryotes where, surprisingly, they comprise about 2% of the genomes of A. thaliana and C. elegans. How could a group of TEs that account for such a large fraction of the genomes of these well-studied organisms remain until now essentially unknown? One answer to this question is that RC transposons have distinct structural features that are not easily detected by computer-assisted searches of DNA sequence databases. Helitron families of elements (as the eukaryotic RC transposons are called) do not generate target site duplications on insertion, as do all other eukaryotic TEs. These short duplications are derived from staggered endonucleolytic cleavage of the target DNA by elementencoded transposase or integrase. Instead, Helitrons target the dinucleotide AT, and insertion does not lead to the duplication of this sequence. Similarly, RC transposons do not have terminal inverted repeats, as do all other class 2 elements. Rather, Helitrons begin with a 59 TC and end with a 39 CTRR (Fig. 1a). Although there is a 16to 20-nt palindrome just upstream of the 39 CTRR, conservation of palindrome structure but not sequence would apparently preclude the use of a consensus sequence in the identification of Helitrons by computer-assisted searches. By analogy to RC mechanisms in prokaryotes, the distinct structural hallmarks of Helitrons are hypothesized to be essential for RC-mediated transposition (Fig. 1). Helitrons may also have escaped classification for so long because the vast majority of family members are nonautonomous, defective elements that resemble internal deletion derivatives of their cognate autonomous element. It is important to note that up to 10 homogeneous subfamilies of nonautonomous Helitrons, with members ranging from 0.5 to 3 kb, were previously identified in the Arabidopsis genome as abundant repeats. These elements were first designated AthE1 (7) and AtREP (8) and, later, Basho (9). However, in the absence of any obvious structural features of either class 1 or class 2 elements, these repeat families remained mysterious and unclassified. It was only when the complete genome sequence of Arabidopsis became available that Kapitonov and Jurka (4) were able to identify the much less abundant but very large (5.5 to 15 kb) Helitrons that have coding capacity for products related to RC replication proteins. Although rare in prokaryotes, nonautonomous elements are common and abundant members of most eukaryotic transposon families. They are usually internally deleted derivatives of autonomous members and lack coding capacity for the transposase. Because most DNA transposon families contain distinct groups of nonautonomous elements that are conserved in both sequence and length, it is likely that most subfamilies arose from a single or a few deleted copies that were subsequently amplified with enzymes encoded in trans by an autonomous element. This seems to be the case for the RC-transposing Helitrons, because homogeneous groups of defective elements sharing their termini with autonomous copies are abundant in the A. thaliana, Oryza sativa (rice), and C. elegans genomes. Although nonautonomous RC transposons have not been reported in prokaryotes, engineered nonautonomous copies of the Escherichia coli RC element IS91 transposed at high frequency when supplied with transposase in trans (5, 6). What is still mysterious is how the RC mechanism generates nonautonomous elements. For other eukaryotic class 2 elements, it has been shown that such defective copies can arise by incomplete double strand gap repair after excision of an autonomous element (10–12). It is unlikely that a similar mechanism can account for the origin of nonautonomous Helitrons because they presumably do not excise as doublestranded molecules and thus do not create a double strand gap at the donor site. Nevertheless, recombination and slippage during the copying of the transposed single strand at the donor site may account for the origin of internally deleted Helitrons (see Fig. 1b). Alternatively, nonautonomous Helitrons may form de novo from host sequences given the minimal cis requirements that appear to be necessary for RC-mediated transposition. Other open questions concern the function and origin of the putative genes encoded by the larger Helitrons. The preliminary analysis of Kapitonov and Jurka (4) suggests that Helitrons from A. thaliana, O.