PIF- and Pong-Like Transposable Elements: Distribution, Evolution and Relationship With Tourist-Like Miniature Inverted-Repeat Transposable Elements

Corresponding author: Susan R. Wessler, University of Georgia, Athens, GA 30602., sue{at}plantbio.uga.edu (E-mail)

Communicating editor: M. J. SIMMONS

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Miniature inverted-repeat transposable elements (MITEs) are short, nonautonomous DNA elements that are widespread and abundant in plant genomes. Most of the hundreds of thousands of MITEs identified to date have been divided into two major groups on the basis of shared structural and sequence characteristics: Tourist-like and Stowaway-like. Since MITEs have no coding capacity, they must rely on transposases encoded by other elements. Two active transposons, the maize P Instability Factor (PIF) and the rice Pong element, have recently been implicated as sources of transposase for Tourist-like MITEs. Here we report that PIF- and Pong-like elements are widespread, diverse, and abundant in eukaryotes with hundreds of element-associated transposases found in a variety of plant, animal, and fungal genomes. The availability of virtually the entire rice genome sequence facilitated the identification of all the PIF/Pong-like elements in this organism and permitted a comprehensive analysis of their relationship with Tourist-like MITEs. Taken together, our results indicate that PIF and Pong are founding members of a large eukaryotic transposon superfamily and that members of this superfamily are responsible for the origin and amplification of Tourist-like MITEs.

TRANSPOSABLE elements (TEs), which are a major component of all characterized eukaryotic genomes, have been divided into two classes according to their transposition intermediate. Class 1 (RNA) elements transpose via an RNA intermediate and most either have long terminal repeats (LTR-retrotransposons) or terminate at one end with a poly(A) tract [long interspersed nuclear elements (LINEs) and short interspersed nuclear elements (SINEs)]. Class 2 (DNA) elements transpose via a DNA intermediate and usually have short terminal inverted repeats (TIRs). DNA elements can be further classified into families on the basis of the transposase (TPase) that catalyzes their movement. A TE family is composed of one or more TPase-encoding autonomous elements and up to several thousand nonautonomous elements that do not encode functional TPases but retain the cis-sequences necessary to be mobilized by the cognate TPase (for reviews see CAPY et al. 1998 ; FESCHOTTE et al. 2002A ).

Miniature inverted-repeat transposable elements (MITEs) were first discovered in the grasses and later found in other flowering plants as well as in animal genomes (for review see FESCHOTTE et al. 2002B ). Structurally, MITEs are reminiscent of nonautonomous DNA elements, but their high copy number and intrafamily homogeneity in size and sequence distinguish them from most previously described nonautonomous elements (WESSLER et al. 1995 ). Since MITEs lack coding sequences, their classification has been based on the sequence similarity of TIRs and target site duplication (TSD). Using these criteria, most of the tens of thousands of plant MITEs have been divided into two groups: Tourist-like (3-bp TSDs, usually TTA/TAA) and Stowaway-like (2-bp TSDs, usually TA) (FESCHOTTE et al. 2002B ).

Two distantly related families of active DNA transposons have recently been associated with Tourist-like MITEs. The maize P Instability Factor (PIF) and a Tourist-like MITE family called miniature PIF (mPIF) share identical TIRs, similar subterminal sequences, and a strong preference for insertion into the 9-bp palindrome CWCTTAGWG with duplication of the central TTA (WALKER et al. 1997 ; ZHANG et al. 2001 ). PIF contains two open reading frames (ORFs), one encoding a TPase, whereas the function of the other ORF is unknown (ZHANG et al. 2001 ; Fig 1A). An even closer relationship was found in rice, where the 430-bp Tourist-like MITE called mPing was shown to be a deletion derivative of a 5.2-kb transposase-encoding element called Ping (JIANG et al. 2003 ; KIKUCHI et al. 2003 ; NAKAZAKI et al. 2003 ). Several lines of evidence, however, led to the conclusion that a related element in the rice genome, called Pong, was the most likely source of TPase mobilizing mPing elements (JIANG et al. 2003 ). Indeed, Pong elements and mPing MITEs were found to be actively transposing in the same cell culture line. Similar to PIF, Pong also contains two ORFs (ORF1 and ORF2). ORF1 may be involved in DNA binding as it includes a domain with weak similarity to the DNA-binding domain of myb transcription factors (JIANG et al. 2003 ; Fig 1B). ORF2 most likely encodes the TPase, as it contains an apparent DDE motif, a signature consisting of three acidic residues found in the catalytic domains of some eukaryotic and prokaryotic TPases (REZSOHAZY et al. 1993 ; MAHILLON and CHANDLER 1998 ; JIANG et al. 2003 ).

View larger version (106K): In this window In a new window Download PPT slide   Figure 1. Structure and conserved coding regions of PIF- and Pong-like elements. Triangles represent element TIRs. Coding regions are represented by dark gray boxes. Regions that are unrelated between different PIF- or Pong-like families are represented by light gray boxes. DDE motifs in TPases are indicated by asterisks, and open boxes localize regions that were used to deduce the phylogeny of PIF- or Pong-like elements (Fig 2 Fig 3 Fig 4 Fig 5 and Fig 7). (a) PIF-like elements. Multiple alignments of the conserved regions of select PIF-like ORF1's and TPases are shown. Horizontal arrows indicate the position of dPCR primers, and the open and solid arrowheads denote the position of intron 1 and intron 2, respectively. Horizontal lines indicate the predicted HTH domain (H) and three blocks of conserved residues (N2, N3, and C1) that likely comprise the catalytic domain (see text). (b) Pong-like elements. Multiple alignments of the conserved regions from ORF1 and TPase of Pong-like elements are shown. A consensus sequence of the Myb domain is shown above ORF1 alignments for comparison. Horizontal lines in the ORF1 alignments indicate three blocks of conserved residues (A, B, and C). Os-PIF1, At-PIF2, and Pong were previously reported (KAPITONOV and JURKA 1999 ; ZHANG et al. 2001 ; JIANG et al. 2003 ). The PIF TPase shown in a is a full-length TPase isolated from maize (X. ZHANG and S. R. WESSLER, unpublished data). Other sequences were named according to the species initials (see Table 1), followed by their GenBank accession number.

PIF and Pong elements share several features, including amino acid sequence conservation in the catalytic domain of ORF2, homology in their ORF1's, nucleotide sequence similarity in their TIRs, and identical TSDs (TTA; JIANG et al. 2003 ). The putative TPases of both PIF and Pong have a large number of homologs in GenBank that have been annotated as unknown/hypothetical proteins. Examination of several PIF TPase homologs from plants, nematodes, and a fungus showed that they resided in PIF-like transposable elements (ZHANG et al. 2001 ). Furthermore, PIF-like and Pong TPases were shown to be distantly related to the TPases of bacterial insertion sequences of the IS5 group (KAPITONOV and JURKA 1999 ; LE et al. 2001 ; ZHANG et al. 2001 ; JIANG et al. 2003 ). Thus, PIF and Pong are the founding members of what appears to be a new and widespread superfamily of DNA transposons called PIF/IS5. Significantly, some of these PIF-like elements were found to be associated with Tourist-like MITEs in their respective genomes (LE et al. 2001 ; ZHANG et al. 2001 ; APARICIO et al. 2002 ; FESCHOTTE et al. 2002B ).

In this study we look at the distribution and evolution of the PIF/IS5 superfamily of transposases and characterize their relationship with Tourist-like MITEs. To this end we conducted a systematic survey (database searches and PCR assays) of putative PIF- and Pong-like TPases in plants and animals. Phylogenetic analyses of >600 TPase fragments from 56 species define three major groups, each represented by multiple ancient and distinct lineages. The availability of virtually the entire sequence of rice (GOFF et al. 2002 ; YU et al. 2002 ) permitted the identification and characterization of all PIF- and Pong-like elements in a single genome. Furthermore, the association between rice PIF- and Pong-like elements and Tourist-like MITEs was explored by performing a genome-wide comparison of these elements. This represents the first comprehensive analysis of the origin of Tourist-like MITEs in any organism.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

PCR amplification of PIF-like TPases:
Degenerate primers were derived from the regions encoding amino acid residues GALDGTH (D1F1, 5'-GGIGCHHTIGATGGHACWCA-3'; I, Inosine; H, A, C, or T; W, A, or T) and ELFNPRH (KR1, 5'-ATGICKMIRRTTRAACAAYTC-3'; K, G, or T; M, A, or C; R, A, or G; Y, C, or T; Fig 1A, positions indicated by arrows). PCR amplifications were performed with 10–100 ng of genomic DNA in 50-µl reactions. Cycling parameters were: 1 cycle at 94° for 3 min, 36 cycles at 94° for 30 sec, 50° for 30 sec, 72° for 1 min, and 1 cycle at 72° for 5 min. Forty microliters of the reaction was resolved on 1% agarose gels, and desired fragments were purified from agarose using the QIAquick gel extraction kit (QIAGEN, Valencia, CA) and cloned using the TOPO-TA cloning kit (Invitrogen, San Diego) according to manufacturers' instructions. Sequencing reactions were performed by the Molecular Genetics Instrumentation Facility of the University of Georgia. The sequences of 28 PIF-like TPase fragments described here were deposited in the GenBank database (accession nos. AY362792, AY362793, AY362794, AY362795, AY362796, AY362797, AY362798, AY362799, AY362800, AY362801, AY362802, AY362803, AY362804, AY362805, AY362806, AY362807, AY362808, AY362809, AY362810, AY362811, AY362812, AY362813, AY362814, AY362815, AY362816, AY362817, AY362818, AY362819).

Database searches and sequence and phylogenetic analyses:
Database searches were performed with blast servers available from the National Center for Biotechnology Information (http://www.ncbi.nlm.nih.gov, databases nr, gss, est, and wgs_Anopheles) as well as the rice genomic database at The Institute for Genomic Research (http://tigrblast.tigr.org/euk-blast/index.cgi?project=osa1). Nucleotide sequences obtained from database searches and degenerate PCR (dPCR) amplifications were conceptually translated into amino acid sequences and aligned with CLUSTALW. Introns were predicted with Netgene2 [available at http://www.cbs.dtu.dk (HEBSGAARD et al. 1996 )]. Putative helix-turn-helix (HTH) motifs were predicted using the NPS@ program (available at http://npsa-pbil.ibcp.fr/cgi-bin/npsa_automat.pl?page=npsa_hth.html). Multiple sequence alignments used to generate the phylogenetic tree in Fig 2 were performed with the CLUSTALW server available at European Bioinformatics Institute (http://www.ebi.ac.uk/clustalw/) with default parameters. Multiple alignments used to generate other phylogenetic trees were performed using the MacVector program. Phylogenetic trees were generated on the basis of the neighbor-joining method, using PAUP* version 4.0b8 (SWOFFORD 1999 ) with default parameters. Pictograms were generated at http://genes.mit.edu/pictogram.html.

View larger version (49K): In this window In a new window Download PPT slide   Figure 2. Phylogeny of PIF and Pong-like TPases. The unrooted tree was generated using the neighbor-joining method from a CLUSTALW multiple alignment of the catalytic domain from 619 PIF- or Pong-like TPases identified by database searches or isolated by dPCR (see text). The maize PIF TPase and rice Pong TPase are represented by the solid red circle in their respective group. Bootstrap values were calculated from 500 replicates.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Distribution and abundance of PIF/Pong-like TPases:
Identification of PIF/Pong-like TPases through database searches: A systematic survey was carried out using the TPases of PIF and Pong as queries in tBlastn searches against several public databases. Significant similarity was detected in >1000 entries from a wide range of eukaryotic species, including 21 plants, 19 animals, and two fungi (listed in Table 1). Six hundred and seventy-three hits (574 were unique) with significant homology to the catalytic domains of PIF or Pong were selected for further analysis.

One striking result from database searches was the abundance of sequences for PIF/Pong-like TPases in some organisms, especially plants. This is most apparent in genomes with large amounts of sequence information: for the catalytic domains alone, there were 80 hits (e-value < 10^-23) in Arabidopsis thaliana, 350 hits (e-value < 10^-10) in rice (Oryza sativa c.v. Nipponbare), and 170 hits (e-value < 10^-30) in Brassica oleracea [30% of its 600-Mb genome available for blast at TIGR (http://www.tigr.org/tdb/e2k1/bog1/)]. In animals, the number of sequences for PIF/Pong-like TPases varies from >100 in the African malaria mosquito (Anopheles gambiae) and 300–400 in zebrafish (Danio rerio) to a few (<3) in Drosophila melanogaster, Caenorhabditis elegans, and human.

Nucleotide sequences of the 574 unique PIF- and Pong-like TPases were conceptually translated into amino acid sequences after removal of introns (see below) and judicious correction of frameshifts caused by small (1–2 bp) insertions or deletions. The resulting amino acid sequences were compared to detect conserved regions that might signify functional domains. Several blocks of highly conserved residues were identified for PIF-like TPases (Fig 1A). Block H corresponds to a predicted HTH domain that may be involved in DNA binding. Blocks N2, N3, and C1 most likely comprise the catalytic domain as they contain an apparent DDE motif with the three acidic residues centered in blocks N2, N3, and C1, respectively. A DDE motif is also present in Pong-like TPases (Fig 1B), but, unlike PIF, no HTH domain was predicted.

PIF/Pong-like TPases are usually adjacent to ORF1 homologs: tBlastn searches using the ORF1's of PIF and Pong as queries also yielded a large number of hits. When located on long contigs, these ORF1 homologs were usually found within 1–2 kb of PIF- or Pong-like TPases, indicating that each "pair" of ORF1 and TPase was encoded by the same element. In fact, when the termini of PIF/Pong-like elements were defined in O. sativa (see below), A. thaliana, and A. gambiae (X. ZHANG, C. FESCHOTTE, and S. R. WESSLER, unpublished data), nearly all elements were found to encode both ORFs. ORF1s are significantly more divergent than the TPases. Two blocks of conserved residues were found in PIF/Pong-like ORF1's (Fig 1, blocks A and B), with the most conserved block (block A) centered in a 100-amino-acid (aa) region that displays weak homology to the DNA-binding domains of myb transcription factors from some plants and animals (JIANG et al. 2003 ). Pong-like ORF1's contain an additional well-conserved block (Fig 1B, block C).

Additional PIF-like TPases from grasses: The majority of PIF-like sequences (80%) were from only a few species (rice, Arabidopsis, B. oleracea, and A. gambiae) since this survey was limited by the availability of DNA sequences in databases. To better resolve the phylogeny of PIF elements, additional TPase sequences were isolated from species with established evolutionary relationships but limited sequence information. To this end, a dPCR procedure was employed to amplify PIF-like TPase fragments from selected grass species. Grasses were chosen for this analysis because their phylogeny is well characterized (KELLOGG 2001 ) and they harbor the only known active PIF and Pong elements (WALKER et al. 1997 ; ZHANG et al. 2001 ; JIANG et al. 2003 ).

dPCR primers were derived from the conserved blocks N2 and C1 in PIF-like TPases (see Fig 1A for positions and MATERIALS AND METHODS for sequences) and used to amplify an 120-aa region from 20 grass species as well as several basal monocots (listed in Table 1). The amplified region included the majority of the catalytic domain in PIF-like TPases, extending from 3 aa upstream of the first Asp to 8 aa upstream of the Glu of the DDE motif (Fig 1A, boxed region). PCR products of the expected size (360 bp) were successfully amplified from all 20 grasses tested and their close relative Joinvillea, as well as several Asparagales (e.g., Gongora ilense; data not shown). In addition to the 360-bp fragments, most species yielded larger PCR products (450 bp) that, when sequenced, were found to contain an intron (see below).

Forty-five fragments from 15 species were sequenced; all were unique, indicating that there are multiple distinct TPases in each species and that only a small fraction had been sampled. No product was amplified from the more basal monocots such as Zamia, Ginkgo, or Gnetum. Failure to amplify TPase fragments by dPCR from these species may be due to nucleotide variation in the primer recognition region or to the absence of PIF-like TPases.

Phylogeny of PIF-like and Pong-like TPases:
Three major clusters of PIF- and Pong-like TPases: The TPase fragments identified by database mining and those isolated by dPCR were pooled and their evolutionary relationships examined. A multiple alignment was constructed from the 45 dPCR products and 574 unique database hits and used to generate an unrooted phylogenetic tree (Fig 2). The majority of the sequences clustered into three groups: the plant PIF-like group, the plant Pong-like group, and the animal group. In addition, the five fungal sequences clustered into two small, species-specific groups.

Clustering of the two plant groups was supported by bootstrap values as well as by several features that were shared within each group but not between groups. First, the spacing (i.e., numbers of residues) between the second Asp and the Glu of the DDE motif differed between PIF-like and Pong-like groups but was consistent within each group. PIF-like TPases exhibit DD47E or DD48E spacing whereas Pong-like TPases exhibit an invariant DD35E spacing. Second, the TIRs of PIF- and Pong-like elements contain sequence motifs that are highly conserved within each group but distinct between the two groups (see below). On the basis of the comparison of TPase sequences, the animal group was related equally to both plant groups.

Phylogeny of plant PIF- and Pong-like TPases: Phylogenetic relationships among plant PIF- and Pong-like elements were determined by analyzing a subset of 99 sequences (63 PIF-like, 36 Pong-like) that were selected to represent the different lineages within each group. A CLUSTALW multiple alignment was constructed from the catalytic domains of these sequences and used to generate a phylogenetic tree (Fig 3). Both plant groups are monophyletic and heterogeneous. In each group, amino acid identity between sequences from distantly related species can be higher than that between two sequences from the same or from a closely related species, suggesting the presence of multiple ancient lineages of both PIF- and Pong-like elements.

View larger version (55K): In this window In a new window Download PPT slide   Figure 3. Phylogeny of plant PIF- and Pong-like TPases. The phylogenetic tree was generated using the neighbor-joining method from a CLUSTALW multiple alignment of 99 catalytic domains of PIF- and Pong-like TPases and rooted with the catalytic domain of a PIF-like TPase from F. neoformans. Presence of intron 1 and/or intron 2 in a certain PIF-like lineage is shown as a "1" or "2" in a gray circle. The intron content of lineage A5 is variable (see text). Loss of intron 1 and/or intron 2 from PIF-like TPases (a "-1" or "-2" in a gray circle) is indicated by arrows. The relative organizations of ORF1/TPase for PIF/Pong-like element lineages are shown (arrows indicate direction of transcription). Sequences are named according to the species initial (see Table 1) followed by the GenBank accession number. Bootstrap values were calculated from 1000 replicates.

The plant PIF-like group is composed of four major lineages (A–D). Lineage A is the largest and most complex with members from both monocots and dicots. It can be further divided into five sublineages (A1–A5). A1 includes five grass subfamilies (Panicoideae, Ehrhartoideae, Bambusoideae, Pooideae, and the ancestral Pharoideae), indicating that this sublineage was present before the diversification of the grasses 70 MYA. Although only two grass subfamilies (Panicoideae and Ehrhartoideae) contributed sequences to A2, this lineage may be even more ancient than A1 as it is also found in the orchid G. ilense (order Asparagales). A3 and A4 are each found in a single dicot family (A3 in Brassicaceae and A4 in Fabaceae). A5 is present in both monocots and dicots and includes the only known active PIF-like element, the maize PIF. B and C are two small lineages from dicots, both restricted to the Brassicaceae family. Lineage D is another monocot-specific lineage found in four grass subfamilies (Panicoideae, Ehrhartoideae, Bambusoideae, and Pharoideae).

Pong-like TPases clustered into three major lineages (O–Q). Lineage O included two sublineages, the dicot-specific O1 and the monocot-specific O2. Lineage P is dicot specific, suggesting that it emerged in dicots after their separation from monocots. P could also be divided into two sublineages (P1 and P2). P1 was found in only the Brassicaceae family and included the majority of Pong-like TPases from A. thaliana (71%) and nearly all from B. oleracea (137 of 139). The P2 sublineage is probably older than P1 as it is also present in the Fabaceae family. Most TPase sequences in lineages Q were from O. sativa. However, the presence of one sequence from Zea mays and two from Lotus japonicus in lineage Q suggests that it is also an ancient lineage.

Introns in plant PIF-like elements: Although the original maize PIF element lacks introns (ZHANG et al. 2001 ), many plant PIF-like TPases contain one or two introns in their catalytic domains. The boundaries of these introns (donor/acceptor sites) were predicted with very high confidence (90–100%), and the coding sequences were restored (compared to intronless TPases) after their removal. Introns in PIF-like TPases can be classified into two classes on the basis of their position (Fig 1A, intron 1 and intron 2). Intron 1 is located 6 aa upstream of the first Asp residue of the DDE motif and intron 2 is located 6 aa upstream of the second Asp residue. Both introns are short (83 bp on average) and A/T rich (71% on average), with little conservation in length or sequence either within or among species. Significantly, the intron number and position from PIF-like TPases were consistent with the lineage designations. Two introns were present in three sublineages of A (A1, A3, and A4), only intron 1 was present in A2, only intron 2 was present in lineages of D, and no introns were found in lineages B and C. Sublineage A5 was an exception: some TPases did not contain any introns while others contained two.

Two models have been proposed to explain the diversity of introns associated with related coding sequences: the "intron-early" model (loss of introns from an intron-rich ancestor; GILBERT et al. 1997 ) or the "intron-late" model (addition of introns to an intron-less ancestor; LOGSDON 1998 ). It is unlikely that the intron-late model explains the distribution of PIF introns as it would require multiple and independent intron acquisitions at identical positions. The intron-early explanation is more parsimonious since the data can be most easily interpreted by hypothesizing the existence of an ancestral PIF TPase with both introns and that multiple independent loss events occurred during evolution (Fig 3). According to this model, both introns were retained in the ancestor of lineage A, but intron 2 was lost in sublineage A4. Intron 1 was lost in the common ancestor of lineages B–D so that none of these lineages contains intron 1. Intron 2 was subsequently lost in the common ancestor of lineages B and C. The predicted stepwise loss of introns from PIF-like TPase genes contrasts with the plant mariner-like TPases, where the data are more consistent with the acquisition of introns during evolution (FESCHOTTE et al. 2002A , FESCHOTTE et al. 2002B ).

TPase/ORF1 arrangements: Several different arrangements of TPase and ORF1 were observed for PIF- and Pong-like elements. All elements within a lineage or sublineage exhibit the same organization. Specifically, TPase and ORF1 in PIF-like elements are transcribed toward the same direction ("tail-to-head") but are organized in two different patterns, with the TPase gene located upstream of ORF1 in lineage A but downstream of ORF1 in lineages B, C, and D. Three different arrangements were found for Pong-like elements. TPase and ORF1 were organized in a "head-to-head" alignment for O1, a "tail-to-tail" alignment for O2, and a tail-to-head alignment for P and Q with TPase located downstream of ORF1 (see Fig 3).

PIF- and Pong-like elements in rice:
The availability of virtually the entire genomic sequence of O. sativa (GOFF et al. 2002 ; YU et al. 2002 ) made it possible to conduct a comprehensive analysis of the relationships between PIF- and Pong-like elements and Tourist-like MITEs. To do this, PIF- and Pong-like TPase sequences were first identified by computer-assisted analysis and then the sequences flanking these hits were searched to define full-length PIF- and Pong-like elements.

PIF and Pong TPases: tBlastn searches using as queries the TPases of PIF and Pong led to the identification of 205 and 145 hits (e-value < -10), respectively, from the TIGR rice database (ssp. japonica, cv. Nipponbare). Duplicate hits located on overlapping regions of bacterial artificial chromosomes were excluded as were severely truncated TPases (containing <50% of the complete coding region). The remaining 116 PIF-like TPases and 80 Pong-like TPases were relatively full-length and contained the entire catalytic domain. After removal of introns and correction of frameshifts caused by small insertion/deletions (1–2 bp), full-length PIF-like TPases were found to range in size from 392 to 432 aa, while full-length Pong-like TPases were from 416 to 549 aa.

The evolutionary relationships of rice PIF- and Pong-like TPases were determined by generating phylogenetic trees from CLUSTALW multiple alignments of their catalytic domains (Fig 4 and Fig 5). Two major lineages for PIF-like TPases that correspond to lineages A (including sublineages A1, A2, and A5) and D were resolved as shown in Fig 3. Correlation between intron content of a PIF-like TPase and the TPase lineage is similar to that described in the broader plant survey (Fig 3) except for two additional intron loss events: OsPIF3 lost intron 1 and OsPIF18 lost intron 2. A tail-to-head alignment for TPase and ORF1 was found for all OsPIF elements (TPase located upstream of ORF1 in lineage A but downstream of ORF1 in lineage D) with one exception: the two ORFs in OsPIF6 are organized in a tail-to-tail alignment, possibly due to a recent rearrangement. Pong-like TPases also clustered into two major groups, corresponding to the sublineage O2 and lineage Q in Fig 3. The ORF1/TPase alignment in lineage Q was found to be tail-to-head while that in lineage O2 is head-to-head.

View larger version (48K): In this window In a new window Download PPT slide   Figure 4. Phylogeny of OsPIF TPases and the structure of the encoding element. The neighbor-joining tree was constructed from a CLUSTALW multiple alignment of the catalytic domains (boxed region in Fig 1A) of 116 OsPIF TPases and rooted with the catalytic domain of OsPong. Bootstrap values were calculated from 1000 replicates. The structure of OsPIF elements is depicted at the right. Black triangles represent element TIRs, colored lines represent noncoding regions, gray boxes represent ORF1s, black boxes represent TPase genes, and open triangles indicate truncation in TPase genes. The positions of intron 1 (pink box) and intron 2 (blue box) are shown. The positions of insertions by other TEs are indicated by gray triangles above the elements (the identity and length of these insertions are described to the right). Dashed lines represent missing regions from elements that are incomplete because of gaps in genomic sequences or rearrangements after insertion (such as deletions or large insertions). The length of elements as well as ORF1 and TPase genes is drawn to scale. OsPIF3d and OsPIF7b were previously reported as Os-PIF1 and Os-PIF2, respectively (ZHANG et al. 2001 ).

View larger version (39K): In this window In a new window Download PPT slide   Figure 5. Phylogeny of OsPong TPases and the structure of the encoding element. The neighbor-joining tree on the left was constructed from a CLUSTALW multiple alignment of the catalytic domains (boxed region in Fig 1C) of 80 OsPong TPases and rooted with the catalytic domain of the maize PIF. Bootstrap values were calculated from 1000 replicates. Structure of OsPong elements is shown at the right. Black triangles represent element TIRs, colored lines represent noncoding regions, gray boxes represent ORF1's, black boxes represent ORF2's, and open triangles indicate truncations in ORF1's or ORF2's. The positions of insertions by other TEs are indicated by gray triangles above the elements (the identity and length of these insertions are described to the right). Dashed lines represent missing regions from elements that are incomplete because of gaps in genomic sequences or rearrangements after insertion (such as deletions or large insertions). The length of elements as well as ORF1 and TPase genes is drawn to scale.

Characterizing full-length elements: PIF- and Pong-like TPases were grouped into families on the basis of TPase sequence identities, with members of the same family being >90% identical. In this way, 27 PIF-like families and 26 Pong-like families (including the Pong family) were defined. These families were designated OsPIF (for O. sativa PIF) and OsPong (for O. sativa Pong), followed by the number of the family. Elements of the same family were further designated with a letter (e.g., OsPIF1a and OsPIF1b, see Fig 4 and Fig 5).

The identification of complete OsPIF and OsPong elements was complicated by the fact that interfamily comparisons indicated that sequence similarity was restricted to the known ORFs. For this reason, full-length elements were identified by comparison of sequences flanking the TPases within the same family where high sequence similarity extended into sequences flanking the ORFs. Sequences marking the boundary of similarity between elements of the same family were then searched for TIRs related to those of PIF or Pong and the flanking 3-bp TSDs characteristic of PIF and Pong elements (TTA/TAA). In this way, TIRs of 21 of the 27 OsPIF families (71 elements) and 20 of the 26 OsPong families (61 elements) were identified (see supplemental data at http://www.genetics.org/supplemental/ for accession numbers and positions).

The TIRs of OsPIF's were of variable length, ranging from 10 bp (OsPIF4) to 45 bp (OsPIF20). In contrast, OsPong TIRs were more uniform: all were 14–18 bp long except for one family (represented by a single-element OsPong5, 66-bp TIRs). Comparison of the TIRs of OsPIF's and OsPong's showed similarities (most began with 5'-GGSC-3', where S represents G or C) as well as differences (the fifth nucleotide was usually A in OsPong's but was rarely an A in OsPIF's; Fig 6, a and b). The inner TIRs contained PIF-specific and Pong-specific motifs [OsPIF, 5'-TGTTTGGTT-3' (positions 6–14); OsPong, 5'-STMCAA-3' (positions 7–12), where M stands for A or C].

View larger version (38K): In this window In a new window Download PPT slide   Figure 6. Comparison of the TIRs of rice PIF-like elements, Pong-like elements, and Tourist-like MITEs, shown in the form of pictograms. The terminal 14 bases from both ends are compared. The nucleotide A is shown in green, C in red, G in blue, and T in yellow. Numbers indicate the position of a nucleotide from the element end (e.g., "1" indicates the terminal nucleotide). The height of each nucleotide represents the relative frequency of that nucleotide at that position. (a) Pictogram of the TIRs from 21 OsPIF families (71 elements). (b) Pictogram of the TIRs from 20 OsPong families (61 elements). (c) Pictogram of the TIRs from 20 previously described rice Tourist-like MITE families [Tourist_Ia, Ib, Ic, III (Gaijin), IV (Castaway), V (Wanderer), VII, VIII, IX, XI, XII, XIV, XV, XVI, Type C, Buhui, Casin, Centre, Stone, and Susu]. (d) Pictogram of the TIRs from eight previously described rice Tourist-like MITE families (Tourist_VI, Helia, Qiqi, ID-2, ID-3, ID-4, Lier, Stola, and Youren; BUREAU et al. 1996 ; TARCHINI et al. 2000 ; JIANG and WESSLER 2001 ; TURCOTTE et al. 2001 ).

Full-length OsPIF's ranged from 2305 bp (OsPIF23a) to 22,169 bp (OsPIF25c) and OsPong's ranged from 2612 bp (OsPong18d) to 18,753 bp (OsPong15a). Most of this variation was due to the insertion of other TEs. These secondary TE insertions were located by searching full-length elements with RepeatMasker and Blastn. Sixteen OsPIF's and 24 OsPong's were found to contain a variety of TEs insertions (see Fig 4 and Fig 5 for their positions and identities), including other DNA elements [Ac-like, Mutator-like (MULEs), and CACTA-like], MITEs (Tourist-like and Stowaway-like), LTR retroelements, solo LTRs (Copia-like and Gypsy-like), non-LTR retroelements (LINEs), and, in one case, a Helitron element. In a few instances, members of OsPIF and OsPong families (e.g., OsPIF3, OsPIF16, and OsPong18) harbored the same MITE insertion at the same position, indicating that the MITE insertion did not prevent further transposition of these elements. When TE insertions were excluded, the length of most OsPIF's (50 of 71) and OsPong's (45 of 61) was found to be in the range of 4–6 kb.

In several instances, OsPIF and OsPong families include elements that are nearly identical, suggesting that they transposed recently and may still be capable of further transposition. For example, OsPIF6 includes five complete elements (4.1 kb) located on four different chromosomes (chromosomes 3, 7, 9, and 10) that are, on average, 99.6% identical over their entire length. In addition, their coding sequences are not interrupted by stop codons or frameshifts. Similarly, OsPong20 includes four elements (5.3 kb) that are on average >99.7% identical.

In contrast, interfamily sequence conservation, even between closely related families, was restricted to coding regions and TIRs. For example, the nucleotide sequences of OsPIF9 and OsPIF7 were 60% identical in their TPases (1.2 kb) and 55% identical in their ORF1's (1 kb), but these two families did not share additional sequence similarity aside from their TIRs. Similarly, the OsPong2 and OsPong3 families share 81 and 85% nucleotide identity in their ORF1 (900 bp) and TPase (1.3 kb) coding regions, respectively, but have completely diverged noncoding regions.

Coevolution of ORF1 and TPase in OsPong's: All 21 OsPIF families and 19 of the 20 OsPong families with defined termini encoded both ORF1 and TPase. The only OsPong family with defined termini that does not harbor an ORF1 is OsPong18, where all eight elements contain only TPase. Absence of ORF1 in these elements is likely due to internal deletion for two reasons. First, the length of these elements is unusually short, ranging from 1365 to 2745 bp. Second, Blastn searches using OsPong18 elements as queries identified several additional family members that do not encode TPase but contain coding sequence for ORF1 (e.g., AP003799; 96,105–99,317). Thus, all OsPong families with defined ends encode both ORF1 and TPase.

As mentioned above, ORF1 and TPase in OsPIF's are arranged in three different alignments and those in OsPong's have two different alignments. Interelement recombination has been shown to be a significant force in the evolution of mobile elements (e.g., ADEY et al. 1994 ; MCCLURE 1996 ; JORDAN and MCDONALD 1998 ; LERAT et al. 1999 ). The presence of two ORFs whose organization varies in different PIF/Pong-like elements prompted us to investigate whether interelement recombination had contributed to the evolution of these elements. To address this question, the phylogenies of ORF1 and the TPase of OsPIF and OsPong elements were compared to determine whether ORFs in the same element were coevolving or whether discrepancies existed that might suggest independent evolution.

Two phylogenetic trees were generated for OsPong elements, one based on an 110-aa region in their ORF1s including all three conserved blocks (Fig 1B, boxed region) and one based on the catalytic domains of their TPases (Fig 7). Comparison of the two trees showed that the phylogenies determined from the two coding sequences were consistent. These results indicate that interelement recombination had probably not occurred in OsPong's. Similarly, the phylogeny of OsPIF ORF1s was found to be consistent with that of the TPases, albeit with less bootstrap support (data not shown).

View larger version (42K): In this window In a new window Download PPT slide   Figure 7. Phylogenetic trees of OsPong ORF1 and OsPong ORF2. Trees were generated using the neighbor-joining method from CLUSTALW multiple alignments of the conserved regions in OsPong ORF1 (boxed in Fig 1B) and the catalytic domain of ORF2 (boxed in Fig 1B), respectively, and rooted with the corresponding regions of ORF1 and ORF2 from an Arabidopsis Pong-like element AtPong_AP000413 (accession no. AP000413; ORF1, 34,047–34,412; ORF2, 32,091–32,453). Where multiple OsPong elements are identical in these regions, only one is shown. The name and color of OsPong's are the same as in Fig 6. Bootstrap values were calculated from 1000 replicates.

Insertion sites of OsPIF's and OsPong's: Prior studies have shown that some plant DNA transposons, such as members of the Ac/Ds and Mutator families, have a preference for insertion into single-copy regions of the genome (CHEN et al. 1987 ; CRESSE et al. 1995 ; DIETRICH et al. 2002 ). Similarly, in a recent study it was shown that the majority of new Pong insertions (9 of 10) were into single-copy sequences of the rice genome (JIANG et al. 2003 ). To test whether this is also true for OsPIF's and other OsPong's, the immediate flanking sequences (100 bp from each end) of all elements with defined termini were searched using RepeatMasker and Blastn. Of the 132 OsPIF's and OsPong's examined, 109 (82.6%) were in single-copy sequences, 13 (9.8%) were in other DNA elements, 6 (5.5%) were in retroelements, and 4 (3.0%) were in unknown repeats.

Relationships between OsPIF's, OsPong's, and Tourist-like MITEs: Identification of full-length OsPIF and OsPong elements permitted the first genome-wide analysis of the relationship between these TPase-encoding elements and Tourist-like MITEs. Sequence identities between OsPIF's and OsPong's and Tourist-like MITEs were determined in two ways. First, we investigated whether Tourist-like MITE families could be associated with OsPIF's or OsPong's on the basis of the sequences of their TIRs. To do this, the TIRs of 31 published rice Tourist-like MITE families were examined (BUREAU et al. 1996 ; TARCHINI et al. 2000 ; JIANG and WESSLER 2001 ; TURCOTTE et al. 2001 ). The TIRs of the majority of MITE families (28 of 31) were of two types: one (20 families) was strikingly similar to the TIRs of OsPIF's (Fig 6C) and the other (9 families) was more similar to OsPong's (Fig 6D). Second, more extensive sequence similarity between individual OsPIF or OsPong families and Tourist-like MITEs was examined. Terminal sequences (200 bp from each end) from the OsPIF and OsPong families with defined ends were used as queries to search a rice repetitive sequence database (N. JIANG, Z. BAO, S. R. EDDY and S. R. WESSLER, unpublished data). Significant nucleotide similarity that extended beyond the TIRs was taken as evidence that an OsPIF or OsPong family was associated with a Tourist-like MITE family. Nine of the 21 Tourist-like MITE families with OsPIF-like TIRs were found to be associated with OsPIF families, while 3 of the 9 Tourist-like MITE families with OsPong-like TIRs were found to be associated with OsPong families.

In some cases a MITE family was clearly identified as a deletion derivative of a particular OsPIF or OsPong family (Fig 8). For example, the high copy number Castaway family (3000 copies; BUREAU et al. 1996 ; JIANG and WESSLER 2001 ) appears to be derived by a simple deletion from the OsPIF6 family. The apparent deletion breakpoints in OsPIF6 occur at a 4-bp direct repeat (TTCC, underlined in Fig 8A) that is present as only a single copy in Castaway. Similar relationships between several other OsPIF families and Tourist-like MITE families are shown in Fig 8B&NDASH;D. In contrast, although sequence similarities between OsPong and Tourist-like MITEs were detected, they were not as extensive as those seen with OsPIF (see Fig 8E and Fig F).

View larger version (69K): In this window In a new window Download PPT slide   Figure 8. Sequence similarity between OsPIF- (a–d) or OsPong-like elements (e and f) and Tourist-like MITEs in rice. Each alignment represents an example of an OsPIF or OsPong family that shares significant sequence similarity with a Tourist-like MITE. TouNJ-30 and mPile10 were identified in this study, while other MITEs were previously reported. Arrows indicate element TIRs, open boxes indicate the regions used to generate the pictograms in Fig 6, and horizontal lines denote the location of direct repeats flanking the deletion breakpoints (discussed in text).

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

PIF- and Pong-like elements are widespread and abundant:
Here we present the first comprehensive analysis of PIF- and Pong-like elements in eukaryotes. Prior studies reported that the TPases of PIF and Pong were distantly related to the bacterial IS5 TPases and noted the presence of PIF-like elements in several plant (rice, sorghum, and Arabidopsis), animal (C. elegans, C. briggsae, and fugu fish) and fungal (Filobasisiella neoformans) genomes (KAPITONOV and JURKA 1999 ; LE et al. 2001 ; ZHANG et al. 2001 ; APARICIO et al. 2002 ; JIANG et al. 2003 ). In this study, >600 PIF- and Pong-like TPases were identified or isolated from 35 plants, 19 animals, and two fungi. Phylogenetic analyses of these TPases defined three major groups, each represented by multiple distinct lineages. Taken together, the PIF/IS5 superfamily of TPases is ancient and its members are widespread. To date, the only other TPase superfamily with such a broad distribution in eukaryotes and prokaryotes is IS630/Tc1/mariner (FESCHOTTE and WESSLER 2002 ; ROBERTSON 2002 ).

PIF- and Pong-like elements are especially abundant in plants, including both monocotyledons and dicotyledons. Large numbers of PIF- and Pong-like TPases were detected in three plants with relatively small genomes: 80 copies in Arabidopsis (130 Mb); 350 copies in rice (450 Mb), and >1000 copies in B. oleracea (extrapolated to 600 Mb). Although significant sequence is not yet available for plants with large genomes, such as maize (2500 Mb) and barley (5000 Mb), the degenerate PCR assay indicates that these genomes also harbor multiple and diverse lineages (Fig 3). Given that amplification of transposable elements is largely responsible for the huge differences in plant genome size (BENNETZEN 2002 ; FESCHOTTE et al. 2002A ), it is reasonable to assume that even larger families of PIF and Pong will be found once these genomes are sequenced.

ORF1 of PIF/Pong-like elements:
The maize PIF and rice Pong elements encode two ORFs (ORF1 and ORF2), of which ORF2 is most likely the TPase while the function of ORF1 is unknown. Database searches revealed a large number of homologs for both ORFs and, where found, they were usually in pairs with an ORF1 homolog located within 1–2 kb of an ORF2 homolog. All OsPIF and OsPong families with defined termini also encoded both ORFs. The amino acid sequence similarity between PIF- and Pong-like ORF1s in blocks A and B (Fig 1) suggests a monophyletic origin, and the presence of this ORF in virtually all PIF/Pong-like lineages suggests that it is necessary for the active transposition of these elements.

A requirement for a protein other than the transposase is unusual for a eukaryotic transposon, having been described previously only for members of the CACTA superfamily (KUNZE and WEIL 2002 ). Although the autonomous Mutator element MuDR from maize encodes two proteins (MURA and MURB), it is so far the only Mutator element shown to encode more than a single ORF among hundreds of MULEs examined (YU et al. 2000 ; SINGER et al. 2001 ; LISCH 2002 ). For CACTA-like elements, multiple proteins are encoded by alternatively spliced transcripts (e.g., TNPA and TNPD of the maize En/Spm element; KUNZE and WEIL 2002 ). In contrast, our data indicate that ORF1 and ORF2 are separate transcription units. First, each ORF has a promoter that was predicted with high confidence by computer programs (data not shown). Second, elements harboring four distinct alignments of ORFs 1 and 2 were detected in plant genomes: head-to-tail (ORF2 either upstream or downstream of ORF1), head-to-head, and tail-to-tail (Fig 3). The fact that ORFs 1 and 2 would be transcribed from opposite strands in the head-to-head or tail-to-tail arrangements rules out the possibility that alternatively spliced transcripts are involved.

Several features of ORF1 provide clues to its possible function(s). Weak similarity between the most conserved region in ORF1 (Fig 1, block A) and the myb DNA-binding domain of some plant and animal transcription factors suggests that ORF1 may encode a DNA-binding protein (JIANG et al. 2003 ). One can envision a model whereby the product of ORF1 binds to the ends of Pong-like elements and recruits ORF2 by protein-protein interactions. Alternatively, products of ORFs 1 and 2 may form a heterodimer that binds to the element ends. If the products of ORFs 1 and 2 interact, it is reasonable to expect that they have been coevolving, a feature consistent with the data presented in this study (Fig 7). That is, the phylogenies of ORF1 and ORF2 are very consistent and no interfamily rearrangement was found (Fig 7). The requirements of PIF and Pong transposition as well as the possible interaction between ORF1 and ORF2 are under investigation.

OsPIF and OsPong elements:
This study identified 116 OsPIF and 80 OsPong TPases representing all of the lineages of PIF and Pong TPases detected in monocot genomes. As such, rice is a suitable model to study the evolution of PIF- and Pong-like elements in plants as well as their relationship with Tourist-like MITEs.

OsPIF and OsPong elements were grouped into 27 and 26 families, respectively, on the basis of sequence identity of their coding regions. These groupings received additional support when it was determined that elements of the same family share extensive sequence similarity in noncoding regions. Several OsPIF and OsPong families (such as OsPIF4, -5, -6, -9, -12, -13, -23 and OsPong8, -17, -19, -20) include members that are nearly identical. Furthermore, each family includes at least one putative autonomous member whose coding region is not interrupted by a stop codon or a frameshift mutation. These features are indicative of recent and perhaps ongoing activity of multiple OsPIF/OsPong families.

OsPIF, OsPong, and Tourist-like MITEs:
In previous studies, PIF and Pong elements were isolated as the TPase sources for two families of Tourist-like MITEs, mPIF and mPing, respectively (ZHANG et al. 2001 ; JIANG et al. 2003 ). In this study, many other PIF- and Pong-like elements were identified, including all the families in rice. Characterization of these elements permitted a comprehensive analysis of the relationships between these TPase-encoding elements and Tourist-like MITEs. The major conclusion from this analysis is that most Tourist-like MITE families are related to either PIF- or Pong-like elements solely on the basis of a comparison of their TIRs. Of the 31 previously described Tourist-like MITE families in rice, the TIRs of 20 were found to be more closely related to the consensus OsPIF TIR while the TIRs of 9 were more closely related to the consensus OsPong TIR (Fig 6).

Attempts to associate individual Tourist-like MITE families with specific OsPIF or OsPong families uncovered many clear-cut relationships (Fig 8). For example, the MITE family Castaway (3000 copies) was found to be derived from the OsPIF6 family by internal deletion and subsequent amplification. Relationships between Tourist-like MITEs and OsPong families are less apparent as sequence similarity is limited to the subterminal regions (as shown in Fig 8E and Fig F). However, detection of sequence identity between subterminal regions of OsPong families and Tourist-MITE families, albeit limited, is significant in light of the fact that even closely related OsPong families display no sequence identity in their subterminal regions.

In summary, the characterization of virtually all full-length PIF- and Pong-like elements in the rice genome has permitted a determination of the extent of their relatedness with most of the 60,000 Tourist-like MITEs residing in this genome. Our data indicate that many Tourist-like MITEs originated from OsPIF and OsPong elements by internal deletion and subsequent amplification. However, 16 of the 28 Tourist-like MITEs examined in this study were not clearly associated with OsPIF/OsPong families. It is possible that their cognate OsPIF/OsPong families were lost from the genome. Such a scenario is not difficult to imagine considering OsPIF/OsPong elements are present at much lower copy number (several per family) than are Tourist-like MITEs (hundreds or thousands per family). Alternatively, some Tourist-like MITE families may have originated by chance events, where, for example, a pair of nearby inverted repeats (and other cis requirements, if any) were mobilized fortuitously by an endogenous PIF- or Pong-like TPase and subsequently amplified to high copy numbers.

The rice genome harbors >90,000 MITEs: 60,000 Tourist-like MITEs and 30,000 Stowaway-like MITEs (FESCHOTTE et al. 2003 ; N. JIANG and S. R. WESSLER, unpublished data). Whereas elements of the PIF/Pong/IS5 superfamily (OsPIF and OsPong) appear to be responsible for the origin and amplification of Tourist-like MITEs, rice elements related to the Tc1/mariner superfamily (called Osmars) have been associated with Stowaway-like MITEs (FESCHOTTE et al. 2003 ). Given that plant genomes are known to harbor many other DNA transposon families (including CACTA-like, hAT-like, and MULEs), one obvious question is, Why are most MITE families associated with only these two superfamilies? A key factor may be the cis requirements for transposition. Transposition of MULEs requires long TIRs (>200 bp; BENITO and WALBOT 1997 ), while that of CACTA-like and hAT-like elements requires both TIRs and subterminal repetitive motifs (reviewed in KUNZE and WEIL 2002 ). Several studies have shown that, for these elements, multiple TPase-binding sites reside in the subterminal repeats. In contrast, the binding sites for animal mariner TPases seem to be restricted to the short element TIRs (28–31 bp; AUGE-GOUILLOU et al. 2001 ; LAMPE et al. 2001 ; ZHANG et al. 2001 ). Indeed, artificial transposons containing just the mariner TIRs have been successfully mobilized by the cognate mariner TPase, in vitro and in vivo in bacteria (e.g., LAMPE et al. 1999 , LAMPE et al. 2001 ; TOSI and BEVERLEY 2000 ). Although the cis requirements for transposition of PIF- and Pong-like elements have yet to be determined, these elements do not contain any apparent repetitive motifs in subterminal regions, suggesting that their TPase-binding sites may also reside in their short TIRs. Thus, the minimal cis requirements for transposition by members of the mariner-like and PIF/Pong-like families may significantly enhance the probability of generating shorter elements by deletion from larger elements or by chance, which could be mobilized by the TPases encoded by these families.

Concluding remarks:
The PIF and Pong elements are founding members of a very large and dynamic superfamily of class 2 elements that are widespread in flowering plants. The impact of these elements is significant, as PIF- and Pong-like families are capable of expansion through the amplification and diversification of both autonomous and nonautonomous members, including very-high-copy-number MITEs. Furthermore, with a demonstrated preference for insertion into genic regions, PIF- and Pong-like elements and their associated Tourist-like MITEs appear to be a major force generating genetic diversity and influencing the evolution of plants.

	FOOTNOTES

Sequence data from this article have been deposited with the EMBL/GenBank Data Libraries under accession nos. AY362792, AY362793, AY362794, AY362795, AY362796, AY362797, AY362798, AY362799, AY362800, AY362801, AY362802, AY362803, AY362804, AY362805, AY362806, AY362807, AY362808, AY362809, AY362810, AY362811, AY362812, AY362813, AY362814, AY362815, AY362816, AY362817, AY362818, AY362819.

	ACKNOWLEDGMENTS

This work was supported by grants from the National Science Foundation and National Institutes of Health to S.R.W.

Manuscript received May 19, 2003; Accepted for publication October 30, 2003.

	LITERATURE CITED

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

ADEY, N. B., S. A. SCHICHMAN, D. K. GRAHAM, S. N. PETERSON, and M. H. EDGELL et al., 1994  Rodent l1 evolution has been driven by a single dominant lineage that has repeatedly acquired new transcriptional regulatory sequences. Mol. Biol. Evol. 11:778-789.[Abstract]

APARICIO, S., J. CHAPMAN, E. STUPKA, N. PUTNAM, and J. M. CHIA et al., 2002  Whole-genome shotgun assembly and analysis of the genome of Fugu rubripes. Science 297:1301-1310.[Abstract/Free Full Text]

AUGE-GOUILLOU, C., M. H. HAMELIN, M. V. DEMATTEI, M. PERIQUET, and Y. BIGOT, 2001  The wild-type conformation of the Mos-1 inverted terminal repeats is suboptimal for transposition in bacteria. Mol. Genet. Genomics 265:51-57.[CrossRef][Medline]

BENITO, M. I. and V. WALBOT, 1997  Characterization of the maize Mutator transposable element MURA transposase as a DNA-binding protein. Mol. Cell. Biol. 17:5165-5175.[Abstract]

BENNETZEN, J. L., 2002  Mechanisms and rates of genome expansion and contraction in flowering plants. Genetica 115:29-36.[CrossRef][Medline]

BUREAU, T. E., P. C. RONALD, and S. R. WESSLER, 1996  A computer-based systematic survey reveals the predominance of small inverted-repeat elements in wild-type rice genes. Proc. Natl. Acad. Sci. USA 93:8524-8529.[Abstract/Free Full Text]

CAPY, P., C. BAZIN, D. HIGUET and T. LANGIN, 1998 Dynamics and Evolution of Transposable Elements. Springer-Verlag, Austin, TX.

CHEN, J., I. GREENBLATT, and S. DELLAPORTA, 1987  Transposition of Ac from the P locus of maize into unreplicated chromosomal sites. Genetics 117:109-116.[Abstract/Free Full Text]

CRESSE, A. D., S. H. HULBERT, W. E. BROWN, J. R. LUCAS, and J. L. BENNETZEN, 1995  Mu1-related transposable elements of maize preferentially insert into low copy number DNA. Genetics 140:315-324.[Abstract]

DIETRICH, C. R., F. CUI, M. L. PACKILA, J. LI, and D. A. ASHLOCK et al., 2002  Maize Mu transposons are targeted to the 5' untranslated region of the gl8 gene and sequences flanking Mu target-site duplications exhibit nonrandom nucleotide composition throughout the genome. Genetics 160:697-716.[Abstract/Free Full Text]

FESCHOTTE, C. and S. R. WESSLER, 2002  Mariner-like transposases are widespread and diverse in flowering plants. Proc. Natl. Acad. Sci. USA 99:280-285.[Abstract/Free Full Text]

FESCHOTTE, C., N. JIANG, and S. R. WESSLER, 2002a  Plant transposable elements: where genetics meets genomics. Nat. Rev. Genet. 3:329-341.[CrossRef][Medline]

FESCHOTTE, C., X. ZHANG and S. WESSLER, 2002b Miniature inverted-repeat transposable elements (MITEs) and their relationship with established DNA transposons, pp. 1147–1158 in Mobile DNA II, edited by N. L. CRAIG, R. CRAIGIE, M. GELLERT and A. M. LAMBOWITZ. American Society for Microbiology Press, Washington, DC.

FESCHOTTE, C., L. SWAMY, and S. R. WESSLER, 2003  Genome-wide analysis of mariner-like transposable elements in rice reveals complex relationships with Stowaway MITEs. Genetics 163:747-758.[Abstract/Free Full Text]

GILBERT, W., S. J. DE SOUZA, and M. LONG, 1997  Origin of genes. Proc. Natl. Acad. Sci. USA 94:7698-7703.[Abstract/Free Full Text]

GOFF, S. A., D. RICKE, T. H. LAN, G. PRESTING, and R. WANG et al., 2002  A draft sequence of the rice genome (Oryza sativa L. ssp. japonica). Science 296:92-100.[Abstract/Free Full Text]

HEBSGAARD, S. M., P. G. KORNING, N. TOLSTRUP, J. ENGELBRECHT, and P. ROUZE et al., 1996  Splice site prediction in Arabidopsis thaliana DNA by combining local and global sequence information. Nucleic Acids Res. 24:3439-3452.[Abstract/Free Full Text]

JIANG, N. and S. R. WESSLER, 2001  Insertion preference of maize and rice miniature inverted repeat transposable elements as revealed by the analysis of nested elements. Plant Cell 13:2553-2564.[Abstract/Free Full Text]

JIANG, N., Z. BAO, X. ZHANG, H. HIROCHIKA, and S. R. EDDY et al., 2003  An active DNA transposon family in rice. Nature 421:163-167.[CrossRef][Medline]

JORDAN, I. K. and J. F. MCDONALD, 1998  Evidence for the role of recombination in the regulatory evolution of Saccharomyces cerevisiae Ty elements. J. Mol. Evol. 47:14-20.[CrossRef][Medline]