Five consensus non-LTR retrotransposons, termed CiI, CiL1, CiL2, CiLOA and CiR2, were derived from five, five, six, five and five C. intestinalis scaffolds, respectively (Figures 1, 2 and see Additional data file 1). TBLASTX comparisons showed that the ascidian elements belonged to the I, LINE1, LINE2, LOA and R2 clades (E-values: 4e^-69 with Biomphalaria glabrata (snail) BGR, 2e^-89 with Nycticebus coucang (slow loris) L1, 2e^-50 with Danio rerio CR1Dr2, e^-146 with Aedes aegypti Lian, and e^-106 with Drosophila melanogaster R2, respectively). CiL1.2 (Figure 2b), derived from five scaffolds, was another ascidian LINE1 element. It showed homology with the Xenopus laevis Tx1 retroelement (E-value: 2e^-40) but was not further analyzed because it was significantly shorter than CiL1 (CiL1.2 only encompassed the reverse transcriptase region).

All the Ciona elements encoded the conserved reverse transcriptase with the distinctive structural hallmarks defined as block 0, 1, 2, 2a, 3, 4, 5, 6 and 7 11 (Figure 2b). Although conservation of the thumb region (block 8 and 9) was weak, preservation of the ascidian sequences defining the CRE/R2/R4/L1/RTE subgroup was still found. The apurinic/apyrimidic endonuclease (APE) region was clearly identified in CiI, CiL1, CiL2 and CiLOA (Figure 2a) on the basis of the reported domains I to VII 17. CiI and CiLOA also contained the RNaseH (RNH) domain at the carboxylic end (Figures 1, 2d). Concerning ORF1, partial sequences were assembled for CiI and CiLOA, but a CCHC motif (single-letter amino-acid code) in this region was only identified in the CiI element (Figure 2e). Finally, for CiR2, a restriction enzyme-like endonuclease (REL-endo) containing the CCHC and KPDI motifs 18 was found in the carboxy-terminal region, and a CCHH domain and a putative c-Myb DNA-binding motif were identified at the amino terminus (Figure 2c). Overall, the structure and organization of the ascidian non-LTR retrotransposons is consistent with those reported for each non-LTR clade.

The reverse transcriptase domain of non-LTR elements was used to establish the phylogenetic relationships of the ascidian elements and the 14 reported non-LTR clades. In the neighbor-joining tree (Figure 3), all clades were supported with significant bootstrap values (≥70%), except clade I, with the lowest bootstrap value (67%) in agreement with previous analyses 111214. Therefore, ascidian sequences clustered within five distinct clades: I, L1, L2, LOA and R2 (bootstraps: 67%, 70%, 97%, 100% and 100%, respectively), as a result of which they were recorded as new members of such groups.

Fragments of about 300 nucleotides of the reverse transcriptase domain of each ascidian element were PCR amplified, cloned, sequenced and used for copy-number estimations and methylation analyses. To quantify the copy number for each element, two independent experimental approaches - slot blot analysis and genomic library screenings (Figure 4a,b) - were combined with in silico scores on the number of Ciona scaffold-containing elements (Table 1). When the reverse transcriptase was considered, the data from the different approaches were consistent and mean values for each element were in the range 3-7, far below the copy numbers of the vertebrate counterparts (Table 2). CiR2 slot-blot analysis did not give a signal, in agreement with the low estimates obtained after in silico searches and library screenings. Indeed, full-length copies could not be assembled for any of the families after database searches. In silico estimates with sequences that also encompassed the 5' and 3' sequences of reverse transcriptase increased the numbers slightly: 9, 22, 24, 69 and 13 for CiI, CiL1, CiL2, CiLOA and CiR2, respectively (Table 1).

Gene density and GC content in the surrounding retrotransposon sequences was estimated from 26 10-kb regions flanking CiI, CiL1, CiL2 and CiLOA elements of 17 scaffolds. Overall, 16.5 genes were found in the 260 kb analyzed. Therefore, the average gene density (1 gene per 15.8 kb) was lower than that of the whole genome (1 gene per 7.5 kb) 1. However, no differences were observed when the GC content of those segments (35.7%) was compared with the overall genomic value (35%). Concerning CiR2, our data confirmed the target specificity for rRNA genes associated with the REL-endo domain: 9 out of 13 CiR2s were indeed linked to rRNA sequences.

Finally, the methylation status of the genomic regions containing the elements was investigated by comparing the hybridization patterns of genomic DNA restricted with the methylation-sensitive enzyme HpaII, and the methylation-insensitive isoschizomer MspI. The identical HpaII and MspI patterns obtained for all the elements (except for CiR2, which gave no signal) supported the location of the ascidian elements in unmethylated genomic segments (Figure 4c).

The analysis of non-LTR elements in the urochordate Ciona provides valuable data for understanding the evolution of early chordate genomes and enlarges the view of the distribution of the non-LTR clades in eukaryotes. The Ciona genome harbors: I, LOA and R2 elements, hitherto restricted to protostomes; L1 elements, formerly uncharacterized in invertebrates; and L2 elements, previously described in protostomes and vertebrates.

Clade I was the least supported branch of our analysis (bootstrap value, 67%). However, ascription of CiI to this clade was unambiguous as it shares with the other I elements the CCHC motif and the APE, reverse transcriptase and RNH domains (Figure 1), and also because it clearly clustered with the I non-LTR retrotransposon BGR of the snail Biomphalaria glabrata (bootstrap 93%). In regard to the LOA clade, representatives in urochordates had previously been identified after BLASTN and BLASTX comparisons. The ascidian Cili-2 retrotransposon gives the closest match with the RNH domain of the mosquito Lian element 3. We have now derived CiLOA, an element that encodes APE, reverse transcriptase and RNH. The phylogenetic analysis of the reverse transcriptase domain together with the other structural hallmarks improved the assignment of the Ciona sequence to the LOA clade (bootstrap 100%). Finally, the phylogenetic analysis and structural features clearly placed CiR2 within the R2 clade (Figures 1-3). As well as the reverse transcriptase, the preservation in the deuterostome lineage of the distinctive R2 structural hallmarks, such as the REL-endo domain and the 5' CCHH and c-Myb DNA-binding motifs, indicate the ancient structural organization of this clade. Additionally, insertions of the element near the Ciona rRNA genes suggest that target specificity through the REL-endo mechanism has been preserved. Overall, not only does the analysis of CiI, CiLOA and CiR2 agree with the origin of these retrotransposons in the Precambrian era 11, but the fact that the urochordate elements resemble the protostome counterparts points to their ancient structural organization.

We derived CiL1 and CiL1.2, whose structural organization and phylogenetic relationship made them cluster within the L1 clade (bootstrap 70%) and supported a previous BLAST analysis of two short Ciona sequences 3. Our data allowed the first structural characterization of the L1 clade in invertebrates. CiL2 clustered within the L2 clade (bootstrap 97%), a novel group of non-LTR retrotransposons closely related to the CR1 and Rex1 clades 14, which includes members previously described in the protostome and deuterostome lineages (Table 2). Interestingly, the CR1 and RTE clades, which are also shared by protostomes and deuterostomes, have not been identified in Ciona. Whether these clades were lost in the whole urochordate subphylum needs further investigation.

Sequence analysis of the scaffolds harboring the non-LTR elements revealed that ascidian transposable elements are flanked by regions of low gene density. However, no differences in GC content with respect to the average genome value were found when comparing these genomic segments. Moreover, Southern analysis showed that ascidian non-LTR retrotransposons are unmethylated. Overall, these data suggest that mobile elements, gene density and methylation status have not influenced the nucleotide composition in urochordates.

Copy-number estimates of the non-LTR elements in the Ciona genome suffer from slight inaccuracies due to the hybridization reaction, which disregards highly divergent elements, and to the fact that computational estimates only refer to the available 90% of the genome. However, the agreement between the in silico and experimental estimations indicates that, in this case, the biases have been minimized. The data show a low copy number per haploid genome of the different ascidian elements: from 9 to 69 copies, which decreased to three to seven copies when estimates were based on the reverse transcriptase domain only. These values are far below the vertebrate counterparts, but similar to numbers reported for protostome genomes. This also seems to apply to another lower chordate genome. In amphioxus (subphylum Cephalochordata) a low copy number has been reported for a non-LTR retrotransposon, BfCR1, 16 and for ATE1, a class II transposable element 19.

The factors involved in retrotransposon control are still an open question. The view that methylation evolved to suppress the activity of transposable elements in vertebrates 20 pointed to DNA methylation as a good candidate for transposition control in lower chordates. However, ascidian transposable elements are clearly unmethylated (21 and this study) and, hence, the genome-defense model cannot be extended to urochordates, and perhaps not to cephalochordates, as the amphioxus BfCR1 element also belongs to the unmethylated genomic fraction (Figure 4c). Among other mechanisms, if required at all for retrotransposon control in lower chordates, co-suppression, which operates on I elements in Drosophila 2223 and in transposable element silencing in plant genomes 24, is a possibility.

The C. intestinalis non-LTRs were identified through a TBLASTX 25 search on the Ciona genome draft deposited in the JGI database 26. The following non-LTR retrotransposons were used as queries: CRE1 from Crithidia fasciculata (accession number M33009), CZAR from Trypanosoma cruzi (M62862), Dong from Bombyx mori (L08889), L1 from Rattus norvegicus (U83119), RTE1 from Caenorhabditis elegans (AF025462), Tad1 from Neurospora (L25662), R1 from D. melanogaster (X51968), Jockey from D. melanogaster (M22874), L1Tc from T. cruzi (X83098), R2 from Porcellio scaber (AF015818), LOA from Drosophila silvestris (X60177), Rex3 from Tetraodon nigroviridis (AJ312226), NeSL-1 from C. elegans (Z82058), CR1 from Gallus gallus (AAC60281) and Maui from Takifugu rubripes (AF086712). The retrieved Ciona sequences were aligned by eye on the basis of the DotPlot comparisons of the MegAlign program from the DNASTAR package, and a consensus composite was assembled. Sequence differences between scaffolds due to nucleotide substitutions or indels were analyzed and the sequence maximizing the similarity to reported elements was selected. The non-LTR nature of each composite sequence was further verified through a TBLASTX search against the GenBank database. The consensus sequence was named after the defined non-LTR clade to which it belonged. The CiI consensus sequence was derived from scaffolds 120, 148, 599, 1398 and 2116; CiL1 from 951, 1407, 1810, 3249 and 7743; CiL1.2 from 388, 890, 1138, 2278 and 2648; CiL2 from 231, 604, 1005, 1177, 1644 and 3322; CiLOA from 398, 925, 1078, 1854 and 4983; CiR2 from 345, 1777, 2388, 2455 and 3439.

The in silico copy number of each repetitive DNA element was estimated from the Ciona database. Two types of search were performed. First, all the derived sequences were used to retrieve scaffold-containing elements that matched with a BLAST expect value of <10^-3 15. To discard wrongly assigned elements, a threshold was defined at the score value of the first match of an element that belonged to another clade. Second, for the sake of comparison with experimental data, only the consensus reverse transcriptase region was used for the search and the scaffolds showing a minimum match of 300 nucleotides with the same BLAST expect value were considered.

PCR amplifications with primers designed from the consensus reverse transcriptase sequence of each identified ascidian non-LTR elements were performed with 250 pg of genomic DNA and 1 U Taq DNA polymerase (BioTherm) in 25 μl of reaction volume containing 0.2 μM for each primer, 32 μM each dNTP and 2 mM MgCl₂. The sequences of the primers were: CiL1-F (forward): 5'-AACTAGTGATACCGCGCC-3', CiL1-R (reverse): 5'-ACACCTCGTTTGATCGG-3', CiL2-F: 5'-GTTGAGGTAAATGGCGC-3', CiL2-R: 5'-CGTTCGTCATTATCTGGG-3', CiR2-F: 5'-TTCCGCAAGGTCGATG-3', CiR2-R: 5'-CAGATAGGGCCCAATCC-3', CiI-F: 5'-CGATCTACCACCGACCAC-3', CiI-R: 5'-GCTTGTCACAGGCAGTTG-3', CiLOA-F: 5'-AACTGCGGAGATCCATGG-3' and CiLOA-R: 5'-GTCGCAGTCTTGATGCGG-3'. PCR conditions were as follows: the initial denaturation step at 94°C for 2 min was followed by 40 cycles at 94°C for 45 sec, 53°C for 30 sec and 72°C for 30 sec and a final extension step at 72°C for 5 min. In each PCR assay, a fragment of approximately 300 bp was amplified and then cloned in a pUC18 plasmid and sequenced using the Big Dye Terminator v3.1 Cycle Sequencing Kit (Applied Biosystems, Foster City, CA) in a 3730 DNA Analyzer (Applied Biosystems).

A C. intestinalis λZapII genomic library (kindly provided by M. Levine) was screened with each of the fragments of the identified Ciona non-LTR elements. The probes were labeled with [α-³²P]dCTP by random-hexamer priming and hybridized to phage DNA transferred on Hybond-N nylon filters (Amersham Pharmacia Biotech, Uppsala, Sweden) in duplicate. Approximately 70,000 phages were screened. Hybridizations were performed in phosphate-SDS solution 27 at 65°C overnight. Two 15-min washes were performed at 65°C in 2 × SSC, 0.1% SDS, 2 × 15 min at 65°C in 1 × SSC, 0.1% SDS and 1 × 15 min at 65°C in 0.2 × SSC, 0.1% SDS. Hybridization signals were detected by autoradiography. Only the signals present in the original and duplicated filters were considered.

For quantitative slot-blot analysis, 500 ng, 250 ng, 50 ng and 25 ng of EcoRI-digested C. intestinalis genomic DNA and serial dilutions of each plasmid-containing probe, EcoRI-restricted and mixed with 1 μg mouse genomic DNA (as nonspecific DNA) were denatured with 0.4 M NaOH and 25 mM EDTA in a final volume of 200 μl and blotted on Hybond-N nylon filters (Amersham Pharmacia Biotech) with a slot-blot device (Minifold II, Schleicher & Schuell, Dassel, Germany). Three genomic DNA replicates of isolated animals were performed. Before sample loading, the membrane was soaked in water and then neutralized with 2 M sodium acetate, pH 5.4 and fixed with UV light. Membranes were hybridized with the same probes used for library screening at the same hybridization and washing conditions. The slot-blot signal was quantified with the GS525 Molecular Imager System (Bio-Rad, Hercules, CA).

For Southern analyses 10 μg of C. intestinalis genomic DNA digested with HpaII or MspI was resolved on 0.8% agarose gels and transferred to nylon membranes. Southern blots were hybridized with the same non-LTR probes used for library screening at identical hybridization and washing conditions.

Gene density and GC content of the retrotransposon insertion sites was assessed. Only the scaffolds that expanded at least 10 kb upstream or downstream from an element were considered. Gene density in the 10-kb flanking regions was calculated by scoring the predicted genes according the Ciona gene model v1.0. When the gene sequence was only partially contained in the region analyzed, it was scored as 0.5.

For phylogenetic analysis the C. intestinalis sequences were added to a previous alignment by Malik 11, updated by adding the NeSL-1 (C. elegans, Z82058), LINE2 (Patella, X77618; Danio rerio, AL591210; Oryzias, AB054295; Fugu, AF086712) and Rex1 (Xiphophorus, AF155728; Batrachocottus, AAA83744) clades. The new alignment was generated using Clustal X 28, maintaining the same pairwise gap penalties and multiple alignment parameters, and adjusted by eye (see Additional data files 2 and 3). Phylogenetic analyses were performed using the neighbor-joining method, rooted with the reverse transcriptase sequence of Neurospora organellar group II intron (accession number S07649) and drawn with the TreeViewPPC program 29. Confidence in each node was assessed by 1,000 bootstrap replicates.