Recently, a new family of bipartite chimeric retroelements termed MINE (Mixed Interspersed Nuclear Elements) has been identified in the genome of the rice blast fungus Magnaporthe grisea 17. The 5' parts of MINE correspond to a full length copy of a 1.1 kb long non-coding transcript of unknown function called WEIRD (Figure. 2A). The 3' parts of MINE correspond to 5' truncated copies of the MGL LINE element 17. The organization and diversity of MINE elements suggest that they are likely formed as a result of a template RNA switch during LINE reverse transcription and integrated into the genome using LINE machinery as proposed for similar mammalian LINE chimeras 313141518. A copy of MINE was found as a recent insertion in ACE1 avirulence gene 1726, showing that this mechanism is still active in fungal cells and could inactivate genes. In order to investigate the distribution of bipartite retrogenes and to find out if other types of chimeras than MINE were generated in M. grisea, we performed a systematic bioinformatic analysis of M. grisea genome sequence for all possible WEIRD and MGL sequences that are not directly flanked by short direct repeats (target site duplications). Detailed analysis of genomic sequences flanking these elements allowed the identification of 33 chimeric retroelements flanked by 10–20 bp long direct repeats (Figure 2). As expected, the majority (31 elements) of these chimeras belong to the MINE family (WEIRD – MGL chimeras). 19 MINEs were flanked by target site duplications while the direct repeats could not be found for the 12 other MINEs as one of their ends correspond to a gap in M. grisea genomic sequence. Compared to a survey performed on a previous version of M. grisea genome 17, we only identified one novel MINE copy.

In contrast to mammalian bipartite chimeric retroelements that have a 5' cellular component corresponding to known nucleolar RNAs (mostly U6 snRNA), the 5' cellular component of M. grisea bipartite chimeric retroelements corresponds mainly to a non-coding RNA (WEIRD) that has no homology to M. grisea U6 snRNA [Gogvadze et al., unpublished data]. The molecular function of WEIRD RNA remains unknown. In a series of in situ hybridization experiments with WEIRD specific oligonucleotide probes, we were not able to define WEIRD intracellular localization due to extremely poor permeability of the M. grisea cell wall [Gogvadze et al., unpublished data]. Expression of WEIRD was assessed in different M. grisea tissues (mycelium, spore) during a barley leaf infection using real-time quantitative reverse transcription-PCR, where expression was relative to two reference fungal housekeeping genes. Levels of WEIRD RNA are comparable to those of the housekeeping gene Ilv5 (35, 20 and 61% of Ilv5 transcript level in different tissues). WEIRD was differentially expressed, being mostly up-regulated during the infection process (2-fold versus mycelium, Table 1), whereas MGL was up-regulated in spores. To address whether WEIRD is a unique genomic sequence or a new transposable element, we performed a bioinformatic analysis of M. grisea genome for WEIRD-like sequences. Surprisingly, no solo WEIRD copies were found, as all of the WEIRD sequences we identified were directly fused with MGL LINE, forming MINE retrogenes. The absence of the initial WEIRD gene in the databases is probably due to the still incomplete state of M. grisea genome sequencing 27. No sequences with significant similarities with WEIRD were found in other available genomes. Therefore, WEIRD (i) is not a transposable element, (ii) is expressed in mycelium/spores and during infection, and (iii) does not encode for a protein. These properties suggest that WEIRD might be a functional housekeeping non-coding RNA.

Apart from the previous 31 MINEs, we identified one novel bipartite chimeric retroelement with a WEIRD sequence fused to 5'-truncated Mg-SINE. This WEIRD-MgSINE retroelement was flanked by 12 bp-long direct repeats (figure 2B, Additional file 1). Mg-SINE is a non-autonomous SINE-family retroelement 28. It likely utilizes a "molecular mimicry" to MGL for its proliferation. Indeed, the 3' end of Mg-SINE is similar to the 3' end of MGL retrotransposon used for priming reverse transcription 2930. This property of Mg-SINE sequence likely results in the capture of Mg-SINE transcripts by the MGL retrotranspositional apparatus.

We have also identified a novel chimera composed of three elements (figure 2C, Additional file 1). Its 5' component is a full length WEIRD fused to a 5'-truncated Mg-SINE that is itself fused to a 5'-truncated MGL. The tripartite chimeric retroelement is flanked by 14 bp-long direct repeats showing that it was integrated into the M. grisea genome as a single chimeric retrogene likely formed by two successive template RNA switches during the reverse transcription of MGL.

Overall, this survey of M. grisea genome showed that most chimeric retrogenes containing MGL sequences are represented by MINEs(WEIRD-MGL), although we also identified another type of bipartite chimeras (WEIRD-MgSINE) and a triple chimera (WEIRD-MgSINE-MGL).

To assess if these fungal bipartite chimeric retroelements were transcribed, we used reverse transcription-PCR (RT-PCR) with MINE specific primers designed to the 5' terminus of WEIRD (in the sense orientation), and to the 3' terminus of MGL (in the antisense orientation; Figure 3). Some MINE chimeras were expressed in mycelium, spores and 24 hours old infected leaves. Sequencing of these RT-PCR products revealed that bands 1 (mycelium) and 5 (24 hours infection) correspond to transcripts from the same MINE chimera as well as bands 2 (mycelium) and 4 (spores) that correspond to another MINE chimera. Overall, we identified three transcriptionally active MINEs in M. grisea: MINE-A (bands 1 and 5), MINE-B (bands 2 and 4) and MINE-C (band 3) that correspond, respectively, to the elements 4, 10 and 3 we identified in the M. grisea genome (Additional file 1; GenBank accession numbers EF585235–EF585237). These three transcribed chimeras are likely an underestimation of the possible MINE transcripts, as many of these elements have a long MGL part (up to 5 kb) that is likely too long to be efficiently amplified by RT-PCR. To further quantify the mRNA levels of these MINEs, we performed quantitative real-time PCR experiments with primers specific for each WEIRD-MGL junctions (primers q1 and q2, Figure 3) to specifically amplify each chimera (Table 2). MINE-A, MINE-B and MINE-C are transcribed in all analyzed tissues, but at different levels. MINE-B is constitutively expressed in all tissues tested including infection, while MINE-A and MINE-C are up-regulated during infection as observed for WEIRD (× 2 vs mycelium) and MINE-A is down-regulated in spores (× 0.5 vs mycelium). These experiments demonstrate that some WEIRD-MGL chimeric elements are transcribed and that the expression pattern differs depending on M. grisea tissues. Consequently, these expressed chimeras may be involved in a particular cell function.

The fungal tripartite chimeric retroelement discovered in this study (Figure 4A) suggests that the suspected mechanism of LINE-mediated in vivo RNA recombination is not limited to a single template switch. The most probable explanation for the formation of a tripartite chimeric retroelement is a double template switch during LINE retrotransposition (Figure 1). Having reverse-transcribed 3'-terminal 4369 nucleotides of MGL LINE (the total length of MGL is ~6 kb), the retrotranspositional complex switched to another RNA (located nearby or captured by the RT/integrase assembly) and added a ~350 nt-long 3' fragment of the Mg-SINE element to the nascent cDNA. RT dissociation from its initial template can be explained, for example, by reverse-transcriptional pausing events 31, or by a damage to the initial RNA strand 21. However, before the completion of Mg-SINE reverse transcription, RT changed template for a second time, and synthesized a full copy of WEIRD RNA (1113 nt-long) on the terminus of the nascent first-strand-cDNA. After ligation, synthesis of the second cDNA strand and genomic DNA repair, the newly formed tripartite retrogene became flanked by 14 bp-long direct repeats resulting from the duplication of the genomic target site.

Previously, two tripartite chimera-like elements were identified in human DNA 1618 (Fig. 4B). One of them was inserted in an expanded microsatellite locus, while the other was flanked by AT-rich sequences. As a result, the identification of direct repeats that should normally flank the chimera was not possible for these elements (Additional file 2). However, a survey of mammalian genomic databases allowed us to identify two new tripartite chimeric retrogenes flanked by perfect 15- and 16 bp-long direct repeats in the mouse genome (Fig. 4C, Additional file 2). These chimeras were composed of (5' to 3') B2 SINE, U6 snRNA and L1 LINE elements. Finding tripartite chimeric retroelements displaying considerable structural similarities in such evolutionary distinct organisms as mammals and fungi suggests that double template switches during LINE reverse transcription is probably not the exceptional case, but might be a conserved mechanism of eukaryotic DNA rearrangement.

The 3' terminal parts of fungal chimeric retroelements are usually composed of 3' fragments from MGL LINE retrotransposon. These fragments vary in length from 230 to 5461 3'-terminal MGL nucleotides. Analysis of the sequence of 33 chimeric retroelements allowed us to retrieve 28 different chimerization sites within the MGL sequence. In six cases, the same chimerization site was shared by two distinct chimeric retroelements (Additional file 1, pairs of elements: 5 & 6; 8 & 9; 10 & 11; 12 & 13; 4 & 15; 17 & 18). This observation showed that template switching occurred exactly at the same nucleotide position of the template RNA during the formation of six different pairs of chimeric retroelements. Additionally, analysis of the 28 chimerization sites revealed that template switching did not occur at random within template RNAs. Furthermore, highlighted hot spots within MGL corresponded to a conserved sequence (Table 3). Importantly, this motif was also found at the chimerization sites of MgSINE chimeric retroelements. The consensus sequence located at the chimerization sites on these template RNAs is GCC*A/U, where * indicates the site of the possible template switching. This observation suggests that MGL RT reverse-transcribes the template RNA until the A/U nucleotide of the motif, where it can jump to another RNA before the GCC triplet. It should be noted that the full-length MGL RNA has 87 GCC(A/U) sites and far more derivatives with single nucleotide substitutions, whereas only a small portion of these motifs were used as chimerization sites in vivo. Taking into account that six of these sites are hot spots for "RNA recombination" we conclude that this motif is necessary but not sufficient for the template switching. It seems reasonable to postulate that the chimerization sites are located in a region of the template RNA with a secondary structure that could lead to a pause in reverse transcription. This situation has already been observed for template switching sites of mammalian chimeric retrogenes that coincide with reverse-transcription pausing sites 2531. In silico prediction of MGL RNA secondary structures around chimerization sites of fungal chimeric retrogenes predicts strong hairpin structures immediately upstream of known template switching sites (Additional file 3) that theoretically could slow down reverse transcription.

It remains unclear why the RT dissociates from the template RNA exactly before the GCC triplet. It is known for many RTs that the enzyme processivity depends greatly on the nucleotide composition of the template RNA. It can be hypothesized that this GCC motif, in particular when located in a hairpin, is a "difficult place" for the MGL RT, inducing a reverse-transcriptional pause or even the RT to dissociate or to jump on another RNA template.

Finally, it should be mentioned that the above chimerization mechanism does not include any kind of specific nucleotide basepairing between the nascent cDNA and the second RNA template. Indeed, we were not able to find extended- or micro-homologies between the templates around the chimerization sites.

Homology searches against GenBank were done using the BLAST Web-server at NCBI 32. Flanking regions of mammalian retroelements were investigated with the RepeatMasker program 33. For fungal genomes, flanking regions were aligned with known filamentous fungal transposable elements reported elsewhere 29. Direct repeats were detected by visual inspection of retroelement flanking sequences. Novel repeats were assigned to subfamilies according to the nomenclature proposed by Daboussi and Capy 29. For multiple alignments, BLAST pairwise search, Vector NTI and Clustal W programs 34 were used.

Two Magnaporthe grisea strains – P1.2 and Guy11 – were used in this study. As all rice pathogenic Magnaporthe grisea isolates, these strains are also pathogenic on barley. Fungal strains were grown and stored as described 35.

Total RNAs were extracted from M. grisea mycelium, spores (strain P1.2) and infected barley leaves at 0 h, 8 h and 24 h (strain Guy11) using Rneasy Plus Mini Kit (Qiagen). cDNA synthesis was performed with 2 μg of total RNA using random primers and ThermoScript RT-PCR System (Invitrogen). The level of WEIRD and MGL transcription was assessed by real-time PCRs with genomic primers corresponding to the 5' and 3'-terminal parts of the elements using an ABI Prism 7700 Sequence Detection System (Applied Biosystems) and Power SYBR Green PCR Master Mix (Applied Biosystems); primers Wfor, Wrev and MGLfor, MGLrev for WEIRD and MGL, respectively. Primer sequences for real-time PCR were chosen using Primer Express software (Applied Biosystems) and are presented in supplementary material (Additional file 4). All measurments were carried out in quadruplicate and expression levels were normalized to Ilv5 and Ef1 using the following formulae 2−ΔCt=2−(Ctsample−Ctrefrence) MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacH8akY=wiFfYdH8Gipec8Eeeu0xXdbba9frFj0=OqFfea0dXdd9vqai=hGuQ8kuc9pgc9s8qqaq=dirpe0xb9q8qiLsFr0=vr0=vr0dc8meaabaqaciaacaGaaeqabaqabeGadaaakeaacqaIYaGmdaahaaWcbeqaaiabgkHiTiabfs5aejabboeadjabbsha0baakiabg2da9iabikdaYmaaCaaaleqabaGaeyOeI0IaeiikaGIaee4qamKaeeiDaq3aaSbaaWqaaiabdohaZjabdggaHjabd2gaTjabdchaWjabdYgaSjabdwgaLbqabaWccqGHsislcqqGdbWqcqqG0baDdaWgaaadbaGaemOCaiNaemyzauMaemOzayMaemOCaiNaemyzauMaemOBa4Maem4yamMaemyzaugabeaaliabcMcaPaaaaaa@50D3@[36]. All RT-PCR experiments were performed against the negative RT "-" controls (no reverse transcriptase added at the stage of the first strand cDNA synthesis) to control the DNA contamination. Only those samples displaying negative results on the RT "-" control experiments were further analyzed. To identify transcriptionally active MINEs a general RT-PCR was performed using primers designed to the 5' terminal part of WEIRD (in the sense orientation), and to the 3' end of MGL (in the reverse orientation) under the following conditions: 95°C – 2 min; 95°C – 25 s, 62°C – 25 s, 72°C – 2 min; 25 cycles. The products of RT-PCR were then diluted into 40 times and used as a template for the nested PCR (95°C – 2 min; 95°C – 25 s, 62°C – 25 s, 72°C – 2 min; 25 cycles). Nested PCR was performed to reduce the background originated due to the use of primers corresponding to repetitive sequences. The obtained products were sequenced by Genome Expressed (Meylan, France). Three transcriptionally active MINEs identified during this study were further analyzed by real-time RT-PCR using primers designed for the specific WEIRD-MGL junctions (q1 and q2 for MINE1; q3 and q4 for MINE2; q5 and q6 for MINE3).