Deletion analysis of the 3' long terminal repeat sequence of plant retrotransposon Tto1 identifies 125 base pairs redundancy as sufficient for first strand transfer.

Retroviruses and many retrotransposons are flanked by sequence repeats called long terminal repeats (LTRs). These sequences contain a promoter region, which is active in the 5' LTR, and transcription termination signals, which are active in the LTR copy present at the 3' end. A section in the middle of the LTR, called Redundancy region, occurs at both ends of the mRNA. Here we show that in the copia type retrotransposon Tto1, the promoter and terminator functions of the LTR can be supplied by heterologous sequences, thereby converting the LTR into a significantly shorter sub-terminal repeat. An engineered Tto1 element with 125 instead of the usual 574 base pairs repeated in the 5' and 3' region can still promote strand transfer during cDNA synthesis, defining a minimal Redundancy region for this element. Based on this finding, we propose a model for first strand transfer of Tto1.

Introduction

Similar to their retrovirus relatives, retrotransposons rely on element-encoded reverse transcriptase to generate DNA copies of their own sequence from an RNA template. The cDNA copy can subsequently be inserted by integrase into the host genome. The general outline of this retroelement life cycle is textbook knowledge. However, many details of the reverse transcription process remain to be elucidated. Furthermore, investigations of individual retroelements frequently uncover unexpected features, pointing to a considerable diversity in many aspects of the replication mechanism. So far, almost all detailed studies of the retrotransposition mechanism were carried out on microbial elements or with tissue culture systems, where low transpositional activity can be compensated by analysis of large numbers of cells. In metazoans, the activity of retroelement promoters is often restricted to specific growth stages or tissues, and defense strategies such as RNA-based silencing further restrict element activity (Beauregard et al., 2008; Jordan and Miller, 2009; Symer and Boeke, 2010; Tenaillon et al., 2010; Wolf and Goff, 2008). Thus, poor expression poses a major impediment to functional analysis of retroelements in higher organisms.

We are interested in Tto1, a copia type long terminal repeat (LTR) containing retrotransposon from tobacco (Nicotiana tabacum). Tto1 has a single open reading frame that encodes a 1338 amino acid poly-protein, consisting of structural virus-like particle (VLP) component gag, followed in frame by enzyme components protease, integrase and reverse transcriptase/RNase H. Tobacco contains approximately 30 copies of Tto1 per haploid genome. Activity is low, presumably due to RNA- (methylation-) based silencing. When tobacco cells are kept in tissue culture for several weeks to months, a drastic increase in copy number is observed, indicating transpositional activity under these specific conditions (Hirochika, 1993). One Tto1 element copy with transpositional competence was isolated and found to be active when introduced into several different plant species (Hirochika, 1993; Hirochika et al., 1996; Okamoto and Hirochika, 2000). In these and similar experiments with the related Tnt1 element, activity upon introduction into a new host is initially detectable, but further transposition is efficiently down-regulated as soon as transposed copies accumulate (Pérez-Hormaeche et al., 2008). This and other characteristics pointed out above make systematic analysis quite challenging even in heterologous hosts, where initial activity can be observed experimentally.

To overcome these problems, and to allow a systematic functional analysis of Tto1, we have adopted a synthetic biology approach. We have replaced the endogenous promoter of Tto1 by a heterologous promoter. Interestingly, the most abundant “natural” mRNA initiation site does not to support translation (Böhmdorfer et al., 2005; Hirochika, 1993), but fusion of heterologous promoters to a further upstream site allows activation of Tto1. After initially testing a constitutive promoter, we have more recently generated an inducible transposition system for Tto1, called iTto1 (inducible Tto1). We have shown that iTto1 allows transposition in whole plants, and that transposed copies of the engineered element are identical to the original element (Böhmdorfer et al., 2010). The latter fact follows from the replication cycle of LTR retrotransposons, in which the 5′ LTR of daughter elements is copied to a large extent from the 3′ LTR of the mother element, so that transposed copies carry the original LTR-borne promoter. The finding that the engineered element displays a complete life cycle supports its use for analysis of those steps of the Tto1 transposition cycle that happen after promoter activation, which differs between engineered iTto1 and unmodified Tto1. A major advantage of the inducible system is that activation can be precisely timed to optimize abundance of transposition intermediates. Another advantage is that transformation of plants with constructs can be separated in time from activation and analysis. This allows the use of identical starting material and/or conditions for repeated experiments, in order to improve reproducibility and comparability. The Tto1 constructs of this work also contain introns, allowing us to distinguish RNA-derived cDNA from the transgenic founder copy.

In this work, we investigate the strand transfer function of the LTR. The LTRs of Tto1 are 574 base pair identical sequences at both ends. In the life cycle, these two identical sequences have distinct roles. The 5′ LTR provides a promoter, contains the transcription initiation site, and part of the nontranslated 5′ leader sequence. In contrast, the LTR at the 3′ end provides transcription termination site(s) and the nontranslated 3′ part of the mRNA. In this way, a portion of the LTR sequence is present at both ends of retrotransposon RNA. This sequence is called Redundancy region and plays an important role in the reverse transcription process: reverse transcriptase has to switch from the 5′ end to the 3′ end of a (usually different) RNA molecule while reverse transcribing this sequence, in order to form a DNA replica of the element (Basu et al., 2008; Boeke and Stoye, 1997; Telesnitsky and Goff, 1997; Wilhelm and Wilhelm, 2001). However, due to the dense arrangement and possible interdigitation of functional domains in the LTR, it is not clear which part of the Redundancy region is actually essential, and which part is only present due to functional constraints with respect to transcription initiation and termination. We show below that transcription termination signals present in the LTR can be replaced by termination sequences from regular plant genes without abolishing the strand transfer function, thereby separating the functions of the 3′ LTR in transcription termination and in strand transfer. Using a series of deletion constructs, we define a 125 bp sequence as sufficient for first strand transfer during the reverse transcription, implying that only a part of the existing redundancy between 5′ and 3′ end of the Tto1 mRNA is essential for the first strand transfer. The 125 bp sequence contains a single-stranded motif that may facilitate homology search by the emerging cDNA. Truncation of LTRs to reduce the length of the sequence repeats that flank a retrotransposon may also have consequences for cellular surveillance, which targets sequence repeats for RNA-based silencing.

Results and discussion

Serial deletion of the 3′ LTR sequence and appendage of a transcription termination sequence to Tto1

In order to facilitate functional analysis of the 3′ LTR of copia-type retrotransposon Tto1, we made use of an engineered version of this element (cf. Introduction; Figs. 1a and b). This engineered element contains an inducible promoter and two introns. Due to the inducible promoter, the element is silent when introduced into Arabidopsis plants. However, induction of the promoter by β-estradiol results in transposition, indicating that the engineered element has all relevant properties necessary for the completion of the life cycle (Böhmdorfer et al., 2010). For the purpose of this work, however, we did not want secondary effects due to newly integrated Tto1 copies. We therefore used a construct with an E to A amino acid change in the integrase active site (amino acid 583 of the Tto1 ORF; Böhmdorfer et al., 2008). The ensuing absence of endonuclease activity abolishes integration, and as a consequence we restrict our observations to a single cycle of cDNA synthesis, controlled by promoter induction.

Fig. 1

DNA constructs used in this work. (a), Schematic drawing of tobacco retrotransposon Tto1, which has one single open reading frame and belongs to the Ty1/copia group of elements. The smaller box with rounded edges symbolizes the single open reading frame, encoding structural gag and enzyme components protease, integrase and reverse transcriptase. (b), Engineered Tto1 element for chemically induced transcription and cDNA formation has an inducible promoter fused to a 5′ long terminal repeat (LTR) shortened by 171 nt, and contains two introns. (c), Constructs with deletions at the 3′ LTR are analogous to (b), and contain a transcription terminator following the truncated 3′ LTR.

We generated a set of 3′ deletion constructs, schematically depicted in Fig. 1c. Fig. 2a shows the sequence of the 3′ end of Tto1 (undeleted control construct). Deletion constructs lack part of this sequence: construct A ends with nt T printed in bold under letter A, and the end points of constructs B, C, D, and E are similarly indicated. As part of the LTR, the sequence depicted also occurs at the 5′ end of all Tto1 constructs. This sequence identity is reduced by the deletions, and the number of base pairs that are identical between 3′ and 5′ end are indicated in parentheses above the deletion end points, namely 336, 222, 125, 25, and 3 bp, respectively, for constructs A to E. Deletion end point A was chosen with specific reference to the Tto1 mRNA. It corresponds to a previously mapped major transcription termination site for the longest Tto1 mRNAs (Böhmdorfer et al., 2005 and Fig. 3b). Deletion construct A therefore contains all sequences present at the 3′ end of these long “natural” Tto1 mRNAs.

Fig. 2

Sequence redundancy at the 5′ and 3′ ends of engineered Tto1 elements. (a), Sequence present at both ends of the engineered Tto1 element without deletion, which corresponds to nt 172–568 of the long terminal repeat (LTR). The sequence at the 3′ end is shortened in the deletion constructs. Deletion end points are written in bold print and marked by letters A to E. The extent of remaining sequence overlap (in base pairs) is indicated in parentheses. In deletion construct E, the 3′ sequence has only 3 nt redundancy with the 5′ end of the construct. The sequence repeat in construct D is 25 nt long, in construct C 125 nt, in construct B 222 nt, and in construct A 336 nt. (b), In all deletion constructs, the last residue of the LTR-derived 3′ sequence as shown in panel (a) is followed by a spacer sequence (bold, small letters), and by the transcription termination sequence of pea rbcS-3A (small letters). (c), In the sequence transcribed as 5′ end of the mRNA of all engineered constructs, a 32 nt extension (small letters) precedes the LTR-derived sequence (capital letters; full sequence shown in (a)).

Fig. 3

Transcript abundance and transcription termination sites of Tto1 constructs. (a), Transgenic lines containing either construct C, or D were used for RNA isolation and subsequent RT-PCR in induced and un-induced state, indicating active transcription after induction. (b), mRNA was enriched by binding to oligo dT matrix and subsequent nested RT-PCR from either non-transformed plants (no Tto1), or from plants containing a Tto1 construct without 3′ deletion (Tto1). After induction of the Tto1 promoter, two major termination sites result in distinct bands (arrowheads to the left). (c), Similar to panel (b), mRNA from induced plants containing deletion constructs was analyzed by RT-PCR. Size determination of PCR products and sequencing of excised bands of constructs C and D indicates that upon deletion of the downstream transcription termination site of the LTR (upper arrowhead for deletion construct A), transcription can proceed further into the appended rbcS terminator (dot for deletion constructs B to D). As in Fig. 1, the stippled box symbolizes the LTR, whereas the dark box symbolizes the appended rbcS terminator. Arrowheads and dots indicated transcription termination positions. “Contr.” lanes of panels (a) and (b) indicate control RT-PCR to amplify AtUBC9 mRNA as a control.

Introduction into Arabidopsis plants and investigation of transcription termination

The constructs described above and in Figs. 1 and 2 were introduced into Arabidopsis by T-DNA transformation. Because the random integration of T-DNA results in a broad range of expression levels, we investigated transcript abundance after induction by β-estradiol. For each construct, we identified transgenic lines with a comparable mRNA abundance, as judged by RT-PCR. The lines finally chosen for further experimentation did not differ by more than a factor of three in transcript abundance after induction (data not shown; see also below). Fig. 3a shows the results for two lines containing deletion constructs C and D. mRNA of construct D is apparently slightly more abundant than mRNA of construct C. It is noteworthy that the PCR primers used flank an intron, so that the RT-PCR signal reflects exclusively mRNA abundance.

We also wanted to assess the effect of appending deletion constructs with the rbcS terminator. Poly A containing mRNA was enriched by oligo dT beads, and used for RT-PCR reactions to determine mRNA end points. The undeleted 3′ LTR contains two major termination sites (Fig. 3b, arrowheads and Böhmdorfer et al., 2005). A major fraction of mRNAs has their poly A tail appended to a position between deletion end points of constructs D and E (Figs. 2a and 3b, lower arrowhead). As mentioned above, another fraction of the Tto1 mRNA terminates around the deletion end point of construct A (Figs. 2a and 3b, upper arrowhead and Böhmdorfer et al., 2005). Fig. 3c visualizes transcript termination sites of constructs A to D. We were particularly interested to assess transcript termination of constructs with more extensive deletions, because these constructs do not contain the downstream termination site. The resultant gels show a strong band common to all constructs, marked by an arrowhead. Sequencing of 8 clones from deletion constructs C and D confirmed the expected transcription termination sites as shown in the symbolic figures above each gel picture. These termination sites lie between nt 4890 and 4920 of the Tto1 sequence, around deletion end points D and E, and correspond to previously mapped abundant termination sites (Böhmdorfer et al., 2005). Similar to the undeleted Tto1, deletion construct A produces a second, longer mRNA species. The size of this mRNA species is consistent with the use of the same termination signals as present in undeleted Tto1 (Fig. 3b, upper arrowhead). In contrast, all other deletion constructs lack this termination region. However, constructs B to D nonetheless have a second termination site that produces longer mRNA molecules (dot in Fig. 3c). The size of the amplified fragments suggests for these constructs that transcription can continue beyond the deletion end points of the LTR into the termination sequence of the rbcS gene, which was used to extend the deleted LTRs. Sequence determination of long cDNAs for constructs C and D showed that the long transcripts indeed ended at the positions of the rbcS terminator previously mapped as poly A addition sites in pea (which was either nt 1658, or nt 1678 of accession X04333; Coruzzi et al., 1984; note that for RT-PCR with mRNA from constructs C and D, we chose a downwards reading primer that binds closer to the LTR, resulting in fragments that are ca. 100 nt shorter compared to the other deletion constructs, or to undeleted Tto1). We want to emphasize that these experiments do not allow conclusions about the relative abundance of the longer vs. shorter transcripts, because reverse transcription and amplification efficiency might differ between fragments of different length, even if flanked by the same primer binding sites. However, the absence of prominent PCR fragments between the two bands discussed above indicates absence of additional termination sites for constructs B, C, and D.

We therefore concluded that in constructs B, C, and D, termination occurs either at the relative beginning of the LTR sequence, at a position also used in an undeleted LTR, or it proceeds beyond the deletion end points, to occur under guidance of the rbcS terminator.

Reverse transcriptase activity of deletion constructs

We expected that all deletion constructs produce the same set of Tto1 proteins in proportion to their mRNA abundance after induction, because translation initiation and mRNA start regions are identical in all deletion constructs. The latter two functional elements are in fact identical to constructs already successfully tested for activity (Böhmdorfer et al., 2010). A previously employed assay for activity of Tto1 proteins uses PCR-based detection of cDNA with primers flanking an intron of the engineered Tto1 element (Figs. 1 and 4a), because a prerequisite for cDNA synthesis is formation of virus-like particles containing enzymes and RNA. This assay is more sensitive than e.g. immunological detection of gag protein (Böhmdorfer et al., 2008). Fig. 4b shows that all constructs allowed formation of reverse transcripts, as judged by the presence of an intron-less PCR band (band denoted by an encircled “−” symbol to the right). The PCR fragment copied from the T-DNA-borne engineered Tto1 element, which is larger due to the intron (band denoted by an encircled “+” symbol to the right), differs significantly in abundance between reactions. This was caused by adjusting the amount of template DNA to obtain equal intensity of amplified cDNA. As indicated before, the selected plant lines may have different copy numbers of the transgene, and slightly differing mRNA abundance, which would be sufficient to explain these differences in band intensity. In particular, these differences do not indicate that the mRNA of any deletion construct is less efficiently reverse transcribed.

Fig. 4

PCR reactions to assess efficiency of first strand transfer by deletion constructs. (a), Schematic depiction of engineered Tto1 (top) and completed cDNA copy (bottom), and of positions of primers for “long” and “short” PCR reactions. (b), “short” PCR reaction with total DNA from plants induced to express deletion constructs A to E, or from undeleted control shows presence of the intron-containing genomic copy (250 bp band; circled + symbol to the right), and of intron-less cDNA (150 bp band; circled − symbol to the right). The template DNA used in the reactions was adjusted such that all transgenic lines gave the same amount of cDNA. (c), To assess whether the cDNA is in all cases the product of an orderly first strand transfer reaction, “long” PCR reactions were carried out with total DNA (same amount of DNA as used in panel (b)). Constructs A to C, but not D or E allow detection of the product of orderly first strand transfer. (d), Same as panel (c), but DNA was prepared from un-induced plants, in which case only T-DNA based complete Tto1 element (Tto1) can be amplified.

The finding of detectable amounts of reverse transcripts with all deletion constructs was somewhat surprising, because we expected a drop in abundance of reverse transcripts for deletion mutants that cannot carry out the first strand transfer. Interestingly, in previous experiments we had described a construct with deletion at the 5′ end, which was capable of reverse transcription, but appeared nonfunctional regarding transposition. We had hypothesized that if an orderly first strand transfer, and therefore generation of a full length cDNA copy with restored 3′ LTR, cannot be achieved, some constructs can nonetheless produce aberrant, smaller reverse transcripts (Böhmdorfer et al., 2005). This hypothesis may also apply to some of the deletion constructs of this work. Reverse transcriptase and mRNA are confined to virus-like particles, and absence of the normal substrate may allow less specific template copying processes after nonspecific foldback-priming of the mRNA. Moreover, the sensitive PCR reaction can also amplify single-stranded DNA molecules, and the section of the 5.3 kb Tto1 sequence amplified by the “short” PCR reaction covers only ca. 250 bp.

We therefore modified the PCR analysis strategy. Instead of a short amplified fragment, PCR primers were chosen that amplified a longer fragment that encompasses the complete 3′ LTR (Fig. 4a, “long” PCR). To normalize the DNA applied to each reaction, we used the same amount of template DNA as for Fig. 4b, which resulted in an equal level of product in the “short” PCR reaction for all constructs. Fig. 4c shows that deletion constructs A, B, and C allowed abundant amplification of the long PCR fragment. In contrast, only a trace amount of this band can be seen for construct D, and no such band was amplified for construct E. We therefore concluded that constructs D and E, the two deletions with the smallest sequence overlap with the mRNA 5′ end, cannot form full length cDNA. As a control reaction (Fig. 4d), DNA from un-induced plant material was used as a template. Fig. 4d shows that the “long” PCR assay indeed detects cDNA, not the T-DNA-borne deletion constructs, due to the lack of a binding site for the upwards reading primer (cf. Fig. 4a). We carried out an additional control reaction, in which the products of the “long” PCR reaction were gel-isolated and used for another “short” PCR. Supplemental Fig. 1 shows that the bands of Fig. 4c, lanes A to C, are predominantly devoid of intron, as expected for PCR amplification of the 3 kb fragment from the intron-less cDNA.

Visualization of cDNA by DNA gel blot analysis

The PCR-based analysis of reverse transcripts was complemented by direct visualization of extra-chromosomal cDNA. DNA was prepared from induced plants containing the deletion constructs, and compared to DNA from plants transformed with the standard construct iTto1 and to DNA from untransformed plants (Fig. 5). The DNA was digested with restriction enzyme NotI for easier handling. The ensuing high molecular weight DNA nonetheless migrates at the size exclusion limit of the conventional electrophoresis gel as a single broad band (see ethidium bromide stained gel pictures, bottom part of Fig. 5). In contrast, the linear Tto1 cDNA, which has no NotI restriction site, migrates at the position corresponding to its length of 5.3 kb. Fig. 5 shows the induction of cDNA in iTto1 containing plants (Fig. 5, lane 2 vs. 3). Likewise, full length cDNA can be detected in induced plants containing deletion constructs A (lane 4), B (lane 5), or C (lane 6). In contrast, cDNA produced by construct D is below detection limit of this experiment (lane 7), and may not have a single defined size. Thus, there is an absolute correlation between the positive signal in the “long” PCR reaction (Fig. 4c), and the presence of full length cDNA as detected by Southern hybridization. We therefore concluded that deletion constructs A to C are capable of producing full length cDNA of Tto1, whereas no such DNA can be visualized for construct D by two different methods.

Fig. 5

DNA gel blot to visualize extrachromosomal Tto1 cDNA. Genomic DNA from plant lines that were either nontransformed (N), contained an undeleted inducible Tto1 construct (iT; lanes 2, 3), or deletion constructs A, B, C, or D (lanes 4–7, respectively), was digested with 8 base cutting enzyme NotI and electrophoresed. Hybridization with a Tto1-specific probe visualized integrated Tto1 constructs as large genomic fragments that migrated at the separation limit of the gel, whereas extrachromosomal Tto1 cDNA had the expected size of ca. 5.3 kb. The cDNA band was detected for positive control iTto and deletion constructs A to C, but not D (black dots). Molecular weight marker sizes are indicated to the left. The bottom panel shows ethidium bromide stain picture of the gel prior to transfer. The strong band in the top panel varies according to both amount of DNA loaded, and copy number of the Tto1 transgene.

A 100 bp sequence is essential for efficient first strand transfer

The results of Figs. 3–5 also imply that transcripts with an “early” termination site, which lies between deletion end points D and E, are not substrates for orderly first strand transfer. However, transcription termination in the artificially appended rbcS terminator can generate templates for first strand cDNA elongation by template switch. Constructs C and D are of particular interest in this respect. They produce mRNA species that differ exactly by a sub-terminal deletion that restricts the length of LTR sequence present at the 3′ end, i.e. the Redundancy region. As construct D does not produce full length cDNA, the template switch of the RT enzyme from the 5′ end of mRNA to the 3′ end (of the same or another mRNA) is compromised for construct D, but not for construct C, implying that the 100 bp sequence difference between the two constructs plays an essential role in the strand transfer.

Mechanistic implications and model building

In addition to sub-terminal deletions discussed above, there is one more feature of the inducible Tto1 constructs with potential influence on first strand transfer: The 5′ ends of the mRNAs transcribed from engineered Tto1 constructs contain a stretch of 32 nt that are not present at the 3′ end (this sequence was introduced with the heterologous promoter; Fig. 2c, sequence stretch in lower case). cDNA produced from this part of mRNA can therefore not base pair with the mRNA 3′ end, precluding productive participation in strand transfer. Experimental comparison of a construct with this mRNA 5′ extension versus a construct with only 6 nt non-LTR sequence at the 5′ mRNA end showed no difference in activity (G. Böhmdorfer, unpublished), suggesting that this extra sequence does not decrease the efficiency of first strand transfer. However, the extension precludes the most simple model of strand transfer, namely that the 5′ region is first copied from mRNA in its entirety, by reverse transcriptase synthesizing along the complete 5′ part, and only then this cDNA (called strong stop cDNA in retroviruses) starts base pairing with the complementary sequence at the 3′ end, for continuation of the reverse transcription process. Strand transfer has been investigated extensively for the HIV retrovirus (Basu et al., 2008), where this mode of first strand transfer was called terminal transfer. In contrast, a template switch before reverse transcriptase has reached the mRNA 5′ end, called invasion transfer, is possible for the engineered Tto1 constructs. With the data available, we cannot rule out the possibility that unmodified Tto1 uses both modes of first strand transfer, but as the invasion mode is apparently sufficient to promote transposition of engineered Tto1 (Böhmdorfer et al., 2010), it may be generally dominant.

In order to generate models for the mechanism of invasion transfer, we used an RNA folding program to reveal differences between the 3′ end structure of mRNA C and mRNA D. Both mRNAs can fold similarly, but mRNA C contains additional structure, formed by the 100 nt present exclusively in mRNA C (Fig. 6). Another in silico test assessed the co-folding of two mRNA C molecules. The purpose of this latter test was to find out whether the 100 nt sequence not present in deletion construct D (but present in construct C) can support a dimer arrangement in which the 5′ and the 3′ ends (of the same, or of two different mRNAs) are in close proximity already before reverse transcription starts. Retrovirus mRNAs usually form dimers. So far, however, dimer formation in retroviruses was functionally linked to packaging, but not to the strand transfer process (dimerization does not usually align the 3′ and 5′ ends of two viral mRNAs; Jewell and Mansky, 2000; Paillart et al., 1996). Interestingly, a complex consisting of two mRNAs and two primer tRNAs was shown to form with yeast retrotransposon Ty3, which belongs to the retrovirus-related gypsy family (Gabus et al., 1998), and binding of tRNA primer to the mRNA 3′ end may facilitate strand transfer in HIV (Brulé et al., 2000). However, no dimer structure with spacially close 5′ and 3′ ends was proposed by the RNAcofold program (data not shown). Tto1 is a comparatively simple representative of the copia group of retrotransposons, and certain features found in retroviruses or their relatives, the gypsy/Ty3 group of retrotransposons, may be missing.

Fig. 6

Potential secondary structure of Tto1 mRNA and single stranded cDNA (strong stop cDNA). (a), Potential structure of strong stop cDNA (left), of the mRNA 3′ end of construct C (middle; starting at the 3′ LTR and ending at poly A tail in the appended rbcS termination region), and of the mRNA 3′ end of construct D (right). A loop sequence with perfect correspondence in the cDNA (encircled) is single stranded in more than 95% of all ensemble structures. (b), Close-up of the structurally conserved hairpins, which are complementary between cDNA (left) and construct C mRNA (right).

We hypothesize that a homology search between emerging cDNA and an mRNA molecule is the relevant and essential step. In our model, the 100 bp sequence in question supports homology search between 5′ generated cDNA and 3′ sequences. As shown in Figs. 6 and 7, a 9 nt loop is present in mRNA C, but absent from mRNA D. It is the sequence with the highest probability of being unpaired in the mRNA C 3′ end (Fig. 7b). No comparable single-stranded region exists in construct D mRNA (Fig. 7c). Likewise, models for the structure of emerging cDNA show a stem-loop formed by the exact reverse complement of the stem-loop sequence in mRNA C, and again the loop in the cDNA has the highest probability of being unpaired in the whole cDNA sequence (more than 90% probability; Fig. 7a). A prerequisite for this model is that the emerging cDNA-mRNA heteroduplex formed at the mRNA 5′ end is efficiently hydrolyzed by RNase H to single stranded cDNA (Champoux and Schultz, 2009; Eickbush and Jamburuthugoda, 2008). Fig. 6b shows a closeup of the relevant regions. We hypothesize that an early and kinetically favored interaction between emerging cDNA of the 5′ end and the 3′ end of an mRNA molecule involves the kissing of these unpaired loop residues. Once these residues form base pairs, pairing is extended to melt the surrounding secondary structure, which is easily possible due to the perfect complementarity of cDNA and mRNA. Quantitative comparison indicated that kissing of the two hairpins contributes ca. − 26 kJ/mol. If cDNA synthesized up to that point continues hybridization with the mRNA 3′ end, both sequences have to give up secondary structure worth − 150 kJ/mol, but the gain from forming a perfect heteroduplex is − 385 kJ/mol. In sum, the annealing of the cDNA from the loop to its 5′ end is therefore strongly favored (− 235 kJ/mol) over separate secondary structure formation of single-stranded cDNA and mRNA. Fig. 8 shows critical steps of the model, who's future testing by generation of Tto1 constructs with deletions in the region of the postulated loop should facilitate further refinement.

Fig. 7

Diagram of the probability of being unpaired (9 nt sliding window) for strong stop cDNA (a), for the construct C mRNA end (starting at the 3′ LTR; (b)), and for the construct D mRNA end (starting at the 3′ LTR; (c)). Both the first strand cDNA, and the construct C mRNA end contain a short sequence with a probability of being single stranded larger than 90%. The two sequences are the exact reverse complement (sequence shown as closeup). This stretch was deleted in construct D and is part of the 100 nt essential for strand transfer.

Fig. 8

Model for first strand transfer of Tto1 reverse transcription. (a), Reverse transcription starts with alignment of the tRNA primer. The RNase H function of reverse transcriptase degrades the mRNA of the emerging RNA–DNA duplex, starting after the RNA–RNA hybrid formed by primer and mRNA, ending at a sliding window of short heteroduplex that is bound to the reverse transcriptase. The emerging single-stranded cDNA engages in secondary structure formation (b), but at least one characteristic sequence remains single-stranded. (c), The 3′ end of the same or of another mRNA molecule contains a complementary loop, which starts base pairing. Perfect complementarity between cDNA and mRNA 3′ end favors formation of heteroduplex, thereby replacing secondary structures (d). In this complex, donor and acceptor mRNA are closely aligned, facilitating template switch of reverse transcriptase (curved arrow) before RT reaches the mRNA 5′ end.

Significance of the “minimal” Redundancy region

In addition to insights into mechanistic details of the reverse transcription process, the definition of a minimal Redundancy region has another potentially interesting implication. Retrotransposons are in most cases repetitive sequences. A generally accepted model suggests that the presence of several to many copies in the genome facilitates down-regulation of their activity by RNA-based silencing mechanisms (Reuter et al., 2009; Tijsterman et al., 2002). A key feature in this process is production of RNA species in antisense orientation, which are more likely to occur from different copies than from a single one. However, the long terminal repeat of a single retrotransposon copy may be a comparable target for antisense transcript production, and thus contribute to retrotransposon silencing. For single copy retrotransposons, the LTR may even be the only structure that provokes silencing. Future experiments will thus aim to investigate whether reduction of the LTR to a minimal size reduces silencing of single copy elements and thus leads to higher activity of the ensuing element.

Materials and methods

Plant transformation and growth

Plants were accession Col-0, transformed with constructs based on β-estradiol-inducible vector pER8 (Zuo et al., 2000) by the floral dip method (Clough and Bent, 1998). Lines were propagated on soil under standard greenhouse conditions. A minimum of 10 lines was tested for each construct to find lines with high, comparable expression levels. For induction of the transgene, seedlings were germinated on Agar medium (MS salts with vitamins, 1% sucrose) and transferred for induction to liquid medium (half strength MS medium with vitamins, 1% sucrose, 5 μM β-estradiol; 23 °C, 16 h light 8 h dark cycles).

DNA constructs

Plasmid pTAsTto3.I (Böhmdorfer et al., 2008) was partially digested with Pme I and ligated with oligonucleotide GGC ATT TAA ATG CC to replace the Pme I site at the end of the Tto1 reading frame by a (unique) Swa I site. An Xho I Pvu II fragment of the ensuing plasmid was cloned into Xho I Pvu II digested vector pER8 (Zuo et al., 2000) to give vector pER8newTtoSwa. Using oligonucleotides ATG CCC TTC GAA CAG CTG GCG AAA GGG GGA TGT GCT and CGG CCC ATT TAA ATC GCC CAC TAG TTG GTC GAT CCA GGC CTC CC, and pER8 as a template, the terminator of rbcS3A was amplified by PCR and cloned into the Sma I site of vector pSK to give pSK-rbcS. Fragments containing deleted versions of the LTR were also generated by PCR, using Tto1 as a template, digested with Swa I and Pfl MI, and cloned into Swa I, Pfl MI digested pSK-rbcS. The oligonucleotides for amplification of LTR sequences were GAT CGG ACA TGT TGA CCA AGA CT as a downwards reading primer, combined with CGG CCC CCA ACT AGT GGA ATA AAT GAC ACA ATA TTT AAC GT (deletion construct A), CGG CCC CCA ACT AGT GGA TAA TTT GGA GGT ACA ACA ATT CC (deletion construct B), CGG CCC CCA ACT AGT GGA TTT TGT CAC TCC CCT GTT AGG AA (deletion construct C), CGG CCC CCA ACT AGT GGA TAC AAT GAA ATG GAA GGG GGT ATT (deletion construct D), and CGG CCC CCA ACT AGT GGA TTT ATA GTT TTG AGA TAG GGA CC (deletion construct E), respectively. Ensuing plasmids were digested with Pvu II and Swa I, and inserted into Pvu II and Swa I digested vector pER8newTtoSwa to give deletion constructs A, B, C, D, and E. Sequences amplified by PCR were sequenced to confirm correctness.

PCR reactions using DNA templates

LA Taq (TaKaRa) and Go Taq (Promega) were used in standard reactions under conditions recommended by the manufacturer. DNA was prepared as described (Böhmdorfer et al., 2005). Oligonucleotides used for “short” PCR (Fig. 4b) were GGT GGA AAG AGA GAC TGG TAA and CCC GTA ATT GAT CAT AAG AGA, oligonucleotides used for “long” PCR (Figs. 4c and d) were GGT GGA AAG AGA GAC TGG TAA and TGT TAG GAT CCG GTG GCA CTA AAC ACT.

RT-PCR

RNA was prepared using the RNeasy plant mini kit (Qiagen). For assessment of mRNA abundance (Fig. 3a), primers AGC TCG AAG AGT TGT ATG CCT CT and CCC GTA ATT GAT CAT AAG AGA were used. These primers flank the first intron, allowing exclusive detection of mRNA. cDNAs of 3′ ends of polyadenylated Tto1 mRNA were enriched and subcloned as described (Böhmdorfer et al., 2005). After association with magnetic oligo dT beads and reverse transcription, nested PCR was carried out with the following primers: first step, CAT CGC AGC AAC GGA GGC TTG C combined with CGG ACG CTC AGC CAG GTT TTT TTT TTT TTT TTT TTT TTT TTT; second step, GTG CTA TCC ACC TTG CGA AGA ATG C combined with CGG ACG CTC AGC CAG GTT T. For deletion constructs C and D of Fig. 3c, a third PCR step was included with primers GAT CGG ACA TGT TGA CCA AGA CT and CGG ACG CTC AGC CAG GTT T. Controls used primers TCC CCC GGG AGA TCT AGG ATG GCA TCG AAA CGG ATT TTG AAG and GGG GTA CCA GAT CTC AGC CCA TGG CAT ACT TTT GGG T to amplify AtUBC9 (At4g27960) mRNA.

DNA gel blot

DNA was prepared using the Illustra DNA extraction kit Phytopure (GE Healthcare) and processed as described (Böhmdorfer et al., 2008; Böhmdorfer et al., 2010), except that genomic DNA was digested with NotI prior to electrophoresis.

RNA structure prediction

All secondary structure predictions were carried out using tools of the ViennaRNA suite (Hofacker et al., 1994), using standard folding parameters including dangling end energies for the bases adjacent to a helix (− d2 option). Cofold analyses were performed using RNAcofold (Bernhart et al., 2006). For prediction of the secondary structure of strong stop cDNA, special energy parameters for DNA folding were used (Mathews et al., 2004). For the computation of the energies of cDNA/mRNA hybrid stacks, parameters of Wu et al. (2002) were used. RNAup (Mückstein et al., 2006) was used to compute the probability that regions of the strong stop cDNA, or of mRNAs, remain unpaired.

The following are the supplementary materials related to this article.

Supplemental Fig. 1

The 3 kb bands of “long” PCR lanes A, B, and C were copied from cDNA. As shown in panel (a) (a copy of Fig. 4c), a “long” PCR was carried out to detect the 3´ half of completed cDNA. Deletion constructs A, B, C, and undeleted Tto1 (Tto1) allow synthesis of this band. In contrast, the band is characteristically absent from panels D and E, indicative of the absence of full length cDNA from plants carrying constructs D and E (see text for further interpretation). In this experiment, the upward reading primer can bind to the genomic copy of the undeleted control construct, but not to the genomic copy of any deletion construct, because its binding site is part of the deleted region. “Long” PCR products of deletion constructs should therefore lack introns. As gel resolution does not allow distinguishing a 100 bp difference of the 3 kb sequence, an additional control reaction was carried out. The visible 3 kb bands of panel (a), and material from the same gel position of lanes D and E, were excised and used for a “short” PCR (panels (b) and (c); cf. Fig. 4a). Panel (c) is identical to panel (b), but a lower number of cycles were used to underscore the predominance of the lower (intron-less) band in lanes A to C. Both panels show that the abundant bands from deletion constructs A, B, and C are almost completely devoid of introns, indicative of their generation from cDNA. In contrast, the intron-less bands for constructs D and E are much less intense and only visible in panel (b), suggesting that they were derived from low amounts of largely aberrant reverse transcripts present in these transgenic lines. Intron-containing (440 bp) fragments are also present in small amounts for all deletion constructs. They vary much less in abundance between lanes A to C and D, E, and may be derived from a low background of genomic DNA present in the 3 kb region of the “long” PCR gel (panel (a)). Molecular weight marker positions are indicated to the right.