Abstract

Formation of RNA-DNA hybrid, or R-loop, was studied in vitro by transcribing an AGGAG repeat with T7 RNA polymerase. When ribonuclease T1 was present, R-loop formation in cis was diminished, indicating that the transcript was separated from the template and reassociated with it. The transcript was found to form an R-loop in trans with DNA comprising the AGGAG repeat, when the DNA was supercoiled. Results of chemical modification indicated that the duplex opened at the AGGAG repeat under negative supercoiling.

1. Introduction

The DNA-dependent RNA polymerase separates product RNA from its template DNA, thereby making the RNA available to ensuing cellular processes [1, 2]. The polymerase possesses structural elements for this separation and the RNA transcript is extruded through a hole of the polymerase molecule [36]. Despite this “separator” function of the polymerase, some transcripts have been known to anomalously form an RNA:DNA hybrid, or an R-loop, with their template, and R-loops have been implicated in a number of biological processes [7]. One classical example is the colE1 replication origin, where the RNA in the R-loop serves as the primer of DNA replication [8]. A more recent example is the FLOWERING LOCUS C of Arabidopsis thaliana, where an R-loop at the promoter of the COOLER gene represses the transcription of the gene [9]. In addition to the “separator” function of the RNA polymerase, cellular processes ensuing transcription, such as splicing and RNA export, also serve to sequester RNA from the template DNA. Compromising such a function has been shown to result in hyperrecombination [7, 10].

R-loops have also been observed at G-rich repetitive sequences in immunoglobulin class-switch regions, although R-loop per se is not considered to trigger class-switching [1116]. R-loop formation at a class-switch region was first shown for the murine region [11]. Transcription of supercoiled plasmid DNA containing a 2.3 kb fragment of the region resulted in relaxation of the DNA, and the relaxation was reversed by RNase H treatment, which indicated R-loop formation. The relaxation was dependent on the direction of transcription. Radiolabeling experiments showed that the R-loop formed also when the template was linearized and that the RNA content in the R-loop was solely A and G [12]. The latter result, together with the length of the RNase A-resistant RNA bound to the DNA, indicated that the R-loop formed at the 28 tandem repeats of AGGAG in the region.

Despite the simple repetitive nature of the sequence, the exact mechanism of the R-loop formation has not been known. In this report, we present experimental results in vitro on an AGGAG repeat to address the following two questions. Firstly, it is not known whether the RNA is separated from and reassociate with the template or a gross structural change takes place in the polymerase molecule for the transcript to stay on the template. The latter possibility cannot be ruled out a priori, because the R-loop was found to form in the presence of ribonuclease (RNase) A in vitro [12]. Secondly, it is not known what structure in the DNA and/or the RNA leads to the R-loop formation: it is not known what allows the transcript to reassociate with its template or what forces the polymerase into an altered conformation in which the transcript can stay on the template.

2. Materials and Methods

2.1. Plasmid Construction

Synthetic oligodeoxynucleotides (GGAGA)5 and (TCTCC)5 were used to synthesize by the use of the nontemplate PCR method [17]. The product was cloned into the EcoRV site of pBluescript SK (Stratagene). Nucleotide sequencing showed that one clone contained GAG(AGGAG)22 in the Kpn I-to-Sac I direction and the clone was designated pSK-(AGGAG)22. A fragment containing the repeat was cut out with EcoR I and Hind III and cloned in between the EcoR I and the Hind III sites of pHC624 [18] to make pHC624-(AGGAG)22. Both plasmids were purified from Escherichia coli DH5α by Cesium-Ethidium banding [19].

2.2. Transcription

Standard transcription reaction was carried out with 6 units of T7 RNA polymerase (Roche Diagnostics) on 0.3 μg DNA (0.3 μg for each DNA, when two DNA species were present) at 37°C for 20 min in 20 μL of 50 mM NaCl, 40 mM Tris·HCl pH 8.0, 6 mM MgCl2, 2 mM spermidine, 10 mM DTT, and 0.6 mM each of ATP, CTP, GTP, and UTP [11, 12]. One μL of 0.5 M EDTA pH 8.0 was added to terminate transcription and 2.5 μL of 1 μg/μL RNase A (Sigma) was added [11, 12]. When the transcript was radiolabeled, [α-32P] ATP was also included. After incubation at 37°C for 1 h, the reaction was extracted with phenol/chloroform/isoamylalcohol. When RNase T1 was included (Figure 1(b)), transcription was carried out at an NaCl concentration of 150 mM for 20 min. One μL of 0.5 M EDTA pH 8.0 was added and the reactions were incubated at 37°C for 5 min and extracted with phenol/chloroform/isoamylalcohol. Samples were electrophoresed in 1% agarose in 50 mM Tris·borate and 1 mM EDTA, DNA was stained with ethidium bromide, and the gel was photographed on a UV transilluminator or scanned on a Typhoon analyzer (GE Healthcare Life Science). For detection of the radiolabel, the agarose gel was dried on Hybond-N+ membrane (GE Healthcare Life Science), the radioactivity was recorded on an Image Plate (Fuji Photo Film), and the Plate was scanned on a Typhoon analyzer.

2.3. Chemical Probing of DNA Structure

Supercoiled pSK-(AGGAG)22 was relaxed with vaccinia topoisomerase I [20]. Five μg of supercoiled or relaxed DNA was subjected to chemical modification in 200 μL of 50 mM sodium cacodylate pH 8.0 and 6 mM MgCl2. Oxidation with potassium permanganate [21] was carried out at 0.9 mM for 2 min at 37°C and terminated by adding 10 μL of β-mercaptoethanol. Modification with diethylpyrocarbonate [22] was started by adding 3 μL of the reagent and vortexing, the reaction was incubated at 37°C for 10 min with one more vortexing during incubation, and 10 μL of β-mercaptoethanol was added to terminate the modification. The DNAs were purified by the use of glass milk (Q-Biogene). Permanganate-oxidized DNA was digested with Xho I and dephosphorylated with calf intestine phosphatase. The reaction was extracted with phenol/chloroform/isoamylalcohol, and the DNA purified with glass milk. The DNA was end-labeled with [γ-32P] ATP and T4 polynucleotide kinase. After inactivating the kinase by heating at 65°C for 20 min, the DNA was digested with Xho I in the presence of 1 mg/mL spermidine [22]. After extraction with phenol/chloroform/isoamylalcohol, the samples were spun in S-400 HR columns (GE Healthcare Life Science) to remove unincorporated radiolabel and the small Xho I-Kpn I fragment. Diethylpyrocarbonate-treated DNA was similarly processed with Not I and Sac I. Recovered DNA was ethanol precipitated with 4 μg of carrier DNA and air dried. The DNA was dissolved in 100 μL of 1 M piperidine and 2 mM EDTA, heated at 90°C for 30 min, and vacuum-dried. The dried pellet was dissolved in 20 μL of water and vacuum-dried. This process was repeated once, and the DNA was dissolved in formamide loading solution [19]. Maxam-Gilbert sequencing reactions were carried out by the standard procedures [19]. The samples were electrophoresed in 5% Long Ranger (Lonza)-urea sequencing gel. After washing-off of urea and drying of the gel, the radioactivity was recorded on an image plate and the plate was scanned on a Typhoon analyzer.

3. Results

3.1. Formation of an R-Loop at (AGGAG)22

In order to create a simplified model of , an AGGAG repeat was synthesized by the nontemplate PCR method [17] and cloned into the EcoR V site of pBluescript SK .One clone contained GAG(AGGAG)22 at the site in the direction that transcription from the T7 promoter would produce an AGGAG-repeat transcript. The plasmid, designated pSK-(AGGAG)22, purified from E. coli and negatively supercoiled, was transcribed with T7 RNA polymerase. Agarose gel electrophoresis showed extensive relaxation of the DNA (Figure 1(a), lane 2) indicative of R-loop formation. The relaxation was orientation-dependent (compare with transcription from T3 promoter; lane 3), as reported previously [11, 12]. In addition, when the repeat was moved into pBluescript KS(+) to reverse the orientation, transcription from the T3 promoter but not from the T7 promoter resulted in relaxation of the DNA (data not shown). These results showed that the AGGAG repeat was sufficient for R-loop formation, although the AGGAG repeat in the region is preceded by a tantalizing GAGCT repeat (GenBank/EMBL/DDBJ entry D11468).

3.2. Inhibition of R-Loop Formation by RNase T1

In order to test whether the transcript was separated from and then reassociated with the template or the RNA stayed on the template, we included a ribonuclease (RNase) in the transcription mixture. If the RNA stays on the template, the RNase should not affect the R-loop formation. Although it was reported that inclusion of RNase A did not affect the formation [11, 12], RNase A is known to be specific to pyrimidines and thus would not be expected to digest the AGGAG-repeat RNA. Therefore, RNase T1, which is specific to purines, was chosen. When RNase T1 was present, the relaxation was diminished (Figure 1(b), lane 3). RNase T1 is known to cleave RNA in R-loops at lower salt concentrations. However, the R-loop was stable under the conditions used, as no change was observed when pre-formed R-loop was incubated with RNase T1 (lanes 4 and 5). Thus, the result indicated that the transcript was separated from and reassociated with the template and that the “separator” function of T7 RNA polymerase was not compromised on the CTCCT-repeat template, in agreement with results reported for other class-switch regions and a telomeric repeat [23].

3.3. R-Loop Formation in Trans and Its Dependence on DNA Supercoiling

The above result indicating reassociation suggested that the R-loop might form in trans. This possibility was explored by including in the transcription mixture a plasmid that contained the AGGAG repeat but no T7 promoter. The plasmid, pHC624-(AGGAG)22, was constructed by inserting the AGGAG repeat in a small cloning vector, pHC624 [18]. An AGGAG-repeat transcript was produced by transcribing linearized pSK-(AGGAG)22 with T7 RNA polymerase, and effects on the target pHC624-(AGGAG)22 were examined. The result is shown in Figure 2. Change in DNA topology was observed for supercoiled pHC624-(AGGAG)22 (Figure 2, lanes 4 and 5). The change was dependent on the presence of the AGGAG repeat both in the transcribed and the target plasmids (lanes 6–8). The change was reversed by treatment with RNase H (lanes 11 and 12), indicating that the change was due to the formation of an R-loop. No change in topology was observed for relaxed pHC624-(AGGAG)22 (lanes 9 and 10), indicating that the R-loop formation in trans depended on the superhelicity of the plasmid DNA.

This R-loop formation in trans was further examined by linearizing pHC624-(AGGAG)22 and radiolabeling the transcript. Transcription of linearized pSK-(AGGAG)22 generated an extra band with a slightly reduced mobility (Figure 2). Radioactivity was found to be associated with this extra band (Figure 3, lanes 4–6), indicating the mobility shift was due to the presence of the R-loop. Transcription of linearized pSK-(AGGAG)22 resulted in relaxation of supercoield pHC624-(AGGAG)22 and radioactivity was associated with the relaxed species (lane 5). By contrast, little radioactivity was associated with pHC624-(AGGAG)22 when the DNA was linearized (lane 6). These confirmed the dependence of the R-loop formation in trans on supercoiling.

3.4. Formation of an Alternative Structure at the AGGAG Repeat by DNA Supercoiling

The results obtained thus far suggested that negative supercoiling induced a structural change and the CTCCT-repeat strand became accessible to the AGGAG-repeat transcript in the altered structure. Chemical modification was employed to detect the structural change. Figure 4(a) shows the result for oxidation of T residues in the CTCCT-repeat strand by permanganate [21]. T residues were found to be oxidized more readily when the DNA was supercoiled than when relaxed. Permanganate is known to oxidize T residues (and C residues to a lesser extent) preferentially in single-stranded forms [21]. Therefore, the result indicated that supercoiling induced, a transition to an alternative structure in which the CTCCT-repeat strand was single stranded. The AGGAG-repeat strand was probed with diethylpyrocarbonate. The chemical is known to readily react with A (and less so with G) preferentially in single-stranded forms [22]. As shown in Figure 4(b), A residues in the central AGGAG units, especially the second A residues in the units, were more reactive to diethylpyrocarbonate when the DNA was supercoiled.

4. Discussion

One possible mechanism that allows reassociation would be zippering of RNA and DNA by backtracking of RNA polymerase [24]. However, in the current model of the elongation complex, backtracking dissociates the 3′ end of the transcript from the template [25]. This will limit the size of the R-loop to that of the transcription bubble, which is only several base pairs [2, 25]. A result on R-loop formation in Saccharomyces cerevisiae cells seems to suggest against backtracking as well, if the mechanism is assumed to be the same. Placing of a ribozyme sequence downstream of a PHO5 was shown to reduce R-loop formation at the PHO5 [10]. The result indicated that at least some population of the R-loop formed after the RNA polymerase had transcribed the ribozyme region, which was 52 bp. This would require the backtracking to start further downstream and to continue through the 52 bp region until reaching the PHO5 to start forming an R-loop, both of which seem very unlikely. The present result of R-loop formation in trans also indicates that backtracking is not necessary, although the result does not exclude backtracking.

The present work showed that an R-loop at the AGGAG repeat formed in trans and the formation in trans was facilitated by negative supercoiling of the target DNA. The results of chemical probing of the repeat DNA under supercoiling were consistent with a structure in which the 5′ side in the CTCCT repeat was single-stranded (Figure 4). Thus, the R-loop formation in trans observed here is most likely to be the association of the AGGAG-repeat transcript with the CTCCT-repeat template that was made accessible by the structural conversion. Formation of alternative structures had previously been studied for (AGGAG)6GA [26]. The previous result of diethylpyrocarbonate modification at low pH was consistent with formation of an intramolecular triple helix. The present result of diethylpyrocarbonate modification was different from the previous one and, therefore, the alternative structure observed here is likely to be different from the previous one, which was a canonical H-form. Another structure had also been observed at neutral pH [26]. The regions of enhanced chemical modification were different between the previous and the present results, and thus, the structures are likely to be different. The difference is likely to be due to the number of the repeats and/or the presence of Mg(II), which was absent in the previous study, as dependence of the formation of alternative forms on the size of the homopurine·homopyrimidine stretch, metal ions, and pH has been well documented [2729]. One structure consistent with our chemical modification results is depicted in Figure 4(c), which is an H-r3 isoform of the *H-form [27, 28]. Alternative fold-back points would result in stronger modification sites in the middle of the AGGAG strand.

Another possible mechanism consistent with the present findings would be one that involves opening of the duplex in the wake of the advancing RNA polymerase and ensuing hybridization of the transcript RNA to the template DNA, which has been proposed by Grabczyk et al. [30]. They have posited that negative supercoiling behind the advancing RNA polymerase induces triplex formation at GAA repeats, however, with no data to show the formation. The present study demonstrated that extensive helix opening does occur at an AGGAG repeat under negative supercoiling, indicating the repeat’s potential of an alternative conformation. There is also circumstantial evidence linking R-loop formation at E. coli rrnB to supercoiling both in vitro and in vivo [31, 32]. It is debatable, however, whether the transient supercoiling behind the polymerase is strong enough to induce the conformation transition on small purified DNA molecules in vitro, especially because R-loop has been found to form on small linearized DNA that cannot be topologically constrained [12, 33]. Duquette et al. [23] studied G-loop formation in immunoglobulin class-switch regions and a telomeric repeat and have proposed that invasion of RNA into the duplex DNA initiates the formation. This general idea of RNA invasion into duplex is supported by the finding that a nick in the nontemplate strand can serve as an initiation site for R-loop at immunoglobulin Sγ3 repeats in vitro: fraying of the nontemplate strand at the nick should increase the accessibility of the template strand to the transcript [33]. Finally, we would like to raise the possibility that the transient duplex separation by the RNA polymerase could be more persistent for some templates, which may involve interaction between DNA single strands and the RNA polymerase, and also the possibility that the nontemplate strand may contribute to the stability of the R-loop by the base interactions involved in triple helix formation [27, 28, 34, 35].

5. Conclusions

In vitro experiments showed that R-loop formed in trans at an AGGAG repeat when the DNA comprising the repeat was negatively supercoiled. Chemical modifications indicated that the duplex at the repeat opened under negative supercoiling. These results were the first to connect structural conversion, duplex opening, and R-loop formation for the same sequence, supporting the importance of the potential of duplex opening in R-loop formation.

Conflict of Interests

The authors declare that they do not have any conflict of interests with the content of the paper.

Acknowledgment

This work was supported in part by the Strategic Research Foundation Grant-aided Project for Private Universities (S1201003) from the Ministry of Education, Culture, Sports, Science, and Technology, Japan.