Design and synthesis of circular DNA templates encoding miRNAs

We designed DNA circles to encode shortened versions of two human pri-miRNAs, miR-19a and miR-122. We based our design on the endogenous human miRNA genomic template-strand sequences, and included DNA encoding ∼3 helical turns of the pri-miRNA stem-loop sequence (Fig. 1B), the minimum stem length required for Drosha/Microprocessor processing [14], [15]. In two of the COLIGOs we minimized the length of the sequences flanking the stem-loops to 4 nt (COLIGO 19a) and 6 nt (COLIGO 122) per repeat. In two other COLIGOS, 19am and 122m, we included more of the natural flanking sequences (taken in both cases from the miR19a genomic template-strand sequence), for a total of 22 and 18 flanking nt, respectively. Additional flanking nt are believed to enhance Drosha processing [15], [16]. However, in a rolling circle transcript all but the terminal stem-loop repeats would be flanked by varying lengths of RNA, so it is not yet clear how much flanking RNA should be encoded by a given COLIGO. The complete monomer transcripts for the four templates - two for each miRNA - used in this study are shown in Fig. 1B. Although the pri-miRNA stem-loop structure forms the basis of our COLIGO design, we note that promoterless transcription may initiate at multiple sites in the DNA circle, and that alternate folding patterns are possible for multimeric transcripts [12].

Some miRNAs have conserved loops that aid in their processing [17], but specific loop sequences are likely not critical to the biogenesis of most miRNAs [15], [18]. Loop sequences at the DNA level in the COLIGO, however, may influence transcription efficiency, as shown for E. coli RNAP RCT [19]. We tested this variable, in combination with the longer flanking sequences, by swapping the native miRNA loops in pre-miR-19a and miR-122 for an unrelated RNA hairpin loop (from HIV TAR RNA), resulting in pre-miR19am and pre-miR122m (monomer transcripts are shown in Fig. 1B).

To choose the site where the linear oligonucleotide is closed to form the COLIGO we predicted the folding of the COLIGO using mFold [20] and selected the region having the least secondary structure, which in the linear form might prevent juxtaposition of the ends. This sequence typically coincided with the flanking nucleotides at the base of the stem (of the transcripts). The DNA circles encoding the transcripts shown in Fig. 1B were therefore ligated between the 5′ and 3′ deoxynucleotides corresponding to the 3′ and 5′ terminal ribonucleotides, respectively, depicted in the monomer transcripts (Fig. 1B).

Rolling circle transcription templates have been circularized by three methods: chemical ligation within a DNA triple-helix [8]; T4 DNA ligase closure mediated by a splint oligonucleotide [21]; and T4 DNA ligase closure within a nicked DNA dumbbell structure, as for the shRNA-encoding templates [12]. Among these methods, splinted DNA ligase closure should be the most general approach for natural miRNA cDNAs, since their stem sequences include multiple mismatches, which might interfere with DNA ligase closure. However, the requirement for a splint to bring the 5′ and 3′ ends together increases the cost of DNA synthesis and requires an additional step to remove the splints after circularization. In preliminary experiments we also found that splint-mediated ligation resulted in unsatisfactory circularization yields and varying amounts of linear multimers.

We obtained better results using a newly discovered thermostable RNA ligase (Rnl) from bacteriophage TS2126 (Fig. 2A). TS2126 Rnl has been reported to efficiently circularize ss DNA without a splint [22]. In a typical COLIGO preparation using this enzyme, the entire 5′ phosphorylated linear template was made in one synthesis, gel purified by DPAGE and circularized with TS2126 Rnl. Most residual linear form was then removed by Exonuclease I treatment. For example, in the case of the 117 nt template encoding the miR19am transcript (Fig. 1B), 5.0 nmol from the desalted DNA synthesis was gel purified to yield 1.2 nmol of the full-length form (23% full-length recovered). Following cyclization and Exonuclease I treatment, 0.87 nmol of the COLIGO was recovered (73% cyclization based on purified linear precursor). The gel profile of the DNA at the various stages is shown in Fig. 2B. A final prep-scale gel purification generally does not improve the circular-to-linear ratio, and leads to a large loss of COLIGO. The circular topology of all COLIGOs used in this study was verified by S1 nuclease nicking (e.g. Fig. 2C), which for COLIGOs leads initially to the linear form with no intermediate gel bands [21].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. Circularization of DNA templates (COLIGOs) for Rolling Circle Transcription.

A. Synthetic 5′ phosphorylated linear DNA sequences were circularized using the thermostable TS2126 RNA ligase. B. Denaturing polyacrylamide gel electrophoresis (DPAGE) at four stages during miR-19am DNA circle synthesis. Lane 1, crude DNA IDT Ultramer synthesis of COLIGO 19am. Lane 2, after preparative DPAGE. Lane 3, crude circularization product. Lane 4, DNA circle template following Exonuclease I clean-up. Visualization using Stains-All. C. Verification of circular topology. Nicking of circular templates by S1 nuclease leads first to linear forms, which are then further digested to successively smaller linear forms.

https://doi.org/10.1371/journal.pone.0016925.g002

In vitro transcription by bacteriophage and bacterial RNA polymerases

While COLIGOs encoding shRNAs [12] and ribozymes [9], [10], [11] have been shown to undergo RCT in vitro by bacterial and bacteriophage RNAPs, COLIGOs encoding native pri-miRNAs have not been investigated. To determine whether native miRNA cDNA secondary structure might pose a general impediment to RCT, we exposed the four COLIGOs encoding miR-19a and miR-122 to T7 and E. coli RNAPs in vitro (Fig. 3A). Transcripts were visualized by [α-³²P]-UTP incorporation. All four COLIGOs generated the large transcripts characteristic of RCT. For each template sequence, the circular topology was required to produce transcripts significantly longer than the COLIGOs' circumference, and withholding one nucleotide triphosphate (ATP, Lane C-) led as expected to the loss of all but very short aborted transcripts. For all templates, more RNA was produced using E. coli RNAP compared to T7 RNAP under similar conditions (see Materials and Methods). Denaturing formaldehyde agarose gel electrophoresis showed that the maximum length of the transcripts typically reached 5 kilobases (5 kb, Fig. 3B). Thus, all four DNA circles served as RCT templates despite the extensive secondary structure predicted for COLIGOs encoding pri-miRNA stem loop structures.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 3. In vitro transcription of miRNA-encoding COLIGOs by E. coli and bacteriophage T7 RNA polymerases.

A. Linear (L) and circular (C) templates 122, 122m, 19a, 19am, encoding either pre-miR-122 or pre-miR-19a, were transcribed in the presence of all NTPs or all except ATP (C-). 9% DPAGE. The Phosphorimager sensitivity setting was increased 5-fold for the weaker T7 RCT reactions. B. Denaturing 1% agarose gel showing the high molecular weight size range of the transcripts. No adjustment in the Phosphorimager sensitivity setting was made.

https://doi.org/10.1371/journal.pone.0016925.g003

The relative amount of RNA produced in these experiments varied with the DNA sequence outside of the miRNA stem sequence. The template encoding the pre-miR-19a with the natural loop and short flanking sequence was noticeably poor (Fig. 3A), but it became a more efficient template when its wild-type (wt) loop was replaced with the TAR RNA loop and the flanking sequences were lengthened (compare 19a and 19am). In contrast, the same loop and flanking sequences, when placed on the miR122 stem, decreased transcription efficiency slightly (compare 122 and 122m). These comparisons demonstrate that the non-essential sequence regions of a pri-miRNA-encoding RCT template can influence the extent to which a COLIGO is transcribed.

Drosha/Microprocessor processing of rolling circle transcripts

Primary miRNA transcripts are first processed in their natural stem-loop context by Microprocessor, a complex that in mammalian cells includes Drosha and DGCR8 [23], [24]. We tested whether pre-miRNA hairpins could be processed from a multimeric rolling circle transcript. Treatment of RCT transcripts (made in vitro with uniform ³²P-labeling using E. coli RNAP) with HEK293T whole cell extract (WCE) led to the release of low levels of the pre-miRNA hairpin (71 nt predicted) and the intervening flanking region (46 nt predicted) (Fig. 4A, lane 2). In order to test whether Drosha was responsible for this processing, we immunoprecipitated (IP'd) FLAG-tagged Drosha from transfected HEK293T cells [16]. The activity of IP'd Drosha (and co-IP'd proteins) was verified on an in vitro transcribed human miRNA cluster whose processing is predicted to produce three pre-miRNAs of 57-nt (pre-miR-23a), 59-nt (pre-miR-24-2) and 62-nt (pre-miR-27a) [13] (Fig. 4B, lane 10, ∼60 nt RNA was produced). When the miR-19a RCT transcript was treated with IP'd Drosha, the same processed RNAs produced by WCE were released from the transcript, but in higher amounts (Fig. 4A, lane 4). The processed transcripts accumulated over the standard in vitro Drosha processing time period of 90 minutes (Fig. 4B, lanes 5, 6, 7), making it unlikely that they resulted from random degradation. These results demonstrated that pre-miRNA hairpins can be accurately processed by human Drosha from RCT transcripts. If RCT were to take place in the nucleus of human cells, then some of the RNA should be capable of entering the natural processing pathway.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 4. In vitro Drosha processing of miRNA Rolling Circle Transcripts (RCT).

A. Deproteinized miR-19am RCTs made in vitro using E. coli RNA polymerase were incubated with HEK293T whole cell extract (WCE, Lane 2); with the Flag-Drosha complex immunoprecipitated from Flag-Drosha-expressing HEK293T WCE (Lane 4); or with Flag immunoprecipitate from WCE made from untransfected HEK293T cells (Mock, Lane 3). 90 min. processing reactions. B. Drosha processing reactions at various processing times using DPAGE gel purified miR-19am RCT transcripts (lanes 5–9) or the miR-23a∼27a∼24-2 cluster transcript as a positive control (lanes 10–12) at the standard 90 min. processing reaction time.

https://doi.org/10.1371/journal.pone.0016925.g004

Materials, reagents and general procedures

Synthetic DNA was made by Integrated DNA Technologies (Coralville, IA) and chemically 5′ phosphorylated. Preparative and analytical denaturing polyacrylamide gel electrophoresis (DPAGE) and denaturing agarose electrophoresis were done according to standard procedures [28]. DNA was located on preparative gels using UV shadowing over fluorescent silica gel plates (EMD Chemicals 5715-7). DNA was recovered by electroelution in 1xTBE within a sealed dialysis membrane tube with a molecular weight cut off of 1000 (Spectrum Laboratories). Eluted DNA was phenol-chloroform extracted and ethanol precipitated with sodium acetate. In some cases, the DNA gel slice was eluted in 0.3 M sodium-acetate at 37°C overnight and ethanol precipitated. The 1% agarose gel was prepared using Ambion's 10X denaturing gel buffer, blotted overnight onto positively charged nitrocellulose (Ambion), baked for 20 min at 55°C, UV crosslinked and exposed to a PhosphorImager® screen (Molecular Dynamics). E. coli RNAP was purchased from USB. T7 RNAP and DNA ligase were purchased from New England Biolabs. RNA markers were purchased from Sigma (R4142) or Ambion (Decade marker, AM7778) and prepared according to the manufacturer's instructions. For the agarose gel (Fig. 3B), sizes were determined by running 8 µg of total cellular RNA extracted from HEK293T cells on the same gel and staining for ribosomal RNA (18S = 1.9 kb and 28S = 5 kb) with methylene blue after blotting.

Synthesis of linear DNA templates

Linear DNA templates encoding shortened forms of pri-miR-19a, -miR-122 and -miR19am (e.g. Fig. 2B) were synthesized by IDT as single 5′ phosphorylated Ultramer sequences. The linear templates for miR122m and miR19am were also made from two half-length oligonucleotides (e.g. 19am-1 and 19am-2) according to the standard splint mediated T4 DNA ligase procedure [21]. The following sequences were used, including splint oligonucleotides where applicable:

19am (117 nt; when the linear form was synthesized as one sequence, it began as in 19am-1 and continued through 19am-2): 19am-1 (63 nt): 5p-GAAGGAAATAGCAGGCCACCATCAGTTTTGCATAGATTTGCACAACGGCAGAGAGCTCCCAGG; 19am-2 (54 nt): 5′p-CTCAGATCTGCCTGCAACTATGCAAAACTAACAGAGGACTGCAAACAAAAACTA, splint 19am (30 nt): 5′-CAGGCAGATCTGAGCCTGGGAGCTCTCTGC; 122m (124 nt): 122m-1 (65 nt): 5′p- AGGAAATAGCCTAGCAGTAGCTATTTAGTGTGATAATGGCGTTTGATAGGGCAGAGAGCTCCCAG, 122m-2 (59 nt): 5′p-GCTCAGATCTGCCCACAAACACCATTGTCACACTCCACAGCTCTGCTAAGGAAACAAAA, splint 122m (31 nt): 5′-TGGGCAGATCTGAGCC-TGGGAGCTCTCTGCC; 19a (86 nt): 5′p-TAGCAGGCCACCATCAGTTTTGCATAGATTTGCACAACTACATTCTTCTTGTAGTGCAACTATGCAAAACTAACAGAGGACTGCAA; 122 (89 nt) 5′p-TTGCCTAGCAGTAGCTATTTAGTGTGATAATGGCGTTTGATAGTTTAGACACAAACACCATTGTCACACTCCACAGCTCTGCTAAGGAA.

TS2126 RNA Ligase I

A synthetic gene (Top Gene Technologies Inc., Canada) encoding the first 393 amino acids of TS2126 RNA ligase (Rnl) I [22] followed by a C-terminal hexahistidine tag was inserted into the EcoRI/SmaI sites of pBluescriptSK(+) plasmid (Stratagene). The synthetic DNA contained codons optimized for expression in E. coli. The entire coding sequence of this plasmid, pTS2126H, was verified by DNA sequencing. E. coli Bl21-CodonPlus(DE3)-RIL (Stratagene) was transformed with pTS2126H and grown in 2 liters of culture. At OD₅₉₅ ∼0.5 the culture was induced to a final concentration of 1 mM IPTG and grown for another 3 hrs at 37°C. The cells were pelleted at 3000 g for 15 min at 4°C, resuspended in 30 ml native lysis buffer (50 mM NaH₂PO₄, 300 mM NaCl, 10 mM imidazole, pH 8 with NaOH), and disrupted using a French press. After centrifugation to remove debris, the lysate was incubated with 1 ml Ni-NTA agarose (Qiagen) with agitation for 2 hrs at 4°C. The beads were washed three times with 5 ml native lysis buffer each time, once with elution buffer without imidazole (20 mM Tris-HCl pH 8, 100 mM KCl, 2 mM DTT) and finally eluted with 500 µl portions of elution buffer containing 250 mM imidazole. The pooled fractions containing the protein were dialyzed (3 times, 2 hrs each) against 10 mM Tris-HCl (pH 8), 50 mM KCl, 0.1 mM EDTA and 1 mM DTT, and stored at 4° C (short term) or in 50% glycerol and –80°C (long term).

Circularization of oligonucleotides

Reactions contained a ratio of 0.015 nmol linear 5′ phosphorylated DNA (0.75 µM final conc.) to 1.5 µg TS2126 Rnl and the following components (final concentrations): 50 µM ATP, 2.5 mM MnCl₂, 50 mM MOPS (pH 7.5), 10 mM KCl, 5 mM MgCl₂ and 1 mM DTT. These conditions were used for reactions ranging from 015 to 1.5 nmol. Circularizations were incubated for 2 hrs at 60°C, followed by phenol/chloroform extraction and ethanol precipitation. In preliminary syntheses, the circularized products and unreacted linear templates were separated on a 1.5 mm DPAGE and visualized by UV shadowing, excised from the gel and electroeluted. In cases where the COLIGO was still contaminated by >5% of the linear oligonucleotide after elution (as determined by gel staining), an Exonuclease I (NEB) digest was done. For most COLIGOs used in this study, we used the more recently adopted procedure shown in Fig. 2B. The crude desalted oligonucleotide was gel-purified, circularized as described above, phenol/chloroform extracted and precipitated. The COLIGO was then treated with 10 u of Exonuclease I per 0.1 nmol DNA in a total volume of 50 µl according to the manufacturer's instructions, phenol/cholorform extracted, precipitated and used without further gel-purification after analytical scale DPAGE showed no multimeric circles and less than 5% linear form by 0.05% Stains-All (Acros) staining.

S1-nuclease assay

To verify the circular topology of the COLIGOs (as in Fig. 2C), the following procedure was used: a 10 µl reaction containing 1 µg of linear or circular oligonucleotide and 0, 0.5 or 1 u of S1 nuclease (USB) in 50 mM sodium acetate, pH 4.6, 1 mM zinc acetate, 250 mM NaCl and 0.5 µg BSA was incubated for 10 min at 37°C, extracted with phenol/chloroform, ethanol precipitated, separated by 10 or 12% DPAGE and visualized with Stains-All.

In vitro transcription

The following conditions were used for in vitro transcription with bacteriophage T7 and E. coli RNAPs: final COLIGO concentration 1 µM, 1unit/µl RNAse inhibitor (Promega), 0.5 mM each ATP, CTP and GTP, 0.05 mM UTP in Fig. 3A and 3B; or 1 mM each NTP in Fig. 4A and 4B; ∼2 µCi [α-³²P]-UTP; and (for T7 RNAP: 40 mM Tris-HCl pH 7.9, 6 mM MgCl₂, 10 mM DTT, 2 mM spermidine); (for E.coli: 40 mM Tris-HCl, pH 8.0, 10 mM MgCl₂, 5 mM DTT, 50 mM KCl, 50 µg/ml BSA). Each 10 µl RCT reaction (Fig. 3) contained the following amount of RNAP as defined by the vendor: 40 units T7, 1 unit E.coli. Reactions were incubated for 1 hr at 37°C, after which time the RNA was extracted with 150 µl TriReagent (Ambion), isopropanol precipitated according to the manufacturer's instructions with 10 µg glycogen added. For all in vitro RNAP reactions, radiolabeled transcripts were separated on a 9% denaturing polyacrylamide gel, dried and exposed to a PhosphorImager® screen. The T7 RNAP reactions shown in Fig. 3A, which were done on the same scale and contained the same amount of [α-³²P]-UTP as the E. coli reactions in the same figure, are shown at 5 times greater sensitivity to adjust for the lower level of incorporation with the T7 enzyme. For comparison, no adjustment was made in Fig. 3B. The transcripts in Fig. 4A were, after transcription at the relatively high concentration of 1 mM cold NTP (see above), deemed sufficiently free of shorter transcripts to be used for processing without gel purification. In Fig. 4B we gel-purified the transcripts by 6% DPAGE and then eluted the RNA for 3 hrs at 37°C in 350 µl 0.5 M ammonium acetate, 1 mM EDTA, 0.2% SDS, extracted with an equal volume of phenol chloroform-isoamyl alcohol and precipitated with 0.3 volumes of 3 M sodium acetate, 2.5 volumes ethanol and 10 µg glycogen. Eluting the RNA from the gel led to some degradation, which produced a wider distribution of RCT transcript sizes in the processing input RNA in Fig. 4B. Some non-specific degradation also occurred during the processing reactions shown in Fig. 4B, since not all of the missing input is accounted for in the processed fragments.

Drosha immunoprecipitation and in vitro RNA processing

The pCK-Drosha-FLAG plasmid (kindly provided by V. N. Kim) was expressed and purified as previously described [16]. Briefly, 8 µg of plasmid were transfected per 6 cm dish of HEK293T cells using 20 µl Lipofectamine 2000 (Invitrogen). After 44 hrs, the cells were lysed (900 µl Sigma FLAG® kit lysis buffer/6 cm dish) and Flag-IP was carried out using 40 µl of a 50% anti-FLAG bead slurry (Sigma FLAG® Tagged Protein Immunoprecipitation Kit) for at least 4 hrs at 4°C. The beads were washed 4 times with 1X Wash Buffer and once in processing buffer (20 mM Hepes-KOH, pH 7.9, 100 mM KCl, 0.2 mM EDTA, 0.5 mM DTT, 0.2 mM PMSF, 5% glycerol). A parallel mock-IP was carried out using Lipofectamine without the Drosha-encoding plasmid. The activity of the immunoprecipitated protein was verified using the miRNA-23a∼27a∼24-2 cluster transcribed in vitro from plasmid pGEM-T-easy_pri-miR-23, 27, 24-2(+) (also provided by V. N. Kim). For HEK293T WCE as the Drosha source, cells were lysed as described above; the crude lysate was agitated at 4°C for 30 min, centrifuged for 30 min at 13000 rpm and the supernatant collected and adjusted to 20% glycerol. For in vitro processing of rolling circle transcripts from COLIGO 19am, 1/10 of an in vitro transcription reaction using E. coli RNA polymerase (described above) or ¼ of the eluted pri-miR-23a∼27a∼24-2(+) transcript was used in a total volume of 30 µl containing 15 µl beads from Drosha Flag-IP (or mock-IP, or processing buffer including 6.4 mM final MgCl₂ and 0.1 u/µl RNase inhibitor). The processing reaction was carried out for 90 min at 37°C (Fig. 4A), or for the times indicated in Fig. 4B. RNA was extracted with TriReagent (Ambion), isopropanol precipitated and separated on a 10% denaturing polyacrylamide gel, dried and exposed to a PhosphorImager® screen.