Branched nucleic acid strands play an integral role in biological RNA splicing. During the first step of intron splicing by the spliceosome (Burge et al. 1999) and also by self-splicing RNAs such as group II introns (Saldanha et al. 1993; Michel and Ferat 1995; Jacquier 1996; Fedorova et al. 2002), a conserved branch-site adenosine within the intron uses its 2′-hydroxyl group to attack the 5′-splice site phosphodiester linkage. The 5′-exon is expelled, forming a lariat intermediate that has the 3′-exon attached (lariat/3′-exon intermediate). During the second step of splicing, the 5′-exon attacks the 3′-splice site, releasing the ligated exons along with the lariat intron. This lariat is degraded via initial branch-site hydrolysis catalyzed by de-branching enzyme, e.g., Dbr1p in yeast (Nam et al. 1994; Ooi et al. 2001). Lariat RNAs are common to all of these biochemical transformations, and studies of splicing pathways will be facilitated by synthetic access to lariat RNAs of wide sequence composition.

Synthesis of lariat RNAs using chemical or biochemical methods is a substantial experimental challenge due to their special topology, which features a closed loop. Solid-phase synthesis methods to prepare even the relatively simple 2′,5′-branched RNA are tedious (Damha and Ogilvie 1988; Damha and Zabarylo 1989; Damha et al. 1992; Ganeshan et al. 1995; Reese and Song 1999), and the closed loop of a lariat provides an added measure of complexity (Carriero and Damha 2003). Group II intron ribozymes can synthesize lariats, but numerous sequence requirements for the RNA substrates limit the applicability of this approach. We recently reported several deoxyribozymes that synthesize 2′,5′-branched RNAs (Wang and Silverman 2003a, b, Wang and Silverman 2005a; Coppins and Silverman 2004 Coppins and Silverman 2005; Pratico et al. 2005). Although some of these branch-forming deoxyribozymes can also synthesize lariat RNAs in one step from linear RNA precursors (Fig. 1A ), in many cases the yield of lariat is low. Presumably this is because the deoxyribozyme and substrate interfere with each other, although the structural basis of such interference is not yet known. In this article, we describe an efficient two-step strategy that sequentially applies a deoxyribozyme and then a protein enzyme (T4 RNA ligase) to construct lariat RNAs (Fig. 1B ). This general approach allows great flexibility in lariat RNA synthesis and also permits the controlled preparation of unnatural lariat isomers that have an alternative connectivity among their three nucleic acid arms.

FIGURE 1. The one-step and two-step strategies for synthesis of lariat RNA. (A) One-step lariat RNA synthesis using a deoxyribozyme to mediate attack of an internal 2′-hydroxyl group on the 5′-triphosphate of the same RNA substrate strand. (B) Two-step (more ...)

FIGURE 2. One-step and two-step syntheses of YBL059W lariat RNA. (A) Attempting one-step synthesis according to the strategy of Figure 1A using the 7S11 deoxyribozyme, which provides an undetectable (<1%) amount of lariat RNA (12% PAGE). The lariat (more ...)

FIGURE 3. Formation of natural and unnatural lariat RNA isomers in the T4 RNA ligase loop-closure reaction, and the blocking and capping approaches to control which isomer is formed. On each structure is marked the cleavage site for debranching enzyme Dbr1p, which (more ...)

In the case of the YBL059W lariat, only one major product band (>97%) was observed during lariat formation using T4 RNA ligase (Fig. 2 ). However, if the natural and unnatural lariat isomers co-migrate on PAGE, the observed product would be a mixture of the two isomers. This possibility was assessed by treating the PAGE-purified lariat product with yeast debranching enzyme Dbr1p (Nam et al. 1994; Ooi et al. 2001), which is expected to debranch the natural lariat isomer to a 69-nt linear RNA but cleave the unnatural lariat isomer into a 52-nt circular RNA and a separate 17-nt linear fragment. Treatment of the YBL059W lariat product with Dbr1p showed clearly that both lariat isomers were indeed present (Fig. 4A ). When the lariat was synthesized with no stem at all in the T4 RNA ligase substrate, Dbr1p treatment revealed that the natural and unnatural lariat isomers were formed in ~1:1 ratio, as expected for essentially no selectivity between the 3′-ends (which are of similar length). In contrast, when a stem was present to hold the proper 3′-end near the 5′-end, the natural lariat isomer was substantially favored, but ~7% of the product was still the unnatural lariat isomer. For some applications and also from the viewpoint of biochemical purity, we sought to control lariat formation more precisely.

FIGURE 4. Characterization of the natural and unnatural YBL059W lariat isomer ratio in lariat RNA synthesized by the variations of the two-step strategy shown in Figure 3 . Each pair of lanes shows lariat RNA without or with debranching enzyme treatment (more ...)

The two approaches developed for this purpose, termed “blocking” and “capping,” are shown in Figure 3 . In the blocking approach, an excess is added of a short DNA oligonucleotide complementary to the RNA 3′-end that would otherwise react to form the unnatural lariat isomer. Because T4 RNA ligase needs a single-stranded 3′ terminus on its RNA substrate (Silverman 2004), the blocking oligo should suppress reactivity of the blocked RNA 3′-end. For lariat synthesis without a stem to direct the reactivity, inclusion of the blocking DNA oligonucleotide increased the ratio of natural to unnatural lariat isomer from ~1:1 to 92:8 (Fig. 4A ). For lariat synthesis when the stem element was present, the percentage of unnatural lariat isomer fell from 7% to only 0.4%, which was barely above the detection limit. In the alternative capping approach, the RNA substrate was synthesized with a 2′,3′-cyclic phosphate cap on its 3′-end, such that the 3′-hydroxyl necessary for loop closure was absent. Regardless of whether the stem was absent or present, the capping approach was entirely effective at preventing unnatural lariat isomer formation, and no unnatural lariat was detected (<0.2%).

Both the blocking and capping approaches could instead be applied to synthesize the unnatural lariat isomer. For these experiments we used the RNA substrate with the stem that favors natural lariat isomer formation, such that unnatural lariat formation must overcome the inherent 93:7 bias towards creating the natural isomer (as shown in Fig. 4A ). Because of this bias, it was unsurprising that the blocking approach is largely but not completely effective at inverting the selectivity to favor the unnatural lariat isomer, particularly when a large excess of blocking oligonucleotide was added (Fig. 4B ). The capping approach is again fully effective at controlling which lariat isomer is formed (Fig. 4B ).

FIGURE 5. Syntheses of ai5γ group II intron lariat RNA. The two-step RNA substrates include a stem to favor natural lariat formation. For the one-step synthesis and for the first step of the two-step synthesis, t = 0, 0.5, and 1.5 h using the 7S11 deoxyribozyme. (more ...)

To verify the structural integrity of the natural ai5γ lariat RNA, we examined its splicing reactions (Nolte et al. 1998). In a direct comparison with the ai5γ lariat prepared independently by self-splicing of the intron, the natural isomer of the synthetic lariat was equally competent in the reverse of the first step of splicing (data not shown). The 3′-exon was not included on the lariats, or the second step of splicing would have occurred instead of the reverse of the first step. Similarly, the synthetic lariat/3′-exon intermediate was able to proceed forward through the second step of splicing when the 5′-exon was added (data not shown).

To enable the most efficient application of the two-step strategy, a ligation site for the loop-closure step can generally be found with a stem-loop structure as shown in the Figure 1B inset. Although T4 RNA ligase is known to tolerate a relatively broad range of ligation substrate sequences, certain nucleotide preferences have been identified. For example, uridine in the acceptor substrate, which reacts directly with the 5′-monophosphorylated nucleotide, is detrimental to ligation activity (Ohtsuka et al. 1977; England and Uhlenbeck 1978; Romaniuk et al. 1982). Whenever possible, the ligation site for the T4 RNA ligase loop-closure step should be chosen with such preferences in mind. For practical application of this strategy, a His₆-tagged version of T4 RNA ligase has been described (Wang and Unrau 2002), in addition to commercial availability of the enzyme.

Most synthetic approaches to lariat RNA are motivated by the natural occurrence of lariats in biological splicing. Indeed, natural biochemical pathways constitute one preparative route to lariat RNAs, but these reactions are limited by sequence requirements for their substrates and also by inherent reactivity of the products. For example, natural splicing reactions cannot generally be used to synthesize a lariat/3′-exon intermediate that is competent for the second step of splicing, because the intermediate will proceed rapidly through exon ligation by nucleophilic attack of the just-released 5′-exon. Therefore, lariat mutants that are competent for the second step cannot usually be synthesized by this route in more than trace amounts. In contrast, our synthetic strategy that uses a deoxyribozyme to synthesize the branched core of the lariat/3′-exon intermediate in the absence of the 5′-exon avoids this concern entirely. Closure of the loop with T4 RNA ligase in the absence of the 5′-exon allows the otherwise reactive lariat/3′-exon RNA intermediate to be prepared efficiently.

The capping strategy completely suppresses formation of the undesired lariat isomer, but for this strategy the 2′,3′-cyclic phosphate must be included in the substrate. Although this adds a synthetic constraint, such cyclic phosphate termini are readily introduced. In particular, cleavage of a 3′-extended RNA substrate precursor with a ribozyme (either in cis or in trans; Grosshans and Cech 1991; Ferré-D’Amaré and Doudna 1996), with a deoxyribozyme (Santoro and Joyce 1997; Pyle et al. 2000; Cruz et al. 2004), or by other means (Lapham and Crothers 1996) is a useful way to introduce a 2′,3′-cyclic phosphate. Cyclic phosphate termini are also readily removed after the reaction if a free 2′,3′-diol is desired (Schürer et al. 2002). As an alternative to capping with a 2′,3′-cyclic phosphate, the 3′-terminal ribose moiety of the substrate could be oxidized with sodium periodate (Proudnikov and Mirzabekov 1996). Such periodate oxidation would be an acceptable capping approach if the resulting 3′-terminal aldehyde functional groups do not interfere with subsequent applications of the RNA.

For synthesis of any particular lariat according to the two-step strategy, the natural and the unnatural isomers may migrate differently on PAGE and may thus be separable during routine purification. The details of PAGE migration will depend sensitively on the details of the substrates. For example, when a 13-nt 3′-exon was included with the YBL059W substrate, lariat formation provided two readily separable PAGE bands that were assigned to the natural and unnatural lariat isomers (data not shown), instead of overlapping bands for the two isomers that were observed when the 3′-exon was absent (Fig. 4 ). In this case, the increased selectivity achieved via the blocking approach, combined with the ability to separate the isomers by PAGE, means that either the natural or the unnatural lariat isomer can be obtained without contamination by the undesired isomer. A similar situation exists for the ai5γ lariat (Fig. 5 ). In this case, the data also emphasize the importance of the stem element to direct proper loop closure, because formation of the unnatural lariat is favored unless the stem is present to guide formation of the natural lariat. To explain this inherent bias in favor of the unnatural lariat, we note that the two possible loop-closure reactions form substantially different sizes of ring (776 nt vs. 98 nt for natural vs. unnatural lariat), and it is well known that larger rings are more difficult to create by macrocyclization.

An alternative approach to lariat RNA synthesis was recently reported, in which chemical ligation with cyanogen bromide is used to close the loop of a branched nucleic acid (Carriero and Damha 2003). It is likely that our blocking and capping strategies would be applicable to loop closure using cyanogen bromide instead of T4 RNA ligase. However, it does not appear that the chemical reagent offers any particular advantages over the protein enzyme for this purpose. Indeed, the chemical reagent does not provide any control over site-selectivity during loop closure, and mixtures of 2′–5′ and 3′–5′ linkages will likely be formed. In contrast, T4 RNA ligase forms only native 3′–5′ linkages.

One-step and two-step lariat RNA synthesis for YBL059W (Fig. 2 )

For two-step YBL059W lariat RNA synthesis (Fig. 2B ), the left-hand (L) RNA substrate was prepared by taking the 69-nt transcript whose sequence is given above and cleaving with a 10–23 deoxyribozyme (5′-GTATGAGGCCGAAGGGCTAGCTACAACGATATTGCCTATGCAT-3′) between nucleotides 17 and 18, providing L as the 3′-fragment. The 52-nt product was purified by PAGE and 5′-³²P-radiolabeled using γ-³²P-ATP and T4 polynucleotide kinase (T4 PNK). Another portion of L was 5′-monophosphorylated using unradiolabeled ATP and T4 PNK. For deoxyribozyme-catalyzed branched RNA formation (step 1), the L substrate (5 pmol, of which 0.5 pmol was 5′-³²P-radiolabeled), deoxyribozyme (15 pmol), and R substrate (30 pmol; nucleotides 1–17 of the sequence given above prepared as an independent unradiolabeled transcript) were annealed in 5.9 μL of 5 mM HEPES (pH 7.5), 15 mM NaCl, and 0.1 mM EDTA by heating at 95°C for 3 min and cooling on ice for 5 min. The sample volume was raised to 10 μL with final concentrations of 50 mM EPPS (pH 8.5), 150 mM NaCl, 2 mM KCl, and 40 mM MgCl₂. The sample was incubated at 37°C; aliquots were quenched and assayed as described above.

For loop closure catalyzed by T4 RNA ligase (step 2), the branched product from step 1 was prepared by a modification of the above procedure. The L substrate (50 pmol, of which 10 pmol was 5′-³²P-radiolabeled), deoxyribozyme (100 pmol), and R substrate (200 pmol) were annealed in 29.5 μL of 5 mM HEPES (pH 7.5), 15 mM NaCl, and 0.1 mM EDTA by heating at 95°C for 3 min and cooling on ice for 5 min. The sample volume was raised to 50 μL with final concentrations of 50 mM EPPS (pH 8.5), 150 mM NaCl, 2 mM KCl, and 40 mM MgCl₂. The sample was incubated at 37°C for 1.5 h and quenched with 60 μL of stop solution. The branched RNA product was purified by 12% PAGE. Then, loop closure catalyzed by T4 RNA ligase was performed as follows: Branched RNA (5 pmol) was annealed in 7.5 μL of 5 mM HEPES (pH 7.5), 15 mM NaCl, and 0.1 mM EDTA by heating at 95°C for 3 min and cooling on ice for 5 min. Portions of 5× ligation buffer and T4 RNA ligase (Fermentas) were added, bringing the final conditions to 50 mM HEPES (pH 7.5), 10 mM MgCl₂, 10 mM DTT, and 50μM ATP with 1 U/μL of T4 RNA ligase in 10 μL total volume. The sample was incubated at 37°C. Aliquots were quenched and assayed as described above. To prepare the lariat RNA samples for the assays of Figure 4 , the entire sample was quenched after 15 min with 20 μL of stop solution, and the lariat RNA product was purified by 12% PAGE.

Controlling and assaying natural and unnatural lariat RNA formation for YBL059W (Fig. 4 )

For the debranching enzyme assays, 0.2–0.5 pmol of PAGE-purified lariat RNA was incubated with 1.5 pmol (75 ng) Dbr1p and 16 U of RNase inhibitor (Boehringer Mannheim) in 10 μL of 20 mM HEPES (pH 7.5), 0.5 mM MgCl₂, 125 mM KCl, and 1 mM DTT for 15 min at 30°C. Aliquots of 4 μL were quenched onto 5 μL of stop solution, followed by PAGE and exposure to a PhosphorImager screen. The linear RNA standards were prepared by 5′-³²P-radiolabeling of appropriate substrates. The circular RNA standard was prepared by circularization of the linear standard with T4 RNA ligase; its structure was verified by partial alkaline hydrolysis, which gave a linear product that co-migrated with the linear standard (data not shown).

One-step and two-step lariat RNA synthesis for ai5γ (Fig. 5 )

For two-step ai5γ lariat RNA synthesis, the L substrate (D123) was prepared as above with 5′-triphosphate and internal ³²P-radio-labels, and the R substrate (D56) was also prepared as above with 5′-monophosphate. For deoxyribozyme-catalyzed branched RNA formation (step 1), the R substrate (D123) (5 pmol), deoxyribozyme (15 pmol), and L substrate (D56) (30 pmol) were annealed in 5.9 μL of 5 mM HEPES (pH 7.5), 15 mM NaCl, and 0.1 mM EDTA by heating at 95°C for 3 min and cooling on ice for 5 min. The sample volume was raised to 10 μL, with final concentrations of 50 mM EPPS (pH 8.5), 150 mM NaCl, 2 mM KCl, and 40 mM MgCl₂. The sample was incubated at 37°C; aliquots were quenched and assayed as described for the YBL059W lariat. For two-step ai5γ lariat RNA synthesis without a stem-loop to control reactivity during step 2, D123 was shortened by 16 nt, such that no stem-loop was present. The remainder of the reaction was performed as described here for step 1.

For loop closure catalyzed by T4 RNA ligase (step 2), the branched product from step 1 was prepared by a modification of the above procedure. The R substrate (D123; 100 pmol), deoxyribozyme (150 pmol), and L substrate (D56; 250 pmol) were annealed in 37.5 μL of 5 mM HEPES (pH 7.5), 15 mM NaCl, and 0.1 mM EDTA by heating at 95°C for 3 min and cooling on ice for 5 min. The sample volume was raised to 50 μL with final concentrations of 50 mM EPPS (pH 8.5), 150 mM NaCl, 2 mM KCl, and 40 mM MgCl₂. The sample was incubated at 37°C for 1.5 h and quenched with 50 μL of stop solution. The branched RNA product was purified by 5% PAGE. Then, loop closure catalyzed by T4 RNA ligase was performed as described above for the second step of the two-step YBL059W lariat synthesis. To prepare the lariat RNA samples for the debranching enzyme assays, the entire 10-μL sample was quenched after 1 h with 20 μL of stop solution, and the lariat RNA product was purified by 5% PAGE.

This research was supported by the Burroughs Wellcome Fund (New Investigator Award in the Basic Pharmacological Sciences), the March of Dimes Birth Defects Foundation (Research Grant No. 5-FY02-271), the National Institutes of Health (GM-65966), the American Chemical Society Petroleum Research Fund (38803-G4), and the UIUC Department of Chemistry (all to S.K.S.). S.K.S. is the recipient of a fellowship from the David and Lucile Packard Foundation. We are grateful to Scott W. Stevens (University of Texas at Austin) for the generous gift of purified yeast debranching enzyme Dbr1p. We also thank members of the Silverman lab for discussions.