Structure of a yeast RNase III dsRBD complex with a non-canonical RNA substrate provides new insights into binding specificity of dsRBDs

The AAGU hairpin binds to and is efficiently cleaved by Rnt1p in the context of snR47 stem sequence

We investigated, using NMR spectroscopy, the interaction of Rnt1p dsRBD with a 32 nt RNA hairpin containing a 14 bp stem and capped by the AAGU tetraloop (AAGU hairpin) found in the snoRNA snR48 precursor (see Fig 3A). To facilitate comparisons between this complex (dsRBD/AAGU) and one with a canonical A/uGNN tetraloop, we used the same stem sequence as in the previously determined structure of a Rnt1p dsRBD-snR47h complex (dsRBD/AGAA), where the hairpin sequence was derived from the snoRNA snR47 precursor [33]. The two base pairs below the tetraloop are the same for snR47 and snR48, but otherwise the stem sequences differ, except for the 1^st and 4^th base pair in our hairpins. Complex formation was initially monitored by chemical shift changes observed in the ¹H-¹⁵N HSQC spectra of the dsRBD upon the addition of AAGU (Figure S1). The dsRBD is at or near fast exchange with the hairpin on the NMR timescale, but the complex is saturated at a small excess of RNA, as observed for the dsRBD/AGAA complex, and amide chemical shift changes are very similar [33].

To compare the relative binding affinity of Rnt1p dsRBD to AGAA and AAGU hairpins, a series of ¹H-¹⁵N HSQC spectra of the dsRBD were acquired at various ratios of added AGAA, AAGU, and UUCG hairpins, under high salt conditions. Previous studies have shown that the dsRBD does not specifically recognize a UUCG-capped hairpin [39], so this hairpin was used as a comparison for non-specific binding of the dsRBD to double stranded RNA. Apparent K_d values calculated from global fitting of the HSQC titration data, assuming a one-site binding model, are 34.1±2.9, 30.1±3.8, and 280±24.6 μM for the AAGU, AGAA, and UUCG hairpins, respectively (Figure 1A,B,C). These values indicate that under the NMR conditions Rnt1p dsRBD binds with the same affinity to both the AAGU and AGAA hairpins. In contrast, non-specific binding to dsRNA capped by a UUCG tetraloop is about an order of magnitude weaker. This result is consistent with gel shift assays using full-length Rnt1p, which show it binds with a 5–10 fold weaker affinity to non AGNN-containing tetraloops than to canonical tetraloop containing substrates [39, 40].

Previously, Rnt1p was reported to require a defined sequence in the stem region of snR48-derived AAGU hairpin substrates for optimal cleavage [46, 48]. To confirm that the snR47-AAGU hybrid hairpin is a good substrate for Rnt1p, we performed Rnt1p cleavage assays with substrates derived from snR47, capped by AGAA, AAGU, or UUCG tetraloops (snR47-AGAA, snR47-AAGU, and snR47-UUCG, respectively) (Figure 2A), under single-turnover conditions (Figure 2B). Both snR47-AAGU and snR47-AGAA showed specific cleavage, while snR47-UUCG was almost unreactive over the assayed reaction time. The fact that the snR47-UUCG substrate shows only a 10-fold reduction of binding (Fig. 1C) but is almost completely refractory to Rnt1p cleavage confirms that, as shown previously [39, 40], the differences in binding affinity between different tetraloop-containing stems are not sufficient to explain the strong cleavage discrimination between these substrates. SnR47-AAGU also shows more additional cleavage products than snR47-AGAA. One corresponds to an intermediate cleaved in one strand only (band below substrate band in Figure 2B), and the others correspond to some alternate cleavage sites. In vivo, cleavage at any of these alternative sites would still be expected to result in correct subsequent processing, because each of these cleaved intermediates are expected to be used at similar efficiencies by the Rat1p exonuclease [49]. Overall, cleavage of the snR47-AAGU substrate occurred at a slightly slower rate than for snR47-AGAA, but the relative cleavage rates for this non-canonical substrate is comparable to the previously reported relative rates for canonical Rnt1p substrates containing AGAA and UGAA tetraloops [43]. These results indicate that the AAGU tetraloop in the context of the snR47 stem sequence is sufficient to support cleavage at a rate comparable to A/uGNN-tetraloop hairpin substrates.

To further investigate the ability of Rnt1p to recognize and process a substrate carrying an AAGU tetraloop, we analyzed the processing of snR47 mutant derivatives with different types of tetraloops in vivo. We introduced AAGU in place of AGAA in the stem-loop sequence upstream from the snR47 [9] chromosomal locus by homologous recombination (see Experimental Procedures). As controls for inefficient processing and complete loss of Rnt1p function, we generated a strain carrying a UUCG tetraloop and used a strain inactivated for Rnt1p (rnt1Δ), respectively. As shown by Northern blot in Figure 2D, the AAGU mutant snoRNA precursor was processed as efficiently as the wild-type precursor in vivo, showing high levels of mature snoRNA and no apparent unprocessed precursor accumulation. In contrast, samples extracted from the UUCG tetraloop mutant or from the rnt1Δstrain exhibited little or no mature snoRNA and a strong accumulation of unprocessed precursors (P, Figure 2D). Primer extension analysis of the wild-type (AGAA), AAGU or UUCG mutant strains confirmed the efficient 5′-end processing of the AAGU mutant and the strong processing defect of the UUCG mutant (Figure 2E). These results indicate that a hairpin capped by a AAGU tetraloop sequence can serve as an efficient recognition site for Rnt1p in vivo, even when placed on a substrate stem that normally contains a canonical AGNN tetraloop. In conclusion, AAGU tetraloop hairpins can be efficiently recognized by Rnt1p in vitro and in vivo, regardless of the stem sequence that they cap.

Overview of the solution structure of Rnt1p dsRBD/AAGU hairpin complex

Protein and RNA resonances in the Rnt1p dsRBD/AAGU complex were assigned following previously established protocols [50]. The protein-RNA interface was well defined by 43 intermolecular NOEs assigned from 2D filtered/edited NOESY [51] (Figure S1). The structure of the dsRBD-AAGU complex has backbone RMSDs to the mean of 0.55±0.11Å and 0.64±0.15Å for the dsRBD and AAGU hairpin, respectively (Figure 3A) (Table 1). The dsRBD adopts the standard α1β1β2β3 α2 fold [27] with the additional helix α3 packed against and stabilizing the C-terminal end of helix α1, as previously observed [33, 45]. The dsRBD binds to one face of the RNA and interacts primarily with the sugar-phosphate backbone in three successive regions: the tetraloop minor groove and top two base pairs with helix α1, the stem major groove with the N-terminal end of helix α2 and the β3-α2 loop, and the stem minor groove with the β1-β2 loop (Figure 3B). All of the bases in the tetraloop are in the anti conformation, and the RNA stem forms an A-form helix.

Analysis of RDCs indicates that the dsRBD adopts the same conformation in both the AAGU and AGAA complexes

We compared the structures of dsRBDs in the dsRBD/AAGU and dsRBD/AGAA complexes using residual dipolar couplings (RDCs) [52, 53]. Structure calculations for the dsRBD/AAGU complex included 83 ¹H-¹⁵N RDCs in the final refinement step. For comparison purposes, we measured and analyzed a comparable set of 81 RDCs for the dsRBD/AGAA complex that had been previously determined and refined with 43 RDCs [33], and re-calculated the structure. We first evaluated the quality of the structures by back-calculating the RDCs from the RDC-refined dsRBD/AAGU and dsRBD/AGAA structures using the program PALES [54] (Figure 4A,B). The Q quality factors (Q) are 11.5% and 9.4% for dsRBD/AAGU and dsRBD/AGAA respectively, indicating an excellent agreement between the structures and their RDC data [53]. We next evaluated how similar the structures of the dsRBDs in the two complexes were to each other by back-calculating the set of RDCs for each complex using the structures of the other complex. Results for the lowest energy structure in each structure ensemble are shown in Figure 4C,D. R factors for the back-calculated dsRBD/AGAA and dsRBD/AAGU complexes are 0.98 and Q factors are <20% for both cases. Similar results are observed for each set of individual structures in the structure ensembles (Figure S2). These values indicate that the dsRBD conformations in the two complexes are highly similar. The dsRBDs in the two complexes have an RMSD between the two ensembles of 1.18 +/− 0.32Å for all heavy atoms (Figure 4E), which is within experimental error of the pairwise RMSDs of each ensemble.

The AAGU tetraloop in the complex adopts a backbone fold similar to that of the AGAA tetraloop

The solution structure of the free AAGU tetraloop is substantially different from the A/uGNN fold [48] (Fig 5C) (see below). Surprisingly however, in the dsRBD/AAGU complex, the backbone fold and shape of the minor groove of the tetraloop is highly similar to that of the AGAA tetraloop (Figure 5A,B), with an RMSD for all backbone heavy atoms of 1.06 Å for the lowest energy structures (Figure 5B). As is the case for the AGNN tetraloops [33], the first two bases in the AAGU tetraloop, A15 and A16, stack on each other and point into the major groove away from the Rnt1p dsRBD helix α1 in the minor groove. The last two bases, G17 and U18 on the 3′ side, point into the minor groove and their backbones interact with helix α1 of the dsRBD (Figure 6). While all point into the minor groove, the positions of these bases, which are not conserved in AGNN tetraloops, differ somewhat among tetraloops.

Although the backbone fold is the same for the AGAA and AAGU tetraloops in complex with the dsRBD, the second base (A16) is in the anti conformation and its position is quite different from that of the syn G in the AGAA tetraloop. The syn conformation positions the G amino group within hydrogen bonding distance to one of the non-bridging oxygen atoms on its 5′ phosphate group, and therefore presumably stabilizes its stacking on the A (or U). An A in the second position does not have a proton donor at the same position as the G, and therefore cannot hydrogen bond to the backbone even in the syn conformation. Thus, in the complex, the stabilization conferred by a hydrogen bond to the backbone from a syn G at position 2 of the tetraloop does not appear to be required for the AAGU tetraloop to adopt the same backbone conformation as an AGAA tetraloop.

In the free AAGU tetraloop the first two As (A15 and A16) stack on each other and point into the major groove and the backbone turns after the second A, as is the case for the bound AAGU tetraloop, however the position of the backbone before the turn is substantially different (Figure 5C). On the 3′ side of the loop, the Watson-Crick face of G17 points into the minor groove and the base is nearly co-planar with A16, while U18 points up above G17 and out into the major groove [48]. In contrast, in the dsRBD/AAGU complex, the base of U18 is in the minor groove, and the positions of both G17 and U18 are significantly different, with the base of G17 pointing into the minor groove from near the top of the tetraloop and U18 below G17. Thus, the AAGU tetraloop in the hairpin undergoes a large conformational change upon binding to the dsRBD, to a conformation that presents a minor groove surface and backbone contacts to helix α1 that are highly similar to those in the AGAA tetraloop.

Since the dsRBD has the same binding affinity for the two hairpins, but the AAGU tetraloop undergoes a larger conformational change, we measured the binding by isothermal titration calorimetry (Figure S3). Although the two complexes have the same ΔG, the dsRBD/AAGU complex has larger negative values for both ΔH and ΔS. For dsRBD/AAGU and dsRBD/AGAA, ΔH = −1.42 × 10⁴ vs −0.91 × 10⁴ kcal/mole and ΔS = −21.8 and −3.94 kcal/moleK, respectively. These results indicate that while the AAGU/dsRBD complex undergoes a larger change in enthalpy upon complex formation, this is offset by a compensatory decrease in entropy.

Comparison of the protein-RNA interfaces in the Rnt1p dsRBD/AAGU and dsRBD/AGAA complexes

In both the dsRBD/AAGU and dsRBD/AGAA complexes, the β1-β2 loop contacts the stem minor groove of base pairs 2–5, the N-terminal end of helix α2 inserts into the major groove between base pairs 5–10, and helix α1 specifically recognizes the minor groove of the tetraloop and top two base pairs (Figure 3 and 6 and Table S1). Almost all of the contacts are to the phosphodiester backbone. Detailed comparison of the dsRBD/AAGU and dsRBD/AGAA complexes shows that the interactions between the protein and RNA stem are nearly identical (Figure 6 and S4). In the minor groove of the three base pairs adjacent to the tetraloop, the D367 side chain carboxyl group interacts with A20 2′OH and A21 2′OH through potential direct or water-mediated hydrogen bonds, and the side chain of K371 forms potential hydrogen bonds to A20 2′OH and A21O4′ for both complexes. On the 3′ side of the tetraloops, the guanidinium group of the R372 stacks onto the ring of the base G17 (in the AAGU tetraloop) or A17 (in the AGAA tetraloop) in the complexes (Figure 6 and S4). On the 5′ side of the tetraloop, the non-polar side chain of M368 stacks onto the A15 ribose and its sulfur group forms a potential water-mediated hydrogen bond with C14 O2. In the dsRBD/AGAA complex, R372 guanidinium group and S376 OH form hydrogen bonds with the 2′OH of A17 and A18 respectively (Figure S4), while in the AAGU complex the R372 guanidinium group forms a hydrogen bond with 2′OH of A15 and G17, and S376 OH forms a hydrogen bond with the 2′OH of U18 (Figure 6B). These small differences can be attributed to the difference in sequence at these positions. Neither the tetraloop sequence nor the top two base pairs are conserved, and consistent with this, no specific contacts to the bases were observed. Mutation of M368, which has a potential interaction with C14 O2, to alanine, does not affect binding [55]. Thus, the dsRBD recognizes the A/uGNN and AAGU tetraloops in the same way, by shape specific binding.

Conformational changes in the dsRBD upon binding to target RNA

Since the structure of the free Rnt1p dsRBD has been reported [45], we were able to examine any conformational changes that take place in the dsRBD upon binding to the AAGU hairpin as well as the AGAA hairpin in detail. We acquired a set of RDC data for the free dsRBD in solution. Of the NMR solution and two crystal structures (from one asymmetric unit) reported, the experimental RDCs fit best to the crystal structure chain A (PDB ID 1T4O) (Figure S5), so this structure was used for comparison to our bound complexes. To quantitatively compare the structures of the free and AAGU tetraloop bound dsRBDs, the experimental RDCs from the secondary structure elements β1, β2, β3, α1, and α2 for the dsRBD/AAGU complex were plotted versus the RDCs calculated from the crystal structure (Figure 7A). Helix α3 was excluded from this analysis, since it adopts three different orientations in the solution and two crystal structures [45]. The correlation gives a Q factor of 32%, but when the RDCs from helix α1 are deleted from the analysis, the Q factor decreases to 16%. Similar results were obtained for dsRBD bound to the AGAA hairpin (Figure 7B). When experimental RDCs from helix α1 only are compared (free dsRBD vs dsRBD/AAGU complex and dsRBD/AAGU complex vs dsRBD crystal chain A), poor correlations are obtained (Q = 32% and 49%, respectively) (Figure S5). Taken together, these data indicate that there is a significant change in helix α1 when the dsRBD binds to target RNA, consistent with structural differences observed by direct comparison of the structures as described below.

Comparison of the structures of the free and AAGU hairpin bound dsRBD revealed that all of the regions of the dsRBD that interact with the RNA show significant changes in position between the free and bound dsRBD (Figure 7C). The β1-β2 loop, which inserts into the stem minor groove in the complex, points away approximately 6Å in the free dsRBD. The N-terminal end of helix α2 and the β3-α2 loop also shift to insert into the major groove. Helix α1 rotates about 18° (Figure 7E) and bends slightly from L374 to S376 (Figure 7D) to fit into the convex surface of the tetraloop. In the free dsRBD, helix α1 begins at N369, while in the complex it begins at L366. Side chains of M368, R372, and S376 all shift positions to align along one face of the helix to form van der Waals interactions and hydrogen bonds to 2′OH in the tetraloop minor groove (Figure 7D and 6A). Thus, helix α1 undergoes a change in helix length and bend and rotates 18° when it binds to the dsRNA hairpin substrate.

Conformational change in the AAGU tetraloop upon dsRBD binding

For the AGAA tetraloop, the positions of the bases in the free vs bound are very similar, although there is some change in the backbone on the 3′ side of the loop (Figure 5D). Because the structures of the hairpins capped by AGAA, AGUU, and UGCA tetraloops, which are all substrates for Rnt1p, all had a similar fold with a syn G [43, 44] and this fold was retained in the dsRBD-AGAA complex [33], it was proposed that the dsRBD recognized the conserved shape of the tetraloop. It was therefore surprising to find that for the AAGU hairpin, the positions of the bases and the backbone trajectory both change significantly in the complex (Figure 5C). Thus, it appears that the AAGU tetraloop and helix α1 of the dsRBD co-operatively fold to form a specific complex with a conserved tetraloop fold. In the complexes, these two different tetraloops provide a rare example of two distinct RNA sequences that adopt the same functional fold [56].

While the conformational changes of the free vs bound AAGU tetraloop are larger than for the AGAA tetraloop, both the free and bound tetraloops have features in common that are likely essential for recognition and binding. In all cases, the backbone turns after the second nucleotide, and the position of the backbone in the turn is the same. On the 5′ side of the tetraloop, the first two bases point into the major groove and are stacked on each other. In the complexes, these two bases have no contacts to the dsRBD and the third base is positioned above the binding site. Finally, we note that the ACAA tetraloop has been proposed to have a similar conformation to the AGAA [57]. However, hairpins capped by ACAA are not cleaved by Rnt1p [43]. In the ensemble of ACAA tetraloop structures [57], about half have a backbone conformation at the turn that is very different from the AGAA and AAGU tetraloops, such that the helix α1 would not be able to insert into the minor groove.

Comparison to other dsRBD-RNA complexes

Although the dsRBD is the second most abundant family of RNA recognition motifs, structures of only a few dsRBDs in complex with RNA have been solved. There are now 6 proteins for which the structures of both the free dsRBD and the dsRBD in complex with RNA have been reported. In addition to Rnt1p dsRBD ([33, 45] and this work), these include Staufen dsRBD [28, 32], TAR RNA binding protein 2 (TRBP2) [35], Arabidopsis HYL1 dsRBD [58], ADAR2 dsRBD1 and dsRBD2 [36, 59], and Aquifex aeolicus RNase III [19, 31, 32, 37]. All of the free dsRBDs, with the exception of ADAR2 dsRBD1 and dsRBD2, superimpose well on each other and have virtually the same angle of helix α1 relative to the other secondary structure elements (Figure S6). Furthermore, the conformations of the free and RNA-bound dsRBDs of HYL1, TRBP, Staufen, A. aeolicus RNase III are the same, respectively, indicating that helix α1 does not change its conformation upon binding RNA. Of the complexes solved to date, only the dsRBD of Rnt1p and dsRBDs of ADAR2 have different helix α1 positions in complex with RNA relative to the free dsRBD (Figure S6 I,J). The dsRBDs of ADAR2 have recently been shown to bind dsRNA in a sequence specific manner, with base recognition via the minor groove from one amino acid each on helix α1 and the β1-β2 loop [36]. These two dsRBDs undergo relatively large conformational changes upon RNA binding, similar to Rnt1p. However, in contrast to Rnt1p dsRBD, the position of helix α1 in the free ADAR2 dsRBD1 and dsRBD2 is different compared to Xlrpba and the other dsRBDs (Figure S6).

Rnt1p requires specific tetraloop structures for substrate cleavage both in vivo and in vitro, while A. aeolicus RNase III, the homologue of E. coli RNase III, cleaves dsRNA in vitro with little apparent sequence specificity. Crystal structures of A. aeolicus RNase III in complex with RNA have revealed that, in addition to non-sequence-specific contacts to the backbone, three bases have direct contacts to the dsRBD, two in helix α1 and one in the β1-β2 loop [37]. One of these, Q157, is conserved in all bacterial RNase IIIs and deletion of it abolished cleavage and binding. The equivalent residue in Rnt1p does not contact the RNA. The other two residues, including A. aeolicus RNase III Q161 in helix α1, have been proposed to give rise to sequence specific binding [36]. Of the sequence- or tetraloop-specific dsRBD-RNA complexes solved to date, A. aeolicus RNase III is the only example where there is no significant change in the orientation of the dsRBD helix α1 upon binding to RNA.

In conclusion, our results show that the non-canonical AAGU tetraloop adopts a canonical fold upon binding to the dsRBD and that reorientation of helix α1 plays a major role in substrate-specific recognition. We propose that the Rnt1p dsRBD initially binds non-specifically to dsRNA and scans along the RNA until it reaches an A/uGNN or AAGU tetraloop. Helix α1 is locked into position by the tetraloop fold like a ball in glove, allowing subsequent positioning of the active site of Rnt1p at the cleavage site 14-16 bp away.

NMR sample preparation

The Rnt1p dsRBD, consisting of residues 366–453, was expressed as a glutathione transferase (GST) fusion protein and purified essentially as described [33] except for the addition of 1 mM DTT to the gel filtration purification step. Details of the purification are given in the Supplementary Methods. NMR samples were ~ 1 mM dsRBD in 20 mM sodium phosphate, pH 6.5, 150 mM NaCl, 1mM DTT. For NMR binding studies, the 32 nt AGAA (Figure 3A), AAGU, or UUCG hairpins were prepared by in vitro transcription using His₆-tagged mutant T7 polymerase (P266L) [60] with a synthetic DNA template and purified on denaturing gels as described [43]. Unlabeled, uniformly ¹³C,¹⁵N-labeled, and A-, U-, G- or C-¹³C,¹⁵N-labeled AAGU hairpins were used for structure determination of the dsRBD/AAGU complex. The dsRBD/AAGU complex was prepared at a 1.1:1 ratio (RNA:protein) by adding the dsRBD to the RNA under dilute conditions followed by concentration in NMR buffer to 1 mM complex.

NMR spectroscopy and structure calculations

All NMR spectra were recorded at 25ºC on Bruker DRX 500 and 600 MHz spectrometers except for 2D NOESY spectra of exchangeable proton resonances of RNA, which were recorded at 10ºC. The assignments of the Rnt1p dsRBD in the complex were derived from the analysis of 3D CBCANH, 3D CBCA(CO)NH, 3D HCCH-TOCSY, 3D HCCH-COSY, 3D ¹³C-NOESY-HSQC and 3D ¹⁵N-NOESY-HSQC experiments [61–63] acquired on ¹³C,¹⁵N-labeled dsRBD in complex with unlabeled AAGU hairpin. The assignments of the AAGU hairpin were derived from 2D HCCH-COSY, 3D HCCH-TOCSY, 2D NOESY, 2D TOCSY [64] and a suite of 2D filtered/edited NOESY [51] using unlabeled dsRBD with ¹³C,¹⁵N-A-, G-, U-, or C-labeled AAGU. Finally, intermolecular NOEs were derived from 2D filtered/edited NOESY experiments as described [51]. One bond ¹H-¹⁵N RDCs were measured from HSQC-IPAP experiments [65] in the presence and absence of C12E5/Hexanol [66] on a 600 MHz spectrometer. A total of 84, 81 and 83 RDCs were obtained for the free dsRBD, dsRBD/AGAA complex [33], and the dsRBD/AAGU complex, respectively. Structure calculations were performed essentially as described [51], and details are given in the Supplemental Experimental Procedures. For comparison purposes, the dsRBD/AGAA complex was re-refined with the larger set of RDCs (81 vs 43).

Determination of apparent K_D from ¹H-¹⁵N HSQC chemical shift titrations

The AAGU, AGAA and UUCG hairpin were individually titrated into 0.1 mM ¹⁵N-labeled dsRBD samples prepared in 500 μl high salt NMR buffer (20 mM sodium phosphate, pH 6.5, 300mM NaCl, 1mM DTT) up to RNA/protein ratios of 2:1 (Figure 1). The higher salt conditions (300mM NaCl), instead of the 150 mM NaCl used for the structure studies, were used in order to minimize non-specific binding. ¹H-¹⁵N HSQC spectra were recorded for each titration point. The apparent dissociation constant Kd was obtained from changes in weighted chemical shift differences Δδ(N,H) = [(Δδ(H_N)²+ (Δδ(N_H)/4)²]^½, assuming a one-site binding model. The titration curves were fitted globally using following equation:

where Δδ(N,H)_max is the average chemical shift difference between the free and bound forms and P₀, L₀ are the total concentration of dsRBD and AAGU hairpin, respectively.

Cleavage kinetics assays

Full-length Rnt1p was expressed with an N-terminal His₆-tag in BL-21 (DE3) Gold cells and purified on a GE Healthcare HisTrap Ni²⁺-affinity column, followed by anion exchange (HiTrap Q) and gel filtration (HiLoad S75) chromatography. Purified Rnt1p was concentrated to ~5 mg/mL. For kinetics assays, 52 nt RNA hairpins snR47-AGAA, snR47-AAGU, or snR47-UUCG (Figure 2A) were ³²P end-labeled with T4 polynucleotide kinase. Cleavage reactions were prepared under single-turnover conditions with 25 nM RNA and 1 μM Rnt1p in 20 mM Tris, pH 8.0, 150 mM NaCl at 25°C, and reactions were initiated by adding MgCl₂ to a final concentration of 5 mM. 10 μl aliquots were removed at time points of 0.25, 0.5, 1, 2, 4, and 8 minutes, and quenched with 10 μl of formamide gel-loading buffer with 20 mM EDTA. Samples were run on a 10% denaturing polyacrylamide gel (19:1 acrylamide:bisacrylamide), dried, and imaged on a Molecular Imager FX Pro Plus (Bio-Rad). Bands in the gel image corresponding to the uncleaved fraction of the substrate were quantified using ImageJ (NIH). Plotted values are the average of three experiments.

In vivo analysis of tetraloop mutants

Tetraloop mutants (AAGU or UUCG) were inserted into the tetraloop upstream from the snR47 snoRNA gene using the delitto perfetto method [68]. A core Kan^R-URA3 cassette was inserted between the second and third position of the snR47 snoRNA tetraloop, and double-stranded DNA oligonucleotides were used to excise the core sequence and introduce the AAGU or UUCG sequence. Genomic DNA sequences were confirmed by sequencing. Strains were grown in YPD and harvested, and Northern blot and primer extension analysis were performed as described [9, 55].

Accession numbers

Coordinates and restraints for the 16 lowest energy structures of the Rnt1p dsRBD/AAGU complex have been deposited in the Protein Data Bank, accession number 2LBS and chemical shifts have been deposited in the BioMagResBank, accession number 17574.