key: cord-0026779-e2ftk68z
authors: Gottstein-Schmidtke, Sina R.; Duchardt-Ferner, Elke; Groher, Florian; Weigand, Julia E.; Gottstein, Daniel; Suess, Beatrix; Wöhnert, Jens
title: Building a stable RNA U-turn with a protonated cytidine
date: 2014-08-03
journal: RNA
DOI: 10.1261/rna.043083.113
sha: 01024dc70f090b164dd1d90136c2bdce883601a4
doc_id: 26779
cord_uid: e2ftk68z

The U-turn is a classical three-dimensional RNA folding motif first identified in the anticodon and T-loops of tRNAs. It also occurs frequently as a building block in other functional RNA structures in many different sequence and structural contexts. U-turns induce sharp changes in the direction of the RNA backbone and often conform to the 3-nt consensus sequence 5′-UNR-3′ (N = any nucleotide, R = purine). The canonical U-turn motif is stabilized by a hydrogen bond between the N3 imino group of the U residue and the 3′ phosphate group of the R residue as well as a hydrogen bond between the 2′-hydroxyl group of the uridine and the N7 nitrogen of the R residue. Here, we demonstrate that a protonated cytidine can functionally and structurally replace the uridine at the first position of the canonical U-turn motif in the apical loop of the neomycin riboswitch. Using NMR spectroscopy, we directly show that the N3 imino group of the protonated cytidine forms a hydrogen bond with the backbone phosphate 3′ from the third nucleotide of the U-turn analogously to the imino group of the uridine in the canonical motif. In addition, we compare the stability of the hydrogen bonds in the mutant U-turn motif to the wild type and describe the NMR signature of the C+-phosphate interaction. Our results have implications for the prediction of RNA structural motifs and suggest simple approaches for the experimental identification of hydrogen bonds between protonated C-imino groups and the phosphate backbone.

In order to carry out their biological functions, many RNA molecules, such as tRNAs, ribosomal RNAs, self-splicing introns, RNaseP-RNA, or riboswitches, adopt intricate threedimensional globular structures rivaling or even surpassing protein structures in their complexity. The steadily rising number of novel X-ray-or NMR-derived structures determined for functional RNAs continuously reveals novel architectures of three-dimensional RNA tertiary structures. While all these structures are very different from each other, many of them share smaller three-dimensional RNA folding motifs or submotifs as common structural building blocks. Such motifs are, for example, stable tetraloops (Woese et al. 1990 ; Allain and Varani 1995; Jucker and Pardi 1995a; Jucker et al. 1996) , tetraloop-receptor interactions (Costa and Michel 1995; Cate et al. 1996) , kink-turn motifs (Klein et al. 2001) , ribose-zippers (Cate et al. 1996) , E-loop motifs (Wimberly et al. 1993; Wimberly 1994) , A-minor motifs (Nissen et al. 2001) , and different types of turn-motifs. A classic example for an RNA folding submotif is the so called "U-turn" motif, which was initially discovered in the early tRNA structures solved by Rich and coworkers and Klug and coworkers in the 1970s (Klug et al. 1974; Quigley and Rich 1976) . The Uturn motif occurs in the anticodon-as well as the TΨCloop of many tRNAs. Subsequently, the U-turn motif was found to be widespread not only in tRNAs but also in many other functional RNAs, including ribosomal RNAs (e.g., Huang et al. 1996; Nagaswamy et al. 2001, pdb entry 1k5i; Zhang et al. 2001 , pdb entry 1hs2), the hammerhead ribozyme (Pley et al. 1994, pdb entry 1hmh; Doudna 1995) , and snRNAs (Stallings and Moore 1997 , pdb entry 2u2a), as well as viral RNA elements (Puglisi and Puglisi 1998, pdb entry 1bvj; Liu et al. 2007 ). As already implied by their name, Uturns induce a sharp turn in the RNA backbone and therefore present one possibility to introduce a hairpin loop in an RNA structure. In general, U-turn motifs share the 3-nt consensus sequence 5 ′ -UNR-3 ′ , where N stands for any nucleotide and R for a purine nucleotide. The change in the direction of the RNA backbone occurs at the phosphate group 3 ′ from the U residue of the consensus sequence. Canonical Uturns are stabilized by two hydrogen bonds. One hydrogen bond is formed between the uridine N3H3 imino group and the phosphate group at the 3 ′ site of the purine nucleotide. The second hydrogen bond involves the 2 ′ -OH group of the uridine as the hydrogen bond donor and the N7 nitrogen of the purine base as the hydrogen bond acceptor group. Despite their 3-nt consensus sequence, U-turns rarely occur as isolated trinucleotide hairpin loops closed by stable Watson-Crick base pairs. In most cases, they are part of larger loop structures often closed by noncanonical sheared G:A, reversed Hoogsteen A:U, or pyrimidine:pyrimidine base pairs (Gutell et al. 2000) . Furthermore, additional nucleotides can be inserted mostly 3 ′ from the U-turn consensus sequence. Importantly, these additional nucleotides are often solventexposed and therefore available for long-range tertiary interactions with other RNA elements. The highly stable GNRA-tetraloop motifs which are also nucleation sites for long-range RNA-RNA interactions share many structural features with the U-turn motif (Jucker and Pardi 1995b; Gutell et al. 2000) , such as a base-phosphate hydrogen bonding interaction, a hydrogen bond between a 2 ′ -OH-group and the N7-nitrogen of the conserved purine base, as well as the backbone geometry at the turning phosphate group.

The neomycin sensing riboswitch is a synthetic gene regulatory element developed by a combined in vitro and in vivo selection procedure for the translational control of gene expression in response to the aminoglycoside neomycin in Saccharomyces cerevisiae (Weigand et al. 2008) . With a minimal size of only 27 nt, it is the smallest known ligand-dependent RNA-based regulatory element. The riboswitch adopts a hairpin-like secondary structure (Fig. 1A) and is able to bind its ligand neomycin with an affinity in the low nanomolar range (Weigand et al. 2011) . Upon insertion into the 5 ′ UTR of mRNAs in front of the AUG-start codon, it interferes with the scanning small ribosomal subunit in a ligand-dependent manner. When no neomycin is present, the scanning small ribosomal subunit reads through the hairpin structure and reaches the start codon. In the presence of neomycin, the hairpin structure is stabilized and presents a road block for the small ribosomal subunit. The hairpin-like secondary structure of the neomycin riboswitch is formally capped by a strongly conserved hexaloop with the sequence 5 ′ -U 13 U 14 U 15 A 16 A 17 U 18 -3 ′ , which adopts a well-defined and rigid structure ( Fig. 1A ; Duchardt-Ferner et al. 2010 , pdb entry 2kxm). The first and the last nucleotide of the hexaloop U13 and U18 form an asymmetric cis Watson-Crick/ Watson-Crick U:U base pair stabilized by two direct hydrogen bonds between the two pyrimidine bases. Nucleotides U14, U15, and A16 fold into a canonical U-turn structure with hydrogen bonds between the N3H3-imino group of U14 and the 3 ′ phosphate group of A16, as well as between the 2 ′ -OH group of U14 and the N7-nitrogen of A16 (Fig.  1B) . A17 is looped out of the structure and reaches down to the ligand binding site. Importantly, it is the only nucleotide in the apical hexaloop involved in direct contact with neomycin, and its conformation is stabilized by stacking and electrostatic interactions between its base and the ligand (Fig. 1A) . In principle, the formation of the U13:U18 base FIGURE 1. Structural and functional characteristics of the neomycin riboswitch and its U14C mutant. (A) Secondary (left) and tertiary (right) structure of the WT neomycin riboswitch bound to ribostamycin. Residues which are involved in ligand binding are shown as open letters, and the residues forming the U-turn are indicated in red. The site of the U14C mutation is indicated. In the 3D structure, the ligand ribostamycin is shown in stick representation. The position of the base of A17, the only nucleotide from the apical loop in contact with the ligand, is marked by an arrow. (B) (Left) 3D-structure of the apical loop of the WT RNA-ribostamycin complex comprising the U-turn motif and the closing U13:U18 base pair . Individual nucleotides are labeled according to the color code and numbering scheme used in A. Hydrogen bonds between the U14 N3H3 imino group and the phosphate group 3 ′ from A16 as well as between the U14 2 ′ -OH group and the A16 N7 nitrogen are indicated by dashed red lines, and the donor and acceptor groups are labeled. U14, U15, and A16 adopt a classical Uturn fold showing all the canonical stabilizing interactions shown in the schematic representation of a consensus U-turn motif on the right. Stacking interactions are marked by elongated red spheres. (C) In vivo gene regulatory activity of the U14C mutant compared to the WT riboswitch as monitored in a GFP reporter gene assay in S. cerevisiae in the presence of neomycin (black bars) or in the absence of ligand (white bars). (D) ITC titration of the U14C mutant with the ligand neomycin. The resulting dissociation constant is given.

pair and the folding of the U-turn motif encompassing nucleotides U14 to A16 can already be observed in the absence of the ligand at low temperatures ). However, ligand binding strongly stabilizes all elements of the loop structure by interacting with A17 and, through indirect effects, by stabilizing the helical element below the loop.

In a recent screen aimed at identifying functional sequence variants of the apical loop of the neomycin riboswitch, we found that a replacement of U14 by a C residue (Fig. 1A ) also results in functional riboswitches (Weigand et al. 2014 ). Such a mutation should interfere with the proper folding and the stability of the U-turn motif because C lacks an imino proton and is thus not able to form a hydrogen bond analogous to the one between the U imino group and the phosphate backbone in the canonical U-turn motif.

Interestingly, function-retaining replacements of the U by a C in biologically important U-turn motifs have been reported previously. For instance, the stem-loop V in the VS ribozyme folds into a U-turn motif thought to be involved in substrate binding (Campbell and Legault 2005; Campbell et al. 2006, pdb entries 1tbk, 1yn2) . The U to C mutation of U696 in the U-turn is the only variant that retains significant ribozyme activity (Rastogi et al. 1996) . Another example is the stem-loop IIa of yeast U2 snRNA (SL2), which contains a conserved U-turn motif as part of the consensus sequence 5 ′ -UAAY-3 ′ (Stallings and Moore 1997, pdb entry 2u2a). The U56C-SL2-mutant with the sequence 5 ′ -CAAC-3 ′ is the only viable alternative. However, the structural basis for the ability of C to replace the U in these RNAs has not been established, and it is unclear if the mutant RNAs retain canonical U-turn structures.

Assuming that the three-dimensional loop structure in the U to C mutants in these systems must be preserved to retain function, there are a number of ways to rationalize these findings. First, it is possible that the remaining hydrogen bonding and stacking interactions are strong enough to preserve the fold of these loops despite the absence of the hydrogen bond between the imino group and the phosphate. Second, a slight shift of the base position in the C-mutant might bring the C4 amino group of the C into a position to hydrogen bond with the phosphate. Third, the C in the mutant could adopt the rare imino tautomeric form and then form a hydrogen bond equivalent to the U imino group. Fourth, the C might become protonated at the N3 position and form a hydrogen bond as well as an ionic interaction with the negatively charged phosphate group. Recent bioinformatics surveys and quantum chemical studies of base-phosphate interactions found that cytidine amino groups are able to form highly stable hydrogen bonds with the phosphate backbone but did not discuss evidence for interactions between protonated cytosine bases and phosphate groups (Zirbel et al. 2009 ).

Here, we investigated in detail the structural basis for the conservation of function in the U14C-mutant of the neomycin riboswitch-ligand complex using NMR spectroscopy. We found that C14 is stably protonated and forms a hydrogen bond which is isosteric to the one between the U imino group and the phosphate in the canonical U-turn motif, thereby preserving the overall fold of the apical hexaloop of the riboswitch. Using NMR methods, we characterized the strength of the hydrogen bond between the protonated C and the phosphate backbone and found that it is highly stable over a wide range of temperatures and pH values. By browsing the pdb, we discovered an additional example for the functional and structural replacement of a canonical UNR-type U-turn by a C+NR-type turn. Furthermore, protonated C-residues might play a role in stabilizing long-range RNA-RNA tertiary interactions.

Thus, our findings have consequences for the prediction of three-dimensional RNA structures from sequences since they extend the U-turn consensus sequence and may serve as an example of how protonated cytidine residues and their interactions can be easily identified by modern NMR techniques.

A cytidine is able to functionally replace the uridine in the U-turn of the neomycin riboswitch

During an in vivo screen of a neomycin riboswitch library with a randomized apical hexaloop (nt 13-18), sequences with a U to C mutation at position 14 were identified as functional despite their apparent deviation from the structurally important U-turn consensus sequence ( Fig. 1C ; Weigand et al. 2014 ). However, compared to the WT riboswitch, their regulatory power is slightly diminished. The WT riboswitch represses gfp expression 8.6-fold in the presence of 100 μM neomycin in the medium, whereas only a threefold reduction of gfp expression is achieved by the U14C-mutant (Fig. 1C) . ITC experiments in conditions as described previously for the WT (Weigand et al. 2011) show that the U14C-mutant binds to neomycin in vitro with a K D of 24.2 nM (Fig. 1D) . Thus, the affinity of the mutant is only slightly diminished in comparison to the WT riboswitch, which binds neomycin with a K D of 9.2 nM (Weigand et al. 2011) . This already suggests that the structure of the ligand binding pocket of the riboswitch is not strongly altered by the mutation in the U-turn motif of its apical loop.

The C14 is N3-protonated and forms a hydrogen bond with the backbone phosphate group as in the canonical U-turn motif While in the in vivo functional studies and in the ITC binding experiments, neomycin was used as the ligand, all NMR experiments were performed with ribostamycin as the ligand. As thoroughly demonstrated in our previous NMR studies of the neomycin riboswitch , the smaller ligand ribostamycin binds to the riboswitch with the same binding mode as neomycin. However, due to the absence of ring IV in ribostamycin, there is reduced signal overlap in some regions of the NMR spectra. In addition, we could directly transfer NMR signal assignments from our earlier NMR studies of the WT riboswitch-ribostamycin complex to the U14C-mutant.

The imino proton 1D-1 H-NMR spectra of the WT and the U14C-mutant in complex with ribostamycin are highly similar, suggesting that the three-dimensional structure of the U14C mutant complex closely resembles the WT ( Fig. 2A,  top) . Importantly, for both complexes the resonances of the U13 and U18 imino protons are present and display very similar chemical shifts and line widths. This indicates that the U13:U18 base pair directly adjacent to the U-turn motif and the mutation site is present in both complexes and is of similar stability (Supplemental Fig. 1 ). The only obvious difference between the two spectra is the absence of the signal corresponding to the U14 imino group at 10.87 ppm in the U14C-mutant, which is to be expected ( Fig. 2A, top) . However, closer inspection of the U14C-mutant spectrum re-veals a novel imino proton signal with a chemical shift of 12.83 ppm at 10°C. Similar chemical shift values are normally observed for guanine imino protons in Watson-Crick base pairs but have also been reported for imino protons of protonated cytidines in base triples when they form hydrogen bonds to guanine carbonyl groups (Nixon et al. 2002a,b; Cornish et al. 2006) . In contrast, the imino protons of protonated cytidines in base triples (e.g., Holland and Hoffman 1996; Brodsky et al. 1998; Cornish et al. 2006; Cash et al. 2013) or in the DNA i-motif (Guéron and Leroy 2000; Lieblein et al. 2012) with hydrogen bonds to nitrogen acceptors have chemical shifts larger than 14.5 ppm. 2D-NOESY experiments show that the novel imino proton resonance has NOE cross peaks to the imino protons of U13 and U18 ( Fig (Middle) Imino region of 1 H, 15 N-HSQC spectra with and without 15 N-decoupling in the direct dimension recorded with a selectively 15 N-cytidine-labeled U14C-riboswitch-ribostamycin complex. The 1 J HN coupling constant is indicated. (Bottom) Imino region of a 2D-NOESY spectrum of the U14C mutant in complex with ribostamycin. Resonance assignments are given for the diagonal peaks. The NOE indicating the spatial proximity between the imino protons of C14 and U13 is highlighted by dashed lines. (B) 2D-H(N)C spectrum recorded with a selectively 13 C, 15 N-cytidine-labeled sample of the U14C mutant bound to ribostamycin. The C4 resonance is correlated to the 1 H chemical shift of the amino groups for all cytidines (gray dashed lines). For C14, the C4 resonance is connected simultaneously to the 1 H imino and amino frequencies (red dashed line). The magnetization transfer pathway for this experiment is indicated by red arrows in the inset showing the structure of a protonated cytidine. (C) 1D-31 P spectra of WT and U14C mutant (top) and 2D-1 H, 31 P-long-range HSQC spectrum of the U14C mutant (bottom). Phosphorous resonances displaying unusual chemical shifts are assigned in the 1D-31 P spectrum of the WT. A projection is shown along the hydrogen dimension of the long-range HSQC and resonances giving rise to cross peaks are connected by dashed lines and labeled. A modeled 3D structure of the U-turn motif containing a protonated cytidine at the first position is shown as an inset. Hydrogen bonds observed as cross peaks in the 1 H, 31 P-long-range HSQC are indicated by dashed orange lines in the structure, and the donor and acceptor groups are colored according to atom type. selectively labeled with 13 C, 15 N-cytidine shows that the imino proton is connected to a cytidine N3 with a nitrogen chemical shift of 144.4 ppm. The 1 J HN scalar coupling constant between the imino proton and the N3 nitrogen is 102 Hz, which is in the same range as the one-bond scalar coupling constants observed in guanosine and uridine imino groups ( Fig. 2A,  middle) . In contrast, the 1 J HN scalar coupling measured for the imino group in C+:C base pairs in the i-motif in DNA is considerably smaller (∼45 Hz), since there the proton is delocalized in the N+H…N hydrogen bond (Lieblein et al. 2012) . NOE correlations, an 1 H, 15 N-HSQC-experiment for amino groups (Supplemental Fig. 3A) , and a 2D-H(N)C-experiment ( Fig. 2B) with the selectively cytidine-labeled sample connect the cytidine imino group with the amino group of the same cytidine with proton chemical shifts of 8.58 and 7.43 ppm (Fig. 2B) and a nitrogen chemical shift of 102.1 ppm (Supplemental Fig. 3A) . Thus, the N3-protonation of C14 is not due to the presence of the neutral imino tautomeric form, but the C14 residue is in the positively charged, N3-protonated state. The amino group chemical shifts for C14 differ significantly from those reported for other protonated cytidine residues (Holland and Hoffman 1996; Brodsky et al. 1998; Cornish et al. 2006 ). There, chemical shifts between 11.7 ppm and 8.5 ppm were reported for the amino protons. In addition, amino nitrogen chemical shifts for protonated cytidines reported previously were larger than 111.0 ppm (Nixon et al. 2002b; Cornish et al. 2005 Cornish et al. , 2006 .

The unusual RNA backbone structure of the apical hexaloop motif of the WT riboswitch ligand complex with its concatenation of the U13:U18 noncanonical base pair, the U-turn containing U14, U15, and A16, and the bulged-out A17 is reflected in its 31 P-NMR spectrum (Fig. 2C, top) . In agreement with previous observations for other U-turn structures (Morosyuk et al. 2001; Campbell et al. 2006) , the 31 P resonances of the phosphate groups 3 ′ from U14 (the turning phosphate of the U-turn), U15, and A17 are shifted downfield compared to the bulk of the backbone phosphate groups. The phosphate group 3 ′ from A16 which is hydrogen-bonded to the U14 imino group according to the Uturn consensus is shifted upfield compared to the bulk of the 31 P-resonances. Importantly, this 31 P chemical shift "fingerprint" is completely conserved in the U14C-mutant RNA (Fig. 2C, top) , indicating that the backbone structure of the WT and mutant RNA is very similar. Remarkably, the 31 P resonance of the phosphate group 3 ′ from A16 is shifted even further upfield compared to the WT (Fig. 2C, top) , likely due to its spatial proximity and hydrogen bonding to the imino group of the protonated C14. A 2D 1 H, 31 P-long-range HSQC spectrum reveals a direct correlation between the C14 H3 imino proton and this 31 P resonance, due to the presence of a small scalar coupling between these two spins (Fig.  2C, bottom) . Thereby, this experiment directly demonstrates the existence of a hydrogen bond between the N3H3 imino group of the protonated C14 and the backbone phosphate 3 ′ from A16. This hydrogen bond conforms exactly to the U-turn consensus structure. The second correlation observed in this spectrum is due to a hydrogen bond between the A16 2 ′ -OH group and the phosphate group 3 ′ from A17. This correlation was previously observed for the WT RNA-ribostamycin complex, further supporting the close structural similarity between the WT and mutant RNA structure. Notably, despite their small line widths and intense NMR signals, the C14 amino protons show no correlations in the 1 H, 31 P-long-range HSQC, suggesting that they are not involved in additional hydrogen-bonding interactions with the phosphate as is possible, e.g., in a bifurcated hydrogenbonding arrangement. The absence of hydrogen-bonding interactions involving the C14 amino group might explain the deviation in chemical shifts from those reported previously for C+ residues in other RNAs containing base triples. In these cases, the amino groups are directly hydrogen-bonded either to a C6 carbonyl oxygen (Holland and Hoffman 1996; Brodsky et al. 1998) or an N7 nitrogen of a guanine (Nixon et al. 2002b; Cornish et al. 2006) .

Previous studies have established that the chemical shifts of both the nitrogen (Wang et al. 1991 ) and those of the carbon nuclei Pardi 1994, 1997) of the bases are excellent reporters of their protonation state and can even be used for the determination of pK a values of individual bases in complex RNA structures (Legault and Pardi 1994; Cai and Tinoco 1996; Ravindranathan et al. 2000; Lupták et al. 2001; Huppler et al. 2002) . In line with these previous observations, we find that the protonation at N3 significantly affects the other chemical shifts associated with C14, in particular those of the hetero-atoms. In agreement with the findings by Legault and Pardi (1997) for protonated 5 ′ -CMP, the most pronounced effects are observed for the chemical shifts of the C4 and C2 carbon nuclei. The C4 carbon resonance frequency is shifted upfield by >4.3 ppm (Fig. 2B) , and the C2 resonance frequency is shifted upfield by >4.8 ppm (Supplemental Fig.  3F ), respectively, compared to the nonprotonated cytidine moieties in the same molecule. While they still display extreme chemical shift values, the C6, C5, N1, and C1 ′ chemical shifts of C14 differ less strongly from those of the nonprotonated cytidines (Supplemental Fig. 3B-E) . The direction of the chemical shift change caused by cytidine protonation is the same as observed previously for protonated CMP (Legault and Pardi 1997) . C6 and C1 ′ of C14 are shifted downfield, whereas C5 of C14 is shifted upfield in comparison to the chemical shifts found for the nonprotonated cytidine residues (Supplemental Fig. 3 ). As expected (Wang et al. 1991) , the N3 nitrogen chemical shift of C14 (144.4 ppm) is also clearly separated from the bulk of the cytidine N3 resonances (195-200 ppm) and is in a similar chemical shift range as observed for protonated cytidines in base triples (142.7 ppm, 138.4 ppm) stabilizing structurally related pseudoknots (Nixon et al. 2002b; Cornish et al. 2005 Cornish et al. , 2006 .

Since the C4 and C2 chemical shifts show the largest changes upon protonation, we attempted to use these chemical shifts to determine the pK a value for C14 protonation by recording H(N)C spectra at different pH values. However, at pH 8.2, we observe virtually the same C4 and C2 chemical shifts for C14 as at pH 6.2 (Supplemental Fig. 4A,B) . Neither did we observe a second set of signals for C14 in a 1 H, 13 C-HSQC spectrum (Supplemental Fig. 4C ) as would be expected when protonated and de-prontonated species would coexist in slow chemical exchange (Cai and Tinoco 1996) . This indicates that even at pH 8.2, C14 exists almost exclusively in the protonated state. Higher pH values could not be tested since then the functional groups of the ligand become significantly de-protonated (Freire et al. 2007 ), leading to changes in its RNA-binding properties. However, our data suggest that the pK a value of C14 is larger than 8.2 and thus more than four pH units above the pK a value of isolated CMP (Legault and Pardi 1997; Wilcox et al. 2011) However, such strong pK a shifts due to involvement of the protonated residue in the formation of stable RNA secondary and tertiary structures are not unprecedented (Wilcox and Bevilacqua 2013) .

To qualitatively assess the stability of the hydrogen bond between the protonated C14 and the phosphate backbone in comparison to the other base pairs in the riboswitch-ligand complex, we recorded imino proton 1D-1 H NMR-and 1D-1 H, 15 N-HSQC spectra using the selectively cytidine-labeled sample at different temperatures and pH values (Fig.  3A,B; Supplemental Fig. 5 ). From these experiments, it is immediately obvious that this hydrogen bond is very stable. The C14 imino proton signal is still detectable at 35°C and at pH 6.2 (Fig. 3A) . At pH 8.2, the imino proton is detectable (Fig.  3B ) at temperatures up to 25°C (Supplemental Fig. 5A ). The line width of the C14 imino proton signal under all these conditions is similar to the one of U14 in the WT as well as to many of the other imino proton signals ( Fig. 3A ; Supplemental Figs. 1, 5) . Thus, the stability of the C14+-phosphate interaction is comparable to those of canonical base-pairing interactions.

The stability of individual hydrogen bonds can be inferred from imino proton solvent exchange experiments (Leroy et al. 1988 ). In these experiments, magnetization of the bulk water protons is transferred to imino protons by chemical exchange. The rate constant for the transfer of the water magnetization to individual imino protons is correlated to the base-pair opening rate, which is a direct measure of basepair stability. The water exchange rates for the U14 H3 proton in the WT riboswitch-ribostamycin complex and the C14 H3 proton of the mutant complex were recorded as a function of temperature (Fig. 3C) . Our data show that the hydrogen bond involving the imino group of the protonated C14 is only slightly less stable than the canonical U-turn hydrogen bond involving the U14 imino group in the WT.

Another measure of hydrogen-bond strength is the size of the two-bond trans-hydrogen-bond scalar coupling between the hydrogen-bond donor proton and the hydrogen-bond acceptor group. The apparent 2h J HP -scalar coupling constant between the H3 imino proton of C14 and the A16 3 ′ phosphate group was determined in a quantitative spin-echo 1D-31 P{ 1 H}-experiment (Fig. 3D, Duchardt-Ferner et al. 2011 ) to be at least 3.05 Hz. Thus, this coupling constant is similar in size to the 2h J HP -coupling constant involving the U14 imino proton in the WT RNA (lower limit of 2.2 Hz), in agreement with comparable hydrogen-bonding strengths in both RNAs. The apparent 2h J HP -scalar coupling between the 2 ′ -OH proton of A16 and the A17 phosphate group in the WT RNA is slightly smaller (1.8 Hz). In comparison, the 2h J HP -scalar couplings measured between protein amide and hydroxyl protons and the phosphate group of bound FMN range from 0.5 to 1.7 Hz (Löhr et al. 2000) , whereas a 1.3-Hz coupling constant was measured between an arginine side chain proton in a small molecule receptor and a noncovalently bound phosphate group (Federwisch et al. 2008) . (D) Cross-(red) and reference (black) spectrum of a quantitative 1D 31 P { 1 H} spin-echo experiment for the U14C mutant with an evolution delay of 40 msec at 22°C. The 3 ′ phosphate groups of U8, A16, and A17 (labeled U8, A16, and A17) are all involved in hydrogen-bonding interactions as indicated by the reduction in signal intensity in the cross experiment whereas the 3 ′ phosphate group of C14 (labeled C14) is not. (E) Quantification of the scalar 2h J HP coupling constants between the phosphate group 3 ′ from A16 and C14 H3 or U14 H3 in the mutant or the WT, respectively. Signal intensity ratios between a coupled cross (I cross ) and a de-coupled reference experiment (I ref ) for the mutant (gray) and the WT (black) are plotted against the duration of the evolution delay τ. Data were fitted using the equation I cross /I ref = cos(π 2h J HP τ) to obtain the 2h J HP coupling constant.

To check for the occurrence of other hydrogen bonding interactions between phosphate backbone groups and potentially protonated cytidine residues, we analyzed all X-ray structures of RNAs and RNPs with a nominal resolution better than 2.5 Å present in the pdb for short (<3.0 Å) cytidine N3-phospate group oxygen distances. One such short cytidine N3-phosphate group oxygen distance (2.6 Å) was found in the highresolution X-ray structure of the large ribosomal subunit from the archaeon Haloarcula marismortui (Klein et al. 2004, pdb entry 1s72) . It is observed between nt C1469 and the phosphate group 3 ′ from nt A1471, which are both located in a 5 ′ -C 1469 AACU-3 ′ pentaloop closed by a G-C base pair (Fig. 4A) . If C1469 were a U, the first 3 nt of this pentaloop would correspond to the classical U-turn consensus sequence. Both the short N3…O=P distance and the angle between the donor and the acceptor group would agree with the presence of a hydrogen bond if C1469 were protonated at the N3 nitrogen. Importantly, a hydrogen bond between the 2 ′ -OH-group of C1469 and the N7 of A1471 is also present in this loop, which corresponds to the second hydrogen bond of the Uturn consensus structure (Fig. 4A) . Measurements of the backbone torsion angles confirm that the turning phosphate in this pentaloop is the phosphate immediately 3 ′ from C1469, as expected for a U-turn structure. Thus, this loop in the 23S rRNA of H. marismortui is most likely another example for a U-turn structure with a protonated C instead of the U at the first position of its consensus sequence. Remarkably, in the 23S rRNA of the thermophilic archaeon Thermococcus celer, the same loop contains the canonical Uturn sequence 5 ′ -UAACU-3 ′ , while Escherichia coli 23S rRNA contains a stable GNRA tetraloop at this position as expected if the 3D structure is conserved at this position in the rRNA (Fig. 4B ).

Hydrogen bonding between protonated cytidines and the phosphate backbone might play a role in mediating long-range tertiary RNA-RNA interactions

The search for short distances between cytidine N3 nitrogens and phosphate group oxygens also revealed a putative hydrogen-bonding interaction between C1321 and the phosphate 3 ′ from nucleotide G1221 in the structure of the small ribosomal subunit from Thermus thermophilus ( Fig. 4C ; Kurata et al. 2008, pdb entry 2vqe) . In this case, C1321 is part of a larger highly structured loop corresponding to a T-loop motif (Nagaswamy and Fox 2002) , while nt G1221 is part of a sequentially and structurally remote A-form helical element (helix 32) (Fig. 4D) . Again, the N3…O=P distance and the angle between the donor group atoms and the acceptor are highly suggestive of a hydrogen bond if C1321 were protonated. Interestingly, an equivalent hydrogen-bonding interaction is present at the same position in the 3.0-Å X-ray structure of the E. coli ribosome (Fig. 4E,F) , where, remarkably, the C is replaced by a U (Borovinskaya et al. 2007, pdb entry 2qal) . Therefore, it appears likely that interactions between protonated cytidines and phosphate backbone groups might play a more general role in stabilizing the tertiary structures of large and structurally complex RNAs. Apparently, protonated cytidines can replace uridines not only in U-turns but also in more complex structure elements.

In summary, we have shown here that a protonated cytidine (C+) can functionally and structurally replace the uridine at the first position of the classical U-turn motif by forming a very stable hydrogen bond with the negatively charged Residues 1469-1472 from the X-ray structure of the 23S large ribosomal subunit of Haloarcula marismortui (pdb entry 1s72), which form a Uturn motif with a cytidine at the first position. Potential hydrogen bonds are indicated by dashed red lines and the distances between the donor nitrogen and the acceptor oxygen are given. Donor and acceptor are colored according to atom type. (B) Primary and secondary structures of the hairpin loop between residues 1466 and 1476 in the 23S RNA in H. marimortui (left) and the respective regions in T. celer (middle) and E. coli (right). (C) Section of the X-ray structure of the small ribosomal subunit from T. thermophilus (pdb entry 2vqe) containing a potential long-range C+-phosphate backbone interaction. Possible hydrogen bonds between the protonated imino group of cytidine C1321 and the amino groups of C1322 and C1320 to the acceptor backbone phosphates are indicated by dashed red lines, and the respective distances are given. (D) Secondary structure of the corresponding region of the 16S RNA from T. thermophilus. The long-range interaction depicted in atomic detail in C is indicated by a red line. (E) Section of the X-ray structure of the small ribosomal subunit from E. coli (pdb entry 2qal). Potential hydrogen bonds of the uridine U1321 imino proton and the amino groups of C1320 and C1322 to the acceptor backbone phosphates are indicated by dashed red lines; distances are given. (F ) Secondary structure of the corresponding section of the 16S RNA from E. coli. The long-range interaction shown in detail in E is indicated by a red line. phosphate backbone in the neomycin riboswitch. Another example of the functional and structural equivalence of classical and C+-containing U-turns is found in 23S rRNA. The possibility of replacing the U with a protonated cytidine residue in U-turn structures would also help to explain mutational data for other functional RNAs, as, e.g., for the VS ribozyme (Rastogi et al. 1996) or the SL2 of the U2 snRNA (Stallings and Moore 1997) . In addition, protonated cytidine residues may play a more general role in stabilizing RNA tertiary structures by forming strong ionic hydrogen bonds with phosphate backbone groups. The unique heteronuclear chemical shift signature of selected hetero-atoms in the nucleobase moieties of protonated cytidines can aid in their rapid experimental identification even in cases where the H3 imino proton is not directly observable due to fast chemical exchange with the solvent or in cases where cytidine residues are only transiently protonated, as recently postulated in DNA (Nikolova et al. 2011) .

The gene regulatory activity of the WT neomycin riboswitch and the U14C mutant in response to neomycin was tested by an in vivo green fluorescent protein reporter gene assay in S. cerevisiae, as described previously (Weigand et al. 2008 ).

Unlabeled, selectively 15 N-cytidine-, and selectively 13 C, 15 N-cytidine-labeled samples for the U14C mutant of the neomycin riboswitch RNA with the sequence 5 ′ -GGCUGCUUGUCCUCUAAGG UCCAGUC-3 ′ were synthesized by in vitro transcription with T7 RNA-polymerase and linearized plasmid-DNA as a template. 15 Nand 13 C, 15 N-labeled nucleotide triphosphates are commercially available (Silantes GmbH). In order to generate uniform 3 ′ ends, the primary RNA transcripts contained a hammerhead ribozyme. Processed transcripts of the RNA were purified by preparative denaturing PAGE according to standard protocols. The purified RNA was folded by heating to 95°C for 5 min, followed by injection into five equivalents of ice cold water. Samples were then rebuffered and concentrated using vivaspin concentrators (MW cutoff 3.000 Da). Sample concentrations were 800 µM for the unlabeled, 616 µM for the 15 N-cytidine-, and 660 µM for 13 C, 15 N-cytidine-labeled sample. All NMR samples were prepared in NMR buffer containing 25 mM potassium phosphate (pH 6.2 or 8.2) and 50 mM KCl. Samples were titrated with 1.2 equivalents of commercially available ribostamycin (Sigma-Aldrich). Saturation of the RNA with ribostamycin was verified by the disappearance of imino resonances of the free riboswitch in the 1D-1 H spectra. Buffer exchange was performed for the 13 C, 15 N-cytidine-labeled sample with NMR buffer at pH 8.2 using vivaspin concentrators (MW cutoff 3.000 Da).

All NMR spectra were collected with 600 MHz and 800 MHz Bruker Avance spectrometers equipped with 5-mm cryogenic triple reso-nance HCN and HCP z-gradient probe heads. For the determination of the 2h J HP coupling constants, quantitative spin-echo 31 P{ 1 H}-1D spectra were obtained on a Bruker DRX300 spectrometer equipped with a 5-mm room temperature BBO z-gradient probe head as described previously . The majority of NMR resonance assignments for the U14C mutant could be transferred from the WT RNA Schmidtke et al. 2010) . Additional NMR resonance assignment experiments for the U14C-mutant RNA were recorded at 10°C in 10% (v/v) D 2 O for the exchangeable proton and in 100% (v/v) D 2 O at 25°C for the nonexchangeable protons. 2D-1 H, 1 H-NOESY-, 2D-1 H, 15 N-HSQC- (Bodenhausen and Ruben 1980) , and 2D-H(C)N- (Sklenar et al. 1993 ) experiments were carried out using standard pulse sequences (Fürtig et al. 2003) . Assignments for the cytidine C4 resonances were obtained from a 2D-H(N)C experiment with the nitrogen offset centered at 130 ppm. The assignments for the C2 resonances of the nonprotonated cytidines were derived from an 2D-H6(C6N1)C2 experiment (Fürtig et al. 2004 ). All NMR data were processed with TOPSPIN2.1 (Bruker Biospin).

Solvent exchange measurements for the WT and U14C mutant imino protons were recorded with a modified 2D-1 H, 15 N-HSQC experiment (Rinnenthal et al. 2010 ) containing a 180°-RE-BURP soft pulse (Geen and Freeman 1991) applied on the water frequency. A variable inversion recovery delay between 4 μsec and 3000 msec and a recycling delay of 6 sec were used. Data were collected within a temperature range between 0°C and 40°C. Curve fitting of signal intensities and determination of the exchange rates was carried out as described (Rinnenthal et al. 2010 ) using OriginLab 8.1.

From a list of nonredundant RNA structures in the protein data bank (release 1.16 -22/06/2013) which was compiled and made publicly available on the RNA 3D Hub web site (http://rna.bgsu. edu/rna3dhub/) by the BGSU RNA Group, all structures with a resolution of 2.5 Å or better were selected. The data set comprised 378 X-ray structures. A macro in the programming language Python was created to visualize the resulting structures in the program PyMol and analyze distances. Hydrogen-bonding patterns between potentially protonated cytidines and phosphate group oxygen atoms of the RNA backbone were analyzed by selecting the N3 atoms of cytidine residues and the OP1 and OP2 phosphate group oxygen atoms of the phosphate backbone. Distances between the nitrogen and oxygen atoms were measured, and all distances smaller than 3.0 Å were stored as putative hydrogen bonds.

Supplemental material is available for this article.

Structure of the P1 helix from group I selfsplicing introns

Natural abundance nitrogen-15 NMR by enhanced heteronuclear spectroscopy

Structural basis for aminoglycoside inhibition of bacterial ribosome recycling

NMR evidence for a base triple in the HIV-2 TAR C-G.C+ mutant-argininamide complex

Solution structure of loop A from the hairpin ribozyme from tobacco ringspot virus satellite

Nuclear magnetic resonance structure of the Varkud satellite ribozyme stem-loop V RNA and magnesium-ion binding from chemical-shift mapping

NMR structure of varkud satellite ribozyme stem-loop V in the presence of magnesium ions and localization of metal-binding sites

Pyrimidine motif triple helix in the Kluyveromyces lactis telomerase RNA pseudoknot is essential for function in vivo

Crystal structure of a group I ribozyme domain: principles of RNA packing

A loop 2 cytidine-stem 1 minor groove interaction as a positive determinant for pseudoknotstimulated −1 ribosomal frameshifting

Dissecting non-canonical interactions in frameshift-stimulating mRNA pseudoknots

Frequent use of the same tertiary motif by selffolding RNAs

Hammerhead ribozyme structure: U-turn for RNA structural biology

Highly modular structure and ligand binding by conformational capture in a minimalistic riboswitch

Rapid identification of noncanonical RNA structure elements by direct detection of OH…O=P, NH…O=P, and NH2…O=P hydrogen bonds in solution NMR spectroscopy

The Hbonding network of acylguanidine complexes: Combined intermolecular 2h J H,P and 3h J N,P scalar couplings provide an insight into the geometric arrangement

A simple NMR analysis of the protonation equilibrium that accompanies aminoglycoside recognition: dramatic alterations in the neomycin-B protonation state upon binding to a 23-mer RNA aptamer

New NMR experiments for RNA nucleobase resonance assignment and chemical shift analysis of an RNA UUCG tetraloop

Band-selective radiofrequency pulses

The i-motif in nucleic acids

Predicting U-turns in ribosomal RNA with comparative sequence analysis

Structural features and stability of an RNA triple helix in solution

Structure of a hexanucleotide RNA hairpin loop conserved in ribosomal RNAs

Metal binding and base ionization in the U6 RNA intramolecular stemloop structure

Solution structure of the CUUG hairpin loop: a novel RNA tetraloop motif

GNRA tetraloops make a U-turn

A network of heterogeneous hydrogen bonds in GNRA tetraloops

The kink-turn: a new RNA secondary structure motif

The roles of ribosomal proteins in the structure assembly, and evolution of the large ribosomal subunit

Conservation of the molecular structure of yeast phenylalanine transfer RNA in two crystal forms

Modified uridines with C5-methylene substituents at the first position of the tRNA anticodon stabilize U.G wobble pairing during decoding

In situ probing of adenine protonation in RNA by 13 C NMR

Unusual dynamics and pK a shift at the active site of a lead-dependent ribozyme

Characterization of base-pair opening in deoxynucleotide duplexes using catalyzed exchange of the imino proton

The nature of hydrogen bonds in cytidine…H+…cytidine DNA base pairs

A Uturn motif-containing stem-loop in the coronavirus 5 ′ untranslated region plays a functional role in replication

Detection of scalar couplings across NH•••OP and OH•••OP hydrogen bonds in a flavoprotein

Direct pKa measurement of the active-site cytosine in a genomic hepatitis δ virus ribozyme

Structure and function of the conserved 690 hairpin in Escherichia coli 16 S ribosomal RNA. III. Functional analysis of the 690 loop

Frequent occurrence of the T-loop RNA folding motif in ribosomal RNAs

NMR structure of a ribosomal RNA hairpin containing a conserved CUCAA pentaloop

Transient Hoogsteen base pairs in canonical duplex DNA

RNA tertiary interactions in the large ribosomal subunit: the A-minor motif

Thermodynamic analysis of conserved loop-stem interactions in P1-P2 frameshifting RNA pseudoknots from plant Luteoviridae

Solution structure of a luteoviral P1-P2 frameshifting mRNA pseudoknot

Three-dimensional structure of a hammerhead ribozyme

HIV-1 A-rich RNA loop mimics the tRNA anticodon structure

Structural domains of transfer RNA molecules

A long-range pseudoknot is required for activity of the Neurospora VS ribozyme

Adenine protonation in domain B of the hairpin ribozyme

Direct observation of the temperature-induced melting process of the Salmonella fourU RNA thermometer at base-pair resolution

NMR resonance assignments of an engineered neomycinsensing riboswitch RNA bound to ribostamycin and tobramycin

Two-and three-dimensional HCN experiments for correlating base and sugar resonances in 15 N, 13 C-labeled RNA oligonucleotides

The structure of an essential splicing element: stem-loop IIa from yeast U2 snRNA

Nitrogen-15-labeled oligodeoxynucleotides. 3. Protonation of the adenine N1 in the A•C and A•G mispairs of the duplexes d

and d[CGGGAATTC( 15 N1)ACG] 2

Screening for engineered neomycin riboswitches that control translation initiation

Mechanistic insights into an engineered riboswitch: a switching element which confers riboswitch activity

Sequence elements distal to the ligand binding pocket modulate the efficiency of a synthetic riboswitch

A simple fluorescence method for pK a determination in RNA and DNA reveals highly shifted pK a 's

Charged nucleobases and their potential for RNA catalysis

A common RNA loop motif as a docking module and its function in the hammerhead ribozyme

The conformation of loop E of eukaryotic 5S ribosomal RNA

Architecture of ribosomal RNA: constraints on the sequence of "tetra-loops

Structural characterization of a six-nucleotide RNA hairpin loop found in Escherichia coli, r(UUAAGU)

Classification and energetics of the base-phosphate interactions in RNA

This work was supported by the Aventis Foundation, the Center of Biomolecular Magnetic Resonance (BMRZ), and the Deutsche Forschungsgemeinschaft (SFB 902 A2, SU 402/4-1, WO 901/2-1).