A critical base pair in k-turns determines the conformational class adopted, and correlates with biological function

Abstract k-turns are commonly-occurring motifs that introduce sharp kinks into duplex RNA, thereby facilitating tertiary contacts. Both the folding and conformation of k-turns are determined by their local sequence. k-turns fall into two conformational classes, called N3 and N1, that differ in the pattern of hydrogen bonding in the core. We show here that this is determined by the basepair adjacent to the critical G•A pairs. We determined crystal structures of a series of Kt-7 variants in which this 3b,3n position has been systematically varied, showing that this leads to a switch in the conformation. We have previously shown that the 3b,3n position also determines the folding characteristics of the k-turn, i.e. whether or not the k-turn can fold in the presence of metal ions alone. We have analyzed the distribution of 3b,3n sequences from four classes of k-turns from ribosomes, riboswitches and U4 snRNA, finding a strong conservation of properties for a given k-turn type. We thus demonstrate a strong association between biological function, 3b,3n sequence and k-turn folding and conformation. This has strong predictive power, and can be applied to the modeling of large RNA architectures.

. The structural environment of Kt-7 in the H. marismortui 50S ribosomal subunit. The local section of 23S rRNA 49-111 is closed by one helix of a three-way helical junction (shown here at the lower end). The longest arm of this junction (on the right-hand side of the image shown) is kinked by Kt-7, the C-helix of which is thus able to make an extensive terminal loop-terminal loop interaction (highlighted red) with the remaining helix (left side). Kt-7 is also bound by the L24 protein. Thus the kturn is constrained both by tertiary interaction and by protein binding. This parallel-eye stereoscopic image was generated using PDB 3CC2.  context shown in array form as a function of the 3b,3n sequence. The composite omit map is shown for each structure, contoured at 1 σ. G2nN3-A2bN7 distances longer than 3.2 Å (i.e. those in the N1 structures) are shown red, indicating that they are too long to be hydrogen bonded.  basepair in each case. The type of 3b,3b basepair is indicated by the symbol type, according to the key shown on the right, and each point is identified by its k-turn. Abbreviations rs = riboswitch, cob = cobalamine, dcGMP = dicyclic GMP. Ribosomal k-turns have the prefix Kt and modified Kt-7 k-turns are indicated by 7_XX where XX designates the 3b,3n basepair. Kt7 ribo designates the Kt-7 k-turn in the context of the ribosome. Figure S7. The distribution of natural k-turn sequences for bacterial Kt-7, Kt-46, SAM-I riboswitches and U4 snRNA according to their 3b,3n sequences. Each cell of the arrays is labelled by their preferred N3 or N1 conformation, and colored by their ability to fold in the presence of metal ions (red good folding, blue poor folding). The sequence distribution is shown as percentages of the total for each 3b,3n sequence, with the single largest written bold for each k-turn. The number of sequences analysed were 2,722 bacterial Kt-7, 3,181 Kt-46, 4,755 SAM-I riboswitch and 9,235 U4 snRNA k-turns.

RNA synthesis
Ribooligonucleotides were synthesized using t-BDMS phosphoramidite chemistry (Beaucage & Caruthers, 1981), as described in Wilson et al. (Wilson et al, 2001 The concentration of RNA was determined by measuring the absorbance at 260 nm.

Preparation of SAM-I riboswitch variants
A plasmid containing a gene encoding the Thermoanaerobacter tengcongensis SAM-I riboswitch (Montange & Batey, 2006) in which the natural k-turn was replaced by HmKt-7 (Daldrop & Lilley, 2013)  measuring the absorbance at 260 nm using extinction coefficients calculated from the nucleotide composition and a correction factor for the hypochromic effect.
Expression and purification of human U1 snRNP protein A U1A-RBD (residues 1-102) (Nagai et al, 1990)  was applied to three tandem 5 ml CM columns (GE Healthcare) and the protein eluted with 200 mM NaCl in buffer T. U1A was then applied to a heparin column (GE Healthcare) and eluted at 400 mM NaCl using a gradient from 50 to 2000 mM NaCl in 20 mM HEPES.Na (pH 7.6). The protein was further purified using a Superset 75 gel filtration column in a buffer containing 5 mM Tris.HCl (pH 8.0), 100 mM NaCl.

Expression and purification of A. fulgidus L7Ae
The gene encoding full-length A. fulgidus L7Ae was cloned into a modified pET- PreScission protease in 20 mM HEPES-Na (pH 7.6), 100 mM NaCl, 0.5 mM EDTA at 4-8°C for 16 h. L7Ae was applied to a heparin column (GE Healthcare) and eluted at 250 mM NaCl in a gradient from 50 to 2,000 mM NaCl in 20 mM HEPES-Na (pH 7.6). The protein was further purified using a Superdex 200 gel filtration column in a buffer containing 5 mM Tris.HCl (pH 8.0), 100 mM NaCl. The protein concentration was measured by absorbance at 280 nm using a molar extinction coefficient of 5,240 M -1 cm -1 for L7Ae. The protein was concentrated to 20 mg/ml in buffer containing 5 mM Tris-HCl (pH 8.0), 100 mM NaCl, and stored at -20 °C as aliquots.
Crystallization, structure determination, and refinement

Kt-7 determined in different structural environments
Two types of design were used in crystallization trials of Kt-7 RNA in the absence of Crystallographic statistics for HmKt-7 are presented in Table S5.

The SAM-I riboswitch variants
The SAM-I riboswitch variants (Table S4) were crystallized using the hanging drop method. Crystal trays were set up by mixing 1 μL of mother liquor with 1 μL of 400 μM RNA plus 1 mM S-adenosylmethionine (Sigma Aldrich) in 40 mM Na-cacodylate (pH 7.0). Drops were seeded using a micro-crystals taken from crystal trays containing the unmodified RNA (corresponding to structure PDB 4B5R). The mother liquor of the drop that yielded the crystal variants used for data collection contained 40 mM Na-cacodylate (pH 7.0), 12 mM spermine-HCl, 80 mM KCl, 10-60 mM BaCl 2 and 8-16% (v/v) MPD. Crystals were grown at 20 °C. For data collection the crystals were cryo-protected using the corresponding well solution with 25% (v/v) ethylene glycol. Cryoprotectant was applied for approximately one minute before freezing the crystal in liquid nitrogen. Diffraction data were collected on different beamlines, including I02, I03, I04, I04-1 and I24 at Diamond Light Sources and ID23-1 and ID29 at ESRF. Data were indexed, integrated and scaled using XDS (Kabsch, 2010) or iMOSFLM and Scala from the CCP4 suite of programs (CCP4, 1994;Winn et al, 2011). Structures were solved by performing molecular replacement using PDB entry 3GX5 (Montange & Batey, 2006) or 4B5R as a preliminary model.
The structures were refined using Phenix refine, and the model was built using COOT (Emsley & Cowtan, 2004). The composite omit map was calculated using Phenix (Adams et al, 2010). Crystallographic statistics for Kt-7 variants in the SAM-I riboswitch are presented in Table S6.
Sequence alignment and analysis.
Bacterial Kt-7 and Kt-46 sequences were taken from the Comparative RNA Web Site.
Specific k-turn regions were aligned manually using Jalview 2.8 (Waterhouse et al, 2009). This resulted in the analysis 2,722 of Kt-7 and 3,181 of Kt-46 sequences. 4,755 SAM-I riboswitch and 9,235 U4 snRNA sequences were taken from the Rfam database (Burge et al, 2013). All sequence composition and covariation analysis was calculated using a modified version of Jalview, that was kindly provided by Dr James Procter (University of Dundee).