Dramatic improvement of crystals of large RNAs by cation replacement and dehydration

Summary

Compared to globular proteins, RNAs with complex three-dimensional folds are characterized by poorly differentiated molecular surfaces dominated by backbone phosphates, sparse tertiary contacts stabilizing global architecture, and conformational flexibility. The resulting generally poor order of crystals of large RNAs and their complexes frequently hampers crystallographic structure determination. We describe and rationalize a post-crystallization treatment strategy that exploits the importance of solvation and counterions for RNA folding. Replacement of Li⁺ and Mg²⁺ needed for growth of crystals of a tRNA-riboswitch-protein co-crystal with Sr²⁺, coupled with dehydration, dramatically improved the resolution limit (8.5 to 3.2 Å) and data quality, enabling structure determination. The soft Sr²⁺ ion forms numerous stabilizing intermolecular contacts. Comparison of pre- and post-treatment structures reveals how RNA assemblies redistribute as quasi-rigid bodies to yield improved crystal packing. Cation exchange complements previously reported post-crystallization dehydration of protein crystals, and represents a potentially general strategy for improving crystals of large RNAs.

Introduction

Novel non-coding RNAs are being discovered rapidly through the application of next-generation sequencing and genomic technologies. Many of these RNAs have been implicated in important cellular processes, but elucidation of their molecular mechanisms of action is often hampered by the paucity of structural information (Wan et al., 2011). X-ray crystallography is the method of choice for structural determination of large RNAs and RNA-protein complexes. However, it is rare for crystals of RNAs with complex three-dimensional structures to diffract X-rays to resolutions useful for biochemical insight, i.e., ~ 3.5 Å or better (Ferré-D’Amaré, 2010; Ferré-D’Amaré et al., 1998). Although several post-crystallization treatments that improve the quality of protein crystals have been described (Heras and Martin, 2005), comparable methods have not been well documented in the challenging arena of RNA crystallography.

Recently, we reported the crystal structure of a ternary complex formed between the complete Stem I domain of a T-box riboswitch, its cognate tRNA, and the RNA-binding protein YbxF (Zhang and Ferré-D’Amaré, 2013). Although structures of some of the isolated components of this complex had been previously reported (Baird et al., 2012; Grigg et al., 2013), the co-crystal structure allowed visualization of a gene-regulatory tRNA-mRNA complex formed outside the context of a translating ribosome for the first time (Chetnani and Mondragon, 2013; Zhang and Ferré-D’Amaré, 2013). Successful structure determination was made possible by a novel post-crystallization treatment that combines cation replacement and dehydration, which improved the diffraction limit of the crystals from ~8 Å to ~ 3 Å resolution. We have now systematically examined the effect of individual treatments alone and in combination. Our analysis reveals unique roles of Sr²⁺ in stabilizing RNA structure and bridging RNA-RNA packing interactions, in particular, through specific binding to RNA terminal cis-diols. To illuminate the underlying physical basis for the dramatic improvement in crystal order, we determined and compared a series of structures of crystals subjected to a variety of post-crystallization treatments, including the untreated crystals (that diffract to only ~8.5 Å resolution). Structural comparison allows us to track how the RNA molecules reorient in the crystals and how such shifts generate improved crystal-packing contacts that ultimately lead to the improvement in crystalline order.

Results

Dehydration and cation replacement synergistically improve crystals

Square, plate-shaped co-crystals containing the ternary complex of Oceanobacillus iheyensis glyQ T-box Stem I RNA, a circularly permuted Bacillus subtilis tRNA^Gly, and B. subtilis YbxF protein grew optimally in the presence of 50 mM Bis-Tris pH 6.5, 0.3 M Li₂SO₄, 20 mM MgCl₂ and 20% (w/v) polyethylene glycol (PEG) 3350. When examined by rotation photography using synchrotron radiation, these crystals diffracted X-rays only to 8 Å resolution. Moreover, the Bragg spots were irregular and streaky, hampering data collection (Figure 1A). In an effort to improve the quality of these crystals, a wide variety of post-crystallization treatment strategies were tested. Ultimately, co-crystals grown in those conditions were incubated in a solution from which the Li₂SO₄ was omitted, 20 mM MgCl₂ was replaced with 40 mM SrCl₂, and the concentration of PEG 3350 was raised to 40–48% (w/v). The combined dehydration and cation exchange substantially improved the diffraction spot profiles and data quality (Figure 1), dramatically extended the resolution of useful data (Table 1), enabled the identification of the two selenium atoms present in the 66 kDa complex for experimental phasing by single-wavelength anomalous dispersion (SAD), and allowed refinement of the structure at 3.2 Å resolution. Presumably in response to the large, sudden osmolarity change due to increased PEG concentration, most crystals developed cracks and disintegrated quickly. To reinforce them, crystals were grown under the same conditions but in the presence of ~ 0.2% (w/v) low melting-point agarose. Although infrequently used, in-gel crystallization using agarose, silica, or other gel matrices has been reported to produce crystals comparable in quality to those grown under microgravity (Chayen, 2004), due to the reduction of nucleation events, suppression of convection, and improved mechanical properties (Lorber et al., 2009). Previously, the inclusion of 0.01% agarose aided the crystallization of a B. stearothermophilus RNase P variant (Kazantsev et al., 2009). The use of agarose not only increased the thickness of our plate-like T-box ternary complex crystals, but greatly reduced the frequency and extent of crystal cracking during the treatment, presumably due to the presence of agarose fibers randomly deposited inside crystal solvent channels (Lorber et al., 2009).

Figure 1. Effect of cation replacement and dehydration on diffraction quality of crystals of a T-box riboswitch-tRNA-YbxF ternary complex.

(A) Portions of rotation photographs demonstrating the diffraction properties of untreated (as-grown) crystals (top, PDB:4TZP), partially treated crystals (left, right and lower right, PDB: 4TZV, 4TZW and 4TZZ), and crystals that were subjected to full cation replacement and dehydration (lower left PDB: 4LCK). Magnified insets demonstrate the improvement in spot profile and order-to-order separation. Arrows indicate progressive additions of treatments. Post-crystallization treatment and crystal properties are summarized in Table 1.

(B) Initially grown plate-like crystals of the T-box riboswitch-tRNA-YbxF ternary complex. The scale bar here and in (C) and (D) represent 200 μm.

(C) Crystals in (B), grown in 20% PEG3350 crack and disintegrate upon exposure to 40–50% PEG 3350 in dehydration solution.

(D) Same crystals grown in the presence of ~0.2% low-melting agarose exhibit drastically reduced cracking when exposed to the same dehydration solution. The crystals in this micrograph have been dissected out of the gellified agarose network.

Table 1.

Select properties of crystals treated with varying degrees of ion replacement and dehydration.

PDB code	Li2SO4 (mM)	MgCl 2 (mM)	SrCl2 (mM)	PEG 3350 (%w/v)	Resolution (Å)	Space Group	Unit cell dimensions (Å)	VM (Å3/ Da)	VS (%)
4TZP	300	20	0	20	8.5	C2221	108.7, 108.8, 291.8*	3.27	62.4
4TZV	0	20	0	20	5.0	P43212	75.7, 75.7, 270.2*	2.93	58.1
4TZW	0	0	50	20	4.7	P43212	75.3, 75.3, 268.9*	2.89	57.4
4TZZ	0	100	0	48	3.6	P21	70.6, 260.7, 70.7†	2.46	50.0
4LCK	0	0	40	40	3.2	C2221	100.3, 108.4, 266.8*	2.75	55.2

^*α = β = γ = 90°

^†α = γ = 90°, β = 92.8°

V_M Matthews coefficient (Matthews, 1968)

V_s Calculated solvent content

To identify the factors underlying the success of our post-crystallization treatment strategy, we analyzed separately the effect of removal of Li₂SO₄, exchange of Mg²⁺ with Sr²⁺, or increase in PEG concentration, and found that either individual treatment alone only modestly improved the diffraction limit (from 8.5 to 5.0 Å), implying synergy between them (Figure 1A and Table 1). Removal of Li₂SO₄ (whose presence was required for growth of robust, single crystals) alone did partially improve the diffraction limit, but not the Bragg reflection profiles. Removal of Li₂SO₄, combined with an increase in PEG and Mg²⁺ concentration (e.g., to 40% and 100 mM, respectively) could, in some cases, increase the resolution of the crystals to ~ 3.2 Å. However, the Bragg spot profiles of crystals treated in this manner remained poor. Despite exhaustive screening of several hundred crystals and a variety of treatment schemes, the only datasets that produced sufficiently resolved Bragg reflection profiles (and correspondingly high quality data) for SAD phasing came from crystals that had undergone exchange of Mg²⁺ with Sr²⁺, in addition to removal of Li₂SO₄ and increase in PEG concentration. Interestingly, the presence of Ba²⁺, a commonly used phasing heavy atom, strongly degrades the diffracting quality of these crystals, presumably due to its larger ionic radius than Sr²⁺. These results suggest that, among the cations tested, Sr²⁺ uniquely improves the long-range order of T-box ternary complex crystals.

Rigid body RNA rearrangements yield improved crystal packing

To visualize the response of the RNA complexes in the crystal to cation substitution and dehydration, we determined the structures of untreated crystals and of crystals subjected to various post-crystallization treatments (Table 1 & 2, Experimental Procedures). Superposition of structures of the ternary complexes in the differently treated crystals demonstrates that the T-box-tRNA-YbxF ternary complexes rearranged essentially as rigid bodies (Figure 2; RMSD = 0.9–1.4 Å for all C1′ atom pairs). The improvement in crystal packing, and the concurrent reduction in solvent content of the T-box RNA crystals resulting from post-crystallization treatments (Table 1) arise from rotations and translations of neighboring complexes (Figure 2). These bring three neighboring ternary complexes related by crystallographic symmetry into closer proximity, enabling formation of intimate stacking interactions between two patches on the rear face of the interdigitated T-loops (opposite the face interacting with the tRNA elbow; Zhang and Ferré-D’Amaré, 2013) of one complex with the apical adenine of the GAAA tetraloop capping the tRNA acceptor stem of a second complex, and also with the terminal base pair of Stem I of a third (compare Figure 3A & 3B). Notably, these movements also enable formation of an energetically favorable class-I A-minor interaction (Nissen et al., 2001) between the last adenine of the GAAA tetraloop (engineered into the tRNA; Zhang and Ferré-D’Amaré, 2014b) of one complex and the minor groove of the Stem I-proximal region of a symmetry related complex (Figure 3D). This crystal contact is present only in the optimally treated crystals, and is likely partly responsible for the dramatic improvement in diffraction limit and Bragg reflection profiles (Figure 1) brought about by full cation replacement and dehydration.

Table 2.

Crystallographic statistics for T-box ternary complex structures

PDB Accession Code	4TZP	4TZV	4TZW	4TZZ	4LCK
Data collection
Space group	C2221	P43212	P43212	P21	C2221
Cell dimensions
a, b, c (Å)	108.7, 108.8, 291.8	75.7, 75.7, 270.2	75.3, 75.3, 268.9	70.6, 260.7, 70.7	100.3, 108.4, 266.8
αβγ(°)	90, 90, 90	90, 90, 90	90, 90, 90	90, 92.8, 90	90, 90, 90
Resolution (Å)	48.63–8.50 (8.81–8.50)*	46.03 – 5.03 (5.20 – 5.03)*	45.79–4.67 (4.84–4.67)*	70.61–3.64 (3.77–3.64)*	28.42–3.20 (3.31–3.20)*
Rmerge (%)	8.3 (107.5)	7.6 (182.8)	5.2 (98.5)	7.4 (46.1)	6.2 (104.8)
<I>/<σ(I)>	10.6 (1.7)	13.5 (1.5)	15.2 (1.9)	16.1 (2.9)	11.1 (1.5)
Completeness (%)	99.9 (100.0)	99.8 (98.9)	98.2 (99.0)	84.5 (85.7)	98.9 (98.8)
Redundancy	6.7(6.7)	13.0 (13.3)	6.7 (7.2)	5.3 (4.8)	6.1 (6.2)
Refinement
Resolution (Å)	48.63–8.50 (8.81–8.50)*	46.03 – 5.026 (5.20 – 5.02)*	45.79–4.67 (4.84–4.67)*	70.61–3.64 (3.77–3.64)*	28.42–3.20 (3.31–3.20)*
No. reflections	1672 (238)	3756 (572)	4428 (616)	24059 (2401)	24325 (2369)
Rwork/Rfree (%)	26.7 (47.6) /37.3 (41.9)	27.9 (39.5) /32.2 (39.2)	26.4 (38.0) /33.1 (40.9)	21.7 (28.3) /26.9 (31.4)	19.6 (33.6) /25.2 (38.2)
No. atoms	8508	4313	4349	17284	8742
RNA	7460	3789	3789	15149	7586
Protein	1048	524	524	2096	1048
Ion	0	0	36	29	86
Water	0	0	0	0	22
Mean B-factors (Å2)	480.4	246.9	243.8	117.2	128.0
RNA	482.1	253.8	238.3	109.6	118.7
Protein	468.2	196.7	274.2	172.9	192.7
Ligand/ion	n/a	n/a	373.5	66.8	176.4
Water	n/a	n/a	n/a	n/a	64.2
R.m.s. deviations
Bond lengths (Å)	0.004	0.007	0.009	0.002	0.001
Bond angles (°)	0.82	1.27	1.68	0.49	0.37

^*Highest resolution shell in parenthesis

Figure 2. In-crystal redistribution of T-box ternary complexes as rigid bodies driven by dehydration and cation replacement.

Overlay of T-box ternary complexes in untreated (as-grown) crystals (light blue, Figure 1B, PDB:4TZP) and fully dehydrated and cation-exchanged crystals (blue, Figure 1D, PDB:4LCK). The corresponding crystallographic unit cells are also shown, indicating close to ~10% compression along both a and c axes. The reference complexes in the center of the panel superimpose well (RMSD for 172 C1′ < 1.4 Å), but the neighboring four complexes shift substantially closer as a result of the post-crystallization treatment (RMSDs range from 3 to 10 Å and 10 to 19 Å, for RNA C1′ and protein Cα, respectively). Red arrows denote directions of displacement (translation and rotation) of the four neighboring complexes.

Figure 3. In-crystal movements of T-box ternary complexes produce superior crystal contacts.

(A) Detail of a major crystal contact in untreated crystals involving three symmetry-related ternary complexes, shown in light blue, pale green, and salmon, respectively. The rear face of the interdigited T-loops of Stem I distal region (opposite the face interacting with the tRNA elbow, Zhang and Ferré-D’Amaré, 2013) form a prominent flat surface available for crystal packing. Two patches of this flat surface (A39 and A60 respectively, in deep salmon) are adjacent to but not in direct contact with the apical adenine of the GAAA tetraloop capping the tRNA acceptor stem (tA73, lower left inset) of a second complex (pale green), and with the terminal base pair of T-box Stem I (G1•C102, lower right inset) of a third complex (salmon), respectively.

(B) Detail of an improved crystal contact found in the same region in optimally treated crystals through cation replacement and dehydration. It is colored as in (A) but with more solid colors. Two neighboring complexes pack closer with the reference complex (top, marine) through translation and rotation (Figure 2), bringing tA73 into direct stacking distance (~3.1 Å) with A39, and the Stem I terminal G1•C102 into direct stacking configuration with A60. Stacking interactions are indicated by parallel lines.

(C, D) The absence (C) and presence (D) of a class-I A-minor interaction between tA75 of the tetraloop of a tRNA and the minor groove of the proximal region of Stem I (G3•C100) in untreated and optimally treated crystals, respectively, colored as in (A) and (B). tRNA residue numbers are preceded by ‘t’. Gray dotted lines indicate hydrogen bonds.

Pervasive Sr²⁺ binding uniquely improves RNA crystal quality

For large RNAs to fold into their functional, compact conformations (and analogously, to pack into crystals), counterions are required to offset the repulsion of the negatively charged backbone phosphates (Draper, 2004). Among divalents, alkaline earth metal ions (Mg²⁺, Ca²⁺, Sr²⁺, Ba²⁺) stand out as enablers of folding or catalysis for most riboswitches and ribozymes (Ferré-D’Amaré and Scott, 2010; Ferré-D’Amaré and Winkler, 2011; Zhang et al., 2010). The ionic radius, coordination number and geometry, hardness, and hydration enthalpy vary substantially among these ions. Our findings from the T-box co-crystals emphasize the importance of screening for the most suitable counterion to stabilize RNA, not only during crystallization, but in post-crystallization treatments as well. Sr²⁺ is considerably softer than Mg²⁺ and Ca²⁺, smaller than Ba²⁺, and has a coordination number up to 10–11 due to its larger radius (Hofer et al., 2006). Importantly, unlike the highly constrained octahedral (hexacoordinate) and pentagonal bipyramidal (octacoordinate) coordination geometries of Mg²⁺ and Ca²⁺, Sr²⁺ exhibits considerable flexibility and ligand mobility, thus tolerates appreciable distortions of its trigonal prism coordination geometry (Hofer et al., 2006); these properties may make it a particularly advantageous cation for RNA crystal stabilization.

Comparison of T-box-tRNA-YbxF co-crystal structures containing Sr²⁺ ions with Mg²⁺-only counterparts reveals that Sr²⁺-binding sites on these RNAs are abundant (40 Sr²⁺ ions per 102 nucleotides), and reproducible between different crystals. Importantly, the structures of the ternary complex in the presence and absence of Sr²⁺ superimpose closely (RMSD < 1.0 Å), suggesting that pervasive Sr²⁺ binding does not distort the local or global structures of either the T-box RNA or tRNA, or their interaction. Interestingly, strong Sr²⁺ electron density is present adjacent to the 3′-terminal cis-diol of Stem I (Figure 4A; the 3′ terminus of the circularly permuted tRNA is intrahelical and does not bind Sr²⁺). This Sr²⁺ bridges two T-box RNA molecules related by crystallographic symmetry through inner-sphere coordination and may thus contribute to improved crystal order. A similarly positioned Sr²⁺ ion bridges two symmetry-related 3′ cis-diols in crystals of a heptanucleotide derived from tRNA^Ala acceptor stem (Figure 4B). Addition of Sr²⁺ improved the diffraction limit of those crystals (Mueller et al., 1999) from 1.7 Å to 1.16 Å. The specific binding of Sr²⁺ to RNA 3′ cis-diols is further exemplified by the crystal structure of a quadruplex-forming RNA (Figure 4C; Pan et al., 2006). Besides at RNA 3′ termini, we observe Sr²⁺ frequently making bidentate innersphere interactions with the Hoogsteen faces of purines, at bulges and junctions where phosphates cluster, or bridging across the narrow major groove (Figure S1). The presence of Sr²⁺ also aided the crystallization of the RNA component of the RNase P ribozyme (Torres-Larios et al., 2005), among others. Taken together, the ability of Sr²⁺ to bind RNA 3′ termini and its flexible coordination geometry are properties that may allow it to improve crystalline packing of RNA.

Figure 4. Interfacial Sr²⁺ ions bridge symmetry-related RNA molecules by binding to their 3′ termini.

(A) A well-defined Sr²⁺ ion (green sphere) is seen bound to the cis-diol of the T-box RNA 3′ terminal nucleotide (C102, marine), and two non-bridging oxygen atoms of a symmetry-related T-box molecule (yellow), through inner-sphere coordination.

(B) Similar inner-sphere coordination between Sr²⁺ and two 3′ cis-diol groups in bridge two symmetry-related heptanucleotide derived from tRNA^Ala acceptor stem. The non-bonded red spheres denote water molecules coordinated by Sr²⁺.

(C) A Sr²⁺ is seen engaging similar inner-sphere coordination with a 3′ cis-diol of a quadruplex-forming RNA.

See also Figure S1.

Sr²⁺ can also facilitate the identification of cation binding sites. Although anomalous scatterers such as metal hexammines or Ba²⁺ have been successfully soaked into RNA crystals for structure determination (e.g. Klein and Ferré-D’Amaré, 2006; Krasilnikov et al., 2004; Reiter et al., 2010; Ren et al., 2012; Serganov et al., 2004; Tereshko et al., 2003), binding of these bulky complex ions can degrade RNA crystals. To further test the general utility of Sr²⁺ in RNA crystallography, we crystallized the adenine riboswitch (Serganov et al., 2004) in the presence of 150 mM Mg²⁺ and 50 mM Sr²⁺ (Experimental Procedures and Table 3). Bound Sr²⁺ ions were clearly visible in anomalous difference and residual Fourier syntheses, even with data collected at the selenium K edge (Figure 5). Compared to Mg²⁺, which scatters X-rays weakly (comparable to water), the electron-rich Sr²⁺ can be readily observed even at low resolution. Data from a Sr²⁺-soaked T-box co-crystal diffracting to only 4.7 Å reveals approximately the same number of well-defined, bound Sr²⁺ ions as does a 3.2 Å dataset, with most Sr²⁺ peaks above 4–6 s.d. in |F_o|−|F_c| residual electron density maps.

Table 3.

Crystallographic statistics of adenine riboswitch structures

PDB Accession Code	4TZX	4TZY
Data collection
Space group	P21212	P21212
Cell dimensions
a, b, c (Å)	49.6, 152.8, 25.0	49.3, 152.3, 25.0
αβγ (°)	90, 90, 90	90, 90, 90
Resolution (Å)	47.21–2.01 (2.13–2.01)*	41.4–2.57 (2.72–2.57)*
Rmerge (%)	6.5 (97.1)	7.7 (64.1)
<I>/<σ(I)>	21.1 (2.1)	15.3 (2.0)
Completeness (%)	99.6 (99.3)	99.8 (99.7)
Redundancy	5.7 (5.2)	3.8 (3.8)
Refinement
Resolution (Å)	47.21–2.01 (2.08–2.01)*	41.4–2.57 (2.66–2.57)*
No. reflections	13466 (1341)	6534 (625)
Rwork/Rfree (%)	22.4 (35.9) /25.7 (37.5)	20.4 (35.6) /25.1 (38.5)
No. atoms	1632	1542
RNA	1502	1502
Protein	0	0
Ligand/ion	19	19
Water	111	21
Mean B-factors (Å2)	45.8	52.6
RNA	46.1	52.6
Protein	n/a	n/a
Ligand/ion	37.7	65.3
Water	42.1	37.9
R.m.s. deviations
Bond lengths (Å)	0.001	0.001
Bond angles (°)	0.35	0.32

^*Highest resolution shell in parenthesis

Figure 5. A representative Sr²⁺ binding site on the adenine riboswitch RNA.

(A) A Mg²⁺ binding site on the add riboswitch RNA, gray mesh is a portion of the composite simulated anneal-omit 2|F_o|−|F_c| electron density map (2.0 s.d.)

(B) Besides the bound Mg²⁺ in the same position as in (a), a new Sr²⁺ binding site appears in both in the composite simulated anneal-omit 2|F_o|−|F_c| electron density (gray mesh, contoured at 2.0 s.d.) and anomalous difference Fourier map (orange mesh, contoured at 3.0 s.d.)

Sr²⁺ also produces robust anomalous signal and thus could facilitate experimental phasing of RNA-only crystals. RNA crystallography lacks a universal de novo phasing strategy comparable to the ubiquitously used selenomethionine substitution (Hendrickson et al., 1990; Yang et al., 1990) developed for proteins. To date, phase information for RNA-only crystal structures are generally obtained by cocrystallization or post-crystallization soaking with a panel of electron-rich or anomalously scattering compounds, such as iridium (III) hexamine, cobalt (III) hexamine, Ba²⁺, etc at various concentrations. While this trial-and-error approach does work for many RNAs, it is laborious, time-consuming, and does not guarantee success. In situations such as crystals grown in the presence of high ionic strength or ammonium ions, soaking these heavy atoms rarely produces adequate occupancy to permit phasing (Keel et al., 2007). Directed soaking strategies, such as the use of engineered G•U motif to recruit hexammine ions, have helped phase several novel RNA structures (Keel et al., 2007). The robust Sr²⁺ anomalous signal at both its K edge (16.1 keV, f″=3.7 e) and CuKα (8.0 keV, f″=1.8 e) wavelengths exceeds the routinely exploited protein sulfur anomalous signal at CuKα edge (f″=0.56 e). This and the observation that moderately sized, soft Sr²⁺ prevalently binds RNA nucleobase, backbone phosphates, and 3′ termini without deforming the structure suggest that SAD phasing using Sr²⁺ may hold promise as a general, straightforward RNA phasing strategy that complements existing methods.

Discussion

Several studies have reported on the utility of controlled dehydration for improvement of the diffraction properties of protein crystals, and investigated how protein molecules reorganized to form improved packing contacts (Deng et al., 2012; Heras and Martin, 2005; Russo Krauss et al., 2012). Using a similar approach, we have previously found that brief soaking in a high osmolarity solution of crystals of the glmS ribozyme-riboswitch improved their diffraction limit (Klein and Ferré-D’Amaré, 2009; Klein et al., 2007) from 3.0 Å to 1.7 Å. Since proteins and nucleic acids are both extensively solvated, dehydration has the potential to improve the crystalline order of both. The cation exchange strategy that we introduce here exploits the fact that, unlike most proteins, nucleic acids are surrounded and stabilized by a cationic counter-ion cloud, as well as site-specifically bound cations (Draper, 2004). Our findings with the T-box ternary complex suggests that the combination of cation exchange, Sr²⁺ soaking and controlled dehydration (possibly assisted by mechanical support from agarose fibers in solvent channels) can be a powerful post-crystallization strategy to improve the crystalline order of large RNAs and also to facilitate de novo phase determination and location of cation-binding sites.

It is noteworthy that in both cases of the glmS ribozyme and T-box-tRNA complexes in which RNA crystals responded favorably to dehydration treatments, PEG3350 is a precipitant and dehydrating agent. In fact, PEGs of various molecular masses are the most commonly used dehydration agents among reports of successful protein crystal dehydration (Russo Krauss et al., 2012). Nonetheless, a number of protein crystals have been successfully improved by dehydrating with non-PEG agents such as 3.5 M (NH₄)₂SO₄, 4.5 M NaCl, 2.0 M KCl, saturated K₂CrO₄, etc. (Heras and Martin, 2005; Russo Krauss et al., 2012). It remains unknown if these salts can also act as dehydration agents to improve crystals containing predominantly RNA. Another consideration in applying this method with other RNAs is the observation that some RNAs may possess specific, structural, or functional Mg²⁺ binding sites that may not accommodate the more bulky Sr²⁺. Post-crystallization treatment solutions containing various ratios of Mg²⁺ and Sr²⁺ may be attempted in these situations. In case of the adenine riboswitch crystals grown in the presence of both Mg²⁺ and Sr²⁺, Sr²⁺ did not replace Mg²⁺ in any of the five specific Mg²⁺ binding sites, but instead bound to a new site proximal to one of the Mg²⁺ sites (Figure 5). Due to the highly restricted octahedral coordination geometry of Mg²⁺, many divalent cations cannot stably substitute Mg²⁺ at specific Mg²⁺ sites.

Despite the empirical successes of post-crystallization treatments such as annealing and dehydration to improve the order of crystals of many proteins (Deng et al., 2012; Russo Krauss et al., 2012), and now RNAs, the structures of pre-treatment crystals have rarely been reported. Therefore, it has largely remained obscure how osmolarity changes trigger in-crystal movements or conformational changes of macromolecules. The precise nature of these molecular transitions, and how newly created crystal contacts are energetically or structurally superior to what was present in untreated crystals, frequently remain undefined. In this work, the structural determination of a series of untreated, partially treated, and fully optimized RNA crystals allowed us to track macromolecular movements within crystals when osmolarity is drastically changed. Our structures suggest that the formation of intimate stacking interactions and a stable A-minor interaction were the key enablers of drastically improved crystalline order for our T-box ternary complex crystals. This finding is consistent with the central role of nucleobase stacking in stabilizing both RNA architecture and crystal contacts (Hermann and Patel, 1999; Zhang and Ferré-D’Amaré, 2014a; Zhang et al., 2010).

In-crystal movements of macromolecules induced by dehydration and ion exchange are subject to constraints of crystal packing, and controlled by kinetic and thermodynamic parameters associated with increased desolvation, disruption of existing packing contacts, and formation of new, energetically more favorable packing interactions. Our comparative analyses show that, upon dehydration, T-box ternary complexes, like the glmS ribozyme-riboswitch, primarily shift as rigid bodies by up to 19 Å (Klein and Ferré-D’Amaré, 2009). In the case of proteins, e.g., the heparan sulfate proteoglycan Glypican-1, dehydration drives rigid-body translation of the proteins by ~11 Å closer to the neighboring molecules, orders a previously disordered protease domain, and creates a new intermolecular interface of as much as 375 Ų (Awad et al., 2013). The observation that both protein and RNA molecules tend to redistribute as rigid bodies upon dehydration makes thermodynamic and structural sense, as large-scale conformational changes not only would have higher activation barriers that may not be overcome by the gain from additional favorable enthalpy from more intimate packing, but may lead to a packing arrangement incompatible with the existing crystal lattice, thus causing mechanical stress to the crystals.

Experimental Procedures

Co-crystallization of the T-box stem I-tRNA^Gly-YbxF ternary complex

Co-crystals were prepared as described (Zhang and Ferré-D’Amaré, 2013). Briefly, O. iheyensis tRNA^Gly [75 nt, engineered by circular permutation (Xiao et al., 2008) and introduction of a terminal GAAA tetraloop (Zhang and Ferré-D’Amaré, 2014b)] was heated to 90°C in water for 3 min and cooled to 4 °C over 2 min, mixed with one equivalent of O. iheyensis glyQ Stem I RNA and incubated in the presence of 50 mM HEPES-KOH pH 7.0, 100 mM KCl, 20 mM MgCl₂, and 5 mM TCEP at 50 °C for 10 min and then at 37 °C for 30 min. One equivalent of selenomethionyl B. subtilis YbxF (Baird et al., 2012) was then added. The solution was adjusted to 200 μM complex, 2 mM spermine, 0.2% (w/v) low melting-point agarose and held at 37 °C. For crystallization at 21 °C by vapor diffusion, this complex solution was mixed 1:1 with a reservoir solution comprised of 50 mM Bis-Tris (HCl) pH 6.5, 300 mM Li₂SO₄, and 20% PEG3350. Plate-shaped crystals grew in 1–4 weeks to maximum dimensions of 300 × 300 × 50 μm³.

Preparation of co-crystals of adenine riboswitch and adenine

Co-crystals were prepared essentially as described (Serganov et al., 2004). Briefly, a 71-nt variant of the add adenine riboswitch RNA was produced by in vitro transcription and purified by denaturing urea-PAGE. For crystallization by vapor diffusion, a solution containing 700 μM RNA, 50 mM KOAc pH 6.8, 2 mM MgCl₂ and 5 mM adenine was mixed 1:1 (v/v) with a reservoir solution comprised of 2.8–3.0 M 1,6-hexanediol, 100 mM Tris-HCl pH 8.5, and either 200 mM MgCl₂ (PDB:4TZX), or 150 mM MgCl₂ and 50 mM SrCl₂ (PDB:4TZY). Crystals grew to maximal dimensions of 250 × 50 × 50 μm³ in five days at 4° C and were directly flash-frozen by plunging into liquid nitrogen.

Post-crystallization treatments and data collection

Crystals of the T-box ternary complex were transferred into glass depression plates filled with 150 μL of various post-crystallization treatment solutions (Table 1) and incubated for 16 hours at 21 °C. Crystals were then dissected out of the agarose using MicroSaws (Mitegen), mounted on 90° bent MicroLoops (Mitegen), and vitrified by plunging into liquid nitrogen. Crystallographic data were collected at 100 K at beamlines 5.0.1 and 5.0.2 of the Advanced Light Source (ALS), and at beamlines 24-ID-C & 24-ID-E of the Advanced Photon Source (APS) and reduced using XDS (Kabsch, 2010) and Scala (Evans, 2006) or HKL2000 (Otwinowski and Minor, 1997). Data collection statistics are summarized in Table 2.

Structural determination and refinement

Structural determination of the T-box ternary complex at 3.2 Å using data from fully ion replaced and dehydrated crystals has been described (Zhang and Ferré-D’Amaré, 2013). Structures of untreated and partially treated crystals were solved by molecular replacement with PHASER (McCoy et al., 2007) using the 3.2 Å structure (PDB: 4LCK) as search model. Unambiguous solutions were obtained for each of the datasets, with overall translation function Z-scores of 12.4, 16.7, 22.4, and 17.2, for the structures with PDB codes 4TZP, 4TZV, 4TZW, and 4TZZ, respectively. Structures of the add riboswitch-adenine complex (non-Sr²⁺, PDB:4TZX; Sr²⁺-containing, PDB:4TZY) were solved by molecular replacement with PHASER using the published structure (Serganov et al., 2004; PDB:1Y26) as a search model. Initial solutions were subjected to manual rebuilding (Emsley et al., 2010) interspersed with iterative rounds of rigid-body, simulated-annealing, and individual isotropic B-factor refinement using PHENIX (Afonine et al., 2012). Because of limited resolution of the data, structure 4TZP was only subjected to rigid body and TLS B-factor refinement. Refinement statistics are summarized in Table 2 and 3.

Highlights.

• Synergistic dehydration and cation replacement improve resolution from 8.5 to 3.2 Å

• Structures compared for pre- and post-treatment RNA-protein co-crystals

• Large RNA assemblies shift as rigid bodies to yield improved crystal contacts

• Pervasive Sr²⁺ binding strongly correlates with improved RNA crystal quality