Metal-ion rescue revisited: Biochemical detection of site-bound metal ions important for RNA folding

Metal-ion rescue experiments have provided a powerful approach to identify and distinguish catalytic metal ions within RNA active sites, but the ability of such experiments to identify metal ions that contribute to tertiary structure acquisition and structural stability is less developed and has been challenged. Here, the authors used the well-defined P4–P6 RNA domain of the Tetrahymena group I intron to reevaluate prior evidence against the discriminatory power of metal-ion rescue experiments and to advance thermodynamic descriptions necessary for interpreting these experiments. The approach successfully identifies ligands within the RNA that occupy the inner coordination sphere of divalent metal ions and distinguishes them from ligands that occupy the outer coordination sphere. These results establish metal-ion rescue as a rigorous tool for identifying and dissecting energetically important metal-ion interactions in RNAs that are noncatalytic but critical for RNA tertiary structure.

Abstract

Within the three-dimensional architectures of RNA molecules, divalent metal ions populate specific locations, shedding their water molecules to form chelates. These interactions help the RNA adopt and maintain specific conformations and frequently make essential contributions to function. Defining the locations of these site-bound metal ions remains challenging despite the growing database of RNA structures. Metal-ion rescue experiments have provided a powerful approach to identify and distinguish catalytic metal ions within RNA active sites, but the ability of such experiments to identify metal ions that contribute to tertiary structure acquisition and structural stability is less developed and has been challenged. Herein, we use the well-defined P4–P6 RNA domain of the Tetrahymena group I intron to reevaluate prior evidence against the discriminatory power of metal-ion rescue experiments and to advance thermodynamic descriptions necessary for interpreting these experiments. The approach successfully identifies ligands within the RNA that occupy the inner coordination sphere of divalent metal ions and distinguishes them from ligands that occupy the outer coordination sphere. Our results underscore the importance of obtaining complete folding isotherms and establishing and evaluating thermodynamic models in order to draw conclusions from metal-ion rescue experiments. These results establish metal-ion rescue as a rigorous tool for identifying and dissecting energetically important metal-ion interactions in RNAs that are noncatalytic but critical for RNA tertiary structure.

INTRODUCTION

When RNA molecules adopt compact three-dimensional architectures, the negative charge of the ribose-phosphate backbone generates enormous repulsive energy. Counteracting this repulsion, cations condense onto all RNA molecules, forming a counterion atmosphere (Manning 1977a, 1978; Bukhman and Draper 1997; Misra and Draper 1998; Das et al. 2005b; Draper et al. 2005; Chu et al. 2008; Leipply et al. 2009). While the vast majority of cations associated with RNA reside in this diffuse ion atmosphere, some ions associate more intimately with the RNA, using their hydration shells to bind to specific sites via hydrogen bonding (outer-sphere coordination) or shedding water molecules to coordinate directly to an RNA heteroatom (inner-sphere coordination). These site-bound metal ions often appear in intricate structures within folded RNAs and, in the case of several ribozymes, participate directly in catalysis (Chen et al. 1997; Sontheimer et al. 1997; Shan et al. 1999; Gordon et al. 2007). Accordingly, understanding RNA structure and function requires knowing the locations of site-bound metal ions and the thermodynamics of their interactions with RNA.

Metal-ion rescue experiments, first used to delineate metal-ion substrate interactions in protein enzyme active sites (Jaffe and Cohn 1978; Darby and Trayer 1983; Lee and O'Sullivan 1985; Pecoraro et al. 1985), have been expanded in ribozyme studies to provide a powerful strategy to identify individual metal ions, define their ligands, and assign their roles in catalysis (Piccirilli et al. 1993; Chen et al. 1997; Sontheimer et al. 1997; Weinstein et al. 1997; Shan et al. 1999; Yoshida et al. 1999; Hougland et al. 2005; Gordon et al. 2007; Forconi et al. 2008; Forconi and Herschlag 2009). The rescue approach relies on the differential preferences of Mg²⁺ and “softer” divalent metal ions like Mn²⁺, Cd²⁺, and Zn²⁺ to coordinate oxygen vs. sulfur or nitrogen ligands. As Mg²⁺ coordinates these ligands more weakly than it coordinates oxygen, sulfur or nitrogen substitution of an oxygen ligand can disrupt the Mg²⁺ interaction, leading to deleterious effects on activity. Inclusion of cations that interact more strongly than Mg²⁺ with sulfur or nitrogen often restores the disrupted interaction, resulting in suppression of the deleterious effect, or “rescue.” In practice, these experiments require quantitative analysis of the energetic effects of the substitution in the presence of Mg²⁺ and in the presence of the rescuing cation M²⁺ to provide the most reliable and extensive information about the number of bound metal ions and their interactions (Shan et al. 1999; Wang et al. 1999; Christian 2005; Forconi and Herschlag 2009).

In principle, this metal-ion rescue approach should also be a powerful biochemical tool for identifying metal-ion sites important for folding and conformational transitions in structured and signaling RNA molecules. However, a previous study of sulfur interferences in the independently folding P4–P6 RNA domain did not correspond well with structural data (Basu and Strobel 1999). Using a nondenaturing gel electrophoresis assay that probes the fraction of folded RNA by the differential migration of compact and extended forms, a population of P4–P6 RNAs containing a random distribution of R_P-phosphorothioates was screened to identify sites within the domain where R_P-phosphorothioate substitution interferes with global compaction (Cate et al. 1997; Basu and Strobel 1999). The identified sites included a subset of the ligands for metal ions observed crystallographically. Within this subset, Mn²⁺ and other soft metal ions suppressed phosphorothioate interference of global compaction at some, but not all, inner-sphere ligands (Basu and Strobel 1999). In addition, Mn²⁺ apparently suppressed phosphorothioate interference of folding by two outer-sphere ligands, a result that called into question the usual inference that metal-ion rescue reveals sites of inner-sphere coordination (Basu and Strobel 1999).

The apparent outer-sphere rescue could have major implications for published and future studies applying metal-ion rescue approaches to both RNA catalysis and RNA conformational transitions. Consequently, we sought to determine whether these results in the P4–P6 RNA identified a fundamental limitation of metal-ion rescue or a limitation of the study's methodology and analysis. Addressing these issues successfully could establish proof-of-concept for adapting the rescue approach to the investigation of metal ions in nonenzymatic, noncoding RNAs (ncRNAs), which likely encompass the vast majority of metal-ion binding sites within the cellular RNAome (Costa 2010).

We have synthesized variants of P4–P6 RNA bearing phosphorothioate modifications at specific locations and used hydroxyl radical footprinting to characterize the Mg²⁺- and Mn²⁺-induced folding transitions of these variants. Rescue is observed at positions identified as inner sphere in the P4–P6 crystal structures, but not at outer-sphere positions. These results clarify limitations of prior metal-ion rescue experiments and provide strong support for the power of the approach for investigating RNA folding and conformational changes, as well as a rigorous template for carrying out such studies.

RESULTS AND DISCUSSION

The “standard” thermodynamic framework for RNA folding and metal-ion rescue

For metal-ion rescue and other experiments that probe interactions, one would ideally make direct comparisons of free energies via thermodynamic cycles. A thermodynamic cycle for folding of unmodified and thio-modified P4–P6 (O and S, respectively) in the presence of Mg²⁺ or Mn²⁺ is shown in Scheme 1A, and the folding free energies are compared via the ΔΔG values defined in Scheme 1B. Rescue would correspond to the observation of amelioration of a folding defect—an increase in ΔG_fold (less favorable folding) for folding of the thio-substituted RNA in Mg²⁺ that is alleviated in the presence of a softer metal ion such as Mn²⁺ (Scheme 1B). In terms of the free energies defined in Scheme 1B, if folding is affected similarly in Mg²⁺ and Mn²⁺, no rescue occurs (ΔΔΔG_rescue = 0), whereas a positive value of ΔΔΔG_rescue would indicate rescue.

SCHEME 1A.

SCHEME 1B.

Evaluating metal-ion rescue quantitatively therefore requires estimating free energies of folding. RNA folding is typically followed as a function of metal-ion concentration, with different RNAs exhibiting different folding transitions. For each RNA, the experimentally observed “fraction folded” is typically fit to the Hill model:

Here, [M²⁺] is the metal-ion concentration, and [M²⁺]_1/2 and n are the transition midpoint and apparent Hill coefficient, respectively (Fang et al. 1999; Rook et al. 1999; Das et al. 2005b). This model corresponds to a free-energy difference between the unfolded and folded state of

where R is the gas constant and T is the temperature. Values for “fraction folded” are readily determined for [M²⁺] around the transition midpoints' [M²⁺]_1/2 values, but extrapolation of inferred folding free energies to higher and lower metal-ion concentrations is necessary to compare RNA variants with different stabilities. If the apparent Hill coefficient n remains constant between unmodified and thio-modified variants, the free-energy difference is directly related to the ratio of the folding midpoints and :

If folding of both the unmodified and thio-modified RNAs is measured in Mg²⁺ and Mn²⁺, and titrations in both metal ions happen to again yield the same Hill coefficient n, this framework gives

In this scenario, describing the metal-ion rescue results in terms of “fold rescue” is equivalent to discussing the free-energy differences ΔΔΔG, with a quantitative translation provided by Equation 4. Nevertheless, as discussed further below, two problems frequently complicate this description. First, fitted Hill coefficients often differ for destabilized variants or when different metal ions induce folding. Equations 3 and 4 are not applicable in these situations, and Equation 2 leads to free-energy estimates that depend on the choice of reference metal-ion concentration. Second, distinct, partially folded states are frequently observed at intermediate M²⁺ concentrations (see below, “Low salt comparisons”), so that a single midpoint and Hill coefficient are not sufficient to describe a folding titration. Instead, multiple folding transitions must be interrogated and characterized.

One approach for minimizing these problems and simplifying analyses is to saturate the ion atmosphere with a high concentration of monovalent cations, such that any associated divalent cations occupy specific metal-ion binding sites nearly exclusively, rather than the ion atmosphere (Bukhman and Draper 1997; Horton et al. 1998). This approach has been directly tested and validated with P4–P6 RNA, where structure mapping and direct ion counting assays showed a congruence of simple two-state folding with a Hill slope of 2 and binding of two Mg²⁺ ions (Scheme 2; Das et al. 2005b). We therefore first investigated P4–P6 folding and metal-ion rescue in a background of 2 M NaCl. We then turned to more standard conditions of low monovalent salt concentration, such as were used in the prior metal-ion rescue study (Basu and Strobel 1999). We first present results in terms of “fold rescue,” qualitatively comparing effects for substitution of inner- and outer-sphere ligands. To take into account observations of varying Hill coefficients (and later, additional intermediates), we developed general thermodynamic treatments that allow systematic estimation of free-energy differences for metal-ion-dependent folding events, and provide quantitative support for the specific conclusions herein.

SCHEME 2.

Mg²⁺- and Mn²⁺-induced folding of P4–P6 in 2 M NaCl

The P4–P6 domain (Murphy and Cech 1993) consists of two sets of coaxially stacked helices—P4–P5–P6 and P5a–P5b—joined by a hinge region (J5/5a) (Fig. 1A,B). Two interactions connect these helical stacks in the folded state: a tetraloop-receptor interaction between the loop of P5b and internal loops in P6 (J6a/b and J6b/a) and an A-minor interaction between an A-rich bulge within P5a and the P4 helix. We refer to these connections as “tertiary interactions” herein. In our studies we have used ΔC209 P4–P6 RNA. The deletion of C209 thermostabilizes the folded RNA; the normally extruded A210 forms a noncanonical base pair with G111 within the P4 helix (Juneau and Cech 1999), but the structure is otherwise unchanged from wild-type P4–P6 (Juneau et al. 2001). We refer to this construct as P4–P6 or unmodified P4–P6 throughout for simplicity, unless noted otherwise.

FIGURE 1.

Structure of ΔC209 P4–P6 RNA. (A) Secondary structure of ΔC209 P4–P6. Nucleotides protected from hydroxyl radical cleavage in the folded state are shown as black circles (tertiary protections) or as black squares (subdomain protections). (B) Crystal structure of ΔC209 P4–P6 (derived from PDB file 1HR2). Green spheres represent magnesium ions observed in the structure. (C) Detail of metal ion binding sites in ΔC209 P4–P6 relevant to this work. Within the A-rich bulge (left) and the trihelical junction (right), magnesium ions make several inner-sphere (yellow dashes) and outer-sphere (blue dashes) contacts to nucleobases and nonbridging phosphate oxygens. The specific phosphate oxygens modified in this work are labeled in yellow text.

Several Mg²⁺ ions are observed in the crystal structures of P4–P6 (Cate et al. 1996; Juneau et al. 2001). Within the A-rich bulge, two Mg²⁺ ions each make three inner-sphere contacts to nonbridging phosphate oxygen atoms contributed by A183, A184, A186, and A187 (Fig. 1C, left). In addition, the pro-R_P nonbridging oxygen of G188 lies within outer-sphere coordination distance of one of these metal ions. A third metal ion directly coordinates the N7 and O6 atoms of G163 and G164, respectively, and makes an apparent outer-sphere contact to the pro-R_P nonbridging oxygen of G163 (Fig. 1C, right). These metal ions bind within the P5abc subdomain, which can undergo divalent metal-ion-dependent folding in isolation (Murphy and Cech 1993; Wu and Tinoco 1998; Silverman et al. 1999; Zheng et al. 2001). Several other metal ions are observed in P4–P6 crystal structures, but are not addressed herein.

In 2 M NaCl, wild type P4–P6 acquires nearly all native long-range tertiary contacts (Takamoto et al. 2004; Das et al. 2005b). The remaining tertiary interactions—those within P5abc and between the A-rich bulge (residues 182–188) and the P4 helix—are in the vicinity of the inner-sphere metal ions and form only upon addition of divalent metal ions (Fig. 2). Initial experiments showed that ΔC209 P4–P6 forms the same tertiary interactions in 2 M NaCl, as reflected by protection from hydroxyl radical cleavage of nucleotides 152–153 (P5b tetraloop), 200–202 (J5/5a hinge), and 222–223 (tetraloop-receptor) (Fig. 1A; data not shown). This state is termed I_TL-TR herein. The addition of Mg²⁺ or Mn²⁺ induces protection of P5abc subdomain nucleotides 164, 176–177, 180–182, and 187, as well as nucleotides 211–212 (the A-rich bulge/P4 A-minor interaction), a transition we term I_TL-TR-to-F (Scheme 2; Fig. 2). In Mg²⁺, these protections appear with a midpoint of 0.24 ± 0.01 mM, whereas in Mn²⁺ the corresponding protections appear at a fourfold lower concentration (0.06 ± 0.01 mM), indicating significant Mn²⁺-specific stabilization (Fig. 3A; Table 1). Both in Mg²⁺ and Mn²⁺, transitions occur with a Hill coefficient of approximately two (e.g., 1.8 ± 0.2 for both Mg²⁺ and Mn²⁺ titrations monitored at nucleotide 164; Table 1; Supplemental Table S1), consistent with prior work that demonstrated binding of two metal ions to induce formation of the fully folded state in 2 M NaCl (Das et al. 2005b; Travers et al. 2007).

FIGURE 2.

Model for equilibrium Mg²⁺- or Mn²⁺-induced folding of unmodified P4–P6 in a background of 2 M NaCl, which promotes the formation of a subset of long-range tertiary interactions. Two Mg²⁺ or Mn²⁺ ions bind within the A-rich bulge to form the final folded structure. (Adapted from Das et al. 2005b with permission from American Chemical Society © 2005.)

FIGURE 3.

Equilibrium Mg²⁺- and Mn²⁺-induced folding of unmodified P4–P6 (A) and phosphorothioate-substituted variants (B–F) in a background of 2 M NaCl (with 50 mM Na-MOPS at pH 7). Mg²⁺ data points are shown as open symbols and Mn²⁺ data points are shown as closed symbols. Values of the fractional protection from hydroxyl radical cleavage derived from two to four independent experiments for subdomain nucleotides 164 (triangles), 176–177 (circles), 180–182 (squares), and 187 (diamonds) were averaged according to the procedure outlined in the Materials and Methods and Supplemental Material. The resulting values were fit to the Hill equation to yield curves whose midpoints and Hill coefficients reflect the appearance of protections over all of the protected nucleotides simultaneously. Data for nucleotide 187 are not included for certain variants due to nonspecific degradation of the phosphorothioate linkage, which interfered with the protection signal at this site.

TABLE 1.

Mg²⁺- and Mn²⁺-dependent folding of P4–P6 in 2 M NaCl

The observation that Mn²⁺ aids folding of P4–P6 even in the absence of phosphorothioate substitutions raised the possibility that prior findings of “anomalous” rescue of outer-sphere metal-ion interactions arose from this enhanced overall stability and not from a specific ability of Mn²⁺ to rescue outer sphere interactions. We therefore revisited these crystallographic outer-sphere ligands with a thermodynamically rigorous approach, focusing on the pro-R_P nonbridging oxygen atoms of G188 or G163. Prior analog-interference analyses have shown that sulfur substitution at either of these sites disrupts folding in 2 mM Mg²⁺ and low concentrations of monovalent cations, as measured by native gel electrophoresis (Cate et al. 1997; Basu and Strobel 1999). Inclusion of 0.3 mM Mn²⁺ with 1.7 mM Mg²⁺ induced an apparent rescue of folding at both of these sites (Basu and Strobel 1999). This rescue by inclusion of a small amount of Mn²⁺ was taken as evidence for an ability of Mn²⁺ to rescue outer-sphere metal-ion interactions, a highly unexpected result as noted above. The single “rescuing” metal-ion mixture was chosen because it did not suppress folding interference of a nonmetal site phosphorothioate (A139 R_P PS) and several 2′-deoxy interference sites. The mixture therefore was thought to minimize any nonspecific stabilizing effects of Mn²⁺. However, no attempt was made to measure nonspecific stabilizing effects of Mn²⁺ on the folding of unmodified P4–P6, the comparison species for which nonspecific Mn²⁺ effects would have meaning in the experiment. In effect, the prior study used A139 R_P PS and the 2′-deoxy interferences as substitutes for unmodified P4–P6, whose folding behavior in Mn²⁺ was never assessed.

Hydroxyl radical footprinting for G188 R_P PS gave protections at a sevenfold higher Mg²⁺ midpoint (1.66 ± 0.05 mM) relative to unmodified P4–P6 (Fig. 3B; Table 1). For G163 R_P PS, the protections appeared at a twofold higher Mg²⁺ midpoint (0.46 ± 0.06 mM) relative to unmodified P4–P6 (Fig. 3C; Table 1). These results indicate that sulfur substitution of either putative outer-sphere ligand destabilizes folding, and that increasing the Mg²⁺ concentration can overcome the effects of the substitution. Saturating Mg²⁺ induced the same protection patterns in G188 R_P PS and G163 R_P PS as for unmodified P4–P6, indicating that at the resolution of the assay, neither substitution changes the folded P4–P6 structure (data not shown).

The corresponding Mn²⁺ midpoints for the protections for G188 R_P PS and G163 R_P PS P4–P6 in 2 M NaCl were 0.32 ± 0.01 mM and 0.11 ± 0.01 mM, respectively (Fig. 3B,C; Table 1). The two modifications increased the midpoint for folding in Mn²⁺ by fivefold and twofold, respectively, nearly the same effects as in Mg²⁺ (ratios of sevenfold and twofold, respectively) (Fig. 3B,C; Table 1). Thus, there is no specific Mn²⁺ rescue of the outer-sphere phosphorothioates under these conditions. Instead, compared with Mg²⁺, Mn²⁺ has a nonspecific and similar stabilizing effect for the unmodified RNA and its variants with outer-sphere phosphorothioates. This conclusion is rendered more quantitative and precise by taking into account lowered apparent Hill coefficients for the mutants (see Table 1) within the thermodynamic framework described in the next section. In particular, Model 1 and Scheme 3 below provide a physical explanation for the change in the apparent Hill coefficient for the variants under these conditions.

SCHEME 3.

We next investigated whether Mn²⁺ could rescue folding of P4–P6 RNAs bearing phosphorothioate substitutions of inner-sphere Mg²⁺ ligands. We focused on A184, whose nonbridging oxygens make inner-sphere contacts to each of the A-rich bulge's metal ions in crystallographic models (Fig. 1C, left). Sulfur substitution of the pro-R_P oxygen of A184 interfered with P4–P6 folding using the aforementioned native gel assay (Cate et al. 1997; Basu and Strobel 1999). Neither Mn²⁺ nor other soft metal ions suppressed this folding interference under the single assay condition used previously (Basu and Strobel 1999). Sulfur substitution of the pro-S_P oxygen of A184 has not been reported. We constructed three P4–P6 RNAs with phosphorothioate substitutions of the pro-R_P, pro-S_P, or both (phosphorodithioate, PS₂) nonbridging oxygens of A184 (Supplemental Fig. S1).

In 2 M NaCl, the Mg²⁺- and Mn²⁺-induced protection patterns for A184 R_P PS, A184 S_P PS, and A184 PS₂ RNAs closely resemble those of unmodified P4–P6 (data not shown). This similarity suggests that the folded structures of these inner-sphere mutants closely resemble the structure of unmodified P4–P6 RNA. However, the phosphorothioate substitution of either or both of the nonbridging oxygens of A184 destabilized Mg²⁺-induced P4–P6 folding dramatically compared with unmodified P4–P6, with the Mg²⁺ midpoints associated with the appearance of protections occurring at 43-, 44-, and 63-fold higher Mg²⁺ concentrations than unmodified P4–P6 (Fig. 3A,D–F; Table 1). In Mn²⁺, the folding midpoints for the three inner-sphere constructs increase by 10-, five-, and ninefold relative to the unmodified RNA (Fig. 3A,D–F; Table 1). The ratios of these Mg²⁺ and Mn²⁺ midpoints give apparent fold rescue values of 4, 9, and 7, respectively, much greater than the values of near unity obtained for the outer-sphere substitutions (Fig. 4A).

FIGURE 4.

(A) Divalent metal-ion midpoint ratios of P4–P6 variants folded in either Mg²⁺ or in Mn²⁺ in a background of 2 M NaCl (OS: outer-sphere metal-ion ligand; IS: inner-sphere metal-ion ligand). (B) Calculated ΔΔG values (relative to unmodified P4–P6) associated with Mg²⁺-induced (left) and Mn²⁺-induced (right) folding of P4–P6 phosphorothioate variants with respect to unmodified P4–P6. The values were calculated according to a thermodynamic framework in which two site-bound metal ions, derived from solution or from the ion atmosphere, bind within the metal-ion core (Model 1, Supplemental Material). (C) Calculated ΔΔΔG values associated with Mg²⁺- vs. Mn²⁺-induced folding of phosphorothioate-subsituted P4–P6 variants in a background of 2 M NaCl.

These results suggest that Mn²⁺ specifically rescues folding of A184 R_P PS, A184 S_P PS, and A184 PS₂ relative to unmodified P4–P6, a result that is typically taken as evidence for inner-sphere coordination between Mn²⁺ and the phosphorothioate, and is consistent with the crystal structure of P4–P6 with Mg²⁺ and unsubstituted phosphates (Fig. 1; Cate et al. 1996; Juneau et al. 2001). These results further suggest that the two divalent metal ions that bind cooperatively with P4–P6 folding in the 2 M NaCl background are the metal ions that make inner-sphere interactions with the phosphoryl oxygen atoms of residue A184. Nevertheless, the apparent Hill coefficients are different for different variants (Table 1; Supplemental Fig. S2); thus, as noted above, “fold-rescue” values do not report directly on ΔΔΔG (i.e., Equations 3 and 4 are invalid) and, at best, give a qualitative summary of the results. To make basic conclusions about the energetic significance of these effects, a thermodynamic framework for computing the free-energy differences of metal-ion rescue is necessary. Such a framework is discussed next.

Thermodynamic models consistent with the rescue data

Estimating free-energy differences from the Mg²⁺ and Mn²⁺ titration data for the P4–P6 RNA and its variants requires extending the “standard” thermodynamic framework in Equations 1–4, which assumes that Hill coefficients for different RNAs are constant at different divalent metal-ion concentrations. In particular, the fitted Hill coefficients for the thio-modified variants are systematically lower than the values of ∼2 for the unmodified RNA (Table 1; Supplemental Table S1; Supplemental Fig. S2) and, indeed, decrease to values near 1 for the more destabilized variants (Table 1; Supplemental Table S1). We therefore sought simple extensions of the thermodynamic framework (Weiss 1997; Garcia et al. 2011) that model changing Hill slopes, while still permitting extraction of well-defined ΔΔG and ΔΔΔG values for evaluating metal-ion rescue. We explored two such models and found them to give numerical free-energy differences for the data herein that were indistinguishable within experimental error.

Model 1. Two site-bound metal ions derived from solution or from the atmosphere

The data herein and from previous work are well fit by a model in which I_TL-TR-to-F folding of unmodified P4–P6 is associated with the specific site binding of two divalent metal ions brought in from solution (Scheme 2). Due to the large excess of Na⁺ saturating the atmosphere (cf. solution concentrations of 2 M NaCl with submillimolar MgCl₂ or MnCl₂), the number of “atmospheric” divalent metal ions is assumed to be either negligible or similar in the I_TL-TR and F states and, indeed, the number of these background ions was verified to be one or less by direct ion counting (Das et al. 2005b). Scheme 2 does not show these diffusely bound divalent metal ions, sodium ions, or chloride ions, as they are negligible in the thermodynamic formulas describing the titration (Equations 1 and 2, with n = 2). However, for the destabilized thio-modified variants herein, higher concentrations of divalent ions are required to probe the I_TL-TR-to-F transition, raising the possibility that some atmospheric divalent metal ions might be present in I_TL-TR under these conditions and that these metal ions might enter the metal-ion core upon folding to F. Because we assumed that F contains a maximum of two site-bound ions, we considered the possibility that I_TL-TR preassociates with a single metal ion to form I_TL-TR•M²⁺ according to a separate equilibrium constant, K₁ (Scheme 3). We also considered a model with two metal ions preassociated with I_TL-TR to form an additional species (I_TL-TR•2M²⁺); a significant fraction of such a species would lead to an admixture of I_TL-TR and F at high divalent metal-ion concentrations. As we observe hydroxyl radical reactivities consistent with complete folding of all the RNAs herein, we do not include such a species. According to Scheme 3, folding to F•2M²⁺ occurs either from the I_TL-TR species via K₂, or from I_TL-TR•M²⁺.

To apply Scheme 3 to the metal-ion rescue experiments, we assumed that the second equilibrium, which involves folding from I_TL-TR•M²⁺ to form F•2M²⁺ via binding of a second divalent metal ion, was perturbed by the assayed thio modifications. We also assumed that the first equilibrium—binding of an “extra” metal ion to I_TL-TR with equilibrium constant K₁—was the same for all mutants, as would be expected if this metal ion was diffusely associated with the RNA and if all variants had similar structural ensembles in the metal core-free I_TL-TR state. Equations for fraction folded, free energies, and M²⁺-dependent Hill coefficients are given in the Supplemental Material. The model predicts that if a variant folds with a midpoint [M²⁺]_1/2, its apparent Hill coefficient is

This equation gives a Hill coefficient of 2 for stable variants, monotonically decreasing to 1 for destabilized variants, resulting from preassociation of a single metal ion with I_TL-TR at higher divalent metal-ion concentrations as prescribed by the model. This model accounts for the entire Mg²⁺ and Mn²⁺ data sets of unmodified P4–P6 and the five variants (Fig. 5, solid lines), with K₁ values of 0.26 ± 0.16 mM (Mg²⁺) and 0.13 ± 0.08 mM (Mn²⁺). This global fit then permits quantitative evaluation of the free-energy differences and metal-ion rescue for each P4–P6 thio substitution (Table 2).

FIGURE 5.

Curves describing the variation of the Hill coefficient (n_Hill) with the divalent metal-ion folding midpoint according to models 1 (solid lines; see equation 5) and 2 (dashed lines; see equation 6) for folding of P4–P6 variants in Mg²⁺ (A) and in Mn²⁺ (B) in a background of 2 M NaCl. The data points are the experimentally estimated values of n_Hill and [M²⁺]_1/2 for each individual RNA construct (Table 1).

TABLE 2.

Folding free energy differences of unmodified vs. phosphorothioate-substituted P4–P6 in 2 M NaCl

For the putative outer-sphere ligand modifications, the Mg²⁺-dependent folding transition I_TL-TR-to-F was destabilized by ΔΔG values of 1.49 ± 0.15 kcal/mol and 0.55 ± 0.13 kcal/mol for G188 R_P PS and G163 R_P PS, respectively (Table 2). When folded in Mn²⁺, the destabilization ΔΔG values for these two variants were 1.47 ± 0.23 kcal/mol and 0.58 ± 0.20 kcal/mol, indistinguishable from the Mg²⁺ values, given experimental error. The metal-ion rescue free-energy values ΔΔΔG were obtained by subtracting the Mg²⁺ and Mn²⁺ ΔΔG values and gave 0.02 ± 0.28 kcal/mol and −0.02 ± 0.24 kcal/mol for G188 R_P PS and G163 R_P PS, respectively. Both values were indistinguishable from zero (P = 0.5). This analysis indicates that substituting phosphorothioates for these crystallographically observed outer-sphere ligands in the P4–P6 system results in negligible metal-ion rescue.

For the putative inner-sphere ligand modifications A184 R_P PS, A184 S_P PS, and A184 PS₂, this framework gave ΔΔG values of 2.64 ± 0.19 kcal/mol, 2.66 ± 0.19 kcal/mol, and 2.87 ± 0.21 kcal/mol for A184 R_P PS, A184 S_P PS, and A184 PS₂, respectively, for the folding transition I_TL-TR-to-F induced by Mg²⁺ (Table 2). In contrast to the outer-sphere ligand experiments above, the ΔΔG values for folding in Mn²⁺ were lower: 1.93 ± 0.27, 1.45 ± 0.25, and 1.83 ± 0.26 kcal/mol. Thus, the results suggest metal-ion rescue, with rescue free-energy values of ΔΔΔG of 0.71 ± 0.33 kcal/mol, 1.22 ± 0.32 kcal/mol, and 1.04 ± 0.33 kcal/mol for A184 R_P PS, A184 S_P PS, and A184 PS₂, respectively. The results give strong evidence for nonzero metal-ion rescue for both A184 nonbridging oxygens (P = 0.02, P = 7 × 10⁻⁵, and P = 8 × 10⁻⁴ for A184 R_P PS, A184 S_P PS, and A184 PS₂, respectively), consistent with their observation as inner-sphere metal-ion ligands in the crystallographic analysis of P4–P6. Figure 4B shows the free energy differences (ΔΔG) for folding in both Mg²⁺ and Mn²⁺ in graphic form, while Figure 4C shows the rescue factors (ΔΔΔG) associated with each phosphorothioate construct.

Model 2. Linear expansion of the apparent Hill coefficient

Model 1 posits a functional form for how the free-energy difference ΔG for an RNA conformational transition varies with background divalent metal-ion concentration and assumes that each RNA modification perturbs this relationship by some constant ΔΔG. However, the model makes use of the special conditions of the experiment (high background sodium ion concentration) and assumptions specific to the P4–P6 metal core transition (I_TL-TR-to-F) to derive the relevant thermodynamic relations. To support the above analysis and to enable analysis under different monovalent ion background conditions (see below), we also explored a framework in which the expression for the free energy was assumed to take a more general form, and in which the data on the unmodified and thio-modified RNA were used to carry out a local mathematical expansion of the free energy as a function of metal-ion concentration. This framework has been applied previously to analyze the thermostabilization of a signal recognition particle RNA motif by substitutions predicted by the Rosetta design algorithm (Das et al. 2010).

Explicitly, we carried out the mathematical expansion on the apparent Hill coefficient function , which is proportional to the derivative of with respect to ln[M²⁺] (see Supplemental Material). This expansion does not require assumptions of high background monovalent ion concentrations or discreteness of the metal-ion binding sites, and thus is applicable as described below to conformational transitions that occur under low monovalent salt concentrations, which are more commonly encountered. We assumed that n varies smoothly with [M²⁺], as it does for computational models or direct measurements of ion atmospheres around fixed or changing RNA conformational ensembles (Manning 1977a,b; Anderson and Record 1990; Bai et al. 2007; Leipply and Draper 2010). We then confirmed that a simple linear expansion fits the observed apparent Hill coefficients herein (Fig. 5, dashed lines):

Here, n₀, the apparent Hill coefficient of the unmodified RNA near its folding midpoint, and the coefficient α_n are fitted parameters, and is the folding midpoint of the unmodified construct. Note that the values of the fitted parameters depend on the different folding midpoints and Hill coefficients of each individual RNA variant (the data points in Fig. 5). The model therefore makes no assumptions about the values of the Hill coefficients, which retain their experimentally determined values.

The corresponding equations for free-energy differences (derived in the Supplemental Material) gave ΔΔG values for all phosphorothioate modifications in both Mg²⁺ and Mn²⁺ that were within error of the values computed using Model 1 above (Table 2). In particular, the final metal-ion rescue values ΔΔΔG were calculated to be 0.37 ± 0.35 kcal/mol (P = 0.1), 0.06 ± 0.29 kcal/mol (P = 0.4), 1.34 ± 0.61 kcal/mol (P = 0.01), 1.72 ± 0.58 kcal/mol (P = 0.002), and 1.68 ± 0.66 kcal/mol (P = 0.005) for the modifications G188 R_P PS, G163 R_P PS, A184 R_P PS, A184 S_P PS, and A184 PS₂, respectively. These values are indistinguishable within experimental error from those obtained using Model 1 above, and further support the conclusion that metal-ion rescue is negligible for outer-sphere sites, but significant for inner-sphere ligands.

Testing rescue of P4–P6 folding in low background Na⁺ concentration

The above experiments carried out in 2 M NaCl support the conceptual underpinnings of metal-ion rescue experiments and indicate that, with proper thermodynamic comparisons, the approach can specifically identify site-bound metal ions within folded RNAs. To more directly compare these results with those obtained by Basu and Strobel (1999), to reveal another complexity in thermodynamic analyses, and to demonstrate how this complexity can be identified and addressed, we have carried out hydroxyl radical footprinting of the constructs described above in low background Na⁺ concentration. We first describe the results in the absence of phosphorothioate substitutions, as these results reveal a critical new feature in the P4–P6 folding pathway.

Mg²⁺- and Mn²⁺-induced folding of unmodified P4–P6 in low background Na⁺ concentration

For P4–P6 folding in low-background monovalent cation concentration, in contrast to the behavior in 2 M NaCl, the RNA initially lacks all of the tertiary interactions, so that extensive three-dimensional structure forms as Mg²⁺ or Mn²⁺ is added. Under these conditions, P4–P6 folds in Mg²⁺ via an apparent single cooperative transition (U-to-F) (Fig. 6A, left). All of the hydroxyl radical protections arise over a narrow Mg²⁺ concentration range, with an average midpoint and apparent Hill coefficient of 0.42 ± 0.01 mM and 4.2 ± 0.1, respectively (Fig. 6B, left; Table 3A). These values agree with those obtained for P4–P6 folding in low background monovalent cation concentration monitored by native gel electrophoresis ([Mg²⁺]_1/2 = 0.46 mM, n_Hill ∼ 4) (Juneau et al. 2001; Schwans et al. 2003), which monitors global molecular compaction.

FIGURE 6.

Mg²⁺- or Mn²⁺-induced folding of P4–P6 variants in a background of 20 mM Na⁺. Folding of unmodified P4–P6 in Mg²⁺ follows an apparent two-state transition modeled as U-to-F (A, left), in which both subdomain protections (B–G, left, open squares and dashed curves) and tertiary protections (B–G, left, open circles and solid curves) appear over approximately the same Mg²⁺ concentration range. In contrast, folding in Mn²⁺ follows the three-state transition U-to-I_P5abc-to-F (A, right), in which the subdomain protections (B–G, right, closed squares and dashed curves) appear at lower concentrations than the protections associated with formation of tertiary structure (B–G, right, closed circles and solid curves). Data points and curve fits were obtained for the subdomain and tertiary protections analogously to the procedure described for Figure 3.

TABLE 3A.

Mg²⁺-induced folding in 20 mM Na⁺

In contrast, P4–P6 folds via two transitions when Mn²⁺ is used as the divalent metal ion. A subset of the protections (152–153 and 200–202) appears at a slightly lower midpoint than in Mg²⁺, but protections of other residues (176–177 and 180–182) occur at a 10-fold lower Mn²⁺ concentration (Fig. 6B, right; Table 3B). The residues that are protected at the lower Mn²⁺ concentration (closed squares) correspond to a previously identified folding subdomain, P5abc, which can adopt structure independent of the overall P4–P6 structure (Murphy and Cech 1993; Wu and Tinoco 1998; Silverman et al. 1999; Zheng et al. 2001). Thus, the first protections that appear with increasing Mn²⁺ concentration correspond to a previously unseen U-to-I_P5abc transition. The protections that arise at higher Mn²⁺ concentrations correspond to formation of the tertiary interactions between the two sets of coaxially stacked helices in the folded P4–P6 structure (I_P5abc-to-F) (Fig. 6B, closed circles; Fig. 1A,B). This three-state, two-transition folding model shown in the right panel of Figure 6A is supported by results with a P4–P6 mutant in which J5/5a is base paired to prevent formation of the overall tertiary structure. This mutant gives P5abc folding at low Mn²⁺ concentration as observed within the context of unmodified P4–P6, but requires considerably higher Mg²⁺ concentrations than unmodified P4–P6 to support acquisition of the final tertiary structure (Murphy and Cech 1994; Wu and Tinoco 1998; data not shown). The prior work using gel electrophoretic measurements (Basu and Strobel 1999; Silverman and Cech 1999; JK Frederiksen and JH Piccirilli, unpubl.) did not detect the U-to-I_P5abc transition, possibly because the structural change is too subtle to be detected at the resolution of a gel mobility assay.

TABLE 3B.

Mn²⁺-induced folding in 20 mM Na⁺

Thus, in 20 mM Na⁺, acquisition of the final tertiary structure requires similar concentrations of Mg²⁺ and Mn²⁺, but the underlying process or folding mechanism is distinct, with Mn²⁺ promoting formation of a predominant folding intermediate that contains the two identified metal-ion binding sites (Fig. 6A, right). This difference has profound consequences. Whereas modifications that significantly weaken Mg²⁺ affinity for these binding sites would result in an increased [Mg²⁺]_1/2 for overall tertiary folding, the same modification in Mn²⁺ might have no influence on the [Mn²⁺]_1/2 for overall tertiary folding. In other words, a modification that increases the [Mn²⁺]_1/2 for formation of the subdomain (which contains the metal-ion binding sites) by ∼10-fold would still allow metal-ion binding and subdomain formation at a Mn²⁺ concentration below that required for formation of the overall tertiary structure (Fig. 6B, open vs. closed symbols) so that no shift in [Mn²⁺]_1/2 for the overall tertiary structure formation would be observed.

Because of the different folding pathways, any mutation or modification that uniformly perturbed the [M²⁺]_1/2 for the single folding transition in Mg²⁺ and for subdomain formation in Mn²⁺, such as modification of an outer-sphere ligand, would give apparent rescue in gel mobility measurements, which monitor only overall tertiary structure formation (i.e., the ratio [Mn²⁺]_1/2/[Mg²⁺]_1/2 would decrease for thio-substituted P4–P6 relative to wild-type P4–P6). Further, rescue of inner-sphere metal-ion ligand perturbations would not be apparent at only a single rescuing Mn²⁺ concentration, as was used in the prior study (Basu and Strobel 1999), if the added Mn²⁺ were insufficient to overcome a strong destabilizing effect from disruption of these metal-ion interactions.

Mg²⁺- and Mn²⁺-induced folding of phosphorothioate-substituted P4–P6 in 20 mM Na⁺

We first investigated folding of the two outer-sphere phosphorothioate-substituted P4–P6 RNAs, G188 R_P PS and G163 R_P PS, in the 20 mM Na⁺ background (Fig. 6C,D). With sufficient Mg²⁺, both constructs yield the same protection patterns as unmodified P4–P6 (left). Subdomain and tertiary protections for the phosphorothioates (open squares and open circles, respectively) appear over a narrow Mg²⁺ concentration range, analogous to the apparent single cooperative folding transition observed for unmodified P4–P6 in Mg²⁺, but at approximately twofold higher Mg²⁺ midpoints and with lower apparent Hill coefficients compared with unmodified P4–P6, reflecting decreased folding stability (Table 3A).

In Mn²⁺, these modified P4–P6 RNAs exhibit separate folding transitions for subdomain formation and overall tertiary structure formation, as is observed for the unmodified RNA (Fig. 6C,D, right). The [Mn²⁺]_1/2 value for formation of the subdomain is increased relative to unmodified P4–P6, but formation of the final tertiary structure is not significantly affected (Fig. 6C,D, right; Tables 3A, 3B). Even though the subdomain is destabilized, it still folds at concentrations lower than those required for overall tertiary folding, so the overall tertiary folding of these variants is unaffected (Table 3B) ([Mn²⁺]_1/2 = 0.3–0.4 mM). Thus, a destabilizing effect on formation of the subdomain manifests as an effect on acquisition of overall tertiary structure in Mg²⁺, as folding of the subdomain is coupled to overall folding in Mg²⁺, but has no effect on overall folding in Mn²⁺ because subdomain folding is complete at lower Mn²⁺ concentrations than overall tertiary structure formation (Fig. 6A). The destabilizing effect in Mg²⁺ and the absence of such an effect in Mn²⁺ corresponds to an apparent rescue of tertiary folding (Fig. 7A), but a rigorous appraisal requires accurately taking into account the additional intermediate observed in Mn²⁺. Complexities from this previously unknown intermediate account for the small, but significant apparent rescue of these outer-sphere constructs observed previously by native gel mobility assays that monitor primarily the global compaction associated with tertiary folding (Basu and Strobel 1999). Analogously, metal-ion rescue of catalysis can be misleading if the same reaction steps are not monitored throughout the comparisons (Shan and Herschlag 2000; Hougland et al. 2005).

FIGURE 7.

(A) Divalent metal-ion midpoint ratios of P4–P6 variants folded in either Mg²⁺ or Mn²⁺ in a background of 20 mM Na⁺, separated into subdomain and tertiary folding midpoints (OS: outer-sphere metal-ion ligand; IS: inner-sphere metal-ion ligand). (B) Calculated ΔΔG values (relative to unmodified P4–P6) associated with Mg²⁺-induced (left) and Mn²⁺-induced (right) folding of P4–P6 phosphorothioate variants with respect to unmodified P4–P6. The values were calculated according to a thermodynamic framework in which the Hill coefficient is assumed to vary linearly with the divalent metal-ion concentration (model 2, Supplemental Material). The values for Mn²⁺ are arranged in a stacked plot that reflects the sum of the contributions from the separate U-to-I_P5abc (subdomain) and I_P5abc-to-F (tertiary) folding transitions. (C) Calculated ΔΔΔG values associated with Mg²⁺- vs. Mn²⁺-induced folding of phosphorothioate-substituted P4–P6 variants in a background of 20 mM Na⁺.

The three inner-sphere constructs (A184 R_P PS, A184 S_P PS, A184 PS₂) also exhibit the same regions of protection as unmodified P4–P6 when folded in a background of 20 mM Na⁺ with sufficient divalent metal ion (data not shown). Like unmodified P4–P6, these constructs appear to fold in Mg²⁺ via a single transition, with subdomain and tertiary protections appearing over narrow concentration ranges (Fig. 6E–G, left; Table 3A). However, the associated midpoints occur at a six- to 10-fold higher Mg²⁺ concentration compared with unmodified P4–P6. As with the outer-sphere phosphorothioates, the increased folding midpoints reflect a destabilization of folding in Mg²⁺, and the effects are greater than the effects from the outer-sphere thio substitutions (Fig. 7A).

Also as for the outer-sphere metal ions, the folding pathway in Mn²⁺ is different from that in Mg²⁺, and this difference leads to apparent rescue. Figure 6E–G (right) shows the two transitions for folding in Mn²⁺. As with the outer-sphere substitutions, the inner-sphere thio substitutions have a destabilizing effect on subdomain formation (closed squares) (cf. Fig. 6, C and D), but this transition still occurs at lower Mn²⁺ concentrations than tertiary folding (Fig. 6E–G, right, closed squares vs. closed circles), resulting in the observed subdomain folding intermediate. As described above for the outer-sphere thio substitutions, the higher [Mg²⁺]_1/2 coupled with the absence of an effect on [Mn²⁺]_1/2 for overall tertiary folding leads to apparent rescue (Fig. 7A). This apparent rescue is larger than that observed for the outer-sphere residues, but only because the inner-sphere substitutions have a larger destabilizing effect than do the outer-sphere substitutions. Any substitution or mutation that has a larger effect on subdomain formation without an additional destabilizing effect on the overall tertiary structure formation (i.e., affecting U-to-I_P5abc without affecting the I_P5abc-to-F transition) would give a still larger apparent rescue. The apparent rescue arises from the inappropriate thermodynamic comparison of folding that occurs via two different pathways—in Mg²⁺, from U to F, and in Mn²⁺, from I_P5abc to F—and, therefore, need not have any relationship to rescue via stronger Mn²⁺–sulfur interactions.

As with the simpler 2 M NaCl data, quantitative evaluation of metal-ion rescue in lower monovalent salt concentration requires calculating and comparing free-energy differences induced by thio modification for identical states in both Mg²⁺ and Mn²⁺ (Scheme 1). To resolve the complexity associated with different transition pathways in Mg²⁺ and Mn²⁺, we calculated free-energy differences for the same overall process U-to-F. To resolve complexities associated with changing apparent Hill coefficients in modified RNAs, we analyzed the data in the general thermodynamic framework derived above based on global fits of the Hill coefficient across all variants (model 2; see Fig. 8 for the corresponding n_Hill vs. [M²⁺]_1/2 plots). The results of these analyses are summarized in Figure 7, B and C and Table 4, and demonstrate Mn²⁺ rescue of the inner- but not outer-sphere ligands.

FIGURE 8.

Linear fits to midpoint and Hill coefficient data according to model 2 (Equation 6) for folding of P4–P6 variants in Mg²⁺ (A,B) and in Mn²⁺ (C,D) in a background of 20 mM Na⁺. (Data from Table 3.)

TABLE 4.

Folding free energy differences of unmodified vs. phosphorothioate-substituted P4–P6 in 20 mM Na⁺

Figure 7B shows the overall folding free energies (between states U-to-F) for each construct in Mg²⁺ and Mn²⁺, taking into account the single-transition vs. two-transition thermodynamics in the two conditions, respectively. The values in Mg²⁺ (Fig. 7B, left; Table 4, Mg²⁺, U-to-F) are obtained from the single folding isotherms shown in Figure 6, B–G (left). The values in Mn²⁺ (Fig. 7B, right; Table 4, Mn²⁺) require summing the free energies from the two sets of folding isotherms (Fig. 6B–G, right), representing the U-to-I_P5abc step (closed squares and dashed lines) and the I_P5abc-to-F step (closed circles and solid lines) to yield the total free-energy difference for U-to-F. Figure 7C plots these differences in the folding free energies in Mg²⁺ vs. Mn²⁺ to allow rescue to be evaluated. As in the analysis of the simpler 2 M NaCl data above, similar values of ΔΔG in Mn²⁺ vs. Mg²⁺ were observed for the outer-sphere substitutions G188 R_P PS and G163 R_P PS, giving no evidence of metal-ion rescue. Further, smaller ΔΔG values in Mn²⁺ vs. Mg²⁺ for inner-sphere thio substitutions (A184 R_P PS, A184 S_P PS, A184 PS₂) provide thermodynamic evidence for rescue of inner-sphere metal-ion interactions in these experiments. The quantitative values for the metal rescue free-energy differences ΔΔΔG are 0.45 ± 0.58 (P = 0.2), −0.66 ± 0.63 (P = 0.9), 0.71 ± 0.67 (P = 0.1), 1.48 ± 0.65 (P = 0.01), and 1.34 ± 0.64 (P = 0.02) kcal mol⁻¹ for the modifications G188 R_P PS, G163 R_P PS, A184 S_P PS, A184 R_P PS, and A184 PS₂, respectively (Table 4). These values have larger errors than the results obtained in the simplifying 2 M NaCl conditions (Table 2), due to the summation of free energies over an additional transition in the Mn²⁺ titrations. For one of the constructs (A184 R_P PS), we cannot yet ascertain whether the measured rescue effect is greater than its associated error. Nevertheless, given these errors, the values in low monovalent salt and high monovalent salt background are in agreement, providing evidence for congruence between these biochemical metal-ion dissection experiments and the inner-sphere/outer-sphere nature of the RNA ligands from crystallography.

SUMMARY AND IMPLICATIONS

Metal-ion rescue: Successes and lingering concerns

Metal-ion rescue experiments have provided a powerful experimental approach for developing models of catalysis by ribozymes and for evaluating the functional relevance of structural models (Wang et al. 1999; Hougland et al. 2006; Nelson and Uhlenbeck 2006). Elucidation of the network of interactions mediated by metal ions and their ligands within ribozyme active sites has relied on the assumption that sites of rescue reveal ligands that occupy the inner coordination sphere of the metal ion. This interpretation follows logically from the known stronger coordination of rescuing metal ions to sulfur than oxygen. While the absence of rescue, a negative result, cannot be used to provide evidence against a direct metal-ion interaction, and while rescue can occur from recruitment of a new interaction not present in the wild-type/cognate pair, metal-ion rescue experiments have been used extensively with functional RNAs, and the results have been predictive of further functional and structural results (Wang et al. 1999; Adams et al. 2004; Guo et al. 2004; Golden et al. 2005; Hougland et al. 2006; Nelson and Uhlenbeck 2006). For example, quantitative metal-ion rescue studies provided evidence for a single catalytic metal ion that bridged positions 20 Å apart in hammerhead ribozyme X-ray structures (Wang et al. 1999), and subsequent X-ray structures of more active forms of the hammerhead provided strong support for this proposed conformational transition and metal-ion binding site (Martick and Scott 2006; Nelson and Uhlenbeck 2006). Extensive studies of group I ribozyme catalysis provided evidence for three active site metal ions and multiple ligand interactions, with all but one of these interactions supported by X-ray structural studies (Shan et al. 1999; Guo et al. 2004; Golden et al. 2005; Stahley and Strobel 2005; Hougland et al. 2006). In this case the structural evidence for two metal ions remains to be resolved with the functional evidence for three metal ions; it is possible that this case presents a rare example of metal-ion recruitment upon thio substitution (Hougland et al. 2006).

The observation from a previous nucleotide analog interference mapping (NAIM) analysis that Mn²⁺ appeared to suppress P4–P6 folding defects caused by sulfur substitution at outer-sphere ligands challenged the “standard” interpretation of metal-ion rescue experiments (Basu and Strobel 1999). It was possible that the combination of sulfur substitution with Mn²⁺ supplanted outer-sphere interactions between Mg²⁺ and the phosphate oxygen with inner-sphere interactions between Mn²⁺ and sulfur, engendered by the longer sulfur–phosphorus bond coupled with the longer sulfur-Mn²⁺ coordination distance. Alternatively, the apparent rescue might have reflected limitations associated with the native gel assay of folding, the single rescuing conditions used, or the thermodynamic assumptions underlying the analysis of the results.

We first used simplifying conditions, investigating P4–P6 folding and rescue in a high background concentration of monovalent cation. We previously showed that these conditions essentially saturate the ion atmosphere with monovalent cations, thereby simplifying divalent metal-ion-dependent folding such that the observed Hill slope of two corresponds precisely to the uptake of two metal ions upon folding (Das et al. 2005b). At the time, we presumed that these two metal ions corresponded to site-bound metal ions in the P4–P6 crystallographic model, likely the most highly coordinated metal ions within the core of the P5abc domain (Fig. 1C, left). The rescue results here provide strong and direct support for this hypothesis and further reveal the identity of these metal ions.

While the simplifying high monovalent concentration conditions have been powerful for P4–P6 and for some other systems, not all RNAs will be as well behaved under these conditions, and it is desirable to also investigate RNAs under more physiologic conditions that allow more normal interactions with proteins and other binding partners. We also wished to investigate rescue under low-salt conditions to provide a more direct comparison to and test of the prior study that had called into question the ability of rescue experiments to discriminate between inner- and outer-sphere structural metal ions.

We found that the thermodynamic folding pathway differed in Mg²⁺ vs. Mn²⁺ in this low monovalent salt concentration background. In Mn²⁺, we observed stabilization and accumulation of a folding intermediate corresponding to the P5abc subdomain of P4–P6 (I_P5abc), a structure that contains the two metal-ion binding sites identified above and has previously been shown to fold independently (Murphy and Cech 1993; Wu and Tinoco 1998; Silverman et al. 1999; Zheng et al. 2001). The presence of two different folding pathways subverts simple thermodynamic analysis of the rescue experiments and was a key element limiting the prior data interpretation. This limitation is akin to the potential to be misled in rescue experiments that probe catalysis when different reaction steps are followed or different rate-limiting steps occur for the constructs being compared (Shan and Herschlag 2000; Hougland et al. 2005). By obtaining residue-by-residue folding isotherms via hydroxyl radical footprinting (Fig. 6) and by developing an approach to compare folding free energies under common conditions, we were able to determine the thermodynamics for the overall folding process and to demonstrate rescue of inner-sphere, but not outer-sphere ligands. Our findings underscore the importance of developing explicit thermodynamic models and carrying out complete thermodynamic characterizations.

Our results show that metal-ion rescue experiments can be successfully applied to identify ligands of structural metal ions, analogous to the use of rescue experiments to identify catalytic interactions of metal ions. As an increasing number of RNAs with biological roles other than catalysis are being identified, tools to dissect folding and metal ion and other interactions within folded structures will be increasingly important. Recent advances that increase the throughput and precision in analysis of nucleic acid footprinting data (Das et al. 2005a, 2010; Merino et al. 2005; Yoon et al. 2011), as well as the conceptual approaches introduced herein, should greatly facilitate analysis of these RNAs.

MATERIALS AND METHODS

Cloning

The plasmid pUC19 ΔC209 P4–P6 was a gift from Kara Juneau and Tom Cech and encodes full-length ΔC209 P4–P6 with a 5′ T7 promoter sequence, cloned between the EcoRI and XbaI restriction sites of the pUC19 vector. The plasmid pUC19 102-178 HH was a gift of Scott K. Silverman and encodes nucleotides 102–178 of P4–P6, followed by a hammerhead ribozyme cleavage sequence, cloned between the EcoRI and HindIII restriction sites of pUC19 (Silverman and Cech 1999). Two additional plasmids, pUC19 175-261 ΔC209 and pUC19 200-261 ΔC209, encoding T7 transcripts 175–261 and 200–261, respectively, of ΔC209 P4–P6 were constructed via PCR from pUC19 ΔC209 P4–P6. The PCR products for these constructs were cloned into pUC19 between the EcoRI and XbaI restriction sites. The sequences of all constructs were verified by forward and reverse automated sequencing.

RNA preparation

Linearization of all plasmids was accomplished by overnight digestion with EarI, except for pUC19 102–178 HH, which was linearized by overnight digestion with HindIII. Typical transcription reactions contained 40 mM Tris-HCl (pH 8), 25 mM MgCl₂, 2 mM spermidine, 0.01% Triton X-100, 4 mM NTPs, 10 mM DTT, 50 μg/mL linearized plasmid, 2 units/mL inorganic pyrophosphatase, and 84.6 μg/mL T7 RNA polymerase (prepared in the Piccirilli lab by Joseph Olvera). Transcription reactions of pUC19 102-178 HH included 30 mM MgCl₂ to ensure sufficient hammerhead ribozyme cleavage (Silverman and Cech 1999). Transcription reactions that used pUC19 175–261 ΔC209 or pUC19 200–261 ΔC209 included 10 mM GMP to ensure the presence of a 5′ monophosphate. All transcriptions were carried out at 37°C for 2 h. Transcripts were purified via 8% denaturing PAGE and electroeluted into TE pH 7.5, extracted with 25:25:1 (v/v/v) phenol/chloroform/isoamyl alcohol, precipitated with ethanol, and stored at –20°C.

RNA nucleotides 102–153 of ΔC209 P4–P6 were obtained by digestion of full-length ΔC209 P4–P6 RNA with the 10–23 DNAzyme 5′-AAGGCCATCTCAAAGGGCTAGCTACAACGATTCCCCTGAGACTTG-3′, purchased from Integrated DNA Technologies (Santoro and Joyce 1997). About 16 nmol of ΔC209 P4–P6 RNA were combined with 25 nmol of DNAzyme, heated at 95°C for 3 min, and placed on ice for 1 min. Following the addition of 5x DNAzyme buffer (750 mM NaCl, 200 mM tris-HCl at pH 8), the mixture was incubated at 37°C for 30 min to allow hybridization. To initiate cleavage, MgCl₂ was added to a final concentration of 60 mM, and the reaction was allowed to proceed at 37°C for 5 h. The reaction was then treated with 25 units of RNase-free DNase (Promega) for 1 h at 37°C to remove the DNAzyme. Following phenol/chloroform/isoamyl alcohol extraction and ethanol precipitation, the desired 52-nt product was purified via 6% dPAGE.

As a result of 3′-end processing, the RNA molecules 102–178 and 102–153 contained 3′-terminal 2′,3′-cyclic phosphate groups. To remove these groups, the RNAs were treated with T4 polynucleotide kinase (T4 PNK; New England Biolabs) based on a protocol described previously (Silverman and Cech 1999). Briefly, ∼5–10 nmol of RNA were incubated with 50–100 units of T4 PNK in 1x kinase buffer (50 mM tris-HCl at pH 7.5, 10 mM MgCl₂, 5 mM DTT) at 37°C for 5 h. The RNAs were then extracted with phenol/chloroform/isoamyl alcohol and precipitated with ethanol.

Synthetic monophosphorothioate RNA oligonucleotides 179–199 A184 PS and 154–174 G163 PS of ΔC209 P4–P6 were purchased from Dharmacon, Inc. Phosphorothioate diastereomers were separated via anion exchange HPLC (Frederiksen and Piccirilli 2009). The RNA oligonucleotide 179–199 containing a phosphorodithioate linkage 5′ of nucleotide A184 was synthesized on a 1-μmol scale on an Expedite 8909 nucleic acid synthesizer. Except at position 183, this oligo was synthesized with standard RNA phosphoramidites. At position 183, an adenosine phosphorothioamidite was coupled that had been synthesized based on methods described previously (Petersen and Nielsen 1990). Following a 30-min coupling or double coupling step, the initial dichlorobenzyl phosphorothioate linkage was converted to a phosphorodithioate linkage by a 400-sec oxidation with 5% sulfur in 1:1 (v/v) pyridine/carbon disulfide or with 0.05 M solution of sulfurizing reagent II in 2:3 (v/v) pyridine/acetonitrile. After synthesis of the full-length oligo had been completed, the dichlorobenzyl protecting group was removed by treatment with 1:1:2 (v/v/v) thiophenol/triethylamine/dioxane at room temperature for 2 h. The solid support was then washed three times each with methanol and diethyl ether and allowed to dry. Base deprotection and cleavage from the solid support was achieved by treatment with 3:1 (v/v) ammonium hydroxide/ethanol for 2 h at 55°C. Following evaporation of the ethanolic ammonia solution, the 2′-silyl protecting groups were removed by treatment with a 6:3:4 (v/v/v) solution of N-methyl pyrrolidinone/triethylamine/triethylamine trihydrofluoride at 65°C for 1.5 h (Wincott et al. 1995). The deprotected phosphorodithioate oligonucleotide was then purified by anion-exchange HPLC and could be clearly distinguished from the R_P and S_P monothioate species, which elute earlier from the column than the dithioate. The mass of the phosphorodithioate oligo was confirmed by MALDI mass spectrometry (calculated for M-H: 6787.9, measured 6787.1).

Modified P4–P6 RNAs prepared by splint ligation

Schemes describing construction of phosphorothioate-modified RNAs are shown in Supplemental Figure S1. When possible, synthetic phosphorothioate oligonucleotides were ordered with 5′-phosphate groups (Dharmacon). The phosphorodithioate oligo was phosphorylated on its 5′ end by using T4 PNK and ATP. RNAs were joined by using standard DNA splint ligation methods (Moore and Sharp 1992). In a typical reaction, 5 nmol of each component (splint, 5′ RNA and 3′ RNA) were combined in TE pH 7.5 (10 mM tris-HCl at pH 7.5, 1 mM Na₂EDTA pH 8), heated to 90°C for 4 min, placed on ice for 1 min, and incubated at 37°C for 30 min. To this mixture were added 10x ligation buffer, 500 pmol T4 DNA ligase (prepared in the Piccirilli lab by Joseph Olvera) and water so that the final volume was as small as possible while still keeping the final concentration of glycerol below 5%. Beside RNA, splint, and T4 DNA ligase, the final ligation reaction contained 40 mM tris-HCl (pH 7.5), 10 mM MgCl₂, 10 mM DTT, 1 mM ATP, and 0.05 mg/mL bovine serum albumin. Ligation reactions were allowed to proceed for 4–5 h at 37°C, after which time the DNA splint was removed by treatment with RQ1 RNase-free DNase (Promega, Inc.). Following phenol/chloroform/isoamyl alcohol extraction and ethanol precipitation, the ligation products were purified via 6% dPAGE and subsequent electroelution into TE pH 7.5. Ligation junctions were confirmed by alkaline hydrolysis and T1 ladders, while iodine cleavage was used to confirm the positions of the phosphorothioate linkages.

Fe-EDTA footprinting

Full-length ΔC209 P4–P6 molecules were 5′-end labeled by treatment with [γ-³²P]ATP and T4 PNK, gel purified by 6% dPAGE, eluted passively overnight into TE pH 7.5 at 4°C, extracted with phenol/chloroform/isoamyl alcohol, and precipitated with ethanol. Prior to storage at –20°C, labeled RNAs were resuspended in water to an activity of ∼100,000 cpm/μL to minimize autoradiolysis.

A mix was prepared and distributed in 8-μL aliquots so that each footprinting reaction initially contained 1 μL of 5′-end labeled RNA, 1 μL of 500 mM Na-MOPS (pH 7), and 6 μL of water. Reactions carried out in the presence of 2 M NaCl contained 4 μL of 5 M NaCl and 2 μL of water. Following the addition of 1 μL of a 10x solution of MgCl₂ or MnCl₂, the RNA was allowed to fold for ∼15 min at room temperature.

A 10x footprinting reagent was prepared separately that contained 1 mM Fe(NH₄)₂(SO₄)₂, 1.25 mM EDTA (pH 8), and 60 mM sodium ascorbate. To initiate the footprinting reaction, 1 μL of this reagent was added to each aliquot to a final volume of 10 μL so that each reaction contained labeled RNA, 50 mM Na-MOPS (pH 7), 2 M NaCl where appropriate, varying concentrations of MgCl₂ or MnCl₂, 0.1 mM Fe(NH₄)₂(SO₄)₂, 0.125 mM EDTA (pH 8), and 6 mM sodium ascorbate. The reaction was allowed to proceed for 30 min at room temperature, and was quenched by the addition of 5 μL of a thiourea stop solution (9 M urea, 300 mM thiourea, 0.1% each xylene cyanol and bromophenol blue).

The footprinting reactions (7 μL of each) were fractionated on 8%–8.5% denaturing gels. The gels were dried, exposed to storage phosphor screens, and scanned on a Typhoon Trio imager (GE Healthcare). Gels were quantified using the SAFA program (Das et al. 2005a). Importantly, for residues that become protected only upon addition of divalent metal ions, no significant difference in cleavage intensity was found between samples folded in 2 M NaCl alone and samples folded in 20 mM Na⁺ alone. This result argues against a significant effect of contaminating divalent metal ions in the 2 M NaCl samples.

Initial data analysis and curve fitting

Regions of protection from hydroxyl radical cleavage were identified as follows. Footprinting in 2 M NaCl yielded clear divalent metal-ion-dependent protections at nucleotides 164, 176–177, 180–182, and 187, corresponding to folding of the P5abc subdomain. In 20 mM Na⁺, protections at nucleotides 152–153 and 200–202, and at 176–177 and 180–182, were chosen to reflect tertiary and subdomain folding, respectively. At each nucleotide position, each individual footprinting experiment generated a set of intensities whose values decreased in proportion to the extent of protection. The intensity data were fit initially to a four-parameter Hill equation as described in the Supplemental Material (Equation S1) and were subsequently normalized based on these fits. Data for protected regions spanning multiple nucleotides (e.g., 152–153 or 180–182) were averaged for each experiment. All of the data corresponding to a given folding transition derived from two to four independent experiments per RNA construct were then plotted on the same graph. For example, each plot in Figure 3 contains all data associated with nucleotides 164, 176–177, 180–182, and 187 (when available) obtained from each footprinting experiment conducted in 2 M NaCl. We then took the average of all of the protection data at each divalent metal-ion concentration and fit the resulting points to the Hill equation to produce the final curves shown in Figures 3 and 6. The midpoints and Hill coefficients derived from these curves were then used to determine ΔΔG and ΔΔΔG values for the constructs according to procedures outlined in the Supplemental Material. Finally, an independent analysis based on a rigorous likelihood formalism was carried out and gave indistinguishable results for the thermodynamic parameters and their errors (see Supplemental Material, Supplemental Figs. S3–S5, and Supplemental Tables S3, S4).

SUPPLEMENTAL MATERIAL

Supplemental material is available for this article.