Functional Analysis of Hairpin Ribozyme Active Site Architecture*

The hairpin ribozyme is a small catalytic motif found in plant satellite RNAs where it catalyzes a reversible self-cleavage reaction during processing of replication intermediates. Crystallographic studies of hairpin ribozymes have provided high resolution views of the RNA functional groups that comprise the active site and stimulated biochemical studies that probed the contributions of nucleobase functional groups to catalytic chemistry. The dramatic loss of activity that results from perturbation of active site architecture points to the importance of positioning and orientation in catalytic rate acceleration. The current study focuses on the network of noncovalent interactions that align nucleophilic and leaving group oxygens in the orientation required for the SN2-type reaction mechanism and orient the active site nucleobases near the reactive phosphate to facilitate catalytic chemistry. Nucleotide modifications that alter or eliminate individual hydrogen bonding partners had different effects on the activation barrier to catalysis, the stability of ribozyme complexes in the ground state, and the internal equilibrium between cleavage and ligation of bound products. Furthermore, substitution of hydrogen bond donors and acceptors with seemingly equivalent pairs sometimes had very different functional consequences. These biochemical analyses augment high resolution structural information to provide insights into the functional significance of active site architecture.

The hairpin ribozyme is a small catalytic motif found in plant satellite RNAs where it catalyzes a reversible self-cleavage reaction during processing of replication intermediates. Crystallographic studies of hairpin ribozymes have provided high resolution views of the RNA functional groups that comprise the active site and stimulated biochemical studies that probed the contributions of nucleobase functional groups to catalytic chemistry. The dramatic loss of activity that results from perturbation of active site architecture points to the importance of positioning and orientation in catalytic rate acceleration. The current study focuses on the network of noncovalent interactions that align nucleophilic and leaving group oxygens in the orientation required for the S N 2-type reaction mechanism and orient the active site nucleobases near the reactive phosphate to facilitate catalytic chemistry. Nucleotide modifications that alter or eliminate individual hydrogen bonding partners had different effects on the activation barrier to catalysis, the stability of ribozyme complexes in the ground state, and the internal equilibrium between cleavage and ligation of bound products. Furthermore, substitution of hydrogen bond donors and acceptors with seemingly equivalent pairs sometimes had very different functional consequences. These biochemical analyses augment high resolution structural information to provide insights into the functional significance of active site architecture.
The well-characterized structure of the hairpin ribozyme offers a valuable framework for investigating the contributions of individual active site interactions to the activation barrier to catalysis and to the stability of ribozyme complexes in the ground state. The hairpin ribozyme catalyzes a reversible selfcleavage reaction in which nucleophilic attack of a ribose 2Ј-hydroxyl on an adjacent phosphorus proceeds through a trigonal bipyramidal transition state that leads to the formation of 2Ј,3Јcyclic phosphate and 5Ј-hydroxyl termini (1). High resolution crystal structures have been solved for hairpin ribozymes in complexes with a noncleavable substrate analog, cleavage prod-ucts and a vanadate mimic of the trigonal bipyramidal transition state, making this ribozyme the subject of more detailed structural analyses than virtually any other catalytic RNA (2-6) (Fig. 1).
Hairpin ribozymes have two essential helix-loop-helix domains, A and B, that associate to form the active site (7). High resolution structures reveal a network of stacking and hydrogen-bonding interactions within the active site that align the reactive phosphate in the appropriate orientation for an S N 2type nucleophilic attack mechanism and orient nucleotide base functional groups near the reactive phosphate to facilitate catalytic chemistry (Fig. 1). Gϩ1 is the conserved nucleotide on the 3Ј-side of the reactive phosphodiester in loop A (Fig. 2). Tertiary interactions between Gϩ1 and nucleotides in loop B define the architecture of the active site (2,8). Guanine at the ϩ1 position has no direct contact with the reactive phosphodiester but loss of this nucleobase virtually eliminates catalytic activity, highlighting the significant contribution of active site architecture to catalytic rate enhancement (8 -12). Positioning and orientation also play critical roles in the internal equilibrium between cleavage and ligation of bound products. Interdomain junction modifications and changes in reaction conditions that destabilize ribozyme tertiary structures shift the internal equilibrium toward cleavage (13)(14)(15)(16). The relationship between tertiary structure stability and the internal equilibrium is consistent with the idea that ligation requires a rigid ribozyme structure to align the 2Ј,3Ј-cyclic phosphate and 5Ј-hydroxyl termini, while these reactive groups are fixed by the diester linkage in a cleavage substrate.
Biochemical studies have shown that interactions that seem to be equivalent in hairpin ribozyme crystal structures can make very different energetic contributions. For example, the N6 exocyclic amines of A9 and A38 both appear to donate hydrogen bonds to the same proR P nonbridging oxygen of the reactive phosphate (3), but substitution of adenine with purine at position 38 inhibits cleavage activity more than 10 3 fold (17) while complete elimination of adenine at position 9 through an abasic substitution reduces cleavage activity just 4-fold (18). The dramatic difference in the functional significance of two seemingly identical interactions illustrates the need for direct analyses of structure-function relationships that might be inferred from crystal structures.
Minimal hairpin ribozymes, in which the A and B domains are joined by a two-way helical junction, partition almost equally between an inactive, extended conformation and a functional structure that juxtaposes loops A and B to create the active site (19). Extensive structure-function studies of minimal hairpin ribozymes have been reported previously (1). However, it is now understood that most modifications that disable minimal ribozymes do so by shifting the conformational equilibrium further toward the inactive, extended structure (19 -21). In viral satellite RNAs, the A and B domains comprise two arms of a four-way helical junction (7,22). Restoring this natural junction stabilizes the docked structure and ribozymes with a four-way junction retain activity despite unfavorable reaction conditions and loss of individual tertiary interactions (14,16,19,21,23). This discovery makes it possible to examine the effects of nucleotide modifications in the context of the native folded structure. We report a systematic analysis of the active site architecture of a four-way junction ribozyme aimed at comparing the energetic consequences of functional group modifications on the stability of ribozyme complexes in the ground state, the activation barrier to catalysis, and the internal equilibrium between cleavage and ligation of bound products.

EXPERIMENTAL PROCEDURES
RNA Preparation-RNAs were prepared by T7 RNA polymerase transcription of linearized plasmid templates or by chemical synthesis ( Fig. 2 and Table 1) (Dharmacon), as described previously (12,14). Ligated ribozyme LR43 3 is a selfcleaving ribozyme that assembles in the context of a four-way helical junction and has three base pairs in the intermolecular helix, H1, that forms between the 5Ј and 3Ј cleavage product RNAs ( Fig. 2A). 5ЈR4 RNA was obtained from LR43 selfcleavage during bacteriophage T7 RNA polymerase transcription of the pTLR43 template. 5ЈR4 was combined with 3ЈP3 RNAs obtained through chemical synthesis to measure cleavage kinetics through ligation-chase assays, as described below. LR46 and LR47 variants are four-way junction ribozymes with six or seven base pairs in H1, respectively, which were produced by ligation of 5ЈR4, the 5Ј product of LR43 self-cleavage, and 3ЈP6 or 3ЈP7 RNAs obtained through chemical synthesis (Fig. 2B). LR46 and LR47 variants were used to measure ligation kinetics and equilibrium dissociation constants for 5ЈR4⅐3ЈP complexes. The pTLR43Gϩ1A,C25U plasmid template encodes the LR43G؉1A,C25U self-cleaving ribozyme variant, and was derived from pTLR43 by using QuikChange mutagenesis to replace the guanine at position ϩ1 with adenine and the cytosine at position 25 with uridine. Self-cleavage of the FIGURE 1. Network of tertiary interactions formed with G؉1. A, three-dimensional structure of the Gϩ1 binding pocket in a ribozyme complex with a vanadate mimic of the transition state prepared from the coordinates described by Rupert et al. (3) (PDB entry 1M5O) with the vanadate indicated in yellow. B, three-dimensional structure of the Gϩ1 binding pocket in a ribozyme complex with cleavage product RNA prepared from the coordinates described by Rupert et al. (3) (PDB entry 1M5V) with the 2Ј,3Ј-cyclic phosphate terminus of the 5Ј cleavage product instead of the vanadate transition state mimic. To help distinguish loop A and loop B residues, Gϩ1 and A-1 nucleotides in loop A are colored light blue and carbon atoms of C25, G36, and A38 nucleotides in loop B are colored green. Tertiary interactions between the essential Loop A and B domains of the hairpin ribozyme define hairpin ribozyme active site architecture and fix the reacting groups in the in-line orientation appropriate for the S N 2-type reaction mechanism. The interdomain interface is created by the extrusion of the Gϩ1 nucleotide from loop A into a binding pocket in loop B. C, diagram of interdomain hydrogen-bonding interactions formed with the Gϩ1 nucleotide. Modifications of nucleotides that donate or accept interdomain hydrogen bonds were designed to probe their functional significance.
Ribozymes with modifications at positions 36 and 38 were prepared through ligation of synthetic RNAs (Table 1 and Fig. 2C). First, LR43⌬A⌬B(5ЈGA 3 )Cϩ6U and B were phosphorylated in reactions with 1 nmol of oligonucleotide, 2 nmol of ATP, and 5 units of T4 polynucleotide kinase in a final volume of 5 l. After a 30-min incubation at 37°C, the kinase was inactivated at 90°C for 2 min. Next, stoichiometric amounts of LR43⌬A⌬B(5ЈGA 3 )Cϩ6U, B and AB(3ЈGA 3 )G4A RNAs were annealed in T4 polynucleotide kinase buffer (NEB) by heating to 85°C for 1 min and cooling to 30°C over 4 min. Incubation at 30°C for an additional 20 -40 min led to nearly complete cleavage of LR43⌬A⌬B(5ЈGA 3 )Cϩ6U for oligonucleotides with noninhibitory modifications of A38 and G36, producing a 2Ј,3Ј-cyclic phosphate terminus. Ribozymes with inhibitory modifications (A38dX) were induced to cleave by adding 20 mM isocytosine to the reaction and extending incubation times to 4 h (17,18). Covalent joining of ribozyme fragments was accomplished by adding 1 mM ATP and 20 units T4 RNA ligase and incubating overnight at 17°C, followed by heat inactivation of the ligase at 90°C for 2 min. Full-length ribozyme products were fractionated by denaturing gel electrophoresis, eluted, and prepared as sodium salts using DEAE ion exchange chromatography as described previously (14) with a typical yield of 120 pmol of ribozyme per 1 nmol of each of the three oligonucleotide components. Chemical synthesis of oligonucleotides that correspond to the 3Ј product of ribozyme cleavage was used to introduce modifications at the Gϩ1 position. Abasic substitutions were introduced as deoxynucleotides because abasic ribonucleotides are not yet available commercially.
3Ј Product RNAs were labeled at the 3Ј terminus by 32 pCp ligation (24). Briefly, 75 pmol of cytosine monophosphate (Cp) was combined with 25 pmol of [␥-32 P]ATP (5,000 Ci/mmol) and 5 units of T4 polynucleotide kinase for 1 h at 37°C to produce 32 pCp. Following heat inactivation of the kinase, the mixture was added to a solution containing 12.5 pmol of 3Ј product RNA, 10% dimethyl sulfoxide, 1 mM ATP, polynucleotide kinase buffer, and 2 units of T4 RNA ligase and incubated overnight at 15°C. Products were separated by denaturing polyacrylamide gel electrophoresis, eluted, then desalted by NAP column chromatography (Amersham Biosciences) and reduced to a volume of ϳ40 l by lyophilization.
Measurement of Equilibrium Dissociation Constants-Equilibrium dissociation constants for 5ЈR⅐3ЈP complexes, K d 5ЈR⅐3ЈP , and values for the internal equilibrium between cleavage and ligation of bound products, K eq int , for 5ЈR⅐3ЈP complexes with rapid ligation kinetics were determined from the RNA concentration dependence of observed ligation rates as described (13,14). Briefly, maximum ligation extents, FL max , were determined FIGURE 2. Hairpin ribozyme variants. The natural form of the hairpin ribozyme consists of six helical regions, H1 through H6, and two unpaired loops, A and B. The arrow marks the reactive phosphodiester in loop A. The two essential H1-loop A-H2 and H3-loop B-H4 elements associate noncoaxially within the folded tertiary structure to create the active site. A, ribozyme variant, LR43, is a self-cleaving ribozyme that assembles in the context of a four-way helical junction and has three base pairs in the intermolecular H1 helix that forms between 5ЈR4, the 5Ј product of self-cleavage, and 3ЈP3, the 3Ј product of LR43 self-cleavage. LR43 was used to measure effects of modifications on self-cleavage kinetics. B, LR46 is a ribozyme variant with six base pairs in the intermolecular H1 helix that was used to measure effects of modifications on ligation kinetics and 5ЈR⅐3ЈP complex stability. C, R4⅐S44 is a ribozyme-substrate complex with four base pairs in H1 that was used to measure effects of modifications on intermolecular cleavage kinetics for Gϩ1A and Gϩ1Pu variants that retained too little ligation activity to support ligation-chase assays. D, scheme for preparation of ribozyme variants with modifications at position 36 and 38 through ligation of three oligonucleotides.
from reactions with varying concentrations of 5ЈR and 0.1-0.25 nM of 32 pCp-labeled 3ЈP6 or 3ЈP7 at 15°C in 10 mM MgCl 2 , 50 mM NaHepes pH 7.5, 0.1 mM EDTA by computing the fit to Equation 1, where FLR obs is the fraction of ligated ribozyme at the end of a reaction, F active is the fraction of active ribozyme complex that is correctly folded and has the necessary 2Ј,3Ј-cyclic phosphate terminus, and K d,app 5ЈR⅐3ЈP is the apparent equilibrium dissociation constant for the 5ЈR⅐3ЈP complex. FLR max is the maximum theoretical fraction of ligated ribozyme calculated from the equilibrium between cleavage and ligation rate constants, K eq int ϭ k lig /k cleav , according to FLR max ϭ K eq int (K eq int ϩ 1) ϭ k lig /(k lig ϩ k cleav ). Within the same set of experiments, F active ϫ FLR max remains constant so that Equation 1 reduces to Equation 2.
values obtained from the cleavage and ligation rate constants that were measured independently, as described below. For ribozyme variants with slow ligation kinetics, equilibrium dissociation constants for 5ЈR⅐3ЈP complexes were determined directly from the concentration dependence of complex formation using gel shift assays, as described previously (25). Briefly, 0.1-0.25 nM of 32 pCp-labeled 3Ј product RNA was incubated with varying concentrations of 5ЈR for 2-4 h at 15°C in 10 mM MgCl 2 , 50 mM NaHepes pH 7.5, 0.1 mM EDTA to allow complex formation to reach equilibrium. Bound and unbound 32 P-labeled 3ЈP were fractionated by electrophoresis in nondenaturing polyacrylamide gels with 15% acrylamide (19:1, acrylamide:bisacrylamide), 10 mM magnesium acetate, 50 mM Tris acetate, pH 7.5 for 12 h at 35 mA at a constant temperature of 5°C in a vertical slab gel electrophoresis unit (Hoefer SE600). Equilibrium dissociation constants were calculated from the fit to Equation Cleavage Kinetics Assays-Cleavage rate constants were determined at 15°C in 10 mM MgCl 2 , 50 mM NaHepes pH 7.5, 0.1 mM EDTA, the same conditions used for equilibrium binding assays, from intermolecular cleavage assays with saturating concentrations of ribozyme, as described (18,25,26). Ligationchase assays were used to monitor self-cleavage rates directly within a fully functional ribozyme created by self-ligation, as described (18). Briefly, 0.1-0.25 nM of 32 pCp-labeled 3ЈP3 was combined with 5Ј ribozyme RNA at a concentration approximately equal to the apparent equilibrium dissociation constant for product binding and allowed to undergo ligation to yield a final fraction of ligated ribozyme of at least 20%. To maximize ligation efficiency in reactions with 5ЈRA38dX, exogenous isocytosine was included at a concentration of 1 mM. Ligation reactions were then diluted 40-and 80-fold into reaction buffer to initiate cleavage. Aliquots were quenched after various times, fractionated on denaturing gels and quantified by radioanalytic imaging (Molecular Dynamics). Cleavage rate constants were computed from fits to the equation ). Each reported value represents the mean of two or more independent measurements; reported errors are standard deviations.
3Ј Product RNA sequences were chosen to form an intermolecular H1 helix with a stability optimized to ensure that prod- Nucleotide modifications that alter active site functional groups are indicated in bold font. Sequence changes that facilitate measurement of kinetics and equilibrium parameters are shown in regular font.
Functional Analysis of Hairpin Ribozyme Architecture MAY 4, 2007 • VOLUME 282 • NUMBER 18 uct binding affinity was high enough to support adequate ligation extents at experimentally accessible RNA concentrations during a ligation pulse but low enough to ensure that cleavage rates observed following a 40-fold dilution were not complicated by slow product dissociation or product re-binding and ligation. Confirmation that similar cleavage rates were observed following 40-and 80-fold dilutions ensured that no significant product rebinding and ligation occurred during the chase phase. For unmodified RNAs and modified RNAs that form stable complexes (dGϩ1, mGϩ1, dA38), we used a 3Ј product RNA, 3ЈP3(ϩ2), that forms an H1 helix with three base pairs. 3ЈP3(ϩ2) contains a 2-nucleotide 3Ј overhang so that 3Ј terminal labeling with 32 pCp does not affect H1 stability. For modified RNAs that form less stable complexes (Gϩ1I, Gϩ1(2AP), A38dX) pulse-chase experiments were carried out using 3ЈP3 RNA, which forms an H1 helix with four base pairs after 3Ј end-labeling with 32 pCp. In each case, results of experiments with different 3ЈP RNAs that have different binding affinities were compared to confirm that observed cleavage rates monitored the cleavage step and not slow product dissociation or re-binding and ligation of bound product. Ligation Kinetics Assays-Ligation kinetics were measured under the same conditions used for cleavage and equilibrium binding assays. 5ЈR, at concentrations between 500 nM and 4 M, first was heated to 85°C for 1Ј in 50 mM NaHepes pH 7.5, 0.1 mM EDTA then slowly cooled to 15°C in the presence of 10 mM MgCl 2 . 32 pCp-labeled 3ЈP6 (or 3ЈP7) was added to a final concentration of 0.1-0.25 nM to initiate ligation. Samples were removed at intervals, quenched, fractionated, and quantified as described above. Observed ligation rates were computed from fits to the equation Comparisons of observed ligation rates and extents in reactions with at least a 2-fold difference in 5ЈR concentrations confirmed that ligation rates were not limited by slow 5ЈR⅐3ЈP complex formation. Ligation rate constants were calculated from the difference between the cleavage rate constant obtained from pulse-chase experiments and the mean observed ligation rate, which reflects the rate of approach to equilibrium between cleavage and ligation and is the sum of ligation and cleavage rate constants, as described previously (13).

RESULTS AND DISCUSSION
The functional consequences of active site modifications were determined from changes in cleavage and ligation rate constants and equilibrium dissociation constants for the complex formed between 5Ј ribozyme and 3Ј product RNAs, 5ЈR⅐3ЈP, as illustrated by the reaction scheme shown in Fig. 3. Kinetic and equilibrium parameters were measured at pH 7.5 in reactions with 10 mM MgCl 2 . These conditions are similar to those previously used for analyses of similar hairpin ribozymes (12,14,17,18), except that a temperature of 15°C was chosen, rather than the standard temperature of 25°C, because the lower temperature facilitated comparisons of ribozyme variants with destabilizing modifications. A ligation-chase protocol was used to measure cleavage rates to ensure that cleavage assays monitored the catalytic activity within properly folded 5ЈR⅐3ЈP complexes and were not complicated by slow refolding of misfolded RNAs (13). Changes in cleavage and ligation rate constants reflect the effects of modifications on the magnitude of the activation barrier to cleavage or ligation, ⌬⌬G ‡ , while changes in equilibrium dissociation constants reflect effects of modifications on the ground state stability of 5ЈR⅐3ЈP complexes, or ⌬⌬G 5ЈR⅐3ЈP . It is important to recognize that observed changes in transition state and ground state energies that result from functional group modifications are not a direct measure of the energetic contribution of the bonds formed between specific functional groups because changes in interactions with solvent also contribute to free energy changes (27).
Functional Consequences of Gϩ1 Modifications-The active site assembles through tertiary interactions between loops A and B (2, 20, 22, 28) (Fig. 1). Crystallographic studies revealed very similar active site structures for a ribozyme complex with vanadate, a mimic of the trigonal bipyramidal transition state in which phosphorus associates with five oxygen atoms, and a ribozyme complex with product RNAs that have 5Ј-hydroxyl and 2Ј,3Ј-cyclic phosphate termini (3) (Fig. 1, A and B). In both vanadate and product complexes, Gϩ1 extrudes from loop A in an unusual syn conformation and stacks between the A38 and A26 nucleobases in loop B. Gϩ1 forms additional hydrogen bonding interactions with loop B nucleotides that include Watson-Crick pairing between Gϩ1 and C25 (8) and hydrogen bonding between N7 of Gϩ1 and the 2Ј-OH of A38 and between the 2Ј-hydroxyl of Gϩ1 and the exocyclic amine of G36. The absence of significant differences between the two structures in the distances between Gϩ1 functional groups and putative hydrogen bonding partners in loop B suggests that Gϩ1 forms virtually the same hydrogen bonding and stacking interactions in the ground state as in the transition state.
Gϩ1 is the 5Ј-terminal nucleotide of the 3Ј cleavage product RNA (3ЈP) so changes in ground state interactions with Gϩ1 will be reflected in equilibrium dissociation constants for 3Ј product RNA binding (14). The length and sequence of the intermolecular H1 helix that forms between 5Ј ribozyme and 3Ј product RNAs can vary without loss of catalytic activity (1). Therefore, 3Ј product RNAs can be designed to form H1 helices with the optimal number of base pairs to facilitate measure- . Reaction scheme for measuring rate and equilibrium constants. The functional contributions of specific features of active site architecture were evaluated from the effects of nucleotide modifications on equilibrium and kinetic parameters measured in self-cleavage and ligation reactions. Changes in cleavage and ligation rate constants, k cleav and k lig , reflect the effects of specific functional group modifications to the magnitude of the activation barrier to catalysis. Gϩ1 is the essential residue at the 5Ј-end of the 3Ј product RNA that mediates interdomain tertiary interactions (8). Tertiary structure stability is a primary determinant of the internal equilibrium, K eq int , between cleavage and ligation of bound products (13,14). Equilibrium dissociation constants, K d 5ЈR⅐3ЈP , for the complex that forms between 5Ј ribozyme and 3Ј product RNAs probe the contributions of functional groups to the stability of product binding in the ground state. Assays used to measure kinetic and equilibrium parameters are described under "Experimental Procedures." ments of cleavage or ligation kinetics or equilibrium dissociation constants for 5ЈR⅐3ЈP complexes (14,15,25) (Fig. 2). The LR43 and LR44 self-cleaving ribozyme variants with three or four base pairs in H1, respectively, were used to measure selfcleavage kinetics ( Fig. 2A). Using variants with small H1 helices ensured that product dissociation was much faster than the ligation step so that observed self-cleavage rates were not complicated by slow product dissociation and re-ligation of bound products. LR46 or LR47 variants with six or seven base pairs in H1, respectively, were used to measure ligation kinetics and equilibrium dissociation constants for ribozyme-product complexes (Fig. 2B). Using variants with six or seven base pairs in H1 ensured that K d 5ЈR⅐3ЈP values fell into the measurable range, between 0.1 and 1,000 nM, even when 3Ј product RNAs contained destabilizing Gϩ1 modifications.
Substitution of Gϩ1 with purine eliminates the possibility of any Watson-Crick hydrogen bonding interactions with C25 but the purine nucleotide retains the ability to form hydrogen bonds with ribose oxygens; N7 of purine is available to accept a hydrogen bond from the 2Ј hydroxyl of A38 and the 2Ј hydroxyl at the ϩ1 position still could donate a hydrogen bond to N6 of G36 (Fig. 1). Purine at the Gϩ1 position also retains the ability to stack between A38 and A26. Comparison of 5ЈR complexes with 3ЈP6 or 3ЈP6G؉1Pu RNAs that both form H1 helices with six base pairs shows that the Gϩ1Pu substitution increased the equilibrium dissociation constant for 3ЈP binding from a K d 5ЈR⅐3ЈP value of about 6 nM for the unmodified 3ЈP6 RNA to a K d 5ЈR⅐3ЈP value of 590 nM for the 3ЈP6 Gϩ1Pu RNA (Table 2). This 100-fold difference in K d 5ЈR⅐3ЈP values corresponds to a ⌬⌬G 5ЈR⅐3ЈP value of ϩ2.6 kcal/mol. A 5ЈR complex with 3ЈP7G؉1Pu RNA, which contains the Gϩ1Pu substitution but forms a seventh base pair in the H1 helix, displayed a K d 5ЈR⅐3ЈP value of 20 nM. This value is just 3-fold greater than the K d value measured for the 5ЈR⅐3ЈP6 complex with an unmodified 3Ј product RNA that forms an H1 helix with six base pairs. Thus, the ϩ2.6 kcal/mol loss of binding energy that results from the purine substitution of Gϩ1 is close to the ϩ2 kcal/mol increase in free energy that results from the loss of a conventional helical base pair in the H1 helix of 5ЈR⅐3ЈP6 relative to H1 of 5ЈR⅐3ЈP7 when both 3Ј product RNAs contain the same Gϩ1Pu modification (compare lines 3 and 4, Table 2). This ⌬⌬G 5ЈR⅐3ЈP value of ϩ2 kcal/mol that results from the loss of a helical base pair subsequently was used to deduce the change in binding energy that can be specifically attributed to a Gϩ1 modification when comparing 5ЈR⅐3ЈP complexes with six or seven base pairs in the H1 helix.
Complete deletion of the Gϩ1 nucleotide in 3ЈP7⌬G؉1 RNA and substitution of guanine with adenine in 3ЈP7G؉1A RNAs reduced 5ЈR⅐3ЈP complex stability by virtually the same amount as the Gϩ1Pu substitution, with ⌬⌬G 5ЈR⅐3ЈP values for the 5ЈR⅐3ЈP⌬G؉1 and 5ЈR⅐3ЈPG؉1A complexes of ϩ2.6 and ϩ2.8 kcal/mol, respectively (Table 2). Thus, the capacity of Gϩ1Pu and Gϩ1A nucleotides to form similar stacking interactions as Gϩ1, and the same hydrogen bonding interactions with ribose oxygens, contributes little to product binding affinity in the absence of Watson-Crick hydrogen bonding between the nucleobases at positions ϩ1 and 25. These results confirm evidence from FRET studies of interdomain docking equilibria that individual hydrogen bonds in the Gϩ1 binding pocket act cooperatively to stabilize the functional structure (29). Results of these FRET studies indicated that Gϩ1A and Gϩ1Pu modifications destabilized the docked conformation of a four-way junction ribozyme by just ϩ0.9 and ϩ0.6 kcal/mol, respectively. Thus, the loss of Watson-Crick hydrogen bonding between Gϩ1 and C25 nucleobases destabilized 3Ј product binding significantly more than it destabilized interdomain docking. Of course, no activity was detected when 3Ј product RNAs lacked the Gϩ1 nucleotide altogether. Gϩ1A modifications previously were reported to eliminate all detectable cleavage activity in minimal two-way junction ribozymes and in ribozymes that assemble in the context of a four-way helical junction (21,30). We found that complexes with Gϩ1Pu and Gϩ1A modifications displayed so little ligation activity that ligation pulse-chase experiments could not be used to measure cleavage kinetics. Cleavage rate constants could only be approximated from intermolecular cleavage reactions with separate ribozyme and substrate RNAs (Fig. 2C). The Gϩ1Pu and Gϩ1A modifications reduced cleavage rate constants by more than 10 3 -fold and reduced ligation rate constants by about 10 4 -fold. This loss of cleavage activity corresponds to increases of about ϩ3.9 and ϩ4.3 kcal/mol in the activation barrier to cleavage for Gϩ1Pu and Gϩ1A modifications, respectively ( Table 2). These changes in catalytic activity are significantly greater than the approximately ϩ2.7 kcal/mol loss in product binding affinity that resulted from the same Gϩ1 modifications (Table 2). Thus, interactions with the Watson-Crick hydrogen bonding face of Gϩ1 lower the activation barrier to catalysis more than they stabilize product binding in the ground state.
Restoration of single hydrogen bond donor and acceptor pairs between nucleobases at positions ϩ1 and 25 increased binding affinity by amounts that ranged from ϩ1.5 to ϩ2.2 kcal/mol ( Table 2). A single amino-imino hydrogen bond could form between the purine N1 in 3ЈPG؉1Pu RNA and N3 of uracil at position 25 in 5ЈRC25U RNA (bond 2 in Fig. 1). The 5ЈRC25U⅐3ЈPG؉1Pu complex was more stable than a complex with just the Gϩ1Pu modification or a complex lacking Gϩ1 completely, displaying a ⌬⌬G 5ЈR⅐3ЈP value of ϩ1.5 kcal/mol relative to an unmodified complex. The 5ЈRC25U⅐3ЈPG؉1Pu complex also displayed significantly more catalytic activity than complexes that are unable to form any Watson Crick hydrogen bonds between nucleobases at the ϩ1 and 25 positions, with a ⌬⌬G ‡ value of just ϩ1.0 kcal/mol relative to an unmodified ribozyme, corresponding to a 5-fold decrease in the cleavage rate constant.
A single hydrogen bond also could form between the keto oxygen of C25 in unmodified 5ЈR and the exocyclic amine of 2-aminopurine in 3ЈPG؉1(2AP) RNA or between the exocyclic amine of Gϩ1 in unmodified 3Ј product RNA and O2 of uridine in ribozyme with a C25U mutation (bond 3 in Fig.  1). 5ЈR⅐3ЈPG؉1(2AP) and 5ЈRC25U⅐3ЈP complexes with the capacity to form this single N2:O2 hydrogen bond displayed K d 5ЈR⅐3ЈP values that were more than 25-fold higher than K d 5ЈR⅐3ЈP values of an unmodified complex, with ⌬⌬G 5ЈR⅐3ЈP values of ϩ2.2 and ϩ2.0 kcal/mol, respectively. Thus, the 5ЈR⅐3ЈPG؉1(2AP) and 5ЈRC25U⅐3ЈP complexes that were able to form bond 3 were somewhat less stable than the 5ЈRC25U⅐3ЈPG؉1Pu complex that was able to form bond 2. They also displayed somewhat lower cleavage activity. Cleavage rate constants for 5ЈR⅐3ЈPG؉1(2AP) and 5ЈRC25U⅐3ЈP com-plexes were reduced by about 15-fold relative to an unmodified ribozyme complex, corresponding to a ⌬⌬G ‡ value of ϩ1.5 kcal/mol (Table 2).
Previously, a minimal ribozyme with a Gϩ1(2AP) substitution was reported to display the same cleavage rate constant as an unmodified ribozyme (30). Thus, it appears that the potential to form a hydrogen bond between the N2 exocyclic amine of a purine nucleobase and the O2 keto oxygen of a pyrimidine lowers the activation barrier to catalysis by somewhat less than it stabilizes product binding in the ground state. A 2-aminopurine substitution for Gϩ1 destabilized the docked conformation of a four-way junction ribozyme by just ϩ0.4 kcal/mol (29), suggesting that the potential to form the pyrimidine O2:purine N2 bond also contributes less to the stability of interdomain docking than to product binding affinity.
A C25U substitution reduced cleavage rate constants by just 15-fold, but reduced ligation rate constants almost 500-fold, shifting the internal equilibrium between cleavage and ligation of bound products, K eq int , from 36 in the unmodified complex to about 1.2 in the 5ЈRC25U⅐3ЈP complex (Table 2). Complexes with a single Gϩ1(2AP) substitution also retained less ligation activity than the unmodified ribozyme, displaying a K eq int value of 2. Shifts in the internal equilibrium toward cleavage were observed for virtually all active site modifications although the magnitudes of the shifts varied, as described below.
The 5ЈRC25U⅐3ЈPG؉1(2AP) complex, which has the potential to form two Watson-Crick hydrogen bonds, (bonds 2 and 3, Fig. 1) showed little change in complex stability or in catalytic activity relative to the unmodified ribozyme even though it lacks the pyrimidine N4-purine O6 hydrogen bond (bond 1, Fig.  1). The 5ЈRC25U⅐3ЈPG؉1A complex also has the potential to form two hydrogen bonds (bonds 1 and 2, Fig. 1) in an A:U Watson Crick pair that replaces the G:C pair in the unmodified ribozyme. This complex was about 10-fold less stable than an unmodified complex, corresponding to a ⌬⌬G 5ЈR⅐3ЈP value of ϩ1.3 kcal/mol. Thus, the loss of stability due to substitution of the Aϩ1:U25 tertiary interaction for the normal Gϩ1:C25 pair is comparable to the difference in stabilizing energy contributed by G:C and A:U pairs in base paired RNA helices (31,32). With 6-fold and 20-fold decreases in cleavage and ligation rate constants, respectively, the 5ЈRC25U⅐3ЈPG؉1A complex displayed a modest shift in the internal equilibrium toward cleavage, giving a K eq int value just 4-fold lower than the value of 36 measured for the unmodified ribozyme complex.
The 5ЈR⅐3ЈPG؉1I complex has the capacity to form the same purine N1:pyrimidine N3 and purine N6:pyrimidine O2 hydrogen bonds as the 5ЈRC25U⅐3ЈPG؉1A complex (bonds 1 and 2,  Fig. 1). However, with a ⌬⌬G 5ЈR⅐3ЈP value of ϩ2.45 kcal/mol, the 5ЈR⅐3ЈPG؉1I complex was only slightly more stable than a 5ЈR⅐3ЈPG؉1Pu complex that lacks the capacity to form any hydrogen bonds between nucleotide bases at positions ϩ1 and 25 (Table 2). Despite the 10-fold difference in K d 5ЈR⅐3ЈP values between 5ЈRC25U⅐3ЈPG؉1A complexes and 5ЈR⅐3ЈPG؉1I complexes, both complexes displayed similar cleavage rate constants of about 0.04 min Ϫ1 . This rate constant is just 6-fold below the cleavage rate constant of 0.24 min Ϫ1 measured for an unmodified ribozyme and more than 10 4 -fold greater than the cleavage rate constant of about 2.7 ϫ 10 Ϫ4 min Ϫ1 measured for the R4⅐S44G؉1Pu complex (Table 2). These results contrast with previous reports that a Gϩ1I substitution eliminates all detectable cleavage activity in the context of a minimal hairpin ribozyme (10,30). Our results show that inosine at the ϩ1 position is nearly as effective as guanine in lowering the activation barrier to catalysis, despite the loss of the pyrimidine O2: purine N6 hydrogen bond, when ribozyme tertiary structure is stabilized in the context of a 4-way helical junction.
This comparison of ⌬⌬G ‡ and ⌬⌬G 5ЈR⅐3ЈP values for 5ЈRC25U⅐3ЈPG؉1A and 5ЈR⅐3ЈPG؉1I complexes indicates that modifications of hydrogen bonding partners that appear to make similar contributions to active site architecture can have very different effects on catalysis and ground state stability. In particular, the interaction between the amidine group of C25 and inosine at the ϩ1 position (bonds 1 and 2, in Fig. 1) contributes little to 3Ј product RNA binding affinity but significantly lowers the activation barrier to catalysis. Uracil at position 25 also has the potential to form hydrogen bonds 1 and 2 with adenine at the 1 position in 3ЈPG؉1A RNA, but with the hydrogen bonding polarity reversed. However, the U25:Aϩ1 pair seems to stabilize product binding as much as it lowers the activation barrier to catalysis. These functional differences between C25:Iϩ1 and U25:Aϩ1 pairs would not be predicted from inspection of the structure alone.
The ligation rate constant for the 5ЈR⅐3ЈPG؉1I complex was reduced by just 40-fold. With a 6-fold decrease in the cleavage rate constant, the internal equilibrium continued to favor ligation relative to cleavage by 5-fold. Thus, the internal equilibrium continued to favor ligation for each of the modified ribozyme complexes that retain the ability to form two hydrogen bonds between nucleobases at positions 25 and ϩ1, 5ЈR⅐3ЈPG؉1I, 5ЈRC25U⅐3ЈPG؉1A and 5ЈRC25U⅐3ЈP6G؉1(2AP), with K eq int values between 5 and 20 ( Table 2 and Fig. 1). Each of the modified ribozyme complexes that retain just a single hydrogen bonding interaction, 5ЈRC25U⅐3ЈP6G؉1Pu, 5ЈRC25U⅐G؉1, and 5ЈR⅐G؉ 1(2AP), displayed a smaller preference for ligation, with K eq int values between 1.2 and 2. Both complexes that lack any potential for interdomain Watson Crick hydrogen bonding, 5ЈR⅐3ЈG؉1Pu and 5ЈR⅐3ЈG؉1A, displayed K eq int values below 1. Thus, K eq int values for modified ribozyme complexes correlated very well with the number of Watson Crick hydrogen bonding interactions that could form between nucleobases at positions ϩ1 and 25.
This relationship between the internal equilibrium and the capacity for interdomain Watson Crick hydrogen bonding is consistent with earlier evidence that tertiary structure stability is a critical determinant of the hairpin ribozyme proficiency as an RNA ligase. Previous work showed that low temperatures and high cation valency and concentration, all features that tend to stabilize RNA structures, shift the balance between cleavage and ligation of bound products in favor of ligation (13,16). Likewise, hairpin ribozymes that assemble in the context of a four-way helical junction display enhanced tertiary structure stability and much higher ligation activity relative to minimal ribozymes with a bulged or two-way helical junction linkage between the A and B domains (14,19).
Recent biochemical studies of an extended form of the hammerhead ribozyme revealed a similar relationship between tertiary structure stability and ligation proficiency (33). In the natural form of the hammerhead ribozyme, an interdomain loop-loop interaction, which is absent from minimal hammerhead ribozymes, stabilizes the functional tertiary structure (34 -36). While the internal equilibrium favors cleavage over ligation by more than 100-fold for minimal hammerhead ribozymes (37), ligation rate constants increased about 2,000-fold in reactions catalyzed by the extended form of the hammerhead ribozyme to give a K eq int value close to 1 (33). A likely explanation for the observation that tertiary structure stability promotes ligation more than cleavage is that ligation depends on active site structure for precise alignment of the 2Ј,3Ј-cyclic phosphate and 5Ј-hydroxyl termini that are to undergo ligation while the same functional groups are fixed relative to each other by the covalent diester linkage in cleavage substrates. According to this view, multiple hydrogen bonds between Gϩ1 and C25 promote ligation in the hairpin ribozyme not only by stabilizing interdomain interactions but also by constraining the propeller twist of the paired nucleobases in the optimal geometry to fix the 2Ј,3Ј-cyclic phosphate and 5Ј-hydroxyl termini in the optimal alignment for ligation.
Nucleobase-Ribose Interactions-The network of interactions in the Gϩ1 binding pocket also includes hydrogen bonds between nucleotide bases and ribose oxygens ( Fig. 1) (2). The 5ЈRG36I⅐3ЈP and 5ЈR⅐3ЈPdG؉1 complexes both lack the potential to form a hydrogen bond between the 2Ј-hydroxyl of Gϩ1 and the N6 exocyclic amine of G36 (bond 5, Fig. 1). Both of these complexes displayed almost the same catalytic rate constants as unmodified complexes. k ligation values fell by just 2-fold and k cleavage values even increased slightly (Table 3). K d 5ЈR⅐3ЈP values for 5ЈR⅐3ЈPdG؉1 and 5ЈRG36I⅐3ЈP complexes also were relatively unperturbed, increasing by just 2-and 4-fold, respectively. The ⌬⌬G 5ЈR⅐3ЈP and ⌬⌬G ‡ values of ϩ0.95 and ϩ0.4 kcal/mol, respectively, measured for complexes that lack the hydrogen bond donor or acceptor involved in bond 5 ( Fig. 1) are similar to the effects of eliminating 2Ј oxygens that participate in tertiary hydrogen bonds in other RNA structural contexts (23,38,39). These small effects also are consistent with a previous report that a dGϩ1 modification destabilized the docked conformation of a four-way junction ribozyme by just ϩ0.5 kcal/mol (29).
A second ribose hydroxyl-nucleobase hydrogen bond forms between N7 of Gϩ1 and the 2Ј-hydroxyl of A38 (bond 4, Fig. 1). A deoxyadenosine substitution for A38 had large effects on cleavage activity of minimal ribozymes, reducing the cleavage rate constant by 50-fold (40,41). We observed smaller effects in the context of a four-way junction ribozyme; a deoxynucleotide substitution of A38 reduced cleavage and ligation rate constants by 2-and 12-fold, respectively. The 6-fold shift in the internal equilibrium in favor of cleavage could reflect a modest contribution of this hydrogen bond to aligning cleavage product termini for ligation. Loss of the 2Ј-hydroxyl of A38 had no significant effect on 3Ј product binding affinity (Table 3).
Gϩ1 Stacking Interactions-Gϩ1 stacks between A38 and A26 nucleobases in the loop B binding pocket (Fig. 1, A and  B). The exocyclic N6 and ring N1 functional groups of A38 interact directly with the 5Ј-oxygen leaving group and nonbridging oxygens of the reactive phosphodiester and A38 appears to participate directly in catalytic chemistry (3,17).
An abasic deoxynucleotide substitution of A38 reduced cleavage and ligation rate constants by 14,000-fold and 370,000-fold, respectively (18), corresponding to a ⌬⌬G ‡ value of ϩ5.4 kcal/mol. The significant decrease in catalytic activity associated with loss of the A38 nucleobase is consistent with a major role in catalysis. However, almost full activity could be restored to an abasic ribozyme variant lacking A38 when certain nucleobase analogs were provided in solution (18), suggesting that the active site structure remains sufficiently intact to allow specific nucleobase recognition and binding in a cavity left by the A38 nucleobase deletion. Each of the exogenous nucleobases that was able to restore catalytic activity to abasic variants that lack A38 (isocytosine, 2-aminopyridine, 3-methyladenine, and 2,6-diaminopurine) is a planar heterocycle with the amidine functional group that corresponds to the Watson-Crick hydrogen bonding face of adenine. Thus, exogenous nucleobase rescue experiments suggest that restoration of catalytic activity to the abasic variant occurs through exogenous nucleobase binding in the cavity left by the adenine deletion (17). Strikingly, the purine nucleobase, without an amidine functional group, displayed the same affinity for an abasic ribozyme lacking A38 as the nucleobases with amidine groups that were capable of rescue but purine was unable to restore catalytic activity. Competitive inhibition of exogenous nucleobase rescue by purine argues that interactions with the amidine group of A38 significantly lower the energy barrier in the transition state but make no detectable contribution to ground state stability.
Comparison of the effects of an abasic substitution on catalysis and on ribozyme⅐product complex stability supports the interpretation that interactions with the amidine group of A38 are more important in the transition state than in the ground state. In contrast to the major contribution of A38 to lowering the activation barrier to catalysis, loss of A38 had only minor effects on stability of the ribozyme-product complex in the ground state. An abasic substitution of A38 reduced 3Ј product RNA binding affinity by 15-fold, corresponding to a ⌬⌬G 5ЈR⅐3ЈP value for the 5ЈRA38dX⅐3ЈP complex of just ϩ1.55 kcal/mol (Table 3). This loss in complex stability is similar to the ϩ1.5 kcal/mol loss of stability observed with partial disruption of Watson Crick bonding interactions with Gϩ1 in a 5ЈRC25U⅐3ЈPG؉1Pu complex ( Fig. 1 and Table 2). The 5ЈRA38dX⅐3ЈP complex was considerably more stable than a complex with a Gϩ1Pu substitution or complexes that lack Gϩ1 altogether. The modest decrease in complex stability that resulted from the loss of A38 confirms evidence from abasic ribozyme rescue experiments that A38 contributes significantly more to lowering the activation barrier to catalysis than to stabilizing active site architecture.
Role of C2Ј-Endo Sugar Puckers in Aligning the Reactive Phosphate-The final feature of active site architecture that we evaluated is the nature of the sugar puckers of the nucleotides flanking the reactive phosphate. Ribonucleosides typically adopt a C3Ј-endo sugar pucker that places the electronegative 3Ј-hydroxyl in the preferred axial orientation (42). However, C2Ј-endo puckers, which are more characteristic of deoxyribonucleosides, were observed in the crystal structure at the A-1 and Gϩ1 positions (2). These unusual puckers define the trajectory of the phosphodiester backbone at the reactive phosphate, fixing the adjacent 2Ј-hydroxyl and 5Ј oxygens in the in-line geometry that is consistent with the S N 2-type reaction mechanism.
A deoxyguanosine substitution for Gϩ1 eliminates hydrogen bond donation by the 2Ј-hydroxyl to the keto oxygen of G36 but favors the appropriate C2Ј-endo conformation. As mentioned above, loss of the 2Ј-hydroxyl from Gϩ1 had very small effects on product binding affinity. Small effects of the deoxynucleotide substitution for Gϩ1 on cleavage and ligation rate constants combined to make the internal equilibrium slightly less favorable for ligation by about 3-fold. The 2Ј-OCH 3 modification of Gϩ1 reduced 5ЈR⅐3ЈPmG؉1 complex stability somewhat more, about 5-fold, and reduced cleavage and ligation rate constants by 2-and 7-fold, respectively. The larger effects of the 2Ј-OCH 3 modification relative to the 2Ј-deoxy modification might be explained by different effects on the Gϩ1 sugar pucker. The 2Ј-OCH 3 modification stabilizes the typical ribonucleoside C3Ј endo pucker while the ribose at the Gϩ1 position adopts the C2Ј-endo pucker characteristic of deoxynucleosides (2,42,43). a 3ЈP6 or 3ЈP7 RNAs were used for k ligation and K d 5ЈR⅐3ЈP measurements. b Measured using ligation chase experiments with 5ЈR and 3ЈP3 RNAs. c k ligation ϭ k obs,ligation Ϫ k cleavage with k obs,ligation measured at saturating RNA concentrations and k cleavage values obtained from ligation chase experiments. d K eq int ϭ k ligation /k cleavage . e K d 5ЈR⅐3ЈP ϭ K d,app 5ЈR⅐3ЈP ϫ (K eq ϩ1). f ⌬G 5ЈR⅐3ЈP (15°C) ϭ ϪRT ln(1/ K d 5ЈR⅐3ЈP ), where R is the gas constant and T is the temperature in degrees Kelvin. g ⌬⌬G 5ЈR⅐3ЈP (15°C) ϭ ⌬G 5ЈR⅐3ЈP6,unmodified (15°C) Ϫ ⌬G 5ЈR⅐3ЈP6,modified (15°C) or ⌬G 5ЈR⅐3ЈP6,unmodified (15°C) Ϫ ⌬G 5ЈR⅐3ЈP7,modified (15°C) Ϫ 2 kcal/mol for complexes with 3ЈP7 instead of 3ЈP6, as explained in the text. h ⌬G ‡ (15°C) ϭ ϪRT ln (k cleavage h/k B T) where h is Planck's constant, k B is Boltzmann's constant, and k cleavage is the self-cleavage rate constant at T ϭ 288°K. i ⌬⌬G ‡ (15°C) ϭ ⌬G ‡,unmodified (15°C) Ϫ ⌬G ‡,modified (15°C). j Determined using gel mobility shift assays. k Determined from the RNA concentration dependence of observed ligation rates. l Value taken from Kuzmin et al. (17).

CONCLUSION
Modifications that changed the potential to form specific interactions that were inferred from crystal structures had different effects on the activation barrier to catalysis, the stability of ribozyme complexes in the ground state, and on the internal equilibrium between cleavage and ligation. All modifications of the Gϩ1 binding pocket inhibited ligation more than cleavage, shifting the internal equilibrium between cleavage and ligation, K eq int , toward cleavage. These results are consistent with previous evidence that tertiary structure stability is the major determinant of the balance between cleavage and ligation. A striking example of differential effects of active site modifications on ground state stability and catalysis was seen with deletion of an active site adenine, A38, which increased the activation barrier to catalysis by more than ϩ5 kcal/mol but reduced the ground state stability of the ribozyme product complex by just ϩ1.6 kcal/mol. Finally, seemingly equivalent functional group substitutions sometimes had very different functional consequences, exemplified by the significant differences in product binding affinity that were seen for two modified ribozyme complexes with the potential to form the same interdomain hydrogen bonds. These quantitative functional studies complement previous analyses of the contributions of Gϩ1 interactions to tertiary structure stability using FRET and provide one of the most detailed views yet of structure-function relationships that contribute to positioning and orientation within a ribozyme active site.