Differential Assembly of Catalytic Interactions within the Conserved Active Sites of Two Ribozymes

Molecular recognition is central to biology and a critical aspect of RNA function. Yet structured RNAs typically lack the preorganization needed for strong binding and precise positioning. A striking example is the group I ribozyme from Tetrahymena, which binds its guanosine substrate (G) orders of magnitude slower than diffusion. Binding of G is also thermodynamically coupled to binding of the oligonucleotide substrate (S) and further work has shown that the transition from E•G to E•S•G accompanies a conformational change that allows G to make the active site interactions required for catalysis. The group I ribozyme from Azoarcus has a similarly slow association rate but lacks the coupled binding observed for the Tetrahymena ribozyme. Here we test, using G analogs and metal ion rescue experiments, whether this absence of coupling arises from a higher degree of preorganization within the Azoarcus active site. Our results suggest that the Azoarcus ribozyme forms cognate catalytic metal ion interactions with G in the E•G complex, interactions that are absent in the Tetrahymena E•G complex. Thus, RNAs that share highly similar active site architectures and catalyze the same reactions can differ in the assembly of transition state interactions. More generally, an ability to readily access distinct local conformational states may have facilitated the evolutionary exploration needed to attain RNA machines that carry out complex, multi-step processes.


Introduction
Molecular recognition is essential for RNA function, allowing structured RNAs to bind and sometimes transform a diverse array of ligands, including nucleic acids, proteins, and small molecules [1][2][3][4] with high specificity. Nevertheless, RNA functional sites are often limited in preorganization, perhaps a reflection of RNA's tendency to form alternative interactions [5,6], and often require conformational rearrangements to accommodate bound ligands (e.g. [1, 7-

Results
To probe active site interactions with G in the E•G and E•S•G complexes, we utilized metal ion rescue experiments, which rely on the preferred coordination of soft metal ions to sulfur and nitrogen compared to oxygen (e.g. [17,[25][26][27][28]). Prior rescue experiments with the Tetrahymena ribozyme provided evidence for metal ion interactions in the transition state, with two active site metal ions, M A and M C , contacting the G 2 0 -and 3 0 -oxygen atoms, the U(-1) 3 0 -oxygen atom on S and the reactive phosphoryl group (Fig 1B, [25,29]). In addition, recent rescue experiments indicate that the G 2 0 -and 3 0 -OH groups, which likely contact M C in the transition state, make these interactions in E•S•G, but make non-catalytic interactions in E•G, as described above and depicted in Fig 1B [19].
The crystallographic data for the Azoarcus ribozyme show two active site metal ions analogous to M A and M C , with M C within coordination distance to the G 2 0 -and 3 0 -OH groups when S is present in the active site [30]. These apparent contacts are equivalent to interactions deduced from metal ion rescue experiments with the Tetrahymena ribozyme in the E•S•G complex ( Fig  1B, [19]). X-ray structures of the Azoarcus E•G complex are not available but the reaction framework shown in Fig 1A indicates that G has the same binding affinity in the presence and absence of S. This observation, most simply, suggests that the catalytic metal ion interactions between M C and the G -OH groups in the E•S•G complex are also made in the E•G complex. Alternatively, G may assume a different configuration within the Azoarcus E•G complex, like the Tetrahymena ribozyme ( Fig 1B, [19]), with the affinity of G to E and E•S being coincidentally the same.
To distinguish these models, we replaced the G 2 0 -and 3 0 -OH groups with an amino group (-NH 2 ) and measured binding of 2 0 -and 3 0 -aminoguanosine to the Azoarcus ribozyme in so-  [20,22]) and Azoarcus (right, [23]) ribozymes. (B) Model for the assembly of catalytic metal ion interactions with M A (blue) and M C (orange) in the Tetrahymena group I ribozyme active site [19,21,24]. Closed dots and hatched lines represent metal ion interaction and hydrogen bonds, respectively. Partial negative charges are represented by 'δ -'. Prior data suggest that a third metal ion, M B , contacts the G 3 0 -oxygen atom instead of M C in the transition state [25] but this interaction is likely an artifact from functional experiments, as described in the Discussion and elsewhere [19]. For simplicity, we show M C contacting the G 3 0 -oxygen atom in the transition state.
called metal ion rescue experiments [17,26,[31][32][33]. Nitrogen prefers to coordinate to comparatively soft metal ions (e.g., Mn 2+ over Mg 2+ ), as has been observed in prior work [19,21]. If a G -OH group forms a metal ion interaction, then the simplest expectation would be that replacing that -OH with a -NH 2 group would result in weak binding in Mg 2+ alone and stronger binding upon addition of Mn 2+ . In contrast, the absence of this signature would provide no evidence for a metal ion / -OH group contact, but, as a negative result, would also not provide strong evidence against such an interaction.
We also measured binding of 2 0 -and 3 0 -deoxyguanosine (G(2 0 H) and G(3 0 H), respectively) to the Azoarcus ribozyme. The hydrogen (-H) substituent, unlike the -OH group, can neither coordinate with a metal ion nor form a hydrogen bond. Thus, disruption of G binding by -H substitution would suggest that a stabilizing interaction is made with the corresponding -OH group, although smaller effects can potentially arise from differential sugar pucker preferences.
Equilibrium dissociation constants for binding of the G analogs to the Azoarcus E•S complex were measured following procedures described in Materials and Methods, and our results are summarized in Fig 2C and 2D. Replacing the 2 0 -or 3 0 -OH group with an -NH 2 group has 17-or !66-fold destabilizing effects on G binding, respectively. The upper limit of !66 was set for G(3 0 NH 2 ) binding due to the dominant contribution of the ÀNH þ 3 species to the observed affinity under experimentally accessible conditions (Materials and Methods, S9 Fig). Mn 2+ enhances binding of G(2 0 NH 2 ) to E•S so that its binding affinity is within error to that of G ( Fig 2C). We could not determine Mn 2+ rescue for G(3 0 NH 2 ), due to the dominant binding of the ÀNH þ 3 species. The weakened binding of the G 2 0 -and 3 0 -H substitution, by 31-fold and 24-fold, respectively, provided additional evidence for interactions with both-OH groups.
The Mn 2+ specificity switch observed for G(2 0 NH 2 ) is consistent with preliminary rescue experiments [34] and crystallographic data [30] and, along with the destabilizing effect of 2 0 -H substitution, provides strong evidence for a direct metal ion interaction with the G 2 0 -OH group in the E•S•G complex. Given the observation that the 3 0 -OH group is within coordination distance of M C in the Azoarcus crystal structure [30], the drop in affinity for G(3 0 H) and G (3 0 NH 2 ) is most simply explained by the disruption of a stabilizing interaction with M C and the G 3 0 -moiety. Nevertheless, we could not obtain direct evidence for this metal ion interaction or more strongly rule out alternative models involving disruption of other interactions because of technical limitations that precluded metal ion rescue experiments at this position (see above and Materials and Methods). The inferred metal ion interactions with the G 2 0 -and 3 0 -OH groups are equivalent to contacts made with the G -OH groups in the Tetrahymena E•S•G complex ( Fig 1B) and suggest that both the Tetrahymena and Azoarcus ribozymes form the metal ion interactions required for catalysis in the presence of bound S. We next probed for interactions made with the G -OH groups in the Azoarcus E•G complex. We measured binding of the G analogs under conditions in which S is not predominantly bound to the ribozyme (see Materials and Methods), and our results are summarized in Fig 2A  and 2B. As observed above for E•S, replacing either the 2 0 -and 3 0 -OH group with an -NH 2 group had strong destabilizing effects on binding of G, 51-and !143-fold, respectively ( Fig  2A). In addition, Mn 2+ restored binding of G(2 0 NH 2 ) to levels similar to that for G binding (Fig 2A). Mn 2+ rescue of G(3 0 NH 2 ) binding could not be tested for the reasons described above. Finally, 3 0 -H substitution weakened binding by 22-fold ( Fig 2B). Unexpectedly, no deleterious effect was observed for binding of G(2 0 H) to E, despite the evidence for a metal ion interaction with the 2 0 -OH group.
The Mn 2+ specificity switch observed for G(2 0 NH 2 ) binding provides strong evidence that the 2 0 -OH group contacts a metal ion in E•G. Although we were unable to probe for an analogous switch for G(3 0 NH 2 ) binding due to technical limitations described above, the observation that the 3 0 -NH 2 and 3 0 -H substitutions weaken G binding is indicative of a stabilizing interaction being made with the G 3 0 -OH group. Although more complex models are possible, analogous functional results were obtained for the E•S•G complex (Fig 2C and 2D), where there is also direct x-ray evidence for 2 0 -and 3 0 -OH interactions with M C [30]. We therefore adopt the simplest model in which these same interactions are present in the Azoarcus E•G complex.
As for the absence of a deleterious effect for 2 0 H substitution to E, we postulate that the deoxyribose ring of the G(2 0 H) binds to the ribozyme in an alternative configuration in which the 2 0 -H atom faces away from the active site metal ions. This configuration would allow the metal ions to remain solvated and thereby avoid a substantial energetic penalty and allow its binding energy to coincidentally match that for G. Such enhanced flexibility may be facilitated by the small size of the-H atom and the alternative sugar conformation of the deoxyribose ring that would be restricted with the bulkier-NH 2 substituent or when S is present in the active site.
For the Tetrahymena ribozyme, there is strong evidence for the absence of an interaction with the 2 0 -OH group of G and for the presence of a nonproductive contact between the adjacent 3 0 -OH group and M A (Fig 1B, [19]). In contrast, the Azoarcus data suggest interactions with both the 2 0 -and 3 0 -OH groups (Fig 2). These results imply that the off-pathway configuration observed for the Tetrahymena ribozyme is not present for the Azoarcus ribozyme.

Discussion
We present a model for molecular recognition of G by the Tetrahymena and Azoarcus group I ribozymes (Fig 3), based on data described here and elsewhere [8,19,23]. The free form of the G binding site, as suggested from the very slow rate of G association to the Azoarcus and Tetrahymena ribozymes (~10 5 M -1 min -1 , [23]), favors one or more G-inaccessible ("rearranged," Fig 3A) configurations for both ribozymes. This scenario is illustrated in the qualitative energy landscape for the G binding site of E of Fig 3A, which shows similar profiles for the Azoarcus and Tetrahymena ribozymes.
Transient formation of the G-accessible ("open", Fig 3A) state allows G to bind to the ribozyme. However, our results suggest that the assembly of catalytic interactions with bound G is different for the Tetrahymena and Azoarcus ribozymes. Prior results with the Tetrahymena ribozyme [19] suggest that bound G is in equilibrium between an active and an inactive configuration in which the G -OH groups do and do not contact M C , respectively. The inactive form is preferred within E•G (Figs 1B and 3B) and the active form is favored once S is bound (E•S•G, Figs 1B and 3C).
In contrast, our results and prior x-ray structural results [30] with the Azoarcus ribozyme indicate interactions with both the G -OH groups and most simply suggest that the conformation in which both -OH groups coordinate to M C , is preferred for both E•G and E•S•G (Fig 3B  and 3C, red). This model implies that the active site of the Azoarcus ribozyme does not significantly populate the alternative off-pathway state present for the Tetrahymena ribozyme and suggests an energy landscape for the Azoarcus ribozyme that is more constrained towards the active form than that for the Tetrahymena ribozyme. This difference occurs despite both introns having resting unbound states that are G-inaccessible (Fig 3A, "rearranged") and highly similar G-occupied binding sites in the E•S•G complexes that overlay with an RMSD of <1 Å [35]. Phylogenetic, mechanistic, and interference studies suggest that the active sites of the Tetrahymena and Azoarcus ribozymes are highly similar [23,34,36,37] and thus favor models in which local rather than global rearrangements occur to assemble the catalytically active conformation.
The ability to access local conformational differences to assemble catalytic interactions suggests a degree of flexibility within the group I ribozyme active site. This flexibility is consistent with metal ion rescue experiments for the Tetrahymena ribozyme that implicate a third metal ion, M B , that contacts the G 3 0 -oxygen atom instead of M C in the transition state (Fig 1B, [25]). Functional studies with the Tetrahymena ribozyme provided strong evidence for a transition state interaction of the G 3 0 -atom with a weakly associating metal ion (i.e. M B ) that is distinct from M A and M C , but M B has not been observed in any of the group I intron x-ray structures [25,30,35]. These functional studies employed metal ion rescue of 3 0 -thio-substituted G [25]. However, recent results suggest that for metal ion rescue of 3 0 -amino-substituted G, M C instead of M B interacts with the 3 0 -atom in the E•S•G ground state complex [19]. Conformational landscapes of the G binding site. Qualitative energy landscapes of the G binding sites of E (A) and the EG (B) and ESG complex (C), for the Tetrahymena and Azoarcus ribozymes (black and red lines, respectively). The G binding site is shown as a surface representation with or without bound G (green). G is also represented in stick form in panels B and C to highlight active site metal ion interactions in the inactive and active states [19]. The energy landscape and the accompanying structures were constructed as described in Materials and Methods. The simplest model to account for these seemingly disparate results is that with the G 3 0 -sulfur the active site of the Tetrahymena ribozyme is sufficiently malleable to recruit M B to form a non-cognate contact. Thus, it appears that the group I active site, while likely utilizing two metal ions for its cognate reaction, can readily toggle between a two metal ion and three metal ion mechanism for catalysis. These observations appear to be a manifestation of RNA's ability to adopt multiple distinct conformational-and functional-states [12,[38][39][40][41].
While the rate of G association to the Tetrahymena and Azoarcus ribozymes is the same [23], G association to the Twort group I ribozymes is~10-fold faster than to the Tetrahymena and Azoarcus. This, together with weaker binding of G to the Anabaena ribozyme relative to Azoarcus, suggests a range in G association rate constants [23,42,43]. It will be fascinating to uncover the structural variation, whether local or global, that differentially impacts these conserved cores.
Our results highlight RNA's ability and tendency to access local conformational differences, even within the conserved catalytic core of a ribozyme. This feature of RNA is common (e.g. [11,26,40,[44][45][46][47][48]), and may have been a valuable asset for Nature in its evolution of complex RNA machines such as the ribosome and splicosome that must access multiple conformational states [12-14, 39, 49]. It will thus be of substantial interest to elucidate the aspects of RNA structure that dictate its ability to restrict or access multiple conformational states, and model systems such as group I ribozymes may provide valuable clues and tools.

General kinetic methods
All cleavage reactions were single-turnover with ribozyme in excess of trace 32 P-radiolabeled substrate (< 0.1 nM), and were carried out at 30°C in the presence of 15 mM MgCl 2 [23] and 50 mM buffer at various pH values (sodium acetate, pH 5.0-5.5; NaMES, pH 6.1-6.7; NaMOPS, pH 7.1; NaEPPS, pH 7.7-8.2; and NaCHES, pH 8.7-9.7). For metal ion rescue experiments, reactions were performed in the absence or presence of 10 mM MnCl 2 . For G (3 0 N), reactions were performed in the presence of 100 mM Mg 2+ to attenuate strong binding of the 3 0 -NH þ 3 form of the analog (S8 and S9 Figs).
Ribozymes were allowed to fold for 30 min at 50°C in 15 mM MgCl 2 and 50 mM buffer at the pH of the final reaction mixture. For reactions performed above pH 8.0, the ribozyme was folded in 25 mM NaMES, pH 6.7, and 15 mM MgCl 2 to minimize degradation of the ribozyme. After cooling to room temperature, the mixture was diluted 10-fold in a solution containing the desired concentration of buffer, G analog, and divalent metal ions. All reactions were allowed to equilibrate for 5-10 min at 30°C before initiating the cleavage reaction by the addition of 32 P-radiolabeled substrate. At specific times, 2 μL aliquots were taken from the 20 μL reaction mixture and quenched in 4 μL quench solution containing 90% formamide, 50 mM EDTA, bromophenol blue and xylene cyanole. 32 P-radiolabeled substrate and product were separated by gel electrophoresis on a denaturing polyacylamide gel (7M urea/20%acrylamide) and quantitated using Phosphorimager analysis (GE Healthcare) with TotalLab.
Reactions were followed to completion and generally exhibited endpoints >95%. Slow reactions were followed for up to 24 hours and rate constants were obtained from initial rates assuming 95% endpoints. Data were analyzed using KaleidaGraph, Synergy Software and errors reported represent the standard error values of the parameters obtained by fitting, unless stated otherwise.

Measuring the affinity of G to E and E•S
Equilibrium dissociation constants for G from E•G and E•G•S (K G d and K G 0 d , respectively, Fig 1A) were determined by plotting the observed rate (k obs ) of cleavage of 5 0 -32 P-radiolabeled CAUA 5 (or CAdUA 5 ) as a function of the concentration of G (0-2.5 mM G), and fitting the data to Eq 1: K G 1=2 , the concentration of G where the observed rate constant (k obs ) is half the maximal rate (k max ), is equal to K G d (or K G 0 d ) when the chemical step is rate-limiting. For the CAUA 5 substrate, the chemical step is rate-limiting at pH <6.3 [23]. Above this pH, we used the CAdUA 5 substrate in which the cleavage site 2 0 -OH group is replaced with a hydrogen atom. This modification slows the reaction rate by~1000-fold so that chemistry is rate limiting at higher pH [23,51].
Binding of G to E•S was monitored under conditions where E is saturating with respect to S ([E] = 500 nM, K S d = 68 nM) [23]. To monitor binding of G to E, a subsaturating enzyme concentration (5-20 nM E) was used. Binding constants of G determined in this work are within~2-3-fold with previously published affinities [23]. Since binding of G is pH independent over the range measured (S1-S3 Figs), affinities for G are represented as an average of independent measurements at various pH with the error bars representing the corresponding standard deviation.
Measuring the affinities of G(2 0 H), G(3 0 H), and G(3 0 N) to E and E•S G(2 0 H), G(3 0 H), and G(3 0 N) are essentially non-reactive [19] so binding of these analogs to E and E•S was measured via competitive inhibition of the reaction with G. At low pH (< pH 7.0), we used the CAUA 5 substrate to accelerate the rate of reaction. Experiments were carried out with subsaturating G (10-30 μM G) and with varying concentrations of G(2 0 H), G(3 0 H), and G(3 0 N) (0-3 mM) at the appropriate [E] to ensure that binding to E or E•S was being monitored (see above). For each G analog (G X ), the data were fit to Eq 2: K i , the inhibitory concentration at which the observed rate constant (k obs ) is half the maximal rate (k max ), is equal to the equilibrium dissociation constant (K G x d or K G x 0 d ).

Measuring the affinity of G(2 0 N) to E and E•S
The affinity of G(2 0 N) to E and E•S was measured using an activity assay following procedures analogous to what we described for G (see above). Equilibrium dissociation constants for G (2 0 N) were determined by plotting the observed rate (k obs ) of cleavage of 5 0 -32 P-radiolabeled CAdUA 5 as a function of the concentration of G(2 0 N) (0-2.5 mM), and fitting the data to Eq 3.
Following Binding of the-NH 2 and-NH þ 3 forms of G(2 0 N) and G(3 0 N) The -NH 2 group on G(2 0 N) and G(3 0 N) ionizes with pK a values of 6.2 and 7.0, respectively [52,53], to form the corresponding -NH þ 3 species. To ensure that we were monitoring binding of the -NH 2 and not the -NH þ 3 species [19,54], we determined the pH-dependence of G(2 0 N) and G(3 0 N) binding to E and E•S. The data were fit to a model for binding of the -NH 2 and-NH þ  [19,54]. Strong binding presumably arises from favorable electrostatic interactions with the -NH þ 3 species in the active site of these ribozymes. The high affinity of G(3 0 NH þ 3 ) obscured our measurements of G(3 0 NH 2 ) at the highest accessible pH value without significant ribozyme degradation in the presence of Mn 2+ (pH 7.7). Metal ion rescue of G(3 0 NH 2 ) binding could thus not be tested.

Construction of qualitative energy landscapes for the Azoarcus and Tetrahymena G binding sites
Here, we describe the rationale behind the conformational landscapes depicted in Fig 3. Within E, prior data suggests that the "open" or G-accessible state is not the most stable species at equilibrium and that the preferred state may correspond to one or more collapsed or rearranged configurations that prevent binding of G [8,23]. The surface representation of the open state was obtained by removing G from the G binding site of the Tetrahymena group I intron (PDB 1X8W), and the rearranged state was built from the structure of the Tetrahymena group I intron using Assemble to alter the position of the top base triple so that G binding is occluded [18]. We emphasize that these structures are models of possible states in the conformational ensemble of the G binding site (see [19] for additional discussion). Binding of G stabilizes the G accessible state (as depicted by the change in relative energies of the rearranged and open states) and, within E•G, bound G can exist in (at least) two configurations referred to as "active" and "inactive". Prior data [19] suggest that the inactive state is the preferred state in Tetrahymena, whereas the work described herein suggests that the active state is preferred in Azoarcus. To represent this, the well for the active state is lower in energy than the well for the inactive state. Within E•S•G, the active site of the Tetrahymena ribozyme undergoes a rearrangement to the active state, which is preferred for both the Azoarcus and Tetrahymena ribozymes.

Supporting Information
report binding of G (2 0 NH 2 ) and G(2 0 NH þ 3 ) to E, respectively. A limit of >5 mM -1 was set for K Gð2 0 NH þ 3 Þ a since binding of G(2 0 N) did not level off at pH 7.7 ((K Gð2 0 NÞ a;obs = 5 mM -1 ). pK Gð2 0 NÞ a and pK EGð2 0 NÞ a are the equilibrium constants for deprotonation of G(2 0 NH þ 3 ) in solution and in E•G(2 0 N), respectively. The value of pK Gð2 0 NÞ a was set to 6.2 [52][53][54], and pK EGð2 0 NÞ a was determined by completing the thermodynamic cycle. Binding affinities are summarized in the S1 File.
report binding of G(2 0 NH 2 ) and G(2 0 NH þ 3 ) to E. A limit was set for K Gð2 0 NH þ 3 Þ a because binding of G(2 0 N) does not level off at the lowest pH measured. pK Gð2 0 NÞ a and pK EGð2 0 NÞ a are the equilibrium constants for deprotonation of G(2 0 NH þ 3 ) in solution and in E•G (2 0 N), respectively. pK Gð2 0 NÞ a was set to 6.2 [52][53][54], and pK EGð2 0 NÞ a was determined by completing the thermodynamic cycle. Binding affinities are summarized in the S1 File.
report binding of G(2 0 NH 2 ) and G(2 0 NH þ 3 ) to E•S, respectively. A limit of >5 mM -1 was set for K Gð2 0 NH þ 3 Þ a as binding of G(2 0 N) did not level off at pH 7.7 (K Gð2 0 NÞ a;obs = 5 mM -1 ). pK Gð2 0 NÞ a and pK ESGð2 0 NÞ a are the equilibrium constants for deprotonation of G(2 0 NH þ 3 ) in solution and in E•S•G(2 0 N), respectively. pK Gð2 0 NÞ a is 6.2 [52][53][54], and pK ESGð2 0 NÞ a was determined by completing the thermodynamic cycle. Binding affinities are summarized in the S1 File.  N) (black) and G (grey) binding to E. The binding affinities (K Gx a;obs (= 1/K Gx d;obs ) for G(3 0 N) and G were obtained from the data in (A) and S3A Fig, respectively. The solid line (black) is a fit of the G(3 0 N) data according to the model shown in (C) which describes binding of the -NH 2 and -NH þ 3 forms of G(3 0 N) to E. For comparison, the data was also fit to a model in which the-NH 2 form of (3 0 N) does not bind to E (dotted line). The solid line (grey) is the average of the measured G affinities to E (21.9 mM -1 ; see S1C Fig). (C) Model for binding of G (3 0 NH 2 ) and G(3 0 NH þ 3 ) to E. K Gð3 0 NH 2 Þ a and K Gð3 0 NH þ 3 Þ a report binding of G(3 0 NH 2 ) and G(3 0 NH þ 3 ) to E, respectively, and were obtained from the fit of the data in (B). The K Gð3 0 NH 2 Þ a value is reported as a limit because binding of G(3 0 N) does not level off at high pH. pK Gð3 0 NÞ a and pK EGð3 0 NÞ a are the equilibrium constants for deprotonation of G(3 0 NH þ 3 ) in solution and in E•G (3 0 N), respectively. pK Gð3 0 NÞ a is 7.0 [52], and pK EGð3 0 NÞ a was determined by completing the thermodynamic cycle. Binding affinities are summarized in the S1 File.  2+ . The data were analyzed as described in Materials and Methods. Binding affinities are summarized in the S1 File. (TIF) S1 File. Binding affinities used in S1-S10 Figs are compiled in the S1 File. (XLSX)