Triphosphate Reorientation of the Incoming Nucleotide as a Fidelity Checkpoint in Viral RNA-dependent RNA Polymerases*

The nucleotide incorporation fidelity of the viral RNA-dependent RNA polymerase (RdRp) is important for maintaining functional genetic information but, at the same time, is also important for generating sufficient genetic diversity to escape the bottlenecks of the host's antiviral response. We have previously shown that the structural dynamics of the motif D loop are closely related to nucleotide discrimination. Previous studies have also suggested that there is a reorientation of the triphosphate of the incoming nucleotide, which is essential before nucleophilic attack from the primer RNA 3′-hydroxyl. Here, we have used 31P NMR with poliovirus RdRp to show that the binding environment of the triphosphate is different when correct versus incorrect nucleotide binds. We also show that amino acid substitutions at residues known to interact with the triphosphate can alter the binding orientation/environment of the nucleotide, sometimes lead to protein conformational changes, and lead to substantial changes in RdRp fidelity. The analyses of other fidelity variants also show that changes in the triphosphate binding environment are not always accompanied by changes in the structural dynamics of the motif D loop or other regions known to be important for RdRp fidelity, including motif B. Altogether, our studies suggest that the conformational changes in motifs B and D, and the nucleoside triphosphate reorientation represent separable, “tunable” fidelity checkpoints.

Genome maintenance and propagation are dependent on faithful and efficient nucleic acid replication catalyzed by a large superfamily of template-directed polymerases (1). In each cycle of nucleotide addition, these polymerases must efficiently select the correct nucleotide that will properly base-pair with the template strand against a large pool of non-cognate nucleotides. The fidelity of nucleotide selection is essential for the integrity and proper expression of the genome. However, evo-lutionary processes require some level of incorrect nucleotide incorporation to generate the genetic diversity that allows a population of organisms to survive challenges from their environment. RNA viruses especially exemplify these ideas. For these viruses, it is now well established that the accuracy of RNA replication is a key determinant of viral virulence (2)(3)(4)(5)(6). Many RNA-dependent RNA polymerases (RdRps) 3 that are responsible for RNA viral genome replication operate near a limit of fine tuned mutation frequency, which optimally balances genetic diversity with overall genome integrity (7). Antiviral compounds like ribavirin that push the RdRp past an "error threshold" lead to the generation of too many replication errors and the loss of functional genetic information (7). However, variant RdRps with more stringent nucleotide selection criteria may not generate the genetic diversity that allows a virus population to escape the bottlenecks of the host's antiviral response. Indeed, it has been shown that viruses encoding variant RdRps with either higher or lower fidelity are attenuated in the host (2)(3)(4)(5)(6). In the case of poliovirus (PV), mice inoculated with these virus strains are also immunoprotected against lethal challenges of wild-type (WT) virus (3)(4)(5). An understanding of the fidelity mechanisms of RdRps would thus provide the framework for broad-spectrum antiviral strategies, including the design of new antiviral compounds and/or the generation of live, attenuated vaccine strains.
RdRps have the canonical "cupped right hand" structure with fingers, thumb, and palm subdomains and include seven highly conserved structural motifs, A-G, that are generally involved in interacting with RNA, nucleotide, and required metal ions (1) (Fig. 1). Recent X-ray crystal structures of PV (8,9) and enterovirus 71 (10) RdRps have been especially informative concerning the conformational changes that take place during the nucleotide selection and incorporation processes. All parts of the incoming nucleoside triphosphate (NTP) are proposed to be involved in binding and selection. These crystal structures suggest that after initial base pairing between the incoming * This work was supported by National Institutes of Health Grants AI104878 (to D. D. B.) and AI45818 (to C. E. C.). The authors declare that they have no conflicts of interest with the contents of this article. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. 1 Both authors contributed equally to this work. 2  NTP and the RNA template, the ribose hydroxyls interact with conserved residues Asp-238, Ser-288, and Asn-297 and other residues in motifs A and B (8 -10) and induce a conformational change in this region. There is then a re-alignment of the NTP triphosphate and conserved residue Asp-233 around the two metal ions (10). Residues in motif F, including Lys-167 and Arg-174, are also involved in binding the triphosphate. We have proposed that structural changes in the motif D active-site loop bring in Lys-359 to act as the general acid to protonate the ␤-phosphate and generate a better pyrophosphate leaving group (11)(12)(13); this may be one of the final steps during nucleotide addition.
Previous mutagenesis experiments have confirmed the importance of conserved residues in motifs A, B, and D to RdRp function and fidelity (10,(12)(13)(14)(15)(16)(17)(18)(19); similar studies for motif F residues have been largely lacking. Residues Lys-167 and Arg-174 probably play important functional roles, including initial nucleotide binding, realignment of the triphosphate for nucleophilic attack, and stabilization of the negative charges in the pentacoordinate transition state (9,16). Previous studies have indicated that the R174K substitution in coxsackievirus B3 RdRp does not lead to viable virus (6), and amino acid substitutions elsewhere in motif F can alter fidelity in bovine diarrheal virus RdRp (20).
In this paper, we have used NMR and kinetic studies to provide insight into the roles of motif F residues in the fidelity mechanism of PV RdRp. 31 P NMR studies have allowed us to probe the NTP triphosphate conformation for both cognate and noncognate NTP, and [methyl- 13 C]methionine NMR studies continue to provide insight into protein structural changes that accompany nucleotide selection, especially those associated with motifs B and D. The results with the motif F and previously identified fidelity variants suggest that triphosphate alignment is an important fidelity checkpoint but may be independent of conformational changes that occur in motifs B and D.

P NMR Spectra Used to Monitor NTP Conformations in PV
RdRp-We propose that there are (at least) four events to NTP selection and incorporation in RdRps: initial Watson-Crick base pairing with the RNA template, interactions between the NTP ribose hydroxyls and residues in motifs A and B to induce further conformational changes, a realignment of the NTP triphosphate involving residues in motifs A and F, and conformational changes in motif D to bring in the general acid Lys-359. Our previous NMR studies of PV RdRp using [methyl-13 C]Met-labeled protein established diagnostic probes for conformational changes in motifs B and D (also see Fig. 1C) (13,21,22). Here, we were interested in establishing probes for monitoring the NTP triphosphate conformation and alignment. Such probes would be important in evaluating the importance of motif F residues in RdRp catalysis and fidelity.
To examine more closely the triphosphate conformation and especially to compare conformations observed when correct versus incorrect nucleotide binds, we used 31 P NMR to probe the microenvironment surrounding the nucleotide and RNA binding pockets. Here, we have focused on the NTP triphos-phate region of the 31 P NMR spectra, which gives insight into the triphosphate binding environment in the various RdRp complexes; only very minor differences were observed for the RNA peaks (0 -1.2 ppm) in the various WT and variant RdRp complexes. Our studies took advantage of a symmetrical 10-mer RNA known as sym/sub-UA, which forms a 6-nucleotide duplex and 4-nucleotide overhangs (23) (Fig. 2A); the "UA" designates the first and second templating nucleobases. Using 31 P NMR, we were able to follow RdRp complex formation using our previously established procedures (13,21) (Fig. 2B). In this procedure, we directly added 3Ј-dATP to our NMR sample, such that PV RdRp catalyzed phosphodiester bond formation to the RNA substrate but then prevented further nucleotide addition reactions due to the lack of the 3Ј-OH. Following the reaction, we passed the RdRp⅐RNA sample across a desalting spin column to eliminate excess 3Ј-dATP that might interfere with the binding of subsequent nucleotides. Following the desalting column, the majority of the signal related to the binding of 3Ј-dATP disappeared (Fig. 2B), indicating that the RdRp⅐RNA complex was now free to bind the next nucleotide. We then added a second nucleotide (i.e. UTP) that binds, but does not react, at the next register of the RNA ( Fig. 2A). Passage across a second desalting column did not lead to the dissociation of the nucleotide (Fig. 2B). X-ray crystal structures had previously identified structural rearrangements that occur to "close" the active site upon nucleotide binding and addition (9). Presumably, these structural rearrangements and new interactions with the incoming NTP were responsible for retaining the UTP.
Consistent with this proposed procedure, additional peaks associated with the ␣-, ␤-, and ␥-phosphates appeared upon adding the second nucleotide, UTP (Fig. 2B). The assignments of these peaks were consistent with those from similar protein-NTP binding studies (24 -28). Close inspection of the 31 P spectrum indicated the presence of three intense peaks (␦ Ϫ4.98, Ϫ10.08, and Ϫ18.65 ppm), which we assigned to the free NTP not bound to the protein because these chemical shifts were identical to NTP (in Mg 2ϩ ) by itself (i.e. no protein, RNA; Fig.  2C), and two minor peaks around the ␣and ␥-phosphate positions (␦ Ϫ9.53 and Ϫ5.38 ppm), which we assigned to the NTP bound to the RdRp⅐RNA complex. After passage of the ternary RdRp⅐RNA⅐UTP complex across the desalting column, only the lower intensity, "protein-bound" peaks were retained (Fig. 2, B and C). The absence of a corresponding ␤-phosphate peak may reflect the low intensity of this peak and/or conformational exchange processes that would broaden out the ␤-phosphate peak associated with the protein-bound NTP.
Incorrect and Correct Incoming NTP Have Different Microenvironments-Some insight into the nature of the triphosphate binding pocket can be gained from comparisons of the 31 P NMR spectra of the RdRp⅐RNA⅐UTP complex and UTP by itself in the presence or absence of Mg 2ϩ (Fig. 2C). As observed previously (28), binding of Mg 2ϩ to free NTP led to similar downfield chemical shifts for both ␥-P and ␣-P resonances (Fig.  2C). In the RdRp⅐RNA⅐UTP complex, the ␣-P peak of the protein-bound nucleotide (␦ Ϫ9.53 ppm) was shifted even further downfield (Fig. 2C), potentially reflecting an even more electropositive environment in the active site with contributions from both Mg 2ϩ ions and basic amino acid side chains like Lys and Arg. Such interactions would also tend to create a more electrophilic ␣-P. In contrast, the ␥-P peak of the protein-bound nucleotide was upfield shifted compared with the ␥-P of Mg 2ϩ -UTP and had a chemical shift similar to that of the ␥-P of free UTP in the absence of Mg 2ϩ (Fig. 2C). Although this finding might suggest that the ␥-phosphate was no longer associated with Mg 2ϩ , we do not believe such a scenario is likely. It is generally believed that the ␤and ␥-phosphates are associated with Mg 2ϩ when binding the polymerase. Instead, a microenvironment may have been created around the ␥-phosphate that was more electron-withdrawing compared with free Mg 2ϩ -UTP, which may also lead to the weakening of the ␣-␤ phosphodiester bond; similar chemical shifts of the ␥-P for free UTP and the RdRp⅐RNA⅐UTP complex were probably fortuitous.
To gain insights into how the binding of incorrect nucleotide differs from correct nucleotide, we also collected 31 P NMR spectra of the RdRp⅐RNA⅐2Ј-dUTP and RdRp⅐RNA⅐CTP complexes ( Figs. 3 and 4). We first performed the RdRp⅐RNA⅐CTP experiments using different concentrations of CTP and at different temperatures. Here, we will focus first on the ␥-P peak(s). As CTP is added, a second ␥-P peak appears (␦ ϳϪ5.8 ppm at 4 mM), and then both ␥-P peaks begin to shift upfield as more CTP is added (␦ ϳϪ5.4 and 6.2 ppm at 12 mM CTP and 293 K) (Fig. 3). As CTP is added, we also observe a broadening of the ␥-P peaks (Fig. 3). To better understand what is happening, it is important to keep in mind that the incorrect nucleotide CTP is is shown with the conserved structural motifs colored (dark green, motif A; tan, motif B; dark cyan, motif C; blue, motif D; golden yellow, motif E; red, motif F; light green, motif G). Also highlighted are motif F residues Lys-167 and Arg-174 (red), motif D residues Lys-359 and Thr-362 (blue), and locations of other fidelity-altering substitutions (i.e. Gly-64 and His-273). B, conformational states of poliovirus RdRp based on X-ray crystal structures and MD simulations. The boxed region in A (left) is flipped 180º to improve clarity. The free state is in the absence of RNA and NTP (PDB entry 1RA6), the "binary" state is in the presence of RNA (PDB entry 3OL6), and postincorporation complex represents the "ternary" state (PDB entry 3OL7). All of the residues shown, including residues in motifs B (i.e. Ser-288 and Asn-297), D (i.e. Lys-359), and F (i.e. Lys-167 and Arg-174), undergo structural rearrangements from the binary to ternary complex. These residues make important interactions with the incoming NTP; Ser-288 and Asn-297 are important for ribose sugar recognition, and Lys-167, Arg-174, and Lys-359 are all predicted to interact with the triphosphate. We have also previously proposed that Lys-359 acts as a general acid to protonate the pyrophosphate leaving group. The "occluded" state is based on our previous MD simulations, which suggested that conserved residues that interact with the incoming NTP make interactions with each other, in the absence of NTP. less tightly associated with the RdRp than correct UTP, and as such, the spectra reflect an exchange between "free" (in solution) and protein-bound CTP. When the time scale of this exchange is similar to that of the chemical shift differences between free and protein-bound CTP, then we observe chemical shift changes for both the apparent free and protein-bound ␥-P peaks. In other words, the chemical shift of the free ␥-P peak is "contaminated" with contributions from the chemical shift of the protein-bound ␥-P peak (and vice versa), so the apparent free ␥-P peak shifts toward the apparent proteinbound ␥-P peak as more CTP is added. For variants with weaker association with CTP (e.g. K167R; see Fig. 4), the faster time scale of exchange can lead to the coalescence of the free and protein-bound peaks, leaving a single peak that represents a weighted population average of the underlying free and protein-bound peaks (see Ref. 29). A similar broadening/shifting of the ␣-P peaks does not occur because of the smaller chemical shift difference between the free and protein-bound ␣-P peaks. This behavior is also not observed for the WT RdRp⅐RNA⅐UTP complex probably because UTP is held more tightly and exchange on/off the enzyme is much slower. It should also be kept in mind that the 31 P NMR experiments were set up to minimize the signal from free (d)NTP, including a short recycle delay, precluding estimates of the proportion of incorrect (d)NTP bound to the protein and free in solution. For the ␥-P peaks for the RdRp⅐RNA⅐CTP complex, we assign the more intense (at ␦ ϳϪ5.4 ppm) and less intense (at ␦ ϳϪ6.2 ppm) peaks to the free and protein-bound CTP, respectively.
There were substantial chemical shift differences in the 31 P NMR spectra of RdRp⅐RNA⅐UTP (Fig. 4A) and the RdRp⅐RNA⅐2Ј-dUTP (Fig. 4B) and RdRp⅐RNA⅐CTP ( Fig. 4C) complexes, strongly suggesting that the orientation and/or binding microenvironments of the correct and incorrect nucleotides were substantially different between the two complexes. In particular, both the ␣-P (␦ Ϫ9.84 ppm for RdRp⅐RNA⅐2Ј-dUTP and Ϫ9.87 ppm for RdRp⅐RNA⅐CTP) and the ␥-P (␦ Ϫ6.21 ppm for RdRp⅐RNA⅐2Ј-dUTP and Ϫ6.16 ppm for RdRp⅐RNA⅐CTP) peaks of the incorrect nucleotide were not as downfield shifted as that of the correct nucleotide (Fig. 4). Besides exchange on/off the enzyme, conformational motions of the triphosphate of the incorrect nucleotide when bound to the RdRp may also have contributed to peak broadening; structural studies of DNA polymerases suggest that incorrect NTP can fluctuate among an ensemble of conformations within the polymerase active site to prevent alignment and reaction with the primer 3Ј-OH (30 -35).
Motif D and F Substitutions Change the Binding Environment of the Nucleotide Triphosphate-To gain more insight into the microenvironment surrounding the triphosphate moiety of the incoming nucleotide and the roles of amino acid residues interacting directly with the triphosphate, we also generated complexes for the K167R, R174K, and K359R variants (Fig. 4). Arg-174, Lys-359, and Lys-167 are proposed to interact primarily with the ␣-, ␤-, and ␥-phosphates, respectively (Fig. 1B). The FIGURE 2. Formation of RdRp RNA and nucleotide complexes can be followed using 31 P NMR. A, the indicated 10-mer RNA can be used as a symmetrical substrate-template ligand. Incorporation of 3Ј-dATP terminates RNA synthesis, allowing the addition of a second nucleotide, UTP, to form the ternary RdRp⅐RNA(3ЈH)⅐UTP complex. Spin desalting columns can be used to remove excess free nucleotide. B, the triphosphates of the nucleotides can be detected using 31 P NMR. Not shown are the peaks for the phosphate backbone of the RNA. The spectra for the RdRp complexes at 293 K (RdRp⅐RNAϩ3Ј-dATP, RdRp⅐RNA(3ЈH), RdRp⅐RNA(3ЈH)ϩUTP, and RdRp⅐RNA(3ЈH)⅐UTP (desalted) shown in golden yellow, magenta, green, and red, respectively) and 303 K (RdRp⅐RNA(3ЈH)⅐UTP (desalted); blue) are shown. C, these spectra are also compared with the spectra for free NTPs in the absence and presence of Mg 2ϩ , where temperatures are indicated on the plots. The assignments of the ␣-, ␤-, and ␥-phosphates are also shown. Spectra for protein complexes were collected with 250 M RdRp, 500 M RNA, 3.1 mM 3Ј-dATP, and 4 mM UTP, where indicated, using buffer consisting of 10 mM HEPES, pH 8.0, 200 mM NaCl, 0.02% NaN 3 , 5 mM MgCl, and 10 M ZnCl. Spectra for free NTPs (i.e. in the absence of protein and RNA) were collected under the same buffer conditions using 2 mM NTP. The spectra represent different concentrations of CTP added and/or different temperature, including ϩ2 mM CTP/293 K (golden yellow), ϩ4 mM CTP/293 K (magenta), ϩ12 mM CTP/293 K (green), ϩ12 mM CTP/298 K (red), and ϩ12 mM CTP/303 K (blue). It should be noted that the incorrect nucleotide CTP has a lower affinity than correct nucleotide UTP, such that CTP is exchanging on/off the enzyme. This behavior is illustrated by the apparent chemical shift changes for the ␥-P peak in response to different amounts of CTP and different temperatures. The spectra were collected with 250 M RdRp, 500 M RNA, 3.1 mM 3Ј-dATP, and the indicated CTP concentration using buffer consisting of 10 mM HEPES, pH 8.0, 200 mM NaCl, 0.02% NaN 3 , 5 mM MgCl, and 10 M ZnCl.
conservative amino acid changes should still allow these positively charged residues to interact with the NTP triphosphate, albeit in an altered fashion that could be monitored by 31 Arg-174 is probably important for the alignment of the ␣-phosphate for reaction with the primer 3Ј-OH. Consistent with the role of Arg-174 in interacting with the ␣-phosphate, the chemical shift for the protein-bound ␣-P was upfield shifted in the R174K variant (␦ Ϫ9.74 ppm) compared with WT RdRp (␦ Ϫ9.36 ppm). The chemical shift positions of the ␣-P and ␥-P for the R174K⅐RNA⅐UTP and R174K⅐RNA⅐CTP complexes were remarkably similar (Fig. 4); only the relative intensities and widths of the peaks were different. The changes in peak intensities may just reflect the difference in the concentrations of the nucleotides used in these experiments (i.e. [UTP] ϭ 4 mM; [CTP] ϭ 12 mM) and different exchange rates on/off the enzyme. Intriguingly, the R174K⅐RNA⅐2Ј-dUTP complex ( Fig.  4B) gave rise to three peaks associated with ␥-P (␦ Ϫ5.28, Ϫ5.89, and Ϫ6.32), suggesting that the ␥-phosphate for the "bound" 2Ј-dUTP might fluctuate between two or more conformations.
We have proposed that binding of correct, but not incorrect, nucleotide results in a structural rearrangement in the motif D loop to bring Lys-359 into the active site to interact with the ␤-phosphate (13). If so, we expected that the K359R substitution would lead to changes in the 31 P NMR spectrum of the correct RdRp⅐RNA⅐UTP complex but not the RdRp complexes bound with incorrect nucleotide. Consistent with this proposal, the K359R substitution led to small upfield chemical shift changes in the ␣-P and ␥-P peaks of the RdRp⅐RNA⅐UTP complex (␦ Ϫ9.49 and Ϫ5.43 ppm; Fig. 4A) but did not lead to substantial chemical shift changes in the RdRp⅐RNA⅐CTP spectrum (␦ Ϫ9.87 and Ϫ6.25 ppm; Fig. 4C) compared with the corresponding WT complexes. However, there was also a change in the ␥-P peaks for the K359R⅐RNA⅐2Ј-dUTP spectrum (␦ Ϫ5.40 and Ϫ6.03 ppm) compared with the WT complex (Fig.  4B). These changes may reflect different nucleotide exchange kinetics and/or changes to the binding microenvironment for 2Ј-dUTP once bound to the K359R variant.
The K167R substitution also led to differences in the spectra for the ternary complexes. In particular, there was one very broad ␥-P peak for the RdRp⅐RNA⅐2Ј-dUTP and RdRp⅐RNA⅐CTP complexes, which probably reflects changes in the nucleotide exchange kinetics. In other words, exchange on/off the enzyme is fast compared with the chemical shift differences between free and protein-bound NTP, such that the single ␥-P peak represents a weighted average of the free and protein-bound states. This proposal is consistent with the higher apparent nucleotide dissociation constants for the K167R variant compared with WT RdRp (see Table 1). The protein-bound peak intensities for the K167R⅐RNA⅐UTP complex were also quite weak, which might be indicative of decreased UTP binding and/or additional exchange processes on the enzyme that would broaden out these signals. These results were consistent with Lys-167 making important interactions with the triphosphate, which might be important for nucleotide binding affinity.
The R174K Substitution Also Prevents Motif D Structural Rearrangements as Monitored by Protein NMR-The 31 P NMR spectra for the K167R, R174K, and K359R variants indicated that these amino acid substitutions lead to changes in how the triphosphate interacts with the RdRp. We also collected NMR spectra of [methyl-13 C]Met-labeled RdRp to gain insight into how these amino acid substitutions change RdRp structural dynamics. In particular, we have previously noted that Met-354 is a good reporter of any structural and/or dynamic changes in the motif D loop (see also Fig. 1C), which we propose are important for repositioning Lys-359 for catalysis. Chemical shift changes to Met-354 are generally accompanied by chemical shift changes to other informative resonances, such as Met-6, Met-74, and Met-225 (Fig. 5). For the K167R variant, there were no substantial chemical shift differences compared with the corresponding WT complexes (Fig. 5, A-C), suggesting that the overall protein conformation was not substantially affected by this amino acid change. In contrast, binding of the correct UTP nucleotide to the R174K variant did not result in the same chemical shift changes as observed for WT RdRp (Fig. 5D). In fact, the chemical shift changes induced by binding of correct and incorrect nucleotides were similar for the R174K variant and similar to those changes induced by the binding of the incorrect nucleotide to WT RdRp (Fig. 5, E and F). These findings suggest that conformational changes in the motif D loop and other associated structural changes were severely impaired in the R174K variant. Altogether, these findings are consistent with our previous results that indicated that the ␣and ␤-phosphates but not the ␥-phosphate of the incoming nucleotide is required for the conformational changes in motif D and surrounding areas (13).
Motif F Substitutions Also Lead to Changes around Motif B-We have also previously used the Met-187 resonance as a probe to support molecular dynamics (MD) simulations suggesting that motif B and nearby regions fluctuate between conformations competent for and occluded from NTP binding (see also Fig. 1C) (22). In the occluded conformation (Fig. 1B), Arg-174 is involved in a salt-bridging interaction with Asp-238, such that this interaction and others must be broken before the RdRp⅐RNA complex can bind NTP (22); Met-187 is a reporter of these interactions and rearrangements (22). For the WT protein, there were distinct chemical shift changes when correct UTP and incorrect CTP were added to the RdRp⅐RNA complex (Fig. 6A). The K167R, R174K, and K359R substitutions all led to changes to the Met-187 resonances, especially for the RdRp⅐RNA and RdRp⅐RNA⅐CTP complexes, suggesting that these substitutions have long range effects on the structure/ dynamics of motif B and surrounding regions (Fig. 6). In contrast, the Met-187 resonances were very similar for the RdRp⅐RNA⅐UTP complexes (Fig. 6). The one exception was for the R174K variant, in which there were two additional resonances, where one of these overlapped that for the R174K RNA⅐CTP complex (Fig. 6C). Altogether, these results suggested that motif B and surrounding regions can attain a similar conformation when correct NTP binds for all variants, but the conformations in the absence of NTP or in the presence of incorrect NTP may differ.
Other changes to the Met-187 resonances were also informative. There were two resonances for the K167R RdRp⅐RNA⅐CTP complex, with one of those resonances more similar to what was observed for the K167R RdRp⅐RNA⅐UTP complex (Fig. 6B). There was no apparent Met-187 resonance for the R174K RdRp⅐RNA complex, suggesting conformational exchange on the intermediate NMR time scale (Fig. 6C). One potential explanation would be that the R174K substitution disrupts the interactions with Asp-238, such that the conformational exchange kinetics between NTP binding-competent and occluded conformations was moved to the intermediate NMR time scale. Finally, the Met-187 resonances for the K359R ternary complexes had similar chemical shift positions, in contrast to what was observed for WT RdRp (Fig. 6D). A comparable  scenario was observed previously for the low fidelity H273R variant, in which the H273R variant was proposed to favor the NTP binding-competent conformation and did not show any differences for the Met-187 resonance between the RdRp⅐RNA⅐UTP and RdRp⅐RNA⅐CTP complexes (22).

Comparison of the kinetic constants of nucleotide incorporation for variant and WT RdRp enzymes
The R174K Substitution Reduces Catalytic Efficiency and Increases Nucleotide Selection Fidelity-Not surprisingly, the 31 P and [methyl-13 C]Met spectra suggested that conserved amino acid substitutions at residues proposed to interact with the NTP triphosphate alter the nucleotide binding kinetics and/or conformation. To gain more insight, we analyzed the ability of RdRp variants to bind RNA and assemble into catalytically competent complexes and measured the maximum polymerase rate constant (k pol ) and the apparent dissociation constant for the incoming nucleotide (K d(app) ). To compare RNA binding between variant and WT RdRp, we monitored interactions with 6-carboxyfluorescein (6-FAM)-labeled sym/sub-UA RNA using a fluorescence polarization assay (Table 2). This assay indicated that most substitutions at Lys-167, Arg-174, and Lys-359 had little effect on RNA binding affinity ( Table 2). The sole exception was the R174L substitution, and the large decrease in RNA binding affinity precluded further assessment of this variant.
To ensure that the variants could assemble efficiently on the sym/sub-UA template and catalyze phosphodiester bond formation with the incoming nucleotide, we initiated the reaction with the addition of RdRp and followed RNA synthesis over various time points (Fig. 7A). The assay indicated that the variants could assemble for catalysis in a similar time frame as WT RdRp. We also compared the stabilities of the RdRp⅐RNA complexes between variant and WT RdRps by incubating RdRp enzyme with labeled sym/sub-UA for 90 s to allow the RdRp⅐RNA binary complexes to form before the addition of excess unlabeled sym/sub-UA to trap free and dissociating enzyme (Fig. 7B). ATP was then added at various times after the addition of the unlabeled sym/sub-UA, and the reaction was quenched 30 s later. This assay indicated that the R174K variant dissociated more quickly from the labeled RNA compared with WT RdRp (Table 2). Nonetheless, the RdRp⅐RNA complex for R174K RdRp was deemed sufficiently stable for our single nucleotide incorporation assays to measure k pol and K d(app) values. It is important to note that we also checked the RdRp catalytic activity under buffer conditions that we used for the 31 P and [methyl-13 C]Met NMR experiments, which did not reveal any substantial differences compared with the buffer system we used for our kinetic assays (Fig. 8).
The K167R substitution had little effect on the k pol and K d(app) values for correct nucleotide incorporation (i.e. ATP, in the presence of Mg 2ϩ ) but led to increases in k pol , K d(app) , and catalytic efficiency (k pol /K d(app) ) for incorrect nucleotide incorporation compared with WT RdRp ( Table 1). As mentioned above, the change in K d(app) for incorrect nucleotide may help to explain the broader peaks in the 31 P NMR spectrum for the K167R⅐RNA⅐2Ј-dUTP and K167R⅐RNA⅐CTP complexes (Fig.  4), which suggests a change in the binding on/off rates. These results also indicate that the K167R variant is less able to discriminate against incorrect nucleotide and has lower fidelity compared with WT RdRp. The K167M variant also had a substantial increase in the K d(app) for correct nucleotide. The higher K d(app) values precluded further kinetic assessment of incorrect nucleotide incorporation for this variant.
Perhaps not surprisingly, the R174K variant led to a 14-fold drop in k pol for correct nucleotide incorporation (i.e. ATP, in the presence of Mg 2ϩ ) compared with WT RdRp (Table 1). More remarkable was the finding that the change in k pol was even more substantial for incorrect nucleotide incorporation, either with incorrect sugar (i.e. 2Ј-dATP) or incorrect nucleobase (i.e. GTP). Specifically, there were 3,700-and 370-fold drops in k pol for 2Ј-dATP and GTP incorporation, respectively, compared with WT RdRp (k pol for GTP was estimated based on using 1000 M GTP (i.e. almost 10-fold greater than the K d(app) for WT RdRp; Table 1). These k pol values also indicated that the R174K variant was more faithful (i.e. k pol(correct) /k pol(incorrect) values were 17,500 and 140,000 for 2Ј-dATP and GTP, respectively) than WT RdRp (k pol(correct) /k pol(incorrect) values were 70 and 5400 for 2Ј-dATP and GTP, respectively).

Metal and Solvent Effects Help to Gauge the Effects of Amino Acid Substitutions on the Prechemistry Conformational Change and Chemical
Steps-To gain more insight into the effects of these amino acid substitutions, we also assayed the effects of  changing the metal ion and the solvent. In the presence of Mg 2ϩ , the incorporation of correct nucleotide catalyzed by the WT enzyme is partially rate limited by both a prechemistry conformational change and the chemistry step itself (36). The prechemistry conformational change and the chemical step also both contribute to discrimination against nucleotide with incorrect sugar or incorrect nucleobase (36). In contrast, correct nucleotide incorporation catalyzed by the WT enzyme is more rate-limited by the chemical step in the presence of Mn 2ϩ , presumably because the correct nucleotide more readily induces the prechemistry conformational change with this metal (37). However, the chemistry step is less discriminatory in the presence of Mn 2ϩ , and the prechemistry step largely governs polymerase fidelity (37). The relative contribution of the chemistry step can also be assayed by performing the assays in D 2 O, because the solvent deuterium kinetic isotope effect  (SDKIE; i.e. k pol(H2O) /k pol(D2O) ) reports on proton transfer steps occurring during phosphodiester bond formation (11). The SDKIEs for the K167R and R174K variants were higher than WT RdRp in the presence of Mg 2ϩ (4.67 Ϯ 0.40, 2.47 Ϯ 0.05, and 2.19 Ϯ 0.03 for K167R, R174K, and WT RdRp, respectively) or Mn 2ϩ (6.77 Ϯ 0.40, 4.25 Ϯ 0.02, and 3.02 Ϯ 0.03 for K167R, R174K, and WT RdRp, respectively) ( Table 1). These results suggest that the K167R and R174K substitutions both decreased the rate of the chemistry step. Nucleotide discrimination for the K167R and R174K variants was more similar to that of WT RdRp in Mn 2ϩ (i.e. k pol(ATP) /k pol(2Ј-dATP) values are 1.4, 3.4, and 2.2 for K167R, R174K, and WT RdRp respectively; k pol(ATP) /k pol(GTP) values are 11, 15, and 20) than Mg 2ϩ . These results suggest that the differences in RdRp fidelity in the presence of Mg 2ϩ are probably due to the K167R and R174K substitutions primarily affecting the chemical step for incorrect nucleotide incorporation, such that most of these effects disappear with Mn 2ϩ .
Kinetic Experiments Indicate That the R174K Variant Exchanges Nucleotides More Readily-The NMR experiments suggested that the NTP and protein structural changes for R174K variant were substantially different from that observed for WT enzyme. To gain more insight, AMP-CPP (or buffer as control) was preincubated with RdRp⅐RNA binary complexes for 5 min to trap the "closed" state by forming RdRp⅐RNA⅐AMP-CPP ternary complexes, and then the reaction was initiated by the addition of an excess amount of ATP (Fig. 9). The reaction occurs only when AMP-CPP is exchanged with ATP. The rate for nucleotide incorporation was determined with and without AMP-CPP. The delay in nucleotide incorporation induced by AMP-CPP is a measure of the reverse rate of the prechemistry conformational change. For WT and K167R RdRps, the rates were delayed by 40 -50-fold (Fig. 9). However, in R174K RdRp, the rate was delayed by only 3-fold (Fig. 9). These results suggested that the R174K substitution destabilized the "closed" state of the ternary complex, consistent with the NMR experiments (Figs. 4 -6). The slower rate of catalysis together with the enhanced ability to exchange nucleotides probably provide additional opportunities for the R174K variant to release incorrect nucleotide, leading to a higher fidelity enzyme.
The Functional Roles of Arg-174 and Lys-359 Are Largely Independent-It was shown previously that the K359R and K359L substitutions lead to decreases in k pol and increases in polymerase fidelity (5,12,13). Because both Arg-174 and Lys-359 (13) appear to be important for the conformational changes in motif D observed by NMR (Fig. 5), we wanted to understand whether the roles of these residues were related or independent from one another. To address this question, we compared the kinetics of the variants with single amino acid substitutions (i.e. R174K and K359L) with the variant with both amino acid substitutions (i.e. R174K/K359L). RdRp with a nonpolar residue at position 359 (i.e. K359M) does not undergo the same conformational changes as WT enzyme (13), similar to what is observed for the R174K variant (Fig. 5). The R174K/K359L variant demonstrated decreased affinity to RNA ( Table 2) and was much more impaired in its ability to incorporate correct ATP nucleotide compared with either of the two single variants (Table 1).
To test whether the amino acid substitutions were thermodynamically coupled and hence whether the function of the residues were interdependent, we compared the free energy changes induced in the catalytic efficiency (i.e. ⌬⌬G ϭ ϪRT ln ((k pol /K d(app) ) variant /(k pol /K d(app) ,app) WT ) by the single and double amino acid substitutions, analogous to the classic double mutant cycle experiments performed by Fersht (38), Mildvan et al. (39), and Serrano et al. (40). The effects of the R174K (⌬⌬G ϭ 1.2 kcal/mol) and K359L (⌬⌬G ϭ 4.0 kcal/mol) substitutions were nearly additive (assuming ϳ20% error) compared with the R174K/K359L double amino acid substitution (⌬⌬G ϭ 6.0 kcal/mol), suggesting that these residues are very weakly thermodynamically coupled and that their roles are largely independent from one another. If the functions of these residues were interrelated, we would have expected a non-additive response. The differences in the NMR spectra for the R174K and K359R variants also suggest that these amino acid substitutions have substantially different effects on the nucleotide and protein structural dynamics.
Fidelity Variants Substantially Alter Triphosphate Binding and Reorientation-Our NMR and kinetic studies indicated that amino acid substitutions at residues that interact with the NTP triphosphate lead to changes in catalytic efficiency and polymerase fidelity. We were interested in testing whether other amino acid substitutions that change polymerase fidelity The black solid lines represent fits of the data to single exponential curves. and attenuate the virus also lead to changes to the triphosphatebinding environment. In particular, the G64S and H273R substitutions are relatively remote from the active site (Ͼ20 Å away from the catalytic center) but generate higher and lower fidelity polymerases, respectively (41)(42)(43). Although these amino acid changes are relatively remote from the active site, interaction pathways suggest how these substitutions might affect RdRp catalysis and fidelity. For example, Gly-64 is involved in a hydrogen bonding network involving Ala-239 and Leu-241 in motif A and Gly-285 in motif B. His-273 makes hydrogen bond interactions with Thr-156 in a ␤-strand that makes interactions with the ␤-strand containing Arg-174. MD simulations have also shown how the G64S and H273R substitutions affect structural dynamics around the active site and more globally (22,44). Another intriguing fidelity variant is T362I RdRp (18). We were initially interested in this amino acid substitution because it is encoded by the Sabin I vaccine strain and is located in motif D, near Lys-359 (45). Our studies indicated that the T362I substitution leads to a lower fidelity polymerase, probably because it can access a more active conformation even in the presence of incorrect nucleotide (18). The T362I substitution by itself (i.e. in the absence of other Sabin I mutations) also leads to some attenuation of the encoding PV strain (18). In the presence of other Sabin I substitutions, these effects are largely mitigated (46). Nonetheless, the T362I variant can serve as an important tool in understanding how the conformational dynamics of motif D are (un)related to the realignment of the triphosphate.
We assessed protein structural changes in the G64S, H273R, and T362I variants by using [methyl-13 C]Met NMR. We were especially interested in tracking the Met-354 resonance, as it responds to structural changes in motif D and surrounding areas (Fig. 1). The G64S substitution induced only very small chemical shift differences in the [methyl-13 C]Met spectra for the RdRp⅐RNA⅐UTP, RdRp⅐RNA⅐2Ј-dUTP, and RdRp⅐RNA⅐CTP complexes (Fig. 10A) compared with WT RdRp (Fig. 5). In contrast, the H273R⅐RNA⅐2Ј-dUTP, T362I⅐RNA⅐2Ј-dUTP, and T362I⅐RNA⅐CTP complexes all led to spectral changes compared with the same WT and G64S complexes (Fig. 10, B and C). We had previously noted that the T362I variant gives rise to two sets of resonances for the RdRp⅐RNA⅐2Ј-dUTP complex (18); one set of resonances overlaps what is observed for the WT⅐RNA⅐UTP complex, and the other set of resonances overlaps what is observed for the WT⅐RNA⅐CTP complex. This finding prompted us to suggest that the T362I variant can access a more active conformation even in the presence of the incorrect 2Ј-dUTP (18), suggesting why it is a lower fidelity variant. The T362I⅐RNA⅐CTP complex also appeared to fluctuate into two (or more) conformations, considering the presence of two Met-354 resonances.
Our previous studies also highlighted the effects of these substitutions on the Met-187 resonance, as a reporter on motif B and surrounding regions (22). The G64S spectra were similar to those of WT RdRp (22). As mentioned previously, the Met-187 resonances for the RdRp⅐RNA, RdRp⅐RNA⅐UTP, and RdRp⅐RNA⅐CTP complexes for the H273R variant all overlapped, consistent with the H273R substitution stabilizing the NTP binding-competent conformation, helping to bypass this fidelity checkpoint (22). For T362I, the Met-187 resonance for the RdRp⅐RNA⅐UTP complex was similar to that for WT enzyme, but the Met-187 resonance for the RdRp⅐RNA⅐CTP complex was at a chemical shift position different from that for WT RdRp (22).
To test the effects of the G64S, H273R, and T362I substitutions on the binding and microenvironment of the NTP triphosphate, we collected 31 P NMR spectra of their correct RdRp⅐RNA⅐UTP and incorrect RdRp⅐RNA⅐2Ј-dUTP and RdRp⅐RNA⅐CTP complexes (Fig. 11). It should be kept in mind that any changes in the 31 P NMR spectra of the G64S, H273R, and T362I variants must be due to indirect effects, because these residues do not make direct interactions with RNA or NTP. Nonetheless, there were substantial changes in the 31 P NMR spectra for these variants compared with WT RdRp, especially for the G64S and H273R variants (Fig. 11). In particular, the H273R substitution induced chemical shift changes for the ␣-P and ␥-P peaks in the RdRp⅐RNA⅐UTP complex (Fig.  11A), suggesting that correct nucleotide may bind in a different manner, and/or the surrounding microenvironment has been altered in the H273R variant. These effects might be due to the series of interactions connecting His-273 to the ␤-strand containing Arg-174, as suggested above, and/or changes in the structural dynamics around the active site, as observed previously in MD simulations (22).
There were also differences in the 31 P spectra for the G64S and H273R variants in their respective RdRp⅐RNA⅐2Ј-dUTP and RdRp⅐RNA⅐CTP complexes compared with WT enzyme (Fig. 11, B and C). Intriguingly, both G64S and H273R variants gave rise to additional ␥-P peaks in the RdRp⅐RNA⅐2Ј-dUTP complexes. We do not think these additional peaks correspond to additional binding pockets created by these variants because these substitutions are not on the surface or near any binding sites. Instead, we suggest that these additional peaks probably correspond to the ␥-phosphate of protein-bound 2Ј-dUTP fluctuating between different conformations and/or microenvironments. In the G64S⅐RNA⅐CTP complex, the ␥-P peak for the G64S variant was downfield shifted (␦ Ϫ5.87 ppm) and appeared to be more well resolved (i.e. the doublet was apparent) compared with what is observed for WT RdRp (Fig. 11C). These findings suggest that the binding environment and/or conformational dynamics of the triphosphate in the G64S variant are different from those of WT RdRp. Likewise, the H273R substitution also led to chemical shift differences in the ␣-P (␦ Ϫ9.77 ppm) and ␥-P (␦ Ϫ6.30 ppm) peaks for the RdRp⅐RNA⅐CTP complex, although these changes were distinct from what was observed in the G64S variant (Fig. 11C). Most notably, there was less of a difference in the chemical shift of the ␣-P peaks between the RdRp⅐RNA⅐CTP and RdRp⅐RNA⅐UTP complexes (␦ Ϫ9.46 ppm) compared with the same complexes for WT enzyme. This result might suggest that the ␣-phosphates for these complexes were in more comparable chemical environments for the H273R variant compared with WT RdRp. If true, this finding would help to explain why the H273R variant has a lower ability to discriminate between correct and incorrect nucleotides. This result is also consistent with the proposal that the H273R substitution stabilizes the NTP-binding competent conformation, helping to bypass this fidelity checkpoint (22).
In contrast to what is observed for the G64S and H273R variants, the T362I substitution did not lead to additional ␥-P peaks for the RdRp⅐RNA⅐2Ј-dUTP complex. In fact, the 31 P spectrum was very similar to that for WT RdRp (Fig. 11), which was in stark contrast to the corresponding [methyl-13 C]Met spectra (Fig. 8). The T362I substitution also induced a chemical shift change in the ␥-P peak (␦ Ϫ5.90 ppm) compared with what was observed for the WT enzyme (Fig. 11C), although this might just reflect a change in the CTP exchange kinetics on/off the enzyme (i.e. the ␥-P peaks are less broad and more intense for the T362I variant). The T362I substitution also induced changes in the RdRp⅐RNA⅐UTP spectra, which was a bit surprising considering that the corresponding [methyl-13 C]Met spectra were very similar to WT enzyme. These findings sug- gest that the T362I substitution can have differing effects on the NTP triphosphate and motif B/D structural dynamics.

Discussion
Enzymes within the nucleic acid polymerase superfamily appear to share a common kinetic mechanism, including a conformational change into a more active form before catalyzing phosphodiester bond formation (1). There has been much debate over the structural details of this conformational change (47). It is likely that the prechemistry conformational change identified in these kinetic mechanisms actually represents a number of conformational events. In RdRps, there are structural rearrangements in motifs A and B to interact with the NTP ribose hydroxyls, a realignment of the NTP triphosphate for proper orientation, and a repositioning of the motif D lysine (Lys-359 in PV RdRp) to act as a general acid to protonate the pyrophosphate leaving group. We have now developed diagnostic NMR experiments to probe each of these structural changes, allowing us to investigate the nucleotide incorporation cycle for WT and fidelity-altering variants. Specifically, the 31 P NMR experiments that we have developed here probe the binding environment of the NTP triphosphate, and our previously developed [methyl-13 C]Met probes monitor structural changes in motifs B and D (Fig. 1C) and surrounding areas. Our studies provide insight into not only which fidelity checkpoints are altered by these amino acid substitutions, but provide information about the connections between different fidelity checkpoints through comparisons of fidelity variants (Fig. 12).
In this report, we have developed a 31 P NMR assay to probe the binding environment of the NTP triphosphate (Fig. 2). Crystal structures of nucleic acid polymerase superfamily members have indicated that there is a reorientation of the triphosphate of the incoming nucleotide following initial binding to properly align the ␣-phosphate for nucleophilic attack by the primer 3Ј-OH (10, 48) (Fig. 12). Having a mispaired primer terminus and/or binding incorrect nucleotide does not result in the same structural rearrangement (32)(33)(34)49). As such, the reorientation of the triphosphate probably reflects an important fidelity checkpoint (10, 37, 50 -52). Our 31 P NMR studies on PV RdRp are consistent with these proposals, where the binding of correct and incorrect NTP led to different 31 P NMR signatures (Figs. 3 and 4). In the interpretation of these experiments, it should be kept in mind that the spectra can represent a mixture of states. If the triphosphate is fluctuating between different conformations on a time scale fast relative to the chemical shift differences between these states, we would FIGURE 12. Nucleotide selection checkpoints altered by PV RdRp fidelity variants. Conformational changes important for nucleotide incorporation include a reorientation of the triphosphate of the incoming nucleotide for direct in-line attack by the RNA 3Ј-OH, a conformational change in structural motif D to reposition conserved residue Lys-359, and conformational changes in structural motif B to allow for interactions with the 2Ј-OH of the incoming nucleotide. We have developed diagnostic NMR probes for each of these rearrangements; our 31 P NMR studies provide insight into the environment surrounding the triphosphate of the incoming nucleotide, and our [methyl-13 C]Met NMR studies provide probes for structural changes in motifs B (i.e. Met-187) and D (i.e. Met-354). Our results indicate that the fidelity-altering amino acid substitutions differentially affect these checkpoints, suggesting that the checkpoints are not strictly dependent on each other. observe only a single set of peaks for protein-bound NTP that represents a weighted average of these different states. Changing residues proposed to interact with the NTP triphosphate led to changes in rates and fidelity of nucleotide addition (Table  1). Interestingly, modification of Arg-174 and Lys-359, proposed to interact with the ␣and ␤-phosphates, respectively, increased fidelity, whereas modification of Lys-167, proposed to interact with the ␥-phosphate, decreased fidelity. The SDKIE and metal effects suggested that these modifications primarily affected the chemistry step itself.
We have also suggested that conformational changes to reposition Lys-359 for catalysis, analogous to the conformational change in the O/P helix in the A/B family DNA polymerases (32,34,(53)(54)(55)(56)(57)(58)(59)(60)(61), represent an important fidelity checkpoint. The Met-354 resonance has been diagnostic of conformational changes in motif D (13,21) (Figs. 1C and 5). The chemical shift changes experienced by Met-354 are absolutely dependent on the protonation state of Lys-359 (13); only under conditions in which the side-chain amine of Lys-359 is protonated do these structural changes occur. We do not observe similar chemical shift changes upon binding incorrect nucleotide. Consistent with this proposal, the K359R substitution changed the 31 P NMR spectrum relative to WT for the correct RdRp⅐RNA⅐UTP complex but not for the incorrect RdRp⅐RNA⅐CTP complex (Fig. 4). This result suggested that this residue was only brought near the NTP triphosphate when correct NTP was present. Unfortunately, PV RdRp crystal structures have failed to capture Lys-359 in a position to protonate the pyrophosphate leaving group. We note here that the 31 P and 1 H-13 C spectra (Figs. 4 and 5) and the AMP-CPP trap assay ( Fig. 9) indicated that the R174K variant was impaired in its ability to fluctuate into a closed conformation. These results might indicate that R174K is capable of catalyzing phosphodiester bond formation through an open-like conformation (i.e. when Lys-359 is not in a position to donate a proton to the ␤-phosphate). Likewise, catalysis in the RdRp crystals may not require the repositioning of Lys-359. It is known that Lys-359 is not absolutely required for catalysis, considering that amino acid substitutions at this position only decrease the catalytic rate by up to 50-fold (12).
Structural rearrangements in motifs A/B upon interacting with the NTP ribose are also well known to be important for RdRp fidelity (62). The Met-187 resonance serves as a diagnostic probe for structure/dynamic changes in motif B and surrounding areas (Fig. 1C). It is intriguing that most of the fidelity variants resulted in some change to the Met-187 resonances (Fig. 6) (22). The pervasiveness of these changes to Met-187 and its surrounding environment are consistent with motif A/Bribose interactions being one of the first steps in NTP binding and selection, necessary to trigger subsequent conformational changes.
One question regarding RdRp fidelity is whether these fidelity checkpoints, involving structural changes in motifs A/B induced by interacting with the NTP ribose, rearrangement of the NTP triphosphate, and repositioning of the general acid Lys-359, are interdependent. Our NMR and kinetic analyses suggest that fidelity-altering substitutions have differing, potentially independent effects on these checkpoints. For example, fidelity variants may change interactions with the NTP triphosphate or lead to structure/dynamic changes in motif D, but not necessarily both (Fig. 12). In these cases, it appears that changes in the NTP triphosphate binding/conformation and structure/dynamics in motif D are not strictly coupled. For instance, there were additional ␥-P peaks for the G64S⅐RNA⅐2Ј-dUTP and H273R⅐RNA⅐2Ј-dUTP complexes (Fig. 11), but these were not accompanied by additional peaks in the [methyl-13 C]Met spectra (Fig. 10). Likewise, there were additional resonances in the [methyl-13 C]Met spectra, but not the 31 P spectra, for the T362I⅐RNA⅐2Ј-dUTP and T362I⅐RNA⅐ CTP complexes. These results were also consistent with the finding that amino acid substitutions at Arg-174 and Lys-359 were not thermodynamically coupled, according to double mutant cycle analysis (Table 1). Likewise, the behavior (i.e. chemical shift, number of resonances) of the Met-187 resonance(s) (Fig. 6) (22) was not coupled to what was observed for the Met-354 resonance (Figs. 5 and 10) or the 31 P triphosphate spectra (Figs. 4 and 11).
Altogether, our results are consistent with the proposal that the realignment of the NTP triphosphate is an important fidelity checkpoint, but this event is not strictly coupled to other fidelity checkpoints in the RdRp (Fig. 12). This finding is especially interesting in comparison with what is found in the A/B family DNA polymerases, in which the positively charged residues that interact with the NTP triphosphate are all located on or near the O/P helix. In RdRps, it appears that the positively charged amino acids interacting primarily with the ␣and ␥-phosphates (e.g. Lys-167 and Arg-174 in motif F) are segregated from the positively charged residue that interacts with the ␤-phosphate and acts as the general acid (i.e. Lys-359 on motif D). Our findings suggest that the binding and alignment of the triphosphate necessary for the bond formation between the ␣-phosphate and the primer 3Ј-O group are effectively separated in time and space from the repositioning of the general acid whose actions facilitate bond breakage between the ␣and ␤-phosphates. The separation of these fidelity checkpoints probably allows independent verification of the incoming NTP, which might be especially important for polymerases that lack proofreading domains.
Site-directed Mutagenesis, Protein Overexpression, and Purification of PV RdRp-All mutants were generated using the QuikChange method (Stratagene) and appropriate primers. Mutations were confirmed by DNA sequencing (Nucleic Acid Facility, Pennsylvania State University). It should be noted that WT and all variant RdRps also have L446D and R455D substitutions to prevent RdRp self-association. Overexpression and purification of PV RdRp followed procedures described previously (21,23,63,64).
NMR Sample Preparation and Spectroscopy-NMR samples of the different complexes (free protein, bound to RNA, bound to RNA and NTP, and so on) were prepared as described previously (13,21). The NMR buffer consists of 10 mM HEPES, pH 8.0, 200 mM NaCl, 0.02% NaN 3 , 5 mM MgCl, and 10 M ZnCl. For ⑀-13 CH 3 Met-labeled protein, samples were in 100% D 2 O, but samples generated for 31 P NMR contained 90% H 2 O and 10% D 2 O for the lock signal. Data collection for the ⑀-13 CH 3 Met-labeled samples followed previous procedures (13,21). The 31 P NMR spectra were collected using a Bruker Avance III 600-MHz NMR spectrometer with a 31 P-selective probe at spectrometer frequencies of 600.07 MHz for 1 H and 242.91 MHz for 31 P. The experiments were performed with fully calibrated 90º pulses and used a recycle delay of 1 s to suppress signal from free-floating RNA and NTP. Protons were decoupled during data acquisition using the Waltz16 sequence. The 31 P chemical shifts are expressed with reference to an 85% phosphoric acid external standard.
RNA Binding Assays-Fluorescence polarization assays were conducted to determine sym/sub-UA binding affinity of variant and WT RdRp. Reactions contained 50 mM HEPES, pH 7.5, 10 mM 2-mercaptoethanol, 5 mM MgCl 2 , 60 M ZnCl 2 , and 10 mM NaCl. Reactions were conducted at room temperature. Enzyme was added immediately from ice, and minipolarization was measured after 10 min from the addition to reach the binding equilibrium. The sym/sub-UA was modified with 6-FAM at the 5Ј-end of the oligonucleotide. The concentration of 6-FAMsym/sub-UA was 1 nM. The volume of enzyme added into the reaction was no more than one-tenth of the total assay volume.