Kinetic Effects of β,γ-Modified Deoxynucleoside 5′-Triphosphate Analogues on RNA-Catalyzed Polymerization of DNA

Abstract

A recently described DNA polymerase ribozyme, obtained by in vitro evolution, provides the opportunity to investigate mechanistic features of RNA catalysis using methods that previously had only been applied to DNA polymerase proteins. Insight can be gained into the transition state of the DNA polymerization reaction by studying the behavior of various β,γ-bridging substituted methylene (CXY; X, Y = H, halo, methyl) or imido (NH) dNTP analogues that differ with regard to the pK_a4 of the bisphosphonate or imidodiphosphate leaving group. The apparent rate constant (k_pol) of the polymerase ribozyme was determined for analogues of dGTP and dCTP that span a broad range of acidities for the leaving group, ranging from 7.8 for the CF₂-bisphosphonate to 11.6 for the CHCH₃-bisphosphonate. A Brønsted plot of log(k_pol) versus pK_a4 of the leaving group demonstrates linear free energy relationships (LFERs) for dihalo-, monohalo-, and non-halogen-substituted analogues of the dNTPs, with negative slopes, as has been observed for DNA polymerase proteins. The unsubstituted dNTPs have a faster catalytic rate than would be predicted from consideration of the linear free energy relationship alone, presumably due to a relatively more favorable interaction of the β,γ-bridging oxygen within the active site. Although the DNA polymerase ribozyme is considerably slower than DNA polymerase proteins, it exhibits a similar LFER fingerprint, suggesting mechanistic commonality pertaining to the buildup of negative charge in the transition state, despite the very different chemical compositions of the two catalysts.

A combination of features in modern biology suggest the possibility of a prior “RNA world”, in which genomes were comprised of RNA rather than DNA and biochemical processes were catalyzed by RNA rather than protein enzymes.^1–4 In vitro evolution has emerged as a powerful tool for obtaining functional RNA molecules thought to be relevant to the RNA world. For example, one of the first in vitro evolved RNA enzymes is the class I ligase, which catalyzes the formation of an RNA phosphodiester linkage by facilitating nucleophilic attack of the 3′-hydroxyl of a template-bound oligonucleotide on the 5′-triphosphate of an adjacent oligonucleotide.⁵ The class I ligase also has a limited ability to catalyze the RNA-templated polymerization of nucleoside 5′-triphosphates (NTPs),⁶ an activity that is relevant to RNA self-replication and has been substantially improved through many additional rounds of in vitro evolution.^7–11 Recently, the most advanced forms of the class I RNA polymerase were found also to function as a reverse transcriptase, copying an RNA template to yield a complementary DNA product through the incorporation of deoxynucleoside 5′-triphosphates (dNTPs).^12,13

As probes for the mechanism of various protein DNA polymerases, a family of dNTP analogues in which the β,γ-bridging oxygen is replaced by a substituted methylene (CXY) or imido (NH) group has been synthesized^14–20 to study the effect of the leaving group structure on the kinetics of templated polymerization reactions. These nucleotide probes generally react in the same manner as standard dNTPs, resulting in the formation of a phosphodiester linkage between the 3′-hydroxyl of a template-bound primer and the α-phosphate of the incoming dNTP. Differences in reaction rates with the various derivatives can plausibly be attributed to the effect of the β,γ-substituent in stabilizing the buildup of negative charge in the transition state. For example, when these analogues were used to study the mechanism of polymerase β (pol β), the rate of polymerization was found to decrease as the pK_a4 of the bisphosphonate leaving group increased.^15,19,21,22 This linear free energy relationship (LFER) can be seen in a Brønsted plot that relates log(k_pol) to the pK_a4 of the leaving group. Similar LFER profiles have been observed in studies of DNA polymerases λ and η.¹⁹

There have been limited mechanistic studies of the class I polymerase, in part due to the lack of high-resolution structural information. There are two reported X-ray crystal structures of the parental class I ligase,^23,24 which lacks the added “accessory domain” that is present in the class I polymerase. Cross-linking studies suggest a close interaction between a loop region of the accessory domain and an unpaired region within the core ligase domain,²⁵ but the overall structure of the polymerase is unknown. Thus, the present study sought to address an aspect of catalytic mechanism without the benefit of structural information, turning instead to kinetic studies using β,γ-modified dNTP analogues and employing the reverse transcriptase ribozyme.

The most recently reported form of the class I polymerase ribozyme is the 38–6 variant,¹¹ which was obtained after 38 rounds of in vitro evolution, selecting for the ability to synthesize functional RNA products. Both the prior 24–3 polymerase¹² and the 38–6 polymerase¹³ have the ability to synthesize DNA, employing an RNA template and either an RNA or DNA primer. The reaction with an RNA primer proceeds so rapidly that it would be challenging to measure the rate of dNTP addition for all the analogues used in this study. Instead a DNA primer was used, which enabled the rate to be measured by manual pipetting methods, also more closely mimicking the extensively studied reactions involving extension of a DNA primer by various DNA polymerase proteins.

The class I polymerase recognizes the RNA template in part through a “processivity tag”, located at the 5′ end of the ribozyme, that binds a complementary region at the 5′ end of the template⁹ (Figure 1). The polymerase requires high concentrations of Mg²⁺, typically 200 mM, and operates in a temperature range of 17–30 °C and pH range of 7.0–8.5.^9–12 For the kinetic assays in this study, a 5′-fluorescein labeled DNA primer was used, and RNA templates were chosen that allow single-nucleotide addition of either dGTP or dCTP. The downstream template sequence was designed to prevent more than a single addition, and only the requisite dNTP was present in the reaction mixture. A variety of β,γ-substituted dGTP and dCTP analogues were tested (Figure 2), measuring the time course of the reaction under pre-steady-state, single-turnover conditions (see Supporting Information for experimental procedures).

Figure 1.

Sequence and secondary structure of the 38–6 form of the class I polymerase.¹¹ The ribozyme is shown in black, the RNA template in brown, the DNA primer in magenta, and the incoming dNTP in cyan. Stem elements P3–P7 of the ligase domain and A3–A4 of the accessory domain are labeled.

Figure 2.

Addition of various β,γ-bridging dNTP analogues to the 3′ end of a template-bound DNA primer, with concomitant release of pyrophosphate or corresponding analogue. The pK_a4 values of the bisphosphonate or imidodiphosphate leaving groups, as previously determined experimentally,²⁰ are 7.8 for CF₂, 8.4 for CFCl, 8.8 for CCl₂, 8.9 for O (pyrophosphate), 9.0 for CHF, 9.3 for CBr₂, 9.5 for CHCl, 9.7 for NH, 9.9 for CHBr, 10.5 for CH₂, and 11.6 for CHCH₃. Where X has a stereogenic center,^18,20 the dNTP analogue is a ~1:1 mixture of diastereomers.²⁰

Employing unmodified dGTP or dCTP, a linear kinetic regime was defined that corresponds to extension of <10% of the template-bound primer by the addition of a single deoxynucleotide. This regime enabled determination of an apparent chemical rate constant (k_pol) under single-turnover, pre-steady-state conditions. The reaction was quenched at various times by adding EDTA in excess of Mg²⁺, and the products were analyzed by denaturing polyacrylamide gel electrophoresis (PAGE), calculating the fraction of primer that had been extended. Both a biotin moiety and fluorescein label were installed at the 5′ end of the primer. The former enabled capture of both the primer and extension products on streptavidin-coated magnetic beads so that the RNA template and ribozyme could be stripped away by a denaturing alkaline wash, thus improving the resolution of PAGE analysis. The fluorescent label enabled quantitation of these materials.

The apparent rate constant, k_pol, was determined for reactions carried out in the presence of various concentrations of either dGTP or dCTP to assess the saturation behavior of these substrates (Figure S1). The apparent K_m for dGTP addition at a C position on the template is 1.8 mM and for dCTP addition at a G position is 0.29 mM. For dGTP addition at a wobble-paired U position, the apparent K_m is 2.0 mM. These values should not be regarded as the binding affinity of dNTP and template but rather the concentration at which the observed rate of reaction is half-maximal. Based on these data, a concentration of 4 mM dNTP was chosen for all subsequent reactions in this study.

For unmodified dGTP or dCTP, k_pol of the 38–6 polymerase ribozyme is 2.3 × 10⁻⁴ or 4.5 × 10⁻⁴ s⁻¹, respectively (Figures S2 and S3). This is considerably slower than the rate of dGTP or dCTP addition catalyzed by pol β, for which k_pol is 14.9 or 5.7 s⁻¹, respectively.²⁶ The ribozyme is ~100-fold faster when operating as an RNA-dependent RNA polymerase than as a DNA polymerase,¹⁰ and when operating as an RNA polymerase it incorporates GTP more readily than CTP.¹³

The difference in k_pol for addition of dGTP versus dCTP is largely explained by the difference in saturation of these substrates when provided at 4 mM concentration (69% and 93%, respectively). When dGTP is added as a wobble pair, k_pol is 2.6 × 10⁻⁶ s⁻¹ (Figure S4), which is ~90-fold slower than the rate of dGTP addition as a Watson–Crick pair. This result cannot be explained by differences in saturation because the apparent K_m for dGTP is very similar in the two reactions. Rather, the altered geometry of the dG·U wobble pair may be less well accommodated by the active site of the ribozyme. A similar decrease in k_pol is seen for pol β, but in that case the rate of wobble addition remains sufficiently fast that it was possible to evaluate the more slowly reacting β,γ-substituted dGTP analogues as wobble pairs,^14,26 which is not the case here.

Taking advantage of a rich toolkit of β,γ-bridging O-substituted dGTP and dCTP analogues that span a broad range of acidity for the bisphosphonate leaving group,^{14,16,20–22,26} the rate of deoxynucleotide incorporation was determined as described above. The dihalo-substituted series, which includes the CF₂, CFCl, CCl₂, and CBr₂ analogues, was especially informative, with the pK_a4 of the leaving group spanning a range of 7.8 to 9.3. A Brønsted plot of log(k_pol) versus pK_a4 demonstrated a clear LFER profile for both the dGTP and dCTP series, with a slope of −0.88 (r = 0.99) and −0.70 (r = 0.99), respectively (Figure 3). The LFER for the corresponding nonenzymatic, base-catalyzed hydrolysis reaction has a slope of approximately −0.9.^27,28 The monohalo-substituted series, which includes the CHF analogue of dGTP and the CHF, CHCl, and CHBr analogues of dCTP, fell off the line of the dihalo series. The three monohalo-substituted analogues of dCTP also demonstrated a LFER profile, with the Brønsted plot having a slope of −0.23 (r = 0.90). The CH₂ analogue of dCTP falls closely on this same line (Figure 3B).

Figure 3.

LFER plots for addition of various dNTP analogues by the polymerase ribozyme. (A) dGTP (O) and various β,γ-substituted dGTP analogues. The slopes of the lines for the dihalo- and non-halogen-substituted analogues are −0.882 ± 0.103 and −0.601 ± 0.003, respectively. (B) dCTP (O) and various β,γ-substituted dCTP analogues. The slopes of the lines for the dihalo- and monohalo-substituted analogues are −0.696 ± 0.068 and −0.227 ± 0.109, respectively. k_pol values were determined based on three replicates, with error bars representing the standard deviation.

For all compounds that were tested as both dGTP and dCTP analogues, k_pol is faster for the dCTP compared to the dGTP analogue. Unsubstituted dNTPs, with a β,γ-bridging oxygen, are incorporated faster than would be expected based solely on the pK_a4 of the leaving group, which is especially the case for dCTP. Three additional non-halogenated analogues that were tested in the dGTP series, the NH, CH₂, and CHCH₃ analogues, demonstrated a distinct LFER profile, with the Brønsted plot having a slope of −0.60 (r = 1.00) (Figure 3A).

DNA polymerases are among the best studied class of enzymes, which is appropriate given their central role in the maintenance and propagation of genetic information. Extensive mechanistic studies have been carried out with various DNA polymerases, utilizing a broad family of β,γ-bridging O-substituted dNTP analogues to probe the electronic character of the transition state.^{14–16,19,21,22,26} By employing a series of bisphosphonate leaving groups with a range of pK_a4 values, it has often been possible to conduct a LFER analysis based on a Brønsted plot of k_pol versus pK_a4 to provide insight into features of the transition state that would not be apparent through crystallographic studies. The present study takes a similar approach, for the first time extending this approach to a DNA polymerase that is composed of RNA rather than protein.

The class I polymerase ribozyme is the product of extensive in vitro evolution studies^6–11 that sought to recapitulate the function of an RNA-dependent RNA polymerase, which has been described as the “first enzyme” of life.²⁹ While not a historical reconstruction of that enzyme, these studies have demonstrated the ability of RNA to catalyze the same biochemical transformation as that of modern polymerase proteins. The most advanced forms of the RNA polymerase ribozyme also have the ability to catalyze the RNA-templated polymerization of DNA,¹² despite the ~100-fold reduced nucleophilicity of the 3′-hydroxyl of DNA compared to RNA.³⁰ Such reverse transcriptase activity is thought to have been crucial for the transition from RNA to DNA genomes during the early history of life on Earth. It also provides an opportunity to examine the mechanism of a polymerase ribozyme using the same tools that have been developed to study DNA polymerase proteins.

As was seen previously with DNA polymerases β, λ, and η,^{16,19,21,22,26} the class I polymerase ribozyme exhibits a LFER profile for both the dihalo- and mono-substituted dNTP analogues (Figure 3). This relationship is seen most clearly for the dihalo analogues, with the Brønsted plots having a slope of −0.88 and −0.70 for dGTP and dCTP, respectively. These values are similar to what was observed for pol λ, where the slope is −0.84.¹⁹ The slope for the nonenzymatic reaction of hydroxide ion with a model diester (methyl phenyl phosphate) is −0.94.²⁷ Taken together, these results are consistent with the hypothesis that, for both the ribozyme and protein polymerase, the charge distribution in the transition state is similar to that of the uncatalyzed solution reaction, indicating that this aspect of the reaction mechanism is conserved between the two evolved biocatalysts despite their completely different chemical compositions.

Notable outliers are the reactions catalyzed by the class I polymerase involving unsubstituted dGTP or dCTP, which are ~100-fold faster than would be predicted based on the LFER plots of the other compounds. This difference is potentially explained by a positioned water molecule that has been observed in the crystal structure of the class I ligase, which contacts both the β,γ-bridging oxygen and the 5′ oxygen of the reactive triphosphate.²⁴ This water molecule may help to align the α-phosphorus of the triphosphate for in-line attack by the nucleophilic 3′-hydroxyl of the primer, an interaction that would be disrupted by the various methylene or imido bisphosphonate substitutions. It would be instructive to determine the crystal structure of the prereacted form of the ribozyme with some of the dNTP analogues, preferably with the advanced polymerase rather than the original ligase.

Although composed of RNA rather than protein, and despite the ~10⁵-fold slower rate of RNA-catalyzed dNTP addition compared to the corresponding reaction with biological polymerase proteins, there is a remarkable convergence of catalytic mechanism among both classes of macromolecules. The class I polymerase is most like pol λ with regard to the electronic properties of the transition state but also similar to polymerases β and η.

RNA has a paucity of functional groups compared to proteins but can capture similar mechanisms of catalysis. For phosphoester transfer reactions, these mechanisms include hydrogen bonding interactions involving a nucleobase or ribose 2′-hydroxyl to assist in orientating the reactive groups, positioning a Mg²⁺ ion to assist in activating a nucleophile, stabilizing the geometry of the transition state, or neutralizing the charge of the leaving group, and general acid–base catalysis by N1 of a purine or N3 of a pyrimidine, when the pK_a of the nitrogen is appropriately perturbed.³¹ Darwinian evolution shapes the catalytic center of an enzyme in a manner that is indifferent to the chemical nature of its component subunits, so long as those subunits help to promote a reaction that confers selective advantage to the evolving entity. Selective advantage is most readily discernible when it relates directly to the copying of genetic information from parent to progeny.