|Home | About | Journals | Submit | Contact Us | Français|
The selection of nucleoside triphosphates by a polymerase is controlled by several energetic and structural features, including base pairing geometry as well as sugar structure and conformation. Whereas base pairing has been considered exhaustively, substantially less is known about the role of sugar modifications for both nucleotide incorporation and primer extension. In this study, we synthesized oligonucleotides containing 2′-fluoro modified nucleosides with constrained sugar pucker in an internucleotide position and, for the first time, at a primer 3′-end. The thermodynamic stability of these duplexes was examined. The nucleoside 2′-deoxy-2′-fluoroarabinofuranosyl uracil (U2′F(ara)) favors the 2′-endo conformation (DNA-like) while 2′-deoxy-2′-fluororibofuranosyl uracil (U2′F(ribo)) favors the 3′-endo conformation (RNA-like). Oligonucleotides containing U2′F(ara) have slightly higher melting temperatures (Tm's) than those containing U2′F(ribo) when located in internucleotide positions or at the 3′-end and when correctly paired with adenine or mispaired with guanine. However, both modifications decrease the magnitude of ΔH° and ΔS° for duplex formation in all sequence contexts. In examining the thermodynamic properties for this set of oligonucleotides, entropy-enthalpy compensation is apparent. Our thermodynamic findings led to a series of experiments with DNA ligase that reveal, contrary to expectation based upon observed Tm's, that the duplex containing the U2′F(ribo) analog is more easily ligated. The 2′-fluoro-2′-deoxynucleosides examined here are valuable probes of the impact of sugar constraint and are also members of an important class of antitumor and antiviral agents. The data reported here may facilitate understanding the biological properties of these agents, as well as the contribution of sugar conformation to replication fidelity.
The accurate replication of nucleic acids requires that polymerases select the correct nucleotide at each successive step of replication. Substantial work has focused on the importance of “base pairing fidelity” in the selection of the correct nucleotide (1-6). Polymerases must also choose among potential nucleoside triphosphates, even when the base pairing condition has been met. For example, RNA polymerases select ribonucleotide triphosphates (rNTP's) whereas DNA polymerases select 2′-deoxyribonucletide triphosphates (dNTP's) upon the basis of sugar structure and conformation. The selection of nucleotide triphosphates (NTP's) based upon differences in sugar structure and conformation has led to a suggested role for “sugar fidelity” among polymerases although the mechanisms have not been extensively explored (7-10).
For a given DNA sequence position, polymerase discrimination must occur in two distinct steps. In the insertion step, a candidate nucleotide is interrogated for its capacity to bind to the primer-template-enzyme complex with sufficient stability in an acceptable geometry (1-6). Differences in the base-pairing and base-stacking energy between correct and incorrect nucleotides, as measured in oligonucleotide melting studies, have been proposed to account for polymerase discrimination at the insertion step. Following insertion, the polymerase must select the next correct nucleotide during the extension step. Surprisingly, polymerase extension beyond a mispair is very difficult, even for the insertion of a correct NTP and even though base-pairing and geometry conditions are met (11-15). Extension fidelity contributes nearly as much to the overall replication fidelity as the initial insertion step. Although polymerase pausing at the extension step following a nucleotide misinsertion event would reduce overall mutation frequency by facilitating proofreading or other repair, the mechanistic basis for extension fidelity is as yet unknown.
Polymerase insertion and extension require a terminal 3′-hydroxyl (3′-OH) in the correct position to attack the α-phosphate of a candidate NTP. In the case of geometrically aberrant base pairs, such as a purine-purine mispair, the 3′-OH would be shifted several angstroms from the correct position, potentially preventing polymerase extension. With purine-pyrimidine mispairs, however, the geometry is closer to that of a normal Watson-Crick base pair, so that more subtle differences, such as sugar conformation, might become important. The furanose sugar component of nucleic acids is non-planar and adopts a number of potential conformations that can be described by the pseudorotation angle, P (16-18). There are several conformations that correspond to energy minima as a function of P, and the value of P can significantly change the position of the 3′-OH. Ribonucleotides in RNA are biased toward a 3′-endo conformation whereas 2′-deoxynucleosides in DNA assume preferentially a 2′-endo pucker, potentially explaining, in part, polymerase sugar fidelity (Figure 1).
While sugar conformation can be biased by sugar structure, sugar conformation can also be influenced by base pair configuration. Sugar conformation, and therefore 3′-OH position, can change due to mispair formation, as well as from changes in the glycosidic torsion angle (19-23). While a correct base pair at the 3′-end of a template-primer complex would likely be found predominantly in a correct conformation with the 3′-OH in the correct position, mispair formation could modify sugar pucker and distort the position of the 3′-OH potentially explaining, in part, polymerase extension fidelity.
Sugar conformation can also be biased by the presence of substituents in the furanose ring. The conformational difference between deoxyribose and ribose sugars is attributed primarily to the presence of the 2′-OH in the ribonucleosides. Other substituents, in particular, electron-withdrawing substituents including fluorine, are known to profoundly influence sugar conformation (24-33). Evidence exists that sugar pucker can influence both nucleotide incorporation and extension by polymerases (34-41). Nucleotides that are constrained to a 3′-endo conformation - for example, 2′-fluororibo nucleotides - are preferentially incorporated by RNA polymerases (34). Conversely, 2′-fluoroarabino nucleotides that prefer the 2′-endo pucker are preferentially incorporated by DNA polymerases, yet surprisingly, are very difficult to extend (38). The physical basis for this selectivity has not as yet been established.
For this study, we constructed oligonucleotides with 2′-deoxyuridine (dU) and the 2′-fluoro analogs 2′-deoxy-2′-fluoroarabinofuranosyl uracil (U2′F(ara)) and 2′-deoxy-2′-fluororibofuranosyl uracil (U2′F(ribo)) in both internucleotide and 3′-end positions (Figure 1). The sugar pucker equilibrium for the U2′F(ara) and U2′F(ribo) analogs studied here has been previously studied by NMR spectroscopy (25-32). The reference nucleoside analog, dU, is in a rapid equilibrium between 2′-endo and 3′-endo conformations, with a preference (61%) for the 2′-endo conformation (32). The U2′F(ara) analog is 57% 2′-endo whereas the U2′F(ribo) analog is 69% 3′-endo (33). When located in oligonucleotides and constrained by internucleotide linkages, the conformational preference of dU and U2′F(ara) shifts more toward 2′-endo whereas U2′F(ribo) shifts more toward 3′-endo (26).
The analogs described here were incorporated into an internucleotide position, as well as on the 3′-end. While both analogs have been incorporated previously into internucleotide positions, this is the first report of incorporation into the 3′-position. The analogs have been incorporated into duplex structures either properly paired with adenine or mispaired with guanine in an internucleotide position, as well as at a 3′-end position creating a model polymerase replication fork or ligase junction. The thermal and thermodynamic stability of these duplex structures has been studied. With this set of analogs, we could probe the energetic advantage or penalty for each analog and base pair, as well as probe for a potential interaction between base pairing and sugar conformation. Thermodynamic results reported here led to a series of experiments with DNA ligase that demonstrate, unexpectedly, that the duplex containing the U2′F(ribo) analog is more easily ligated. The thermodynamic parameters and results obtained are discussed within the context of the available literature on polymerase preferences for both nucleotide insertion and extension. Due to the importance of sugar-modified nucleosides as anticancer and antitumor drugs, the results reported here may provide new insight into the mechanisms of activity and the potential adverse effects of these analogs.
The solvents dichloromethane (CH2Cl2), methanol (MeOH), ethyl acetate (EtOAc) and hexanes were purchased from Fisher Scientific (Pittsburgh, PA). Pyridine, triethylamine (TEA) and acetonitrile (MeCN) were purchased from Sigma-Aldrich (St. Louis, MO). Dimethoxytrityl chloride (DMT-Cl) and 2-cyanoethyl tetraisopropyl phosphorodiamidite were purchased from Sigma-Aldrich (St. Louis, MO). 1,3,5-tri-O-benzoyl-2′-deoxy-2′-fluoro-D-arabinofuranose was purchased from MP Biomedicals (Aurora, OH). Thin layer chromatography (TLC) was performed on precoated silica gel 60 F254, 5×20 cm, 250 μm thick plates purchased from EMD (Gibbstown, NJ).
Commercially available 1,3,5-tri-O-benzoyl-2′-deoxy-2′-fluoro-D-arabinofuranose was brominated to the corresponding bromosugar in 100% yield (42,43). The bromosugar was then coupled to 2,4-bis-O-trimethylsilyluracil to give 1-β-D-(3,5,-di-O-benzoyl-2-fluoroarabinofuranosyl)uracil as a solid residue (44). The dibenzoyl derivative was then deprotected to give 2′-deoxy-2′-fluoro-1-β-arabinofuranosyluracil (U2′F(ara)) in a 30% yield (45). U2′F(ara) was then tritylated to give 5′-dimethoxytrityl-2′-deoxy-2′-fluoro-1-β-arabinofuranosyluracil in 48% yield and subsequently converted to its phosphoramidite derivative, 5′-dimethoxy-2′-deoxy-2′-fluoro-1-β-D-arabininofuranosyluracil, 3′-[(2-cyanoethyl)-(N,N-diisopropyl)]-phosphoramidite in 49% yield (46).
Standard oligonucleotide synthetic procedures (47) were used to produce oligonucleotides with normal and modified analogs including dU, U2′F(ara) and U2′F(ribo) residues located at an internucleotide site. Oligonucleotide synthesis was conducted with a Gene Assembler Plus (Pharmacia LKB) automated DNA synthesizer. Oligonucleotides were deprotected with concentrated NH3 (aq) at 60°C for 24 hours. In general, following synthesis and deprotection, oligonucleotides were purified by HPLC using a Hamilton PRP-1 column and a gradient of 10 – 40% MeCN in potassium phosphate buffer (10mM, pH 6.8) and examined by MALDI-TOF-MS. Oligonucleotides were then detritylated with 80% aqueous acetic acid at room temperature for 30 minutes. Following detritylation, oligonucleotides were purified by HPLC using a C-18 Vydac column and a gradient of 0 – 20% MeCN in water. Oligonucleotide purity was examined by MALDI-TOF-MS (48), and the free base composition was verified by HPLC, following enzymatic digestion (49) using a Supelcosil LC-18-S column and a gradient of 0 – 15% MeCN in water.
To insert dU, U2′F(ara) and U2′F(ribo) residues at the primer terminus, three synthetic approaches were investigated using the following universal supports available from Glen Research: 1) Glen UnySupport CPG 500, 2) Universal Support II and 3) Universal Support III PS (50,51). In each of the three approaches, it was necessary to increase the coupling times to 10 min (from 3 min) for insertion of U2′F(ara) and U2′F(ribo) residues at the primer terminus. Following detritylation, overall purity of the oligonucleotides produced using each of the three universal supports was determined by MALDI-TOF-MS and denaturing polyacrylamide gel (20% (v/v) polyacrylamide, 8M urea). Upon the basis of MALDI-TOF-MS analysis, oligonucleotides synthesized using Universal Support III PS were of greater purity than those synthesized using Universal Support II or Glen UnySupport CPG 500. In particular, the mass spectra for oligonucleotides containing U2′F(ara) and U2′F(ribo) residues, produced using Universal Support III PS, revealed single peaks corresponding to the expected oligonucleotide masses. Following synthesis using Glen UnySupport CPG 500, several unidentified impurities in U2′F(ara) containing oligonucleotides (M-50, 221, 307, 360 and 619) and in U2′F(ribo) containing oligonucleotides (M+18 and M-227) were observed. Upon the basis of purity assessment following gel electrophoresis, Universal Support III PS was again determined to produce oligonucleotides of greater purity. In all, synthesis using Universal Support III PS produced oligonucleotides in greater quantity and of higher purity and was thus used exclusively for subsequent syntheses of oligonucleotides with dU, U2′F(ara) and U2′F(ribo) residues at the 3′-end.
Samples containing non self-complementary oligonucleotides were prepared in buffer containing 0.1 M NaCl, 0.01 M sodium phosphate, and 0.1mM EDTA, pH 7.0. Complexes were prepared by mixing equimolar amounts of interacting strands, and concentration dependent Tm measurements were conducted with total strand concentration (CT) between 2 and 60 μM in cuvettes with path lengths between 1 and 10 mm. Molar extinction coefficients of oligonucleotides were calculated (52) to determine single strand concentrations. Oligonucleotide melting temperatures (Tm) were determined using a Varian Cary 300 Bio UV-visible spectrophotometer (Varian, Walnut Creek, CA). Five temperature ramps were performed on each sample per run while observing the absorbance at 260 nm: 1) 12 °C to 90 °C at a rate of 0.5 °C/min, 2) 90 °C to 12 °C at a rate of 0.5 °C/min, 3) 12 °C to 90 °C at a rate of 0.5 °C/min, 4) 90 °C to 12 °C at a rate of 0.5 °C/min, and 5) 12 °C to 90 °C at a rate of 0.5 °C/min. The sample was held for 3 min when the temperature reached 90 °C and for 10 min when it reached 12 °C and started the next cycle. Data were collected at 0.5 °C intervals while monitoring the temperature with a probe inserted into a cuvette containing only buffer. The Tm of each duplex was determined using Cary WinUV Thermal software (Varian). Theoretical Tm values for control duplexes (A:dU and G:dU) were determined (53,54) and compared against values obtained using the Cary WinUV Thermal software. Thermodynamic parameters for non self-complimentary duplexes were calculated in two ways: 1) averages from fits of individual melting curves at different concentrations using Van't Hoff calculation in the Cary WinUV Thermal software; 2) the 1/Tm versus ln (CT/4) plots fitted to the following equation for the non self-complementary sequences examined here.
Both methods assume a two-state model and ΔCp = 0 for the transition equilibrium. The two-state approximation was assumed to be valid for sequences in which the ΔH° values derived from the two methods agreed within 15% (54). The ΔH ° values derived from the two methods agree within 15%, indicating that the two-state approximation is valid for all other sequences employed in this study.
The E. coli DNA ligase was obtained from New England Biolabs (Ipswich, MA) and human DNA ligase III was obtained from Enzymax (Lexington, KY). Oligonucleotide 5′-end radiolabeling was performed using adenosine 5′-[γ-32P]-triphosphate ([γ-32P]-ATP) (MP Biomedical, Costa Mesa, CA) and T4 polynucleotide kinase (New England BioLabs) under conditions recommended by the enzyme supplier. Labeled mixtures were subsequently centrifuged through G-25 Sephadex columns (Roche Applied Science, Indianapolis, IN) to remove excess unincorporated nucleotide. Duplex oligonucleotides containing a ligase junction were generated by mixing the labeled single strand (5′-GGCCACGACGG-3′) with a 2-fold molar excess of the strands CTTTGCCCGAAX, where X is dU, U2′F(ara), U2′F(ribo), and CCGTCGTGGCCATTCGGGCAAAG in the appropriate enzyme buffer as previously described (55). The E. coli DNA ligase buffer contained 30 mM Tris-HCl pH 8.0, 4 mM MgCl2, 1 mM DTT, 26 μM NAD+, 50 μg/ml BSA. The human DNA ligase III buffer contained 50 mM Tris-HCl pH 7.5, 10 mM MgCl2, 10 mM DTT, 1 mM ATP. Annealing mixtures were heated at 95°C for 5 min and then cooled slowly to room temperature. Standard E. coli DNA ligase assays were performed using 50 nM substrate with 500 nM E. coli DNA ligase in buffer, as above, in a total volume of 10 μl at 16°C for selected time periods. Substrates (50 nM) were incubated with 50 nM human DNA ligase III in buffer, as above, in 10 μl total volume at 26.5°C for selected time periods. The reactions were terminated by adding an equal volume of Maxam-Gilbert loading buffer (98% formamide, 0.01 M EDTA, 1 mg/ml xylene cyanole and 1 mg/ml bromophenol blue). Samples were denatured by heating at 95°C for 5 min and quickly placed on ice for 2 min before electrophoresis on 20% denaturing polyacrylamide gels (8 M urea). The bands corresponding to substrate and products were visualized and quantified using a Molecular Dynamics PhosphorImager (Molecular Dynamics, Sunnyvale, CA, now part of GE Healthcare) and quantified using ImageQuant software. Reaction rate constants (kobs) for ligation reactions were determined by fitting time course data to a single exponential (y = a(1-a-bx)) using Sigma Plot 10.0, where “a” is the maximum product ratio and “b” is the reaction rate constant, kobs.
The phosphoramidite analog of U2′F(ribo) is commercially available; however, the corresponding phosphoramidite of U2′F(ara) is not available and was prepared in this laboratory by previously described methods as shown in Figure 2A. Oligonucleotides containing both analogs were prepared by standard solid phase synthesis methods. Sequences of oligonucleotides used in this study are shown in Figure 3.
Although the synthesis of oligonucleotides with U2′F(ara) and U2′F(ribo) have been previously reported, oligonucleotides with these analogs at the 3′-end are reported here for the first time. We considered two methods: the synthesis of solid phase supports linked to the analogs of interest or the use of solid supports containing linkers or “universal supports” for the preparation of 3′-end modified oligonucleotides (Figure 2B). As the needed phosphoramidites were available in our lab, we proceeded to test a series of commercially available solid supports. Although we did not exhaustively examine all of the supports, we found that Universal Support III PS provided the highest consistent coupling yields and purity.
Synthetic oligonucleotides were characterized by MALDI-TOF-MS following deprotection and purification by HPLC. The mass spectra of the oligonucleotides containing the U2′F(ara) analog in both an internucleotide position and at a 3′-end are shown in Figures 4A and B, respectively. The observed mass in each case was consistent with the expected mass and demonstrated that the 3′-phosphate of the original phosphoramidite had been removed (Figure 2B). Oligonucleotides were also characterized by enzymatic digestion and analysis of the liberated nucleosides by HPLC. We considered this important as the mass of the two 2′-fluoro analogs is identical, and we needed an additional method to confirm that the oligonucleotides contained the correct isomer. As shown in Figure 5A, dU, U2′F(ara) and U2′F(ribo) are separable by HPLC and resolvable from the standard DNA nucleosides (Figure 5B).
Oligonucleotide duplexes were prepared in buffered solution by mixing equimolar amounts of the two strands, as indicated in Figure 3. Strands were annealed by heating to 90°C, followed by slow cooling. Melting profiles were obtained by observing UV absorbance at 260 nm as a function of temperature as described in Materials and Methods. Melting temperatures and corresponding thermodynamic parameters were obtained by analysis of the UV-temperature profiles as previously described (53, 54). Melting temperatures and thermodynamic parameters are presented in Table 1. Example melting curves are shown in Figure 6.
Thermal and thermodynamic data obtained for the ensemble of oligonucleotides examined here were expressed as the corresponding differences by comparing the measured value for the substituted duplexes with the standard A:dU containing duplex for the 3′-end and internucleotide series. This data and the corresponding values of ΔTm, ΔΔG°37, ΔΔH° and ΔΔS° are presented in Table 1. Values of ΔΔH° and ΔΔS° appeared to correlate with one another, and this relationship is presented in Figure 7. Energy differences between base pairs examined in this study are shown diagrammatically in Figure 8.
Oligonucleotides were assembled as shown in Figure 3C to create a ligase joint. The model ligase junction was incubated with human DNA ligase III and E. coli DNA ligase and the reaction products were examined by gel electrophoresis as shown in Figure 9. Human DNA ligase III and E. coli DNA ligase were able to ligate junctions containing all three analogs. Human DNA ligase III was found to ligate junctions containing dU, U2′F(ara) and U2′F(ribo) with kobs values of 1.83 ± 0.21 × 10-2 s-1, 4.70 ± 0.43 × 10-3 s-1 and 2.18 ± 0.25 × 10-2 s-1, respectively. In addition, E. coli DNA ligase was found to ligate junctions containing dU, U2′F(ara) and U2′F(ribo) with kobs values of 6.34 ± 0.73 × 10-3 s-1, 8.64 ± 1.33 × 10-4 s-1 and 2.81 ± 0.39 × 10-2 s-1, respectively.
These properties might help explain why polymerases initially insert correct nucleotides (insertion) and insert correct nucleotides following mispairs (extension) with such low efficiency. The impact of constrained sugar conformation on mispairs has not been previously examined. We considered the possibility that mispair geometry might be coupled with changes in sugar pucker at a duplex 3′-end which might help explain the inefficiency of mispair extension. Previous structural studies have suggested that aberrant base pair geometry could induce changes in sugar conformation (19-23), however, these effects have not been previously studied at a replication fork. The sugar-substituted nucleosides examined here are also members of an important class of nucleoside analogs with antitumor and antiviral properties (34-41), and thus the data reported here might facilitate a greater understanding of the biological activity of this class of nucleoside analogs. Our experimental approach was to construct oligonucleotides containing dU and 2′-fluoro analogs constrained to either the 2′-endo (south, DNA-like) sugar pucker or the 3′-endo (north, RNA-like) sugar pucker and to measure thermodynamic stabilities of duplex oligonucleotides.
In the studies reported here, uracil was selected as the pyrimidine rather than thymine so that the data set examined here could be used in future studies to compare a series of 5-substituted pyrimidines. The replacement of T by dU does not change base pairing geometry when paired with A (56) or mispaired with G (57) and results in only a modest decline in Tm due to reduced base stacking (58). Oligonucleotide duplexes were assembled as shown in Figure 3. Oligonucleotide melting temperatures were obtained from the temperature dependence of the UV absorbance as shown in Figure 5. Thermodynamic parameters were extracted from the melting curves and the corresponding data is presented in Table 1. The observed Tm of the duplex containing an internucleotide A:dU base pair was 43.8 ± 0.2 °C. The expected Tm for a duplex of the same sequence except dU would be replaced by thymine is 43.8 to 44.2 °C, depending upon which basis set and method of calculation is used (53). The observed Tm of the duplex in which the dU residue on the 3′-end was paired with A was observed to be 49.4 ± 0.3 °C, slightly below the 52.8 to 53.1 °C calculated range when dU is replaced by T (53).
The conversion of a Watson-Crick base pair to a wobble mispair is known to significantly decrease melting temperatures (1,11,58). The formation of the internucleotide G:dU mispair results in a Tm of 36.6 ± 0.2 °C, 7.2 °C lower than the A:dU duplex. The calculated range for a duplex of the same sequence containing a G:dT mispair is 36.8 to 38.2 °C (54). When the G:dU mispair is moved from an internucleotide position to the 3′-end, the observed Tm is 48.1 ± 0.2 °C, only 1.3 °C lower than the Tm for the sequence containing a correct A:dU base pair. The expected Tm for an oligonucleotide of an otherwise same sequence, but with a G:dT rather than G:dU mispair at the 3′-end, is 50.2°C to 50.8°C (54). The data reported thus far is consistent with expectations based upon literature precedents and confirms that the impact of a mispair on Tm is substantially less when the mispair is located at the 3′end.
The Tm of the duplex containing the internucleotide A:U2′F(ara) base pair is observed to be 43.8 ± 0.3 °C, experimentally indistinguishable from that of the A:dU duplex. Previously, other investigators have observed that the placement of 2′-fluoroarabino analogs in duplex structures increases their melting temperatures by roughly 1°C per nucleotide, and this stabilizing effect has been attributed to constraining the sugar into the more DNA-like 2′-endo pucker (28,32). In most of the previous studies, however, the oligonucleotides included multiple substitutions. The observed Tm for the formation of the internucleotide G:U2′F(ara) mispaired duplex is 37.5 ± 0.2 °C, which is 0.9 °C higher than that of the G:dU duplex. The 3′-terminal A:U2′F(ara) base pair duplex exhibited a Tm 0.9°C higher than that of the 3′-terminal A:dU base pair duplex whereas the Tm of the terminal G:U2′F(ara) mispair duplex is indistinguishable from the Tm of the 3′-terminal G:dU duplex. These results demonstrate that incorporation of the U2′F(ara) analog stabilizes duplexes in some cases; however, the observed effect on Tm is modest and substantially less than the impact of mispair formation.
Substitution of dU with the U2′F(ribo) analog at an internucleotide position has the opposing effect, slightly decreasing Tm 's when paired with A or mispaired with G. Again, the impact of the sugar constraint on Tm is less than that of mispair formation at the internucleotide position. At the 3′-end, the effect of sugar constraint on Tm shows similar decreases. One of the initial hypotheses considered here is that mispair geometry might be related to sugar pucker at a duplex 3′-end, and therefore, a nucleotide with constrained sugar pucker might have an opposing impact on a mispair versus a normal pair. The evidence thus far on Tm's does not support this hypothesis in that mispairs with G are approximately 2 °C less stable than base pairs with A for both U2′F(ara) and U2′F(ribo) analogs.
Although oligonucleotide duplexes might have similar Tm's, they can have different thermodynamic parameters. For this reason we examined the free energy changes at 37°C (ΔG°37), enthalpy (ΔH°) and entropy (ΔS°) changes for each of the duplex structures reported here (Table 1). In Figure 6B, the melting curves for oligonucleotides with A:dU and A:U2′F(ara) are shown. Although the midpoint for the temperature-dependent UV transition (Tm) is similar for each, the shapes of the curves, and corresponding thermodynamic parameters, are different. The magnitude of the experimental errors for ΔS°, ΔH° and ΔG° observed here are in accord with previously reported studies (58,60).
In order to assess the impact of the substitution on the thermodynamic parameters, the free energy, enthalpy and entropy changes are expressed as the corresponding differences (ΔΔG°, ΔΔH° and ΔΔS°) relative to the A:dU base pair for the internucleotide or 3′-terminal series. In all cases, positive values are observed for ΔΔG°37, ΔΔH° and ΔΔS°, indicating that constraining the sugar pucker with either the U2′F(ara) or U2′F(ribo) analog has a destabilizing effect, as does mispair formation. Previously, positive values for ΔΔH° and ΔΔS° upon substitution with U2′F(ara) have been observed (25), and the magnitudes of these changes per substitution are similar to those reported here. The impact of the sugar constraint on ΔΔS° has been attributed to a conformational preorganization, reducing the net conformational entropy change upon duplex formation (25). The impact on ΔΔH° would be attributed to the constrained sugar preventing the formation of the most favorable base-stacking geometry.
In this study, we have assumed a two state equilibrium between 2′-endo and 3′-endo sugar puckers as supported by previous NMR studies with the U2′F(ara) and U2′F(ribo) analogs studied here (25-32). The reference nucleoside analog, dU, is in a rapid equilibrium between 2′-endo and 3′-endo conformations, with a preference (61%) for the 2′-endo conformation (32). The U2′F(ara) analog is 57% 2′-endo whereas the U2′F(ribo) analog is 69% 3′-endo (33). When located in oligonucleotides and constrained by internucleotide linkages, the conformational preference of dU and U2′F(ara) shifts more toward 2′-endo whereas U2′F(ribo) shifts more toward 3′-endo (26).
Enthalpy and entropy differences for the oligonucleotide duplexes examined here are shown to be proportional. In previous studies of oligonucleotide stability in which mispairs and constrained sugar pucker were considered separately, enthalpy and entropy contributions were shown to correlate (60). As shown in Figure 7, enthalpy and entropy are shown to correlate for the series examined here, as well. The relatively large size of the error bars when the values are presented as ΔΔS° and ΔΔH° are consistent with previous studies as discussed by McTigue et al. (60). The slope of the line in Figure 7 is 2.9, and this plot includes normal base pairs, those with sugar constraints and unconstrained mispairs, and mispairs with sugar constraints in both internucleotide and 3′-terminal positions. This value compares favorably with the value of 2.8 from a previous study on mispairs with no sugar constraint (11) and with the value of 2.95 from a study that examined constrained sugars but no mispairs (60), suggesting that this value of ΔΔS° / ΔΔH° is broadly applicable to nucleic acids.
Previous studies have established that sugar pucker can also influence nucleotide insertion by DNA polymerases. DNA polymerases must distinguish between NTP's with different sugars, and it is known that the discrimination is between one and three orders of magnitude depending upon the template and polymerase, and several apparent Km values have been reported. Astatke et al. (9) demonstrated that the mechanism for E. coli DNA polymerase discrimination between deoxy and dideoxunucleotides involves a direct interaction with the 3′-OH present on the deoxy, but not dideoxynucleotide, accounting for several orders of magnitude higher Km for the dideoxynucleotide. Richardson et al. (8) demonstrated that human polymerase α and polymerase γ accept U2′F(ara) nucleotides with Km's similar to dUTP. However, both polymerases discriminate against ribonucleotides and 2′-deoxy-2′-fluororibonucleotides with Km's that are increased between one and two orders of magnitude, although all of these nucleotides contain the necessary 3′-OH.
Previously, Goodman and coworkers (11) explained how polymerases could discriminate against mispairs by amplifying free energy differences (ΔΔG°) between correct and incorrect base pairs. Free energy differences are determined by the relative magnitude of the enthalpy and entropy contributions (ΔΔG° = ΔΔH° - TΔΔS°). If ΔΔH° and ΔΔS° are proportional, large values of ΔΔH° might be cancelled by large values of ΔΔS°, giving small values of ΔΔG°. On the other hand, if the polymerase active site accepts only nucleotides with sugar conformations that approximate the correct conformations, ΔΔG° would approach ΔΔH° providing sufficient energy for the observed discrimination. One of the surprising findings of this study is that, despite modest differences in Tm, (ΔTm = -0.2 °C), the measured difference in free energy change between A:dU and A:U2′F(ribo) (ΔΔG° = 0.5 kcal/mol) is as large as the corresponding difference in free energy change between a correct A:dU base pair and a G:dU mispair (ΔΔG° = 0.5 kcal/mol) although the mispair has a substantially larger impact on the observed Tm (ΔTm = -1.3 °C). These results are consistent with a thermodynamic contribution to sugar fidelity for polymerase insertion.
Our initial expectation was that constraining the sugar pucker to 2′-endo (U2′F(ara)) would stabilize a correct base pair, but destabilize an incorrect base pair whereas the 3′-endo sugar (U2′F(ribo)) would destabilize the correct base pair and stabilize the mispair. Our experimental results are inconsistent with this expectation. When examining the impact of mispair formation and sugar pucker on ΔΔG°, the contributions of each appear to be additive for both 3′-terminal and internucleotide positions. Neither constrained pucker appears to selectively stabilize or destabilize either the correct base pair or the mispair, in accord with observations of Tm discussed above. The contributions of mispair formation and constrained sugar pucker also appear to be additive for ΔΔH° and ΔΔS° in all cases presented here, as indicated in Figure 8. For example, for the internucleotide A:dU base pair, conversion to a G:dU mispair is associated with a ΔΔH° of 6.1 kcal/mol. The energy penalty associated with constraining the sugar conformation of the mispairs by comparing G:dU and G:U2′F(ara) is associated with a ΔΔH° of 5.3 kcal/mol. The combined effect of mispair formation and constraining the sugar with the U2′F(ara) analog would be expected to be 11.4 kcal/mol if they were additive and not interacting, which is the observed value obtained when comparing the thermodynamic properties of the A:dU and G:U2′F(ara) duplexes.
The observation that the influences of mispair geometry and sugar constraint are simply additive strengthens the proposal that aberrant base pair geometry and constrained sugar pucker are not linked. Therefore, it is unlikely that the altered base pairing geometry of a mispair induces a shift in equilibrium for the 3′-residue, placing the 3′-OH in a position inconsistent with insertion of the next nucleotide. The data presented thus far suggests that polymerases are unlikely to exploit induced changes in sugar conformation to increase replication fidelity.
The magnitude of the observed ΔΔH° and ΔΔS° for the 3′-terminal G:dU mispairs are 3.0 kcal/mol and 7.0 cal/mol K, respectively, whereas for the internucleotide mispair, the values increase to 6.1 kcal/mol and 13.8 cal/mol K, respectively. Interestingly, the enthalpic and entropic destabilization approximately doubles when moving from a 3′-terminal mispair to an internucleotide mispair, and this observation likely has important implications for understanding why polymerases have substantial difficulty in extending beyond mispairs. Previously, Goodman and coworkers (11) argued that polymerases can exploit differences in ΔG°, ΔH° and ΔS° between correct and incorrect base pairs for nucleotide insertion, and the altered base pairing and stacking energy associated with mispair formation increases proportionately the apparent Km for insertion of the incorrect nucleotide. A comparatively larger Km for insertion of a modified nucleotide is associated with a higher tendency for a candidate nucleotide to dissociate from the enzyme-primer-template complex.
Extension past a mispair, however, involves inserting a correct nucleotide, yet the Km is also substantially higher than when extending past a correct pair. When comparing the energetic penalty of mispair formation for the sequences examined here, it is apparent that an internucleotide mispair induces destabilization on both the 5′-side and on the 3′-side of the mispair. At the misinsertion step, an incorrect base pair does not stack as well on the 5′-primer-template complex and is more likely to dissociate, resulting in an increased apparent Km. At the extension step, an incoming correct nucleotide would stack - but only poorly - on the mispair. Thus, the destabilizing impact of the mispair is transmitted in the 3′-direction as well, perhaps explaining a significant component of polymerase extension fidelity.
When comparing thermodynamic parameters for base pairs located at the 3′-terminal position with those obtained for the same base pair located in an internucleotide position, some trends are apparent that could be important for understanding the activities of enzymes that act upon nucleic acids including polymerases and ligases. The thermodynamic results presented here establish that the U2′F(ribo) substitution is more destabilizing, with larger magnitude ΔH° and ΔS° when on the 3′-end, relative to the internucleotide linkage. In contrast, U2′F(ara) is substantially less destabilizing than U2′F(ribo) on the 3′-end. However, when comparing the impact of U2′F(ara) substitution on the 3′-end to the internucleotide position, U2′F(ara) is more destabilizing in the internucleotide position than on the 3′-end. The conversion of the 3′-end modification to the internucleotide linkage could be accomplished by members of the DNA ligase family. Conversion of the 3′-end U2′F(ara) to an internucleotide linkage would increase the destabilizing impact of the substitution, however, conversion of the 3′-end U2′F(ribo) to an internucleotide linkage would substantially reduce the destabilizing impact of the substitution. We therefore predicted that the rate of ligation of the analog substrates examined here would be A:U2′F(ribo) > A:dU > A:U2′F(ara).
Data obtained with E. coli DNA ligase and human DNA ligase III, as shown in Figure 9, is consistent with the above prediction, in that the order of ligase efficiency is U2′F(ribo) > dU > U2′F(ara). As seen in the panel for the U2′F(ara) data, initial interaction of the ligase with the 5′-phosphate of the linker transfers an adenosine monophosphate group (App-5′-DNA). The same intermediate forms with both the U2′F(ribo) and dU substrates; however, with the U2′F(ara) substrate, the intermediate is less efficiently converted to the internucleotide linkage. Sugar pucker does impact ligase efficiency, however, the relative impact of the U2′F(ara) and U2′F(ribo) substitutions is opposite expectation based upon Tm's, but likely explained by the thermodynamic differences described here. We note that previously, Mikita and Beardsley (35) observed that an oligonucleotide containing araC is ligated by T4 ligase three times more slowly than one with dC, although both oligonucleotides had similar Tm's. It was suggested that the presence of the 2′-OH in the arabino configuration could directly interfere with the ligase. In the studies reported here, the A:U2′F(ara) containing oligonucleotide is ligated by human DNA ligase III approximately four times slower than the A:dU oligonucleotide. Likewise, the rate of ligation of the A:U2′F(ara) oligonucleotide by E. coli DNA ligase is approximately seven times slower than the A:dU oligonucleotide. Here, the difference in ligation rates is attributed to net thermodynamic differences between a 3′-end A:U2′F(ara) and the corresponding base pair in an internucleotide linkage. Although U2′F(ara) is less disruptive thermodynamically when placed at a duplex 3′-end, it is more disruptive thermodynamically when in an internucleotide linkage, perhaps accounting for decreased ligase efficiency.
In the ligase studies reported here, it was assumed that the primary effect of the 2′F substitution was to constrain the sugar pucker equilibrium. It is important to note, however, that the larger fluorine substituent could induce additional steric effects as well as electronic effects upon the nucleophilicity of the 3′ OH, and these effects might be different for each isomer.
Multiple studies have been published on the impact of nucleotide analogs on polymerase incorporation and extension (7,8,34-41). The various studies examining different sets of analogs and polymerases make it difficult to draw specific conclusions. However, one of the consistent findings, and as yet unexplained paradoxes, with these nucleotide analogs is that some DNA polymerases strongly discriminate against ribonucleotides and 2′-deoxy-2′-fluoroarabino analogs at the insertion step, but incorporate arabino and 2′-fluoro-2′-arabino analogs with kinetics similar to normal dNTP's. Yet, once incorporated, the arabino and 2′-deoxy-2′-fluoroarabino analogs prevent further elongation and act as chain terminators. In contrast, some DNA polymerases discriminate against ribonucleotides at insertion, yet efficiently extend 3′-ribonucleotides. Indeed, DNA polymerase α preferentially adds dNTPs to ribonucleotide primers as part of the polymerase α/primase complex. Chain termination underlies the mechanism of toxicity for arabino analogs, yet the mechanism for the difference in polymerase preference between insertion and extension is as yet unknown.
Above, we discussed surprising results with DNA ligase which revealed that the U2′F(ara) analog was more thermodynamically destabilizing when in an internucleotide linkage as opposed to the 3′-end, and the net thermodynamic disadvantage with incorporating the 3′-OH of U2′F(ara) into an internucleotide linkage could provide an energetic barrier for polymerase extension. However, insufficient data currently exists to resolve this issue. Further studies are currently in progress to understand the role of sugar constraint on thermodynamic properties, and the results of these studies might provide mechanistic insights into the mechanism of action of an important class of antitumor and antiviral compounds.
†This work is supported by the National Institute of Health, National Institute of General Medical Sciences (GM451336).