|Home | About | Journals | Submit | Contact Us | Français|
Non-standard nucleotides are ubiquitous in RNA. Thermodynamic studies with RNA duplexes containing non-standard nucleotides, whether incorporated naturally or chemically, can provide insight into the stability of Watson–Crick pairs and the role of specific functional groups in stabilizing a Watson–Crick pair. For example, an A-U, inosine•U and pseudouridine•A pair each form two hydrogen bonds. However, an RNA duplex containing a central I•U pair or central Ψ•A pair is 2.4 kcal/mol less stable or 1.7 kcal/mol more stable, respectively, than the corresponding duplex containing an A-U pair. In the non-standard nucleotide purine, hydrogen replaces the exocyclic amino group of A. This replacement results in a P•U pair containing only one hydrogen bond. Optical melting studies were performed with RNA duplexes containing P•U pairs adjacent to different nearest neighbors. The resulting thermodynamic parameters were compared to RNA duplexes containing A-U pairs in order to determine the contribution of the hydrogen bond involving the exocyclic amino group. Results indicate a loss of 1.78 kcal/mol, on average, when an internal P•U replaces A-U in an RNA duplex. This value is compared to the thermodynamics of a hydrogen bond determined by similar methods. Nearest neighbor parameters were derived for use in free energy and secondary structure prediction software.
RNA is an important biomolecule adopting a broad range of structures, which in turn provides a broad array of functionality. Whether found as mRNA, tRNA, or rRNA, the folding and structure of RNA play a crucial role in its observed biological function. The function of RNA is linked to its tertiary structure that extends from its secondary structure. Knowing or being able to predict the secondary and tertiary structure of RNA may therefore provide insight into the function or role of a specific RNA molecule. Prediction of tertiary structure from secondary structure can be a useful tool as the number of solved RNA structures climbs slowly, while the need for structure information grows quickly. In order to aid in the prediction of tertiary structure, there is a need to more fully understand secondary structure motifs, as well as non-canonical or non-Watson–Crick pairs.
Non-standard nucleotides occur naturally throughout RNA, but they can also be inserted purposefully into RNAs of interest. The use of non-standard nucleotides and their effect on duplex stability has been studied to help further understand the secondary structure of RNA (1–13). The approach of using non-standard nucleotides that have altered bases can also highlight the role of certain functional groups. Grohman et al. (6) explored the mechanism of folding directed by an RNA chaperone. The study made use of the non-standard nucleotide inosine in place of the native guanosine, and thus the loss of the exocyclic amino group, to help explain the mechanism of the folding process.
Purine (P), also known as nebularine, is a purine ring possessing no additional functional groups (Figure (Figure1).1). Due to its structural similarity to (A), P has been used to probe structures to highlight the effects of base modifications on function. For example, P has been used to investigate the activity of adenosine deaminase enzymes (14,15). A-U pairs in RNA duplexes were replaced with P•U pairs, and the effect on enzyme activity was determined (14–18). In these studies, both with P and its analogues, the loss of the functional group inhibited the function of enzymes that recognize A (14–18). This inhibitory effect could be the result of the enzyme requiring the exocyclic amino group for function or could be the result of decreased stability in the target RNAs due to the A to P replacement.
Similarly, P has also been used to investigate hammerhead ribozyme cleavage (19–21) by replacement of A with P. The goal of replacement is to correlate a single functional group with a catalytic event (22). Commonly, the modified nucleotides selected to measure the cleavage rate have multiple functional group changes or deletions (22). P was chosen as a good candidate for replacement of A, since the only structural difference between P and A is the exocyclic amino group of A, which only slightly changes the distribution of partial charges (7). The loss of the amino group at specific positions within the structure of the hammerhead ribozyme revealed a decrease in the cleavage reaction of the ribozyme (19–21). It is noted that the replacement took place at non-base pairing positions in the ribozyme. One study reported that the base stacking of the adjacent nucleotides with P was expected to be unaffected, although this was never tested or confirmed (19).
Blount et al. (22) acknowledged that a nucleotide modification may have consequences beyond hydrogen bonding ability and could alter properties such as stacking affinity and pKa. This has been confirmed in a study by Znosko et al. (7) where the contribution of the A amino group was explored in tandem A•A pairs. In the study, P replaced A at either one or both positions in the tandem pair. The pKa values for A and P were indeed different and were reported as 3.5 and 2.1, respectively. Although the study resulted in hydrogen bonds that were longer in 5΄-PA-3΄/3΄-AP-5΄, the study also reported that the duplex containing 5΄-PA-3΄/3΄-AP-5΄ was 1.3 kcal/mol more stable than the duplex containing 5΄-AA-3΄/3΄-AA-5΄. This thermodynamic difference can be explained by differences in the conformations adopted by these two motifs and differences in stacking interactions. The favorable stacking interaction in the duplex containing 5΄-PA-3΄/3΄-AP-5΄ may be strong enough to compensate for the longer hydrogen bonding. Performing a systematic thermodynamic investigation would aid in understanding the effect of P incorporation on duplex stability, specifically for P•U pairs.
Although not systematic, one previous study did look at the thermodynamic stability of RNA duplexes that incorporated non-standard nucleotides (23). Modified nucleotides, which included P, were introduced in place of various nucleotides for improved potency of siRNA (23). Addepalli et al. (23) set out to explore the effect of thermally destabilized RNA strands in order to find duplexes that dissociate prior to cleavage (23). Although this study did thermodynamically investigate the incorporation of P into RNA duplexes, it only investigated four duplexes (23). These data were not included in the analysis done here as each of the oligoribonucleotides had additional deoxynucleotides at the 3΄ terminus.
In order to further the understanding of secondary structures of RNAs containing non-standard nucleotides, the thermodynamic contribution of P in RNA duplexes was systematically investigated. The results of this study highlight the effect P has on the stability of RNA duplexes, and in turn highlight the importance of the exocyclic amino group of A to duplex stability. The value of a hydrogen bond in the context of the duplexes studied here is compared to values derived from similar methods. Through this process, nearest neighbor parameters that can be implemented in secondary structure prediction software were also derived.
Duplexes were designed to give all combinations of nearest neighbors containing a P•U pair. These combinations were inserted into G-C rich stems to prevent fraying, resulting in duplexes that are 7 base pairs in length for the 16 duplexes containing P•U internal pairs and 6 base pairs in length for the 8 duplexes containing P•U terminal pairs. The sequences were entered into the structure prediction software RNAstructure (24) to check for possible competing structures. As RNAstructure (24) does not recognize P, A was used in its place for each of the sequences. Oligoribonucleotides containing P were synthesized and deprotected by the Keck lab at Yale University (New Haven, CT, USA), while complementary strands were obtained from Integrated DNA Technologies, Inc. Purification of the oligoribonucleotides followed standard TLC and column chromatography procedures (1,2).
Following a previously described method (1,2), RNA single strand concentrations were determined using Beer's law. Extinction coefficients for single strands were calculated using RNACalc (25), substituting the extinction coefficient of A for P. This replacement is justified as in some cases, the practice of replacing the extinction coefficient of a non-standard nucleotide with the extinction coefficient of the corresponding canonical nucleotide is common (1,2,23), yet in other cases, an average of the extinction coefficient of each standard nucleotide is used in place of the nonstandard nucleotide (23). Chen et al. (26) performed UV melting experiments with a duplex containing a P in place of an A. Here, the extinction coefficient of A was used for the extinction coefficient of P with the explanation that although there are differences in extinction coefficients due to functional group changes or deletions, the individual nucleotides only slightly contribute to the overall oligomer extinction coefficient and so the effect on thermodynamic measurements are minimal (26). Absorbance readings for the single strands and duplexes were taken at a temperature above 80°C to minimize unwanted folding of the RNA. Duplexes were formed in a 1:1 molar ratio of single strands. Sample volumes, designed to give absorbance values between 0.2 and 2, were dried down and reconstituted in 100 μl of melt buffer (1 M sodium chloride, 20 mM sodium cacodylate, and 0.5 mM disodium EDTA at pH 7.0).
Optical melting experiments followed a scheme consisting of nine melts resulting from three concentrations subjected to a series of dilutions. This melt scheme gives a concentration range of at least 50-fold. The experiments were performed on a Beckman–Coulter DU800 spectrophotometer at 280 nm with a Beckman–Coulter high performance temperature controller using a heat rate of 1°C/min from 10–90°C.
The optical melt curves were analyzed in Meltwin (25), which fits the melt curves to a two-state model, assuming linear sloping baselines and temperature independent values for enthalpy and entropy (25,27). Meltwin (25) requires the sequence in order to calculate thermodynamic parameters. The sequences were entered into Meltwin (25) with all P's being replaced with A's, therefore using the A extinction coefficient for P (see explanation above). TM values were calculated at various concentrations to determine thermodynamic parameters, according to a previously described method by Borer et al. (28). Values from the melt curve fits were in good agreement with the van't Hoff plots (TM versus log CT), and so only values from the van't Hoff plots were used for further analysis.
The nearest neighbor model was used to isolate the thermodynamic contribution of P•U pairs from the total RNA duplex contribution. Each duplex's thermodynamic contribution can be written as a sum of the nearest neighbor interactions. For example,
is the Gibbs free energy change for the duplex determined from optical melting, is the free energy change for duplex initiation determined to be 4.09 kcal/mol (29) and all other terms are the incremental nearest neighbor interactions. The sum of the P•U nearest neighbor interactions was determined by subtracting the canonical incremental interactions from the experimental duplex free energy (29). For example,
Similar calculations were performed for terminal P•U pairs, yielding one P•U nearest neighbor parameter instead of the summation of two. This same analysis was performed to derive ΔH° and ΔS° contributions from the P•U nearest neighbors.
The eight possible internal P•U nearest neighbor combinations plus a terminal combination were used as variables for linear regression using the LINEST function in Microsoft Excel, with the P•U contributions to ΔG°37 derived from Equation (2) used as constants. The LINEST function simultaneously solved for each variable and resulted in linearly independent nearest neighbor parameters. This same analysis was performed to derive P•U nearest neighbor values for ΔH° and ΔS°.
Table Table11 lists the thermodynamic parameters of duplex formation derived from both van't Hoff plots (TM−1 versus log (CT/4)) and fitting the optical melting curves to a two-state model. Two duplexes, and , did not show single, sharp transitions during melting and were not included in any analysis including trends, averages or linear regression. The remaining duplexes did show single, sharp transitions and the enthalpy values derived from the average of curve fitting and the van't Hoff plots agree within 15% of each other, suggesting that a two-state transition occurred during duplex melting (30). A representative melt curve and van't Hoff plot can be found in Supplementary Data Figures S1 and S2.
Table Table22 lists the ΔG°37 measured from optical melting experiments of duplexes containing a P•U pair compared to the ΔG°37 from nearest neighbor calculations of the same duplex with an A-U pair replacing the P•U pair. Duplexes containing an internal P•U pair were, on average, 1.78 kcal/mol less stable than the corresponding duplex with an internal A-U pair, while duplexes containing a terminal P•U pair were 1.42 kcal/mol less stable, on average, than the corresponding duplex with a terminal A-U pair. Similar calculations and comparisons were done for enthalpy and entropy values, and these can be found in Supplementary Data Tables S1 and S2.
The contribution of P•U pairs to duplex thermodynamics is given in Supplementary Table S3. These results include those described by Equations (1) and (2) and the same calculations for enthalpy and entropy. Table Table33 lists the nearest neighbor parameters derived for P•U pairs, as described in the Materials and Methods section. All P•U nearest neighbors contribute a negative enthalpy and negative entropy to duplex thermodynamics. and both have a positive free energy contribution, while all other generated free energy nearest neighbor parameters were negative. The penalty of 0.86 kcal/mol for a terminal P•U pair is twice as destabilizing than the penalty of 0.45 kcal/mol for a terminal A-U pair (29).
The newly derived nearest neighbor parameters were used in conjunction with the nearest neighbor model (29) to give predictive thermodynamic values for duplexes containing P•U pairs. These predicted values can be compared to the experimental values and are given in Table Table1.1. The average deviations between the predicted and experimental values are 4.5%, 5.8% and 6.1% for ΔG°37, ΔH°, and ΔS°, respectively. These deviations are comparable to previous model deviations reported for Watson–Crick (3.2%, 6.0%, and 6.8%, respectively) (29), I•U (5.1%, 4.6%, and 5.1%, respectively) (1), and Ψ•A (1.7%, 6.7%, and 8.0%, respectively) (2) parameters.
The thermodynamic parameters derived from the optical melting experiments are given in Table Table1.1. An enthalpy agreement of 15% between the average of curve fitting and the van't Hoff plots indicates a two-state transition during melting. All duplexes studied met this criterion. There were, however, two duplexes, and , that were excluded from all analyses due to a lack of single, sharp transitions during melting and the possibility of competing unimolecular interactions. For example, the single strand 5΄-GCGCGU-3΄ could form a duplex with itself consisting of four G-C pairs and two terminal G-U wobble pairs. The melting of the remaining duplexes did show single, sharp transitions and the van't Hoff plots showed a positive correlation between melting temperature and RNA concentration, which suggests duplex formation.
The stability of duplexes containing P•U pairs was compared to the stability of duplexes containing A-U pairs, and the results can be found in Table Table2.2. The duplexes containing internal P•U pairs were, on average, 1.78 kcal/mol less stable than the corresponding duplexes containing an internal A-U pair. The deletion of the amino group, a hydrogen bond donor, in the P•U pair shows more of a destabilizing effect here than the effect reported for the deletion of the amino group in G-C to I•C pair substitutions (~1 kcal/mol) (6). This difference may be attributed to cooperative effects between hydrogen bonds within a pair and cooperative effects between neighboring base pairs (9). More specifically, when substituting an I•C pair for a G-C pair, there is a loss of one hydrogen bond but retention of two hydrogen bonds. When substituting a P•U pair for an A-U pair, there is the loss of one hydrogen bond but retention of only one hydrogen bond. The remaining lone hydrogen bond may experience a greater loss of stability due to a cooperativity effect. This greater loss of stability could also be attributed to greater flexibility and rotation of the base due to the lost hydrogen bond. Stacking and hydrogen bonding have been shown to compete in base pairs (10). With only one hydrogen bond holding the P•U pair together, versus two in an A-U pair or I•C pair, the bases may move into an orientation aimed at maximizing the hydrogen bonding strength resulting in weaker stacking interactions. Similarly, the bases could move into an orientation aimed at maximizing the stacking interactions resulting in weaker hydrogen bonding. Structural studies could explain if different orientations are contributing to the destabilization. As of yet, there are no structures in the PDB with a P•U pair surrounded by Watson–Crick pairs.
There are differences in stability for duplexes containing terminal P•U pairs depending on if the P is at the 5΄ or 3΄ end of the strand. Incorporating a terminal P•U pair with P at the 5΄ end of a duplex (see 5΄ terminal duplexes in Table Table2)2) was, on average, 0.64 kcal/mol less stable than an A-U pair, while incorporating a terminal P•U pair with P at the 3΄ end (see 3΄ terminal duplexes in Table Table2)2) was, on average, 1.81 kcal/mol less stable than the corresponding duplex with an A-U pair. This result shows differences based on position within the helix that can be attributed to differences in stacking. This result is similar to results seen in the G-C to I•C pair substitutions where a hydrogen bond is worth −1.6 kcal/mol or −0.7 kcal/mol depending on whether the I is the at the 5΄ or 3΄ end (10). Again, without structural information, it is hard to determine the source of the differences seen here.
Similar to nearest neighbor trends for Watson–Crick/Watson–Crick pairs, all derived enthalpy and entropy nearest neighbor parameters for P•U pairs are negative. Conversely, the derived free energy nearest neighbor parameters have both positive and negative parameters, ranging from −1.94 kcal/mol to 0.39 kcal/mol. Both and contribute a positive, unfavorable, term to duplex ΔG°37. This has been seen for previously derived free energy parameters for I•U pairs, both and contributed a positive term to duplex ΔG°37 (1). These nearest neighbor combinations are similar to the and nearest neighbor combinations, two of the least stabilizing Watson–Crick nearest neighbor parameters. It is not surprising that when a hydrogen bond is removed (when A is replaced by P), the nearest neighbor parameters become destabilizing. It is important to remember, however, that the incorporation of a P•U pair into a non-terminal position in a duplex involves two nearest neighbor parameters involving P•U. Therefore, the contribution of one of these nearest neighbor parameters will be summed with a second P•U nearest neighbor parameter. In almost all cases, this will result in a stabilizing contribution from the inclusion of a P•U pair.
The general trend when A-U is the 3΄ pair in a nearest neighbor combination from most stable to least stable is > > > (20). This is the same trend for when a P•U pair is the 3΄ pair in a nearest neighbor combination, or ≈ > ≈ . The similarity of stability trends between Watson–Crick pairs and P•U pairs can also be seen when a P•U pair is the 5΄ pair in a nearest neighbor combination, ≈ > ≈ , except here, the two least stable nearest neighbor parameters are switched from their Watson–Crick counterparts (although the P•U parameters in question are within experimental error of each other). In both instances where a P•U pair is in a nearest neighbor combination, the nearest neighbors with a G-C pair are more stable than the nearest neighbors with an A-U pair, as is seen for canonical pairs (29). This is likely due to the extra hydrogen bond in a G-C pair versus an A-U pair. Also, each P•U nearest neighbor combination contributes less to duplex stability than the corresponding A-U nearest neighbor pair. This result is consistent with the decrease in stability due to the removal of a hydrogen bond. It is also possible that differences in stacking contribute to the differences between P•U nearest neighbor combinations containing G-C pairs versus A-U pairs and contribute to the differences between P•U nearest neighbor combinations versus A-U nearest neighbor combinations.
It is also noted that the ΔG°37 term is more negative (i.e. more stable) when the P•U pair is the 5΄ pair in a nearest neighbor combination, or , than when the P•U pair is the 3΄ pair, or . This result is similar to results seen with A•A mismatches where A was replaced with P giving P•A pairs (7). Although the P is paired with A instead of U, the duplex with the P•A as the 5΄ pair or , showed an increase in stability and the duplex with P•A as the 3΄ pair, or , showed a decrease in stability in comparison to the A•A mismatch (7). However, some of this difference for P•A pairs may be attributed to differing base pair orientations, as the hydrogen bonding is trans Hoogsteen/sugar-edge for the P•A pair and trans Watson–Crick/Watson–Crick for the A•P pair (7).
The effect on the stability of RNA duplexes reported for hydrogen bonds can vary depending on the location and type of the hydrogen bond. Table Table44 summarizes the free energy differences in duplexes from the literature where nucleobase replacements have removed hydrogen bonds (31). The results in Table Table22 give energy differences equivalent to the loss of a hydrogen bond in the context of duplexes where an A-U pair was replaced with a P•U pair. The values included in Table Table44 for the duplexes with a P•U pair replacement are the average values given in Table Table2.2. Position in the helix, number of hydrogen bonds, and type of hydrogen bond play a role in contributing to duplex stability. For example, 2,6-diaminopurine (D)•U and G-C pairs have the same number but different type of hydrogen bonds. G-C and A-U pairs have different number and different type of hydrogen bonds, and A-U and P•U pairs have different number but same type of hydrogen bonds, as do G-C and I•C pairs.
The free energy value of a hydrogen bond, given in Table Table4,4, ranges from −0.4 to −3.4 kcal/mol. The free energy value of a hydrogen bond in a P•U pair with various nearest neighbor combinations, given in Table Table2,2, is similar and ranges from 0.3 to −3.0 kcal/mol. Also, the effect of a hydrogen bond on the stability of a duplex is similar for DNA versus RNA. The energy values for a hydrogen bond in duplexed DNA versus duplexed RNA range from −0.5 to −1.8 kcal/mol and −0.4 to −1.9 kcal/mol, respectively.
Still, with these similarities there are differences. For example, the range for the value of a hydrogen bond in RNA hairpins is −1.8 to −3.4 kcal/mol, whereas the range for duplexed RNA is −0.4 to −1.9 kcal/mol. Also, there is a difference in energy when an A-U pair replaces a G-C pair depending on the location of the base pair and the sequence composition of the duplex. If the replacement occurs at the end of a duplex with a -GCGC- core, the hydrogen bond value is −0.5 kcal/mol (entry 8 of Table Table4).4). If the replacement occurs at the end of a duplex with an -AUGCAU- core, the hydrogen bond value is −1.5 kcal/mol (entry 9 of Table Table4).4). This difference is large considering that both duplexes have a G-C to A-U replacement at the helix termini. However, if the replacement (G-C to A-U) occurs at the center of a duplex, the hydrogen bond value is −1.6 kcal/mol (entry 7 of Table Table4)4) or −1.9 kcal/mol (entry 11 of Table Table4)4) depending on the sequence composition. Here, the difference is relatively small.
The internal and terminal average values from this work are included in Table Table44 as entries 15, 16, and 17. All internal replacements for duplexed RNA without P•U pairs range from −1.5 kcal/mol to −1.9 kcal/mol (Table (Table4).4). The average value for internal A-U to P•U pair replacement is −1.8 kcal/mol (entry 17 of Table Table4),4), which is within the range for internal replacements for duplexed RNA without P•U pairs. Although this average value falls within the range observed for internal replacements of duplexed RNA without P•U pairs, there is variation in the individual values for internal A-U to P•U pair replacement, and some of the individual values (Table (Table2)2) are outside of this range.
On the other hand, all terminal replacements for RNA without P•U pairs (entries 1, 2, 4, 5, 8, and 9 of Table Table4)4) range from −0.4 kcal/mol to −1.6 kcal/mol. The terminal average values for duplexes with A-U to P•U pair replacement (entries 15 and 16 of Table Table4)4) are −1.8 kcal/mol for the terminal pair with the 3΄ P and −0.6 kcal/mol for the terminal pair with the 5΄ P. Although the value for the terminal pair with the 5΄ P is within the range observed for terminal replacements of RNA without P•U pairs, the value for the terminal pair with the 3΄ P is not. The range of values for these entries highlights the effects from not only the location within a duplex, but also the neighboring base pairs, on the value of a hydrogen bond. The type of hydrogen bonds can also have an effect on the value of a hydrogen bond. A G-C to A-U replacement at the terminus of a duplex results in a pair with three hydrogen bonds being replaced by a pair with two hydrogen bonds, just as in a D•U to A-U replacement. Yet, the value for a terminal G-C to A-U replacement is −1.4 kcal/mol (entry 4 of Table Table4)4) and the value for a terminal DA•U replacement is −0.4 kcal/mol (entry 5 of Table Table4).4). The type of hydrogen bond having an effect on the value of the hydrogen bond is even further highlighted when comparing an A-U pair to a 2-aminopurine•U pair. Here, the number of hydrogen bonds has not changed but the type of hydrogen bonds has changed. An RNA hairpin where an A-U to 2-aminopurine•U pair replacement occurred at the stem region had a difference in free energy of 0.8 kcal/mol (8). The range of values found in Tables Tables22 and 4 suggests that a single value representing the energy of a hydrogen bond may be insufficient for understanding the contribution to duplex stability by functional groups that participate in hydrogen bonds.
The results of this study indicate that duplexes containing P•U pairs are less stable than the same duplexes containing A-U pairs, which is as expected. The derived nearest neighbor parameters can be used in structure prediction software to determine the stability of RNA duplexes containing P•U pairs. These results highlight the destabilization to a duplex upon the loss of one hydrogen bond. The destabilization effect can vary depending on the location within the duplex and the nearest neighbors to the P•U pair. Structural studies or computational studies may further explain the source of the destabilization due to P•U pair substitution.
Present address: Brent M. Znosko, Department of Chemistry, Saint Louis University, St Louis, MO 63103, USA.
Supplementary Data are available at NAR Online.
National Institutes of Health (NIH) [2R15GM085699-02]. Funding for open access charge: National Institutes of Health (NIH) [2R15GM085699-02].
Conflict of interest statement. None declared.