Non-Nearest-Neighbor Dependence of Stability for RNA Bulge Loops Based on the Complete Set of Group I Single Nucleotide Bulge Loops†

RNA Synthesis and Purification

Most oligomers were synthesized on CPG solid supports (Applied Biosystems 392 DNA/RNA Synthesizer) utilizing phosphoramidites with the 2’ hydroxyl protected as the tert-butyl dimethylsilyl ether from Glen Reseach (Sterling VA) (23, 24). Oligomers underwent ammonia and fluoride deprotection, and crude sample was purified using preparative TLC (n-propanol:ammonium hydroxide:water, 55:35:10) and Sep-Pak C18 (Waters) chromatography. Some oligomers were ordered from Dharmacon and deprotection of the oligomers was carried out using the manufacture’s instructions. The oligomers were then purified as described above. Sample purity was determined through analytical TLC or HPLC (C-8), and was greater than 95%.

Melting Curve and Data Analysis

For non-self complementary sequences, individual strand concentrations were calculated from high-temperature single strand absorbance at 260 and 280 nm using nearest-neighbor extinction coefficients (25, 26). Single strands were then annealed in a 1:1 molar ratio. Optical melting experiments were performed using a Beckman DU 640 Spectrophotometer and High Performance Temperature Controller at 260 or 280 nm. Absorbance changes for oligomers in 1 M NaCl melt buffer (1 M NaCl, 0.01 M cacodylic acid, 0.5 mM EDTA, pH 7.0) were recorded as a function of temperature from 95-10°C at a rate of 1°C/min as described previously (27). The experiment was repeated at ten varying sample concentrations to give at least a 50-fold concentration range (10 μM-1mM) for each sample. Absorbance versus temperature profiles were fit to a two-state model with sloping base lines using a nonlinear least squares program (28). Thermodynamic parameters for duplex formation were obtained by two methods: (1) enthalpy and entropy changes from the fits of the individual melting curves were averaged and (2) plots of the reciprocal melting temperature, T_M^-1, versus log C_t/4 gave enthalpy and entropy changes (29):

Here, C_t is the total concentration of oligomer. Parameters derived from the two methods generally agreed within 15%, consistent with the two-state model (30, 31). The Gibbs free energy change at 37 °C was calculated as:

Determination of the contribution of bulge loops to duplex thermodynamics

The free energy of duplex formation can be approximated by the nearest-neighbor model (32). The free energy contribution of each bulged nucleotide was calculated from the experimental data and the nearest-neighbor model according to equation 3, where ΔG°_37(measured) is the experimentally determined value from the melts and

ΔG°_37(duplex) is calculated from the nearest-neighbor model for the duplex as if it did not contain the bulge. ΔH° values are calculated in a similar fashion.

Phylogenetic Analysi

A database of phylogenetically determined RNA secondary structures (33) of 305 SSU rRNAs, 169 LSU rRNAs, 16 Group I intron RNAs and 7 Group II intron RNAs, was searched for group I single nucleotide bulge loops. The loops were characterized by nearest-neighbor base pairs and bulge identity.

Statistical Analysis

Statistical analysis of the data was done using the statistical software available with GraphPad Prism and GraphPad Instat.

Nearest-Neighbor Influences on the Thermodynamics of Bulge Loops Adjacent to Watson-Crick Base Pairs

Since the range of thermodynamic contributions of bulge loops adjacent to Watson-Crick base pairs to duplex formation (Table 3) is relatively large, 1.3-5.4 kcal/mol, we reinvestigated the influence of the bulge nearest neighbors on the thermodynamic contribution of bulge loops on duplex formation. Figure 1 displays the plot of the free energy of the nearest-neighbor base pairs versus the free energy of the bulge. Initially, the purine and pyrimidine bulge data were plotted separately and the linear regression for each set determined. The two linear regression lines were found not to be statistically different for each other. Therefore, the two data sets were combined and the linear regression of the entire set of group I single nucleotide bulge loop data determined. The slope of the regression line (0.05 ± 0.16) was not significantly different from zero, further indicating that neither the identity of the bulge nor its nearest neighbors influence the thermodynamics for bulge loop insertion on duplex formation.

Non-Nearest-Neighbor Influences on the Thermodynamics for Bulge Loops Adjacent to Watson-Crick Base Pairs

The data highlighted in red in Table 3 were originally singled out because they seemed to represent single nucleotide bulge loops that were much less destabilizing than the remainder of the bulges examined. For example, in Table 3, the bulge sequence $\left(\begin{array}{cccc}5’& G& C& U\\ 3’& C& & A\end{array}\right)$, imbedded within two different duplex sequences influences the stability of the duplex by either 1.3 or 4.3 kcal/mol. In an attempt to understand the origin of the difference thermodynamic contribution of these bulges to duplex formation, we examined a number of non-nearest- neighbor factors which might contribute to the difference. The first difference we noted was that the red data points were derived from bulge loops imbedded within relatively short helix stems. Figure 2a examines the influence of helix size upon the thermodynamic contribution of the bulge to duplex formation. Although the red data points do indeed represent short helices, many of the small helices (hexamer) had bulges which were not unusually destabilizing. Next we investigated whether the stability of the parental helix was related to the thermodynamic contribution of the bulge to helix formation. This analysis is shown in Figure 2b. There was a clear relationship between the stability of the duplex and the thermodynamic contribution of the bulge to helix formation. A linear fit of the data in figure 2b gives in the following: ΔG°_{37bulge loop} = (-0.27*ΔG°_37duplex) + 1.0 (r² = 0.46). Where ΔG°_{37bulge loop} is the free energy increment (in kcal/mol) for insertion of a bulge loop into a duplex and ΔG°_37duplex is the free energy of the duplex without the bulge. Finally, we compared the stability of the less stable duplex stem with the thermodynamic contribution of the bulge to helix formation (Figure 2c). A linear fit of the data in figure 2c gives in the following: ΔG°_{37bulge loop} = (-0.51*ΔG°_37duplex) + 1.0 (r² = 0.32). The bulge loops which are less destabilizing to duplex formation (in red in Table 3) are found inserted into duplexes where the duplex or one of its stems are relatively unstable. While the fit to the data in Figure 2c is not as good as in Figure 2b, inclusion of the data for bulge loops adjacent to wobble base pairs gives a more meaningful relationship (see below).

Non-Nearest-Neighbor Influences on the Thermodynamics for Bulge Loops Adjacent to Wobble Base Pairs

A similar non-nearest-neighbor analysis of the data for bulge loops adjacent to a wobble base pair is presented in Figure 3. The data for the bulge loops adjacent to Watson-Crick base pairs is included in Figure 3 for comparison. Figure 3a examines the influence of helix length upon the thermodynamic contribution of the bulge to duplex formation. For the bulges adjacent to wobble base pairs, several of the less destabilizing bulges were found within longer duplexes of seven or eight base pairs and as seen for the bulges adjacent to Watson-Crick base pairs, many of the small helices (hexamer) had bulges which were not unusually less destabilizing. Next we investigated whether the stability of the parental helix was related to the thermodynamic contribution of the bulge to helix formation. This analysis is shown in Figure 3b. As for the bulges adjacent to Watson-Crick base pairs, there was a relationship between the stability of the duplex and the thermodynamic contribution of the bulge to helix formation. A linear fit of the data for bulge loops adjacent to wobble base pairs gives the following: ΔG°_{37bulge loop} = (-0.42*ΔG°_37duplex) - 0.7 (r² = 0.37). This line was not significantly different than the line for the bulge loops adjacent to Watson-Crick base pair, so both sets of data were combined. The linear fit for the combined data give the following: ΔG°_{37bulge loop} = (-0.33*ΔG°_37duplex) -0.7 (r² = 0.41). Where ΔG°_{37bulge loop} is the free energy increment (in kcal/mol) for insertion of a bulge loop into a duplex and ΔG°_37duplex is the free energy of the duplex without the bulge. Finally, we compared the stability of the less stable duplex stem with the thermodynamic contribution of the bulge adjacent to a wobble base pair to helix formation (Figure 3c). A linear fit of the data for bulge loops adjacent to wobble base pairs in figure 3c gives the following: ΔG°_{37bulge loop} = (-0.66*ΔG°_{37less stable stem}) + 0.0 (r² = 0.37). This line was not significantly different than the line for the bulge loops adjacent to Watson-Crick base pair, so both sets of data were combined. The linear fit for the combined data give the following: ΔG°_{37bulge loop} = (-0.62*ΔG°_{37less stable stem}) -0.25 (r² = 0.38). While the fit of the data in figure 3c I slightly better than the fit in figure 3b, the difference is so small as to be meaningless. So in considering the two relationships, we found using the less stable stem as more meaningful, because in Figure 3c, all of the less destabilizing bulge data points are now clustered. They all appear to be inserted adjacent to a stem that had a stability of less than 5 kcal/mol. These results are summarized in Table 6.

In an earlier report on bulge loops (15), non-nearest-neighbor interactions were observed for the thermodynamic of bulge loop insertion into a duplex. Substitution of a CG base pair for an AU base pair, two nucleotides removed from the bulge caused the bulge to be more destabilizing. It was also observed that the addition of a 3’ dangling end also caused the bulge to have more of a destabilizing effect. Both of these observations are in concordance with the relationship described above. In both cases, substitution of a GC base pair for an AU base pair and the addition of a 3’ dangling end increases the stability of the stem adjacent to the bulge and therefore should increase the destabilizing effect of the inserted bulge.

Similar non-nearest-neighbor effects have been observed with other structural motifs. In particular, there are several reports where non-nearest-neighbor effects have influenced the thermodynamics of single mismatches in RNA duplexes (42, 43). The position of the single mismatch with a duplex has been shown to influence the stability of the duplex. As the bulge was moved closer to the terminus (decreasing the stability of the stem), the stability of the duplex increased. This effect was somewhat sequence specific as it was observed for UU and AA single mismatches but not for GG (42).

The influence of the adjacent stem on the thermodynamics of bulge loop insertion into an RNA helix may be related to the tight packing and electrostatic stain present in the RNA duplex (43). The lower the stability of the adjacent stem, the easier the structural perturbation from the insertion of the bulged nucleotide can be dissipated along the neighboring base pairs. Therefore, as the stability of the neighboring stem decreases, it becomes easier (less energetically costly) to insert a bulge loop, or other structural motif (e.g. single mismatch).

Nearest-Neighbor Influences on the Thermodynamics for Bulge Loops Adjacent to Wobble Base Pairs

In our analysis thus for, we have considers the bulge loops adjacent to wobble base pairs as a single group, but in fact, the position and orientation of the nearest-neighbor wobble base pair does influence thermodynamic of bulge loop formation. These influences are examined in figure 4, where the free energy increment for bulge loop formation is graphed as a function of the position and orientation of the wobble base pair in relation to the bulge. In figure 4, the free energy increment for bulge loop formation as a function of the less stable stem for each family of bulge loops is graphed relative to the least squares fit for all of the data points. For example, if we focus on the bulge sequences of the type $\left(\begin{array}{cccc}5’& U& B& X\\ 3’& G& & Y\end{array}\right)$, all 10 of the bulges data points fall below the line by an average of 0.6 kcal/mol. Therefore, these bulges have a smaller destabilization on duplex formation than other bulge loops. This may be caused by the geometry of the GU base pair which produces a change in the twist angle of the RNA duplex (35). For sequences of the type $\left(\begin{array}{ccc}5’& U& X\\ 3’& G& Y\end{array}\right)$, the twist angle between the base pairs is increased to about 40° which leads to unstacking of the wobble base pair with the adjacent base pair in the helix (Figure 3A). $\left(\begin{array}{ccc}5’& U& X\\ 3’& G& Y\end{array}\right)$ nearest-neighbor base pairs are among the less stable. For sequences of the type $\left(\begin{array}{cccc}5’& X& B& G\\ 3’& Y& & U\end{array}\right)$, five of the seven bulges have values that fall above the line although the average is only 0.2 kcal/mol greater, within the error of the experiment. For sequences of the type $\left(\begin{array}{cccc}5’& G& B& X\\ 3’& U& & Y\end{array}\right)$ and $\left(\begin{array}{cccc}5’& X& B& U\\ 3’& Y& & G\end{array}\right)$, 11 of the 15 bulges have values that fall above the line, indicating that they are more destabilizing than other bulges. Excluding the very less destabilizing $\left(\begin{array}{cccc}5’& G& U& G\\ 3’& U& & C\end{array}\right)$ bulge (first value in Table 5), the average value is 0.4 kcal/mol above the line. In a similar analysis of the geometry of a $\left(\begin{array}{cccc}5’& G& X& \\ 3’& U& & Y\end{array}\right)$ base pair, the wobble base pair decreases the twist angle between the base pairs (Figure 3B). In this instance, the base pair still stacks well with the wobble pair, and these nearest-neighbor pairs are among the more stable, so the bulge has a greater potential to disrupt the nearest-neighbor base pairs. For the bulges inserted between two wobble base pairs, all five of the data points fall below the line by an average of 0.6 kcal/mol. Consecutive wobble pairs also leads to unstacking of the bases as seen in Figure 3C. That a bulge between two adjacent wobble base pairs is the least destabilizing may be due to the fact that the tandem GU pairs are among the least stable nearest-neighbor base pair. In fact they contribute a positive free energy to duplex formation (36). These results are summarized in Table 4. The parameters in Table 4 can be combined with the nearest-neighbor parameters for duplex formation (32) to predict the thermodynamic stability of RNA duplexes containing bulge loops.

In Figure 4, five data points are circled, three that correspond to bulge loops that are significantly less destabilizing than predicted from the linear regression and two that are more destabilizing than predicted. The three sequences that are less destabilizing are: $\left(\begin{array}{ccccccccc}5\text{'}& A& C& U& G& U& G& C& G\\ 3\text{'}& U& G& A& U& & C& G& C\end{array}\right)$, $\left(\begin{array}{cccccccc}5\text{'}& G& C& U& A& G& A& C\\ 3\text{'}& C& G& G& & C& U& G\end{array}\right)$, and $\left(\begin{array}{cccccccccc}5\text{'}& G& A& C& A& U& A& G& U& C\\ 3\text{'}& C& U& G& & G& U& C& A& G\end{array}\right)$. There does not appear to be any common nearest-neighbor nor non-nearest-neighbor features that would explain their free energy increment. The position and orientation of the bulge is different in each case as well as the differences in the length and stability of the stems relative to the position of the bulge. In other sequence contexts, all three of the bulges have approximately the predicted value for the free energy increment of duplex destabilization. The two bulges that were circled as more destabilizing than predicted have the sequences: $\left(\begin{array}{ccccccccccc}5\text{'}& C& A& G& U& A& U& A& G& U& C\\ 3\text{'}& G& U& C& A& & G& U& C& A& G\end{array}\right)$ and $\left(\begin{array}{ccccccccccc}5\text{'}& C& A& G& U& G& U& A& G& U& C\\ 3\text{'}& G& U& C& A& & G& U& C& A& G\end{array}\right)$. In this case both bulges are within the same duplex context. Again, there is no obvious explanation for the larger free energy increment for these bulges.

Enthalpic Contributions of Group1 Single Nucleotide Bulge Loops to Duplex Stability

Since the free energy increment of the insertion of a bulge loop into a duplex was dependent upon the stability of the stem adjacent to the insertion, we investigated the influence of stem stability on the enthalpy of bulge loop insertion. Figure 6 displays the enthalpy of bugle loop insertion as a function of the stability of the less stable stem of the duplex for bulges adjacent to both Watson-Crick and wobble base pairs. The least squares fit for the Watson-Crick and wobble data were first considered separately.

Neither line had a slope significantly different than zero so the data sets were combined. The slope of the line for the combined Watson-Crick and wobble enthalpy values were again not significantly different than zero suggesting that the average enthalpy value is a reasonable approximation for the enthalpic energy contribution. The average enthalpic value for the combined data is 16.5 ± 11.4 kcal/mol. The enthalpy value can be used in conjunction with the free energy and enthalpy parameters for other nearest-neighbor motifs to determine the stability of RNA structures at temperatures other than 37 °C (34).

Phylogenetic Analysis of Single Nucleotide Bulge Loops

The database examined in this study contained 4330 group I single nucleotide bulge loops, 3520 closed with Watson-Crick and 810 closed wobble base pairs. The frequency of occurrence for the group I single nucleotide bulge loops are listed in Tables 3 and 6. For the Watson-Crick closed group I single nucleotide bulge loops, nearly 70 % of the total represent adenosine bulge loops. No other bulge sequence occurs more frequently than 4% except than (GUG/C C). Since the stability of group I single nucleotide bulge loops is independent of the identity of the bulge, there is no correlation between the thermodynamic contribution of the bulge and its frequency of occurrence. Therefore, the selection of naturally occurring bulge nucleotide sequences must be related to factors other than stability.

When the bulge has a wobble nearest neighbor, the position and orientation of the bulge does effect its contribution to stability. Nearly 90% of the naturally occurring bulged nucleotides are purine residues, with adenosine representing nearly 80%. The most frequent (76%) type of purine bulges have a 5’ U/G, the more stable orientation; while the less stable 3’U/G represents less than 15% of the naturally occurring bulge sequences. However, the most stable bulge loop (with two adjacent wobble base pairs) represent less than 2% of the naturally occurring sequences and several bulge combinations were not observed in the database.

Single Nucleotide Bulge Loops in Natural Contexts

The effect of stem stability on the influence of single nucleotide bulge loop on duplex formation is particularly important to the analysis of naturally occurring RNAs where the average duplex is less than 7 nucleotides. In fact, of the 20 single nucleotide bulge loops found in the small and large E. coli ribosomal RNA (33), 19 have a stem with stability less than 5 kcal/mol. The average stability for the less stable stem is approximately 3 kcal/mol, excluding the influence of terminal mismatches. Therefore, because of the low stability of duplex stems in natural RNAs, the influence of single nucleotide bulge loops on duplex stability is less than previously thought (16).

Although the experiments from this and previous bulge loops studies (15, 16) are by agreement performed at 1M NaCl and show that bulges are generally destabilizing to double-stranded regions of RNA. The conformations of bulged nucleotides can afford them stabilization through binding of other biologically relevant ligands. Bulges can be stabilized via either direct or water-mediated interactions with divalent cations (42), often within the context of tertiary interactions (43-46). Examples include interactions of the A-rich bulge of P4-P6 of the group I intron (43), metal binding in the bulge region of HIV-1 TARN NRA (44, 45), or within the bulged region of the leadzyme (46). In addition, bulges function as critical portions of recognition elements for RNA-protein interactions (47-50). Thus, the destabilization of double stranded RNA by bulge loops is often necessary in order for RNA to properly fold into its functional structures or form RNA-protein complexes which serve to offset the destabilization of the secondary structure.