Influence of a Hairpin Loop on the Thermodynamic Stability of a DNA Oligomer

DSC was used to evaluate the mechanism of the thermally induced unfolding of the single-stranded hairpin HP = 5′-CGGAATTCCGTCTCCGGAATTCCG-3′ and its core duplex D (5′-CGGAATTCCG-3′)2. The DSC melting experiments performed at several salt concentrations were successfully described for HP and D in terms of a three-state transition model HP↔I (intermediate state) ↔ S (unfolded single-stranded state) and two state transition model D↔2S, respectively. Comparison of the model-based thermodynamic parameters obtained for each HP and D transition shows that in unfolding of HP only the HP↔I transition is affected by the TCTC loop. This observation suggests that in the intermediate state its TCTC loop part exhibits significantly more flexible structure than in the folded state while its duplex part remains pretty much unchanged.


Introduction
Hairpin loops are a common form of nucleic acid secondary structure and are crucial for tertiary structure and function [1]. They are known to play a key role in a number of biological processes such as gene expressions, DNA recombination, and DNA transposition [2][3][4]. In RNA molecules hairpins act as nucleation sites for RNA folding into final conformations [5][6][7] and play a critical role in RNA-protein recognition and gene regulation [8,9]. Furthermore, due to the specificity of probe/target hybridization determined as a match-versus-mismatch discrimination, hairpin DNA oligomer probes have become an important tool in modern biotechnology and diagnostics [10,11]. The thermodynamics and kinetics of hairpin formation, hairpin binding to complementary nucleic acids, and hairpin-ligand associations have been studied extensively [12][13][14][15][16][17][18][19][20][21]. There is no doubt that studies of hairpin-to-coil transitions and hairpinligand binding affinity and specificity have greatly enhanced our understanding of structural features and function of the naturally occurring nucleic acids [22,23]. However, despite extensive biophysical research on the systems involving hairpin structures that produced a number of high-quality explanations and evaluations on properties and behavior of nucleic acids containing hairpin formations, there are still many unresolved questions.
As pointed out by Marky et al. [24] the most suitable hairpin molecules for studying the thermodynamics of their conformational transitions and ligand binding are the singlestranded hairpin molecules. They form stable partially paired duplexes that tend to melt in simple monomolecular transitions. Furthermore, their conformational stability and ligand binding properties are easily compared with those of the corresponding core duplexes. In this way one can evaluate the contributions of the loops to the thermal stability of the hairpins. Despite the simple structure of single-stranded hairpins it is not clear whether their monomolecular folding/unfolding transitions occur in a two-state or multistate manner. Measurements of their thermally induced unfolding transitions followed by UV, CD, and/or fluorescence spectroscopy as a rule result in sigmoidal melting curves suggesting that they may be considered as two-state processes. The same conclusion can be reached also on the basis of DSC measurements performed on the same sample solutions in the older generation of less sensitive DSC instruments (e.g., Microcal MC-2) which resulted in single-peak DSC thermograms. Recent measurements of conformational transitions of DNA quadruplex structures have shown, however, that the sigmoidal shape of UV or CD melting curves may be misleading. Namely, the DSC measurements performed on samples for which sigmoidal UV and CD melting curves were 2 Journal of Nucleic Acids observed using the DSC of the latest generation (CSC, Microcal) resulted in thermograms containing two or three well-distinguished peaks thus indicating that the observed DNA melting process occurs in a multi-state manner [25,26]. Furthermore, recent T-jump experiments performed on small hairpin molecules have produced a direct evidence that their unfolding transitions involve intermediate structures and thus cannot be considered as two-state processes [19,[27][28][29].
In our DSC study of the unfolding mechanism and stability of the 5 -CGGAATTCCGTCTCCGGAATTCCG-3 hairpin we performed the DSC melting experiments on the hairpin and its core duplex, (5 -CGGAATTCCG-3 ) 2 (see Figure 1), at several salt concentrations using an extremely sensitive microcalorimeter (CSC). To see to what extent the TCTC loop affects the hairpin unfolding process we attempted to describe for each oligonucleotide the measured DSC thermograms in terms of the simplest possible unfolding model. We derived the corresponding model functions and by fitting them to the experimental data we tested the appropriateness of the suggested models and obtained for each transition the characteristic thermodynamic quantities of transition ΔG 0 (T) , ΔH 0 (T) , ΔS 0 (T) , and Δc 0 P . By comparing these values determined for the hairpin and the core duplex we tried to estimate the contribution of the TCTC loop to the stability of the hairpin.

Materials and Methods
2.1. Materials. Self-complementary oligonucleotide 5 -CG-GAATTCCG-3 and oligonucleotide 5 -CGGAATTCCGT-CTCCGGAATTCCG-3 that in solution at room temperature form a duplex (D) and a single-stranded hairpin structure (HP), respectively, were purchased HPLC pure from Invitrogen Co., Germany and used without any further purification. Their concentrations in buffer solution (10 mM phosphate buffer and 1 mM Na 2 EDTA adjusted to pH = 7.0)) in the presence of 100 mM NaCl were determined at 25 • C spectrophotometrically in the Cary Bio 100 UV-spectrophotometer. The molar extinction coefficients were determined using the nearest neighbor data of Cantor et al. [30] for single-stranded DNA at 25 • C and the absorbance at 260 nm of thermally unfolded oligonucleotide extrapolated back to 25 • C (ε D260 = 84600 M −1 cm −1 , ε HP260 = 216000 M −1 cm −1 ). The phosphate buffer solutions used in all experiment contained 0, 0.1, 0.3, or 1.0 M NaCl.
Differential scanning calorimetry (DSC). Thermally induced unfolding of duplex (D) and hairpin (H) in buffer solutions with different added NaCl concentrations was followed between 5 and 95 • C in a Nano-II DSC calorimeter (CSC; UT) at the heating rate of 1 • C/min and essentially the same results were obtained from several test-experiments performed at the heating rate of 0.25 • C/min. The thermally induced unfolding of both oligonucleotides was monitored in terms of c Pex = c p2 − c pD,F versus T thermograms in which the differences between the partial molar heat capacity of the measured oligonucleotide c p2 (raw signals corrected for the solvent contributions) and the partial molar heat capacities of the corresponding folded states extrapolated from low temperatures over the whole measured temperature interval, c pD,F , are normalized for the duplex or hairpin concentration. The total enthalpy of unfolding, ΔH cal (T) ,was obtained from the measured thermograms as the area under the c Pex(T) versus T curve.

Analysis of the DSC Thermograms.
The thermally induced conformational transitions can be experimentally followed in a model-independent way only by DSC. At relatively low concentrations used in DSC experiments the measured solute-normalized heat capacity of the sample solution, c P(T) , with the subtracted baseline may be equalized with the oligonucleotide partial molar heat capacity, c P2(T) . Thus, the overall heat effect that accompanies the measured conformational transition from its initial folded state at the temperature T 1 to its final unfolded state at T 2 can be expressed as Since the enthalpy is the state function, the enthalpy change ΔH (T1 → T2) may be expressed also as where (c p2(T) ) F and (c p2(T) ) U are the partial molar heat capacities of the folded and unfolded DNA conformation, respectively, ΔH (Tref) is the enthalpy of unfolding at T ref which can be any temperature between T 1 and T 2 . By choosing T ref = T 1/2 where T 1/2 is the melting temperature at which a half of oligonucleotide molecules undergo unfolding transition (2) transforms into  14) and (18)).
where ΔH cal T1/2 is a model-independent enthalpy of transition at T 1/2 that can be easily determined by the appropriate integration of the experimental [c p2(T) − (c p2(T) ) F ] and [c p2(T) − (c p2(T) ) U ] curves as presented in (3).
According to the DSC thermograms of the measured hairpin (HP), its thermally induced unfolding involves at least two conformational transitions ( Figure 2) can be expressed at given P and T as where K HPI and K IS are the corresponding equilibrium constants, the quantities in brackets are the equilibrium molar concentrations of HP, I, and S, n 1 is the number of moles of solvent, and n 2 is the number of moles of solute (oligonucleotide) that can be further expressed as: n HP in (6) (5) that Finally, by introducing α HP = 1 − α I − α S into (7) and taking the temperature derivative of the modified (7) one obtains the model function for the measured DSC signal, c P,ex , expressed as in which at any temperature c P,2 is the measured c P (with subtracted baseline), c P,HP is the partial molar heat capacity of HP extrapolated from low-temperature region over the entire measured temperature interval, ΔH c P,ex can be obtained experimentally simply by subtracting the hairpin c P,HP versus T curve extrapolated from low-T region over the entire measured temperature interval from the corresponding measured c P,2 versus T curve. The modelbased c P,ex , however, can be calculated from the right-hand side term of (8). According to the suggested model (4) For the description of the DSC experiment with the model function (8) one needs also the temperature derivatives of α S and α I . By using for each transition, i, the van't Hoff relation one obtains Assuming that for each transition, i, the corresponding Δc 0 Pi does not depend on T the standard free energy of that transition, ΔG 0 i(T) , can be obtained at any T from the integrated form of the Gibbs-Helmholtz relation as where T i,1/2 is the temperature at which the α i values sof species participating in transition i are the same. The corresponding equilibrium constant, and for the suggested mechanism of the hairpin unfolding (4) it can be easily seen that for each suggested monomolecular transition ΔG 0 i(Ti,1/2) = 0. Finally, according to the DSC experiments performed at different oligonucleotide concentrations the ΔH i(T) values appear to be concentration independent thus indicating that one may assume for each transition that ΔH i(T) = ΔH 0 i(T) and Δc Pi = Δc 0 Pi . Using these assumptions and (8)- (13) one can express the model function (14) c P,ex = α I Δc 0 p,HPI + dα I dT ΔH 0 HPI + α S Δc 0 p,HPI + Δc 0 only in terms of parameters T i,1/2 , ΔH 0 i(Ti,1/2) , and Δc 0 P,i , characteristic for each of the suggested transitions. Their "best fit" values are obtained by fitting the model function (14) to the experimental c P,ex versus T curves. Furthermore, since for each transition, i, the corresponding ΔH 0 i(T) and ΔS 0 i(T) quantities can be expressed as the "best fit" parameters T i,1/2 , ΔH 0 i(Ti,1/2) and Δc 0 P,i can be used also to obtain the ΔH 0 i(T) and ΔS 0 i(T) values at any T. In contrast to HP unfolding, the measured thermally induced duplex (D) to single strand (S) transition appears to be a simpler, all-or-none process that can be described in terms of the total oligonucleotide concentration, c T , the concentrations of the duplex form [D] and the single strands [S], the fraction of duplex molecules that undergo the unfolding transition at a given temperature, α S , and the equilibrium constant K DS interrelated as A similar, though much simpler derivation of the model function than the one presented for unfolding of the hairpin structure (14) leads for the suggested D ↔ 2S transition to where c P,2 is the measured c P of the sample solution with subtracted baseline, c P,D is the heat capacity of the duplex form extrapolated from the low-T region over the whole measured temperature interval, Δc 0 P,DS = Δc P,DS = 2c PS − c PD and ΔH 0 DS = ΔH DS = 2H S − H D . From the suggested model (16) and (17) it follows that ΔG 0 DS(T1/2) = −RT ln(2c T ) and The corresponding expressions for ΔG 0 DS(T) ΔH 0 DS(T) and TΔS 0 DS(T) are the same as those shown for each transition in the suggested hairpin unfolding mechanism (12), (13) and (15). Similarly, in deriving (18) the ΔH DS(T) and Δc P,DS are assumed to be independent on the oligonucleotide concentration and thus equal to ΔH 0 DS(T) , and Δc 0 P,DS . Inspection of (18) and (19) shows that the model function (18) is expressed in terms of adjustable parameters, T 1/2 , ΔH 0 DS(T1/2) and Δc 0 P,DS that can be determined by fitting the model function to the experimental c P,ex versus T curve and further used to determine the ΔG 0 DS(T) , ΔH 0 DS(T) and ΔS 0

DS(T) values at any T.
To obtain a set of the "best fit" adjustable parameters T i,1/2 , ΔH 0 i(Ti,1/2) and Δc 0 P,i describing the hairpin and duplex thermal unfolding at each of the added salt concentrations the iterative nonlinear Levenberg-Marquardt χ 2 regression procedure was used [31]. Furthermore, assuming that for the observed transitions the accompanying Δc 0 P,i quantities do not depend on the added NaCl concentration their values may be determined also from the slopes of the ΔH 0 i(T1/2) versus T i,1/2 curves constructed from the "best fit" ΔH 0 i(T1/2) and T i,1/2 parameters determined at different added salt concentrations [32]. These data can be also used to construct the corresponding T i,1/2 versus ln[Na + ] plots from which the amount of the Na + ions released upon thermal unfolding of the hairpin or duplex structure can be estimated (see discussion, (20)).

Results and Discussion
According to the measured DSC thermograms presented in Figure 2 the thermally induced unfolding of the hairpin HP consists of at least two conformational transitions while the one observed for the duplex D occurs in a simpler "all or none" manner. In addition, UV melting experiments (not shown) resulted for HP in biphasic melting curves that exhibit transitions independent on HP concentration (monomolecular transitions) while for D monophasic melting curves dependent on D concentration (nonmonomolecular transition) were observed. Moreover, excellent repeatability of the consecutive measured heating and cooling c P versus T curves and the observed independence of the measured DSC peaks on the applied heating rate (several test experiments) clearly shows that the studied thermal unfolding events may be considered as reversible processes. Model analysis of the measured thermograms shows that the hairpin thermal unfolding can be well described in terms of a three state model involving H (hairpin) ←→ I (intermediate state) ←→ S (unfolded single-stranded structure) transitions and the corresponding model function (14) characterized for each of the suggested transitions with the corresponding "best fit" adjustable parameters T i,1/2 , ΔH 0 i(Ti,1/2) , and Δc 0 P,i (Table 1). However, analysis of the applied fitting procedure indicates that the parameter Δc 0 P,IS is highly correlated to some other adjustable parameters. Thus, to obtain safe estimate of Δc 0 P,IS another method of its determination has to be used. Assuming that it does not depend on the simple salt concentration Δc 0 P,IS was estimated as a slope of the ΔH 0

IS(T1/2)
versus T IS,1/2 plot (Figure 3(a)) constructed from the "best fit" parameters determined at different NaCl concentrations (Table 1). This method of determining Δc 0 P,i was justified by a good agreement between the Δc 0 P,i values for other transitions obtained by the described fitting procedure and the Δc 0 Pi values determined as the slopes of the corresponding ΔH 0 i(Ti,1/2) versus T i,1/2 plots (Table 1). By using the parameters presented in Table 1 one can calculate for the duplex and hairpin solutions the relative populations of the modelpredicted species in the measured temperature interval and at all added salt concentrations ( Figure 2). Evidently, the thermal stability of the folded state of the measured duplex and the hairpin is substantially enhanced by increasing the added salt concentration. At low salt concentrations, however, a small fraction of the hairpin molecules undergoes transition into the intermediate state already at physiological temperatures.
A standard way of testing the quality of a suggested model is to compare the enthalpy of unfolding determined at a given temperature directly by an appropriate integration of the experimental c P,ex(T) versus T curve (ΔH cal HS , see (3)) with the corresponding model-based value ΔH 0 HS calculated at the same temperature using the reported "best fit" adjustable parameters. As shown in Table 1 a good agreement was obtained which clearly supports the appropriateness of the suggested H ←→ I ←→ S unfolding model.
It is well known that DNA unfolding is accompanied by release of counterions. The number of the released Na + ions, Table 1: Thermodynamic parameters a obtained from fitting the model functions ( (14) and (18)) to the duplex (D) and hairpin (HP) DSC thermograms presented in Figure 2.   Δn Na + ,i , upon each HP and D transition, expressed per mole of oligonucleotide, may be estimated from [33] dT i,1/2 d ln Na + = in which T i,1/2 is the melting temperature at a given Na + concentration, [Na + ], ΔH 0 i(Ti,1/2) is the corresponding enthalpy of transition at T i,1/2 . The Δn Na + ,i values presented in Table 1 were determined from the slopes of the ln[Na + ] versus T i,1/2 . plots (Figure 3(b)).
At any T in the measured range of physiological temperatures the difference between the given property characterizing the total unfolding of the hairpin H and the duplex D (for ex. ΔΔH 0 (T) = ΔH 0 HS(T) − ΔH 0 DS(T) ) reflects the contribution of the TCTC loop to that property relative to the core duplex ( Figure 4, Table 2). Thus, the observed ΔΔH 0 (T) > 0 indicates a favorable energy contribution of the TCTC loop to the stability of the hairpin that results, very likely, from the increased number of stacking interactions (in the first place from those occurring at the core duplexloop connections) [34]. The corresponding ΔΔS 0 (T) > 0 is consistent with the highly positive ΔΔH 0 (T) indicating that the unfavorable entropy contribution of the TCTC loop to the hairpin stability arises largely from a substantial disruption of the loop structure accompanying the unfolding of the hairpin. The observed ΔΔc 0 P > 0 and ΔΔn Na + > 0 show, however, that the loop contributions to ΔΔH 0 (T) and ΔΔS 0 (T) may be, to a certain extent, determined also by hydration [35] and electrostatic interactions. The ΔΔc 0 P > 0 suggests that within the folded hairpin conformation, not only the core duplex but also the TCTC loop are less exposed to water than in the unfolded state. In addition, the observed ΔΔn Na + > 0 may be ascribed to a decrease in the surface charge density accompanying the unfolding of the oligonucleotide which is significantly more pronounced in the case of hairpin unfolding. Comparison of the ΔΔ values for ΔG 0 T , ΔH 0 T , ΔS 0 T , Δc 0 P , and Δn Na + quantities determined for the HP → I and I → S transitions of the hairpin with the corresponding ΔΔ values determined for the D → 2S transition of the duplex shows that the ΔΔ values for the I → S and D → 2S transitions are very close (Figure 4, Table 2). Evidently, one may speculate that in the hairpin structure only the HP → I transition is influenced by the TCTC loop. In other words, it seems ±0.07 ±0.3 a Units: kcal mol −1 (ΔΔG 0 , ΔΔH 0 , TΔΔS 0 ), kcal mol −1 K −1 (ΔΔc 0 P ); for temperature dependence of ΔG 0 , ΔH 0 , and TΔS 0 see Figure 4.  that the observed HP → I transition reflects mainly the changes in the TCTC conformation. Thus, the intermediate state I may be considered as a state in which the core duplex remains more or less unchanged while the TCTC loop occurs as a more flexible structure characterized by the additional stacking interactions and the freedom of the neighboring water molecules and ions similar to the one in the unfolded state.
To the best of our knowledge this is the first time that a three-state unfolding of a simple hairpin structure, observed 8 Journal of Nucleic Acids by DSC, has been reported and characterized thermodynamically. We believe that the main reason for this is that in most studies of thermal unfolding of hairpins too high starting temperatures have been chosen and therefore the low-temperature transitions have been overlooked.