Differential Effects of Strand Asymmetry on the Energetics and

Nov 15, 2017 - Internal loops within structured nucleic acids disrupt local base stacking and destabilize neighboring helical domains; however, these ...
0 downloads 14 Views 2MB Size
Article Cite This: Biochemistry 2017, 56, 6448−6459

pubs.acs.org/biochemistry

Differential Effects of Strand Asymmetry on the Energetics and Structural Flexibility of DNA Internal Loops Thao Tran and Brian Cannon* Department of Physics, Loyola University Chicago, Chicago, Illinois 60660, United States S Supporting Information *

ABSTRACT: Internal loops within structured nucleic acids disrupt local base stacking and destabilize neighboring helical domains; however, these structural motifs also expand the conformational and functional capabilities of structured nucleic acids. Variations in the size, distribution of loop nucleotides on opposing strands (strand asymmetry), and sequence alter their biophysical properties. Here, the thermodynamics and structural flexibility of oligo-T-rich DNA internal loops were systematically investigated in terms of loop size and strand asymmetry. From optical melting experiments, a thermodynamic prediction model is proposed for the energetic penalty of internal loops that accounts for diminishing enthalpic and increasing entropic contributions due to loop size and strand asymmetry for bulges, asymmetric loops, and symmetric loops. These single-stranded domains become less sequence-dependent and more polymeric as the loop size increases. Single-molecule fluorescence resonance energy transfer studies reveal a gradual transition in conformation and structural flexibility from an elongated domain to an increasingly flexible bend that results from increasing strand asymmetry. The findings provide a framework for understanding the thermodynamic and conformational effects of internal loops for the rational design of functional DNA nanostructures.

U

metal ion coordination sites for DNAzymes.6−9 Although extensive work has been conducted on RNA internal loops to elucidate their sequence-dependent, thermodynamic parameters and improve nearest neighbor and other predictive models,29−39 DNA internal loops have not deeply studied, with energetic predictions primarily based on entropy-based extrapolations15,37 and the effect of strand asymmetry (i.e., relative distribution of unpaired nucleotides on opposing strands) on loop stability largely unknown. The presence of single-stranded regions within one or both strands of continuous helical domains alters the overall conformation and flexibility. Ensemble and single-molecule fluorescence studies show that the presence of single and tandem mismatches can enhance the flexibility and induce sharp bends.40−42 Larger, single-stranded domains confined to one strand (i.e., bulges) produce a bent conformation that acts as a flexible hinge in which the degree of bending and flexibility is sensitive to the size and sequence of the unpaired nucleotides in the bulge.2−5,43 In contrast, symmetric loops (same number of unpaired nucleotides on both strands) produce an elongated form relative to a fully paired duplex.44,45 Additional structural studies of RNA36,46,47 and DNA48 symmetric internal loops show that the overall helical structure within these domains is maintained by significant loop stabilization through noncanonical pairings and base stacking interactions, yet they

npaired nucleotides occur in DNA and RNA through multiple pathways, including inherent sequence, replication errors, base modifications, thermal excitation, and protein interactions. These unpaired regions give rise to secondary structure motifs that contain single-stranded (or loop) domains, such as single mismatches, tandem mismatches, bulges, internal loops, and hairpins. These loop motifs destabilize neighboring helical regions and can disrupt genetic processing;1 however, they also greatly expand the functional and structural capabilities of structured nucleic acids beyond that of duplexes, with roles in enhancing flexibility,2−5 binding metal ions,6−9 targeting by proteins,10 and forming tertiary contacts.11 These attributes have led to their incorporation as common elements in the rational design and synthesis of functional DNA molecules (i.e., aptamers and DNAzymes) and nanostructures.12−14 The structural traits of these motifs and the properties conferred by them are highly dependent on sequence and size; therefore, accurate nucleic acid structure prediction relies on the accumulation of sequence-specific data to calculate the energetic and topological effects of these motifs. Optical melting studies have been an important approach for quantifying the energetics of DNA structural motifs, particularly with regard to the development of nearest neighbor models.15 The thermodynamic parameters of DNA loop motifs have been reported for single mismatches,16,17 hairpins,18 bulges,19−22 and tandem mismatches.23,24 DNA internal loops remain largely unexamined, although they have significant use in synthesized functional DNA as toeholds to “fuel” DNA assembly,25−27 linkages in multijunction nanostructures,14 binding sites for the malaria biomarker Plasmodium lactate dehydrogenase,28 and © 2017 American Chemical Society

Received: September 18, 2017 Revised: October 31, 2017 Published: November 15, 2017 6448

DOI: 10.1021/acs.biochem.7b00930 Biochemistry 2017, 56, 6448−6459

Article

Biochemistry have larger end-to-end distances and greater flexibility relative to those of fully paired duplexes because of partial destabilization of local helical structure and base stacking.44 These conformational differences for bulge and symmetric loops highlight the structural implications of the contribution of strand asymmetry on loop entropy and enthalpy. Here, we systematically characterized the thermodynamic parameters of DNA-containing pyrimidine-rich, internal loops that vary in size and strand asymmetry by optical melting, and changes in conformation and flexibility that result from differences in strand asymmetry were evaluated by singlemolecule FRET (smFRET). All of the internal loops were oligo-T. This homopolymeric loop design was chosen to minimize enthalpic contributions from Watson−Crick base pairs that occur in random sequences49 and base stacking interactions that arise in poly(A) domains.50 In addition, all noncanonical contributions are of the same type. The optical melting experiments show that loop energy increases with loop size and is strongly affected by the distribution of the unpaired nucleotides on the opposing loop strands. For all loop sizes, the bulge has the largest energetic penalty, and the symmetric distribution has the lowest penalty, which arises from added stabilization due to noncanonical contributions; however, as the size of the loop increases, the energetic difference between the different symmetries decreases as the loop becomes less sequence-dependent and more polymeric. An empirical thermodynamic model is proposed for loop stability that accurately accounts for the effects of loop size and strand asymmetry. The results of the smFRET experiments show that conformation and flexibility depend on the asymmetry between the strands, with greater asymmetry markedly increasing the flexibility and bend angle. The combination of optical melting and single-molecule approaches advances the understanding of the contribution of different loop factors to the energetics and conformation of DNA internal loops.

12a, 5′-TGCTCTGCGTCACCGTACACGTCCAGTGC/ Cy3-3′, and 0 × 12b, 5′-Cy5/GCACTGTTTTTTTTTTTTGACGTGT-3′; 6 × 6a, 5′-TGCTCTGCGTCACCGTACACGTCTTTTTTCAGTGC/Cy3-3′, and 6 × 6b, 5′-Cy5/GCACTGTTTTTTGACGTGT-3′. A tether strand (T) for immobilization was complementary to the a strand. For the bending experiments, the T strand was 5′-labeled with Cy5 and 3′-biotinylated with the sequence: 5′Cy5/CGGTGACACGGAGCACTG-biotin. The number of nucleotides separating the dyes (16) remained constant for all of the smFRET constructs. The tether for the fraying experiments was 3′-biotinylated with the sequence 5′-CGGTGACGCAGAGCA-3′. All constructs were prepared by annealing the appropriate a, b, and T strands at a ratio of 5:5:1, with a T strand concentration of 2 μM. Figure S2 shows the alignment of the strands for the annealed constructs used in the smFRET experiments. The strands were heated to 94 °C for 2 min and then underwent slow, light-protected cooling to room temperature. The annealing conditions included 25 mM Tris-HCl (pH 8.0) and 50 mM NaCl. The annealed DNAs were kept on ice until they were measured. Optical Melting Experiments. For the melting experiments, absorbance measurements (260 nm) versus temperature were taken with a Thermofisher Evolution 260 spectrophotometer equipped with an eight-cell Peltier heating unit. The heating rate was 0.5 °C/min across the temperature range of 15−85 °C. The DNA constructs were measured across a 200fold total DNA concentration range from 0.5 to 100 μM (8−10 concentration values per sample) by serial dilution and varying quartz cuvette lengths. The melting buffer conditions included 20 mM sodium cacodylate, 0.5 mM EDTA, and 1.0 M NaCl (pH 7.0). The buffer solution was degassed prior to measurement. All samples were measured three times. Melting data were analyzed through two conventional approaches. In the first method, the melting curves were fit by a multiparametric, two-state Ising model:5



MATERIALS AND METHODS Materials. All constructs for the optical melting experiments were based on the sequence 5′-CTCGTCTN1CAGTGC-3′ and 5′-GCACTGTN2GACGAG-3′, where N1 ≥ 0 and N2 ≥ 0. All of the loops were centrally positioned. The sequences were designed to melt in the range of 40−60 °C. All oligonucleotides were purchased from IDT (Coralville, IA) with polyacrylamide gel electrophoresis purification, with purity confirmed by 3% agarose electrophoresis with SYBR Green I staining. Sample concentrations were determined from 260 nm absorbance values at 90 °C and single-stranded extinction coefficients. All samples were annealed at an equimolar ratio for the two complementary strands by being heated to 94 °C for 2 min and then slowly cooled to room temperature. The sequences for the bending and fraying smFRET experiments were as follows: 0 × 0a, 5′-CAGTGCTCCGTGTCACCGTCACTCGTCCAGTGCA-3′, and 0 × 0b, 5′-Cy3/ TGCACTGGACGAGTG-3′; 0 × 6a, 5′-CAGTGCTCCGTGTCACCGTCACTCGTCCAGTGCA-3′, and 0 × 6b, 5′-Cy3/ TGCACTGTTTTTTGACGAGTG-3′; 1 × 5a, 5′-CAGTGCTCCGTGTCACCGTACTCGTCTCAGTGCA-3′, and 1 × 5b, 5′-Cy3/TGCACTGTTTTTGACGAGT-3′; 2 × 4a, 5′CAGTGCTCCGTGTCACCGTACTCGTCTTCAGTGC-3′, and 2 × 4b, 5′-Cy3/GCACTGTTTTGACGAGT-3′; 3 × 3a, 5′-CAGTGCTCCGTGTCACCGTCTCGTCTTTCAGTGCCy3-3′, and 3 × 3b, 5′-Cy3/GCACTGTTTGACGAG-3′; 0 ×

A total (T ) = αA ss(T ) + (1 − α)Ads(T )

(1)

where α is the single-stranded fraction, Ass is the single-stranded absorbance, and Ads is the double-stranded absorbance. This model includes independent baseline corrections to Ass and Ads, and it assumes temperature-independent ΔH° and ΔS° values. Marquardt−Levenberg optimization (Origin Pro, Northampton, MA) was used to determine the best fits to the melting temperature (TM), enthalpy (ΔH°), and entropy (ΔS°). In the second approach, the thermodynamic parameters were determined from the inverse melting temperature (TM−1) and total strand concentration (Ct) by van’t Hoff analysis for noncomplementary duplexes according to51 TM −1 = (R /ΔH °) ln(C t /4) + ΔS°/ΔH °

(2)

The free energy change at 37 °C was calculated as ° ΔG37 = ΔH ° − 310.15ΔS°

(3)

The ΔH°, ΔS°, and ΔG° values extracted from the two analyses had average differences of 0) and no thymines were located on strand 1 (N1 = 0). Increasing Loop Size and Strand Asymmetry Lower the Melting Transition Temperature and Decrease the Overall Stability. The melting behavior and thermodynamic

energy for DNA internal loops, oligo-T domains were positioned centrally within the duplex and were flanked by 6 bp helical domains. The loops were terminated by C-G base pairs at each end to stabilize the helical domains. The internal loops were varied in terms of (1) loop size, the total number of unpaired thymines in the loop (Ntotal = N1 + N2, where N1 and N2 refer to the numbers of central thymines within each strand), and (2) strand asymmetry, the difference in the number of unpaired thymines on the two strands (ΔN = |N1 − N2|). The loop sizes ranged from 1 to 24 nucleotides. The internal loops were classified into three categories based on 6451

DOI: 10.1021/acs.biochem.7b00930 Biochemistry 2017, 56, 6448−6459

Article

Biochemistry

Figure 3. Variation in the measured ΔG37 ° (loop) values by differences in loop size (Ntotal) and strand asymmetry (ΔN = |N1 − N2|). (a) ΔG37 ° (loop) values for symmetric loops (○), asymmetric loops (▲), and bulges (□) vs loop size for Ntotal = 1−16. (b) Change in the magnitude of the loop penalty with strand asymmetry within each Ntotal as a function of loop size. The decrease in variation with loop size follows a power-law dependence, best fit by the equation y = 4.17/Ntotal1.42 (R2 = 0.91; black line).

measured ΔG37 ° of 15.32 kcal/mol, in excellent agreement with nearest neighbor calculation (15.2 kcal/mol) using the DINAmelt software.54,55 The free energies for the symmetric constructs decrease from 13.38 kcal/mol for the 1 × 1 loop to 7.79 kcal/mol for the 12 × 12 loop. The bulge constructs have ΔG37 ° values ranging from 12.93 kcal/mol for the onenucleotide bulge construct to 7.55 kcal/mol for the 12nucleotide bulge construct. Strikingly, the 12-nucleotide bulge causes a destabilization of the construct that is approximately equivalent to that caused by the 12 × 12 construct, which has 24 unpaired nucleotides distributed equally on the two strands. In terms of the gradual effect of strand asymmetry with Ntotal = 6 as a representative example, the ΔG37 ° values for the 3 × 3, 2 × 4, 1 × 5, and 0 × 6 constructs are 10.95, 10.52, 9.67, and 8.52 kcal/mol, respectively. Increased asymmetry produces greater destabilization with an approximately 2.2 kcal/mol difference between the 3 × 3 and 0 × 6 distributions. The enthalpy and entropy components show that the decreases in free energy with increasing strand asymmetry occur with a concomitant decline in enthalpy and a rise in entropy. The higher-enthalpy component for the symmetric loops likely reflects the formation of noncanonical T-T interactions that occur through two hydrogen bonds. The added stabilization from these interactions constrains the loop, limiting loop entropy. Calculation of Loop Free Energies, ΔG37 ° (loop). To determine the effect of the internal loop on the total free energy of the constructs, the loop energies were calculated for all of the measured loop sizes and asymmetries. For the single-nucleotide bulge, the loop energy was calculated with the assumption that the flipped-out conformation of the bulge nucleotide does not interrupt base stacking in the duplex, so the following expression, which has been applied to RNA19 and DNA,15 was used:

parameters for DNA duplexes containing T-rich internal loops were systematically investigated by ultraviolet (UV) melting experiments for loops of varying size and asymmetry in 1 M NaCl. Figure 2a shows normalized melting curves acquired by monitoring changes in UV absorbance (260 nm) at increasing temperatures for DNA constructs with symmetric internal loops of size Ntotal = 0−16. The melting temperature (TM) decreases with an increase in loop size as shown in the TM−1 versus ln(Ct/4) plots (Figure 2b). The linear characteristic of the TM−1 versus ln(Ct/4) plots for all of the studied DNA constructs and the agreement of the extracted thermodynamic parameters (ΔH°, ΔS°, and ΔG°) by van’t Hoff analysis (eq 2) with curve-fit analysis support the two-state melting transition model for DNA containing these loops. The measured thermodynamic parameters (TM, ΔH°, ΔS°, and ΔG°) are summarized for symmetric internal loops (Table 1), bulges (Table 2), and asymmetric internal loops (Table 3). Overall, the melting temperature decreases with loop size for all strand asymmetries, and for any given loop size, greater strand asymmetry corresponds to a lower melting temperature, with bulges having the lowest TM and symmetric loops having the highest TM. The reference duplex (0 × 0) has a measured TM of 66.0 °C, and the TM decreases with an increase in symmetric loop size, from 59.7 °C for the 1 × 1 loop to 39.6 °C for the 12 × 12 loop. The concentration dependence of the inverse TM generally decreases with an increase in loop size up to 3 × 3, reflecting added stabilization between noncanonical T−T interactions. For larger loop sizes, the dependence increases, indicating the diminishing contribution of the noncanonical interactions and the increasing level of destabilization with loop size. The difference in concentration dependence between the 0 × 0 and symmetric loop constructs may also partly arise from differences in the sizes of the constructs.53 The measured TM values for the bulge constructs range from 59.2 °C for the onenucleotide bulge to 41.8 °C for the 12-nucleotide bulge construct. In terms of the effects of strand asymmetry with Ntotal = 6 as an example, the melting temperatures decrease in the order of the strand asymmetry: 3 × 3 (ΔN = 0), 2 × 4 (ΔN = 2), 1 × 5 (ΔN = 4), and 0 × 6 (ΔN = 6) with TM values of 52.1, 51.7, 49.3, and 46.2 °C, respectively. Similarly, the overall free energies (ΔG°37) for the loop constructs decrease as the loop size increases, and the loop becomes more asymmetric. The reference duplex (0 × 0) has a

° ΔG37 (1‐nucleotide bulge) ° = ΔG37 (duplex with 1‐nucleotide bulge)

(5)

This equation yields a ΔG37 ° (loop) value of 2.4 kcal/mol for a single-thymine bulge within GG/CC stacking, and this result agrees with the reported value of 2.6 kcal/mol.21 The change in free energy occurs with no significant change in enthalpy, consistent with a single-nucleotide bulge not disrupting helical stacking of the intervening GG/CC base pairs. For all other 6452

DOI: 10.1021/acs.biochem.7b00930 Biochemistry 2017, 56, 6448−6459

Article

Biochemistry

Figure 4. (a) Comparison of the measured and predicted ΔG37 ° (loop) values. The predicted values were calculated from the empirical model (eq 7). The solid line shows a y = x line for reference. The average deviation between the predicted and measured values was 0.22 kcal/mol (R2 = 0.98). (b) ° (loop) values from the empirical model as a function of loop asymmetry for varying loop sizes (Ntotal = 4−50). The Ntotal values are Predicted ΔG37 shown in parentheses.

loops (Ntotal > 1), the loop energy, ΔG37 ° (loop), was calculated according to15

with loop size. This smaller change in entropy likely corresponds to the symmetric loops having the most extensive noncanonical TT interactions for any loop configuration. Second, for each Ntotal value, the loop energy increased correspondingly with strand asymmetry. As noted above, for every Ntotal value, the bulge exhibited the largest loop energy, and the symmetric loop had the smallest loop energy. Importantly, the energetic difference between the symmetric and bulge distributions decreased with an increase in loop size. For instance, the two-nucleotide loops, 0 × 2 (bulge) and 1 × 1 (symmetric), differ by 3.17 kcal/mol, compared to a 1.51 kcal/ mol difference between the 0 × 12 and 6 × 6 loops. Closer examination of the change in loop energy with strand asymmetry revealed that for each Ntotal value, the loop energy increases monotonically with ΔN. A plot of the linear fits to the change in ΔG 37 ° (loop) with strand asymmetry (ΔN) ⎡ ΔΔG37° (loop) ⎤ ⎣⎢ ΔN ⎦⎥ as a function of Ntotal shows that the energetic difference arising from strand asymmetry decreases with loop size following a power-law dependence with an exponent of −1.42 (Figure 3b). The stability of hairpins also shows a powerlaw dependence on loop size.18 On the basis of the entropydominated behavior of bulges, the results here suggest that the enthalpic contribution to loop stability from noncanonical interactions within the loop diminishes with loop size and that the behavior of these single-stranded domains becomes more polymeric and less sequence-dependent. The Proposed Thermodynamic Model Accurately Predicts Loop Energies. On the basis of our findings that (1) the loop energy of bulges increases in an entropically defined manner with loop size, (2) energetic differences for any loop size change linearly with ΔN, and (3) the effects of strand asymmetry decrease with loop size, we propose the following model to describe the loop energies for DNA internal loops:

° ° ΔG37 (loop) = ΔG37 (duplex with loop) ° − ΔG37 (ref. duplex) ° − ΔG37 (interrupted NN)

(6)

where ΔG37 ° (interrupted NN) = 1.84 kcal/mol for the disrupted GG/CC stacking due to the internal loop.15 The TT mismatch had the weakest energy penalty of 0.1 kcal/mol, consistent with previous measurements for a single TT mismatch within the GTG/CTC helical context, which had a reported value of 0.3 kcal/mol.16 This result supports the view that the effects of a TT mismatch do not propagate along the helix and weakly perturb the stability of flanking helices.17,21 Figure 3 summarizes the measured ΔG°37 values for loop sizes of Ntotal = 1−16 and for different asymmetries using eqs 5 and 6. This plot contains several important features with regard to the effects of loop parameters on the energetics. First, the loop energy increases with Ntotal for all strand asymmetries, with the bulge and symmetric loops appearing as the upper and lower boundaries for all sizes. The energetic penalties for DNA and RNA internal loops have been proposed as entropic extrapolations;15,30,38 therefore, the bulge and symmetric internal loops were fit using the Jacobson−Stockmayer model. The bulge energies were fit according to the equation ΔG37 ° (Ntotal) = cRT ln(Ntotal/1) + ΔG37 ° (Ntotal = 1), where ΔG°37(Ntotal = 1) = 2.4 kcal/mol is the experimental value measured for the single-nucleotide bulge. This model yielded a best-fit coefficient c of 2.19 ± 0.08 (R2 = 0.93), which gives excellent agreement with values obtained from kinetic measurements (c = 2.44) for poly(T) loops50 and entropic models (c = 2.10−2.20) of DNA melting bubbles,56−58 suggesting that the loop energy of the bulge conformation is primarily entropic. This value is greater than the theoretical value of 1.75 for an isolated loop, which is reasonable because internal loops have two closure points, which reduces the total number of possible conformations. Similar treatment of the symmetric loops by using the equation ΔG37 ° (Ntotal) = ΔG37 ° (Ntotal = 2) + cRT ln(Ntotal/2), where ΔG°37(Ntotal = 2) = 0.1 kcal/mol is the measured 1 × 1 loop energy, yields a best-fit coefficient c of 3.73 ± 0.04 (R2 = 0.99). The larger coefficient represents a smaller entropy gain (e.g., smaller number of conformations)

° ° ΔG37 (loop) = ΔG37 (initiation) + cRT ln(Ntotal) ° − ΔG37 (symmetry)

(7)

⎛ N − |ΔN | ⎞ ° ΔG37 (symmetry) = D⎜ total α ⎟ Ntotal ⎝ ⎠

where ΔG37 ° (initiation) = 2.19 kcal/mol represents the measured loop energy for a single extruded nucleotide 6453

DOI: 10.1021/acs.biochem.7b00930 Biochemistry 2017, 56, 6448−6459

Article

Biochemistry

Figure 5. Cooperativity of the melting transition affected by strand asymmetry. (a) Differentiated melting curves for DNA (10 μM) with an internal loop size Ntotal = 6 for different loop distributions: 3 × 3 (black), 2 × 4 (red), 1 × 5 (purple), and 0 × 6 (green). The solid lines are Gaussian fits to each curve. (b) Plot of full width at half-maximum (FWHM) values determined from the Gaussian fits to the differentiated melting curves in panel a vs strand asymmetry (ΔN = |N1 − N2|). The differential curves were five-point smoothed.

ΔG37 ° (Ntotal = 1). The second term accounts for the entropic contribution from the loop size. The symmetry term acts as a stabilizing penalty and accounts for the maximum number of potential noncanonical interactions for any given structure. The contribution from this term decreases as the loop size increases. This symmetry term vanishes for bulges because Ntotal = ΔN to reflect the fact that bulge energy increases entropically with loop size. The coefficient D represents the maximum stabilizing effect due to strand−strand interactions. Nonlinear fitting of the model to the measured loop energies yielded the following parameter values: c = 2.17, D = 4.32 kcal/mol, and α = 1.5. A comparison of the measured and predicted ΔG°37(loop) values shows excellent agreement (Figure 4a). The average deviation between the measured and predicted values is 0.22 kcal/mol. According to eq 7, Figure 4b shows how strand asymmetry modulates the loop energies for different loop sizes. The energetic difference between the different distributions decreases, becoming less than 1 kcal/mol for loop sizes with >20 nucleotides. Any loop beyond this length essentially behaves as an entropic loop with two points of closure and is independent of the distribution of unpaired nucleotides on the two loop strands. As a test of the general applicability of this model, examination of a thermodynamic study of DNA hairpins containing internal loops sizes of 1 × 1, 3 × 3, and 5 × 5 shows an increasing level of destabilization with loop size,59,60 similar to the observations in the current study for duplexes. The measured hairpin destabilization with loop size increases in a manner consistent with our model with an average fractional difference of 0.12 between the measured and predicted values from eq 7. In terms of limitations for the model, nearest neighbor calculations have demonstrated that sequence-specific considerations are critical for accurate prediction of free energy values.15 For small internal loops, the free energy contribution of the loop will likely depend on the local sequence as the free energy for single-nucleotide bulges and mismatches can vary greatly according to the neighboring sequence. The initiation term ΔG°37(initiation) may require adjustment based on local sequence considerations for the initial extrusion; however, the increments in free energy for longer loops should increase with loop size according to the proposed model.

This work supports observations from RNA studies that identified differences in loop energies for symmetric and asymmetric loops with asymmetric loops as more energetically penalizing.37,39,61 Prediction models for RNA internal loops have extrapolated loop energy through use of the Jacobson− Stockmayer approximation with an entropy coefficient smaller than that measured for DNA.30,38 Small- and medium-sized RNA loops have been shown to deviate from predictions on the basis of Jacobson−Stockmayer approximations due to complex sequence-dependent effects from extensive GA and UU interactions and formation of stabilized kink turns that may dominate the entropic component of the loops.33,36,62 As the size of the loops increases, such sequence-dependent effects may decrease similarly for RNA and DNA loops. Future experiments with more complex loop sequences for RNA and DNA will be important in examining the size-based transition of loop behavior from enthalpy- to entropy-based. The Distribution of Unpaired Nucleotides Affects Melting Cooperativity. The results presented above show the degree to which helical destabilization by internal loops is affected by strand asymmetry. To gain further insight into additional helical properties influenced by strand asymmetry, the cooperativity of the melting transition was examined. Helical destabilization often corresponds to a loss of cooperativity, which can be observed by an increase in the half-width of the differentiated melting curve;63 therefore, the full width at half-maximum (FWHM) values from the differentiated melting curves were determined to provide a qualitative measure of melting cooperativity. Figure 5 shows the differentiated melting curves and a plot of the fwhm values for the four Ntotal = 6 constructs (0 × 6, 1 × 5, 2 × 4, and 3 × 3). The FWHM values increase monotonically with strand asymmetry, indicating that cooperativity decreases as the loop becomes more asymmetric. This loss of cooperativity likely corresponds to an increased level of intermediate states in the melting transition64 such that a larger number of intermediate states may be populated by the more asymmetric states because there are fewer noncanonical interactions that decrease the conformational entropy. Asymmetric anucleosidic linkages also reduce melting cooperativity.65 These findings together support a loop entropy mechanism for loss of cooperativity between neighboring helices, such that the greater amount of entropy in 6454

DOI: 10.1021/acs.biochem.7b00930 Biochemistry 2017, 56, 6448−6459

Article

Biochemistry

Figure 6. Single-molecule FRET histograms for DNA constructs with Ntotal = 6 in 1 M NaCl. Representative time traces and histograms from individual molecules for (a) the 0 × 0 duplex and for the four strand asymmetry configurations: (b) 0 × 6, (c) 1 × 5, (d) 2 × 4, and (e) 3 × 3. Heat maps showing the distributions of the standard deviation vs the average FRET value for all analyzed molecules: (f) Nmol = 445 for 0 × 0, (g) Nmol = 488 for 0 × 6, (h) Nmol = 529 for 1 × 5, (i) Nmol = 592 for 2 × 4, and (j) Nmol = 668 for 3 × 3. The images were smoothed by nearest neighbor averaging.

6 and varying strand asymmetries. The Ntotal = 6 series was selected because this size had a large energetic difference between the bulge and symmetric forms with distinct entropic and enthalpic contributions. This size then allowed any variations in flexibility or conformation for the different constructs to be evaluated in terms of these components. DNA constructs with the four possible Ntotal = 6 strand asymmetries [0 × 6, 1 × 5, 2 × 4, and 3 × 3 (see Figure 1)] were studied by smFRET. For all constructs, the Cy3 and Cy5 dyes were located 16 nucleotides apart on opposing strands with the loop regions flanked by the same 6 bp sequences as the melting constructs. A loop-free, fully formed duplex served as the control. The presence of co-solutes such as polyethylene glycol and salt concentration can produce contrasting effects on the stability and conformation of the nucleic acid due to changes in local ion concentration and volume exclusion effects.66−70 Single-molecule experiments require substantial

the separating domain and not its physical size leads to greater decoupling between the loop-linked helical domains. This loss of cooperativity coupled with the energetic penalty of the loops likely gives a minimum size limit of approximately 5 bp per helical domain for these structures to remain fully intact when linked by extensive loop sizes under the given experimental conditions. Flexibility of Loops Sensitive to Strand Asymmetry. The optical melting experiments show how the size and asymmetry of internal loops influence the enthalpic and entropic components of the free energy for internal loops. To examine the modulation of the conformation and dynamics by internal loop asymmetry, single-molecule FRET experiments (smFRET) were performed with immobilized, loop-containing duplexes labeled with the FRET pair, Cy3 and Cy5. Specifically, the smFRET experiments examined changes in the interdye distances for DNA duplexes with an internal loop size of Ntotal = 6455

DOI: 10.1021/acs.biochem.7b00930 Biochemistry 2017, 56, 6448−6459

Article

Biochemistry levels of glucose as part of the oxygen-scavenging systems that prolong the observation time for the fluorescent dyes; therefore, optical melting experiments were conducted with the 0 × 0 duplex as a control for glucose-mediated effects because prolonged exposure of DNA to glucose can lead to strand breakage and depurination due to glycation reactions between glucose and glycosidic bonds.71,72 The experiments reveal that glucose does destabilize the duplex with decreases in ° of 0.83 kcal/mol (Figure S2 and TM of 0.7 °C and ΔΔG37 Table S1); however, these effects likely do not significantly interfere with the integrity of the DNA constructs in the smFRET experiments because the DNA does not have prolonged exposure to glucose (