Base-pair mismatch can destabilize small DNA loops through cooperative kinking
BBase-pair mismatch can destabilize small DNA loops through cooperative kinking
Jiyoun Jeong and Harold D. Kim ∗ School of Physics, Georgia Institute of Technology,837 State Street, Atlanta, GA 30332-0430, USA (Dated: January 3, 2019)Base pair mismatch can relieve mechanical stress in highly strained DNA molecules, but how itaffects their kinetic stability is not known. Using single-molecule Fluorescence Resonance EnergyTransfer (FRET), we measured the lifetimes of tightly bent DNA loops with and without basepair mismatch. Surprisingly, for loops captured by stackable sticky ends, the mismatch decreasedthe loop lifetime despite reducing the overall bending stress, and the decrease was largest whenthe mismatch was placed at the DNA midpoint. These findings show that base pair mismatchtransfers bending stress to the opposite side of the loop through an allosteric mechanism known ascooperative kinking. Based on this mechanism, we present a three-state model that explains theapparent dichotomy between thermodynamic and kinetic stability of DNA loops.
Cellular DNA is constantly exposed to the possibil-ity of mispairing (i.e. non-complementary base pairing)[1]. Most commonly, mismatched base pairs result frombase misincorporation during gene replication [2] andheteroduplex formation between slightly different DNAsequences during homologous recombination [3]. Theycan also arise from exposure to DNA damaging agentsthat modify nucleobases [4, 5]. Due to less favorable basepairing and stacking [6], mismatched base pairs can in-crease local flexibility of double-stranded DNA [7–9], andconsequently the capture rate of tightly bent loops [10].For example, 1 to 3 bp-mismatch near the center of ashort DNA fragment ( <
150 bp) was shown to increasethe rate of DNA loop formation by one to two orders ofmagnitude [11, 12]. The kinetics of loop formation or cap-ture is intuitively understood by a one-dimensional freeenergy curve with the end-to-end distance as a singlereaction coordinate (Figure 1(a)). Base pair mismatchwould reduce the mechanical work required to bring twodistant DNA sites to proximity, more so for a shorterend-to-end distance. Therefore, the base pair mismatchwould lower the transition state relative to the unloopedstate (dotted line, Figure 1(a)).Base pair mismatch is also expected to affect the break-age or release rate of small DNA loops that are capturedby protein complexes [13] or by sticky ends of the DNAitself [14]. Looped DNA segments on the order of onepersistence length are subject to a high level of mechan-ical stress; therefore, the free energy of the looped stateis significantly lowered in the presence of the mismatch.According to the free energy diagram in Figure 1(a), thetransition state, being at a slightly longer end-to-end dis-tance by ∆x ‡ , would be lowered to a lesser degree (Fig-ure 1(a)). Therefore, the one-dimensional model predictsthat the rate of loop release would decrease in the pres-ence of base pair mismatch.Such prediction of mismatch-dependence seems plausi-ble considering the success of the model in predicting thelength dependence of loop capture and release rates. Inthe length regime where the free energy of loop formationis dominated by bending energy, increasing DNA lengtheffectively reduces the tilt in the free energy curve be- cause states at shorter end-to-end distances receive morestress relief, similar to the dotted line in Figure 1(a). Thischange predicts that loop capture and release rates mea-sured at different DNA lengths would be anti-correlated;loops associated with higher mechanical stress are cap-tured more slowly and released more quickly. This pre-diction has been confirmed for both DNA loops capturedby Lac repressor [15] and DNA loops captured by sticky (a)(b) End-to-end distance F r ee ene r g y No mismatchWith mismatch
Time (min) F R E T e ff i c i en cy ∆ t unloop High-salt buffer
Time (min) ∆ t loop Low-salt buffer ∆ x ‡ FIG. 1. (a)
One-dimensional free energy landscape for DNAloop capture and release. The two minimum free energy statescorrespond to the looped and unlooped states. The transi-tion state (vertical line) is separated from the looped stateby a small distance ∆x ‡ , which is equal to the capture ra-dius. The base pair mismatch is expected to increasingly un-tilt the solid curve toward shorter end-to-end distances, whichresults in the dotted curve. (b) Typical FRET trajectoriesof a DNA molecule undergoing loop capture (left) and looprelease (right). The DNA molecule labeled with Cy3 (green)and Cy5 (red) is in the low FRET state when unlooped, andin the high FRET state when looped. A sudden increase ordecrease in NaCl concentration at the 20-second time point(marked by a vertical dotted line) triggers the transition. a r X i v : . [ phy s i c s . b i o - ph ] J a n No mismatch0 20 40 60 80 100
Relative defect stiffness (%) B end i ng ang l e ( deg r ee s ) (c)(b) (f)(e) No defectDefect at 1 P r obab ili t y d i s t r i bu t i on ( N k i n k = ) (d) DefectOpposite to defect (a) un l oop ( s e c ) l oop ( s e c ) l oop ( s e c ) Center10bp20bp
Cy3Cy5BiotinSticky endsHairband loop
Center
FIG. 2. (a)
Schematic of a hairband loop captured by sticky ends. The schematic on top shows base-paired overhangs, Cy3(green circle), Cy5 (red circle), and the biotin linker (black circle). In this geometry, the overhangs on opposite strands forma duplex that can stack at both nicks of the loop. Different positions of base pair mismatch tested in our experiments aremarked on the linear form at the bottom. Only the bases on the overhangs are shown. (b)
Loop capture time of the hairbandmolecules (108 bp) as a function of the central mismatch size (circles). Data with an off-center 3-bp-mismatch are also shownas triangles. The upright and flipped triangles represent the loop capture times for base pair mismatches placed at 20 and 10bp away from the center of the molecule, respectively. Error bars, the standard errors of the mean, are smaller than the sizeof the symbols. (c)
Hairband loop lifetime (loop release time) as a function of the central mismatch size. Error bars representthe standard errors of the mean. (d)
Probability density of spontaneous kink positions along the coarse-grained minicircle (105bp) with (red) and without a pre-existing flexible defect (black), which is placed at position 1. (e)
Bending angle calculatedfrom the minimum-energy conformation of a DNA minicircle (105 bp) with a defect. Top and bottom figures show bendingangles at the defect and the site opposite to the defect, respectively, as a function of the defect stiffness relative to an intactbase pair. The minimum-energy conformations of the two extreme cases of the defect stiffness (0 and 100%) are also shownalong the curves with the defect position marked by X. (f )
Hairband loop lifetime as a function of the mismatch position (3-bpin size). For comparison, the horizontal dotted line shows the loop lifetime without the mismatch. Error bars represent thestandard errors of the mean. ends [16, 17]. While increasing DNA length evens out thebending stress over the entire DNA molecule, the basepair mismatch tends to localize sharp bending. There-fore, the effect of base pair mismatch might be quite dif-ferent from that of increasing DNA length.In this Letter, we investigated how base pair mismatchaffects the stability of small DNA loops. As a model sys-tem for DNA loop capture and release, we used shortdouble-stranded DNA molecules with sticky ends. Tomonitor loop capture and loop release events, we used thesingle-molecule FRET assay as previously published [14].Briefly, DNA molecules labeled with Cy3 and Cy5 neartheir sticky ends were immobilized to a NeutrAvidin-coated glass surface through a biotin linker, and loop cap-ture or release was triggered by exchange of buffers withdifferent NaCl concentrations (see Supplemental Mate-rial for more details). The first transition times in theFRET signals (∆ t ) of ∼
150 individual DNA moleculeswere collected. The mean of ∆ t spent in the unloopedstate before looping is defined as the loop capture time( τ unloop ), and the mean of ∆ t spent in the looped state before unlooping is defined as the loop release time orloop lifetime ( τ loop ). All DNA molecules used in thisstudy were shorter than 150-bp, the length regime wherethe free energy of loop formation is dominated by bend-ing energy.We first tried the loop capture geometry used in DNAcyclization, which we term as the “hairband loop” (Fig-ure 2(a)). In this geometry, the complementary over-hangs protrude from different strands so that the stickyends can anneal in trans and stack upon each other. Ina previous study, we showed that this end stacking, orequivalently nick closing, substantially increases the hair-band loop stability [18]. Using the single-molecule FRETassay, we measured the hairband loop capture times withand without base pair mismatch in the center. As shownin Figure 2(b), hairband loop capture took less time inthe presence of the mismatch as expected. The loop cap-ture time further decreased with increasing mismatch size(circles, Figure 2(b)). The base pair mismatch in the cen-ter position led to the largest decrease in the loop cap-ture time, and the decrease dropped as the mismatch wasplaced further from the center (triangles, Figure 2(b)).These observations confirm previous findings that mis-matched base pairs reduce the energy barrier for loopformation by increasing DNA bendability [8, 11, 19, 20],and this barrier reduction is most effective when the mis-match is in the center [21].Next, we measured the hairband loop release times orloop lifetimes ( τ loop ) with and without the mismatch inthe center. Since a mismatch could relieve the bendingstress of the hairband loop, we thought that the loop life-time would become longer. To our surprise, we observedthe exact opposite effect where the central mismatch de-creased the hairband loop lifetime (Figure 2(c)). Increas-ing the size of the mismatch from 1 bp to 3 bp led toa further decrease in the lifetime. This effect seemed toplateau past the mismatch size of 3 bp (Figure 2(c)). Thisresult suggests that the mismatch-containing hairbandloop is more kinetically unstable than the mismatch-freeloop, which seems paradoxical through the lens of theone-dimensional model presented in Figure 1(a).We thus considered the possibility that the transitionstate depends on other reaction coordinates besides theend-to-end distance, such as the closing angles at theloop junction. Since base stacking at the nick(s) in thehairband loop is a key determinant of decyclization ki-netics [18], we asked whether the central mismatch coulddestabilize the hairband loop by allosterically inducingnick opening. To investigate such allosteric coupling,we calculated the curvature profile of a kinkable semi-flexible loop [22] containing a defect with zero rigidityfrom a Monte Carlo simulation (see Supplemental Mate-rial for details). As shown in Figure 2(d), a kink witha sharp bending angle appeared most frequently at thefurthest end of the loop from the defect. We also calcu-lated the minimum energy conformation of a semiflexibleloop while varying the rigidity of the defect and foundthat the bending angles of furthest points were highlycorrelated (Figure 2(e)). This loop-mediated correlationof sharp bending angles between most distant sites istermed cooperative kinking [23], and has been observedin torsionally strained DNA minicircles by cryo-electronmicroscopy and molecular dynamics simulations [23–25].We hypothesized that the enhanced flexibility of thecentral mismatch destabilizes the hairband loop prevent-ing nicks(s) on the opposite side from closing. This hy-pothesis provides a few testable predictions. First, if themismatch were displaced from the midpoint of the DNA,the degree of destabilization would be dampened. Inagreement with this prediction, we observed a longer looplifetime when the mismatch was placed at a quarterpointinstead of the center (Figure 2(f)). Second, the cooper-ative kinking hypothesis requires nicks that can buckleunder the bending stress, and therefore the mismatch-induced destabilization would be eliminated in a loopcapture geometry free of end-stacking. We thus tested adifferent loop geometry referred to as the “hairpin loop”,where the complementary overhangs protrude from thesame strand (Figure 3(a)). In this geometry, the sticky ends anneal in cis and cannot stack upon each other. Us-ing these new DNA constructs with a central mismatchof various sizes, we repeated loop capture and release ex-periments. Similar to hairband loop capture, the hairpincapture time decreased with the size of base pair mis-match (Figure 3(b)). However, in sharp contrast to thehairband loop, the hairpin loop lifetime increased withmismatch size (Figure 3(c)). The effect of the base pairmismatch on the hairpin loop stability is therefore con-sistent with the prediction of the one-dimensional model.Overall, the lifetimes of hairpin loops were shorter thanthose of hairband loops, which is consistent with easierrupture of DNA duplex in an unzipping geometry thanin a shearing geometry [26–28]. These results lend strongsupport to the idea that cooperative kinking governs thekinetic stability of a mismatch-containing hairband loop.The mismatch-dependence of the hairband loop releasekinetics reveals the limitations of the one-dimensionaltwo-state model (Figure 1(a)) and invites us to consideradditional states and alternative reaction paths along an-other dimension. Here, we present two different paths( k (0) and k ( m ) ) that are likely to be the dominant onesfor mismatch-free and mismatch-containing DNA (Fig-ure 4(a)). Each path goes through three different states:unlooped, unstacked, and stacked. The loop capture rateis much greater in the presence of a central mismatch dueto its enhanced flexibility ( k ( m )1 (cid:29) k (0)1 ). The reverse rateis expected to be slower with the mismatch ( k (m)2 < k (0)2 )because of the weaker loop tension. Mismatch-free DNA (b)(c) un l oop ( s e c ) l oop ( s e c ) (a) Sticky endsHairpin loop
FIG. 3. (a)
Schematic of a hairpin loop. The schematic showsthe FRET pair (green and red circles), the biotin linker (blackcircle), and base-paired overhangs. In this geometry, the over-hangs on the same strand form a duplex like a zipper. (b)
Loop capture time of the hairpin (105 bp) molecules as a func-tion of the central mismatch size. Error bars are omitted dueto their small sizes. (c)
Hairpin loop lifetime as a function ofthe central mismatch size. Error bars represent the standarderrors of the mean. (a)(b) unloop (sec)01020304050 l oop ( s e c ) Pearson corr. = 0.74 P a t h State
No mismatchWith mismatch k k k k k k k k FIG. 4. (a)
The three-state model for hairband loop closureand release. The three states from left to right are unlooped,unstacked, and stacked states. The looped state is a mixedstate between the unstacked and stacked states. Therefore,the apparent loop capture rate ( k loop ) is equal to k , but theapparent loop release rate ( k unloop ) depends on k , k , and k . For the hairpin loop, k = 0, and therefore, k unloop isequal to k . Two representative paths for central mismatchsize 0 and m are highlighted with arc-like (top) and tweezers-like (bottom) motions, respectively. The vertical dotted linesimply the continuum of paths running parallel to the twoextreme ones shown. (b) Correlation between loop captureand release times of 16 unrelated hairband DNA molecules ofthe same size (94bp). The loop capture and release times weremeasured in equilibrium (i.e. no buffer-exchange) at slightlyelevated temperature of 34 ◦ C with [NaCl] = 700mM. undergoes small bending fluctuations uniformly through-out its contour, and therefore, follows an arc-like trajec-tory toward the looped state where end-stacking (nickclosing) and end-unstacking (nick opening) transitions may occur. In comparison, DNA with a mismatch inthe center can be sharply bent at a much lower energycost, and therefore, the most dominant path toward thelooped state will resemble a tweezers-like motion. As aresult of this motion, the sticky ends anneal at a sharpangle, and the hairband loop with the mismatch faces ahigher energy barrier for end-stacking (nick closing) thanwithout ( k ( m )3 (cid:28) k (0)3 ). The mismatch not only sup-presses end-stacking, but also promotes end-unstacking(nick opening) through cooperative kinking, which im-plies k ( m )4 (cid:29) k (0)4 . Hence, the apparent release rate ofthe hairband loop ( k unloop ) becomes faster with the mis-match than without because the looped state with themismatch is heavily biased towards the unstacked state.In comparison, for the hairpin loop that cannot proceedto the stacked state, the three-state model is reduced tothe two-state model, and the loop release rate is slowerwith the mismatch ( k (m)2 < k (0)2 ).The two paths boxed in Figure 4(a) represent the twomost extreme paths in terms of kinetics, the top path forthe slowest hairband loop capture and release, and thebottom for the fastest. In reality, there exists a contin-uum of paths going through the three states with inter-mediate rates, and the flexibility profile of DNA deter-mines the relative weights at which individual paths aretaken. Therefore, any changes to the flexibility profileof DNA would lead to correlated changes in the hair-band loop capture and release rates. To test this idea,we measured hairband loop capture and release times of16 unrelated sequences, all of the same length. Althoughlimited in sample size, we observed a significant degree ofcorrelation between the two times (Pearson correlation =0.74, Figure 4(b)). This result suggests that cooperativekinking is a general mechanism that governs the kineticsof hairband loop capture and release.In conclusion, we demonstrate that base pair mis-match can constrain the geometry and interactions forDNA loop capture through cooperative kinking, and theclose coupling between hairband loop geometry and end-stacking can give rise to correlated changes between loopcapture and release times (“easy come, easy go”). Wepropose a three-state model that correctly describes theeffect of mismatched base pairs on the apparent kineticsof loop capture and release. We expect the effect of mis-matched base pairs on protein-mediated DNA loops tobe more complex because of the diversity in loop capturegeometry [29]. Beyond passively captured DNA loops,it would be interesting to investigate whether base pairmismatches can also influence the kinetics of DNA loopextrusion [30, 31] through cooperative kinking. ∗ Corresponding author.Email: [email protected][1] N. Chatterjee and G. C. Walker, Environmental and Molecular Mutagenesis , 235 (2017).[2] T. A. Kunkel and D. A. Erie, Annual Review of Genetics , 291 (2015). [3] K.-C. Tham, R. Kanaar, and J. H. G. Lebbink, DNARepair , 75 (2016).[4] J. Cadet and J. R. Wagner, Cold Spring Harbor Perspec-tives in Biology , a012559 (2013).[5] A. Granzhan, N. Kotera, and M.-P. Teulade-Fichou,Chemical Society Reviews , 3630 (2014).[6] J. SantaLucia and D. Hicks, Annual Review of Biophysicsand Biomolecular Structure , 415 (2004).[7] G. Rossetti, P. D. Dans, I. Gomez-Pinto, I. Ivani, C. Gon-zalez, and M. Orozco, Nucleic Acids Research , 4309(2015).[8] M. Sharma, A. V. Predeus, S. Mukherjee, and M. Feig,The Journal of Physical Chemistry B , 6194 (2013).[9] S. Chakraborty, P. J. Steinbach, D. Paul, H. Mu,S. Broyde, J.-H. Min, and A. Ansari, Nucleic Acids Re-search , 1240 (2017).[10] A. Dittmore, S. Brahmachari, Y. Takagi, J. F. Marko,and K. C. Neuman, Physical Review Letters (2017).[11] J. D. Kahn, E. Yun, and D. M. Crothers, Nature ,163 (1994).[12] R. Vafabakhsh and T. Ha, Science , 1097 (2012).[13] C. Tardin, Biochimie , 80 (2017).[14] J. Jeong, T. T. Le, and H. D. Kim, Methods , 34(2016).[15] Y.-J. Chen, S. Johnson, P. Mulligan, A. J. Spakowitz,and R. Phillips, Proceedings of the National Academy ofSciences , 17396 (2014).[16] T. T. Le and H. D. Kim, Biophysical journal , 2068(2013).[17] T. T. Le and H. D. Kim, Nucleic Acids Research ,10786 (2014). [18] J. Jeong and H. D. Kim, bioRxiv (2018),https://doi.org/10.1101/503490.[19] K. A. Schallhorn, K. O. Freedman, J. M. Moore, J. Lin,and P. C. Ke, Applied Physics Letters , 033901 (2005).[20] C. Yuan, E. Rhoades, X. W. Lou, and L. A. Archer,Nucleic Acids Research , 4554 (2006).[21] P. Ranjith, P. S. Kumar, and G. I. Menon, Physicalreview letters , 138102 (2005).[22] A. Vologodskii and M. D. Frank-Kamenetskii, NucleicAcids Research , 6785 (2013).[23] T. A. Lionberger, D. Demurtas, G. Witz, J. Dorier,T. Lillian, E. Meyhfer, and A. Stasiak, Nucleic AcidsResearch , 9820 (2011).[24] R. N. Irobalieva, J. M. Fogg, D. J. Catanese Jr, T. Sut-thibutpong, M. Chen, A. K. Barker, S. J. Ludtke, S. A.Harris, M. F. Schmid, W. Chiu, et al. , Nature communi-cations , 8440 (2015).[25] T. Sutthibutpong, C. Matek, C. Benham, G. G. Slade,A. Noy, C. Laughton, J. P. K. Doye, A. A. Louis, andS. A. Harris, Nucleic acids research , 9121 (2016).[26] M. Mosayebi, A. A. Louis, J. P. K. Doye, and T. E.Ouldridge, ACS Nano , 11993 (2015).[27] J. Zhang, Y. Yan, S. Samai, and D. S. Ginger, TheJournal of Physical Chemistry B , 10706 (2016).[28] S. R. Tee and Z. Wang, ACS Omega , 292 (2018).[29] A. R. Haeusler, K. A. Goodson, T. D. Lillian, X. Wang,S. Goyal, N. C. Perkins, and J. D. Kahn, Nucleic acidsresearch , 4432 (2012).[30] M. Ganji, I. A. Shaltiel, S. Bisht, E. Kim, A. Kalichava,C. H. Haering, and C. Dekker, Science , 102 (2018).[31] J. F. Marko, P. De Los Rios, A. Barducci, and S. Gruber,bioRxiv (2018), https://doi.org/10.1101/325373. upplemental Material to “Base-pair mismatch can destabilize small DNA loopsthrough cooperative kinking” Jiyoun Jeong and Harold D. Kim ∗ School of Physics, Georgia Institute of Technology,837 State Street, Atlanta, GA 30332-0430, USA (Dated: January 3, 2019)
I. MATERIALS AND METHODSA. Preparation of DNA molecules
A 105-bp-long DNA molecule was extracted from yeastgenomic DNA by polymerase chain reaction (PCR) toserve as a control DNA template without any struc-tural defect. To probe the effect of a permanent de-fect, we planned to introduce a DNA mismatch to thecontrol molecule by mixing it with its mutant followedby a strand-exchange reaction. To do so, we addition-ally prepared a set of mutated DNA molecules that dif-fer from the control only in a certain location in whichwe put a mutation of size equal to 1bp, 3bp, or 5bp.To make such molecules, first, the mutated templatesof the control DNA were synthesized from Eurofins Ge-nomics (EXTREMer oligos) and duplexed via PCR. Eachof the duplexed products was then incorporated into apJET1.2 \ blunt vector (ThermoFisher) and cloned intoDH5 α Escherichia coli cells. Finally, the cloned frag-ments of DNA were extracted via colony PCR from thecells and were sequenced to ensure the correct mutationwas made at the desired location.To modify these molecules to carry a FRET pair (i.e.Cy3 and Cy5), biotin, and single-stranded sticky ends,we followed our standard preparation protocol [1], whichinvolves a series of PCR and strand exchange reactionsthat can be found in elsewhere. For introducing a DNAmismatch in the final construct, we mixed the Cy3-labledcontrol molecule with one of the Cy5-labeled mutatedmolecules with a ratio of 4:1 in the strand-exchange re-action.The final DNA construct generated by this protocolcarries a 5 protruding sticky end on each end and makesa hairband loop upon end-annealing as shown in Figure2(a) of the main text. We also made hairpin loops byhaving sticky ends on the same DNA strand (Figure 3(a)of the main text). A complete list of all DNA sequencescan be found in Tables S1 and S2 below. B. single-molecule FRET looing and unloopingassay
We followed our previous single-molecule FRET assaythat employs the sudden salt-exchange protocol [2, 3].For cyclization, DNA molecules were deposited on a pas-sivated surface of a flow-cell and were incubated at a lowsalt (10 mM [NaCl) imaging buffer containing the PCD- PCA oxygen scavenging system [4] for 10 minutes. Wethen injected a high salt (1 M [NaCl]) imaging bufferinto the flow-cell to promote sticky ends to capture theloop configuration. Decyclization measurements weredone similarly, except that the NaCl concentration waschanged from 2 M to 75 mM. The immobilized moleculeswere excited by a 532-nm laser continuously through anobjective-type TIR microscope from the beginning of thebuffer exchange. The time trajectories of FRET signals(Figure 1(b) of the main text) from the molecules wererecorded by an EMCCD camera (DU-897ECS0-
C. Minicircle simulations
The Monte Carlo simulation of a minicircle was im-plemented as previously described [2, 5]. A set of 105connected nodes was used to create a coarse-grained rep-resentation of a DNA minicircle of 105 bp. The bendingenergy at each node was described by the kinkable worm-like chain model [5] with the parameters of b = 0.3 andh = 12 following the same notation used in Ref. [6]. Weperformed the simulation with and without a flexible de-fect of zero bending energy placed at a fixed location. Forthe case of no flexible spot, we first initialized the sim-ulation without allowing the kink formation. Once thekink-free simulation was equilibrated, we allowed sponta-neous kinks to appear. To construct the probability den-sity of kink positions, we ran the simulation and stop atthe first appearance of a kink. We then recorded the po-sition of this kink and equilibrated back to the kink-freestate. This procedure was repeated until we collected adistribution of 1000 kink positions. The same procedurewas repeated in the presence of the hyperflexible spotto predict the effect of a flexible spot on the probabilitydistribution of kink. a r X i v : . [ phy s i c s . b i o - ph ] J a n Supplementary Table S1: DNA sequences of hairband molecules.No mismatch 5 − TGAATTTACG T GCCAGCAACAGA[T]AGCCGCGATCGCCATGGCAACGAGGTCGCACACGCCCCACACCCAGACCTCCCTGCGAGCGGGCATGGGTACAATCATTCGAGCTCGTTGTAG-3 − CACGGTCGTTGTCTATCGGCGCTAGCGGTACCGTTGCTCCAGCGTGTGCGGGGTGTGGGTCTGGAGGGACGCTCGCCCGTACCCATGTTAGTAAGCTCGAGCAACA T CACTTAAATG-5 − TGAATTTACG T GCCAGCAACAGA[T]AGCCGCGATCGCCATGGCAACGAGGTCGCACACGCCCCAGACCCAGACCTCCCTGCGAGCGGGCATGGGTACAATCATTCGAGCTCGTTGTAG-3 − CACGGTCGTTGTCTATCGGCGCTAGCGGTACCGTTGCTCCAGCGTGTGCGGGGTCTGGGTCTGGAGGGACGCTCGCCCGTACCCATGTTAGTAAGCTCGAGCAACA T CACTTAAATG-5 − TGAATTTACG T GCCAGCAACAGA[T]AGCCGCGATCGCCATGGCAACGAGGTCGCACACGCCCCGGGCCCAGACCTCCCTGCGAGCGGGCATGGGTACAATCATTCGAGCTCGTTGTAG-3 − CACGGTCGTTGTCTATCGGCGCTAGCGGTACCGTTGCTCCAGCGTGTGCGGGGCCCGGGTCTGGAGGGACGCTCGCCCGTACCCATGTTAGTAAGCTCGAGCAACA T CACTTAAATG-5 − TGAATTTACG T GCCAGCAACAGA[T]AGCCGCGATCGCCATGGCAACGAGGTCGCACACGCCCGCGCGCCAGACCTCCCTGCGAGCGGGCATGGGTACAATCATTCGAGCTCGTTGTAG-3 − CACGGTCGTTGTCTATCGGCGCTAGCGGTACCGTTGCTCCAGCGTGTGCGGGCGCGCGGTCTGGAGGGACGCTCGCCCGTACCCATGTTAGTAAGCTCGAGCAACA T CACTTAAATG-5 − TGAATTTACG T GCCAGCAACAGA[T]AGCCGCGATCGCCATGGCAACGAGGTCGTGGACGCCCCACACCCAGACCTCCCTGCGAGCGGGCATGGGTACAATCATTCGAGCTCGTTGTAG-3 − CACGGTCGTTGTCTATCGGCGCTAGCGGTACCGTTGCTCCAGCACCTGCGGGGTGTGGGTCTGGAGGGACGCTCGCCCGTACCCATGTTAGTAAGCTCGAGCAACA T CACTTAAATG-5 − TGAATTTACG T GCCAGCAACAGA[T]AGCCGCGATCGCCATGGCGGTGAGGTCGCACACGCCCCACACCCAGACCTCCCTGCGAGCGGGCATGGGTACAATCATTCGAGCTCGTTGTAG-3 − CACGGTCGTTGTCTATCGGCGCTAGCGGTACCGCCACTCCAGCGTGTGCGGGGTGTGGGTCTGGAGGGACGCTCGCCCGTACCCATGTTAGTAAGCTCGAGCAACA T CACTTAAATG-5 Supplementary Table S2: DNA sequences of hairpin molecules.No mismatch 5 − TGAATTTACG(CT)G T GCCAGCAACAGA[T]AGCCACATCGCCATGGCAACGAGGTCGCACACGCCCCACACCCAGACCTCCCTGCGAGCGGGCATGGGTTGCATGTCAGCTATGGATCCATTCGTAAATTCA-3 − CACGGTCGTTGTCTATCGGTGTAGCGGTACCGTTGCTCCAGCGTGTGCGGGGTGTGGGTCTGGAGGGACGCTCGCCCGTACCCAACGTACAGT(CG)ATACCTAGGT-5 [Cy3]1bp-mismatch(central) 5 − TGAATTTACG(CT)G T GCCAGCAACAGA[T]AGCCACATCGCCATGGCAACGAGGTCGCACACGCCCCAGACCCAGACCTCCCTGCGAGCGGGCATGGGTTGCATGTCAGCTATGGATCCATTCGTAAATTCA-3 − CACGGTCGTTGTCTATCGGTGTAGCGGTACCGTTGCTCCAGCGTGTGCGGGGTCTGGGTCTGGAGGGACGCTCGCCCGTACCCAACGTACAGT(CG)ATACCTAGGT-5 [Cy3]3bp-mismatch(central) 5 − TGAATTTACG(CT)G T GCCAGCAACAGA[T]AGCCACATCGCCATGGCAACGAGGTCGCACACGCCCCGGGCCCAGACCTCCCTGCGAGCGGGCATGGGTTGCATGTCAGCTATGGATCCATTCGTAAATTCA-3 − CACGGTCGTTGTCTATCGGTGTAGCGGTACCGTTGCTCCAGCGTGTGCGGGGCCCGGGTCTGGAGGGACGCTCGCCCGTACCCAACGTACAGT(CG)ATACCTAGGT-5 [Cy3] Continued on next page − TGAATTTACG(CT)G T GCCAGCAACAGA[T]AGCCACATCGCCATGGCAACGAGGTCGCACACGCCCGCGCGCCAGACCTCCCTGCGAGCGGGCATGGGTTGCATGTCAGCTATGGATCCATTCGTAAATTCA-3 − CACGGTCGTTGTCTATCGGTGTAGCGGTACCGTTGCTCCAGCGTGTGCGGGCGCGCGGTCTGGAGGGACGCTCGCCCGTACCCAACGTACAGT(CG)ATACCTAGGT-5 [Cy3] Both top (5 to 3 ) and bottom (3 to 5 ) sequences are shown. The underlined sequences represent sticky ends.A Cy5 fluorophore is internally attached at the thymine base colored in red. A Cy3 fluorophore is either at thegreen thymine base or the 5 end of the bottom strand. A biotin molecule is linked to the thymine base shownas [T]. Hairpin molecules includes a 2-nt gap (indicated by sequences in parentheses) near each end of the topstrand before sticky ends. ∗ Corresponding author.Email: [email protected][1] T. T. Le and H. D. Kim, Journal of Visualized Experiments (2014), 10.3791/51667.[2] T. T. Le and H. D. Kim, Nucleic Acids Research , 10786 (2014).[3] J. Jeong and H. D. Kim, bioRxiv (2018), https://doi.org/10.1101/503490.[4] C. E. Aitken, R. A. Marshall, and J. D. Puglisi, Biophysical Journal , 1826 (2008).[5] X. Zheng and A. Vologodskii, Biophysical journal , 1341 (2009).[6] A. Vologodskii and M. D. Frank-Kamenetskii, Nucleic Acids Research41