Base-pair mismatch can destabilize small DNA loops through cooperative kinking

TL;DR

This study reveals that base pair mismatches in small DNA loops can destabilize their kinetic stability by promoting cooperative kinking, which transfers bending stress and affects loop lifetime.

Contribution

It introduces a three-state model explaining how base pair mismatches influence DNA loop stability through cooperative kinking mechanisms.

Findings

Mismatches decrease DNA loop lifetime despite reducing bending stress.

Mismatches at the loop midpoint cause the largest decrease in stability.

A three-state model explains the thermodynamic and kinetic stability dichotomy.

Abstract

Base pair mismatch can relieve mechanical stress in highly strained DNA molecules, but how it affects their kinetic stability is not known. Using single-molecule Fluorescence Resonance Energy Transfer (FRET), we measured the lifetimes of tightly bent DNA loops with and without base pair mismatch. Surprisingly, for loops captured by stackable sticky ends, the mismatch decreased the loop lifetime despite reducing the overall bending stress, and the decrease was largest when the mismatch was placed at the DNA midpoint. These findings show that base pair mismatch transfers bending stress to the opposite side of the loop through an allosteric mechanism known as cooperative kinking. Based on this mechanism, we present a three-state model that explains the apparent dichotomy between thermodynamic and kinetic stability of DNA loops.

Full text

††preprint: APS/123-QED

Cellular DNA is constantly exposed to the possibility of mispairing (i.e. non-complementary base pairing) Chatterjee and Walker (2017). Most commonly, mismatched base pairs result from base misincorporation during gene replication Kunkel and Erie (2015) and heteroduplex formation between slightly different DNA sequences during homologous recombination Tham et al. (2016). They can also arise from exposure to DNA damaging agents that modify nucleobases Cadet and Wagner (2013); Granzhan et al. (2014). Due to less favorable base pairing and stacking SantaLucia and Hicks (2004), mismatched base pairs can increase local flexibility of double-stranded DNA Rossetti et al. (2015); Sharma et al. (2013); Chakraborty et al. (2017), and consequently the capture rate of tightly bent loops Dittmore et al. (2017). For example, 1 to 3 bp-mismatch near the center of a short DNA fragment ($<$150 bp) was shown to increase the rate of DNA loop formation by one to two orders of magnitude Kahn et al. (1994); Vafabakhsh and Ha (2012). The kinetics of loop formation or capture is intuitively understood by a one-dimensional free energy curve with the end-to-end distance as a single reaction coordinate (Figure 1(a)). Base pair mismatch would reduce the mechanical work required to bring two distant DNA sites to proximity, more so for a shorter end-to-end distance. Therefore, the base pair mismatch would lower the transition state relative to the unlooped state (dotted line, Figure 1(a)).

Base pair mismatch is also expected to affect the breakage or release rate of small DNA loops that are captured by protein complexes Tardin (2017) or by sticky ends of the DNA itself Jeong et al. (2016). Looped DNA segments on the order of one persistence length are subject to a high level of mechanical stress; therefore, the free energy of the looped state is significantly lowered in the presence of the mismatch. According to the free energy diagram in Figure 1(a), the transition state, being at a slightly longer end-to-end distance by $\Delta\textrm{x}^{\ddagger}$, would be lowered to a lesser degree (Figure 1(a)). Therefore, the one-dimensional model predicts that the rate of loop release would decrease in the presence of base pair mismatch.

Such prediction of mismatch-dependence seems plausible considering the success of the model in predicting the length dependence of loop capture and release rates. In the length regime where the free energy of loop formation is dominated by bending energy, increasing DNA length effectively reduces the tilt in the free energy curve because states at shorter end-to-end distances receive more stress relief, similar to the dotted line in Figure 1(a). This change predicts that loop capture and release rates measured at different DNA lengths would be anti-correlated; loops associated with higher mechanical stress are captured more slowly and released more quickly. This prediction has been confirmed for both DNA loops captured by Lac repressor Chen et al. (2014) and DNA loops captured by sticky ends Le and Kim (2013, 2014). While increasing DNA length evens out the bending stress over the entire DNA molecule, the base pair mismatch tends to localize sharp bending. Therefore, the effect of base pair mismatch might be quite different from that of increasing DNA length.

In this Letter, we investigated how base pair mismatch affects the stability of small DNA loops. As a model system for DNA loop capture and release, we used short double-stranded DNA molecules with sticky ends. To monitor loop capture and loop release events, we used the single-molecule FRET assay as previously published Jeong et al. (2016). Briefly, DNA molecules labeled with Cy3 and Cy5 near their sticky ends were immobilized to a NeutrAvidin-coated glass surface through a biotin linker, and loop capture or release was triggered by exchange of buffers with different NaCl concentrations (see Supplemental Material for more details). The first transition times in the FRET signals ($\Delta t$) of $\sim$150 individual DNA molecules were collected. The mean of $\Delta t$ spent in the unlooped state before looping is defined as the loop capture time ($\tau_{\textrm{unloop}}$), and the mean of $\Delta t$ spent in the looped state before unlooping is defined as the loop release time or loop lifetime ($\tau_{\textrm{loop}}$). All DNA molecules used in this study were shorter than 150-bp, the length regime where the free energy of loop formation is dominated by bending energy.

We first tried the loop capture geometry used in DNA cyclization, which we term as the “hairband loop” (Figure 2(a)). In this geometry, the complementary overhangs protrude from different strands so that the sticky ends can anneal in trans and stack upon each other. In a previous study, we showed that this end stacking, or equivalently nick closing, substantially increases the hairband loop stability Jeong and Kim (2018). Using the single-molecule FRET assay, we measured the hairband loop capture times with and without base pair mismatch in the center. As shown in Figure 2(b), hairband loop capture took less time in the presence of the mismatch as expected. The loop capture time further decreased with increasing mismatch size (circles, Figure 2(b)). The base pair mismatch in the center position led to the largest decrease in the loop capture time, and the decrease dropped as the mismatch was placed further from the center (triangles, Figure 2(b)). These observations confirm previous findings that mismatched base pairs reduce the energy barrier for loop formation by increasing DNA bendability Sharma et al. (2013); Kahn et al. (1994); Schallhorn et al. (2005); Yuan et al. (2006), and this barrier reduction is most effective when the mismatch is in the center Ranjith et al. (2005).

Next, we measured the hairband loop release times or loop lifetimes ($\tau_{\textrm{loop}}$) with and without the mismatch in the center. Since a mismatch could relieve the bending stress of the hairband loop, we thought that the loop lifetime would become longer. To our surprise, we observed the exact opposite effect where the central mismatch decreased the hairband loop lifetime (Figure 2(c)). Increasing the size of the mismatch from 1 bp to 3 bp led to a further decrease in the lifetime. This effect seemed to plateau past the mismatch size of 3 bp (Figure 2(c)). This result suggests that the mismatch-containing hairband loop is more kinetically unstable than the mismatch-free loop, which seems paradoxical through the lens of the one-dimensional model presented in Figure 1(a).

We thus considered the possibility that the transition state depends on other reaction coordinates besides the end-to-end distance, such as the closing angles at the loop junction. Since base stacking at the nick(s) in the hairband loop is a key determinant of decyclization kinetics Jeong and Kim (2018), we asked whether the central mismatch could destabilize the hairband loop by allosterically inducing nick opening. To investigate such allosteric coupling, we calculated the curvature profile of a kinkable semiflexible loop Vologodskii and Frank-Kamenetskii (2013) containing a defect with zero rigidity from a Monte Carlo simulation (see Supplemental Material for details). As shown in Figure 2(d), a kink with a sharp bending angle appeared most frequently at the furthest end of the loop from the defect. We also calculated the minimum energy conformation of a semiflexible loop while varying the rigidity of the defect and found that the bending angles of furthest points were highly correlated (Figure 2(e)). This loop-mediated correlation of sharp bending angles between most distant sites is termed cooperative kinking Lionberger et al. (2011), and has been observed in torsionally strained DNA minicircles by cryo-electron microscopy and molecular dynamics simulations Lionberger et al. (2011); Irobalieva et al. (2015); Sutthibutpong et al. (2016).

We hypothesized that the enhanced flexibility of the central mismatch destabilizes the hairband loop preventing nicks(s) on the opposite side from closing. This hypothesis provides a few testable predictions. First, if the mismatch were displaced from the midpoint of the DNA, the degree of destabilization would be dampened. In agreement with this prediction, we observed a longer loop lifetime when the mismatch was placed at a quarterpoint instead of the center (Figure 2(f)). Second, the cooperative kinking hypothesis requires nicks that can buckle under the bending stress, and therefore the mismatch-induced destabilization would be eliminated in a loop capture geometry free of end-stacking. We thus tested a different loop geometry referred to as the “hairpin loop”, where the complementary overhangs protrude from the same strand (Figure 3(a)). In this geometry, the sticky ends anneal in cis and cannot stack upon each other. Using these new DNA constructs with a central mismatch of various sizes, we repeated loop capture and release experiments. Similar to hairband loop capture, the hairpin capture time decreased with the size of base pair mismatch (Figure 3(b)). However, in sharp contrast to the hairband loop, the hairpin loop lifetime increased with mismatch size (Figure 3(c)). The effect of the base pair mismatch on the hairpin loop stability is therefore consistent with the prediction of the one-dimensional model. Overall, the lifetimes of hairpin loops were shorter than those of hairband loops, which is consistent with easier rupture of DNA duplex in an unzipping geometry than in a shearing geometry Mosayebi et al. (2015); Zhang et al. (2016); Tee and Wang (2018). These results lend strong support to the idea that cooperative kinking governs the kinetic stability of a mismatch-containing hairband loop.

The mismatch-dependence of the hairband loop release kinetics reveals the limitations of the one-dimensional two-state model (Figure 1(a)) and invites us to consider additional states and alternative reaction paths along another dimension. Here, we present two different paths ($k^{(0)}$ and $k^{(m)}$) that are likely to be the dominant ones for mismatch-free and mismatch-containing DNA (Figure 4(a)). Each path goes through three different states: unlooped, unstacked, and stacked. The loop capture rate is much greater in the presence of a central mismatch due to its enhanced flexibility ($k^{(m)}_{1}\gg k^{(0)}_{1}$). The reverse rate is expected to be slower with the mismatch ($k^{(\textrm{m})}_{2}<k^{(\textrm{0})}_{2}$) because of the weaker loop tension. Mismatch-free DNA undergoes small bending fluctuations uniformly throughout its contour, and therefore, follows an arc-like trajectory toward the looped state where end-stacking (nick closing) and end-unstacking (nick opening) transitions may occur. In comparison, DNA with a mismatch in the center can be sharply bent at a much lower energy cost, and therefore, the most dominant path toward the looped state will resemble a tweezers-like motion. As a result of this motion, the sticky ends anneal at a sharp angle, and the hairband loop with the mismatch faces a higher energy barrier for end-stacking (nick closing) than without ($k^{(m)}_{3}\ll k^{(0)}_{3}$). The mismatch not only suppresses end-stacking, but also promotes end-unstacking (nick opening) through cooperative kinking, which implies $k^{(m)}_{\textrm{4}}\gg k^{(0)}_{\textrm{4}}$. Hence, the apparent release rate of the hairband loop ($k_{\textrm{unloop}}$) becomes faster with the mismatch than without because the looped state with the mismatch is heavily biased towards the unstacked state. In comparison, for the hairpin loop that cannot proceed to the stacked state, the three-state model is reduced to the two-state model, and the loop release rate is slower with the mismatch ($k^{(\textrm{m})}_{2}<k^{(\textrm{0})}_{2}$).

The two paths boxed in Figure 4(a) represent the two most extreme paths in terms of kinetics, the top path for the slowest hairband loop capture and release, and the bottom for the fastest. In reality, there exists a continuum of paths going through the three states with intermediate rates, and the flexibility profile of DNA determines the relative weights at which individual paths are taken. Therefore, any changes to the flexibility profile of DNA would lead to correlated changes in the hairband loop capture and release rates. To test this idea, we measured hairband loop capture and release times of 16 unrelated sequences, all of the same length. Although limited in sample size, we observed a significant degree of correlation between the two times (Pearson correlation = 0.74, Figure 4(b)). This result suggests that cooperative kinking is a general mechanism that governs the kinetics of hairband loop capture and release.

In conclusion, we demonstrate that base pair mismatch can constrain the geometry and interactions for DNA loop capture through cooperative kinking, and the close coupling between hairband loop geometry and end-stacking can give rise to correlated changes between loop capture and release times (“easy come, easy go”). We propose a three-state model that correctly describes the effect of mismatched base pairs on the apparent kinetics of loop capture and release. We expect the effect of mismatched base pairs on protein-mediated DNA loops to be more complex because of the diversity in loop capture geometry Haeusler et al. (2012). Beyond passively captured DNA loops, it would be interesting to investigate whether base pair mismatches can also influence the kinetics of DNA loop extrusion Ganji et al. (2018); Marko et al. (2018) through cooperative kinking.