Twinkle-catalyzed Toehold-mediated DNA strand displacement reaction

Abstract

Strand exchange between homologous nucleic acid sequences is the basis for cellular DNA repair, recombination, and genome editing technologies. Specialized enzymes catalyze cellular strand exchange; however, the reaction occurs spontaneously when a single-stranded DNA toehold can dock the invader strand on the target DNA to initiate strand exchange through branch migration. Due to its precise response, the spontaneous toehold-mediated strand displacement (TMSD) reaction is widely employed in DNA nanotechnology. However, enzyme-free TMSD suffers from slow rates resulting in slow response times. Here, we show that human mitochondrial DNA helicase Twinkle can accelerate TMSD up to 6000-fold. Mechanistic studies indicate that Twinkle accelerates TMSD by catalyzing the docking step, which typically limits spontaneous reactions. The catalysis occurs without ATP, and Twinkle-catalyzed TMSD rates remain sensitive to base-pair mismatches. The simple catalysis, tuneability, and speed improvement of the catalyzed TMSD can be leveraged in nanotechnology, requiring sensitive detection and faster response times.

Graphical Abstract

INTRODUCTION

Strand exchange between homologous DNA sequences is the primary way cellular DNA is repaired^1–3. This process is catalyzed by specialized enzymes in the cell, but strand exchange between homologous DNA sequences can occur spontaneously if short single-stranded DNA regions on the target DNA called ‘toehold’ are accessible to an invader strand to dock and displace the complementary protector strand through spontaneous branch migration reaction^{4, 5} (Figure 1a). The simplicity and precise response of the toehold-mediated strand displacement (TMSD) reaction has been exploited widely in DNA and RNA nanotechnology for constructing dynamic nucleic acid-based devices, logic gates in DNA circuits, and biosensors^5–13. The TMSD reaction is energetically favorable due to the additional base-pairing energy gained in the product from the toehold region¹⁴. However, the diffusive nature of the docking step and the low stability of the toehold duplex make the spontaneous TMSD reaction kinetically slow and inefficient^{14, 15}. The slow response time of the spontaneous TMSD is a major limitation of its application in complex computational circuits and diagnostic devices^{7, 9, 12, 16}.

Figure 1:

Comparison of spontaneous and Twinkle-catalyzed TMSD on the 34-bp target substrate.

a. Schematic depicting basic steps of TMSD reaction with fluorescence-based detection. A 68-nt target strand (TS) is annealed to a 34-nt protector strand (PS) to create a 34-bp branch migration domain (BMD) for TMSD. 6-nt in the ssDNA tail of the target strand adjacent to the BMD forms a toehold to dock the 40-nt invader strand (IS). Three dG residues at the 5’ end of the target strand quench the fluorescence of FAM moiety at the 3’ end of the protector strand. FAM fluorescence intensity will increase upon displacement of the protector strand to provide a real-time assay for measuring the TMSD reaction rate.

b. Representative time courses of TMSD in the absence (black) and presence of Twinkle (red) with 10 nM 34-bp target dsDNA, 40 nM Twinkle hexamer, and 40 nM invader strand.

c. Spontaneous TMSD kinetics measured at increasing invader strand concentrations (circles) to fit a single exponential function (lines) to provide the TMSD rates (k_obs).

d. Twinkle-catalyzed TMSD kinetics measured using a rapid mixing stopped-flow instrument at increasing invader strand concentrations with 10 nM 34-bp target dsDNA and 40 nM Twinkle fit a single-exponential function to provide the catalyzed TMSD rates (k_obs).

e-f. Panel e shows the linear increase (grey solid line) in spontaneous TMSD rates (k_obs, grey circles) with increasing invader concentrations. The slope estimated the bimolecular rate of spontaneous TMSD (k^TMSD). Panel f shows the hyperbolic increase (red line) in Twinkle-catalyzed TMSD rates (red circles) with increasing invader concentrations. The fit to the hyperbolic Equation 2 provided the K_M and V_max.

g. Comparison of the bimolecular TMSD rate constants k^TMSD for the spontaneous and Twinkle-catalyzed (V_max/K_M) reactions. Error bars represent standard errors from data fitting.

h. Stopped-flow fluorescence time traces of TMSD on the substrates with 3’ or 5’ toehold measured at 10 nM target dsDNA and 40 nM invader strand with or without 40 nM Twinkle hexamer.

i-j. Panel i shows native PAGE analysis of TMSD reactions conducted using 5 nM 34-bp target dsDNA, 20 nM Twinkle hexamer, and 20 nM invader strand. Panel j compares the protector strand displacement kinetics with and without Twinkle.

Twinkle is the replicative helicase in the mitochondria that assembles into a hexameric/heptameric ring-shaped structure encircling DNA through its central channel^17–20. Under cellular conditions, Twinkle works cooperatively with the replicative polymerase γ and human mitochondrial single-strand binding protein to catalyze mitochondrial DNA replication^20–22. Twinkle exhibits a range of enzymatic activities, including ATP hydrolysis²³, DNA unwinding^21–24, Holliday junction resolution²⁵, and DNA annealing²⁴ (Figure S1). The DNA unwinding and Holliday junction resolution activities depend on ATP hydrolysis. In contrast, the DNA annealing activity of Twinkle does not require ATP hydrolysis^{24, 25}. Twinkle utilizes a forked DNA substrate to initiate the unwinding process of the dsDNA²³. It encircles the 5’-ssDNA tail and subsequently translocates along this strand, powered by NTP hydrolysis, to unwind the duplex DNA through a mechanism known as strand exclusion (Figure S1). On its own, Twinkle is poor at separating the strands of even short stretches of duplex DNA; however, the unwinding activity is stimulated when single-strand binding protein or a ‘trap’ DNA homologous to one of the unwound strands is added to the reaction^{23, 24}. Our prior studies showed that the trap DNA undergoes a strand exchange reaction to aid the DNA unwinding process, and when the trap contained a homologous region to the 3’-tail strand, the exchange occurred without NTP²⁵.

In this study, we show that Twinkle can catalyze strand exchange reaction using TMSD on a non-forked DNA that cannot be unwound in a canonical way. Depending on the substrate and toehold length, we observed a ~6000-fold acceleration of strand exchange rate in the presence of Twinkle relative to the spontaneous reaction. We show that rate acceleration is attributed to Twinkle’s DNA annealing activity, which catalyzes the crucial toehold formation step that limits the spontaneous TMSD reaction. Our mechanistic studies reveal that Twinkle has multiple DNA binding sites, which bring the invader and target strands in proximity, promoting the annealing of the toehold, analogous to the ‘Jencksian Circe’²⁶ effect in enzyme catalysis. The catalyzed reaction’s rapid and regulated response times and simplicity offer distinct advantages for leveraging the catalyzed TMSD in broader applications for DNA nanotechnology.

EXPERIMENTAL PROCEDURES

Oligonucleotides.

All oligonucleotides (Table S1) were ordered HPLC purified from Integrated DNA Technologies, USA, and dissolved in 1xTE buffer to desired concentrations. Target dsDNA was made by annealing the target and protector strand in the annealing buffer (1xTE buffer, 100 mM NaCl). Annealing of oligos was confirmed on 20% native TBE gel.

Twinkle protein expression and purification.

Human mitochondrial replicative helicase Twinkle lacking the N-terminal 42 amino acid long mitochondrial localization signal sequence (Δ1–42) and with a C-terminal histidine tag was expressed in Sf9 insect cells as described elsewhere²⁷. Briefly, Twinkle-expressing cells were gently lysed by stirring the cell pellets in lysis buffer (50 mM sodium phosphate, pH 7.0, 90 mM sodium chloride, 5 mM magnesium chloride, 0.5% v/v NP-40, 10% glycerol, 0.1% Triton X-100, 10 mM beta-mercaptoethanol and 1 Roche cOmplete EDTA-free protease inhibitor tablet per 50 ml volume) for 2 hours at 4 °C at a low speed. Sodium chloride concentration was increased to 1 M by adding salt slowly to the lysate while gently stirring on ice, and the lysate was further stirred at low speed for 30 min on ice. Lysate filtered through a 0.22 μm filter was loaded on a Ni affinity chromatography column (HisTrap HP, Cytiva, USA) equilibrated with 50 mM sodium phosphate buffer (pH 7.0) containing 400 mM sodium chloride, 10% glycerol, 10 mM imidazole, and 10 mM beta-mercaptoethanol. The column was washed twice with the above buffer containing 40 mM imidazole, and the bound protein was eluted by running a continuous imidazole gradient (40 mM to 500 mM imidazole). Pure fractions were pooled, and sodium chloride concentration was reduced from 1 M to 300 mM by diluting the protein with a dilution buffer (50 mM sodium phosphate, pH 7.0, 10% glycerol, 10 mM beta-mercaptoethanol). Twinkle was further purified on a Heparin column (HiTrap Heparin HP affinity column, Cytiva, USA) equilibrated with HEPES buffer (pH 7.2) of 300 mM sodium chloride, 10% glycerol, 0.5 mM EDTA, and 1 mM DTT. Bound protein was eluted with a continuous sodium chloride gradient (300 mM to 2 M sodium chloride). Pure Twinkle peak eluted at around 800 mM sodium chloride. Twinkle fractions were pooled, concentrated, and stored at −80 °C.

Stopped-flow measurement of TMSD and toehold docking kinetics.

A stopped-flow instrument allowed fast kinetic measurements of TMSD reactions from milliseconds to the minute’s time range at 25°C. Reaction Mixture A (20 nM target dsDNA with or without 80 nM Twinkle hexamer) was filled in syringe A, and Reaction Mixture B (invader strand) was in syringe B. The reaction buffer contained 50 mM Tris-acetate, pH 7.5, 50 mM sodium acetate, 0.01% Tween 20, 1 mM EDTA and 5 mM DTT. Unless mentioned in the figures and text, 80 nM invader strand concentration was used, and the target dsDNA was incubated with Twinkle for 20 min. After mixing equal volumes of Reaction Mixtures A and B, the final concentrations were 10 nM target dsDNA, 40 nM Twinkle hexamer, and 40 nM invader strand. The reaction mixture was excited with a 495 nm light, and FAM fluorescence emission was measured in real-time using a long pass 515 nm cut-off filter. The time courses were fit to a single exponential (Equation 1) to obtain the k_obs with minimized standard errors.

$$F=A\left(1-e^{-kt}\right)+C$$ Equation 1

Where F is the fluorescence intensity at the time ‘t’, A is the fluorescence signal amplitude, C is the initial fluorescence at time zero, and k_obs the observed rate of TMSD. For better visualization, the fluorescence traces were normalized by subtracting the initial fluorescence. The traces that showed exponential increase were further normalized to unity by dividing the fluorescence intensity data by the respective amplitudes (A).

The k_obs were measured at various invader strand concentrations and fit to a linear equation with intercept or to a hyperbola (Equation 2).

$$k_{obs}=V_0+\frac{(V_{max}-V_0)\left[IS\right]}{K_M+\left[IS\right]}$$ Equation 2

Here k_obs is the observed rate of TMSD at invader strand concentration [IS], V₀ is the k_obs at zero invader concentration, and V_max is the maximum reaction rate.

Variations of the above strategy were used to conduct different reactions. For reactions conducted with Mg²⁺ and ATP, 11 mM magnesium acetate was added to both the Reaction Mixtures while 4 mM ATP was added to Reaction Mixture B. TMSD reactions under high salt conditions were conducted in a buffer containing 300 mM sodium acetate rather than 50 mM sodium acetate. For the TMSD reactions with very slow kinetics, Reaction Mixtures A and B were mixed manually, and fluorescence intensity was measured on a plate-based Spark fluorometer (excitation wavelength = 485 nm, emission wavelength = 535 nm) (Tecan Group, Switzerland).

The toehold docking kinetics was measured in the stopped-flow instrument with Reaction Mixture A containing 20 nM FAM labeled target dsDNA with or without 80 nM Twinkle and Reaction Mixture B with various concentrations of BHQ1-labeled ssDNA with (toe-dT₁₁dA₄) or without (dT₂₂) toe sequence.

Gel-based TMSD assay.

A discontinuous gel-based assay measured the displacement of the protector strand directly. Mixtures A and B in equal volumes were mixed at 25°C to initiate the TMSD reaction, and fixed reaction volumes were drawn at the stated time points, mixed with 0.5% SDS and immediately loaded on a 20% TBE native gel running continuously at a low voltage of 30 V with 1xTBE gel running buffer. Once all the time points were loaded, the gel was run at 120 V for around 2 hours to resolve the substrate and product DNAs. The gels were run with a cold buffer while keeping the gel apparatus on ice. Gels were scanned using the GE Typhoon 9500 imaging system, and fluorescence intensity of FAM labeled target dsDNA and displaced protector strand were quantified. The fraction of displaced protector strand at each reaction time was determined and plotted as a function of time. Data were fitted to a single exponential trend (Equation 1), which provided the first-order rate constant and amplitude of the strand displacement reaction.

Equilibrium binding of dsDNA and ssDNA to Twinkle.

The equilibrium dissociation constant (K_D) of FAM-labeled target dsDNA or FAM-labeled protector ssDNA complex with Twinkle was measured using fluorescence polarization titrations. All binding reactions were conducted at 25°C, and fluorescence polarization was measured on a plate-based fluorometer (Tecan Spark, Switzerland). Briefly, 5 nM DNA was incubated with a range of Twinkle concentrations (T) in 1x reaction buffer (50 mM Tris-acetate, pH 7.5, 0.01% Tween 20, 1 mM EDTA and 5 mM DTT) with 50 mM sodium acetate (low salt) or 300 mM sodium acetate (high salt). Equilibrium binding was detected by measuring fluorescence polarization (p) (excitation wavelength = 485 nm, emission wavelength = 535 nm). The titration data were fit to Equation 3 to determine the K_D values.

$$p=\frac{p_{max}\text{*}\left[T\right]}{\left(K_{\text{D}}+\left[T\right]\right)}+c$$ Equation 3

Here, p is polarization at Twinkle concentration [T], p_max is the maximum polarization when all the DNA is bound to Twinkle, c is the polarization of free DNA, and K_D is the dissociation constant of the Twinkle-DNA complex.

Observation of the secondary binding sites.

To determine the dissociation constant (K_D) of the DNA bound to the secondary DNA binding site on Twinkle, we incubated 40 nM of FAM-labeled target dsDNA and 40 nM of Twinkle at 25°C for 20 min and added to increasing concentrations of BHQ1-dT22 or BHQ1-toe-dT₁₁dA₄ DNA. After further incubation for 20 min, each sample’s fluorescence intensity and polarization were measured on a plate-based fluorometer at 25°C (Tecan Spark, Switzerland). Decreasing fluorescence was fitted to Equation 4 to determine the K_D for the secondary site DNA binding.

$$F=F_{\mathit{\text{final}}}+\frac{F_{\mathit{\text{initial}}}-F_{\mathit{\text{final}}}}{\left(1+\frac{\left[DNA\right]}{K_D}\right)}$$ Equation 4

Where F is the fluorescence intensity at BHQ1-ssDNA concentration [DNA], F_initial and F_final are fluorescence intensities of the target dsDNA-Twinkle complex in the absence of BHQ1-ssDNA and presence of saturating concentrations of BHQ1-ssDNA, respectively, and K_D is the dissociation constant of BHQ1-ssDNA bound at the secondary binding site of Twinkle.

Measurement of the dsDNA product dsDNA off-rate.

To estimate the off-rate of the TMSD product dsDNA from Twinkle, 40 nM Twinkle was rapidly mixed with different concentrations of FAM labeled product dsDNA (FAM labeled target strand annealed to invader strand) in the stopped flow device set at 25°C and the rate of fluorescence increase upon binding of Twinkle to the labeled dsDNA was measured. The observed rates were fit to linear trend and the intercept on the Y-axis was used to estimate the off-rate of the complex. Alternatively, the off-rate of the dsDNA-Twinkle complex was directly measured using a stopped-flow fluorescence assay. Briefly, 80 nM Twinkle (hexamer concentration) and 20 nM FAM labeled TMSD product were incubated for 20 minutes at 25°C and loaded in one of the two stopped flow syringes. A 10-fold excess of unlabeled product dsDNA (200 nM) was loaded in the second syringe. Equal volumes of the contents in the two syringes were rapidly mixed, and FAM fluorescence was measured in the flow cell in real time. The fluorescence intensity decreases as the FAM-labeled product falls off the Twinkle. The excess of unlabeled product DNA traps the available Twinkle binding sites to prevent the rebinding of labeled DNA to Twinkle. Fluorescence traces were fitted to a single-exponential function (Equation 1) to determine the off-rate (k_off). Fluorescence anisotropy kinetics of just the dsDNA and dsDNA-Twinkle complex in the presence and absence of unlabeled product dsDNA were also measured on a plate-based fluorometer to confirm complex dissociation.

RESULTS

Twinkle catalyzes the TMSD reaction.

We discovered Twinkle’s role in catalyzing the TMSD reaction when measuring strand exchange on a 34-bp DNA substrate. In this substrate, the target strand (TS) contains a 34-bp strand exchange region with a 6-nt toehold region flanked by a 31-nt 3’- end ssDNA tail (Figure 1a). The protector strand (PS) anneals to the 34-nt strand exchange region and contains a 3’-end fluorescein (FAM). When the two strands are annealed, FAM fluorescence is quenched by the string of three dG residues in the target strand. Upon addition of an invader strand (IS), complementary to the 34-nt target strand and the toehold region, we observed a fast displacement of the protector strand through FAM fluorescence increase in the presence of Twinkle, but there was no fluorescence increase in its absence (Figure 1b, Figure S2a–b). A slight decrease in fluorescence over the time course was observed, possibly due to photobleaching. No ATP or Mg²⁺ ions were added to the reaction; hence, the reaction was not occurring through Twinkle’s ATPase-driven unwinding activity.

To measure the spontaneous TMSD reaction in the absence of Twinkle, we had to increase the invader strand concentration to micromolar amounts (Figure 1c, Figure S2c). The spontaneous TMSD rate increased linearly with invader strand concentrations, providing a bimolecular TMSD rate constant of 18 M⁻¹s⁻¹ (Figure 1e). The Twinkle-catalyzed TMSD rates also increased with increasing invader strand concentrations (Figure 1d, Figure S2d); however, unlike the spontaneous reaction, the rates increased hyperbolically, showing a saturation behavior (Figure 1f). The hyperbolic invader concentration dependency provided an invader K_M of 271 nM and TMSD V_max of 0.029 s⁻¹. The V_max/K_M ratio provided the apparent second-order rate constant of catalyzed-TMSD as 10.8 × 10⁴ M⁻¹s⁻¹. Comparison of the TMSD rate constants in the presence and absence of Twinkle (10.8 × 10⁴ M⁻¹s⁻¹/18 M⁻¹s⁻¹) indicated that Twinkle accelerates the TMSD rate on the 34-bp target dsDNA by ~6000 fold (Figure 1g). The toehold region was essential for the observed TMSD catalysis in the presence of Twinkle (Figure S2e–f).

The 34-bp TMSD substrate contains a 3’-end toehold region; hence, the directionality of the branch migration reaction from the invader’s perspective is from the 5’-end to the 3’-end. Twinkle moves along the ssDNA in the 5’−3’ direction²³. If Twinkle’s directional movement on DNA somehow influences the branch migration reaction, we would observe a different TMSD rate on an opposite polarity substrate. Our experiments show that Twinkle catalyzes TMSD on the substrate with a 5’-toe region at a ~1.4-fold slower rate of 0.013 s⁻¹ compared to the 3’-toe forming substrate (Figure 1h, Figure S2g–h). If translocation played a significant role, we would expect this difference to be larger. Hence, we argue that Twinkle is catalyzing the TMSD on both 3’-toe and 5’-toe substrates through a mechanism that does not require its directional translocation activity. The 1.4-fold difference could be an effect specific to these DNAs, and a more detailed study on a diverse set of DNA substrates is needed to confirm how the directionality of branch migration affects Twinkle-catalyzed TMSD.

Having observed TMSD catalysis for the first time with Twinkle, we wanted to verify this activity by an independent assay. The TMSD activity of Twinkle was verified using a gel assay that resolves the FAM-labeled target dsDNA and the protector strand product (Figure 1i). Spontaneous TMSD produced < 3% of the protector strand at the end of one hour. However, in the presence of Twinkle, the product was observed within a minute, with ~97% completion after 60 min. The TMSD rate in the presence of Twinkle was estimated as 0.007 s⁻¹ (Figure 1j), which closely resembles the rates measured with the real-time stopped-flow assay under similar conditions (Figure S2i–j).

Substrates with 6-nt toehold regions typically show TMSD rates of ~10⁶ M⁻¹s⁻¹ on different-sized target DNAs^{14, 15}, whereas TMSD on our 34-bp substrate occurred slowly at a rate of 18 M⁻¹s⁻¹, which was puzzling to us. Such slow spontaneous TMSD rates are observed on substrates with one nucleotide toehold¹⁵. Analysis of the DNA sequence revealed that both the target DNA and invader strand have unintended secondary structures (Figure S3a–e), reducing the docking region from 6-nt to ~1-nt (Figure S3f). One of the secondary interactions was from the 3’-ssDNA tail, whose removal increased the spontaneous TMSD rate by 6-fold (Figure S3g, left panel). While the removal of the secondary structure-forming 3’ tail in the target DNA substrate partially relieves the difficulties in toehold formation, the secondary structure in the invader, which obstructs 3-bp in the toehold formation still hinders the reaction.

The Twinkle-catalyzed rate increased by only 2-fold (Figure S3g, right panel), which indicates catalyzed TMSD is less sensitive to secondary DNA structures than spontaneous TMSD. Hence, Twinkle-catalyzed TMSD would be more effective on naturally occurring DNA sequences. We also asked if Twinkle could catalyze TMSD from toehold regions in gaps between two flanking duplexes. Twinkle-catalyzed TMSD was observed on two such DNA configurations, where annealing to the invader strand would result in a nick or no nick, indicating flexibility in the Twinkle catalysis of TMSD (Figure S3h–i).

Mechanism of Twinkle-catalyzed TMSD.

We wished to understand how Twinkle accelerates the TMSD reaction, and for such mechanistic investigations, it was necessary to compare the kinetics of spontaneous TMSD to the Twinkle-catalyzed TMSD. Because the secondary structures in the 34 bp DNA were inadvertently introduced and drastically slowed down the spontaneous TMSD, we designed a new set of 15-bp substrates (Figure 2a) lacking secondary structures¹⁵. The TMSD kinetics was measured in real-time using the FAM and dG quencher pair, as described for the 34-bp substrate. At 10 nM 15-bp substrate with 7-nt toehold and 40 nM invader, the spontaneous TMSD reaction occurred with a rate of ~0.01 s⁻¹, and the Twinkle-catalyzed at 0.38 s⁻¹ (Figure 2b).

Figure 2:

Comparison of spontaneous and Twinkle-catalyzed TMSD on the 15-bp target substrate

a. Schematics of the TMSD reaction on a target dsDNA having 15-bp branch migration domain (15-bp BMD) with a 7- nt toehold. Three dG residues at the 3’ end in the target DNA quench the FAM fluorescence at the 5’ end of the 15-nt protector strand. FAM fluorescence increases as the 22-nt invader strand with a 7-nt toehold region and a complementary 15-nt BMD displaces the protector strand through TMSD.

b. Representative stopped-flow fluorescence time traces showing spontaneous (dark grey) and Twinkle-catalyzed (red) TMSD on the 15-bp substrate. Control reactions show that TMSD does not occur without an invader strand.

c. Spontaneous TMSD rates (k_obs) (dark grey) increase linearly with increasing invader concentrations with the slope providing the spontaneous k^TMSD. The Twinkle-catalyzed rates (red) increase hyperbolically and fit Equation 2 provided the V_max and K_M values. The V_max/K_M ratio estimated the Twinkle-catalyzed k^TMSD.

The observed rates are means from at least two independent measurements.

d-e. Panel d compares the TMSD rates (Mean, N = 3) of substrates with different toehold lengths (1-nt, 3-nt, 5-nt or 7-nt) at 10 nM target dsDNA, 40 nM invader strand with (red circles) and without (gray circles) 40 nM Twinkle hexamer. Gray and red bars show data from two modified 7-nt invaders which create 1-nt mismatches in the toehold domain at different positions, Mismatched TD 1 (dark red and black) and Mismatched TD 2 (light red and light black). Panel e plots the fold increase in TMSD rates due to Twinkle catalysis on different substrates (mean, N = 3).

f-h. Panel f shows a schematic of substrate design to study the effect of a 2-nucleotide mismatch (depicted as triangle) in the target dsDNA which is repaired by a fully matched invader when it displaces the mismatched protector strand. Panel g shows spontaneous and Twinkle-catalyzed TMSD reaction rates (mean ± SD, N = 3) measured with 10 nM 15-bp or 25-bp matching target dsDNA or a 15-bp target dsDNA with 2-nucleotide mismatches and 40 nM invader strand. Panel h compares the fold increase in TMSD rates due to Twinkle catalysis on substrates with different branch migration domains (mean ± SD, N = 3).

i-k. Panel i shows a schematic of substrate design to study the effect of modified invader strands creating 1, 2, or 3 nucleotide mismatches in the branch migration domain. Panel j compares spontaneous and Twinkle-catalyzed TMSD reaction rates (mean ± SD, N ≥ 2) measured with 10 nM 15-bp target dsDNA and 40 nM invader strands with 0, 1, 2, or 3-nucleotide mismatches (depicted as triangles). Spontaneous TMSD on the 3-nucleotide mismatch was undetectable. Panel k shows a linear relationship between the Log₁₀ of the mean TMSD rates and nucleotide mismatches. Spontaneous and Twinkle-catalyzed reactions show similar trending slopes.

The spontaneous TMSD rate increased linearly with invader strand concentrations, providing a second-order TMSD rate constant of 1.7 × 10⁵ M⁻¹s⁻¹ (Figure 2c, Figure S4a–b). In contrast, the Twinkle-catalyzed rates increased hyperbolically with invader concentration, similarly to the 34-bp substrate (Figure 2c, Figure S4c). The hyperbolic fit provided a TMSD V_max of 2.4 s⁻¹ and invader strand K_M of 232 ± 34 nM. Comparison of the V_max/K_M of 1.03 × 10⁷ M⁻¹s⁻¹ to the spontaneous second-order rate indicates that Twinkle accelerates TMSD on the 15-bp substrate by ~59-fold (Figure S4d).

Next, we asked how the toehold domain length affects the spontaneous and catalyzed TMSD rates. Increasing the toehold domain length from 1-bp to 7-bp affected both the spontaneous and Twinkle-catalyzed TMSD rates. However, the spontaneous TMSD was more drastically affected (~16000 fold) than the catalyzed rates (~150 fold) from toehold length increase from 1-bp to 7-bp (Figure 2d, Figure S4e). Furthermore, the spontaneous TMSD rate increased exponentially with toehold domain length from 1 to 7-bp, whereas Twinkle-catalyzed rates increased between 1 to 5-bp length toehold and showed no change between 5 and 7-bp. Comparing the rates revealed that the TMSD rate acceleration by Twinkle was the highest on the shortest toehold substrate (~4000 for 1-bp toehold) and then progressively decreased as the toehold length increased (~38 for 7-bp toehold) (Figure 2e).

We also tested the TMSD reaction with an invader strand that had no toe region or a non-complementary 7-nt toe region. No TMSD reaction was observed under spontaneous conditions and was very slow with Twinkle (Figure S4f). We had to use a high invader concentration of 500 nM to measure the TMSD rate of ~0.001 s⁻¹ (Figure S4f). The TMSD rate on the 7-bp toehold reaction was ~0.38 s⁻¹ at 40 nM invader concentration (Figure 2c). Assuming the zero toehold TMSD rate increases linearly between 40 nM and 500 nM, we estimate the TMSD on the zero toehold substrate is ~5000 fold slower than the 7-bp toehold.

Next, we asked how mismatches in the toe region affect the TMSD rates. Introducing a 1-nucleotide mismatch at two different positions in the 7-bp toehold domain decreased the spontaneous TMSD rate by 30 and 22-fold, whereas the Twinkle-catalyzed rates remained nearly unaffected, slowing down by 1.27 and 1.13-fold (Figure 2d, Figure S4g). This is consistent with the toehold length dependence, where we observed TMSD rate saturation after a 5-bp toehold length. Thus, the rate acceleration by Twinkle on the 7-bp toehold with a 1-nucleotide mismatch was 720 to 830-fold in comparison to the 38-fold observed with the matched toehold (Figure 2e).

Increasing the branch migration domain from 15-bp to 25-bp while keeping the 7-bp toehold domain constant did not affect the rates of the spontaneous TMSD, while the Twinkle-catalyzed TMSD decreased by ~1.5-fold (Figure 2f–h). Furthermore, TMSD reactions on a mismatched 15-bp branch migration domain with a matching ‘repairing’ invader did not affect the rates of spontaneous reactions, but the Twinkle-catalyzed TMSD rate increased by 1.7-fold (Figure 2f–h). These results are consistent with the different rate-limiting steps of the catalyzed and spontaneous TMSD reactions. The spontaneous TMSD, limited by toehold docking, is not affected by changes in the branch migration domain, whereas Twinkle-catalyzed TMSD, partly limited by branch migration, shows changes (Figure 2c).

We wondered if Twinkle would catalyze branch migration at a faster rate when provided with Mg²⁺ and ATP. Twinkle uses its ATPase activity to unwind DNA and catalyze branch migration on Holiday junction^{24, 25}. However, Mg²⁺ and ATP did not affect the Twinkle-catalyzed rate (Figure S4h–i). Interestingly, MgATP doubled the spontaneous TMSD rate from 0.01 s⁻¹ to 0.02 s⁻¹, probably due to Mg²⁺ ions shielding the DNA backbone and accelerating toehold annealing.

We asked if Twinkle-catalyzed TMSD is sensitive to mismatches in the branch migration domain, as shown for the spontaneous TMSD^{28, 29}. We designed three 15-nt invader strands with 1, 2, or 3 contiguous nucleotide changes that resulted in mismatches between the invader and the target strand (Figure 2i). Both TMSD rates were equally sensitive and progressively decreased with the increased number of mismatches (Figure 2j). A log plot of rates versus the number of mismatches shows comparable slopes for the spontaneous and catalyzed reactions (Figure 2k). This suggests Twinkle may not be actively catalyzing the branch migration reaction under these conditions.

Twinkle catalyzes TMSD by promoting toehold docking.

There are two essential steps in the TMSD reaction: the toehold docking step and the strand displacement step catalyzed by branch migration (Figure 1a). The linear increase in spontaneous TMSD rate with invader strand concentrations is clear evidence for the docking step limiting the spontaneous TMSD. In contrast, Twinkle-catalyzed rates show a hyperbolic behavior, indicating that TMSD is limited by the docking step only at very low invader concentrations, but higher invader (> invader K_M), toehold docking is fast, and branch migration steps and strand displacement limit TMSD. This is consistent with the 1.5-fold slower rate of Twinkle-catalyzed TMSD on the 25-bp versus 15-bp substrate and the increased rate of the catalyzed TMSD reaction with the repairing invader (Figure 2f–g).

To investigate if Twinkle catalyzes the toehold docking step, we developed a new assay that exclusively measured the toehold annealing kinetics independent of branch migration. The invader strand contained the 7-nt toehold sequence, but the branch migration sequence was replaced by a non-complementary dT₁₁dA₄ segment (Figure 3a). A FAM and BHQ1 pair were introduced in the toehold-forming regions in the target and invader strands, respectively. The BHQ1 strongly quenches the FAM fluorescence in a distance-dependent manner; hence, toehold docking can be monitored in real time through FAM quenching by the BHQ1.

Figure 3:

Twinkle catalyzes TMSD by promoting the toehold docking step

a-b. Panel a shows the experimental design to measure the toehold docking step independent of branch migration. The design uses a 5’ FAM labeled 15-bp target dsDNA and a 3’ BHQ1 labeled invader containing a complementary toe sequence and a non-complementary branch migration domain (BHQ1-toe-dT₁₁dA₄). FAM fluorescence quenches upon toehold formation. Panel b shows representative fluorescence time traces of toehold docking with and without 40 nM Twinkle hexamer. The spontaneous docking reaction was performed with 2.56 μM BHQ1-toe-dT₁₁dA₄ and the Twinkle-catalyzed with 40 nM.

c-f. Panel c shows the linear increase in toehold docking rates with increasing BHQ1-toe-dT₁₁dA₄ (solid lines) with and without Twinkle. Data are means of at least two independent measurements. Panels d and e show the docking step on-rates (k_{on toehold}) from slopes and off-rates (k_{off toehold}) from Y-intercepts. Panel f compares the K_D of the toehold duplex (k_{off toehold}/k_{on toehold}) with or without Twinkle. The error bars are the standard errors from data fitting.

g. Free energies of toehold duplex show stabilization by Twinkle (ΔG^+Twinkle- ΔG^−Twinkle). The predicted value of the toehold duplex’s free energy (ΔG) was obtained from an online resource (https://arep.med.harvard.edu/cgibin/adnan/tm.pl), and the experimental values were calculated from the measured K_D values depicted in f. The standard errors from the estimates are shown.

h-i. Comparison of docking and TMSD reactions under spontaneous (h) and Twinkle-catalyzed conditions (i).

Upon mixing the target with the invader DNAs, we observed a time-dependent decrease in FAM fluorescence (Figure 3b). The amplitude of fluorescence change was much greater in the presence of Twinkle than in the absence. Nevertheless, we could measure the kinetics of toehold docking with increasing invader concentrations. The linear increase in docking kinetics with invader concentration provided the toehold docking rate, k_{on toehold} (slope), and the dissociation rate, k_{off toehold} (Y-intercept) (Figure 3c–e, Figure S4j–l). Comparing these rates with and without Twinkle demonstrated that Twinkle not only increases the rate of toehold docking (by 25-fold) but also decreases the rate of dissociation of the toehold duplex (by 50-fold). The toehold docking rate in Twinkle’s absence was 0.53 × 10⁶ M⁻¹s^−1, and the dissociation rate was 4.7 s⁻¹. The toehold docking rate with Twinkle was 1.34 × 10⁷ M⁻¹s^−1, and the dissociation rate was 0.1 s⁻¹.

The ratio k_{off toehold}/k_{on toehold} provided the toehold duplex K_D of 8.9 μM in the absence of Twinkle and 7.8 nM in the presence of Twinkle (Figure 3f). Thus, the ΔG of toehold duplex formation is −6.88 ± 0.06 kcal/mole in the absence of Twinkle, consistent with the predicted value from base-pair annealing (−6.26 kcal/mol) (Figure 3g). In the presence of Twinkle, the ΔG of toehold formation is −11.05 ± 0.25 kcal/mol, which indicates that Twinkle stabilizes the toehold duplex by −4.17 kcal/mole (Figure 3g). These results demonstrate that Twinkle catalyzes TMSD by promoting the docking step and preventing the toehold duplex from falling apart. Increased lifetime of toehold duplex in the presence of Twinkle is expected to increase the probability of initiating the branch migration reaction.

Comparing the toehold docking rates with the measured TMSD rates indicates that different steps limit the spontaneous and Twinkle-catalyzed TMSD reactions. The docking step limits the spontaneous TMSD because the toehold docking rate (0.53 × 10⁶ M⁻¹s⁻¹) is similar to the spontaneous TMSD rate (0.17 × 10⁶ M⁻¹s⁻¹) over the entire range of invader concentrations examined here (Figure 3h). In the presence of Twinkle, the TMSD rates are slower than the toehold docking rates (Figure 3i), which indicates that branch migration steps most likely limit the catalyzed TMSD rates.

Twinkle employs its multiple DNA binding sites to catalyze toehold docking.

Having established that Twinkle accelerates the TMSD reaction by promoting the toehold docking step, we asked how Twinkle increased the kinetics and stability of the toehold duplex. One mechanism by which Twinkle can catalyze toehold docking is by binding and bringing the target and invader DNA together. However, there is no evidence that Twinkle can bind two DNAs simultaneously. Studies indicate that Twinkle binds one DNA in the ring’s central channel^{17–19, 30}. Nevertheless, it is possible to bind additional DNAs as recent studies show Twinkle’s N-terminal domain has a DNA binding activity²¹.

Fluorescence polarization-based titrations show that Twinkle has a high affinity for target dsDNA (3 nM) and ssDNA (5.48 nM) in low salt or 50 mM sodium acetate buffer but not in high salt or 300 mM sodium acetate (Figure 4a, Figure S5a–b). Additionally, a 30-bp blunt-ended dsDNA binds Twinkle with high affinity (2.4 nM), which confirms that Twinkle can bind to dsDNA with no ssDNA overhangs (Figure S5a). Consistent with the DNA binding ability, Twinkle catalyzed TMSD at 50 mM but not 300 mM salt (Figure 4b–c, Figure S5c). Twinkle-catalyzed TMSD rates at the higher salt were the same as the spontaneous TMSD rate, which interestingly increased by 4-fold in the presence of higher salt, likely due to the Na⁺ ions shielding the negatively charged DNA backbone to facilitate DNA annealing. These results indicate that Twinkle’s DNA binding activity is essential for TMSD catalysis.

Figure 4:

Twinkle uses multiple DNA binding sites to catalyze toehold docking

a-c. Panel a shows the K_D values of FAM-labeled ssDNA and dsDNA on Twinkle in 50 and 300 mM sodium acetate buffer. Panel b shows the TMSD kinetics in low and high salt buffer conditions obtained with 10 nM 15-bp target dsDNA, 40 nM invader strand with and without 40 nM Twinkle. Panel c shows the fold increase in TMSD rates by Twinkle under low and high salt conditions (mean, N=2).

d-e. Panel d shows the scheme to measure the K_D of the secondary DNA binding site on Twinkle using FAM-labeled 15-bp target dsDNA and BHQ1-labeled dT₂₂ ssDNA. Panel e shows the progressive decrease in FAM fluorescence intensity with increasing BHQ1-dT₂₂ (orange circles) or BHQ1-toe-dT₁₁dA₄ ssDNA (black circles). Data were fit to hyperbolic function (Equation 4) to estimate the K_D of Twinkle’s secondary DNA binding site with standard fit errors. Data are means of two independent repeats.

f-h. Panel f compares the rate dependency of Twinkle binding to the toehold forming BHQ1-toe-dT₁₁dA₄ and the toehold domain lacking BHQ1-dT₂₂ at the secondary DNA binding site. Panel g compares the estimated on-rates, and panel h compares the off-rates obtained from slopes and the Y-intercepts. Data are means of two independent measurements.

To investigate if Twinkle can bind more than one DNA, we mixed a complex of Twinkle and FAM-labeled target dsDNA with increasing concentration of a non-complementary BHQ1-labeled dT₂₂ DNA. If Twinkle binds to both DNAs, then the BHQ1-labeled ssDNA will quench the fluorescence of FAM-labeled target dsDNA (Figure 4d). As the concentration of BHQ1-dT₂₂ was increased from 0 to 100 nM, the FAM fluorescence intensity decreased hyperbolically (Figure 4e) while the FAM polarization remained high (Figure S5d and S5f). The decrease in FAM fluorescence was not observed in the absence of Twinkle (Figure S5d). These data provide evidence that Twinkle brings the BHQ1-labeled DNA in proximity to the FAM-labeled DNA. Thus, Twinkle has multiple DNA binding sites. The DNA binding site in the central channel of the Twinkle ring is well-established and referred to as the primary binding site. The newly discovered binding site here is the secondary DNA binding site. Data fitting indicates that the secondary DNA binding site has a K_D of 27.7 ± 3.8 nM. The DNA binding experiments were also carried out with the toe-forming BHQ1- toe-dT₁₁dA_4, which binds with a K_D value of 24.7 ± 3.9 nM (Figure 4e, Figure S5e). As expected, the toe-forming ssDNA strongly quenches the FAM fluorescence, whereas the DNA that cannot form the toehold duplex quenches to a lower extent.

Next, we asked if the rate of DNA binding to Twinkle limits the toehold annealing rate. The on and off-rates of the non-complementary BHQ1-dT₂₂ were measured using the FAM quenching signal. The DNA binding k_on (0.96 × 10⁷ M⁻¹s⁻¹) closely matches the toehold annealing k_on (1.34 × 10⁷ M⁻¹s⁻¹) (Figure 4f–g, Figure S5g). The off-rate or dissociation rate of BHQ1-dT₂₂ from Twinkle (0.37 s⁻¹) is ~4 times greater than the dissociation rate of the toe-forming DNA (0.1 s⁻¹) (Figure 4h), most likely due to additional stabilization from toehold duplex (Figure S5g–h). These results demonstrate that the rate of DNA binding to Twinkle dictates the rate of toehold formation in the presence of Twinkle.

Mechanism of spontaneous and Twinkle-catalyzed TMSD.

Based on the results of this study, we propose the following mechanisms for the spontaneous TMSD and Twinkle-catalyzed TMSD reactions (Figure 5). We divide the TMSD reaction into two major steps: the toehold docking step, resulting in the 7-bp toehold duplex, and the strand displacement step, leading to the release of the protector strand. On the 15-bp substrate with no secondary structures, the toehold docking reaction occurs at a fast rate constant of 0.53 × 10⁶ M⁻¹s⁻¹; however, the resulting 7-bp toehold duplex is unstable and falls apart at a fast rate of 4.7 s⁻¹. The short lifetime of the toehold duplex decreases the chances of initiating the branch migration reaction. Thus, the overall rate of the spontaneous TMSD reaction is slow.

Figure 5.

Mechanistic models for spontaneous and Twinkle-catalyzed TMSD reactions

TMSD occurs in two general steps: toehold docking and protector strand displacement. The upper panel shows the kinetic mechanism of spontaneous TMSD, and the lower panel Twinkle-catalyzed TMSD. The rate constants are derived from experiments with the 7-nt toehold 15-bp target dsDNA substrate. Uncatalyzed TMSD reaction initiates when the invader docks on the toe region of the target strand. This diffusive step occurs at a bimolecular rate constant, k_{on toehold} = 0.53 × 10⁶ M⁻¹s⁻¹, resulting in the 7-bp toehold duplex. The 7-bp duplex is unstable and dissociates with a k_{off toehold} = 4.7 s⁻¹. The TMSD reaction occurs with a rate constant k^TMSD of 0.17 × 10⁶ M⁻¹s⁻¹, which is close to the on-rate of the docking step. The instability of the toehold duplex limits the rate of spontaneous TMSD.

Twinkle has multiple high-affinity DNA binding sites. The primary DNA binding site in the central channel has a K_D of ~3 nM and a secondary site of ~25 nM. Like enzyme catalysis, the TMSD reaction initiates after the target and invader DNAs bind to Twinkle’s primary and secondary binding sites. Toehold annealing on Twinkle is 25-fold faster than spontaneous reaction and is limited by the DNA binding rate. The observed rate of toehold annealing on Twinkle k_{on toehold} = 1.34 × 10⁷ M⁻¹s⁻¹ is similar to the DNA binding k_on of 0.96 × 10⁷ M⁻¹s⁻¹. The dissociation of the toehold duplex at k_{off toehold} = 0.1 s⁻¹ is 50-fold slower than the spontaneous rate. The increased lifetime of the toehold duplex on Twinkle likely facilitates the branch migration reaction, which occurs at 2.4 s⁻¹ over the 15-bp duplex region. This mechanism explains the hyperbolic dependence of the TMSD rates on invader strand concentrations.

In Twinkle-catalyzed reactions, the toehold docking step does not limit the TMSD reaction. We show that Twinkle has multiple DNA binding sites, which brings the target DNA and invader in close proximity to catalyze the toehold annealing step like the ‘Jencksian circe’²⁶ effect in enzyme catalysis. In other words, Twinkle increases the effective concentrations and the lifetime of the target and invader DNAs near each other to promote toehold annealing. We estimate that the primary DNA binding site in the ring’s central channel has a K_D of ~3 nM, and the secondary binding site has a K_D of ~27 nM. When target dsDNA is bound to Twinkle’s primary binding site, the invader strand binds to the secondary site with a fast k_on of 1.3 × 10⁷ M⁻¹s⁻¹ and dissociates slowly at a rate of 0.1 s⁻¹. The long lifetime of the joint molecule between invader strand and target dsDNA increases the chances of initiating the branch migration reaction.

In contrast to the spontaneous TMSD, the Twinkle-catalyzed TMSD rates increased hyperbolically with invader concentration, which provided an estimate of the overall branch migration rate of ~2.4 s⁻¹ on the 15-bp stretch of DNA. Because the spontaneous rate did not saturate, we could not estimate the spontaneous branch migration rate. However, a recent single-molecule study measured the spontaneous branch migration rate on 14-bp DNA substrates³¹. Depending on the DNA sequence, the displacement time varied by a factor of 13 from the mean displacement time of 35 ms (~2.69–100 ms). The Twinkle-catalyzed branch migration rate of ~2.4 s⁻¹ on the 15-bp substrate suggests a displacement time of 412 ms (1/2.4 s⁻¹), which is ~4 times slower than the slowest reported rate of spontaneous branch migration. This difference could be due to the different methods used to estimate the branch migration rate, bulk vs. single-molecule, different buffer conditions or sequences. Alternatively, Twinkle may hinder the branch migration step. Additional studies are needed to investigate these possibilities.

After completing the strand exchange reaction, we expect that Twinkle will release the fully duplexed DNA product at 0.21–0.47 s⁻¹ (Figure S6). There is a possibility that Twinkle catalyzes multiple rounds of TMSD reactions, but it remains to be confirmed.

DISCUSSION

Twinkle is a replicative helicase that uses its ATPase activity to unwind DNA and assist the mitochondrial DNA polymerase in strand displacement DNA synthesis^{20, 22, 23}. On its own, Twinkle is poor at unwinding the strands of the dsDNA but has a strong DNA annealing activity^{24, 25}, which we show here, helps catalyze the TMSD reaction. The biological role of Twinkle’s DNA annealing or the TMSD activity is unknown. Given that TMSD is widely and creatively used in logic circuits, nucleic acid amplification, detection, and diagnostics, we explored the mechanism of Twinkle in TMSD catalysis to provide a foundation for its application in DNA nanotechnology. Spontaneous TMSD reactions are typically very slow and take hours to complete, which limits their applications^{7, 8, 14, 15}. We found that Twinkle can increase the TMSD reaction rate by 6000-fold, particularly on short toehold substrates that show very slow spontaneous TMSD. Such rate acceleration improves the response times in DNA amplification and diagnostics and would aid in building complex DNA-based devices. However, further studies are needed to investigate the scope of Twinkle-catalyzed TMSD for biotechnology applications.

We show here that Twinkle-catalyzed TMSD is less sensitive to secondary DNA structures. This property would be advantageous in nanotechnology as it increases the DNA sequence space for designing DNA devices and applications involving natural DNA and RNA sequences. The sensitivity of TMSD rates to single-nucleotide mismatches has been exploited in biotechnological applications to detect single nucleotide mutations and polymorphisms^32–34. Twinkle-catalyzed TMSD remains highly sensitive to single and multiple nucleotide mismatches like spontaneous TMSD reactions. Thus, Twinkle-catalyzed TMSD can be developed into an efficient genotypic platform for detecting mutations and polymorphisms.

TMSD catalysis by Twinkle may not be unique. Many proteins involved in genomic DNA repair and telomere maintenance have demonstrated DNA annealing activity. Examples include RecA, Rad52, Pif1, Dna2, Werner’s, and HelQ helicases^35–40. These proteins may potentially serve as catalysts for the TMSD reactions. One study explored RecA as a catalyst for toehold-mediated catalytic hairpin assembly but found only a moderate rate acceleration of 9-fold but at high micromolar concentrations of the protein⁴¹. In contrast, Twinkle can catalyze TMSD at nanomolar protein concentrations. RecA is not ideal for TMSD because it catalyzes strand exchange without requiring a single-stranded toehold region, which increases nonspecific TMSD. Although Twinkle catalyzes such a reaction, it is ~5000 times slower than the TMSD on toehold-containing substrates. Thus, a DNA annealing protein that depends on the toehold for strand exchange is a more desirable candidate for achieving specificity in TMSD reactions.

Many proteins, like HelQ and Rad52, require single-strand DNA binding proteins to catalyze DNA annealing^{40, 42}, which adds complexity. Twinkle-catalyzed TMSD requires no accessory protein or cofactors like Mg²⁺ and ATP. Thus, Twinkle has very simple requirements. Human Rad52 recombinase forms oligomeric rings^43–45 reminiscent of Twinkle rings and wraps ssDNA around its outer surface to catalyze complementary strand annealing⁴². Rad52 has been shown to catalyze TMSD^{36, 46} but needs further exploration to establish it as a viable catalyst. Adapted CRISPR-Cas systems perform strand displacement in blunt-ended dsDNA substrates lacking a toehold and, thus, demonstrate the potential to catalyze classical TMSD reactions⁴⁷. The enzyme-like ‘circe’ effect observed with Twinkle is also mimicked by DNA-functionalized polymer surfaces, which act as a scaffold to create a ‘circe’ effect to bring DNA substrates together and catalyze TMSD reactions⁴⁸.