Extended DNA Tile Actuators

Abstract

A dynamic linear DNA tile actuator is expanded to three new structures of higher complexity. The original DNA actuator was constructed from a central roller strand which hybridizes with two piston strands by forming two half-crossover junctions. A linear expansion of the actuator is obtained by fusing two actuators of different sequence designs with a third central roller strand. This structure spans 35 nm and its integrity was verified by PAGE analysis. Owing to sequence homology around the crossovers the actuator can obtain 12 different states. The states of the actuator are controlled by a lock strand inserted at one end of the actuator and monitored by Förster resonance energy transfer (FRET) spectroscopy between a fluorophore pair which is located at the other end of the actuator. Two other designs were made where the linear actuator monomer is expanded into two dimensions by forming triangular and quadrilateral actuators. This could be attained by extending the central roller to three or four repeated 32 nucleotide regions that are complementary to the same piston. The triangular and quadrilateral actuators were characterized by PAGE analysis and it was shown that they could be locked in states 0, 5, and 10 and furthermore that they could be switched between the different states by strand displacement reactions.

Introduction

DNA nanotechnology offers the opportunity to form complex nanoscale structures by self-assembly of well-defined DNA structures in sizes ranging from a few sequences to several hundreds.1–3 DNA nanostructures can also be programmed to undergo switching between different well-defined geometrical shapes. An early example was described by Seeman and co-workers where an ion induced transition between B and Z DNA caused a 180 ° rotation of the structure.4 The most common mechanism behind conformational switching in DNA nanostructures is achieved by strand displacement reactions.5, 6 One example is the extension and contraction of hairpin loops by strand displacement,7–10 but also the so-called PX-JX crossover has proven to be an important method to switch between states by strand displacement.11–13 The switching between different states has also been observed for different DNA actuator structures.14–18 Previously reported structures have typically enabled switching between two or three discrete structures.

Recently, our research group described a DNA actuator tile that could be switched between 11 discrete states (Figure 1).19 This actuator tile is a relatively small DNA nanostructure which consists of a central roller strand (R), that hybridizes with two pistons (A and B) and two additional identical lock strands (L) that are specific to each of the 11 states and that fix the actuator in the desired state. The mechanism of switching of the actuator is related to the Holliday junction known from nature where a junction between two dsDNA strands can migrate owing to sequence homology around the junction.20, 21 In the actuator there are 11 nucleotides (nt) of sequence symmetry around the two half-crossover junctions which allows a sliding/rotating motion around the two half-crossovers. The sliding motion arises by rotation of the two helices of the actuator in opposite directions unwinding and winding the two helices respectively, around the half-crossovers.

Figure 1. DNA tile actuator with 11 discrete states. From top to bottom the open actuator, actuator locked in state S0, state S5, and state S10. The blue and orange cubes indicate the position of two FRET dyes.19

By inserting so-called lock strands, the actuator could be specifically locked in any one of the 11 states. The relative position of the piston strands could be monitored by Förster resonance energy transfer (FRET) spectroscopy. Furthermore, the actuator could be switched between states by replacing the lock strands by strand displacement reactions.

Results and Discussion

The extended actuator

Herein, we report on the extension of the scope of the DNA actuator by demonstrating that the actuator structure and mechanism can be applied to more complex devices. First, we have investigated the linear extension of the actuator. The extended actuator design and three-dimensional (3D) structure is shown in Figure 2 and it consists of two actuators each with their unique sequence design. The two actuators are fused by a third bridging roller strand. The bridging roller strand (Roller X) shares sequence homology with each of its neighbors and thus allows sliding motion of the actuator.

Figure 2. The extended linear DNA tile actuator in (a) a 3D illustration and (b) schematic illustration of the system locked in state 0, (c) state 5, and (d) state 11. Black marks regions of the strands with unique base sequences that only hybridize in one way that is static regions, or in the case of black piston termini, locking regions. Colors mark dynamic regions where the roller strands can base pair with each piston region of the same color. Vertical lines only illustrate connections, not nucleotides. Positions of fluorophores for FRET experiments are marked with stars in the schematic illustration and with cubes in the 3D rendering. Green is for Cy3b and magenta is for Cy5.

While the original actuator was designed with identical sequences in the two locking regions to enable use of the same lock strand (L) in both ends, the two ends of the extended actuator have different designs. Here we want to explore if locking the DNA actuator in one end of the structure will induce motion at the other end of the structure. Thus, in our experiments, we input a lock strand in the right hand end of the actuator (Figure 2) and monitor a change in conformation of the structure through a change in the FRET signal at the other end of the actuator.

The sequences were constructed utilizing a Python 2.7 script written for the purpose (see the Supporting Information). The script was designed to minimize unwanted complementarities while also suppressing the sequential occurrence of the same base more than twice in otherwise randomly chosen base sequences. All dynamic base regions (colored regions in Figure 2 b–d) were set to be 10 nt in length and Rollers 1 and 2 were fixed to a total length of 64 nt each. These lengths were consistent with the design of the original actuator.19 Roller X was given a length of 62 nt, and the lock strands a length of 20 nt each with the crossover set to be halfway into the sequence (see the Supporting Information for further information and sequences). The total length of the double stranded region of the extended actuator, including the lock strand is 10 helical turns, corresponding to 35 nm and the minimal distance from the dyes to the lock strand is 7 helical turns corresponding to 18 nm.

To verify that the extended actuator was correctly assembled it was analyzed by non-denaturing polyacrylamide gel electrophoresis (PAGE; Figure 3). The unlocked actuator monomer (Piston 2, 4, and Roller 2) is shown in lane 1 and the locked actuator monomer (S0) is shown in lane 2 and there is a clear mobility shift. The unlocked extended actuator can be observed in lane 6 and the locked version in lane 7. However, owing to the size of the extended actuator the insertion of the lock strand only results in a minor mobility shift. For the data shown in Figure 3, lanes 2, 3, and 7, the lock strands were nicked at the half-crossover and one of the two lock strands functionalized with biotin (lane 4). Upon binding to streptavidin the mobility of the lock strand is significantly retarded. When the lock-biotin-streptavidin strand was used with the actuator monomer (lane 3) and the extended actuator (lane 8) the binding of the lock strand was clearly verified by the decreased mobility of the complexes.

Figure 3. PAGE analysis (7.5 %) of the actuator monomer with Roller 2 and the extended actuator locked with biotin-modified nicked lock strands and streptavidin was added: Lanes from left to right contain: (1) unlocked actuator monomer (Roller 2), (2) Locked monomer, (3) Locked monomer with streptavidin, (4) Lock strand, (5) Lock strand with streptavidin, (6) unlocked extended actuator, (7) Locked extended actuator, (8) Locked extended actuator with streptavidin, and (9) a 100–1000 bp ladder on the far right. PAGE analysis (6 %) of the extended actuator locked with biotin-modified U-lock strands and incubated with streptavidin: lane 10) state S0, 11) unlocked, 12) state 5, 13) unlocked, and 14) state S10. The last lane 15) contains 25–500 bp DNA ladder.

The locking of the extended actuator in states S0, S5, and S10 is shown in Figure 3, lanes 10–14. The actuator was locked with biotin-modified lock strands and incubated with streptavidin before PAGE analysis. In this case the non-nicked U-lock strands were applied. In particular for the locked actuators (lanes 10, 12, and 14), independent of the locked state, we observed some weaker bands with lower mobility, which may arise from dimerization and/or aggregation of the extended DNA actuator.

To ascertain that the different lock strands lead to the organization of the extended actuator in the expected states, FRET experiments were performed for the actuator locked in each of the possible twelve states. The fluorophores Cy5 and Cy3b were conjugated to piston strands 1 and 3 in positions placing them opposite each other when the dimer was locked in state S0 and furthest apart in state S11 (Figure 2). Because the lock strands hybridize with piston strands 2 and 4, the FRET data would thus only reflect a change in states if the locking induced a sliding motion which propagates across the Roller X link between the two actuators.

The observed FRET values are shown in Figure 4 a along with the theoretical FRET data in Figure 4 b, (see the Supporting Information for calculation details). There is generally a good qualitative agreement between the experimentally observed and theoretical FRET data. The amplitude changes of the FRET signal are relatively low compared to the expected values and this may be because only a certain share of the actuators undergoes the expected structural changes. This may, at least in part, be due to aggregation, as some undefined structures and aggregates are seen in lanes 10, 12, and 14 in Figure 3.

Figure 4. a) Relative FRET values for the 11 different states of the actuator using the dyes Cy5 and Cy3B that are each positioned on the Pistons 1 and 3 and located 16 nt from the left hand half crossover of the roller. b) Theoretical FRET values calculated based on our theoretical model, using R₀=7.0 nm.

The FRET data show a non-monotonic dependence when locked from states S0 to S11 with a local minimum in state S6. The smallest FRET value is found for state S11 as expected. The local minimum in the FRET data relates directly to the rotational element in the movement of the fluorophores. A minimum at S6 is achieved in the model when fluorophores are rotated approximately 30 ° away from pointing directly toward each other (see the Supporting Information). In addition, a twist in the actuator plane resulting from torsional tension can contribute to a shift in the minimum.

The increase in signal from S0 to S4 in the observed FRET data may arise from the electrostatic repulsion between the DNA helices. If the model only takes into account the rotation of the fluorophores around the helices and the displacement along two parallel helices, the fluorophore–fluorophore distance is approximately constant in the first four states, S0–S3. However, the theoretical model where the helices are assumed to touch at the half-crossovers (an approximate helix center distance of 2 nm) and be 4 nm further apart at the midway point between half-crossovers results in an increase in the FRET values from state S0 to S2 as illustrated in Figure 4 b. In other words, the repulsion between the DNA double strands leads to a larger distance between the fluorophores in state S0 than in state S2. Our simplified model reproduces the observed FRET values which increase from states S0 to S3. The observed and modeled FRET values show similar trends which are consistent with the designed actuator structure at the molecular level.

The FRET results show that it is indeed possible to regulate the position of the actuator by using different lock strands that mechanically regulate the relative position of a FRET pair located 18 nm from the closest parts of the lock strands.

To further explore the scope of the actuator principle,19 we have investigated the possibility of expanding the structural features of the linear DNA tile actuator into triangular and quadrilateral actuators. The designs of the two new actuators are shown as 3D models in Figure 5. For simplicity and to form the actuators with a minimum of sequences, the two actuators were designed with maximal symmetry applying piston strands that are identical in all edges of the structure. The triangular actuator was designed with a roller strand of 96 nt containing three repeated regions of 32 nt complementary to the piston, and the quadrilateral actuator was formed with a roller strand of 128 nt containing four repeated regions of 32 nt complementary to the piston. In both cases the piston strands are 64 nt long and for each half-crossover there is 11 homologous bases around the junction providing the structures with their dynamic rolling ability.

Figure 5. 3D models of the triangular and quadrilateral actuators. In both models the roller sequence is shown in yellow, the identical piston strands are shown in blue, and the lock strands are illustrated in green. Symmetry around the half-crossovers allows the actuator to slide between states, defined in the absence of the lock strands and the lock strands enable the locking of the two actuators in 11 different positions.

To lock the actuators in specific positions and to demonstrate the ability to shift the actuator between states we have used lock strands (L) with a toehold for positions S0, S5, and S10. Each lock strand can hybridize with a part of both pistons extending from the half-crossovers. The toehold of the lock strand makes it possible to remove the lock strand again by applying a strand complementary to the lock strand (LC). This makes it possible to switch between states. The locked triangular and quadrilateral actuators are designed so that the helices are organized in one plane, which implies that the helices of the pistons are forced to bend at the junctions to allow for the half-crossover of the lock strand that locks the actuator. As shown below the triangular and quadrilateral actuators are indeed formed, however it should be noted it is not clear if the actual structures are planar as shown in Figure 5 or if they are twisted/bended owing to the intrinsic strain.

First, we investigated the formation of the triangular actuator. A number of titration experiments were performed to investigate the exact stoichiometry. Second, the magnesium concentration was varied from 1 mM to 30 mM (see Figure S4 in the Supporting Information). At low magnesium concentrations significant amounts of the complex with only two pistons (Piston/Roller=2:1) were observed. At magnesium concentrations of 12.5 mM and above, the complex with three pistons (Piston/Roller=3:1) was abundant and 12.5 mM was chosen to avoid unnecessarily high salt concentrations in PAGE analysis, as well as aggregation problems.

To test the dynamic locking and unlocking of the actuator, the initial piston and roller 3:1 complex (PR) was first formed separately, followed by consecutive addition of L0. The mixture was then heated to 37 °C for 20 minutes while shaken, resulting in the formation of the locked complex, S0, with a lower electrophoretic mobility. Subsequently, the complementary LC0 was added, unlocking the system and regenerating the PR complex. The unlocked actuator is locked in S5 by addition of L5, and subsequently it is unlocked by addition of LC5. Finally, L10 generated the locked complex S10 and LC10 regenerated PR. Initially, the experiments were performed with lock sequences containing a 10 nt toehold extension, however, this resulted in problems with aggregation of the structures and low quality of the PAGE. As an alternative, we used lock strands with only 5 nt toeholds in the 3’ end, which resulted in successful formation of the locked and unlocked complexes. The resulting assembly of the triangular actuator is shown in Figure 6. Lanes 2 and 3 contain the piston and the roller sequences, lane 4, the PR complex. The band just below the formed three-way complex is a complex with only two pistons. Additional unhybridized pistons were observed in the bottom of the lane. Lanes 5–10 demonstrate the locking and unlocking of the actuator progressing with the same sample from state 0 to state 10. In lanes 5–10 some additional bands were observed, and as indicated in the right of the figure, these are believed to be excess piston and piston hybridized with excess lock strands.

Figure 6. Non-denaturing PAGE (7.5 %) of the triangular actuator. Lock strands with 5 nt toehold regions were used, resulting in the formation of all the locked and unlocked complexes. Lanes from left to right contain: (1) 25–500 bp ladder, (2) piston, (3) roller, (4) unlocked PR, (5) Locked PR complex S0, (6) unlocked PR, (7) Locked S5, (8) unlocked PR, (9) Locked S10, and finally (10) unlocked PR.

The locked and unlocked complexes were formed successfully as shown in Figure 6, however, for some of the locked structures, various aggregation products were observed. This may result from lock strands inter-linking the actuator complexes. Indeed, the inherent bending of the piston helices at the roller half-crossovers is caused by the lock strand half-crossover, and this strain may favor the sharing of the lock strands with another actuator, causing dimerization and polymerization of the actuators. To test this hypothesis, we nicked the lock strand at the half-crossover and thus used two half lock strands to lock the actuator. PAGE analysis of the actuator formed with the nicked lock strands showed significantly less aggregation (Figure S5). However, the actuator is not securely locked when the lock strand is nicked due to the homology around the half-crossover. In this case, the crossover can be considered as a traditional Holliday junction in which one sequence is nicked. The position of the actuator that places the nick at the junction is probably the most stable; however, we have no evidence to confirm this.

The quadrilateral actuator was formed by extending the roller sequence to 128 nt, introducing an additional 32 nt unit with complementarity to the piston. The symmetric four-way junction was expected to be formed from four pistons and one roller stand. As for the triangular actuator, the quadrilateral actuator has sequence homology around the half-crossovers allowing the actuator to slide between states defined by the lock strands. The dynamics of the quadrilateral actuator was demonstrated using lock strands with 5 nt toehold regions (Figure 7 a). Formation of the unlocked actuator complex and insertion and removal of lock strands was demonstrated by PAGE. The non-denaturing PAGE showed the formation of all the locked and unlocked complexes, however, it was accompanied by a significant degree of aggregation and partly folded actuators. Thus, it was also attempted to form the actuator with the nicked lock strands for the quadrilateral actuator and as shown in Figure 7 b the PAGE analysis revealed much cleaner products through the insertion and removal of lock strands.

Figure 7. Non-denaturing PAGE (7.5 %) of the quadrilateral actuator (a) formed with full length locks with 5 nt toehold regions (b) formed with the nicked lock strands each with 5 nt long toehold regions. Lanes from left to right contain: (1) 25–500 bp ladder, (2) piston, (3) roller, (4) unlocked PR, (5) Locked PR complex S0, (6) unlocked PR, (7) Locked S5, (8) unlocked PR, (9) Locked S10, and finally (10) unlocked PR.

Conclusion

We have demonstrated that the previously reported DNA tile based actuator19 could be extended to larger and more complex structures, and these structures maintain their ability to be locked in the different states. The extended actuator is two to three times larger than the original structure, and is locked by a single lock strand. It was clearly demonstrated that the insertion of a short lock strand in on end of the linear structure, could through mechanical motion of the actuator control the relative positioning of a FRET pair located at least 18 nm from the lock strand. For the triangular and quadrilateral actuators it was not possible to characterize the relative positioning of the lock strands by FRET, however it was demonstrated by PAGE that specific lock strands would bind to the structure and to do this the actuator has to adjust to the actual state. Furthermore, the consecutive replacement and insertion of lock strands in positions S0, S5, and S10 could be performed by strands displacement reactions. Owing to the 2D design of the triangular and quadrilateral actuators, there may be some strain in the state locked by the U-Lock strands, presumably leading to partial dimerization and polymerization of the actuators by sharing lock strands, whereas the use of nicked lock strands resulted in less by products. In future studies we will investigate if non-symmetric versions of the triangular and quadrilateral actuators have potential as signal splitting devises, where an input in one site leads to two or three outputs.

Compared to other dynamic DNA nanostructures, the actuator devices can obtain a much higher number of different states as the 12 individual states reported here. The extension of the actuator both in one and two dimensions, demonstrate the great potential of the actuator method for controlling motion at the nanoscale. In particular the extended actuator is proof-of-concept of a nanoscale mechanical device that can be assembled in modular fashion extending and synchronizing the motion in the assembled modules. Future studies will show if this can be extended to polymeric actuator structures.

Experimental Section

Materials and general methods

All oligonucleotides were purchased from either TAG Copenhagen A/S or DNA Technology A/S in Denmark. The sequences are listed in the Supporting Information. Purification by HPLC was done by the company directly after synthesis. The monoreactive Cy3b-NHS and Cy5-NHS dyes were purchased from GE Healthcare. All other chemicals were purchased from Sigma–Aldrich Co. and used without further purification, unless otherwise stated. All utilized organic solvents were HPLC grade and all water was purified on a Milli-Q Biocell system by Millipore. All purifications on RP-HPLC were performed on a Hewlett Packard Agilent instrument with autosampler and fraction collector and fitted with a Phenomenex Clarity 3 μm Oligo-RP 50×4.60 mm column. The eluent was a mixture of acetonitrile and triethylammonium acetate (TEAA; 0.1 M, pH 7.0) ramping from 0 to 20 % acetonitrile in the first 3 minutes, then to 30 % in the next 13 minutes. Yields of products conjugated to DNA were determined by UV absorption (λ=260 nm) in aqueous solution under the assumption that the effect of the DNA on the extinction coefficient at the given wavelength are insignificant. The UV absorption measurements were conducted on a NanoDrop ND-1000 spectrophotometer.

Conjugation of fluorophores to DNA

The monoreactive N-hydroxysuccinimide (NHS) ester of the dye to be conjugated (0.1 mg) was dissolved in N,N-dimethylformamide (5 μL) and added to an aqueous solution of the C6-amino-modified DNA (2 nmol, 5 μL) along with acetonitrile (5 μL) and triethylamine (0.2 μL). The reaction mixture was covered in aluminum foil and agitated for 2 h. Subsequently, the DNA was precipitated from the reaction mixture by addition of sodium acetate (1.5 μL, 3 M, pH 5.2) and ethanol (45 μL, 96 %, cold) and cooling on dry ice for 45 min. The precipitated DNA was isolated by centrifugation (1 h, 14 000 rpm) and washed with ethanol (45 μL, 70 %, cold). The isolated DNA was dissolved in TEAA buffer (0.1 M, pH 7) and purified by RP-HPLC. Conjugation with Cy3b proceeded with very poor yield (69 pmol, 3.5 %). Conjugation with Cy5 showed close to full conversion. Only products purified by HPLC were used in the following experiments.

Assembly of the extended DNA actuator

A polymerase-chain-reaction (PCR) tube was charged with piston and roller strands in stoichiometric amounts. If a locked complex was desired, the appropriate lock strands were added in five equivalents. TAE/Mg²⁺ buffer and water were added until a final buffer concentration of 1x(40 mM tris(hydroxymethyl)aminomethane (Tris)-HCl, 20 mM acetic acid, 2 mM ethylenediaminetetraacetic acid (EDTA), and 12.5 mM magnesium acetate, pH 8.0) and a DNA concentration of 2 μM for each of the piston and roller strands. The samples were mixed thoroughly and annealed on an Eppendorf Mastercycler Thermal Cycler (80 °C→55 °C in 30 min and then 55 °C→20 °C by 1 °C min⁻¹).

Assembly of the triangular DNA actuator

Formation of the triangular actuator was achieved by mixing in exact stoichiometric ratio P/R_T/L=1:3.5:3.5 and P/R_T/L_A/L_B=1:3.5:3.5:3.5 in a 1xTAE/Mg²⁺ buffer. The samples were mixed thoroughly and annealed on an Eppendorf Mastercycler Thermal Cycler (80 °C→55 °C in 30 min and then 55 °C→20 °C by 1 °C min⁻¹).

Quadrilateral DNA actuator

Formation of the quadrilateral actuator was achieved by mixing in exact stoichiometric ratio P/R_Q/L=1:5:5 and P/R_Q/L_A/L_B=1:5:5:5 in TAE/Mg²⁺ buffer. The samples were mixed thoroughly and annealed on an Eppendorf Mastercycler Thermal Cycler (80 °C→55 °C in 30 min and then 55 °C→20 °C by 1 °C min⁻¹).

Polyacrylamide gel electrophoresis (PAGE)

Gels contained either 7.5 % or 6 % polyacrylamide (from 19:1 acrylamide/bisacrylamide), TBE buffer (89 mM Tris base, 89 mM boric acid, 2 mM EDTA, pH 7.5), magnesium chloride (12.5 mM) and N,N,N′,N′-tetramethylethylene-1,2-diamine (10 μL) and ammonium persulfate (10 w/v % in water, 100 μL). Prior to loading of samples onto the gel, the samples were added an equal volume of loading dye (40 % or 60 % glycerol in TAE/Mg and trace amounts of bromophenol blue tracking dye). Electrophoresis was performed at 70 V, 4 °C in TBE buffer (Tris-HCl, boric acid, EDTA) running for 1.5 or 2 h, using Bio-Rad PowerPac 1000 power supply. Subsequent staining was performed with ethidium bromide (0.5 μg mL⁻¹) in water (20 min.) and followed by imaging on a Syngene InGenious GVM20 Transilluminator.

Streptavidin binding

Streptavidin was bound to the locked actuator complex in some experiments to increase retention in PAGE analysis compared to the unlocked complex. In these experiments, biotin modified lock strands were used. In the case of “nicked” locks, only the strand designed to hybridize with Piston 2 was biotin modified. The locked complexes were assembled using the general procedure and subsequently adding five equivalents (20 μM in TAE/Mg buffer) streptavidin relative to the biotin modified lock strand. The samples were kept at 4 °C for 1 h before subsequent PAGE analysis.

FRET measurements

Samples of the extended actuator intended for FRET analysis were assembled following the general procedure with Cy3b-modified Piston 1, Cy5-modified Piston 3, and unmodified Pistons 2 and 4 and Rollers 1, 2, and X. For locking, unmodified U-lock strands were included in excess (5 equivalents). A sample locked in each of the states 0 through 11 was produced, each containing 2 pmol complex, and following hybridization, TAE/Mg buffer was added to a total volume of 70 μL. The FRET measurements were performed on a scanning spectrofluorometer (Fluoro-Max-3, HORIBA Jobin Yvon Inc.) in a quartz cuvette, which was washed twice with TAE/Mg between different samples. Excitation was performed at 530 nm (Cy3b) and at 600 nm (Cy5), and spectra of the locked actuator dimers were recorded with 0.5 s integration time and 1 nm wavelength intervals. Spectral intensities were corrected for the excitation lamp intensity fluctuations and for instrumental change of detection efficiency as a function of wavelength. Temperature was set to 25 °C during measurements. FRET values were calculated using the ratio A method22 as , where I_AD is the acceptor peak fluorescence intensity after donor excitation from which contribution from donor fluorescence was subtracted and I_AA is the acceptor peak fluorescence intensity after acceptor excitation and ε_AA and ε_DA have values of 0,07 and 1,03 respectively.

Acknowledgements

This study was supported by grants from the Danish National Research Foundation to the CDNA Center, the Danish Council for Independent Research′s research career program Sapere Aude, and the Lundbeck Foundation.