A poly(thymine)–melamine duplex for the assembly of DNA nanomaterials

Abstract

The diversity of DNA duplex structures is limited by a binary pair of hydrogen-bonded motifs. Here we show that poly(thymine) self-associates into antiparallel, right-handed duplexes in the presence of melamine, a small molecule that presents a triplicate set of the hydrogen-bonding face of adenine. X-ray crystallography shows that in the complex two poly(thymine) strands wrap around a helical column of melamine, which hydrogen bonds to thymine residues on two of its three faces. The mechanical strength of the thymine–melamine–thymine triplet surpasses that of adenine–thymine base pairs, which enables a sensitive detection of melamine at 3 pM. The poly(thymine)–melamine duplex is orthogonal to native DNA base pairing and can undergo strand displacement without the need for overhangs. Its incorporation into two-dimensional grids and hybrid DNA–small-molecule polymers highlights the poly(thymine)–melamine duplex as an additional tool for DNA nanotechnology.

INTRODUCTION

The formation of a DNA duplex, a central molecule for all living systems, is predicted by the Watson-Crick base pairing rules: guanine (G) typically pairs with cytosine (C); adenine (A) with thymine (T). Variation in the arrangement and orientation of hydrogen bonds (H-bonds) leads to alternative DNA structures including triplexes,¹ G-quadruplexes,² and i-motifs.^{3, 4} Organic molecules or polymers can also interact with DNA^{5, 6} via the formation of hydrogen bonds with individual bases.^{7, 8} For example, the McLaughlin group designed a Janus-Wedge DNA triplex by using a polymer wherein each wedge residue formed a H-bonded triplet with a target base pair,⁹ while Bong and coworkers showed recognition between poly(thymine) (poly(T)) and a poly(melamine)-peptide to generate a triplex.¹⁰ Recently, Sleiman and co-workers found that cyanuric acid molecules could mediate the formation of a poly(A) triplex, where three cyanuric acid molecules bind to three adenines to generate a hexameric rosette.¹¹ Metal ions have also been shown to mediate DNA chain interactions.^{12, 13} For example, Hg2+ can bridge two T residues¹⁴ and Ag+ can bridge two C residues, generating metal-DNA duplexes.¹⁵ Cu2+ can bridge between artificial oligonucleotides to generate biologically-inspired, ferromagnetic polymers.¹⁶ These alternative structures present rich chemistry for DNA and provide opportunities for materials applications and new regulatory routes in biological processes.

Herein, we present experimental data to show that melamine (MA), an inexpensive, small molecule, can mediate the association of two poly(T) strands to form antiparallel homoduplexes via T-MA-T H-bonding. This interaction allows sensitive MA detection and can programme DNA tiles to assemble into designed nanostructures. These DNA-small molecule assemblies are inherently more dynamic than Watson-Crick base pairs, enabling tunable stabilities that demonstrate promising roles in developing stimuli-responsive DNA-based assemblies.

RESULTS AND DISCUSSION

Structure of the poly(T)-MA duplex

Melamine (MA) contains three, symmetrical edges that are H-bond complements to thymine. Thus, we hypothesized that MA molecules may interact with poly(T) in water, assembling into duplexes^{17, 18} or triplexes^{19, 20} (Figure 1). After annealing a mixture of poly(T) and MA, we performed native polyacrylamide gel electrophoresis (PAGE, Figure 2a–c). At 25 °C, poly(T) strands (T10, T20, and T30, which are 10, 20, and 30 bases long, respectively) migrate with expected mobilities in the absence of MA. However, in the presence of 10 mM MA, both T20 and T30 migrate more slowly, which suggests that T20 and T30 form multi-stranded complexes with MA. At elevated temperature (37 °C), the multi-stranded complexes dissociate and all poly(T) strands migrate as single strands. T10 retains the same electrophoretic mobility at 25 °C before and after MA addition, due to the low melting temperature of short strands with MA (see below). This length and temperature dependence of mobility is mirrored by the native B-form of DNA.

Figure 1. Potential hydrogen bonding complexes between melamine (MA) and thymine (T).

(a) A T₂MA hetero-triplet and (b) a T₃-MA hetero-quadruplet will lead to a poly(T) homo-duplex and -triplex, respectively. There are three H-bonds between each thymine and MA. Experimental evidence supports the structure in (a).

Figure 2. Native PAGE analysis of the poly(T)-MA interaction.

a-c, Varying the concentration of MA or the temperature (indicated below the gel images). d-e, Monitoring DNA folding of a DNA hairpin, mediated by MA. Sample compositions and the band identities are indicated above and beside the gel images, respectively.

To determine the number and relative orientation of DNA strands in poly(T)-MA conjugates, we prepared a 52-base-long hairpin molecule, H (Figure 2d–e). This strand contains a Watson-Crick hairpin domain in the middle (blue) and two T20-domains (black) at both ends. We hypothesized that H should form a three-stranded complex containing multiple H strands if poly(T)-MA was a triplex, and that H would remain monomeric (for antiparallel duplex) or form a two-stranded complex (for parallel duplex), should poly(T)-MA assemble as a duplex. In the presence of MA, strand H showed a sharp band by PAGE, which ran slightly faster than strand H in the absence of MA (Figure 2d–e). This mobility difference indicated that strand H formed a compact monomer in the presence of MA, suggesting that poly(T)-MA assembled as an antiparallel duplex. Furthermore, when H was combined with T20 and 10 mM MA, PAGE analysis showed the presence of only H-MA and (T20)2-MA, suggesting that antiparallel duplexes are formed exclusively in solution. We attribute the preferential formation of a duplex over a triplex to steric hindrance and strong electric repulsion among the DNA chains.

Circular dichroism (CD) spectra suggested that poly(T)-MA exhibited a chiral secondary structure (Figure S1). While all poly(T)-MA conjugates and poly(T) single strands displayed a positive Cotton effect centered at ~275 nm and a negative Cotton effect at ~250 nm, the poly(T)-MA complexes exhibited a higher peak and a shallower trough than single poly(T) strands. When compared with three major DNA duplex conformations (A-, B-, and Z-forms), poly(T)-MA and B-DNA duplexes exhibit similar features, although their molar ellipticities are distinct: while B-DNA shows similar intensities for both positive (at 275 nm) and negative (at 245 nm) signals, poly(T)-MA displays a high peak and a shallow trough, suggesting the formation of a distinct secondary structure for this poly(T)-MA duplex.

Dynamic light scattering (DLS) confirmed that no large particles were present in a solution of poly(T)-MA, further indicating that poly(T)-MA was a discrete duplex, as opposed to a polymer, in solution (Figure S2). To verify the 2:1 T:MA stoichiometry in the native (i.e., not exclusive to the hairpin) structure, we varied the mole fraction of T50 to MA (where 1 equivalent corresponds to 50 molecules) and analyzed the solutions by CD spectroscopy. A Job plot^{21, 22} indicated that the ratio of T50:MA was 2:1 in solution (Figure S3), reinforcing that MA is complexed at only two of its three H-bonding faces. Interestingly, the third face of the complexed MA could be accessed by other hydrogen-bonding molecules. The addition of an equimolar portion of uric acid to T50–MA generated a unique CD spectrum, with an induced CD signal from uric acid (Figure S4). No substantial increase in particle size was observed by DLS, which suggests that T50–MA maintains its structure on uric acid binding to the third face of MA to generate a duplex that incorporates two small molecules.

X-ray crystallography unambiguously revealed the detailed structure of poly(T)–MA (Figure. 3). Single crystals were obtained from solutions containing T6 and MA at 20 °C by the hanging-drop vapor-diffusion method²³ (see Figure S6 for a crystal containing T5 and MA) and the structure of T6–MA was resolved to a resolution of 2.42Å by molecular replacement (Table S2). Two T6 strands complexed with six MA molecules to form a half-turn, right-handed, antiparallel duplex, which was composed of six consecutive T–MA–T triplets (Figure S7). MA interacted with two T residues by three hydrogen bonds to each T (O2, N3 and O4 of thymine), as shown in Figure 3d. In each T–MA–T segment, all the atoms were approximately coplanar. The geometry and hydrogen-bond lengths within a single T–MA–T triplet resemble those of a C-G-C+ triplet with T in the places of C and C+ and MA in the place of G (Figure S8).¹

Figure 3. Crystallographic study of the T₆-MA complex.

a, Schematic drawing of the T₆-MA duplex. Arrowheads indicate the 3’ ends of the DNA strands. b, An optical image of the T₆-MA crystal. c, Structural model of the T₆-MA duplex. Two DNA strands are shown with golden ribbon backbones and MA moieties are shown as blue/green spheres. d, Details of an MA-mediated T-T base pair. Hydrogen bonds between Ts and MA are represented by blue dashed lines. e, A stereo image showing that the poly(T)-MA duplexes fit within the major grooves of each other in the crystal. f-g, Crystal packing in the T₆-MA crystal: f, viewing along the 3-fold screw axis and g, viewed along the 4-fold screw axis. Melamine molecules are highlighted blue.

Interestingly, the morphology of the T6–MA duplex is similar to that of a DNA triplex, wherein the two T6 strands resemble the orientation of two outer poly(T)/(C) strands and the central MA column resembles the poly(A)/(G) strand1. The morphology is thus distinct from the geometries of A- and B-form DNA duplexes (Table S3).²⁴ With an average twist of 30° for the T6–MA duplex, one turn of an extended poly(T)–MA duplex has a length of 12bp. In the crystal, the duplexes were observed to stack in a tail-to-head fashion, resulting in uninterrupted, infinite columns of discrete duplexes across the crystal lattice (Figure 3f–g). Duplexes in neighbour columns fit into the major grooves of each other (Figure 3e). Such groove fitting leads to a relative 60° rotation between two interacting duplexes.

Tunable thermal stability and cooperativity

We observed that both DNA length and MA concentration significantly and positively impacted the stability of poly(T)-MA duplexes. Upon formation, the poly(T)-MA duplex was accompanied by an absorbance decrease at ~260 nm (Figure S9). This hypochromic effect provided a convenient method to measure the stability of the poly(T)-MA complex (Figure S10). In the presence of 10 mM MA, the melting temperatures (Tm) were 21.5, 36.0, or 39.8 °C for T10-MA, T20-MA, or T30-MA, respectively (Figure S10a). For T30-MA, Tm increased from 19.8 to 33.0 to 39.8 °C when the MA concentration was increased from 1 to 5 to 10 mM (Figure S10b). Without MA, no clear hypochromic transition was observed.

Solution pH was also observed to be a determinant in poly(T)-MA duplex stability (Figure S10c). When pH changed from 8.0 to 5.0, the thermal stability of the T30-MA duplex increased, with the Tm increasing from 39.8 to 53.3 °C. We attribute this pH-induced thermal stability to the protonation of MA (pKa = 5.1)²⁵ under acidic conditions (Figure S11). The resulting electrostatic attraction between the positive charge on protonated MA and the electron-rich keto oxygens in the thymine bases or the negative charges of the DNA backbones might help to stabilize the poly(T)-MA duplex at lower pH. Interestingly, Mg2+, a divalent cation, only moderately enhanced the poly(T)–MA duplex stability (Figure S10d and S12). The cooperative nature of native DNA folding was also observed in poly(T)-MA duplexes. By varying concentrations of MA, CD spectroscopy showed a short induction period in the association of MA with poly(T), followed by exponential growth (Figure S13). These curves suggest a cooperative assembly process,²⁶ wherein elongation of the central MA column is preceded by a nucleation step, as opposed to a stepwise MA association.^{27, 28} This phenomenon scaled with the length of poly(T): longer strands exhibited a shorter induction period. Longer strands also produced a more-intense CD signal (for the same concentration of thymine bases), which suggests that the secondary structure propagates throughout the poly(T)–MA duplex. Consistent with a cooperative folding mechanism, melting curves also displayed hysteresis (Figure S10e).

Dynamic strand displacement

The poly(T)–MA duplex was less stable thermodynamically than poly(A)–poly(T) duplex, but more stable than a poly(A)–poly(T)– poly(T) triplex of the same length (Figure S14). For instance, when poly(A), poly(T) and MA were combined, poly(T) associated with poly(A) to form a conventional poly(A)–poly(T) duplex instead of a poly(T)–MA duplex. Likewise, when the poly(T)–MA duplex and poly(A) were mixed, poly(A) displaced the MA, generating a poly(A)-poly(T) duplex and release MA (Figure S14a). In no experiments was triplex DNA observed (Figure S15a). A thermal denaturation experiment confirmed that a conventional DNA duplex was more stable than poly(T)–MA (Figure S15b): the Tm of a A20T20 duplex was 52.5 °C, nearly 20 °C higher than the Tm of a T20–MA duplex (36.0 °C).

MA forms a two-dimensional (2D) hydrogen-bonding network in the presence of cyanuric acid (CA).²⁹ We hypothesized that strand displacement akin to the addition of poly(A) would also occur upon the addition of CA: upon addition of CA, poly(T)-MA would lead to the precipitation of CA-MA, regenerating free poly(T) in solution. Indeed, when CA (5 mM) was added to a solution of T15-MA (5 mM MA), a white solid precipitated; however, no signal corresponding to free T15 or T15-MA was observed in the CD spectrum of the supernatant (Figure S16a). An infrared spectrum of this precipitate, compared to independently generated CA-MA, suggested that T15 was incorporated into the CA-MA matrix (Figure S16b). The addition of A15 to the solid containing T15-MA-CA led to a CD trace consistent with T15A15 in the supernatant. Poly(T) thus precipitates within the MA-CA matrix upon addition of CA to poly(T)-MA, and can be re-extracted into solution with the addition of poly(A) (Figure S14b).

Mechanical properties

Mechanical properties of the poly(T)-MA duplex were examined by single-molecule mechanical unfolding experiments in a laser tweezers instrument (Figure 4).^{30, 31} We synthesized a DNA construct that contained an unstable DNA hairpin consisting of a 4-bp G-C stem with a tetraloop (Figure 4a). The G-C stem was extended by 20 T-T mismatch pairs (see Fig S17 and SI for sequences), which were expected to be stabilized by melamine (MA) via the formation of T-MA-T complexes. The entire construct was sandwiched between two long double-stranded DNA (dsDNA) handles, which were anchored to two optically trapped beads via affinity linkages (Figure S17, see SI for experimental details). By moving one optically trapped bead away from another with a loading speed of ~5.5 pN/s (10–30 pN range) in a microfluidic chamber filled with 40 mM Tris buffer (supplemented with 12.5 mM magnesium acetate at pH 8.0), we increased the tensile force of the construct until structures were unfolded. We collected force–extension (F–X) curves for each single-molecule construct with an incubation time of 30 seconds between two successive F-X curves. The unfolding forces thus reflect mechanical stabilities of folded structures in the 20-TT DNA construct with and without melamine.

Figure 4. Mechanical properties of poly(T)-MA duplexes.

a, Schematic of the DNA construct that contains 20 T-T mismatched pairs for the T-MA-T formation. b, Schematic of the control DNA construct that contains 20 A-T base pairs in the hairpin stem (20-AT). c, Representative force-extension (F-X) curves of the 20-TT and 20-AT DNA constructs without and with melamine (100 μM). Stretching and relaxing events are represented by colored and black traces, respectively. In the bottom panel, the two sets of F-X curves are offset in x-axis for clarity. d, Rupture force (RF) and the change-in-contour-length (ΔL) histograms of the 20-TT construct in the presence of 100 μM melamine (top) and those of the 20-AT construct in the absence of melamine (bottom). Solid curves depict Gaussian fittings. e, Percentage of folded species in different constructs without and with melamine.

In presence of 1 or 100 µM melamine in the same Tris buffer, the DNA construct revealed two types of unfolding features in the force-extension (F-X) curves. In the first type, the construct showed one unfolding feature, whereas in the second type, two sequential unfolding features were observed (Figure 4c, middle two panels, see Figure S18 for more traces). To confirm that these features were due to the unfolding of hairpins, we analyzed the unfolding force (RF) and change-in-contour-length (ΔL). The observed unfolding force (average 26.2 pN) (Figure 4d, top left) was higher than that for A-T rich DNA hairpins (17.0 pN),30 suggesting that the formation of poly(T)-MA complexes stabilize the hairpin stem. Both RF and change-in-contour-length (ΔL) showed a wide range of values (8–38 pN for RF and 6–20 nm for ΔL) (Figure 4d, top panels). The species with larger ΔL (~20 nm) was close to that expected for a fully folded hairpin (see SI for calculation), suggesting most T-T mismatch pairs formed complexes with MA in this population. The species with smaller ΔL suggests that not all T-T mismatch pairs formed complexes with MA, especially those at the base of the stem. This result is consistent with the reduced probability of hairpin formation at lower MA concentrations (Figure 4e). In the absence of melamine, no unfolding features were observed (Figure 4c, top).

To directly compare the mechanical stability of poly(T)-MA with A-T base pairs, we evaluated the mechanical properties of a control DNA construct in which the 4-bp G-C stem was extended to include 20 A-T base pairs (Figure 4b and Figure S17b). The F-X curves showed single unfolding features (Figure 4c, bottom panel). The unfolding force (~17 pN) was in good agreement with reported values,30 while the change-in-contour-length (ΔL) (18.7 nm) was close to that of expected (see SI). In the presence of 100 µM MA in the same 40 mM Tris buffer (12.5 mM Mg2+ at pH 8.0), F-X curves identical to those without MA were observed (Figure 4c, bottom panel, see Figure S19 for histograms). These control experiments indicated that MA did not bind to A-T pairs while poly(T)-MA complexes were mechanically more stable than the A-T pairs with identical length. Mechanical anisotropy,³² which imparts structures with different mechanical stabilities as a function of the direction of applied force, explains the disparity between the measured mechanical and thermodynamic stabilities (see Figure 15). During unfolding of the T-MA-T pairs, external force is perpendicular to the long axis of the stem. The increased unfolding force likely reflects the need to unzip extra π–π interactions between neighboring T-MA-T pairs. In the X-ray structure (Figure 3), the T-MA-T pairs indeed showed elongated π–π stacking along the long axis of the hairpin stem.

The mechanical unfolding of the T-MA-T pairs can be used in single-molecule mechanochemical sensing ³³ of MA molecules. The MA binding efficiency is low (Figure 4e). We rationalize that when more T residues in a sensing DNA would increase the chance for the MA to bind to the T-T mismatch pairs and improve the limit of detection (LOD). We used rolling circle amplification (RCA)^{34, 35} to introduce multiple poly-thymine (T44) repeats in a single-stranded DNA template (Figure 5a). The RCA product was labelled by biotin and digoxigenin (see SI for details) at its two ends for attachment to the two optically trapped particles (Figure 5a). Binding of the MA to any of the T-T mismatch pairs in the sensing template prepared by the RCA is expected to produce unfolding features similar to those observed in Figure 4c. Indeed, at 10 pM MA, 100% sensing templates showed MA binding within 20 minutes (Figure 5b). At lower concentrations (e.g. 1 pM MA), the percentage of sensing molecules that bind MA decreases in the same time period (only 33% molecules showed the MA binding, see Figure 5c for an exemplary sensing molecule and Figure S20 for other traces). Since LOD was defined as the concentration of ligand at which 50% of the sensing molecules showed analyte binding, ³⁶ 3 pM was estimated as the LOD for the MA detection within 20 minutes. Such an LOD represents 3 orders of magnitude improvement compared to that obtained by the most sensitive bulk methods employing the same T-MA-T strategy within a similar time window.^{37, 38, 39}

Figure 5. Mechanochemical sensing of MA using polyT templates.

a, Schematic of the polyT sensing construct prepared by RCA. Unfolding F-X curves at 10 pM (b) and 1 pM (c) of the Melamine in 10 mM Tris and 100 mM KCL (pH 7.4) buffer. Blue and red traces indicate unfolding curves in the buffer and in the Melamine target, respectively. Time interval of 30 sec was carried out between two F-X traces. Blown-up section in b shows the unfolding of an MA binding event. See Figure S20 for F-X traces of additional molecules.

Nanostructure assembly

To explore the possible regulation effect of this poly(T)–MA DNA duplex on the self-assembly of DNA nanostructures,⁴⁰ we investigated the 1D assembly of a double-crossover (DX) motif (Figure 6a–c).⁴¹ Each side of the DX motif contained a blunt end and a single-stranded T10 overhang. In the absence of MA, DX association was observed in neither PAGE (Supplementary Figs. 21 and 22) nor atomic force microscopy (AFM) imaging (Supplementary Figure 23a,b). On the addition of 10mM MA, association of ssT10 overhangs assembled the DX tiles into 1D arrays, which were visualized by AFM imaging (Figure 6b,c, and Figure S23c,d). To confirm the function of ssT10 overhang association, we prepared a control DX molecule (DX*) that was isomorphous to the aforementioned DX tile without ssT10 overhangs. DX* did not assemble into 1D arrays in the absence or presence of MA; instead, short DNA structures were observed in AFM imaging due to blunt-end stacking (Figure S24).^{42, 43}

Figure 6. Applying polyT-MA to program self-assembly of DNA nanostructures.

a-c, DNA double crossover (DX) assembles into 1D chains. d-f, DNA 3-point-star (3PS) tile assembles into hexagonal 2D arrays. g-i, DNA 4-point-star (4PS) tile assembles into tetragonal 2D arrays. a, d, and g, Schemes of MA-mediated DNA self-assembly. Solid lines (red, green and black) and blue dashed lines represent DNA strands and melamine associated with poly-T strands, respectively. b & c, e & f, and h & i: pairs of atomic force microscopy (AFM) images at different magnifications of the 1D chains, hexagonal 2D arrays, and tetragonal 2D arrays, respectively.

We extended this strategy to regulate the formation of 2D DNA arrays from 3- and 4-point-star motifs (3PS (Figure 6d–f) and 4PS (Figure 6g–i), respectively)^{44, 45}. As with the DX motif, the outside end of each branch of the star motifs contained a blunt end and a single-stranded T10 overhang (Figures 6d,g). In the absence of MA, both motifs remained as individual motifs (Figures S22, S25 and S26). Upon the addition of 10 mM MA, both motifs assembled into 2D arrays (Figure 6e,f,h,i). Although most DNA constructs exhibit specific thermal stabilities, the stabilities of structures that incorporate T–MA–T motifs can be altered by varying the MA concentration and poly(T) length.

CONCLUSIONS

We have shown that MA can guide poly(T) to form an antiparallel, poly(T)–MA duplex in water. This structure has strong mechanical and tunable thermal stabilities during the unzipping of stacked T–MA–T pairs, which allow it to regulate the self-assembly of DNA nanostructures. While poly(T) has previously been observed to interact with positively-charged poly(melamine), this assembly relies on multivalency and electrostatic attraction.¹⁰ In the present study, individual MA molecules are involved instead, and added entropic barriers are overcome to generate the poly(T)-MA duplex.

The ability to modulate the properties of DNA with a small molecule extends our knowledge of DNA chemistry beyond Watson– Crick base pairing and provides an alternative, inexpensive method to generate DNA-based architectures. It is conceivable that other small molecules with the correct hydrogen-bonding capability may interact with other bases in a similar fashion. Understanding how DNA interacts with other organic molecules may provide opportunities for applications in programmed supramolecular chemistry⁴⁶ and biosensing,⁴⁷ to name a few, and may provide answers to why DNA evolved into the universal, genetic materials for life.⁴⁸

METHODS.

Materials.

Melamine was purchased from Sigma-Aldrich. Streptavidin and anti-digoxigenin-coated polystyrene beads were purchased from Spherotech. DNA sequences were designed by the computer program SEQUIN and all the schemes of DNA nanomotifs were drawn with the computer sofware Tiamat. All the oligonucleotides were purchased from IDT and purifed by 10–20% denaturing PAGE. Oligonucleotides from all the tiles are listed below with their designated ratio and concentration during in situ assembly.

DNA sequences (5′–3′):

T5: TTTTT

T6: TTTTTT

T10: TTTTTTTTTT

T15: TTTTTTTTTTTTTTT

T20: TTTTTTTTTTTTTTTTTTTT

T30: TTTTTTTTTTTTTTTTTTTTTTTTTTTTTT

T50: TTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTT TTTTT

T20C10: TTTTTTTTTTTTTTTTTTTTCCCCCCCCCC

H: TTTTTTTTTTTTTTTTTTTTCGCCATTAGGCGTTTTTTTTTTTTTTTTTTTT

(hairpin sequence is underlined)

L2: CCAGGCACCATCGTAGGCTTGCCAGGCACCATCGTAGGCTTG

L3: AGGCACCATCGTAGGTTTCTTGCCAGGCACCATCGTAGGTTTCTTGCCAGGCACCATCGTAGGTTTCTTGCC

L4:

AGGCACCATCGTAGGTTTTCTTGCCAGGCACCATCGTAGGTTTTCTTGCCAGGCACCATCGTAGGTTTTCTTGCCAGGCACCATCGTAGGTTTTCTTGCC

M: TAGCAACCTGCCTGGCAAGCCTACGATGGACACGGTAATGAC

M′: ACTATGCAACCTGCCTGGCAAGCCTACGATGGACACGGTAACG

S: GTCATTACCGTGTGGTTGCTATTTTTTTTTT

S′: CGTTACCGTGTGGTTGCATAGT

Oligo1:CTAGTGCATTAGGAAGCAGCCCAGCTAACCTTTTTTTTTTTTTTTTTTTTCGCCATTAGGCGTTTTTTTTTTTTTTTTTTTTCCAAAGAGCAAGACGTAGCCCAGCGCG

Oligo 2: TTTTCTGGGCTGCTTCCTAATGCA

Oligo 3: GGCCCGCGCTGGGCTACGTCTTGCTTTTT

Oligo 4: CTAGTGCATTAGGAAGCAGCCCAGCTAACCAAAAAAAAAAAA

Oligo 5: AAAAAAAACGCCATTAGGCGTTTT

Oligo 6: TTTTTTTTTTTTTTTTCCAAAGAGCAAGACGTAGCCCAGCGCG.

Motifs:

DX: L2+M+S (1:2:2)

DX*: L2+M′+S′ (1:2:2)

3PS: L3+M+S (1:3:3)

4PS: L4+M+S (1:4:4).

Buffer:

TAE/Mg2+ buffer.

The TAE/Mg2+ buffer consisted of 40mM tris base, 20mM acetic acid, 2mM EDTA, and 12.5mM magnesium acetate; pH was adjusted to 8.0. This was used in the PAGE, thermal denaturation, AFM and mechanical stability experiments.

TA/Mg2+ buffer.

The TA/Mg2+ buffer consisted of 40mM tris base, 7.6mM magnesium chloride; pH was adjusted to 8.0 with acetic acid. This was used in the CD and DLS experiments and those shown in Fig. S3.

Assembly of poly(T)–MA complexes.

The DNA strand (T10, T20, T30 or T50) and MA were mixed in TAE/Mg2+ or TA/Mg2+ buffer and cooled from 50 °C to 4 °C over 2h. DNA strand H was employed in TAE/Mg2+ buffer and heated to 95 °C for 3min and cooled on ice for 10min before being mixed with MA and cooled from 50 °C to 4 °C over 2h.

Native PAGE.

Native PAGE that contained 20% polyacrylamide (19:1 acrylamide:bisacrylamide) was run on a FB-VE10–1 electrophoresis unit (FisherBiotech) at 25 °C (90V, constant voltage). Each lane contained 1 μg of DNA in 20 μl of solution. TAE/Mg2+ buffer was used both as the running buffer and the buffer in the gel. After electrophoresis, the gels were stained with Stains-All (Sigma) and scanned with an HP scanner (Scanjet 4070 Photosmart). The duplex size markers were homemade by hybridizing complementary strands together.

Assembly of poly(T)–MA complexes at different pH.

Before the annealing process, 1.0M HCl was added to the mixture and the solution pH was adjusted to a desired value. Samples were then annealed from 50 °C to 4 °C over 2h.

Thermal denaturation.

The DNA strand and MA were dissolved in TAE/Mg2+ buffer (150 µl) and annealed from 50 °C to 4 °C over 2h. TAE/Mg2+ or TA/Mg2+ buffer (with MA) was used as a blank. The thermal denaturation was monitored by an Agilent Technologies CARY Series UV-Vis-NIR spectrophotometer and the temperature was increased at a rate of 1 °Cmin–1 from 25 to 60 °C and then decreased at 1 °C min–1 from 60 to 4 °C. In all the measurements, the concentration of the thymine nucleotide was kept at 60 µM.

CD.

The DNA strand and MA were dissolved in TAE/Mg2+ buffer (500 µl) and annealed from 50 to 4 °C over 2h. TAE/Mg2+ buffer (MA) was used as a blank. The CD spectra were measured on a Jasco J-1500 CD spectrometer at room temperature (230–350nm range, 50nmmin–1 scan rate, digital integration time 2 s and bandwidth 1nm over three accumulations). Cooperativity of the DNA assemblies. Studies to monitor the cooperativity were carried using a quartz cuvette with a path length of 1mm on a Jasco-810 spectropolarimeter equipped with a xenon lamp, a Peltier temperature control unit (at 5 °C) and a water recirculator. TA/Mg2+ buffer was employed in the samples, and used as a blank for all experiments.

Ultraviolet spectra.

The DNA strand and MA were dissolved in TAE/Mg2+ buffer (150µl) and annealed from 50 to 4 °C over 2h. TAE/Mg2+ buffer (MA) was used as a blank. The ultraviolet spectra were measured on an Agilent Technologies CARY Series UV-Vis-NIR spectrophotometer at room temperature (230 to 350nm range, 50nmmin–1 scan rate).

DLS spectra.

DLS experiments were performed on a Brookhaven photon correlation spectrometer equipped with a BI9000 AT digital correlator. A Compass 315M-150 laser (Coherent Technologies) was used at 532nm as an incident light source. A refrigerated recirculator was used to control the sample temperature. Autocorrelation functions were analysed using the CONTIN algorithm. Borosilicate glass sample vials were purchased from Canadawide Scientific. Samples were prepared in TA/Mg2+ buffer, and the reagents were filtered through 0.45-µm-nylon syringe filters prior to incubation.

Job’s plot procedure.

A series of solutions that contained T50 (1 equiv.) and MA (50 molecules=1 equiv.) were prepared such that the total sum of the poly(T) and MA concentrations remained constant. The mole fraction of MA was varied from 0.1 to 1.0 and the solutions analysed by CD spectroscopy. The corrected ellipticity (mole fraction×ellipticity) at 278nm was plotted against the molar fraction of MA. The intersection of two linear lines was used to deduce the product stoichiometry.

Crystallization.

DNA oligonucleotides were purchased from IDT. Crystals were grown using the hanging drop vapour diffusion method by combining 4μl of the oligonucleotides (0.5μgμl –1) solution that contained 20mM MA with 5μl of growing buffer (pH 7.0) that containing 0.002M magnesium chloride hexahydrate, 0.005M MOPS, 0.2M ammonium sulfate and 0.00005M spermine. Ammonium sulfate (2M) was used as the reservoir buffer. Cubic-shaped crystals with dimensions as large as 150μm×150μm×150μm were obtained after incubation for 4–7days at 20 °C. Crystals were then transferred with cryoloops into a drop of cryoprotectant and allowed to incubate for 1min before being frozen by liquid nitrogen. Cryoprotectant was prepared with one drop of 4μl of 0.2× TAE/Mg2+ buffer with 5μl of growing buffer (pH 7.0) that contained 0.002M magnesium chloride hexahydrate, 0.005M MOPS, 0.2M ammonium sulfate and 0.00005M spermine (and 3% 2-methyl-2,4-pentanediol) and was incubated against 600μl of 2M ammonium sulfate overnight (the volume shrunk to 1/10).

Data collection, processing and structure solution.

X-ray diffraction data were collected on a Rigaku RU-H2R rotating anode X-ray machine at 1.54Å. All the data were indexed, refined, integrated and scaled using HKL2000. The crystal belonged to the space group I23. The structure was solved using the molecular replacement method. A model of six nucleotides as a DNA single strand (triplex-forming strand, TCTCTC, then the sequence was mutated to TTTTTT in COOT by the ‘Calculate— Model/Fit/Refine—Simple Mutate’) cut from a triplex model (1d3x)⁴⁹ was used as the initial search model. The Phaser-MR program in the PHENIX package was used for molecular replacement. The resulting Fo−Fc difference map was used to determine the positions of the MA molecules. MA molecules were added manually one by one using COOT, followed by rounds of refinements using phenix.refine after the addition of each MA molecule. Finally, the poly(T)–MA triplet was treated as one rigid body during refinement. In later rounds, real-space, atom occupancies, B-factor and XYZ coordinates were refined. In further refinements, magnesium atoms and a sulfate ion contained in the crystallization buffer were modelled into large difference peaks in the difference map. Final statistics are given in Table S2. The atomic coordinates and structure factors have been deposited in the Protein Data Bank with accession codes 6WK7. All of the figures were generated with COOT and PyMOL. The DNA structure parameters (twist and rise) were analysed with 3DNA (Web 3DNA 2.0 for the analysis, visualization and modelling of 3D nucleic acid structures, http://web.x3dna.org/).

Assembly of individual DNA tiles.

DNA single strands were mixed in their designated ratios in TAE/Mg2+ buffer to give a final 1 μM DNA solution. The DNA solutions were then incubated at 95 °C for 5min, 65 °C for 30min, 50 °C for 30min, 37 °C for 30min, 22 °C for 30min and then 4 °C for 60min. PAGE was conducted in the presence of 10mM MA, if necessary. The running buffer was the TAE/Mg2+ buffer. Gels were run on a FB-VE10–1 electrophoresis unit (FisherBiotech) at 4 °C (300V, constant voltage) for 2 or 3h. After electrophoresis, the gels were stained with Stains-All dye (Sigma) and scanned.

Assembly of DNA arrays.

DNA single strands were mixed in their designated ratios in TAE/Mg2+ buffer (supplemented with 10mM MA) to give a final 1μM DNA solution. The DNA solutions were then incubated at 95 °C for 5min, 65 °C for 30min, 50 °C for 30min, 37 °C for 30min, 22 °C for 30min and then 4 °C for 60min.

AFM.

DNA solution (20 µl) was deposited onto a freshly cleaved mica surface and incubated for 5min. Without further treatment, AFM images were captured by a MultiMode 8 (Bruker) using ScanAsyst-fluid mode with ScanAsyst-fluid+ probes (Bruker) in the DNA sample buffer. The tip–surface interaction was automatically adjusted to optimize the scan set point.

Synthesis of DNA constructs for mechanochemical experiments.

The DNA constructs that contain a DNA hairpin were sandwiched between two long dsDNA handles (2,028 and 2,690 bp) (Fig. S10). The-biotin labelled 2,028-bp dsDNA handle was prepared by PCR amplification using a pBR322 template (New England Biolab (NEB)) and a 5′-biotinylated primer 5′-GCA TTA GGA AGC AGC CCA GT AGTA GG-3′ (IDT). The PCR product was subsequently digested with XbaI restriction enzyme (NEB). The 2,690-bp dsDNA handle was prepared from pEGFP plasmid by sequential digestion using SacI (NEB) and EagI (NEB) restriction enzymes. This handle was subsequently labelled at the 3′-end (SacI) by digoxigenin using 18μM DIG–dUTP (Roche) and terminal transferase (Fermentas).

To synthesize the 20 T-T DNA construct (Fig. S10a), oligo 1 was annealed with oligo 2 and oligo 3. The final DNA construct was synthesized using T4 DNA ligase (NEB) through sequential two-piece ligations (1:1 molar ratio), starting with the ligation between the 2,028 dsDNA handle and the hairpin-containing fragment, followed by ligation to the 2,690 dsDNA handle via the respective cohesive ends.

To synthesize the 20 A-T DNA construct (Fig. S10b), oligo 4, which contained a part of the hairpin stem (12nucleotides, underlined), was annealed with oligo 2 at 97 °C for 5min and slowly cooled to room temperature for 6h. This fragment was ligated with the 2,028-bp DNA handle by T4 DNA ligase (NEB) and gel purified using a kit (Midsci). On the other side of the DNA construct, oligo 6, which contained another part of the hairpin stem, was annealed with oligo 3. This fragment was ligated with the 2,690-bp handle and gel purified. The final DNA construct was synthesized using T4 DNA ligase (NEB) through three-piece ligation of the 2,028 and 2,690-bp DNA handles and oligo 5, which contained a 4-nucleotide loop with underlined regions that represented the complementary regions of the hairpin stem.

To synthesize the RCA constructs used for mechanical chemical sensing (Fig. 6), we first circularized the 5′-phosphorylated linear template (5′-GTCGTGATA44-CAATCCTG) with a splint (5′-GCA TTA GGA AGC AGC CCA GTA GTA GGA TCA CGA CCA GGA TTG) using T4 DNA ligase at 16 °C for 16h. The splint was removed by a splint remover that contained sequence complementary to the splint. The obtained circular DNA templates were then annealed with the biotinylated primer (/5′-Bio/-GCA TTA GGA AGC AGC CCA GTA GTA GGA TCA CGA CCA GGA TTG). RCAs were carried out using Phi 29 enzyme (NEB) at 37 °C for 10min to obtain linear single-stranded DNA. The resulting linear DNA strands were labelled with digoxigenin using DIG–dUTP (Roche) and terminal transferase (Thermo Fisher) at 37 °C for 3h.

Single-molecule mechanical unfolding experiments.

The experiment was conducted on a homebuilt laser tweezers instrument for the single-molecule experiment.⁵⁰ To perform the mechanochemical experiment, 2μl of a diluted DNA construct (~1ngμl –1) was incubated with 1μl of a 0.1% solution of streptavidin-coated polystyrene beads (diameter, 1.87μm; Spherotech) for about 30min at room temperature (25°C), which immobilized the DNA construct on the bead surface through streptavidin–biotin complex formation. The incubated sample was further diluted to 1ml in TAE/Mg2+ buffer and injected into the top channel of a three-channel microfluidic chamber. Polystyrene beads (2μl, diameter 2.10μm) coated anti-digoxigenin antibodies were also dispersed into the same buffer (1ml) and injected into the bottom channel of the three-channel microfluidic chamber. Two separate laser beams derived from the same diode-pumped solid-state laser (1,064nm, 5W; IPG Photonics) were used to trap two different types of beads. The DNA construct was tethered between the two beads in the middle channel by escorting one of the trapped beads closer to another using a steerable mirror (Madcity Labs Inc.). In the ramping F–X mode, the steerable mirror that controls the streptavidin-coated bead was moved away from the antibody-coated bead with a loading speed of ~5.5pNs–1. The hairpin structure underwent an unfolding transition when the tension inside the tether was gradually increased. This transition was manifested with a sudden change in the end-to-end distance during the process. The single tether was confirmed by a single breakage event or a DNA overstretching plateau at 65pN in the F–X curves. The F–X curve for each tether was recorded in a Labview 8 program (National Instruments Corp.), and data treatment was performed using Matlab (The MathWorks) and Igor (WaveMetrics) programs. The unfolding force was measured directly from the F–X curves, whereas ΔL due to the unfolding was calculated by the two data points that flanked a rupture event using an extensible worm-like chain model (equation S1):

$$\frac{\Delta x}{\Delta L}=1-\frac{1}{2}{\left(\frac{k_{B}T}{FP}\right)}^{\frac{1}{2}}+\frac{F}{S}$$ (S1)

where Δx is the change in extension between the data points of the stretching and relaxing curves at the same force (F), kB is the Boltzmann constant, T is absolute temperature, P is the persistent length (50.8±1.5nm), F is the force and S is the elastic stretch modulus (1,243±63pN). When the molecule was relaxed with the same loading speed, the hairpin refolded, manifested by a sudden change in force or end-to-end distance in the F–X curve.

Single- molecule mechanochemical sensing experiments.

Single-molecule mechanochemical sensing using the poly(T) RCA product was performed in a microfluidic chamber. The buffer only and buffer with target (MA) were flown in the top and bottom microfluidic channels, respectively. Poly(T) RCA products were tethered between a streptavidin-coated bead and an anti-digoxigenincoated bead in a 10 mM Tris buffer (pH 7.4 supplemented with 100 mM KCl). After the two different beads were trapped separately by the two laser foci in the optical tweezers instrument (see previous section), the force ramping experiment was carried out in the buffer channel (top channel). The same molecule was then taken to the MA target channel (bottom channel). The binding of MA was detected by the sudden drop of the force signal in F–X traces.

Calculation of the expected ΔL.

ΔL can be calculated using the equation S2:

$$\Delta L=L-\Delta x=N\times L_{\mathrm{nt}}-\Delta x$$ (S2)

where N is the number of nucleotides contained in the structure, Lnt is the contour length per nucleotide (~0.45nm) and Δx is the end-to-end distance (~2nm, the diameter of dsDNA). According to equation (S2), the theoretical ΔL for a hairpin (52nt=24-bp stems and 4-nucleotide loops) unfolding was calculated as ~21.4nm (52nt×0.45nm/nt–2nm=21.4nm).