Information propagation through enzyme-free catalytic templating of DNA dimerization with weak product inhibition

Abstract

Information propagation by sequence-specific, template-catalysed molecular assembly is a key process facilitating life's biochemical complexity, yielding thousands of sequence-defined proteins from only 20 distinct building blocks. However, exploitation of catalytic templating is rare in non-biological contexts, particularly in enzyme-free environments, where even the template-catalysed formation of dimers is challenging. Typically, product inhibition-the tendency of products to bind to templates more strongly than individual monomers-prevents catalytic turnover. Here we present a rationally designed enzyme-free system in which a DNA template catalyses, with weak product inhibition, the production of sequence-specific DNA dimers. We demonstrate selective templating of nine different dimers with high specificity and catalytic turnover, then we show that the products can participate in downstream reactions, and finally that the dimerization can be coupled to covalent bond formation. Most importantly, our mechanism demonstrates a design principle for constructing synthetic molecular templating systems, a first step towards applying this powerful motif in non-biological contexts to construct many complex molecules and materials from a small number of building blocks.

Fig. 1. DNA strand displacement topologies, catalysis mechanism of the template and system design.

a, TMSD. Binding to the toehold (t) domain in the target DNA strand (R) mediates displacement of the incumbent (C) by the invader (I). After displacement, the toehold is cooperatively sequestered in duplex IR. b, HMSD. When I binds to the handhold (h) domain in C, the effective concentration of I increases in the vicinity of R, enhancing displacement. The reversible nature of handhold binding allows IR to detach. c, The DNA-based catalytic templating system. The DNA monomers (M_xL and N_y) can dimerize after binding to a DNA template (T_xy), exploiting first toehold exchange (a TMSD variant) then HMSD. Dimerization between the monomers weakens the interaction with T_xy, allowing M_xN_y to detach and for T_xy to undergo another dimerization cycle. d, The specific-sequence domains of T_xy can trigger the dimerization of a specific M_xL, N_y pair from pools of monomers in solution. The result is a product distribution enriched in M_xN_y dimers with t and h domains (red boxes) complementary to T_xy, propagating the sequence information in the template. Any x,y combination is possible, with the dimerization domain a initially hidden by L, inhibiting any direct reaction in the absence of T_xy. The edges of the M_xL duplex have additional bases—‘clamps’—suppressing any leak reactions. The two mismatched base pairs in the a domain of M_xL ensure that dimerization is thermodynamically favoured. The DNA strands are represented by domains (contiguous sequences of nucleotides considered to hybridize as a unit). The domains are labelled with a lowercase letter; a prime symbol indicates complementarity; for example, a′ binds to a.

Fig. 2. Initial TOF is optimized for toeholds and handholds of moderate length.

a, The experimental setup. A small concentration of template (1 nM) is combined with a larger pool of M₁L and N₃ monomers (10 nM) for a range of primary toehold and handhold lengths. The catalytic turnover of M₁L is reported by an increase in fluorescence signal. b, The example trajectories showing the concentration of reacted M₁L over time, for a range of handhold lengths and a primary toehold of 6 nt. Increasing the handhold length above 9 nt results in a decrease of the M₁L catalytic turnover due to increased product inhibition. These results illustrate how the overall reaction rate is a balance between the displacement and M_xN_y detachment from T_xy. The leak reaction in the absence of template could not be detected. Its magnitude for a monomer concentration of 100 nM is shown for M₁N₃ in Fig. 3c, and for all monomer combinations in Supplementary Fig. 12. The concentration of reacted M₁L is inferred from the fluorescence data as outlined in Supplementary Note 5. c, The initial rate of reaction per unit of template (TOF) for each primary toehold and handhold condition. An optimum is obtained for a system with a primary toehold of 6 nt and a handhold of 9 nt (6t/9h) (1.01 ± 0.03 h⁻¹) followed by condition 6t/8h (0.622 ± 0.009 h⁻¹).

Fig. 3. The optimal design of the HMSD-based catalyst experiences only moderate competitive product inhibition and achieves high turnover.

a, The reacted monomer concentration [M₁L] in a system with 10 nM N₃, 10 nM M₁L, 1 nM T and an initial non-fluorescent pool of products M₁N₃ at a range of concentrations ${\left[M_1{N}_3\right]}_0$. The condition 6t/8h, considered as optimal, is highlighted in red. b, The initial TOF at different ${\left[M_1{N}_3\right]}_0$ conditions for 6t/8h, 6t/9h and 6t/10h, obtained from the kinetics depicted in a. The symbols are centred on the best fit of the TOF to a single trajectory, with the height indicating a 95% confidence interval on that fit. Although 6t/9h has a higher TOF in the absence of [M₁N₃], 6t/8h combines rapid growth with a higher resistance to rate reduction at high ${\left[M_1{N}_3\right]}_0$. c, The turnover of M₁L as inferred from fluorescence data, in experiments with 100 nM M₁L, 100 nM N₃ and variable concentrations of the template ${\left[T_{13}\right]}_0$ (6t/8h). A large proportion of M₁L is observed to react, even for a concentration of ${\left[T_{13}\right]}_0$ 400 times lower than the number of monomers, reaching turnovers above 20 products per template. The template-free leak reactions are essentially negligible (0.32 ± 0.06 M⁻¹ s⁻¹) even compared with the lowest template concentration regimes. d, The initial rates of reactions from c and an additional set of replica experiments (Supplementary Note 7.2), as a function of ${\left[T_{13}\right]}_0$. The symbols are centred on the best fit of the rate to a single trajectory, with the error bars giving a 95% confidence interval on that fit. The red line is the linear fit of the system TOF (3.6 ± 0.3 h⁻¹) to the 11 independent kinetic measurements, the red dashed line is the 95% confidence interval of that fit and the black dashed line is the untemplated rate for monomers at 100 nM = 0.012 ± 0.002 nM h⁻¹.

Fig. 4. Information propagation by sequence-specific catalytic dimerization.

a, The design of monomers to demonstrate accurate information propagation in catalytic dimerization. We consider three types of M_x, differentiated by their primary toehold, and three types of N_y, differentiated by their handholds. Fluorescent labelling, using poly(T) linkers of variable length, allows the identification of all M_xN_y complexes through gel electrophoresis. Templates T_xy are intended to selectively template the formation of M_xN_y from a pool of all six monomer species. b, The fluorescent scan of gel electrophoresis demonstrating sequence-specific templating. Products control: the signal produced by each possible M_xN_y dimer produced by annealing 75 nM of each monomer. Templated reactions, the reaction mixture in which a low concentration of a single T_xy (5 nM) is combined with 100 nM of each M_xL monomer and 75 nM of each N_y monomer. The observed products and signal from unreacted monomers in each well after 40 h of reaction is consistent with the intended M_xN_y production. The false colours include: blue, Alexa 488; green, Alexa 546; red, Alexa 647; cyan, FRET 488/546; yellow, FRET 546/648; purple, FRET 488/648.

Fig. 5. Real-time kinetics of dimer formation for sequence-specific dimerization.

A mixture of six monomers and a specific template T_xy were mixed together to react. We plot inferred concentrations of eight of the nine possible dimers for all templates except T₂₂ (M₂N₂ is not distinguishable from its constituent monomers via fluorescence alone). Note that the PAGE results in Fig. 4b show that M₂N₂ does indeed form as intended. Top: each dimer is represented by the same colour in each plot, indicated by the key. Middle: an estimate of the percentage of correct dimer formation after 24 h for each template.

Fig. 6. Extension of templating to trimerization and covalent bond formation.

a, A schematic of a trimerization process A_x + B_y + C_z → ABC_xyz templated by two dimerization catalysts $T_{A{B}_{xy}}$ and $T_{B{C}_{yz}}$. Each stage is analogous to the catalytic dimerization cycles of Fig. 1. For simplicity, in this diagram, we have assumed template 1 first joins A_x and B_y before template 2 joins C_z to A_xB_y. b, A non-denaturing PAGE analysis of reaction products after mixing either A₁, B₁ and C₁; A₂, B₂ and C₂; or both, with various combinations of templates. The formation of intended products is minimal in the absence of the relevant templates but visible when the templates are present. The bands, including intermediates and unintended products, can be identified by comparing the fluorescence in three channels and migration speed to controls (Supplementary Results 5 and Supplementary Figs. 33–35). c, A schematic illustrating the coupling of HMSD-based dimerization to the formation of a covalently linked dimer. The moieties for a click reaction (Cu-catalysed alkyne azide cycloaddition) are conjugated to monomers M₁ and N₁. d, Denaturing PAGE, which disrupts the duplex formed by HMSD, is used to detect which systems have formed covalent bonds. The monomers are converted to dimers after hybridization of M₁ an N₁ (‘M-alkyl + N-aza’) but not if binding is inhibited by the presence of lock strands (‘blocked M-alkyl + N-aza’). The action of a template T₁₁ (at a ratio of 1:10 with M₁L) during 24 h restores dimerization. Top: the labels indicate the initial concentrations of M₁L and N₁.