Extended Data Fig. 6: A pair of reusable DNA catalysts.
From: Heat-rechargeable computation in DNA logic circuits and neural networks
a,b, Design, simulations, and fluorescence kinetics experiments of two distinct DNA catalysts before and after reset. c, Simulations and experiments of the two DNA catalysts in one test tube. Two reporters with distinct fluorophores and quenchers were used to simultaneously measure the two output signals. Reset was done by heating to 95 °C for 5 minutes and cooling to 20 °C in one minute. Compared to the design shown in Fig. 3c, the hairpin gates here have two nucleotides removed from the open toehold (changed from T to t), leading to slower kinetics that allows for faster annihilation and desired winner-take-all behavior in neural networks. Toehold dissociation rate kd1 = 0.018 and 0.03 per second were used in simulations for GY1 and GY2, respectively, to explain the kinetics observed in the experimental data. The small difference between the two rates can be explained by the sequence-dependent variations in spurious binding between the hairpin loop and the toehold. A 90% and 86% reset success rate were applied in simulations for GY1 and GY2, respectively, to explain the reset performance observed in experimental data. The initial concentrations of the hairpin gates, fuel strands, and reporters were decreased by 10 and 14 nM, respectively, whereas the initial concentrations of the output-reporter complexes and quencher strands were increased accordingly after reset. This non-ideal reset behavior was due to historical reasons, where reporters with a 6-nt toehold near the quencher and fluorophore modifications were used. As shown in Fig. 2c, compared to the design with a mismatched reporter (6-nt loop toehold and 7-nt reporter toehold) that result in a 1-nt bulge in the output-reporter complex, the design with a matching reporter exhibited a roughly 3.6-fold worse reset performance. Nonetheless, the kinetics was sustained after reset: simulations using the same set of rate constants before and after reset quantitatively agreed with the experimental data. When the two gates operated and reset simultaneously in one test tube, the observed kinetics was slightly slower than simulations, indicating the impact of increased toehold occlusion in a larger circuit.
