Research Article

A cargo-sorting DNA robot

See allHide authors and affiliations

Science  15 Sep 2017:
Vol. 357, Issue 6356, eaan6558
DOI: 10.1126/science.aan6558
  • Conceptual illustration of two DNA robots.

    The robots are collectively performing a cargo-sorting task on a DNA origami surface, transporting fluorescent molecules with different colors from initially unordered locations to separated destinations. Considerable artistic license has been taken.

    ILLUSTRATION: DEMIN LIU (WWW.MOLGRAPHICS.COM)
  • Fig. 1 The cargo-sorting algorithm.

    (A) Schematic diagram of sorting arbitrarily distributed molecules into distinct piles at specified destinations. (B) Flowchart of a simple cargo-sorting algorithm. In the molecular implementation, choices for picking up and dropping off cargos are not always taken as designed—the robot may instead return to random walking with a small probability. Mechanism of the three building blocks for the (C) random walk, (D) cargo pickup, and (E) cargo drop-off. (F) Composability of the three building blocks. Three types of outlines highlight the components used in the three building blocks. (G) Implementation for sorting multiple types of cargos. Squiggled lines indicate short toehold domains and straight lines indicate long branch migration domains in DNA strands, with arrowheads marking their 3′ ends.

  • Fig. 2 The random-walk building block.

    (A) 3D and 2D schematic diagrams of an eight-step long track on a double-layer DNA origami. The lines between adjacent track locations indicate possible moves of the robot: The two types of track strands are in a checkerboard pattern, and for each step, the robot can only move between two distinct types of tracks. Thus, the hexagonal grid is functionally a square grid for the movement of the robot (fig. S4A). (B) Mechanism of protecting the robot from interactions with tracks and activating the robot only at the beginning of an experiment. The activation reaction is biased forward by using trigger strands at 20× higher concentration than the inhibited robot. (C) Mechanism of the robot reaching a goal location. (D) AFM image of the double-layer DNA origami with a track of length 8. (E) Fluorescence kinetics data of random-walk experiments with eight distinct track lengths and a negative control with no track. A 20-fold excess of free-floating robot strands, relative to the origami concentration, was added at the end of the experiments to measure the maximum possible completion level. The two-thirds completion time (T2/3) is plotted against the track length (l). The least-squares fit of a quadratic function is T2/3 = 0.38 + 0.055 × l2. (F) Mass action simulations of the random walk and the negative control. In this model, the robot walks from an arbitrary track location to its neighboring location at kw = 3.5 × 10–3 s–1. The robot is initially inhibited and triggered at kt = 3.2 × 104 M–1 s–1. These two rate constants were determined on the basis of the quadratic fit of the two-thirds completion time versus track length obtained from the experimental data. The negative control was simulated with the robot on one DNA origami interacting with the goal on another at ks = 5 × 102 M–1 s–1.

  • Fig. 3 Demonstration of cargo sorting.

    (A) Mechanism of protecting a goal from interactions with cargos and activating the goal only at the beginning of an experiment. The layout of the two types of tracks in all cargo-sorting systems is shown in fig. S8A. (B) Fluorescence kinetics data of cargo-sorting experiments with two distinct types of cargos. In the initial states, cargo1-F and cargo2-F indicate cargos labeled with fluorophores, and goal1-Q and goal2-Q indicate goals labeled with quenchers. The final states show a random choice of the locations of the robot and an unoccupied goal. (C) AFM images of each type of cargos at their initial locations and delivered to their goal locations, respectively. All images are at the same scale, and the scale bar in the bottom right image is 50 nm.

  • Fig. 4 Exploring the parallelism with mixed populations of DNA origami and with multiple robots on individual DNA origami surfaces.

    (A) Fluorescence kinetics experiments with two mixed populations of DNA origami, each having two types of cargos that can be sorted separately. (B) Stochastic simulation of sorting two types of cargos as a continuous-time Markov chain. Robotx,y indicates a robot at an arbitrary track location (x, y). (x*, y*) is a neighboring location of (x, y). Cargoi and Goali indicate specific types of cargo and goal, respectively. d is the Euclidean distance between (x1, y1) and (x2, y2). dMin is the Euclidean distance between a robot and a cargo or goal at its immediate neighboring location. Because there are 16 base pairs (bp) between the closest staple extension locations and there is a 1-bp deletion every three staple columns for origami twist correction, dMin is calculated as (16 × 3 – 1)/3 bp × 0.34 nm per bp = 5.33 nm. The model allows the robot to pick up a cargo from (or drop off a cargo to) a location that is not its immediate neighbor, if the distance is within the reachable range, but the rate of the reaction decreases quadratically with the distance. Because the total number of base pairs in the double-stranded foot, leg, and cargo (or goal) attacher domains is 41 bp, and the total number of nucleotides in the single-stranded foot and linker domains is 16 nucleotides (nt), the maximum reachable distance is calculated as 41 bp × 0.34 nm per bp + 16 nt × 0.5 nm per nt = 21.94 nm. The rate of random walk is kw = 3.5 × 10–3 s–1, and the rate of closest cargo pickup and drop-off is kc = 100 × kw. We assume that only 80% of the cargos can be successfully delivered to a goal location due to a fraction of origami missing a functional robot. This fraction was determined on the basis of the experimental data shown in Fig. 3B. (C) Fluorescence kinetics experiments with multiple robots collectively performing a single cargo-sorting task.

Supplementary Materials

  • A cargo-sorting DNA robot

    Anupama J. Thubagere, Wei Li, Robert F. Johnson, Zibo Chen, Shayan Doroudi, Yae Lim Lee, Gregory Izatt, Sarah Wittman, Niranjan Srinivas, Damien Woods, Erik Winfree, Lulu Qian

    Materials/Methods, Supplementary Text, Tables, Figures, and/or References

    Download Supplement
    • Materials and Methods
    • Supplementary Text
    • Figs. S1 to S14
    • Tables S1 and S2
    • References
    Correction (22 September 2017): The following revisions and clarifications have been made: (i) The Materials and Methods section has been removed, as it was an exact duplicate of the Materials and Methods in the main text. (ii) A new section, "Additional notes on simulations," has been added. This new section clarifies that in the cargo-sorting experiments, each of the six cargo initial locations could have either a cargo1 or a cargo2. For exposition, a canonical example order of cargos in all initial state diagrams is shown, but it should be understood that other initial cargo arrangements are possible. The implications of this understanding on simulations when compared with experimental data are further discussed. (iii) Missing axis labels in Fig. S3B have been added.
    The original version is accessible here.

Navigate This Article