Molecular robotic agents that survey molecular landscapes for information retrieval

Abstract

DNA-based artificial motors have allowed the recapitulation of biological functions and the creation of new features. Here, we present a molecular robotic system that surveys molecular environments and reports spatial information in an autonomous and repeated manner. A group of molecular agents, termed 'crawlers', roam around and copy information from DNA-labeled targets, generating records that reflect their trajectories. Based on a mechanism that allows random crawling, we show that our system is capable of counting the number of subunits in example molecular complexes. Our system can also detect multivalent proximities by generating concatenated records from multiple local interactions. We demonstrate this capability by distinguishing colocalization patterns of three proteins inside fixed cells under different conditions. These mechanisms for examining molecular landscapes may serve as a basis towards creating large-scale detailed molecular interaction maps inside the cell with nanoscale resolution.

Fig. 1. Design and mechanism of the molecular crawler system.

a Schematic showing the crawler concept. A crawler roams around a molecular landscape and generates a record that reflects the trajectory. b Anatomy of a probe. See text for details. c Basic mechanism of operation. The top row depicts the unit operation in a single probe. A primer (strand ‘a’) binds the primer-binding domain (a*) and gets elongated by a polymerase along the template. The newly synthesized part competes with the existing strand and can be displaced, exposing a new primer (domain ‘b’); the a–a* pair (16 bp) is stable (T_m ~= 60 °C) at the operating temperature (room temp.) and remains bound. The new primer can initiate a next reaction, as shown in the middle row, with another probe nearby (typically within tens of nanometers; tunable). When three probes are in proximity as in the bottom row, a series of reactions yields an extended crawler spanning across the three probes. Upon binding and extension of a release primer, a record can be released into the solution, which also returns the probes to their original state.

Fig. 2. Basic demonstration.

a Schematic of a three-point track created on DNA origami for basic tests. b Models and averaged (across different molecules) AFM images of before (i, ii) and after (iii, iv) a crawling reaction. Before reaction, probes appear as dots because the sweeping action of an AFM tip pushes around the tethered probes and can only capture faint images of the anchor points. After ~1 h reaction, but without the release primer added, the crawler now connects and holds the three probes together limiting their movements, and thus appear accordingly in the AFM image. Ref., reference. c Gel confirms the correct length of the full record. Lanes were rearranged to allow easy comparison; see Supplementary Fig. 4 for full gel. nt, nucleotides. d Record generation quickly exceeds the added origami amount (marked by the dashed orange line) through the non-destructive and repeated recording mechanism. e The full record sequence was confirmed by Sanger sequencing. f Extended track with ten probes, with repeated arrangements of ‘b–c’, ‘c–d’, and ‘d–b’ probes after the Start (a–b) site, demonstrating the scalability of the crawling reaction. Representative AFM images before (i) and after (ii) a reaction (with no release primer added), along with an averaged (across different molecules) image (iii). Scale bars, 25 nm. Source data are provided as a Source Data file.

Fig. 3. Random crawling mechanism and counting.

a Schematic and detailed diagram of the universal probe, designed to allow the initiation, crawling and release of a crawler at any site, thereby enabling random crawling. b Strand diagram showing a crawler over three universal probes as an example, along with a schematic depicting four kinds of possible records from streptavidin. c Gel image showing the results of ‘counting’ the number of subunits in streptavidin (+SA) and in its absence (−SA). Ladder lane is from a different gel with the same ‘+SA’ sample run in parallel; see Supplementary Fig. 4 for sample lane next to ladder. nt, nucleotides. d Schematic showing artificial molecular complexes with tunable size designed on DNA origami. e Strand diagram showing a crawler over three probes; the middle probe (b–b) is a variable unit. Actual probes used had an alternative architecture (Supplementary Fig. 6), but diagram here was drawn with the basic design for clarity. f Gel image showing the counting results for the artificial complexes. Note the number of bands is the complex size minus one in this design. g Schematic of a square-shaped track for path ‘deconvolution’ tests. The middle probes with repeating primers (b–b) had barcodes embedded to be used to distinguish different paths by high-throughput sequencing. h Gel data showing possible records with three different lengths. Records of length-3 (denoted len3) and of 4 (len4) can be produced from two different paths, respectively. Lanes were rearranged to allow easy comparison; see Supplementary Fig. 4 for full gel lanes or Source Data for full gel. The sample lane is from a Cy5 channel scan. i Sequencing analysis reveals the relative populations of different paths, unveiling the ‘preference’ of the crawlers for shorter paths. j Sequencing read portions replotted across all record lengths for easy comparison. Source data are provided as a Source Data file.

Fig. 4. Multivalent interaction detection.

a Schematic showing the setup and mechanisms of trivalent and monovalent records. Note for β tubulin and EB1, labeling used secondary antibodies but was simplified here for clarity. b Three groups of cell populations and their respective treatments. c Gel image showing two kinds of monovalent records and the trivalent record for the three groups. While there is no noticeable difference in the level of monovalent records between groups, the level of trivalent records drops for group (2) and recovers for group (3). nt, nucleotides. d Plot of normalized gel intensities of trivalent records for multiple experiments (n = 8; dotted lines) along with the average (solid line). Error bars are the standard errors of the means. ****P ≤ 0.0001, ***P ≤ 0.001. e Schematic showing a crawler over three probes and a fluorescent strand used for imaging confirmation. Note we used strand ‘d*−3*’ for increased stability. f Representative fluorescence images showing the characteristic differences between the three groups. DAPI stains nuclei. The straight yellow line is an example trace for intensity profile measurement; see g. Scale bars, 10 μm. g Fluorescence intensity profiles (example trace and its profile shown in yellow in f and g, respectively, for Group (1)) normalized both in the distance and in intensity (average profiles in orange) show peaks near the periphery of cells in Groups (1) and (3) only. Source data are provided as a Source Data file.