DNA origami is a robust assembly technique that folds a single-stranded DNA template into a target structure by annealing it with hundreds of short ‘staple’ strands1,2,3,4. Its guiding design principle is that the target structure is the single most stable configuration5. The folding transition is cooperative4,6,7 and, as in the case of proteins, is governed by information encoded in the polymer sequence8,9,10,11. A typical origami folds primarily into the desired shape, but misfolded structures can kinetically trap the system and reduce the yield2. Although adjusting assembly conditions2,12 or following empirical design rules12,13 can improve yield, well-folded origami often need to be separated from misfolded structures2,3,14,15,16. The problem could in principle be avoided if assembly pathway and kinetics were fully understood and then rationally optimized. To this end, here we present a DNA origami system with the unusual property of being able to form a small set of distinguishable and well-folded shapes that represent discrete and approximately degenerate energy minima in a vast folding landscape, thus allowing us to probe the assembly process. The obtained high yield of well-folded origami structures confirms the existence of efficient folding pathways, while the shape distribution provides information about individual trajectories through the folding landscape. We find that, similarly to protein folding, the assembly of DNA origami is highly cooperative; that reversible bond formation is important in recovering from transient misfoldings; and that the early formation of long-range connections can very effectively enforce particular folds. We use these insights to inform the design of the system so as to steer assembly towards desired structures. Expanding the rational design process to include the assembly pathway should thus enable more reproducible synthesis, particularly when targeting more complex structures. We anticipate that this expansion will be essential if DNA origami is to continue its rapid development1,2,3,17,18,19 and become a reliable manufacturing technology20.
Your institute does not have access to this article
Open Access articles citing this article.
Nature Communications Open Access 28 April 2022
Nature Communications Open Access 06 December 2019
Nature Communications Open Access 05 March 2019
Subscribe to Journal
Get full journal access for 1 year
only $3.90 per issue
All prices are NET prices.
VAT will be added later in the checkout.
Tax calculation will be finalised during checkout.
Get time limited or full article access on ReadCube.
All prices are NET prices.
Rothemund, P. W. K. Folding DNA to create nanoscale shapes and patterns. Nature 440, 297–302 (2006)
Douglas, S. M. et al. Self-assembly of DNA into nanoscale three-dimensional shapes. Nature 459, 414–418 (2009)
Dietz, H., Douglas, S. M. & Shih, W. M. Folding DNA into twisted and curved nanoscale shapes. Science 325, 725–730 (2009)
Sobczak, J. P. J., Martin, T. G., Gerling, T. & Dietz, H. Rapid folding of DNA into nanoscale shapes at constant temperature. Science 338, 1458–1461 (2012)
Seeman, N. C. DNA in a material world. Nature 421, 427–431 (2003)
Arbona, J.-M., Aimé, J.-P. & Elezgaray, J. Cooperativity in the annealing of DNA origamis. J. Chem. Phys. 138, 015105 (2013)
Song, J. et al. Direct visualization of transient thermal response of a DNA origami. J. Am. Chem. Soc. 134, 9844–9847 (2012)
Levinthal, C. How to fold graciously. Mössbauer spectroscopy in biological systems. Univ. Illinois Bull. 67, 22–24 (1969)
Anfinsen, C. B. Principles that govern the folding of protein chains. Science 181, 223–230 (1973)
Dobson, C. M. Protein folding and misfolding. Nature 426, 884–890 (2003)
Baker, D. A surprising simplicity to protein folding. Nature 405, 39–42 (2000)
Martin, T. G. & Dietz, H. Magnesium-free self-assembly of multi-layer DNA objects. Nature Commun. 3, 1103 (2012)
Ke, Y., Bellot, G., Voigt, N. V., Fradkov, E. & Shih, W. M. Two design strategies for enhancement of multilayer-DNA-origami folding: underwinding for specific intercalator rescue and staple-break positioning. Chem. Sci. 3, 2587–2597 (2012)
Castro, C. E. et al. A primer to scaffolded DNA origami. Nature Methods 8, 221–229 (2011)
Douglas, S. M., Bachalet, I. & Church, G. M. A logic-gated nanorobot for targeted transport of molecular payloads. Science 335, 831–834 (2012)
Perrault, S. D. & Shih, W. M. Virus-inspired membrane encapsulation of DNA nanostructures to achieve in vivo stability. ACS Nano 8, 5132–5140 (2014)
Han, D. et al. DNA gridiron nanostructures based on four-arm junctions. Science 339, 1412–1415 (2013)
Ke, Y., Ong, L. L., Shih, W. M. & Yin, P. Three-dimensional structures self-assembled from DNA bricks. Science 338, 1177–1183 (2012)
He, Y. et al. Hierarchical self-assembly of DNA into symmetric supramolecular polyhedra. Nature 452, 198–201 (2008)
Kershner, R. J. et al. Placement and orientation of individual DNA shapes on lithographically patterned surfaces. Nature Nanotechnol. 4, 557–561 (2009)
Wolynes, P. G., Onuchic, J. N. & Thirumalai, D. Navigating the folding routes. Science 267, 1619–1620 (1995)
Onuchic, J. N., Wolynes, P. G., Lutheyschulten, Z. & Socci, N. D. Towards an outline of the topography of a realistic protein-folding funnel. Proc. Natl Acad. Sci. USA 92, 3626–3630 (1995)
Fink, T. M. A. & Ball, R. C. How many conformations can a protein remember? Phys. Rev. Lett. 87, 198103 (2001)
Jacobsen, J. L. & Kondev, J. Field theory of compact polymers on a square lattice. Nucl. Phys. B 532, 635–688 (1998)
Zhang, D. Y. & Winfree, E. Control of DNA strand displacement kinetics using toehold exchange. J. Am. Chem. Soc. 131, 17303–17314 (2009)
Zhang, P. H. et al. Engineering BspQI nicking enzymes and application of N.BspQI in DNA labeling and production of single-strand DNA. Protein Expr. Purif. 69, 226–234 (2010)
Douglas, S. M. et al. Rapid prototyping of 3D DNA-origami shapes with caDNAno. Nucleic Acids Res. 37, 5001–5006 (2009)
Wickham, S. F. J. et al. Direct observation of stepwise movement of a synthetic molecular transporter. Nature Nanotechnol. 6, 166–169 (2011)
Gillespie, D. T. Exact stochastic simulation of coupled chemical reactions. J. Phys. Chem. 81, 2340–2361 (1977)
SantaLucia, J., Jr & Hicks, D. The thermodynamics of DNA structural motifs. Annu. Rev. Biophys 33, 415–440 (2004)
Jacobson, H. & Stockmayer, W. H. Intramolecular reaction in polycondensations. I. The theory of linear systems. J. Chem. Phys. 18, 1600–1606 (1950)
Ouldridge, T. E., Louis, A. A. & Doye, J. P. K. Extracting bulk properties of self-assembling systems from small simulations. J. Phys. Condens. Matter 22, 104102 (2010)
Rayleigh On the problem of random vibrations, and of random flights in one, two or three dimensions. Phil. Mag. 37, 321–347 (1919)
Chandrasekhar, S. Stochastic problems in physics and astronomy. Rev. Mod. Phys. 15, 1–89 (1943)
Dijkstra, E. A note on two problems in connexion with graphs. Numer. Math. 1, 269–271 (1959)
Morrison, L. E. & Stols, L. M. Sensitive fluorescence-based thermodynamic and kinetic measurements of DNA hybridization in solution. Biochemistry 32, 3095–3104 (1993)
Gao, Y., Wolf, L. K. & Georgiadis, R. M. Secondary structure effects on DNA hybridization kinetics: a solution versus surface comparison. Nucleic Acids Res. 34, 3370–3377 (2006)
SantaLucia, J., Jr A unified view of polymer, dumbbell, and oligonucleotide DNA nearest-neighbor thermodynamics. Proc. Natl Acad. Sci. USA 95, 1460–1465 (1998)
Peryet, N. Prediction of Nucleic Acid Hybridisation: Parameters and Algorithms. PhD thesis, Wayne State Univ. (2000)
Owczarzy, R., Moreira, B. G., You, Y., Behlke, M. A. & Walder, J. A Predicting stability of DNA duplexes in solutions containing magnesium and monovalent cations. Biochemistry 47, 5336–5353 (2008)
Smith, S. B., Cui, Y. & Bustamante, C. Overstretching B-DNA: the elastic response of individual double-stranded and single-stranded DNA molecules. Science 271, 795–799 (1996)
Rivetti, C., Walker, C. & Bustamante, C. Polymer chain statistics and conformational analysis of DNA molecules with bends or sections of different flexibility. J. Mol. Biol. 280, 41–59 (1998)
Mills, J. B., Vacano, E. & Hagerman, P. J. Flexibility of single-stranded DNA: use of gapped duplex helices to determine the persistence lengths of poly(dT) and poly(dA). J. Mol. Biol. 285, 245–257 (1999)
Murphy, M. C., Rasnik, I., Chang, W., Lohman, T. M. & Ha, T. Probing single-stranded DNA conformational flexibility using fluorescence spectroscopy. Biophys. J. 86, 2530–2537 (2004)
Chen, H. et al. Ionic strength-dependent persistence lengths of single-stranded RNA and DNA. Proc. Natl Acad. Sci. USA 109, 799–804 (2012)
Saenger, W. Principles of Nucleic Acid Structure (Springer, 1984)
Hagerman, P. J. Flexibility of DNA. Annu. Rev. Biophys. Biophys. Chem. 17, 265–286 (1988)
We thank K.V. Gothelf, M. Dong, A.L.B. Kodal, S. Helmig and S. Zhang (Department of Chemistry and Interdisciplinary Nanoscience Centre iNano, Aarhus, Denmark) for assistance with AFM imaging. This research was supported by Engineering and Physical Sciences Research Council grants EP/G037930/1 and EP/P504287/1, a Human Frontier Science Program grant RGP0030/2013, a Microsoft Research PhD Scholarship (F.D.), the ERC Advanced Grant VERIWARE (F.D. and M.K.) and a Royal Society–Wolfson Research Merit Award (A.J.T.).
The authors declare no competing financial interests.
Extended data figures and tables
a, Well-folded, planar states can be considered as two adjacent monomer tiles linked by a single reciprocal template crossing at any of the locations marked with a triangle and numbered (centre). This gives a set of 6 unique shapes, as indicated (periphery). b, With the exception noted below, there are four ways to make each of these shapes, distinguished by the nucleotide sequence at the template crossing but not resolved by AFM imaging. In the example shown, crossings made at positions 5 and 8 correspond to the fold 4:1, and crossings made at 17 and 20 correspond to fold 1:4, as indicated in the circle diagrams (left). All give the same shape with fractional short edge offset w/W of 3/6 (right). The exception is that there are only two variants of state 5:0i as configurations formed by linking tiles at positions 1 and 24 are not distinguishable, nor are links at positions 12 and 13. c, Detailed view of the connection between monomer tiles in this case, for which the long-edge offset is not precisely defined (it can range from 0 to 2/7 depending on the conformation of the long edge staple). For the purpose of predicting geometry for model configurations, we take an average value of l/L = 1/7. The set of 22 well-folded, planar states thus consists of two folds for the shape shown in c and four folds for each of the other five shapes.
a, b, The set of legal folds permitted by the model consists of the 22 planar folds defined in Extended Data Fig. 1 and an additional 52 non-planar folds, four for each of the 13 shapes shown here in a, b. Shapes in a are formed by allowing three reciprocal crossings between two tiles, those in b are formed by allowing 5 reciprocal crossings. These non-planar folds form only rarely in simulation. c, An example of a misfolded shape: the part-folded domains are, individually, well-formed but cannot be joined to give a legal fold.
a, AFM images were flattened by line-by-line subtraction of a second-order polynomial. Image processing and fitting were performed using custom MATLAB programs. Image 1.2 × 1.2 µm. b, A histogram of pixel heights was used to set the threshold for the generation of a binary image. The threshold was found by calculating the average of the means of the two peaks corresponding to background and tiles; if this failed because the image was noisy the threshold was set manually. c, Well-separated objects in the binary image which have the approximate area of a dimer tile were flagged for fitting (numbered). d, Tile outlines were generated using a Sobel edge-finding filter. e, Representative fitted outlines (two equal, offset parallelograms) were used to classify dimer tiles as described in the text (compare Fig. 3d).
Panels a–e show a 1.5 µm field of view containing structures folded from each of the five staple sets of Fig. 4a–e. Shapes that were flagged for fitting are marked with a dot, green if the shape was successfully fitted and red otherwise. The fitted outlines are superimposed on the image. f, Examples of structures that were either not flagged for fitting or not successfully fitted. AFM images are shown alongside the outline of a suggested structure. The collection of shapes that were not successfully fitted includes crowded areas where shapes are touching and shapes where the two component monomer tiles are distorted, perhaps during deposition on the mica surface, but can be clearly assigned to one of the predicted shapes. Part-folded (or damaged) shapes are also observed, often with one well-folded monomer attached to a part-folded monomer; sometimes a portion of unfolded template can be observed.
The structure labelled 1:1 is a part-folded intermediate in which four pairs of seam staples are bound. If the seam staples remain in place this intermediate could progress to a fully folded structure with seam configuration 4:1 or 3:2 (as indicated by arrows to the right), but fold 5:0 is inaccessible unless two pairs of seam staples dissociate. Circle diagrams in the upper panel show seam connections corresponding to the structures below.
Assembly of a monomer tile (Fig. 1) was monitored using fluorescently labelled staples. The positions of the labelled strands in the folded tile are shown in a: the seam staple was labelled with 5′ Cy3 and 3′ Black Hole Quencher 2, and the body staple with 5′ Cy5 and 3′ Black Hole Quencher 2. Reactions containing the monomer template at 50 nM and staples at 100 nM in a buffer containing 12.5 mM MgCl2, 10 mM Tris-HCl and 0.5 mM EDTA pH 8.0 were held at 96 °C for 10 min, cooled from 96 °C to 25 °C at 0.3 °C min−1, held at 25 °C for 10 min then heated to 96 °C at 0.3 °C min−1. The fluorescence signal for Cy3 and Cy5 was recorded at 0.3 °C intervals during cooling and heating cycles. Staple binding increases the separation between fluorophore and quencher and therefore increases the fluorescence intensity. b, Fluorescence intensities (F) and c, their derivatives (dF/dT) as functions of temperature during origami annealing and melting. Sharp transitions, corresponding to narrow ranges of staple incorporation temperatures, are consistent with cooperative origami assembly. In the case of the unmodified tile the seam staple is incorporated into the tile at the same temperature as the body staple. Hysteresis (marked *) is consistent with the cooperative binding of the seam staple. When one half of the seam is broken the hysteresis observed for seam staple binding is reduced and the seam staple is incorporated at a lower temperature than the body staple. Weakening the seam has little effect on the incorporation of the body staple.
a–e, Heat maps showing the predictions of the model for the number of reconfiguration events during assembly for each of the staple sets shown in Fig. 4a–e. A ‘reconfiguration event’ occurs when a contact between two template domains is released and replaced by an alternative contact. Domains omitted from the map are those which would generate an illegal fold if reconfigured.
a, The original staple set (Fig. 4a); b, the broken-seam variant (Fig. 4b). In each case, average data from 1,600 simulations are presented (‘all’) together with subsets sorted by final fold (5:0, 4:1, 3:2 and misfold). The simulation count for each subset is indicated below each panel. Simulations resulting in well-folded, non-planar structures (NP) are included in ‘all’ but not presented separately: such structures occurred 65 times in a and 5 times in b. Circular icons with internal connections of different lengths represent links across the seam (‘seam links’) connecting points on the template spanning (that is, that are separated by) 28, 56, 84, 112 and 140 template domains (as in the ‘circle’ diagrams of Fig. 3). A ‘seam link’ represents a connection across the seam mediated by at least one seam staple (with the original staple set, a, it may also represent a pair of staples). Data are presented at seven different temperatures as the system is cooled. Correlations between seam links are represented graphically by three 5 × 5 blocks. Each pixel represents a correlation between a pair of seam links which are identified by two icons. The orientation of the icons has the same significance as in the ‘circle’ diagrams: two icons related by 180° rotation represent one internal link in each of the two halves of the template; a 90° rotation represents one internal link and one cross link. Only relative orientation is significant so, for example, fully folded state m:n is not distinguished from n:m (Extended Data Fig. 1b). Each pixel represents the average number of pairs of links present with the specified spans and relative orientations (range 0–8; colour coded, key at top right). The bar on the right of the figure, labelled ‘B’, represents the average occupancy of body staples (range 0–2). For staple set a, folding is substantially complete at 62 °C: at this temperature the patterns of correlation that are characteristic of the fully-folded structures can be seen clearly. For example, the presence of the longest (140–domain) link with no cross-link to the other half of the template is characteristic of fold 5:0. (A 140-domain link with a cross-link only occurs in misfolds and NP structures.) A 112-domain link with a 28-domain cross-link is characteristic of 4:1, and the presence of two 56-domain links including a cross-link is characteristic of 3:2. These and other correlations that are characteristic of the final folds are already visible in the averaged correlation maps (when simulations are sorted by final fold) at very early stages of folding. The pattern of seam staples at an early stage of folding is therefore predictive of the final fold (Extended Data Fig. 9). For the broken-seam staple set b, intact seam staples are incorporated later in the folding pathway (the 50% incorporation temperature for seam staples is 64.2 °C for a, 62.3 °C for b). The 50% body staple incorporation temperature is unchanged (63.9 °C for a, 64.0 °C for b). The same characteristic patterns of seam staples that, with the full seam, are associated with different final folds are also visible at high temperatures for the broken-seam staples. However, 90% of broken-seam simulations result in fold 5:0, as designed. Additional evidence for the influence of strong seam contacts on the folding pathway in the model is provided by the dramatically different yields of misfolds: 52% for full-seam staples a, 1% for broken-seam staples b. Stable incorporation of incompatible seam staples in a prevents the formation of well-folded structures.
Extended Data Figure 9 Seam–staple correlations at early stages of folding are predictive of the final fold.
Data shown correspond to the original staple set (see Extended Data Fig. 8a and Fig. 4a). Three tests were applied at the temperature at which, on average, half of all seam staples are incorporated (64.2 °C). These tests were designed to discriminate between patterns of seam staples characteristic of different final folds. For simulations that satisfy each test, the table records the distribution between final folds. Test 1: a 140-domain seam link with no cross-link to the other half of the template (characteristic of fold 5:0). Test 2: a 112-domain link with a 28-domain cross-link (characteristic of fold 4:1). Test 3: two 56-domain links, including one internal link and one cross-link between halves of the template (characteristic of fold 3:2). Highlighted entries correspond to the fold that each test was designed to predict. The last row of the table records the final distribution between folds of all 1,600 simulations.
See Methods section ‘Example rate calculations’ for the worked examples. a, A half-bound seam staple (brown) can bind to one of two sites on the template (green). Distances along the template to each of the two possible binding sites for the second domain of the staple, measured in nucleotides and base pairs, are marked on the template. In the example shown, the closer binding site is connected by a 448-nt ssDNA chain and the further by a composite chain comprising a 2,208-nt single-stranded chain and one rigid 16-bp double stranded segment. The local concentration of the closer domain at the half-bound staple is estimated to be 11 times higher than that of the more distant domain with a correspondingly greater hybridization rate. b, The previous incorporation of staples changes the physical properties of the loops connecting staple binding sites and thus staple incorporation rates. In the absence of staple A, the shortest path between the binding domains of staple B shown consists of a 864-nt ssDNA chain. In the presence of staple A the path is shortened: it passes through the link formed by staple A and comprises 384 nt ssDNA, 3 rigid 16-bp dsDNA segments and a staple crossover. The effect of the previous insertion of staple A, shortening the link between the two binding sites, is to accelerate the hybridization of the second domain of staple B by a factor of 2.6.
About this article
Cite this article
Dunn, K., Dannenberg, F., Ouldridge, T. et al. Guiding the folding pathway of DNA origami. Nature 525, 82–86 (2015). https://doi.org/10.1038/nature14860
Nature Communications (2022)
Nature Reviews Methods Primers (2021)
Nature Materials (2021)
Nature Communications (2019)
Programming chain-growth copolymerization of DNA hairpin tiles for in-vitro hierarchical supramolecular organization
Nature Communications (2019)