Programmable materials and the nature of the DNA bond

See allHide authors and affiliations

Science  20 Feb 2015:
Vol. 347, Issue 6224, 1260901
DOI: 10.1126/science.1260901

Valency and bonding on a larger scale

In molecular systems, valency describes the number of bonds an atom can make with its neighbors. Larger objects such as colloids can be linked together to make connected structures in which the number of connections, or valency, is controlled by the central object. Jones et al. review the two main approaches to creating stiff bonds, based on DNA-based materials synthesis. These approaches allow the construction of molecular-like objects from building blocks much larger than single atoms.

Science, this issue 10.1126/science.1260901

Structured Abstract


Nucleic acids are ubiquitous in biology because of their ability to encode vast amounts of information via canonical Watson-Crick base-pairing interactions. With the advent of chemical methods to make synthetic oligonucleotides of an arbitrary sequence, researchers can program entire libraries of molecules with orthogonal interactions, directed to assemble in highly specific arrangements. Early attempts to use DNA to make nanostructures led to topologically defined architectures, but ones that were too conformationally flexible to be used to guide the construction of well-defined nanoscale materials from the bottom up. In this Review, we discuss the key discoveries that have overcome this limitation and distill common design principles that have since led to a revolution in materials sophistication based on DNA-directed assembly.


The experimental realization of DNA-based constructs that are sufficiently rigid so as to impart directionality to hybridization interactions marks a major milestone in the development of programmable materials assembly. This feat was accomplished simultaneously by the Mirkin Group and Seeman Group in 1996, but through chemically and conceptually distinct pathways. In one approach, rigidity is derived from multiple strand crossover events and the hybridization that stabilizes them to create a conformationally restricted DNA tile. In the other approach, a rigid non-nucleic acid–based nanoparticle (inorganic or organic) core acts as a template to organize functionalized DNA strands in a surface-normal orientation. It is appealing to draw the analogy between DNA-based constructs of this sort with the concepts of “bonds” and “valency” found in atomic systems. Just as understanding the nature of atomic bonding is crucial for chemists to manipulate the formation of molecular and supramolecular species, so too is an understanding of the nature of these DNA bonding modes necessary for nanoscientists to build complex and functional architectures to address materials needs.


The interest in nanoscale materials constructed by using DNA bonds has continued to grow steadily, but has seen a noteworthy explosion in relevance over the past several years. This is due in large part to the development of methods to move beyond simple clusters and crystals to more sophisticated nanostructured materials that are dynamic and stimuli responsive, are macroscopic in spatial extent, and exhibit emergent physical properties that arise from specific arrangements of matter. These techniques offer perhaps the most versatile way of organizing optically active materials into architectures that exhibit unusual and deliberately tailorable plasmonic and photonic properties. In addition, prospects include the use of these materials in biological settings, being that they are constructed, in large measure, from nucleic acid precursors. The ability to manipulate gene expression, deliver molecular payloads via DNA-based binding events, and detect relevant markers of disease with nanoscale spatial resolution represent some of the most fruitful avenues of future research.

Differentiating nanoscale DNA bonds.

(A) Multiple strand crossover events and DNA hybridization produce a conformationally constrained molecule with a rigid core. (B) A rigid nanoparticle acts as a scaffold for the immobilization and organization of DNA strands in a surface-normal direction.


For over half a century, the biological roles of nucleic acids as catalytic enzymes, intracellular regulatory molecules, and the carriers of genetic information have been studied extensively. More recently, the sequence-specific binding properties of DNA have been exploited to direct the assembly of materials at the nanoscale. Integral to any methodology focused on assembling matter from smaller pieces is the idea that final structures have well-defined spacings, orientations, and stereo-relationships. This requirement can be met by using DNA-based constructs that present oriented nanoscale bonding elements from rigid core units. Here, we draw analogy between such building blocks and the familiar chemical concepts of “bonds” and “valency” and review two distinct but related strategies that have used this design principle in constructing new configurations of matter.

A grand challenge in the fields of chemistry and materials science is the ability to construct materials with absolute control over the placement of each component in order to tailor properties for a given application. Synthetic chemists regularly wield this degree of control over atoms by manipulating the formation of covalent bonds, and supramolecular chemists control the organization of larger molecular species through the manipulation of noncovalent interactions. A key requirement for these bonds is that their interactions be sufficiently directional so that the final arrangement and orientation of molecules may be predicted with reasonable accuracy (1, 2). When this condition is not met—when interactions are conformationally flexible—it is difficult for a system to arrive at a singular thermodynamic product that is well-defined (for example, inherent nonuniformity found in polymer systems), and rational control over the final material is greatly diminished. The synthesis of nanomaterials and their assembly into larger well-defined architectures has conceptually similar goals. We foresee the recent advances in nanomaterials synthesis facilitated by DNA-based assembly processes as capable of one day producing a synthetic methodology that may rival, and in certain cases exceed, at the nanoscale what small-molecule chemists have achieved at the molecular scale (3, 4). Therefore, we find it useful to explore the concepts of “valency” and the “bond” when applied to nanoscale building blocks whose interactions are governed by DNA hybridization.

Aside from their obvious role as carriers of genetic information, nucleic acids have also been used by biological systems to generate natural nanostructures such as ribozymes (5) and Holliday junctions (6) that serve crucial roles in a variety of cellular processes. Perhaps the most salient feature of DNA that can explain its versatility in biological settings is the specificity of canonical Watson-Crick base-pairing interactions (A-T and G-C). Permutation of the nucleobase sequence of particular DNA strands, even those that are relatively short, results in an enormous library of orthogonal interactions that can direct hybridization to occur with high selectivity and specificity.

The concept of controlled valency or directional DNA bonding in programmable materials synthesis can be traced to two seminal papers (7, 8) and several patents (acknowledgements, this paper) published circa 1996 (Fig. 1). These examples were the first to use rigid nanoscale building materials that retained the tailorability of DNA-mediated interactions, as opposed to structures defined only by topology that were explored in early efforts to gain structural control with DNA (9). Although rigidity of a central building block is essential to the valency control in both of these approaches, they differ in how such rigidity is attained and the types of architectures one can envision and construct. The first methodology uses branched DNA architectures (molecules containing multiple crossover junctions between double helical domains) (8), which results in constructs that lacked the flexibility seen previously with only a single crossover junction (2, 10, 11) and form much of the basis for what is called “structural DNA nanotechnology” [Summary figure, (A)] (12). In this approach, carefully designed hybridization and intertwining of DNA strands create a rigid building block with programmable bonding characteristics and allow one to make functional architectures with well-defined geometries. Although these molecules, commonly known as DX tiles, were intriguing for a variety of fundamental reasons (13), it was the demonstration of their conformational rigidity (8) and later their assembly into large crystals (14) that proved their ability to function as two-dimensional (2D) nanoscale building blocks with programmable bonds.

Fig. 1 The development of nanoscale DNA bonds.

Although nature has provided several examples of evolutionarily selected DNA nanostructures, they are conformationally flexible and often do not result in well-defined thermodynamic products when used as building blocks for materials assembly. The development of rigid nanoscale constructs that present directional DNA hybridization interactions via two contrasting approaches in 1996 is largely responsible for the diversity of DNA-programmable materials available to researchers today. [Modified from (12, 41, 47, 58, 60, 64, 81, 82, 91), with permission]

The second approach introduced the concept of a programmable atom equivalent comprised of a rigid non-nucleic acid core, densely functionalized with a layer of highly oriented single strands of DNA (7). The valency in these structures, now termed spherical nucleic acids (SNAs) (15), is dictated by both the central particle and the dense loading of oligonucleotides on the surface of the structure [Summary figure, (B)]; crowding directs the oligonucleotide bonding elements and provides subnanometer control with respect to particle-particle binding events. They do not require hybridization to create a functional building block, and they permit building hybrid materials not attainable via the approaches that rely on nucleic acids to attain valency. Although the prototypical example was a spherical gold particle chemically functionalized with alkylthiol-modified DNA (7), there is now a large table of element equivalents consisting of particles that vary in size, composition, shape, and type of functionalized nucleic acid (3). With this approach, a number of assembled structures, first with short-range order (16, 17) and ultimately with extended 3D periodicity (18, 19), demonstrated the power of this nanoparticle building block to imbue DNA with bonding properties.

Contemporaneous with these aforementioned contributions to DNA valency, a more complete fundamental understanding of the thermodynamics of DNA hybridization allowed for quantitative predictions of duplex melting temperatures that included empirically relevant conditions such as sequence and salt dependencies (20). In addition, a number of important materials possessing only topological order were reported that used DNA to assemble proteins (21) and nanoparticles on discrete molecular templates (22). This approach was expanded by using organic molecules or transition metal complexes whose inherently well-defined bonding geometries allow for DNA hybridization events to be somewhat oriented in space (23, 24). Although these structures do not present rigid DNA bonds and are not useful for programming the formation of macroscopic materials, they are valuable for labeling nucleic acid architectures and building certain molecular analogs to the 3D materials that are the focus of this manuscript.

The field of nucleic acid–guided programmable materials has been bifurcated into two subdivisions that achieve the goal of rigid, directional, DNA-based bonds through different fundamental chemical interactions: (i) the use of intricately woven oligonucleotides participating in hybridization to produce rigid architectures such as tiles and scaffolds, and (ii) the use of rigid nanoparticle cores, which act to template directional interactions on the basis of the core geometry (Fig. 1). We will commonly differentiate these methodologies by referring to each as “hybridization-based DNA bonds” or “nanoparticle-templated DNA bonds,” respectively. These methods represent contrasting but powerful approaches at manipulating matter at the nanoscale through DNA bonds and the principle of valency. Just as the character of different atomic bonds dictates the types of materials that can be constructed from atoms, each type of nanoscale building block presented here has distinct properties that allow access to different materials that are constructed using DNA.

Hybridization-based DNA bonds: Tiles

Some of the first DNA hybridization-based nanostructures used the principle of multiple crossover junctions to impart sufficient rigidity to achieve directional interactions [Summary figure, (A)]. Typically in these molecules, coplanar double helices—often called helical domains—contain two or more locations at which the component single strands switch their connectivity from one helix to the other (Fig. 2A) (12). Although there are numerous forms of these molecules, those used for construction typically have their crossovers formed between strands of opposite polarity. These crossover events between helical domains impart substantial structural rigidity by greatly reducing the number of possible conformations that still produce a hybridization-driven topology. Structures that have two helical domains and two crossover junctions between them are known as DX (double crossover) (12), and those that have three helical domains (often, but not necessarily coplanar) with two crossover junctions between each are known as TX (triple crossover) (12, 25). The rigidity imparted by this strategy has been quantified through measurement of the persistence length of these molecules; DX structures have a persistence length that is roughly double that of an ordinary DNA duplex (26), whereas more complex analogs built of six helices surrounding a central vacancy rotated 120° from each other (27) have even larger values (28). Although the antiparallel and coplanar arrangement of DNA helices in these molecules roughly mimics a linear coordination geometry, because each helical domain can be terminated in a different sticky-end sequence, multicomponent systems of two to eight DX or TX molecules have been designed to tile in an alternating brickwork-like pattern to produce 2D crystalline lattices (Fig. 2A) (14, 25).

Fig. 2 2D DNA hybridization-based tiles.

(A) The DX and DX+J tiles assembled into a brickwork 2D lattice. Scale bar, 300 nm. [Modified from (12, 14), with permission] (B) The 4 by 4 tile assembled into a 2D alternating square lattice. Scale bars, 100 nm. [Modified from (39), with permission]

This ability to rationally assemble synthetic DNA-based tiles has enabled a number of important advances in both the construction of scaffolds for the immobilization of other nano-objects and for the development of dynamic nanoscale materials. Slight modifications to the DNA design allow for DX and TX tiles in which hairpin loops or sticky-end sequences protrude approximately perpendicular to the plane of the molecule (25). In addition to being useful as a topographical marker in atomic force microscopy (AFM) characterization, these strands can be used to capture proteins or nanoparticles and organize them on the 2D lattice assembled from these tiles (29, 30). By linking two DX molecules with a DNA sequence capable of undergoing a rotational structural transition, the concept of using rigid DNA tiles in the formation of nanomechanical devices was introduced (31). This work marks the start of the field of nanorobotics, which has made extensive use of rigid DNA bonds to create dynamic systems that can carry out physical tasks of impressive sophistication (3235). One tile design that has become particularly useful in systems requiring DNA-based rotary motion is the PX molecule (paranemic crossover) (34, 35), in which crossover events occur at every possible location where the major or minor grooves of the antiparallel helical domains meet (36).

Although many of the early tile-based structures presented linear or pseudolinear bonding modes, more advanced tessellating structures constructed by using DNA hybridization have expanded directional assembly interactions to include more complex symmetries. This concept was first explored by combining several four-arm branched junction molecules into a single parallelogram-shaped structure (37). Although individually too flexible to be considered bonds (38), when these DNA junctions are combined into larger well-defined molecules, it was shown that they could assemble into 1D ribbons or 2D lattices, depending on the placement of sticky-end groups, each with a rhombus-shaped repeat unit. Yan, Reif, and coworkers later showed a cross-shaped DNA structure, known as the 4 by 4 tile, that presented four coplanar arms oriented 90° to one another (Fig. 2B) (39). Each arm in this structure consists of two helical domains, much like the DX molecules, the antiparallel orientation of which imparts sufficient rigidity to maintain the orientational relationship between each of the four arms. When sticky-end sequences are placed on the ends of each arm, these structures are able to assemble into 1D rolled-up tubes or periodic 2D square lattices, depending on the number of helical turns of DNA that are in each arm (Fig. 2B). Because the central strand in this tile can be readily modified to present capture sequences or protein binding domains (such as biotin), these tiles have been used extensively to organize proteins and nanoparticles into single-component and alternating 2D arrays (40).

Mao and coworkers presented directional binding domains in three dimensions using structures known as tensegrity triangles (Fig. 3A) (41). Like the parallelogram structures discussed previously, these constructs derive their rigidity from the linking of several four-arm branch junctions and conceptually consist of three double-helical domains oriented to form a triangular-shaped tile. However, unlike previous structures, the arms of this tile are not all coplanar and point in three separate directions in 3D space like the axes of a rhombohedral coordinate system. Thus when two of the three arms of the tensegrity triangle are assembled via sticky-end hybridization, 2D arrays form with the third arm pointing at a non-coplanar angle. In later work, a more rigid version of this molecule was synthesized in which each arm consisted of a DX molecule, rather than a single linear duplex, which allowed for the generation of ordered 2D arrays of nanoparticles (42). Recently, tensegrity triangles have been assembled into macroscopic 3D crystalline materials with considerable long-range order so as to be characterized by means of x-ray diffraction, demonstrating their power as DNA-bonds (Fig. 3A) (43). Mao and coworkers have shown an alternative method using DNA tile-based building blocks to construct 3D wireframe polyhedral structures (Fig. 3B) (44). In this work, they use a single, versatile structure known as the three-point-star motif, which consists of three DX tile arms linked via a central cyclic strand. The presence of hairpin loops on this central strand imparts intrinsic curvature to the tile and allows one to tailor the flexibility of the structure via the length of the loop. When these tiles were designed to present sticky-end sequences on each arm, it was found that the tile concentration and arm flexibility (controlled through loop length) dictated the morphology of the assembled structure; tetrahedra were observed with flexible tiles at low concentrations, and dodecahedra and truncated icosahedra were observed with more rigid tiles at low and high concentrations, respectively (Fig. 3B). The yield of each structure is considerable (69, 76, and 90% for the truncated icosahedra, dodecahedra, and tetrahedra, respectively), and it is likely that more complex 3D objects can be made with a more diverse library of tiles from which to build. With added sophistication, however, comes an increased likelihood of undesirable side-products, suggesting that methods to mitigate error propagation must go hand-in-hand with increases in structural complexity.

Fig. 3 3D DNA hybridization-based tiles.

(A) The tensegrity triangle assembled into a macroscopic 3D rhombohedral lattice. Scale bar, 500 μm. [Modified from (12, 43), with permission] (B) The three-point star motif assembled into a variety of wireframe polyhedra based on the intrinsic curvature of the tile and its concentration. Scale bars, 20 nm. [Modified from (44), with permission] (C) Single-stranded DNA assembled into arbitrary 3D brickworklike nanoscale objects. Scale bar, 20 nm. [Modified from (46), with permission]

In an interesting departure from the usual rigid tile approach, Yin and coworkers demonstrated in 2012 extremely complex discrete 2D and 3D objects using single-stranded DNA units (Fig. 3C) (45, 46). In this case, a library of oligonucleotides that are conceptually split into four sequence domains are designed to hybridize to four independent neighbors. This forms a rigid bricklike building block that serves the crucial purpose of enabling directional hybridization interactions. Because of the well-defined helical twist of a DNA duplex, the number of bases in a domain determines the dihedral angle between adjacent hybridized segments. Consequently, strands can be designed to form a 2D (45) or 3D (46) brickwork pattern. Conceptually, one then imagines selectively removing pixels or voxels from this molecular canvas to create an arbitrary 2D (45) or 3D (46) object, respectively. A computer program then considers a number of important design rules and generates a set of thousands of single-stranded oligonucleotides that, when combined and annealed for a day or more, form extremely complex aperiodic patterns, including 2D symbols and images and 3D block letters and shapes (Fig. 3C) (45, 46). Although the relative yield of these nanoscale objects is low, the technique is comparable with DNA origami and uses a modular approach by computationally selecting the desired tiles from a master set. This method does not seem to require a tight control over the stoichiometry of the strands; like other large multicomponent systems, including DNA origami and crystals, failures to include an individual component results in a flaw, sometimes not detectable, rather than complete failure to make the target.

Hybridization-based DNA bonds: Origami

The premise of DNA origami is to fold a multithousand-base circular single-stranded DNA “scaffold,” obtained from a viral genome, by using short helper or “staple” oligonucleotides into a desired nanoscale shape (Fig. 4A) (47). Because each staple strand is different, some can be designed to present DNA sticky ends at programmed locations on the periphery of the final assembled object, allowing these structures to act as scaffolds with spatially prescribed DNA bonds that capture and organize other nano-objects. These materials, therefore, derive the structural rigidity of their DNA bonds from the same chemical source as DNA tiles: numerous crossover events between neighboring duplexes and the hydrogen bonds that link them. However, this approach differs in that it is less focused on the development of a small number of building blocks that can arrange themselves into large periodic structures, and is more interested in constructing discrete nanoscale objects whose size and shape is well-defined. The advantage over tile-based assembly is that hybridization occurs frequently via intramolecular interactions on the scaffold strand, resulting in a high local concentration of complementary oligonucleotides that drives the system toward the intended sequence-specific thermodynamic product (48). This allows for relatively high yields of particularly complex objects and obviates the need for a precise stoichiometry of all components in the system. As in all these systems, one must be extremely careful to control the sequences of the staple oligonucleotides to fold the scaffold strand into a desired shape, necessitating computational methods to be applied for the design of each structure that is to be synthesized. To address this problem, a number of open-source software packages are now available that aid researchers in developing the strands and sequences required for a particular DNA origami structure (49, 50).

Fig. 4 DNA hybridization-based origami structures.

(A) Illustration of the rasterlike pattern of the scaffold strand folded by staple strands, which are collectively used to generate arbitrary 2D patterns. Scale bar, 20 nm. [Modified from (47), with permission] (B) Extension of the origami principle to 3D by using staple strands which promote the formation of pleated sheets of duplexes, which ultimately pack into a honeycomb lattice. Scale bars, 20 nm. [Modified from (58), with permission] (C) Insertion or deletion of bases at specific sites in a block of honeycomb duplexes can be used to selectively create regions of local twisting or curvature. Scale bars, 20 nm. [Modified from (59), with permission] (D) Concentric rings of duplexes can be encouraged to extrude out into the third dimension to form hollow spherical and pseudospherical objects. Scale bars, 50 nm. [Modified from (60), with permission]

Although this principle was experimentally demonstrated with barcode arrays (51) and in the construction of a nanoscale wireframe octahedron (52), the power of the technique was best illustrated with the formation of numerous 2D shapes of impressive complexity (47). Conceptually, these patterns are generated by imaging the scaffold strand being rastered back and forth to fill an arbitrarily shaped 2D area (Fig. 4A). Computational methods are then used to select the appropriate staple strands so that the desired folding path is stabilized by DNA hybridization and numerous crossover junctions. The method requires that several important design parameters be considered, including the number of helical twists of DNA between crossover points and the elimination of strain, nicks, and seams. Although the majority of structures formed in greater than 60% yield, two alternate strategies for forming similar triangular patterns resulted in vastly different yields of ~1 and 88%, highlighting the need to carefully design and optimize sequences because the number of potential undesired interactions is quite large.

DNA origami has become a powerful tool for building discrete nanoscale materials, generating a variety of complex and dynamic structures, including a box with a sequence-specific opening lid (53) and a barrel-shaped structure that can carry molecular payloads and release them based on intracellular logic-gated aptamer binding events (54). Origami methods have found particular utility in constructing 2D scaffold materials because individual sticky-ends and other DNA-based features can be placed in nearly any arrangement. This has facilitated the demonstration of DNA walkers that can pick up nanoscale cargo (35), studies of distance-dependent bivalent ligand-protein binding (55), and the controlled placement of nanoparticles for plasmonic applications (56, 57).

A conceptual leap in DNA origami was made in 2009 when these principles were extended from building 2D objects to those that fill 3D space (58, 59). This was accomplished by imagining the scaffold strand being rastered into an elongated 2D sheet, which is then encouraged to fold back and forth onto itself via the interactions of staple strands to form pleated sheets of antiparallel double helices (Fig. 4B) (58). Because of the helical twist of DNA, crossover events between helical domains are selected to occur at 120° dihedral angles, as in the six-helix bundle (27), resulting in a dense honeycomb lattice of DNA helices that fill a 3D volume. From this block of DNA, one must imagine removing segments in order to “carve out” the desired nanoscale object, a process that requires careful attention to a variety of design considerations but is now simplified by computer-aided design (CAD) software (49). Using this approach, a variety of complex 3D shapes were synthesized, including a monolith, a square cross, and a genie bottle (Fig. 4B); in some cases, these objects could even be assembled hierarchically into larger structures, such as a wireframe icosahedron approximately 100 nm in size (58). A second enhancement introduced the concept that insertion or deletion of bases in a block of honeycomb duplexes creates local regions of strain because of underwinding or overwinding of the helix, respectively (Fig. 4C) (59). This strain is accommodated by the system through simultaneous global twists and global bends of the origami structure. By properly selecting the insertion or deletion sites in a block of duplexes, one can isolate a single deformation mode and create structures that either twist or bend in a controllable manner. This approach greatly expands the types of materials available to origami methods because structures no longer need to be imagined as comprising voxel-type, stepwise components but instead can experience smooth spatial transitions from one feature to another. To demonstrate the power of this approach, a variety of impressive structures were formed, including twisted ribbons of both chiralities, bent arms with a full range of internal angles, notched gears of different sizes, and 3D spherical wireframe objects.

In a conceptually similar approach, one imagines folding a scaffold strand in such a way as to create a series of concentric circular rings of DNA duplexes that are coplanar (Fig. 4D) (60). Through appropriate design of crossover points and staple sequences, each ring can then be positioned at non-coplanar dihedral angles with respect to its neighboring rings. In the simplest case, this has the effect of pulling the concentric circular rings out of their coplanar arrangement to generate a hemisphere or bowl-shaped object (Fig. 4D). Additional designs for incorporating gradual curvature allow for hollow sphere and vase-shaped objects to be formed.

In all of these examples, such a high degree of structural complexity comes at a price: Mixtures of scaffold and staple strands require precise recipes for counterion concentrations (Na+ and Mg2+) and long annealing times of up to a week for equilibrium materials to be formed (58). Even under these ideal and optimized conditions, scaffold strands are estimated to incorporate into the desired monomeric species at between 7 and 44% yield and therefore necessitate a gel purification step to isolate (58). Conversely, Dietz and coworkers have demonstrated non-equilibrium folding of complex origami structures with near quantitative yield in minutes, suggesting that kinetic routes to idealized materials may be an important synthetic tool for future research (61). Nonetheless, these objects have already been used by other groups to generate DNA bonds that facilitate the construction of interesting plasmonic (62) and metamaterial (63) architectures.

Both brick structures and origami structures are objects and not periodic lattices. Although both can fill large areas (about 104 nm2) in two dimensions, they are finite. Both of these types of structures have been facilitated by the drop in effective DNA cost, largely because the methods used to analyze structures require much less material.

There is another phenomenon that has been revealed when using both simple tiles and DNA origami structures as building blocks to construct larger periodic (or perhaps aperiodic) arrays: The success of an assembly depends on the presence of helix axes pointing in the direction of crystal propagation. The first 2D crystals formed from DX or TX tiles tended to be long and narrow (14, 25) but certainly passed for crystals. The assembly algorithm was diagonal complementarity along the tile. When this approach was applied to forming larger crystals from origami building blocks, the arrays were even thinner, relative to the size of the tiles (64). Only when the helix axes pointed in the direction of propagation were substantially isotropic crystals of origami tiles produced (64). The same is true of 3D crystals from small tiles: The six-helix bundle and the TX tile are capable of forming 3D-ordered materials (65) but fail to produce crystals that diffract adequately. Only the tensegrity triangle (41), with three linearly independent helix axes, has produced substantial 3D crystals thus far (43). However, although the rule of connecting objects along the direction of the helix axis is a necessary condition, it is not a sufficient condition. Numerous examples of very small arrays have been found in systems that ought to have done better (65). This is an avenue of investigation that needs to be explored further.

Nanoparticle-templated DNA bonds

A fundamentally different approach to generating nanoscale DNA bonds is to use nanoparticles as templates for the immobilization and orientation of surface-bound oligonucleotides [Summary figure, (B)] (66). Rather than use hybridization and intertwining of DNA strands, it is most often the metallic or ionic bonds that form the crystalline lattice of an inorganic core material that provide the necessary rigidity for these species to generate directional interactions. Thus, many of the design considerations for DNA hybridization-based systems, like the helical twist of a DNA duplex or the necessity of several crossover junctions, become unimportant so long as the inorganic material can be functionalized with a dense shell of the desired oligonucleotide ligands. One of the more notable differences of this approach is that nanoparticle-templated DNA bonds typically hybridize with other constructs through the collective interactions of tens to hundreds of individual, densely packed DNA strands. The result of spatially confining these strands is profound; enormous changes to the thermodynamics of hybridization lead to sharper melting transitions (67), enhanced binding constants (68), and elevated thermal stability (69), the enthalpic and entropic underpinnings of which remain the subject of continuing research effort. This focus on engineering individually weak but multivalent DNA hybridization interactions results in substantial differences in how these constructs participate in bonding, the symmetries of their interactions, and ultimately the materials that can be constructed from them.

Nanoparticle-based structures using DNA hybridization to govern their association were developed by using oligonucleotides presenting 3′ or 5′ terminal alkyl-thiol moieties that were added to solutions of colloidal gold nanoparticles ~13 nm in diameter (7). By gradually increasing the ionic strength of the solution, the negatively charged phosphate backbone of the DNA strands could be effectively screened, facilitating the formation of extremely dense monolayers of oligonucleotides on the gold nanoparticle scaffolds (70). Because the DNA nucleobases themselves have a considerable affinity for gold surfaces (71), ordinary oligonucleotides are able to lie down and wrap around gold nanoparticles in a random and nondirectional manner. Therefore, the presence of the thiol moiety as a robust anchoring group on one end of the DNA strands (either 3′ or 5′, depending on synthetic details) serves the nontrivial role of orienting all of the oligonucleotides in a common surface-normal direction, making them available for hybridization to their complements. Initial work focused on exploring the use of linker oligonucleotides that contained regions complementary to the strands anchored to the particles to hybridize them into macroscopic networks (7, 72). It was found that these materials exhibited considerable short-range order, with interparticle distances proportional to the length of DNA used to link them (16, 17). Significant correlation between the particle positions was only observed when the DNA linking them was primarily double-stranded; introduction of single-stranded regions resulted in significant a loss of ordering (17). These results demonstrate that it is a combination of the rigidity of the inorganic core, coupled with a dense, oriented monolayer of primarily duplexed oligonucleotide ligands that create the conditions necessary for these constructs to generate directional DNA bonds.

These building blocks can be crystallized into well-ordered superlattices with considerable long-range periodicity (Fig. 5A) (18, 19). The key advance was the use of a thermal annealing step at temperatures just below the particle-particle dehybridization transition and the design of short DNA sticky ends that are strong enough to hold the particles together but weak enough to allow for rapid equilibration (18). The implication is that the thermodynamic state of the system is one in which the particles are assembled into ordered superlattices that maximize the number of DNA hybridization events, and the network structures observed initially represent kinetically trapped, metastable states (73). This hypothesis was tested and confirmed in numerous subsequent investigations and now forms the basis for a series of design rules that guide and explain the formation of dozens of different superlattice symmetries (3, 74, 75). In order to determine the thermodynamic stability of an arbitrary arrangement of particles, one can simply account for the surface-area contact between spherical particles baring complementary linkers as a proxy for counting the number of hybridization events (74). This formalism, known as the complementary contact model (CCM), explains the effects of experimental parameters such as linker sequence, DNA flexibility, linker number ratio, and particle size ratio and has even been used in the a priori design of more complex three-component nanoparticle superlattices (76).

Fig. 5 Nanoparticle-templated spherical DNA constructs.

(A) Spherical nanoparticles functionalized with the appropriate DNA strands can assemble into a variety of superlattices (AB6-type structure shown in the scheme and transmission electron microscopy (TEM) image with tomographic reconstruction, inset), some of which form large faceted single crystals. Scale bars, TEM image, 100 nm; scanning electron microscopy image, 1 μm. [Modified from (74, 82), with permission] (B) Superlattices can be grown from a substrate with a preferred crystal orientation, allowing for textured nanoparticle films. Scale bars, 200 nm. [Modified from (80), with permission]

Although understanding the fundamental properties of these nanoparticle-templated DNA bonds was crucial, the technique presented several limitations: Lattices had only been constructed from gold nanoparticles, were only stable in buffered solution conditions amenable to DNA hybridization, and were polycrystalline in nature. Because the only requirement for these DNA bonds is that they derive their structural rigidity from a solid material core, inorganic nanoparticles with a variety of different compositions—including catalytic noble metals, semiconductors, and magnetic oxides—were subsequently functionalized with DNA and assembled by use of this technique (7779). In addition, lattices have since been grown from substrates with a preferred crystallographic direction (Fig. 5B) and encapsulated in glass by using a molecule that provides an initiation site for silica growth (80, 81). These advances have allowed DNA-nanoparticle superlattices to become solid-state materials, greatly expanding their potential use in a variety of applications. It may be possible to use similar approaches to create solid-state analogs of the tile- and origami-based structures. Last, these programmable atom equivalents, when cooled slowly, have been shown to form large single crystals with a well-defined crystal habit (a rhombic dodecahedron) indicative of the minimum-energy Wulff polyhedron of the parent superlattice (Fig. 5A) (82). When these faceted microcrystals are assembled from plasmonic nanoparticles, they interact strongly with light through hybrid plasmonic-photonic modes, demonstrating the importance of using nanoparticle-templated DNA bonds that can assemble materials that are approaching macroscopic length scales (83).

Although these spherical nanoparticle-based DNA bonds allow for regularly spaced network structures and extremely well-ordered superlattices to be assembled via fairly isotropic interactions that mimic metallic or ionic bonding, considerable effort has been applied to develop methods to break the spherical symmetry and achieve a greater directionality of the DNA interactions that more closely mimic covalent bonding (Fig. 6A). One conceptually straightforward, albeit experimentally challenging, way to achieve this is to dictate an asymmetric spatial distribution of oligonucleotide ligands on the surface of a spherical nanoparticle. Two alternate strategies have been demonstrated for achieving DNA-functionalized gold nanoparticles having one hemisphere hybridized with linker strands to make them asymmetric (84, 85). Both methods rely on the hybridization of small DNA–gold nanoparticles (AuNPs) to larger DNA-coated colloidal particles in order to “capture” new oligonucleotide functionality at the hemispherical interface between the two particles. In one case, the strands of the smaller particle are enzymatically ligated to short oligonucleotides, extending their length and sequence (84), and in the other case, the smaller particle hybridizes and acquires linker oligonucleotides that offer new binding functionality (85). An alternative method has been used to produce the same asymmetric DNA functionality by hybridizing small DNA-AuNPs to larger DNA-modified particles that act to sterically and electrostatically block access to the hemisphere nearest the larger particle while allowing hybridization of linker strands to the hemisphere exposed to solution (86). Building blocks of this variety resemble linear or unidirectional bonding modes and therefore assemble into discrete dimer clusters or can be combined with particles of differing size to create hierarchically assembled core-satellite structures.

Fig. 6 Nanoparticle-templated anisotropic and asymmetric DNA constructs.

(A) Control over the spatial distribution of surface-bound DNA strands allows for particles that are asymmetric and can form core-satellite clusters. Scale bars, 50 nm. [Modified from (84, 87, 88), with permission] (B) The formation of DNA patches at symmetrically arranged positions on spherical colloids allows for particle analogs of certain molecular species. Scale bars, 2 μm. [Modified from (91), with permission] (C) By using core nanoparticles that are anisotropic (triangular prisms, rods, and rhombic dodecahedra), while leaving the DNA design unchanged, directional interactions that mimic the symmetry of the underlying particle facilitate the formation of different superlattice architectures. Scale bars, 100 nm. [Modified from (93), with permission]

Although asymmetrically functionalized particles are appealing for achieving nanoparticle-based DNA bonds, reliable methods for moving beyond simple linear geometries have proved elusive. One approach to overcome this hurdle has been to use anisotropic nanostructures that present regions of greater chemical reactivity that can be functionalized selectively with DNA. It is thought that the native surfactant coating required for the synthesis of these structures is less dense at regions of high curvature (tips and edges) and can therefore be replaced by alkylthiol-functionalized DNA ligands more readily. This has been experimentally demonstrated by the selective incorporation of DNA at the tips and edges of platelike triangular nanoprisms (87), or at the tips of gold nanorods (88), resulting in anisotropic core-satellite clusters (Fig. 6A). Several reports have demonstrated dynamic forms of asymmetric bonding in which large colloids or emulsions are functionalized with DNA ligands that are mobile around the particle surface and can pool together at bridge points between complementary structures, forming valence clusters (89, 90). One impressive alternative strategy generates micrometer-sized colloidal particles with between 1 and 7 DNA patches arranged in highly symmetric geometries that mimic the multivalent bonding modes of atomic orbitals (Fig. 6B) (91, 92). The process starts by creating clusters with different numbers of amine-modified polystyrene spheres packed into highly symmetric arrangements. A chemically distinct polymer is then spherically grown from the center of any cluster by using a swelling and cross-linking process to the extent that only small islands or patches remain from the original amine-modified spheres. DNA strands can then be selectively conjugated to these amine-terminated domains, resulting in spherical colloids with sequence-selective patches oriented in well-defined geometries (91). Combinations of particles with different bonding coordination modes allow for the assembly of colloidal analogs to common molecules, including linear CO2, triangular BF3, and tetrahedral CH4, to name a few (Fig. 6B) (92). It remains a challenge to extend these principles of selective DNA functionalization to rigid nanoparticles of smaller size or with interesting compositions.

One interesting feature that differentiates particle-templated DNA bonds is that the role of the DNA in mediating sequence-specific interactions is conceptually decoupled from the role that the nanoparticle plays as the rigid scaffold necessary for the formation of well-defined products. Consequently, one can imagine an alternative strategy for accessing directional oligonucleotide interactions that adopt nonspherical geometries: Replace the typically spherical nanoparticle cores with those that are anisotropic, leaving the DNA design and functionalization unchanged (Fig. 6C). This modular approach uses several different anisotropic gold nanostructures (rods, triangular prisms, rhombic dodecahedra, and octahedra) uniformly functionalized with oligonucleotides and allowed to assemble via DNA-linker–mediated hybridization (93). In each case, it was found that nanoparticle superlattices formed whose symmetry and dimensionality were dictated by the geometry of the particle used in their construction. In particular, regions of each nanoparticle shape with the least curvature (flat facets) showed strong preferences for interacting such that rods formed 2D hexagonal close-packed lattices, triangular plates (prisms) formed 1D columnar stacks, and rhombic dodecahedra and octahedra formed face-centered-cubic (FCC) and body-centered-cubic (BCC) superlattices with face-to-face orientational correlations, respectively (Fig. 6C). It was later shown that the collective interaction of numerous oligonucleotide ligands oriented perpendicular to these flat surfaces could exhibit binding constants six orders of magnitude larger than similarly functionalized curved surfaces (94). The implication of this work is that when bundles of densely packed DNA strands are collectively oriented in conformations ideal for binding (such as flat face-to-face interactions), the shape of the underlying inorganic nanoparticle can play an enormous role in dictating the directionality of interparticle interactions (95). Consistent with results from Chaikin and others (96, 97), this highlights the importance of considering the entropic consequences of tethering and otherwise conformationally restricting particle-bound strands when using particle-based DNA bonds. Indeed, when loosely packed single-stranded oligonucleotides (78) or extremely long duplexed linker strands (93) are used to direct anisotropic particle association, assemblies that indicate isotropic interactions are observed because the DNA is too conformationally flexible to retain the symmetry of the rigid scaffold to which it is anchored; these conditions do not favor the use of anisotropic particles to create directional nanoscale bonds.


Although DNA-based nanoscale construction continues to produce materials with an impressive degree of control, the hybridization-based and nanoparticle-templated subfields remain relatively isolated, with very few examples of overlap between disciplines (98). It is likely that there are fruitful areas of research that make use of the advantages of each approach. For example, DNA nanotechnology and DNA origami have been used extensively in the creation of dynamic nanostructures such as DNA machines (31, 32), walkers (33), a nanoscale assembly line (35), and complex hybridization-based amplification or reaction networks (99101). Although some of these concepts have recently been applied to nanoparticle-based bonds to create superlattices that are dynamically tunable or have programmable phase behavior (102104), there is considerable room for improvement in the creation of reconfigurable particle-based structures. On the other hand, although nanoparticle-based structures have been used in a number of biological applications as tools for gene regulation (105) and as probes in more than 2000 medical diagnostic products available worldwide (106109), tile- or origami-based DNA structures are only now being used for cellular applications, including drug delivery (54) and biological imaging (110). Last, it remains a considerable research challenge to mimic the complexity and versatility of discrete DNA origami structures with inorganic materials such as noble metal and semiconductor nanostructures.

Although periodic structures assembled efficiently by using tile- and nanoparticle-based DNA bonds and have allowed the rational design of large, faceted single crystals of each (43, 82), more robust methods at generating discrete DNA objects in high yield with techniques such as origami are necessary. This advance may be accomplished through methods that allow for synthetic oligonucleotides whose sequence and chemical structure are more pure, or perhaps there are alternative methods that are superior at creating large objects that challenge conventional thoughts in DNA origami (45, 46). Improvements in understanding the complex thermodynamic and kinetic influences of various counterions and thermal cycling may also aid in allowing these systems to arrive at the desired product.

Historically, this field has been driven primarily by the control of matter at the smallest length scales, with functional applications being demonstrated only after materials can be constructed in high-yield and in a well-defined manner. This notion persists today, and consequently, an enormous focus still rests on the development of DNA bonds that are versatile and powerful tools for programming the construction of sophisticated nanoscale materials. Although enormous progress has been made in the short time that these building blocks have been available, the field is still in its infancy and has much more to offer the scientific community. Just as chemists have learned to manipulate atomic bonds to synthesize astonishingly complex and functional molecules, an understanding of the nature of these DNA bonds and an ability to control them may one day allow nanoscientists to create similarly complex and functional nanoscale materials.

References and Notes

  1. Nanosphere;
  2. EMD Millipore;
  3. Acknowledgments: C.A.M. thanks the Air Force Office of Scientific Research under award FA9550-11-1-0275. N.C.S. thanks the following grants: N000141110729 and N000140911118 from the Office of Naval Research; grants CMMI-1120890, EFRI-1332411, and CCF-1117210 from the National Science Foundation (NSF); Multidisciplinary University Research Initiative W911NF-11-1-0024 from the Army Research Office; DE-SC0007991 from the U.S. Department of Energy for partial salary support; and grant 3849 from the Gordon and Betty Moore Foundation. C.A.M. and M.R.J. thank E. Auyeung, J. Cutler, and J. Wu for providing the tomographic reconstruction of the AB6 nanoparticle superlattice provided in Fig. 5A. M.R.J. thanks NSF for a Graduate Research Fellowship. The authors acknowledge the following patents based on the work described in this document: C.A.M., R. L. Letsinger, R. C. Mucic, J. J. Storhoff, R. Elghanian (Nanosphere), Nanoparticles having oligonucleotides attached thereto and uses therefor, W.O. Patent 1998004740, February 5, 1998; C.A.M., R. L. Letsinger, R. C. Mucic, J. J. Storhoff, R. Elghanian (Nanosphere), Nanoparticles having oligonucleotides attached thereto and uses therefor, U.S. Patent 8,323,888, December 4, 2012; N.C.S., X. Li, X. Yang, J. Qi (New York University), Nanoconstructions of geometrical objects and lattices from antiparallel nucleic acid double crossover molecules, W.O. Patent 1997041142, November 6, 1997; N.C.S., E. Winfree, F. Liu, L. W. Savin (New York University), Polynucleotide lattice for use as tools in genetic engineering, U.S. Patent 6,255,469, July 3, 2001.
View Abstract

Navigate This Article