DNA nanotechnology offers unparalleled precision and programmability for the bottom-up organization of materials. This approach relies on pre-assembling a DNA scaffold, typically containing hundreds of different strands, and using it to position functional components. A particularly attractive strategy is to employ DNA nanostructures not as permanent scaffolds, but as transient, reusable templates to transfer essential information to other materials. To our knowledge, this approach, akin to top-down lithography, has not been examined. Here we report a molecular printing strategy that chemically transfers a discrete pattern of DNA strands from a three-dimensional DNA structure to a gold nanoparticle. We show that the particles inherit the DNA sequence configuration encoded in the parent template with high fidelity. This provides control over the number of DNA strands and their relative placement, directionality and sequence asymmetry. Importantly, the nanoparticles produced exhibit the site-specific addressability of DNA nanostructures, and are promising components for energy, information and biomedical applications.
Since the power of DNA as a chaperone for the assembly of gold nanoparticles (AuNPs) was first exemplified1,2, it has been successfully applied to the one-dimensional (1D), two-dimensional (2D) and three-dimensional (3D) organization of AuNPs3,4. DNA-mediated AuNP assembly typically employs polyfunctionalized nanoparticles, and has yielded a range of functional AuNP superlattices5,6,7,8,9,10. However, the construction of discrete nanoparticle assemblies of arbitrary geometry requires the use of DNA scaffolds to position the nanoparticles precisely11,12,13,14,15,16,17,18,19,20. Often, complex DNA origami scaffolds are used to organize a few nanoparticles21,22,23,24,25. Importantly, the information is provided solely by the scaffold, with the nanoparticle either isotropically functionalized or monofunctionalized. DNA nanostructures have been employed previously as templates for the growth of nanowires and inorganic surface patterning. In these instances the overall shape of the structures is transferred to the substrate, but the complex DNA-sequence information encoded in the template is lost26.
Recently, the ability to impart particles themselves with the geometries and valences that we take for granted in molecular self-assembly has been proposed as an alternative approach27,28, and it may open doors to higher levels of control over nanoparticle assembly with a minimum number of DNA strands. With this in mind, the creation of AuNPs that possess anisotropic DNA patterns has been highly sought28,29. For example, controlled spacing between two or three DNA strands on the AuNP surface has been achieved using 1D DNA templates30,31,32. Microsphere clustering with in situ polymerization has been used to create DNA patches on polymer colloids33. Surface immobilization has also been used, and creates Janus particles with localized patches of DNA strands34,35,36,37. The combination of surface techniques with a stepwise approach that exploits electrostatic repulsion has produced geometrically defined DNA–AuNPs with a controlled number of up to six strands38. More recently, amphiphilic polymers were used to allow regioselective DNA functionalization on gold nanoparticles39. Although these methods produced AuNPs with a significant measure of anisotropic regioselectivity, the ability to position many DNA strands with different sequences into geometrically controlled patterns has remained elusive.
Herein we describe a general method to transfer DNA-strand motifs from a parent 3D DNA template to gold nanoparticles (Fig. 1). We show that the nanoparticle substrates acquire the DNA-sequence information encoded in the parent template with high fidelity. Importantly, this produces nanoparticles that exhibit one of the defining properties of DNA nanostructures themselves: site-specific addressability. To our knowledge, this represents the first example of the direct ‘printing’ of a DNA pattern onto gold nanoparticles from a 3D nanoscaffold, and provides arbitrary control over the number of strands and their relative placement, directionality and sequence asymmetry. It creates directionally functionalized AuNPs27, which are very promising components for the self-assembly of complex DNA-programmed architectures. This also represents a first step in chemically copying the complex information contained in DNA nanostructures to another material, and offers a potential approach to the problem of cost and scalability of DNA-based technologies. The process is related to the lithographic replication of electron-beam patterns onto other materials, but here the molecular information of the template is retained in the replication process.
Results and discussion
The goal of our approach is to produce a robust patterning of DNA strands on the nanoparticle surface. This requires anchoring groups appended to the oligonucleotide that can bind strongly to the gold surface and remain so after downstream processing and manipulation of the samples29. With this in mind, a novel DNA conjugate was synthesized40,41, and was terminated with two cyclic disulfide moieties to allow AuNP–DNA conjugation (Supplementary Section IIIa–f).
The bisdisulfide–DNA (Dx, where x refers to the unique sequence) binds to AuNPs more efficiently and with faster binding kinetics than simple cyclic disulfide (CD)-terminated DNA (Supplementary Section IIIh). The extending C12 portion in Dx may help to orient the disulfide moieties away from the highly charged DNA backbone, and thus reduce the electrostatic repulsion, which is a key factor in the binding of DNA strands to gold nanoparticles42. Furthermore, AuNPs functionalized with Dx were considerably more stable than with the CD, as evidenced by a displacement assay using 1,4-dithiothreitol43 (see Supplementary Section IIIi). These binding and stability data suggest that the Dx conjugates are well suited to the patterning experiments to follow.
The criteria we set for the DNA nanostructure used as a template are as follows: it has the potential for (1) geometric variation, (2) the introduction of sequence asymmetry, (3) positioning of different numbers of Dx and (4) can be removed easily after pattern transfer. We showed previously that DNA minimal ‘clip-by-clip’-based structures can be used efficiently to site-specifically position DNA–amphiphile and DNA–polymer conjugates41,44,45. Here the same strategy was employed to arrange precisely the gold-binding DNA conjugates for pattern transfer to a nanoparticle.
The first step in the patterning process was to prepare the template, in this case a 3D DNA scaffold decorated with Dx for transfer to the AuNP substrate. As such, the organization of different numbers of Dx on the DNA scaffold was carried out to produce a range of different letters for the printing process (Fig. 2a). The single-stranded regions of the cubic scaffold Cb are 20 nucleotides (nt) in length. Based on previous designs, only the central 14 nt were used for the binding of the Dx and these have a spacer of 5 nt to orient them away from the crowded corners of the structure. This design also lowers the melting temperature of this region to facilitate the removal of the template after pattern transfer. The structures shown in Fig. 2b were assembled by one-pot thermal annealing of the appropriate molar equivalents of the component DNA strands in a buffer solution that contained 100 mM Na+. Quantitative yields of the desired products Cb–Dax (x = 1–4, cubes with one to four Da strands positioned on one face) eliminated the need for purification of the DNA nanostructure prior to AuNP patterning (Fig. 2b).
With the Da-decorated cubic scaffolds Cb–Dax in hand, we proceeded to investigate their ability to bind to AuNPs and transfer the desired number of Da strands. Based on geometry calculations (Supplementary Section Va), 10 nm AuNPs (NP10) were used, as they represent the best size match for the cubic scaffold. Gold nanoparticles were incubated with Cb–Dax at a 1:1 molar ratio for 16 hours at room temperature. The resulting hybrid DNA cage–nanoparticle assemblies, NP10–Cb–Dax (Fig. 2c), were analysed by agarose gel electrophoresis (AGE) with reference to a control AuNP sample (Fig. 2d).
Electrophoretic mobility data for the reaction mixtures of cubes Cb–Dax with 10 nm particles NP10 (Step 1) are shown in Fig. 2d. Although the binding of Cb–Da1 and Cb–Da2 (cubes with one or two Da conjugates, respectively) to NP10 is inefficient, both Cb–Da3 and Cb–Da4 show a product that exhibits a mobility consistent with the NP10–[Cb–Dax] complex (see below for an elucidation). An increasing conjugation yield is observed with increasing Da number positioned on the cube, which suggests a cooperative effect because of the spatial organization on the template (see below and Fig. 3 for the identity of the minor product of lower gel mobility in Fig. 2d, Lane 4, as the AuNP with two bound cubes). Control experiments using cubes without Dx strands revealed no interaction with AuNPs, which confirms that binding was mediated by the Dx organized on the DNA cube (Supplementary Fig. F9).
To further stabilize the structures for downstream processing, passivation of the remaining surface of the AuNPs with a stable ligand was carried out (Fig. 2c). Carboxyl-terminated octaethylene glycol disulfide (OEG) chains were used to cover the AuNP surface. Interestingly, after this stage the patterned AuNPs were found to be remarkably stable in sodium buffer, and even in magnesium-containing buffers. This passivation step could be carried out before or after removal of the template scaffold with no observable difference to the assemblies (see Supplementary Fig. F11). The samples were then run on AGE and the product bands excised and isolated by electroelution. The template scaffold was removed by disrupting all the DNA hybridization under denaturing conditions (3 M urea, 1 × Tris-borate-EDTA (TBE)), followed by centrifugation to isolate the AuNPs. This process afforded purified NP10–Dax (see Step 2 in Fig. 2) with average isolated yields of up to 47% with respect to NP10. In principle, the constituent DNA strands of the cube may be isolated at this stage, and the cube reassembled for repeated use as a transient template.
To determine the number of Dx strands transferred to the AuNP, and thus the fidelity of the patterning process, extension strands (EXT-X) were used to provide greater gel mobility differences (see Step 3 in Fig. 2). Importantly, these also probe the addressability of the Dx strands bound to the AuNP surface. The EXT-X strands have a 17 nt region of complementarity with each Dx and an extending single-stranded DNA tail of 43 nt.
The gold nanoparticle that contains four transferred DNA strands, NP10–Da4, was incubated with an excess of EXT-A (Fig. 2f, Lane 3). The resulting sample exhibited a decreased mobility, which reveals that the Da strands transferred to the AuNP surface remain available for duplex formation and retain practical hybridization kinetics. The process was repeated for NP10–Da3 and NP10–Da2 in which the cube is bound to the nanoparticle via three or two DNA strands, respectively, and a higher-fidelity pattern transfer was observed with higher numbers of Da on the scaffold. In contrast, a non-templated control sample, prepared by incubating four molar equivalents of Da with 10 nm AuNPs, displayed a different behaviour. Hybridization of this structure to EXT-A showed a statistical distribution of products with between one and four Da strands bound to the AuNP (Supplementary Fig. F10). Dynamic light scattering (DLS) analysis of NP10–Da4 showed a structure with less than 10% polydispersity, compared with the non-templated control with more than 20% polydispersity, which corroborates the electrophoresis results and suggests regioselective control via the cube template (see Supplementary Section Ve for further details).
DNA prismatic cages can be varied readily with respect to their geometry. Each geometry should transfer a different pattern of DNA strands to the nanoparticle, with different numbers and spacing of the Dx on the AuNP surface (Fig. 3a). The preparation of Janus particles as a strategy to introduce anisotropy to spheres has been used to produce a range of asymmetrically functionalized AuNPs34,35,37,39,46. Although hemispherical separation can be obtained, the number and placement of DNA strands on the AuNP is not precisely controlled.
To investigate this geometric control, triangular prism (TP) and pentagonal prism (PP) templates were synthesized in the same manner as for the cube (Fig. 3b). The scaffolds were decorated with Da strands to produce TP–Da3 and PP–Da5, which were incubated with 10 nm AuNPs and then passivated with OEG (Fig. 3a).
The prism–AuNP constructs were analysed by AGE and a decrease in electrophoretic mobility was seen with increasing DNA scaffold size, consistent with the binding of the prisms to the AuNPs. In each case the second bands, highlighted in Fig. 3c, were determined to be 1:1 complexes, for example, NP10–[TP–Da3]1. The band of lowest mobility was assigned as the 2:1 product, for example, NP10–[TP–Da3]2, based on gel mobility and titration assays (Fig. 3e and Supplementary Fig. F12).
After purification and removal of the prism strands by denaturation, EXT-A was added to resolve the number of strands transferred to the AuNP. In Fig. 3d, electrophoretic mobility patterns reveal that three, four and five strands were robustly transferred by TP–Da3, Cb–Da4 and PP–Da5, respectively. Additionally, hydrodynamic radii of the patterned products, determined by DLS (Supplementary Section Ve), revealed a steady increase with an increasing number of strands on the nanoparticles, which is accentuated by the addition of extension strands EXT-A. Polydispersities were lower than those of non-templated gold nanoparticle conjugates (Supplementary Figs F17–F21 and Supplementary Table T5). These data, obtained for each step of the patterning process, demonstrate that well-defined numbers of DNA strands can be transferred to the AuNPs by geometric variation of the parent template.
Finally, we were interested in the potential of this strategy to transfer strands of different sequences to the nanoparticle. This confers particles with the ability for site-specific addressability and allows for anisotropic functionalization, and thus produces building blocks with great potential for the self-assembly of advanced materials27,28. A cubic scaffold with eight unique ssDNA-binding sites—a′, b′, c′, d′—was assembled and decorated with four unique gold-binding conjugates, Da, Db, Dc and Dd, on one face to give Cb-Asym (Fig. 4a). Incubation with 10 nm AuNPs followed by OEG surface passivation, isolation of the major product, NP10–[Cb-Asym]1, and subsequent removal of the cubic scaffold produced sequence-asymmetric NP10–Da1–Db1–Dc1–Dd1 (NP10-Asym). To confirm that the correct four-strand DNA sequence code had been transferred to the AuNP, stepwise addition of four unique EXT strands, complementary to Da, Db, Dc and Dd (EXT-A to EXT-D), was carried out as before. In Fig. 4b, the stepwise binding and site-selective addressability of NP10-Asym is shown. In Lane 3 of Fig. 4a, the addition of EXT-A produced a particle with one extension strand bound and Db, Dc and Dd available for further hybridization, E1. The subsequent addition of the remaining three extension strands revealed a ladder of decreasing mobility as each additional EXT binds to its specific site on NP10-Asym. This confirmed that the asymmetric pattern had been correctly transferred to the AuNP and the Dx are individually addressable because of their unique sequences.
A control sample of AuNP incubated with one molar equivalent of each Dx strand (Da–Dd) without organization on a template was prepared under the same conditions as the patterned sample. This should produce a statistical mixture of number and identity of Dx on the AuNP. Figure 4c reveals the stepwise addition of the EXT strands A–D. This highlights the control over information transfer possible through the use of the DNA nanoscaffold template.
Next, to probe both the sequence identity and relative positioning of the Dx strands, we used the cubic scaffold that contains both the sequence and geometry information needed to probe the exact configuration of the Dx strands on NP10-Asym. A sample of NP10-Asym was then incubated separately with two different cubic scaffolds that have either the correct four-sequence binding motif, matching the pattern of strands on NP10-Asym, or a mismatch control, which has four a′ and four b′ sequences, shown in Fig. 4d. The incubation was carried out at 75 nM AuNP, with 5 equiv. DNA structure with respect to AuNP, in a 1 × TAMg (Tris-acetate-magnesium) buffer, for 16 hours at room temperature. The resulting mixtures were then analysed by AGE. The results demonstrate that the patterned particle retains the molecular recognition (DNA sequence, shape and spacing) information granted by the template. In comparison, a non-templated control did not exhibit the same specificity (Supplementary Fig. F14), which strongly suggests that the pattern of DNA strands on the AuNP surface is anisotropic. Furthermore, the same experiment was carried out over multiple days, and the pattern was found to be stable for up to one month (see Supplementary Fig. F15).
To further characterize the asymmetrically functionalized nanoparticles, NP10-Asym, transmission electron microscopy (TEM) was used. First, to visualize the asymmetric pattern and regioselectivity by TEM, polyvalent 6 nm AuNPs (NP6) bearing complementary sequences (a′–d′) to the patterned NP10-Asym were used as site-selective probes. The parent particle NP10-Asym was incubated with an excess of NP6–Dx′, each with a sequence complementary to one of the strands on the nanoparticle (16 hours, room temperature, 1 × TAMg) to produce the structures T1–T4 shown in Fig. 5a. To aid microscopy analysis, the assemblies were purified by AGE prior to TEM analysis (see Supplementary Section VIf). TEM revealed that the desired satellite structures were generated successfully, with yields that decreased for the more-complex assemblies T3 and T4. The side products observed probably result from the ability of the polyvalent particles to act as crosslinks between NP10-Asym particles. Furthermore, interparticle distances were consistent with the expected values for T1 and T2; however, T3 and T4 showed mostly collapsed structures, possibly because of surface drying effects and Mg2+-mediated aggregation between the particles in such close proximity.
Although these assemblies highlight the sequence-specific addressability of the parent particle, NP10-Asym, inherited from the asymmetric cube template, an improved design was developed to aid the microscopy characterization. This uses rigid double-stranded DNA (dsDNA) arms of different lengths on monofunctionalized nanoparticles to probe each addressable site on the NP10-Asym. Complementary sequences to the four strands on the nanoparticles had different arm lengths, denoted A–D in Fig. 5b. These should produce interparticle distances of 30.4, 23.6, 16.8 and 10.0 nm. Assemblies were carried out with an excess of the 5 nm monoconjugates A–D, and purified by AGE for TEM. Figure 5c shows representative images of some of the anisotropic satellite structures that can be obtained from NP10-Asym and the four unique arm lengths. These assemblies exhibited fewer collapsed structures and misassembled products than the structures T1–T4. This is consistent with the longer, more-rigid dsDNA arms that distance the particles from each other, and with the DNA monoconjugated nanoparticles. Additionally, interparticle distances were measured for each of the assemblies and were found to correlate well with the expected values. A full statistical overview, including population histograms, is given in Supplementary Section VIh. Overall, this method represents a simple modular strategy to give access to many unique architectures using a small number of building blocks. More importantly, these experiments also reveal the potential of the information-rich patterned particles, NP10-Asym, to undergo selective DNA-mediated self-assembly, and showcases their viability for the production of unique DNA–AuNP assemblies.
An additional investigation into the asymmetric nature of the patterned AuNP, NP10-Asym, was carried out using the TEM data for the trimer species T2 and AB. The presence of four unique binding sites on NP10-Asym allows the positioning of two smaller nanoparticles in either adjacent or opposite relative positions. This produces a different spacing of the nanoparticles and angle between them, which can then be measured experimentally from the TEM data. A good correlation was found between the calculated and observed angles, which further confirms the retention of asymmetry of the patterned AuNPs. A full statistical analysis of angle measurements is given in Supplementary Section VIh and Supplementary Figs F24 and F25.
To further investigate the sequence specificity imparted to the NP10-Asym particles, a fluorescence assay was employed. A ‘barcoded’ particle with four different dye-labelled DNA strands was generated. Two fluorophores were used; 6-carboxyfluorescein (6-FAM) and cyanine 3 (Cy3), which have discernible excitation/emission properties. We synthesized four 5′-dye-labelled DNA sequences a′–d′, Cy3-a′, Cy3-b′, 6-FAM-c′ and 6-FAM-d′. Incubation of 1.5 molar equivalents of fluorophore to each unique binding site on NP10-Asym was carried out at room temperature for 12 hours. This was followed by three centrifugation and washing cycles, under native conditions, to remove any traces of unbound dyes.
Steady-state fluorescence spectra of the nanoparticle bearing all four dye-labelled DNAs, F4 in Fig. 6a, were measured and compared with control samples with the same concentration of free dyes in solution and in the presence of unbound AuNP, NP10(OEG). A decrease in fluorescence is seen in the presence of NP10(OEG) compared with the free dyes, which suggests some nonspecific binding. For the sample F4, a large decrease in emission is observed when the dye-labelled DNAs are bound to the nanoparticle by hybridization to their surface-bound complements. The fluorescence quenching observed is probably caused by both the AuNP and the close proximity of the dyes to each other47.
The efficient fluorescence quenching of the AuNP-bound dyes allowed a displacement assay to be monitored directly. Four specific eraser strands, Ea to Ed, were designed to remove the dye-tagged strands one-by-one from F4 by strand displacement, utilizing a 10 nt toehold region for specificity and rapidity. Figure 6d shows the cumulative addition of the different eraser strands, which release the dyes from the AuNP sequentially to give an increase in fluorescence emission of a specific dye at each step. For example, Ea is added to F4 to liberate Cy3-a′ and produce the structure F3. At this step, an increase in Cy3 fluorescence is observed with negligible change in the 6-FAM fluorescence. The following displacement steps were carried out by alternating between the two dyes, and in each case an increase in fluorescence for the specific displaced dye was seen. These results further support that four DNA strands with unique sequences were transferred to the AuNP by the patterning method and present a platform with unique properties well-suited to biosensing. In particular, this method can result in a very large and diverse set of ‘barcoded’ gold nanoparticles from a minimum number of dyes of different colours, which allows efficient multiplexing and direct detection in bioanalyte-sensing platforms48.
In summary, we have developed a method to encode gold nanoparticles with complex DNA-strand patterns comprising number, geometry and positioning of different unique strands. The use of DNA nanostructures as transient templates for the creation of patterned AuNPs is a general strategy that could be applied to a variety of inorganic nanoparticles. For example, the application of our approach to non-spherical particles could be used as a strategy for furthering complexity. This method also presents the advantage of potential recycling of the template or solid-support immobilization for the high throughput, scalable generation of patterned structures. From a DNA nanotechnology standpoint this represents an interesting avenue to explore for the role of DNA scaffolds in nanofabrication, as a potential way to reduce costs by using DNA nanostructures in a transitory fashion. As well as allowing the transfer of different geometrical patterns to the AuNP, it is possible that the size ratio between the different DNA scaffolds and the AuNP could be used for further control over surface spacing and angle between DNA strands. In principle, the transfer should produce a chiral arrangement of the DNA strands on the gold nanoparticles, useful for the development of chiral plasmonic structures; we are currently working to test this stereospecificity.
The patterned particles described have the distinct advantage that multiple strands with unique sequences can be placed on the nanoparticle. This opens the potential for molecular computation, such as algorithmic self-assembly, in which each nanoparticle could be treated as a ‘Wang tile’49,50. Alternatively, DNA computing could be applied in which strand-displacement cascades can be used to carry out logic operations51. As well as applications in colloidal assembly and biocomputing52, we predict that this method could also be applied to surfaces to guide nanoelectronic or photonic circuitry or create robust anchors for positioning other functional components. Additionally, the method is compatible with DNA/RNA aptamers, whose molecular recognition properties can also be exploited for targeted cellular delivery and diagnostics53. The ability to organize aptamers in specific orientations is known to be important for polyvalent receptor recognition54; this is an avenue that we will be exploring in future work. As AuNPs coated with nucleic acid have been shown to enact efficient gene silencing55, the ability to pattern different gene-silencing strands on a nanoparticle is an exciting prospect for nanomedicine.
Preparation of DNA nanostructure patterned AuNPs
A generic protocol was developed and applied for all prism/sequence variants; here Cb–Da4 is provided as an example. In a typical patterning experiment, bis(p-sulfonatophenyl)phenylphosphine) (BSPP)-coated AuNPs (20 pmol) were incubated with Cb–Da4 (20 pmol (3D construct)). The final sample concentrations were about 300 nM in a Tris (90 mM), boric acid (90 mM), EDTA (2 mM) and NaCl (100 mM) buffer (1 × TBEN, pH 8.3) with 2 mM BSPP. After incubation at room temperature overnight, 20,000 equiv. OEG (about 0.2 M in 1 × TBEN) was added to passivate the surface stably. After a 30 minute incubation with OEG at room temperature, samples were centrifuged (12,000g) at 4 °C, the supernatant was removed and fresh buffer added. This process was repeated at least twice before loading the samples on gel. Gel bands were assigned (see Figs 2 and 3) and the desired band carefully excised. Samples were extracted by electroelution in 1 × TBE buffer and subsequently concentrated by centrifugation (15,000g) at 30 °C and removal of supernatant. The ‘pellet’ was then washed with 3 M urea in 0.625 × TBE three times. This was followed by three washes with sterile deionized water. Finally, sample solutions were quantified by absorption spectroscopy, monitoring absorbance at 450 nm. Further experimental details, protocols and characterization data are given in the Supplementary Information.
The authors acknowledge the Natural Sciences and Engineering Research Council of Canada (NSERC), the Canada Foundation for Innovation, the Centre for Self-Assembled Chemical Structures and the Canada Research Chairs Program for financial support. T.G.W.E. thanks the Canadian Institutes of Health Research for a Drug Development Training Program scholarship. D.B. thanks the NSERC for a Bionano scholarship and C.J.S. thanks the NSERC for a Banting Postdoctoral Fellowship. H.F.S. is a Cottrell Scholar of the Research Corporation.