As colloidal self-assembly increasingly approaches the complexity of natural systems, an ongoing challenge is to generate non-centrosymmetric structures. For example, patchy, Janus or living crystallization particles have significantly advanced the area of polymer assembly. It has remained difficult, however, to devise polymer particles that associate in a directional manner, with controlled valency and recognition motifs. Here, we present a method to transfer DNA patterns from a DNA cage to a polymeric nanoparticle encapsulated inside the cage in three dimensions. The resulting DNA-imprinted particles (DIPs), which are ‘moulded’ on the inside of the DNA cage, consist of a monodisperse crosslinked polymer core with a predetermined pattern of different DNA strands covalently ‘printed’ on their exterior, and further assemble with programmability and directionality. The number, orientation and sequence of DNA strands grafted onto the polymeric core can be controlled during the process, and the strands are addressable independently of each other.
Through their reliable, reversible hybridization, nucleic acids can guide the assembly of nanostructures with unparalleled precision1,2,3. The predictability of Watson–Crick DNA base pairing makes it possible to generate anisotropic and monodisperse nanostructures with a great degree of complexity, from two-dimensional patterns to three-dimensional objects such as tubes or polyhedra, but the functionality of these structures remains limited4,5,6,7,8,9,10,11. Synthetic polymers, on the other hand, provide stability and ease of functionalization, yet lack the high degree of programmability, predictability and monodispersity that DNA offers. As such, DNA amphiphilic polymer conjugates have recently received much attention as promising hybrid materials arising from the combination of the programmability and predictability of DNA with the stability and functionality of polymers, and have found use in numerous applications ranging from drug delivery to nanoelectronics12,13,14,15,16.
The self-assembly of DNA amphiphilic polymers mostly gives rise to symmetrical structures such as micelles and vesicles based on microphase separation of the two polymeric blocks17,18,19. Conversely, the creation of discrete and asymmetric nanostructures using DNA–polymer building blocks has not been examined, due to the difficulty in controlling the number, directionality and relative orientation of DNA strands grafted onto the polymeric core. Although polymeric micelles are now being manipulated with greater precision, both by varying their shapes and introducing anisotropy (for example, Janus, patchy particles or particles obtained by living crystallization), there is an ongoing challenge in generating a library of polymer nanoparticles capable of directional bonding20,21,22,23,24,25,26. This has been pursued recently with DNA-functionalized inorganic nanoparticles (NPs), with the development of new strategies to control the valency and bond anisotropy of NPs27,28,29,30,31,32,33,34,35,36,37,38,39,40,41. Our group, as well as Fan et al., have recently proposed strategies to transfer DNA patterns from three-dimensional DNA nanostructures or DNA origami onto gold NPs (AuNPs)42,43. The number of DNA strands and relative arrangement of DNA patterns in two dimensions can be controlled based on the Au–sulfur interactions. However, to our knowledge, three-dimensional DNA pattern transfer (and, more generally, DNA pattern transfer to polymeric particles) has not been reported so far.
Here, we describe a method to transfer three-dimensional DNA motifs from DNA cube structures, generating DNA–polymer particles using covalent chemistry. The DNA-imprinted particle (DIP) is ‘moulded’ on the inside of a DNA nanostructure and is composed of a monodisperse polymer core and a prescribed number of DNA strands in specific orientations (Fig. 1). As a proof of concept, we have ‘printed’ a specific pattern of six unique strands directly from a DNA cube scaffold to a polymeric core. We show that DIPs are stable under thermal, denaturing conditions and can be precisely controlled in terms of the number of DNA strands and their directionality, while preserving sequence anisotropy. These polymeric particles can self-assemble into well-defined structures using DNA base-pair recognition. Recent studies have demonstrated significantly increased control over polymer self-assembly via microphase separation and crystallization25. DIPs provide additional directional and programmable control and can potentially find numerous biological and materials applications.
Results and discussion
Design and working principle
Our patterning process relies on a DNA cube scaffold (Cb) as a template (Fig. 1a,c) and DNA–polymer amphiphiles complementary to the sides of this cube. We have previously shown that, when eight amphiphiles are positioned on Cb, they undergo an intrascaffold ‘handshake’ to create an internal hydrophobic pocket44,45,46. We hypothesized that, if the polymer portions on the inside of the cube were covalently crosslinked, this would produce polymeric particles that are covalently conjugated to the DNA strands hybridized to the cube. This DNA-imprinted particle (DIP) can then be released from the cubic scaffold by denaturation. Its core is defined by the number of strands and the nature and length of the polymer, which are precisely controlled (see Fig. 2 and Supplementary Section XXII), and it is functionalized with DNA strands of different sequences, presenting on its exterior in well-defined numbers and directions.
The cube scaffold (Cb) was prepared using a ‘clip-by-clip’ approach as previously reported (Fig. 1c)45,46. This scaffold was chosen because it presents several attractive features: (1) it has up to eight different binding regions (four on top and four on the bottom) and (2) it is possible to generate DIPs with different patterns (for example, number of DNA strands, sequences and directionality) using only one scaffold.
Another component required in the process comprises DNA amphiphilic polymers. These were prepared using automated solid-phase phosphoramidite synthesis18. We functionalized DNA strands with hydrophobic alkyl chains and reactive amine moieties in precise numbers and positions. The units consist of a hexa-ethylene (HE) segment and an amino group (Am), both of which feature an end phosphate group through which they can connect to each other as well as to the DNA strand (Fig. 1b). The resulting DNA–polymer amphiphile structure was optimized (by changing the numbers and positions of HE chains and Am units) to maximize the yield of the desired crosslinked product (Supplementary Section IV). The optimal strands contain six HE units and three Am groups in the sequence 5′-Am-(HE)3-Am-(HE)3-Am-DNA-3′ (that is, one at the end, one in the middle of the hydrophobic part and one between the DNA and polymeric portions) and are called ‘reacting strands’ (Fig. 1b). Recently, we showed that the hydrophobic cores of DNA micelles can act as nanoreactors that facilitate the conjugation of DNA with hydrophobic organic molecules in aqueous buffer47. We predicted that the hydrophobic core inside Cb could serve as a platform to carry out organic reactions that crosslink this core. The result is a polymeric nanoparticle that is covalently bonded to different DNA strands. To make a hexavalent particle inside the cube, six reacting strands (with Am groups) are needed, leaving two single-stranded regions on the scaffold. To form a hydrophobic environment within the cube, all eight binding sites on Cb need to be filled with amphiphiles. Therefore, two additional amphiphiles, which contain only six HE units but without any functional groups, called ‘filler strands’, are also introduced (Fig. 1b). The six DNA reacting strands (with Am groups) and two non-reacting filler strands (without Am groups) on Cb are expected to fold in, creating a hydrophobic pocket in the cube with multiple Am groups46. A crosslinking reagent (sebacic acid bis(N-succinimidyl) ester (C10-bi)) is then added to covalently ‘lock’ the core via amide bond formation at room temperature (Fig. 1a).
Figure 1d shows a native PAGE as the outcome of this process: (1) quantitative assembly of Cb decorated with eight amphiphiles (lane 3); (2) the structure remains intact after the crosslinking reaction (lane 4); and (3) no higher-order structures are formed, demonstrating that the crosslinking reactions occur only inside one cube. Interestingly, we observed previously that when eight amphiphiles with six HE units (HE6-DNA) are placed on the cube scaffold, its electrophoretic mobility is higher than that of a cube with eight unfunctionalized complementary DNA strands (Fig. 1d, lanes 2 and 3). This is most probably due to efficient folding of the hydrophobic chains upon introduction of the HE units on the decorating strands, which ultimately leads to a more compact structure (Fig. 4b)46. Finally, after the crosslinking process, the crude mixture was run on a denaturing PAGE to separate the desired hexavalent product (6x) from side products (divalent (2x), trivalent (3x), tetravalent (4x) and pentavalent (5x)) (Fig. 2a). Although 2x–5x are expected to be composed of different isomers, 6x should be a single product. The unreacted filler strands dissociate from the particle, leaving behind the prescribed number of DNA arms and HE core (Figs 2a and 5b). The core of 6x is equivalent to a polyethylene chain with exactly 216 repeat units, -(CH2-CH2)-216; thus, it is monodisperse.
Characterization of the formation of 6x
The denaturing PAGE in Fig. 2a shows that, after crosslinking, a mixture of products with different numbers of DNA strands (from divalent to hexavalent) is formed. The yield of the formation of 6x is 30% based on band intensity analysis, corresponding to at least 88% efficiency in individual reactions between amino groups and NHS moieties (Supplementary Section IX; 40% yield for a DIP with four DNA strands, see Fig. 5b), and 6x can be recovered at up to 92% by electroelution after gel extraction. With higher-yielding coupling reactions, we expect this efficiency to increase further. In addition, the formation of the crosslinked 6x product was confirmed by atomic force microscopy (AFM) imaging in air and dynamic light scattering (DLS). Our previous studies showed that HE6-DNA amphiphiles can only self-assemble into spherical micelles in the presence of Mg2+, which compensates for the electrostatic repulsion between phosphate groups18. In contrast, 6x in water (without Mg2+) could still maintain its integrity with a diameter of 22.2 ± 3.8 nm (measured by AFM; Fig. 2b) and 30.6 ± 5.7 nm (measured by DLS; Supplementary Section XVI), which strongly supports that 6x is covalently crosslinked. The difference in size of 6x obtained by AFM and DLS could be due to repulsion between phosphate backbones in a salt-free environment. Despite its high polyethylene content, the particle appears compact in the AFM images, most probably due to its phospholipid-like folding/compaction and from its collapse on the mica surface. Moreover, the ‘printed’ particle was still intact after incubation at 95 °C for 2 h, as analysed by denaturing PAGE (Supplementary Fig. 3).
To further characterize the number and addressability of the DNA strands transferred onto the crosslinked micelle, six fully complementary strands (comp 1–6), which hybridize with each of the six DNA unique strands, were sequentially added to the printed particles. A sequential decrease in electrophoretic mobility of the structures indicated the successful hybridization of individual comp 1–6 to 6x (Fig. 2b, lanes 1–7). In addition, as the complementary strands were added, the width of the band narrowed, probably because the printed particle's polymeric core interacts less with the gel matrix. Furthermore, the approach can be expanded to different cage geometries (for example, trigonal prism or pentagonal prism), which tunes the degree of polymerization of the polymer core, demonstrating its versatility (-(CH2-CH2)-n where, for example, n = 180, 216 or 252, Supplementary Section XXII). To our knowledge, this is the first example of a monodisperse DNA–polymeric particle featuring a specific number of unique and addressable DNA sequences.
Rebinding 6x to ‘correct’ and ‘incorrect’ scaffolds
Following the formation of 6x, we sought to determine if the relative orientations of DNA strands were maintained after our patterning process. To do this, we carried out a series of rebinding experiments of crosslinked 6x to the cube scaffold42.
First, the correct cube scaffold (Cb), with sequences and geometry matching 6x, was prepared separately, then incubated with 6x (in excess) at room temperature for 16 h. Two HE6-DNA strands complementary to the remaining cube sides were also added (filler strands) to fill all eight binding regions on the Cb and trigger the intrascaffold handshake (step 1, Fig. 3a). The 6% native PAGE in Fig. 3d shows a single band for the resulting rebinding product (RP) (lane 2), indicating high rebinding efficiency. Moreover, RP has the same mobility as a control crosslinked cube structure (Cb-A), implying that it has the same compaction degree and a similar structure (see Supplementary Fig. 6 for additional verification). This suggests that the printed particle has its six DNA strands bound to their correct positions on the cube.
As the amphiphiles have a 5T spacer and are hybridized to only 14 bases of the 20-base edges on the DNA cube, a strand displacement strategy can be used. DNA strands with fully complementary sequences to the strands of 6x were added to the RP sequentially. This resulted in the liberation of 6x hybridized with six complementary strands (6x-6) from Cb (Fig. 3c). Indeed, only after the addition of all six complementary strands (comp 1-6) to RP did we observe full dissociation of the particle from the cube scaffold (Fig. 3e, lane 10). This unambiguously confirmed that 6x had rebound completely to each of its six binding regions on Cb. Interestingly, we observed that 6x could fully recognize the cube scaffold, even after being heated at higher temperature (Supplementary Figs 3 and 4). This implies that the DNA pattern of 6x is still maintained at higher temperature.
To further verify the asymmetric nature of our ‘printed’ particle (6x), we designed a rebinding experiment in which 6x was incubated with an ‘incorrect’ scaffold. This scaffold (Cb-wrong) contains six binding regions complementary to 6x, but they are presented in a spatial arrangement different from that of the correct template cube (Cb), so only three contiguous binding sites on Cb-wrong can be accessed by 6x. Interestingly, 6x could not bind completely to the wrong scaffold due to its different spatial arrangement; instead, two 6x particles partially bound to the incorrect cube, as characterized by non-denaturing PAGE, DLS and several control experiments (Fig. 4a and Supplementary Sections XIV–XVI).
Molecular mechanism of polymer self-assembly and crosslinking inside the DNA cube
To gain insight at the molecular level, we performed computer modelling with molecular dynamics simulations of our cube scaffold (Cb) decorated with amphiphiles. Our simulations started from two distinct conformations: the Cb scaffold with a prefolded hydrophobic portion (control), and the scaffold with an extended hydrophobic portion of amphiphiles. The latter may better represent the annealing conditions of the experiment. Within 20 ns, we observed simultaneous fast folding and self-assembly of the HE chains in the simulations, starting from the extended hydrophobic portion, but relatively slow self-assembly in the simulations starting from the prefolded hydrophobic portion (Fig. 4b). With the help of the cube, it is likely that the extended HE chains can easily form a hydrophobic core inside the cube, which is further stabilized by rearrangements of the charged and hydrophobic groups during polymer folding (Supplementary Section XXI and Supplementary Movie). To our surprise, this phenomenon is quite different from many natural peptides, in which the self-assembly occurs faster than folding48. We also showed, by significantly reducing the diffusion, that the cube facilitates polymer self-assembly as well as crosslinking. Interestingly, during the self-assembly process, we observed distortion in the length of the DNA cubic edges, which further supports our discussion earlier about compaction of the cube (Figs 1 and 4b and Supplementary Fig. 20). Along with the self-assembly process, there is an increasing probability that the Am groups from hydrophobic portions have a separation within 15 Å, a distance that allows crosslinking. Toward the last 1 ns of the simulations starting from the extended HE chains, almost every Am group on the reacting strands has another in the distance range that is ready for crosslinking (Supplementary Fig. 21).
Controlling the valency
We were interested in controlling the number of DNA strands transferred onto the polymeric core. Although this has been examined previously with gold nanoparticles29, it is still a great challenge to precisely control the DNA valency and sequences grafted onto a polymeric core. The ability to control the number and sequences of DNA strands on a nanostructured object is of great importance (for example, in drug delivery, in controlling the nature and number of cell-targeting ligands on a nanomaterial)29. By deliberately hybridizing a predefined number of reacting strands (for example, x strands where x ≤ 8) and 8 − x non-reacting filler strands (HE6-DNA) on the cube scaffold, after crosslinking using C10-bi, an x-valent product was formed and purified (Fig. 5a). Our cube scaffold has the potential to be recovered and reused for ‘printing’ to increase the scalability of this approach. We were particularly interested in making printed patterns with two to six unique DNA arms because the synthesis of these structures is not trivial. The divalent (2x), trivalent (3x), tetravalent (4x) and pentavalent (5x) products were formed with 60, 51, 40 and 35% yields, respectively (Fig. 5b; see Supplementary Section XVII for gel characterization and Supplementary Section XXIII for octavalent particle (8x) formation).
Self-assembly of printed particles 6x into higher-order discrete structures
An important challenge is to generate complex and well-defined polymeric structures by self-assembly49,50. We were thus motivated to explore the possibility of using them as new building blocks in assembling nanostructures based on polymeric particles. We set out to create a dimer and a trimer from the printed particle 6x as a proof of concept. Connector 1, which is a 42-base-pair DNA duplex with 14-base sticky ends at both ends, was introduced to bring together two DIPs with a designed configuration, separated by 70 bases, leading to the formation of dimer product (Fig. 6a). To make DIP trimers, we added more DIP and connector 2 (same design as connector 1). This allowed the positioning of two other DIPs in adjacent relative positions on the central particle (see Fig. 6a,c-1 for positioning). Native PAGE revealed successful assembly, as demonstrated by the decrease in mobility shifts (Supplementary Fig. 15). Dimer and trimer assembled structures were further characterized by AFM (in air) in two separate experiments with yields of 68% (N = 196) and 45% (N = 345), respectively (Fig. 6a and Supplementary Section XVIII). The difference in yield when compared to PAGE is probably a result of AFM sample preparation. Moreover, the distance between two printed particles was measured to be 22.0 ± 2.5 nm (N = 100), which is in accordance with the expected length of a 70-DNA-base-pair duplex connecting the particles (∼23.8 nm).
The asymmetric nature of the 6x was further examined using AFM for the construction of trimer structures. The central particle allows us to place two other particles in positions ‘opposite’ to one another (by using two DNA arms transferred from two opposite cube edges), resulting in wider observed angles of the trimers (Fig. 6b and Supplementary Fig. 16). We expect that if the relative placement is maintained, we should observe a wider angle in type B trimer compared to type A trimer (Fig. 6c). Indeed, by analysing angular distributions in each type of trimer structure, we observed that the trimers with configuration B have wider angles than trimers with configuration A (Fig. 6a,b).
Finally, self-assembled tetramer structures based on DIPs were prepared (yield of 40%; N = 183) with three particles connected to the central particle, by adding more DIP and connector 3. Unambiguously, we observed a wider angle between particles 1 and 3 (two DNA arms are originally from opposite edges of the cube) than between particles 1 and 2 or between particles 2 and 3 (two DNA arms are originally from adjacent edges of the cube) in the tetramer (Fig. 7 and Supplementary Section XX). Moreover, the angle between particles 1 and 2 is statistically similar to the angle between particles 2 and 3, reflecting the adjacent positions of the DNA strands on the printed particles. This further confirms retention of the relative geometric patterns after being transferred from the cube scaffold. For our statistical analysis, we only counted tetramer species presenting all three arms on the mica surface. The yield of 40% tetramer may reflect the manner in which the structures land on the mica surface, as their arms can be turned towards the surface and may thus appear to have fewer arms (for example, they may look like trimers, although they are tetramers).
We have demonstrated the first general method to transfer DNA patterns in three dimensions, directly from a DNA cubic scaffold to well-defined polymeric particles. We have shown that our printed particle (6x) has exactly six unique DNA strands grafted on its HE core and that its valency can be controlled exactly. The polymer nanoparticle is ‘moulded’ inside the cage, with a precise and tunable number of repeat units. More importantly, the hexavalent printed particles preserve the orientation and sequence anisotropy obtained from the DNA cubic scaffold, which was demonstrated using scaffold rebinding as well as hierarchical assembly experiments. We anticipate that this method will allow the self-assembly of colloidal particles in which the printed particles can serve as precisely-defined ‘six-arm junctions’ to create highly complex nanostructures in a predictable manner. Furthermore, this method can be used in targeted delivery and diagnostics in the future due to the addressability and monodispersity of the resulting particles. More specifically, our ‘printed’ particle could potentially be functionalized with targeting moieties with chosen three-dimensional orientations, which is of great importance in polyvalent receptor recognition51. Applications such as drug delivery vehicles, asymmetrically substituted nanomaterials to control cellular processes, barcoded diagnostics or building blocks for non-centrosymmetric polymer patterning are anticipated. These directions of research as well as scaling up of the printed product (for example, by immobilizing the scaffold on bead) will be our main focus in the future.
Preparation of hexavalent printed particle (6x)
Four clip strands (1AB, 2CD, 3AE, 4IH; for sequences see Supplementary Sections I–IV) and DNA–polymer conjugates (six reacting strands at B, C, D, E, I and H sides and two non-reacting strands at A sides) in the appropriate ratio (1 equivalent each with respect to binding site) were mixed in a solution of 1× TAMg to obtain a final concentration of 500 nM. The 1× TAMg buffer was composed of 45 mM Tris and 12.5 mM MgCl2.6H2O, with pH adjusted to 8.0 using glacial acetic acid. The sample was mixed and annealed from 95 to 4 °C over 4 h to form the cage-micelle structure. Then, 10 mM of sebacic acid bis(N-hydroxysuccinimide) (C10-bi) dissolved in THF was added to the cage-micelle solution. The reaction mixture was stirred at room temperature for 16 h before purification by 12% denaturing gel (500 V, 2 h). The desired band was assigned as in Fig. 2a, excised carefully, and recovered by electroelution in 0.5× TAE buffer (20 mM Tris, 10 mM acetate and 0.5 mM EDTA).
The data collected and reported in this study, including all LC-MS, PAGE, DLS, AFM and modelling, are available upon request from the correspondence author (including data presented in the main text and in the Supplementary Information).
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
The authors acknowledge the Natural Sciences and Engineering Research Council of Canada (NSERC), the Canadian Institutes for Health Research, the Centre for Self-Assembled Chemical Structures (CSACS), the Qatar Research Foundation (project no. NPRP 5-1505-1-250) and the Canada Research Chairs Program for financial support. H.F.S. is a Cottrell Scholar of the Research Corporation.
Supplementary Movie 1
About this article
Nature Nanotechnology (2018)