Photo-printing of faceted DNA patchy particles

Significance

Powerful new approaches to assemble colloidal particles have emerged over the past decade, including the introduction of shape complementarity and heterogeneous patchy colloid surfaces. While proteins use both approaches simultaneously, achieving this for colloids has proven difficult. Here, we introduce a scalable method to synthesize solid colloids with localized faceted DNA patches with different symmetries. Our synthetic approach starts with liquid droplets coated with DNA, which is then organized and localized on the droplets, producing facets, and then finally is photopolymerized to produce the desired solid faceted patchy particles. We show that these can self-assemble into a variety of new structures. This method adds a significant element to colloidal self-assembly which should find wide use.

Abstract

Patchy particles with shape complementarity can serve as building blocks for assembling colloidal superstructures. Alternatively, encoding information on patches using DNA can direct assembly into a variety of crystalline or other preprogrammed structures. Here, we present a tool where DNA is used both to engineer shape and to encode information on colloidal particles. Two reactive oil emulsions with different but complementary DNA (cDNA) brushes are assembled into CsCl-like crystalline lattices. The DNA brushes are recruited to and ultimately localized at the junctions between neighboring droplets, which gives rise to DNA-encoded faceted patches. The emulsions are then solidified by ultraviolet (UV) polymerization, producing faceted patchy particles. The facet size and DNA distribution are determined by the balance between the DNA binding energy and the elastic deformation energy of droplets. This method leads to a variety of new patchy particles with directional interactions in scalable quantities.

Shape complementarity and directional interactions are useful elements for controlling colloidal self-assembly (1–3). Both schemes are used by proteins, for example, to assemble complex structures such as viral capsids. Over the past decade, these two schemes have increasingly been employed to assemble synthetic colloids and nanoparticles into intricate new structures, as various synthetic methods to control particle shape and interactions have been developed.

For nanoparticles, shape complementarity and directional interactions are achieved by controlling which facets are exposed during particle growth (4, 5). Directional interactions and patchiness can also emerge in larger colloids as a consequence of lock-and-key complementarity (6), by compression molding (7), by liquid- or vapor‐phase deposition (8, 9), or by site-selective chemical modification (10). For example, encoding DNA on patchy particles offers highly selective directional interactions that expand the kinds of structures that can be made by self-assembly (11–13).

Using liquid droplets (14–17) as building blocks offers unique opportunities, as they consist of a deformable canvas, where mobile DNA brushes can be tethered and organized. Two such droplets can bind to each other through DNA hybridization, just as solid colloids do. If the binding is weak, the droplets remain spherical. If the binding is sufficiently strong, however, the contact zone between two droplets distorts and becomes essentially flat. Moreover, because the DNA tethered to the droplet surfaces is mobile, more DNA can move into the contact zone and bind, which causes the flat contact zone to grow. The equilibrium size of the contact zone is determined by balancing the entropic cost of concentrating DNA in the contact zone, the free energy gained by DNA binding between the two droplets, and the energy cost of distorting the interface from its natural spherical shape (18). DNA binding between droplets can thus cause droplet faceting (19). If the recruitment of DNA to the contact zone is sufficient, well-defined DNA patches are produced on the facets of droplets. Both the DNA patches and facets on the droplets are lost if the DNA strands dissociate and the droplets unbind. In that case, the droplets return to their spherical shape with uniform DNA coatings.

Here, we introduce polymerizable liquid droplets with DNA coatings that form faceted droplet arrays in CsCl-like crystalline lattices and then are converted into faceted DNA patchy particles by photo-polymerization in bulk. By varying the strength of the DNA binding, the degree of faceting and patch formation on liquid droplets can be precisely tuned. The approach is distinct from previous approaches in which particles were prepared with patches that were subsequently functionalized with DNA (10, 20); here, patchy particles are prepared and coated with DNA simultaneously. Our method consists of four steps (Fig. 1): First, two sets of monodisperse polymerizable liquid-emulsion droplets were prepared with mobile DNA coatings. Each set of droplets has DNA with sticky ends that are complementary to the DNA on the other set. Second, the two sets of emulsions were mixed and held at a temperature about 1 to 2 °C below the DNA freezing temperature T_f, such that colloidal crystals formed and annealed, similar to their solid counterparts (21, 22). Third, the temperature was lowered by 10 to 20 °C below T_f, increasing the strength of the DNA interactions, which led to the recruitment of the mobile DNA with sticky ends that were distributed around each droplet into the adhesion zones where the DNA on different droplets binds. The emulsions developed faceted patches at the adhesion zones, where, ultimately, nearly all of the DNA with sticky ends was localized. Fourth, the emulsion droplets were photo-polymerized, freezing in the facets functionalized with immobile DNA. The solid particles dissociated when the temperature was raised above the melting temperature T_m (T_m > T)_f, releasing the individual solid particles for further use. The method gives rise to faceted DNA patchy particles whose size, symmetries, and number of patches can be designed by controlling the crystal geometry and binding energies of the bulk emulsion colloidal crystal.

Fig. 1.

Schematic representation of faceted DNA patchy particle production as described in the text. Droplets containing photoinitiator are coated with cDNA to drive the binary crystallization of emulsions at an annealing temperature just below T_f. Mobile DNA on the surface of droplets enables duplex recruitment and droplet deformation at the adhesion zones (red circular areas) by cooling from annealing temperature. Rapid photo-polymerization of droplets captures recruited DNA, producing faceted DNA patchy particles.

Results

Monodisperse Droplets with DNA Coating.

We prepared monodisperse oil droplets of 3-trimethoxysilylpropyl methacrylate (TPM) monomers (23, 24), whose size, monodispersity, and liquidity (viscosity) can be tuned precisely (25, 26) by balancing monomer concentration, stirring rate, and pH.

Hydrolysis of TPM monomers and their condensation to form oligomers accelerate when the pH deviates from neutral (27). We used NH₃ to bring pH to ∼10, creating oligomers and reducing the solubility of the TPM, which produced droplets within 20 min that subsequently grew to the desired size of about 1 μm in less than 1 h. Excessive condensation of hydrolyzed TPM monomers made the droplet interior viscous and impeded droplet deformation, whereas weak polycondensation led to droplet polydispersity through Ostwald ripening, facilitated by the migration of water-soluble, low-molecular-weight oligomers between droplets. Our protocol for TPM droplet preparation (Methods) minimized polycondensation while maintaining high levels of monodispersity. The reactive acrylate groups in TPM droplets remained thermally stable during this process.

We coated fresh monodisperse TPM droplets with a dense, mobile layer of amphiphilic polystyrene-b-polyethylene oxide-b-DNA (PS-b-PEO-b-DNA) triblock copolymers, as illustrated in Fig. 2. The hydrophobic polystyrene (PS) blocks inserted into the droplets, but only if a volatile organic solvent (28, 29), dichloromethane (DCM), was added to the droplets. The hydrophilic polyethylene oxide (PEO)-DNA blocks remained within the aqueous medium. Once the PS blocks inserted into the droplets, the DCM was removed by evaporation, leaving behind DNA-functionalized TPM droplets (SI Appendix, Fig. S1). When prepared in this way, the stability and mobility of DNA-tethered diblock copolymers are easily preserved for weeks.

Fig. 2.

Schematic protocol for preparing UV-reactive droplets with DNA triblocks (PS-b-PEO-b-DNA). Highly monodisperse droplets are loaded with UV photo-initiator. Swelling the droplets with a volatile organic solvent, DCM, allows insertion of PS hydrophobic blocks. Subsequent removal of DCM deswells the droplets and traps the hydrophobic anchors, while keeping DNA triblocks fully mobile. DNA strands uniformly distribute on the droplet surface.

Typically, we coated droplets with two types of DNA strands, some with complementary sticky ends and others that were inert (noncomplementary). Coating droplets with a second inert set of DNA strands promoted droplet stability by assuring DNA brush coverage over the entire droplet surface, even when the DNA with sticky ends was recruited to the adhesion patches, as described above. Here, we used inert poly-T DNA. To promote stability, droplets were further coated with ionic and nonionic surfactants, sodium dodecyl sulfate (SDS) (1 mM) and Pluronic F108 (0.4 weight [wt] %), respectively. Thus, droplets holding complementary DNA (cDNA) can readily bind, melt, and rebind by adjusting temperature (Fig. 3A), while keeping DNA stably anchored to their surface.

Fig. 3.

Patch symmetry via binary crystallization of monodisperse emulsion droplets. (A) Freezing and melting profiles of droplets holding cDNA: Rapid DNA hybridization creates disordered assemblies with random DNA patches located on the surface of the droplets. At high temperatures (state 1), disordered assemblies melt (T above T_m), and droplets recover their uniform binding potentials. At state 2, the droplet mixture binds uniformly by slowly cooling, narrowing the melting profile and ordering assemblies (crystalline lattices) for T ∼ 1 to 2 °C below T_f. Further cooling of ordered assemblies from the annealing temperature activates uniform recruitment of DNA on the surface of the droplets (state 3). (B) Protocol for symmetric DNA patch design in bulk: 1) Droplets holding cDNA rapidly bind at room temperature to form random assemblies; however, for temperatures above DNA melting temperature, binding potential is weak compared with thermal energy. Image shows monodisperse DNA-coated droplets holding cDNA freely diffusing above melting temperature. 2) Keeping the mixture near the annealing temperature promotes crystallization of a BCC lattice—isostructural to CsCl atomic crystals. Just a few DNA strands hybridize to hold the crystal. Bright field shows the square symmetry of the BCC crystal. (Inset) Confocal fluorescence images show the (100) BCC plane. 3) Quenching a CsCl fluid superlattice from annealing to room temperature triggers recruitment of dsDNA and the formation of localized patches. The spatial distribution of patches for one droplet set within the CsCl lattice provides eight patches. (Inset) Confocal fluorescence images show patch localization on (110) and (100) planes. (Scale bars, 5 µm.)

DNA-Mediated Colloidal Crystallization of Droplets.

Fig. 3 illustrates our protocol to create colloidal crystals of DNA-coated droplets. In a typical experiment, we prepared a set of droplets with two orthogonal single-stranded DNA (ssDNA) sequences: One type, A or A′, drove the programmed binding interactions (A is complementary to A′), while the second, T, stabilized the rest of the droplet surface. Both DNA types (A or A′ and T) were mobile, uniformly distributed, and densely covered the droplet surfaces when no droplets with cDNA were present (∼1 ssDNA/20 nm²; Methods).

When droplets with complementary sticky ends, A and A′, were brought together, they bound to each other by forming double-stranded DNA (dsDNA) duplexes when T < T_f. For our samples, the DNA freezing temperature T_f was typically in the range of 35 to 45 °C, as shown in Fig. 3A. At room temperature, well below T_f, the binding energy between complementary strands (A to A′) easily surpassed the thermal energy. Hence, droplets coated with cDNA, one with (A′, T) and the other with (A, T), rapidly bound to each other, forming flexible clusters, owing to the high surface mobility of the tethered DNA. For solid colloids, cluster flexibility occurred only when a sample was held near T_f. As time proceeded, on the order of several seconds, the binary emulsion formed a dense arrested gel, and DNA migrated to small adhesion patches where binding to neighboring droplets with sticky ends was maximized. Typically, in a matter of minutes, all cDNA (A to A′) was recruited to these patches. Because DNA was recruited and concentrated at the adhesion patches, the particles unbounded (melted) only at temperatures well above the freezing temperature T_f. Freezing and melting curves are shown in Fig. 3A. Thus, at temperatures above DNA melting, dsDNAs melted, and droplets recovered their uniform isotropic binding potential (Fig. 3B).

Alternatively, when cooled from above T_f to about a degree below T_f, DNA-coated droplets (A, T) interacting with complementary droplets (A′, T) of the same size formed binary CsCl-like crystals with 1:1 stoichiometry, as confirmed by fluorescence imaging (SI Appendix, Fig. S2). Annealing of the crystal was rapid, typically on the order of tens of minutes, facilitated by the mobility of the DNA triblocks, which near, but below, T_f remained nearly uniformly distributed over the surfaces of the droplets. Crystal formation in DNA-coated droplets proceeded by nucleation and growth (30) (Movie S1). Fig. 3A shows a freezing curve that extends over about 10 °C, broader than the 2 °C typically observed for the crystallization of solid DNA-coated colloids.

Droplet polydispersity (>15%) altered the quality of crystals (31) and, therefore, the coordination of droplets within the crystalline emulsion. For instance, large droplets can irregularly nucleate crystals on the surface, whereas smaller droplets can be partially engulfed by the forming crystal (Movie S1). The production of monodisperse crystalline emulsion lattices requires careful control of droplet monodispersity and physical conditions for binding (e.g., salt concentration, surfactants, etc.; Methods). Movie S2 shows the lattice fluctuations of crystalline DNA emulsions with different droplet coordination. For instance, binary CsCl-like crystals formed from droplets with eight nearest neighbors exhibited greater lattice fluctuations than the closed packed face-centered cubic (FCC) structure crystals formed of droplets with self-complementary (palindromic) DNA with 12 nearest neighbors.

DNA Patches in Colloidal Crystals of Droplets.

After crystals were annealed near T_f, a quench to room temperature (T << T_f) caused the number of DNA linkages to increase and, thus, strengthened the binding between droplets within the crystalline lattice. In addition, as we shall see more clearly further on, the adhesion patches between droplets expanded and flattened to form facets as the DNA binding energy became comparable to the droplet-deformation energy. In Fig. 3 B, Bottom, an increase in fluorescent intensity can be seen at the patches where particles bind, although the flattening of droplet at the patches is not yet evident.

The DNA patches on each droplet reflect the coordination symmetry of the crystalline lattice. For example, small droplets (500 nm to ∼1 µm) forming CsCl-like crystals generate two sets of droplets, each having eight DNA patches, resembling “swollen” cubes whose vertices are functionalized with DNA. CsCl-like colloidal crystals made from large droplets (>1.5 µm) are only one or two layers thick, as the gravitational height is reduced to about 1.3 µm, or slightly less than a particle diameter (ρ_TPM droplet ∼ 1.18 g/cm³). When the lattice formation is restricted to a single layer, only four DNA patches are formed, as further described in the Faceted DNA Patchy Particles section.

Interestingly, bidisperse droplets of different size create a whole new family of patches (SI Appendix, Fig. S3). For instance, bidisperse droplets with a 1:2 size ratio and 1:1 stoichiometry produce NaCl-like colloidal crystals of droplets. SI Appendix, Fig. S3, II–IV shows how NaCl-like colloidal crystals of droplets produce two sets of droplets, both exposing four patches on their surface within a two-dimensional (2D) lattice. Likewise, bidisperse droplets with a 1:2 size ratio and 2:1 stoichiometry produce AlB₂-like colloidal crystals of droplets. SI Appendix, Fig. S3, V–VIII shows how large and small droplets create six and three patches, respectively, at the contact points within a 2D lattice. Patches in large and small droplets resemble hexagonal and triangular swollen prisms, respectively. In DNA-coated solid colloids, AlB₂-like crystals are preferentially formed over NaCl-like lattices when colloid size increases (32, 33). For DNA-coated droplets, however, both lattices are readily obtained, probably due to DNA mobility and improved crystal plasticity, which facilitates more open configurations with droplets adjusting their relative position within the lattice more freely. Forming crystals with two or more droplet sizes requires additional control over monodispersity. For instance, binary crystals can be formed with low monodispersity droplets; however, contact points and facets on patchy particles exhibit small variations in patch symmetry (i.e., patch-to-patch distance). If sufficient DNA is used for recruitment, binding energy on patches should remain constant.

Faceted DNA Patchy Particles.

Colloidal crystals of droplets can be photo-polymerized, which both solidifies the droplets and simultaneously fixes in place the hydrophobic anchors of the DNA triblocks (PS-b-PEO-b-DNA) so that they are no longer mobile (Fig. 4A). We used photo-polymerization, given the fast droplet hardening achieved by the rapid cross-linking of vinyl groups within TPM droplets. Here, a free radical cascade reaction is initiated by small quantities of light-activated initiator present in the interior of the droplets, which hardens all droplets exposed to ultraviolet (UV) in the order of minutes. The photo-initiator (Darocur 1173, Ciba) was loaded within the droplets at an early stage of droplet preparation (Methods). Thermally initiated hardening methods (e.g., 2,2-azobisisobutyronitrile and potassium persulfate) (SI Appendix, Fig. S4) are not useful, as the DNA patches are modified or completely disappear upon melting when the temperature is increased to initiate polymerization.

Fig. 4.

Faceted DNA-coated patchy particles from photo-polymerization of patchy crystalline emulsions. (A) Exposure of patchy crystalline emulsions to UV light triggers droplet polymerization capturing hydrophobic anchors embedded in the droplets to create patchy particles with active DNA. Bright-field image in A, Center shows a large CsCl-like colloidal crystal after photopolymerization. (Scale bar, 10 µm.) Fluorescent image in A, Right shows that DNA fluorescence is preserved when a radical scavenger is used. (A, Right, Inset) Red channel fluorescence image shows distribution of patches on the surface of the droplets. (Scale bar, 5 µm.) (B) Patches on particles reflect the symmetry of the patchy crystalline emulsion. SEM image of typical faceted DNA patchy particles obtained after photo-polymerization of droplets forming large CsCl-like crystals is shown. Patchy particle monodispersity is preserved after rapid droplet photo-polymerization. (Scale bar, 2 µm.) (C) Photo-polymerized 2D CsCl crystalline emulsions produce faceted DNA patchy particles with four selective energy sites. Fluorescence coupled with bright-field image shows 600-nm PS particles coated with cDNA (small green and gray particles) and four-patch faceted patchy particles (large gray particles) holding DNA complementary (red patches) to gray PS colloids. Small gray PS particles bind preferentially to faceted patchy particles due the high concentration of DNA on patches. (C, Insets) Bright-field images show rapid DNA assembly of a chain-like hybrid structure. (Scale bars, 2 µm.) (D) Fluorescence image of faceted patchy particles reveal the cubic distribution of contact points obtained from the original CsCl superlattice (eight patches). Binary mixture of DNA patchy particles can be separated by using DNA-coated magnetic particles (SI Appendix, Fig. S5). Fluorescence image and bright-field (D, Inset) image show 600-nm PS particles coated with cDNA selectively binding to DNA patches on faceted patchy colloid. (Scale bars, 1 µm.)

Photo-polymerization proceeded under low-UV irradiation (365 nm, ∼3.5 mW/cm²) to maximize internal cross-linking without damaging the colloid [e.g., structural defragmentation (34)] or DNA. UV-sensitive functionalizations [e.g., p-carbamoylvinyl phenol (35) and 3-cyanovinylcarbazole (36)] on DNA strands should be carefully avoided to eliminate intrastrand irreversible binding.

A small, but significant, fraction of the free radicals (37) generated by UV irradiation can escape the interior of the droplet during photo–cross-linking, so it becomes important to protect the DNA strands, which can be damaged by the free radicals, mostly by attacking pyrimidines (38). The free radicals can also cause bleaching of fluorophores. To address this problem, we neutralized the free radicals outside the droplets by incorporating radical scavengers in the aqueous medium. Sodium nitrate (NaNO₂) (86 mM) offers sufficient protection, as it can efficiently inhibit free radical polymerization of water-miscible monomers in the aqueous phase (39) and does not interfere with DNA binding like other inorganic radical scavengers (e.g., CuCl₂) (40, 41). Fig. 4 A, Right shows that fluorescence is preserved when a radical scavenger is used.

Fig. 4B shows patchy particles obtained after photo-polymerization of CsCl-like colloidal crystals of droplets. After droplet cross-linking, two sets of patchy particles are produced in bulk. Each set has two types of bound ssDNA, one concentrated on patches (binding DNA) and the other (nonbinding DNA) stabilizing nonpatchy regions. Two types of DNA patchy particles with four symmetric patches (Fig. 4C) were obtained by photo-polymerizing 2D CsCl-like colloidal crystals of relatively large droplets. One of the DNA patchy particle types can be separated from the other by mixing cDNA-coated magnetic particles and using a magnet. Alternatively, centrifugation can also be used (SI Appendix, Fig. S5).

The DNA on the faceted patches is highly concentrated (∼1 ssDNA/25 nm²), which is estimated assuming that all sticky DNA strands are recruited equally to all patches. More precise estimates are presented in the Tuning Facet Size section for several different conditions. The density of the DNA coatings on the patches is comparable to the density obtained for DNA-coated PS particles (∼1 ssDNA/14 nm²). Indeed, when two sets of PS particles coated with cDNA (Fig. 4C) were mixed with our solid faceted patchy particles (coated with DNA complementary to only one set of PS particles), assemblies of patchy particle and PS particles formed over PS–PS ones (Fig. 4C). Moreover, binding to DNA patchy areas of faceted particles was highly directional, as illustrated in Fig. 4D. Eight small PS particles exclusively bound to faceted patchy particles that hold eight active cDNA patches.

In principle, all DNA-coated droplets that form colloidal crystals become patchy particles after photo-polymerization. Nevertheless, the presence of edges and vacancies reduces the number of patches on droplets and on the subsequently extracted patchy particles. We minimized such effects by nucleating crystals with excess of one type of droplet (∼20%), followed by thermal annealing at shallow quenches. Thus, target patchy particles in the interior of the crystal were surrounded by complementary particles in excess. Using complementary particles in excess increased the yield of the targeted faceted patchy particle and prevented target patchy particles from having an incomplete number of neighbors. Magnetic separation readily removed particles in excess at an ∼1:1.2 ratio. Therefore, the yield of the faceted patchy particles was about 45% of the entire colloidal crystal; that is, almost half of the produced patchy particles were used in templating the bulk and surface of the colloidal crystal.

A typical capillary tube containing 10 µL of sample enables full emulsion crystallization, producing ∼10⁷ patchy particles. The method can be scaled up by crystallizing multiple samples simultaneously in larger capillary tubes. This can increase the sample volume up to ∼100 µL, or ∼10⁸ particles. Our typical DNA stock contains ∼50 nmol of DNA that can be used to coat ∼10¹⁰ particles. It should be possible to produce larger volumes (>100 mL) by using large containers with precise control over crystallization and UV reactors with low flux (∼3.5 mW/cm²) to preserve quality of fabricated patchy particles. Future strategies (e.g., TPM droplet size below 250 nm or density matching) can target lowering the use of templating or particles in excess in bulk. For instance, the production of large equilibrium crystals minimizes the surface-to-volume ratio, lowering the number of particles in excess needed and maximizing the number of patches on them. The histogram in SI Appendix, Fig. S6A shows the patch distribution in CsCl crystals grown to their equilibrium shape (Wulff polyhedron) for various crystal sizes. More than 98% of droplets produced eight patches after quenching CsCl-like colloidal crystals of droplets when crystal size reached ∼100 × 100 × 100. CsCl-like colloidal crystals exposed preferentially their {110} family of planes to minimize the surface energy of the crystals faceting rhombic dodecahedra (32). Thus, {110} planes will expose droplets with six patches, whereas droplets exposed in {100} and {111} planes will produce droplets with four and three patches, respectively. Patch distributions for NaCl and AlB₂ superlattices typically required fewer droplets to reach high yields of fully coordinated patchy droplets (SI Appendix, Fig. S6 B and C).

Tuning Facet Size.

The scanning electron microscope (SEM) photographs in Fig. 5A show typical solid particles with flat patches. Small DNA patches that are difficult to resolve with optical microscopy were easily identified by using an SEM. The size of the facets on the droplets reflects the competition between the total DNA binding energy and the excess surface energy required to locally flatten the spherical droplet interface. Increasing the DNA binding energy, say, by lowering the temperature to increase the number of hybridized DNA duplexes that span two droplets or by increasing the length of the DNA sticky ends, tends to flatten the interface between two droplets, as more DNA is recruited to the adhesion zone. Decreasing the surface tension, say, by increasing the amount of surfactant, including the surface density of inert DNA strands, also tends to flatten the interface at the adhesion patches.

Fig. 5.

Tuning facets on DNA patches. (A) SEM of typical faceted patches after droplet photopolymerization. (Scale bars, 200 nm.) (B) SEM images of small CsCl-faceted patchy particles. Increasing the colloid size from 250 nm (I) to 500 nm (II) in radius while keeping the same DNA strength (8-bp sticky end) produces an increase in patch size from 150 nm (I) to 295 nm (II). The connection of particle radius and DNA energy with patch size can be used to produce patches with similar energy and same angular size. (C) Recruitment of binding DNA on patches and body of faceted patchy particles can be estimated numerically (SI Appendix, Model). Contour lines represent predictions for patch sizes under simultaneous control of DNA strength and colloid size. Green circles represent experimental values of faceted patches holding eight patches. Model predictions are read as illustrated by red lines for patchy particle (I). (D) Increasing the length of the DNA sticky ends from 8 bp (III) to 16 bp (IV) while keeping the same colloid size (750 nm in radius) produces an increase in patch size from 169 nm (III) to 204 nm (IV). This illustrates a pathway to increase patch angular size on same-sized patchy particles. (E) Model predictions for patch size and DNA distributions for faceted particles with four patches. Blue squares represent the experimental values of faceted patches holding four patches.

Patch formation requires control over several parameters, including surface diffusion, binding strength, grafting density, surface tension, droplet size, and coordination with other particles. In equilibrium, the size of the DNA patches reflects a competition between the increased binding energy associated with increased DNA binding compensated by the energetic cost of deforming the spherical droplet and the entropic cost of localizing mobile DNA strands in the patch areas.

As already noted, we grew colloidal crystals from droplets with high-density DNA coatings just below the DNA freezing temperature, where droplet deformation is negligible. This promotes annealing and the growth of large single crystals, so that every particle has the same number of patches. The crystals were then cooled to room temperature, typically 10 °C to 20 °C below the DNA freezing temperature, which caused the patches to grow and form facets. To better understand facet formation, we adapted a simple thermodynamic model previously used to model DNA-coated droplets (18) to account for the different factors contributing to the patch size R_p:

$$R_p\approx {\left(\frac{{\text{Γ}}_{DNA}}{\gamma N_A}\right)}^{1/2}{\left[-\left(\mathrm{\Delta }G{°}_{DNA}-T\mathrm{\Delta }{S}_p-k_{B}T\hspace{0.17em}\text{ln}\langle g_{\beta }\rangle \right)+k_{B}T\hspace{0.17em}\left(1-q_{\beta }+2N_p\hspace{0.17em}\text{ln}\frac{q_{\alpha }}{q_{\beta }}\right)\right]}^{1/2}R.$$ [1]

tHere, R is the radius of the droplet; Γ_DNA is the number of DNA strands, both binding and inert, per unit area on the droplets; γ is the surface tension of the TPM droplets, N_A is Avogadro’s number; k_B is Boltzmann’s constant; and T is the temperature. The Gibbs free energy of DNA hybridization is ΔG°_DNA, the configurational entropy cost of forming a DNA duplex from two flexible strands is ΔS_p, and g_β is the average number of possible DNA bonds within the patch region. N_p is the number of patches, q_β is the fraction of binding DNA within a patch, and q_α is the fraction of binding DNA outside the patches. A more detailed description of the thermodynamic model is available in SI Appendix.

The solution of this nonlinear model yields a full description of patch size and DNA partition, patch vs. body, on the surface of a patchy particle. Experimentally, we started by prescribing simple parameters like the DNA distribution on a droplet (i.e., areal density and number of patches), DNA binding energy, and particle size. Numerically solving the model with these input parameters enabled us to predict various patchy particle designs. For instance, starting with two populations of droplets with R = 250 nm that have complementary eight-base pair (bp) DNA sticky ends, we assembled a CsCl crystal, which recruited DNA to eight patches when the temperature was lowered to 22 °C (Fig. 5B). When the radius of the droplets was increased to R = 500 nm, the patch size R_p increased from 75 to 168 nm, holding the areal density of DNA sticky ends constant at ∼1 ssDNA/25 nm² (Fig. 5B). In this case, the patch opening angle θ, defined in Fig. 5A, remained approximately constant (θ ∼ 35°) for colloids with different sizes.

Fig. 5C shows model predictions captured by Eq. 1 together with experimental points for patchy particles presented in Fig. 5B. Contour lines in Fig. 5C represent predictions for several patch radii under simultaneous control of DNA binding energy and colloid size. The fraction of binding DNA on a patch and on the body of patchy particles in Fig. 5B can be obtained directly from Fig. 5 C, Upper and Lower, respectively. Red arrows on Fig. 5C illustrate this for patchy particle (I) in Fig. 5B. Overall, Eq. 1 captures the linear increase of patch width with colloid size when all other parameters are kept approximately constant (e.g., number of patches, DNA areal density, and binding energy).

We also formed a bilayer of a CsCl crystal using two populations of large droplets (R = 750 nm), in order to restrict the number of patches on each droplet to four. When the DNA sticky-end length increased from 8 to 16 bp, the patch size increased by ∼20% (169 to 204 nm), as illustrated in Fig. 5D. In this case, colloids with the same size can produce different patch opening angles by controlling the DNA strength (e.g., 8 bp for θ ∼ 26° and 16 bp for θ ∼ 32°). Contour lines remained in good agreement with the experimental dependency of patch size with DNA binding energy (power ∼ 1/2) when the number of patches, areal density, and particle diameter were kept approximately constant (Fig. 5E).

Contour lines in Fig. 5 C and E illustrate additional ways to visualize patch design. For instance, contour lines represent multiple conditions where ΔG°_DNA, and R produced the same patch size. Hence, precisely adjusting the number of base pairs in DNA sticky ends along with colloid size offers unique and subtle control over patch sizes. Both full and partial DNA recruitment on patchy regions can prove relevant for future particle designs, where weak or strong interactions with the patchy particle body are desired.

While temperature annealing brings unique symmetries to our patchy particles, it also defines the limits of our tool. A natural way to produce particles with large facets (θ > 40°), which is key for the packing of nearly polyhedral particles (42), would require annealing TPM droplets coated with rather long DNA sticky ends (>16 bp). Small droplets (R = 250 nm) tend to be reaction limited if coated with 16-bp DNA sticky ends (i.e., binding is hindered, as the time to hybridize long DNA sticky ends is longer than the dwell time allowed by the rapid diffusion of small droplets). By contrast, large droplets (R = 1,500 nm) formed extremely strong DNA patches that never melted below 90 °C (TPM droplet stability limit). Nonetheless, opportunities for precise patch design within our experimental conditions (R = 200 to 1,300 nm, four to 16 DNA sticky end lengths) remain vast, in particular for patches with θ ∼ 20°, where assembly becomes highly directional (43). In the absence of temperature annealing, droplets formed random packings with a broad distribution of contact points and patch sizes (SI Appendix, Fig. S7). Yet, large facets (θ > 40°) and multitype DNA of patchy particles are still possible.

Besides providing a route to manufacture patchy particles, our method introduces additional advantages in the production of large colloidal assemblies. For instance, we can enzymatically ligate DNA on crystalline emulsions and thus remove the salt-screening dependence and confer temperature stability to the crystal. After DNA ligation, droplets can freely move in space with the entire lattice remaining connected by long flexible ligated “coils” (PS-b-PEO-DNA-PEO-b-PS). A solid crystalline lattice can be rapidly obtained upon photo-polymerization of ligated crystalline patchy emulsions (SI Appendix, Fig. S8). Likewise, templating emulsion coordination with external forces (e.g., acoustic, magnetic) could assist in circumventing DNA annealing to open further avenues for future exploration.

Conclusions

Here, we have presented a way to use DNA to fabricate faceted patchy particles. While DNA strands are better known for driving highly specific interactions, their hybridization energy can be high enough to deform liquid colloids and, as we have shown, can be used to program facets on solid particles with nanometer resolution. The directional interaction on droplet facets is not driven by chemical or compositional gradients like in model membranes (44, 45), but, instead, by energetically and geometrically designing colloidal interactions. The use of UV light to capture mobile DNA by droplet polymerization offers a rapid and versatile method to produce a wide range of patchy particles in large scale.

Methods

Droplet Synthesis.

We homogenously nucleated fresh monomers of TPM (≥98%; Sigma-Aldrich), 3-dimethoxymethylsilyl propyl methacrylate (DPM) (≥95%; Gelest, Inc.) in basic deionized water to produce liquid monodisperse droplets of desired sizes in less than 40 min. In a typical experiment, 2-µm TPM droplets were produced by adding 900 µL of TPM monomer to a solution of 100 µL of ammonia (NH₃) (28 wt %) in 100 mL of deionized water under mild stirring. One-micrometer TPM droplets were produced by using half the amount of monomer (450 µL). TPM droplets with radii below 250 nm exhibited decreased monodispersity when prepared via this sol-gel protocol in the absence of surfactant. DPM produced larger-sized droplets than TPM under the same synthesis conditions. The presence of fresh monomer is crucial in preparing monodisperse droplets and large crystalline emulsions. Droplets can be solidified by adding a small amount of initiator (∼1 wt %) to trigger radical polymerization.

DNA Strands.

High-performance liquid chromatography-purified DNA strands were purchased from IDT. Strand modifications (dibenzyl cyclooctane [DBCO], DBCO-triethyleneglycol [TEG], fluorophores [Cy5, maximum (max) emission 668 nm; Cy3, max emission 564 nm; Alexa 488, max emission 525 nm; and Alexa 647, max emission 665 nm], and 5′ phosphorylation [Phos]) were introduced in desired sequences as follows:

• A8: 5′(DBCO)–Cy3–(T)₁₂GGGACATA–3′

• A8′: 5′(DBCO-TEG) –(T)₁₂ TATGTCCC–Alexa647–3′

• B8: 5′(DBCO)–Cy5–(T)₁₂GGTGAGAG–3′

• B8′: 5′(DBCO)–Cy3–(T)₁₂CTCTCACC–3′

• A16: 5′(DBCO) –(T)₄GGG​ACA​TAG​GGA​CAT​A–3′

• A16′: 5′(DBCO) –Cy5–(T)₄TATGTCCCTATGTCCC–3′

• T20: 5′(DBCO) –(T)₂₀–3′

• P4: 5′(DBCO) –(T)₂₀-GCGC–3′

• LB6′: 5′–(A)₈CCA​CTC​TCC​TCT​CA–3′

• B8″: 5′(Phos) –GAGAGTGG(T)₁₂–3′(DBCO)

Clicking DNA to Azide Functionalized Diblock Copolymer.

Amphiphilic diblock copolymers, PS (2,200 g/mol)-b-PEO (11,000 g/mol) (Polymer Source Inc.), were first functionalized with azide (N₃) groups at the end of PEO blocks (PS-b-PEO-N₃) (28). Long DNA-tethered diblock copolymers (PS-b-PEO-b-DNA) were obtained by orthogonally reacting stoichiometric amounts of DBCO-functionalized DNA strands (100 µM 1× Tris–ethylenediaminetetraacetic acid [TE]) with PS-b-PEO-N₃ (1 mM, deionized water) at 55 °C for 24 h in a buffered medium (1× TE and NaCl [200 mM]). TE-buffered PS-b-PEO-b-DNA strands remain stable for weeks if stored at 4 °C.

DNA Droplet Coating.

In a typical experiment, 300 µL of freshly prepared droplets (1 wt %) in deionized water was swollen with 5% DCM (Sigma Aldrich) for 5 min before adding 10% PS-b-PEO-b-DNA strands (10 µM). The mixture was gently agitated for 10 min, before DCM removal at room temperature in a fume hood. In droplets holding multiple DNA sequences, target PS-b-PEO-b-DNA brushes were mixed before addition to droplets. The DNA excess was removed to eliminate DNA cross-talk during self-assembly (body-centered cubic [BCC] structure transfers to FCC under self-complementary interactions). SDS (1 mM) was added to preserve monodispersity during three consecutive centrifugation steps (Eppendorf MiniSpin) at 1,000 rpm for 5 min each. Droplets left to stand at room temperature can also sediment (12 to 36 h) to avoid shear forces during centrifugation. We quantified the number of DNA strands coating TPM droplets using flow cytometry (BD LSR II flow cytometer, BD Biosciences). DNA coverage was obtained by comparing the fluorescent output of DNA-coated droplets (8 bp, Cy5) with five cytometry standards (Quantum Cy5 Molecules of Equivalent Soluble Fluorochrome, Bangs Laboratories Inc.). Samples and standards were stabilized in 1× TE, F108 [0.4%].

In Situ Crystallization and Patch Distribution.

DNA-coated droplets were mounted in treated glass capillary tubes to self-assemble targeted crystalline lattices (NaCl [100 mM] [Sigma-Aldrich] and Pluronic F108 [0.4 wt %] [Sigma-Aldrich]). Glass capillary tubes (100 µm × 2 mm × 5 cm; Vitrocom) were etched by oxygen plasma for 30 min and coated with hexamethyldisilazane (Sigma-Aldrich) overnight. Capillary tubes were threefold washed with 1% Pluronic F108 and deionized water before sample mounting. Nonionic surfactants such as F68 and F127 decreased DNA surface coverage. Sample-loaded capillaries were sealed and attached to glass microscopy slides by using a UV fast-cure resin (Ludoc, Thick). Capillary tubes were protected with aluminum foil and edges with short strips of insulator tape during resin curing. Samples were mounted in a homemade microscopy thermal stage for crystal formation. Droplet aggregates (formed for rapid DNA binding at room temperature) were first melted and kept at high temperature (T_m +10 °C) for 10 min before slowly cooling (0.1 °C/min) down to crystal annealing temperature (T_m −1 °C). Targeted crystalline structures started forming in less than 10 min and were left to grow for 24 h. The same protocol was applied to different studied systems (SI Appendix, Fig. S9). Crystallizing droplets more than a week old can cause a small fraction of secondary nucleation during annealing. However, newly formed droplets did not interfere with DNA binding, as DNA remained strongly anchored in the original set of droplets. Freshly prepared droplets gave the best-quality crystals. Quenching grown crystals (10 °C/min) from annealing to room temperature triggered the recruitment of dsDNA at the surface of droplets, producing uniformly distributed DNA patches in space in less than 5 min. Crystalline lattices and DNA patches were stabilized by storing postquenched samples at 4 °C before imaging.

Crystal Ligation.

Five microliters of LB6′ strands (100 µM) were preannealed with 40 µL of droplets (1 wt %) coated with B8″ strands (phosphorylated DNA). Duplex formation (LB6′–B8″) proceeded at room temperature in buffered medium (1× TE, NaCl [500 mM], and 0.4 wt % Pluronic F108). DNA excess was removed by two cycles of sedimenting and redispersing in phosphate-buffered saline (PBS) 1×. BCC crystals were formed by mixing complementary A-coated droplets and (LB6′–B8″)-coated droplets in a buffered medium (1× TE, NaCl [100 mM], and 0.4 wt % Pluronic F108). We enzymatically ligated 17 µL of droplets forming open BCC lattices (0.1 wt %) by adding 1 µL of T4 Ligase (catalyzing the formation of phosphodiester bonds between 5′ phosphate and 3′ hydroxyl ends in dsDNA) and 2 µL of T4 Ligase buffer (50 mM Tris⋅HCl, 10 mM MgCl₂, 10 mM dithiothreitol, and 1 mM adenosine triphosphate, pH 7.5 at 25 °C) (New England Labs). Ligation reaction took place in less than 3 h at room temperature (22). Ligated droplets forming BCC crystals were threefold washed with SDS (1 mM) before sample mounting for imaging in treated capillary tubes. Ligated flexible crystals were stable without salt and did not melt with temperature. Ligated structures can be broken by vigorous shaking.

Droplet UV Photo–Cross-Linking and Faceted Patchy Particle Assembly.

The photo-initiator (1 wt %; Darocur 1173, Ciba) was loaded to droplets before the addition of PS-b-PEO-b-DNA brushes. In a typical photo-polymerization experiment, 17 µL of TPM droplets holding cDNA were mixed with 2 µL of NaCl (1 M), F108 (4 wt %), and 1 µL of NaNO₂ (86 mM). Once target patches were produced, samples were stored overnight at 4 °C. Samples were then exposed to UV (365 nm, 3.5 mW/cm²) at 5 cm for 15 min. Patchy particles were threefold washed in SDS (1 mM) in 1× TE buffer. For self-assembly studies, DNA-coated faceted patchy particles of interest were introduced in pretreated (see above) capillary tubes together with small PS colloids (600 nm) coated with cDNA. PS colloids were coated following the swelling–deswelling method (28). PS magnetic colloids (superparamagnetic 2.7 µm M-270 Dynabeads) were purchased from Invitrogen and threefold washed in deionized water before assembly. Patchy particle–PS assemblies were buffered in PBS 1×, F127 (1 wt %).

Microscopy Imaging.

Bright-field images were captured in a Nikon TiE widefield fluorescence microscope. Fluorescent confocal images were captured in a Leica SP8 confocal fluorescence microscope. Patchy particles were immersed in 70% glycerol, 1× TE, pH 8.4, and propyl gallate (10 mM) before imaging. SEM images were taken by using a MERLIN (Carl Zeiss) field-emission SEM. Dilute colloidal suspensions were dropped on silicon wafer cuts attached to SEM stubs, dried under vacuum, and coated with a 3-nm iridium layer for SEM imaging. Contrast and brightness of captured images were adjusted by using ImageJ.

Data Availability.

All data are included in the paper and SI Appendix.