Impact of Heterogeneity and Lattice Bond Strength on DNA Triangle

Impact of Heterogeneity and Lattice Bond Strength on DNA Triangle Crystal Growth Evi Stahl,† Florian Praetorius,† Carina C. de Oliveira Mann,‡ Karl-Peter Hopfner,‡ and Hendrik Dietz*,† †

Physik Department and Institute for Advanced Study, Technische Universität München, Am Coulombwall 4a, 85748 Garching near Munich, Germany ‡ Department of Biochemistry and Gene Center, Ludwig-Maximilians-Universität, Feodor-Lynen-Str. 25, 81377 Munich, Germany S Supporting Information *

ABSTRACT: One key goal of DNA nanotechnology is the bottom-up construction of macroscopic crystalline materials. Beyond applications in fields such as photonics or plasmonics, DNA-based crystal matrices could possibly facilitate the diffraction-based structural analysis of guest molecules. Seeman and co-workers reported in 2009 the first designed crystal matrices based on a 38 kDa DNA triangle that was composed of seven chains. The crystal lattice was stabilized, unprecedentedly, by Watson−Crick base pairing. However, 3D crystallization of larger designed DNA objects that include more chains such as DNA origami remains an unsolved problem. Larger objects would offer more degrees of freedom and design options with respect to tailoring lattice geometry and for positioning other objects within a crystal lattice. The greater rigidity of multilayer DNA origami could also positively influence the diffractive properties of crystals composed of such particles. Here, we rationally explore the role of heterogeneity and Watson−Crick interaction strengths in crystal growth using 40 variants of the original DNA triangle as model multichain objects. Crystal growth of the triangle was remarkably robust despite massive chemical, geometrical, and thermodynamical sample heterogeneity that we introduced, but the crystal growth sensitively depended on the sequences of base pairs next to the Watson−Crick sticky ends of the triangle. Our results point to weak lattice interactions and high concentrations as decisive factors for achieving productive crystallization, while sample heterogeneity and impurities played a minor role. KEYWORDS: DNA nanotechnology, DNA crystal, DNA tensegrity triangle, crystallization

P

crystal matrices could host organic and inorganic objects for applications in various fields such as photonics, plasmonics, or catalysis.17−21 Moreover, the ability to grow crystals would open access to atomic-resolution structural analysis of the building blocks by X-ray diffraction. High-resolution structural feedback would help designing DNA origami mechanisms and realizing complex functions such as molecular recognition or even catalysis with designed DNA objects. However, crystallizing DNA origami into well-ordered macroscopic 3D lattices remains an unsolved problem. Negative results are rarely reported in the scientific literature, but we have attempted crystallization of DNA origami in various ways and have not been successful so far. A typical outcome in our experiments was the formation of aggregates or assemblies such as spherulites that lacked long-range order. One of the potential issues that could affect the ability to crystallize arising with the multichain objects such as DNA

rogrammable molecular self-assembly with nucleic acids offers a path to fabricate custom-shaped objects with nanometer dimensions.1 Various approaches to design and assemble such objects have been developed that enable the construction of increasingly large and complex objects. With DNA origami2−4 and the tile−brick approach,5,6 the sequences of hundreds of DNA strands may be programmed to selfassemble into discrete objects that include several thousand DNA base pairs at user-defined positions and that reach molecular weights on the megadalton scale. Using DNA origami, prototypical shapes and mechanisms have been developed for various purposes, for example, for encapsulating and controlling the accessibility of molecules,7,8 for scaffolding enzymes to enhance the activity of reaction cascades,9 to interact with membranes,10,11 and for various purposes in single-molecule biophysics12−14 and in structural biology.15,16 DNA origami and tile−brick objects represent attractive candidate building blocks for the bottom-up assembly of macroscopic 3D crystalline materials. By engineering the molecular building blocks, properties such as lattice geometry, chemical composition, and porosity could be controlled to achieve desired macroscopic functionalities. Designer DNA © 2016 American Chemical Society

Received: July 18, 2016 Accepted: September 1, 2016 Published: September 1, 2016 9156

DOI: 10.1021/acsnano.6b04787 ACS Nano 2016, 10, 9156−9164

Article

www.acsnano.org

Article

ACS Nano origami concerns sample heterogeneity. Self-assembly reactions for DNA origami may yield heterogeneous products of various types, including partially folded or geometrically different structures having missing or incorrectly integrated strands, in addition to higher-order objects such as oligomers and aggregates. Heterogeneity may also be induced by chemical imperfections such as truncations of the DNA oligonucleotides used in the self-assembly reactions. Another issue that could affect the ability to crystallize concerns the strength of the interactions between particles. Oligomerization in one or two dimensions is commonly achieved by hybridization of several complementary DNA motifs, each multiple base pairs long.22,23 Such interactions may be too strong for crystal growth in three dimensions (3D). In addition, single-stranded defects24 that are randomly dispersed across the particles’ surfaces may introduce additional sticky sites that could disturb the growth of ordered crystals. This study is therefore concerned with the impact of chemical, geometrical, and thermodynamical sample heterogeneity on the ability to crystallize in the first place (and less with the diffractive properties of crystals thus obtained, if any) and with marking out a range of Watson−Crick bond strengths that allow crystal growth. To carry out these investigations, we rely on a model system that crystallizes via Watson−Crick hybridization bonds. Crystals suitable for X-ray diffraction have been grown successfully for several smaller DNA objects with molecular weights below 50 kDa.21,25−38 Among those objects is a class of multistranded objects known as the DNA “tensegrity triangle”,31,39 which is one of two systems31,38 that crystallize in 3D lattices with designed lattice geometry. The triangle consists of three different DNA single strands, termed L, M, and S that occur in the triangle in a 3:1:3 stoichiometry (Figure 1a). Once triangles have formed, they can interact via Watson−Crick complementary sticky ends to form a 3D crystalline lattice. Here, to study the impact of heterogeneity, we used a variant of the triangle having 5′ G and C single-base overhangs as the crystallizing model multichain object and studied crystal growth in mixtures containing the original and also modified triangle DNA oligonucleotides. To mark out a window of suitable Watson−Crick bonding strengths, we modified the interfacial sequences of the triangle.

Figure 1. DNA triangle crystallization as a function of strand concentration and strand stoichiometry. (a) Schematic two-step representation of the triangle crystallization reaction pathway. The varying line thicknesses in the schemes indicate structural elements that are nearer (greater line thickness) or further away (smaller line thickness) from the observer, to add a three-dimensional perspective, for example, as in a Haworth projection. The structure was drawn based on 3gbi.pdb31 using UCSF Chimera.56 (b) Typical images of crystals grown at the indicated concentrations (M strand concentrations), acquired after 10 days (top) versus 150 days (bottom) of incubation. Scale bar is 100 μm. (c) The crystal score considers factors such as crystal size, aspect ratio, defects in edges, and defects in faces (Figure S1) but does not reflect the diffractive properties of the crystals. The highest score of 100 signifies a rhombohedral crystal with at least 200 μm edge length, without obvious defects and with straight edges and faces. X-axis gives effective concentration of M strands in the equilibrated droplets. Top axis label is included to facilitate comparison to results in Figures 2 and S10 where “mock” (here water) was replaced by competing strands. Symbols (circles, squares, triangles pointing to the left, upward, or downward) give crystal scores from five independent crystallization assays; each score value is an average from at least three droplets studied at identical conditions. Bars give standard deviation from mean score. Dashed line is a guide to the eye, to indicate the transition to the regime of stochastic crystal appearance at low strand concentrations. (d) Typical images of crystals grown at varying ratios of the L, M, and S strands in crystallization droplets. Three titration series for each strand (L, M, S) were prepared, in which the concentration of the particular strand was adjusted according to the indicated stoichiometric factors (0.3−2) with respect the default ratio L/M/S = 3:1:3. Images were acquired after 150 days. Scale bar is 100 μm. (e) Symbols give crystal scores for data in panel d as achieved after 10 days of incubation: diamonds, L; triangles, M; squares, S. Error bars give the standard deviation of the crystal score in four independent drops within the same chamber.

RESULTS AND DISCUSSION All crystallization assays performed herein were carried out in the same fashion: components were mixed at various concentrations for droplet crystallization assays with 10-fold volume reduction via vapor exchange equilibration of the droplets with the buffer reservoir. If not otherwise noted, the reservoir contained 250 mM Tris-HCl at pH 8.5 and 200 mM MgCl2, but no additional cofactors (Note S1). The crystal plates were first annealed for 4 days and then incubated at constant temperature to monitor crystal growth for up to 150 days via bright-field and fluorescent imaging. We employed a crystal-scoring scheme that evaluates the overall shape and size of crystals obtained to facilitate comparison of the results of several crystallization experiments. The score makes no statement about the diffractive properties of the crystals (Figure S1). We also studied a subset of the crystals in diffraction experiments (Figure S2). To investigate whether particular design modifications could self-assemble into triangles, we relied on gel-electrophoretic mobility analysis (Figures S4−S9). To analyze whether modified strands were integrated into the crystal lattice, we used fluorescence tracing experiments in

which the modified strands were additionally fluorescently labeled (Figures S3−S9). In the case of strand incorporation, fluorescent crystals were obtained. However, the presence of 9157

DOI: 10.1021/acsnano.6b04787 ACS Nano 2016, 10, 9156−9164

Article

ACS Nano

Figure 2. Studying DNA triangle crystallization in heterogeneous environments. (a) Specifically modified triangle strands (red) were added in addition to the default triangle strands to triangle crystallization droplets to induce heterogeneity (also see Figure S10). (b−f) Typical images of crystals grown in the presence of increasing content of modified strands (mod) and decreasing content of corresponding original strands (original). The total strand concentration was kept constant and the stoichiometry of triangle strands was maintained to meet a ratio of 3(L + L′)/1(M + M′)/3(S + S′). Schematics give details of the triangle modifications to create truncations (b), sequence complementarity mismatches (c), base pair insertions and deletions (d), bulges induced by base insertions in one strand (e), and bulges induced by base deletions in the opposite strand (f). Scale bars are 100 μm. 9158

DOI: 10.1021/acsnano.6b04787 ACS Nano 2016, 10, 9156−9164

Article

ACS Nano

Figure 3. Impact of particular types of heterogeneity on DNA triangle crystallization. Table provides a classification based on the results of Figure 2, taking into account the concentration dependence of triangle crystallization as in Figure 1. Gel electrophoretic mobility analysis indicated whether a particular modified design self-assembled into triangles in the absence of original triangle strands. Asterisks indicate that the particular variant was studied for integration into the crystal lattice by fluorescence tracing experiments (Figures S4−S9). Red, light red, or gray asterisks indicate strong integration, weak integration, or no integration, respectively. The stylized diffraction patterns indicate that crystals grown in a mixture of original and modified triangle were also analyzed in X-ray diffraction (Figure S2). Top left schematic shows the original triangle for comparison. The 3′-ends of the strands are marked with arrow-tips.

the fluorescent label itself sometimes affected the crystallization behavior (Figure S3). To induce heterogeneity in a controlled fashion, we studied mixtures containing the original and also modified triangle DNA oligonucleotides. Since the preparation of such mixtures will reduce the absolute concentration of the original triangle, we first studied the dependence of crystal growth on the concentration of the original triangle in the absence of any additional species. After 10 days of incubation, crystals had robustly formed in all droplets having at least 20 μM effective triangle concentration (Figure 1b,c). At lower concentrations, the appearance of crystals became stochastic, meaning that in a number of droplets having identical conditions only a few droplets had crystals, and the crystals were also smaller. However, after approximately 150 days of incubation, crystals appeared also in droplets having only 6 μM effective triangle concentration (Figure 1b). Hence, the crystallization of the DNA triangle was robust with respect to triangle concentration, but the rate of crystallization can become impractically slow at low concentrations. For comparison, crystallization of proteins

with comparable molecular weight (80 μm grew only for a subset of the variants, predominantly at intermediate temperatures (25 °C) or when using a thermal ramp that cools slowly from 60 to 20 °C. We can attempt an estimate of the interaction energies of the different variants using the nearest-neighbor base pairing46 or 9161

DOI: 10.1021/acsnano.6b04787 ACS Nano 2016, 10, 9156−9164

Article

ACS Nano base pair stacking parameters.47 When we sort the variants according to the stacking energies, we observe clustering of the variants that give good crystals at intermediate temperatures and many small crystals at low temperatures (Figure 5c). The clustering is less obvious when using the nearest-neighbor base pairing parameters as estimators (Figure S11). Altogether, the data indicates a narrow “crystallization window” with respect to the range of interaction energies per triangle crystal contact. The window is just ∼1 kcal/mol wide and corresponds to 4−5 units of thermal energy per bond according to the stacking strength estimation. The actual absolute free energy difference for integrating a triangle into a growing lattice will depend on the buffer conditions and on the monomer concentration, and it will also include entropic penalties. In addition, the stacking parameters used for our estimation hold for 20 mM MgCl2, but our experiments were carried out in the presence of 200 mM MgCl2. Nonetheless, our results reflect a trend that indicates that some sequence combinations were too weak for crystal growth, while others were already too strong, despite the fact that we merely changed the neighbors of the single base pair that is formed per triangle arm.

of crystallization phenomena including nucleation,49 influence on crystal growth of impurities, concentration, and temperature,50 and effects from gravitation,51 magnetic,52 and electric fields.53 Phase diagrams54 and thermodynamic parameters55 have been investigated in detail using the lysozyme system. The DNA triangle crystallizes similarly robustly, but in contrast to the lysozyme it may also be altered and engineered in various and more rational ways in order to study the impact on crystal growth, as our study and those of others show.

EXPERIMENTAL METHODS DNA strands were prepared by solid-phase chemical synthesis without phosphates at the 5′-ends (Eurofins MWG, Ebersberg, Germany, HPLC grade). Strands were dissolved at a concentration of approximately 300 μM in nuclease free water (Qiagen). Actual strand concentrations were determined by UV/vis detection of DNA absorbance at 260 nm. Crystallization. One microliter of premixed sample (see Note S1 for details on sample preparation) was added to 4 μL of buffer in sitting drop plates (Intelli-plates, Art Robins Instruments, 24-well or 48-well format) to obtain droplet conditions of 25 mM Tris-HCl (pH 8.5) and 20 mM MgCl2. The droplets were annealed in sealed chambers supplied with a reservoir solution of 250 mM Tris-HCl and 200 mM MgCl2 (pH 8.5) using a thermal ramp decreasing from 60 to 20 °C with 0.4 °C/h, and then incubated at a constant temperature of 20 °C. Crystallization conditions for experiments shown in Figure 4f and Figure 5 are described in separate notes (Note S1). Imaging. Droplets were typically imaged under polarized light in a bright field light microscope approximately 10 days after experiment setup. A subset of droplets was monitored again after approximately 150 days of incubation at 20 °C. Fluorescence imaging was performed with 545/25 nm excitation filters and 605/70 nm emission filters at a Leica M165FC stereomicroscope coupled to a metal-halide light source. Twelve-bit grayscale images were acquired with an exposure time of 1.5 ms in monochromatic camera mode. Fluorescence in the overall droplets was analyzed by comparing the average intensity per pixel in the area covered by crystals with the average intensity per pixel in the background of the drop. Diffraction Data Collection. Crystals were transferred to a cryosolvent containing 30% glycerol, 250 mM Tris-HCl, and 200 mM MgCl2 and were flash-frozen in liquid nitrogen. X-ray diffraction data were collected at 1 Å on beamline SLS X06SA (Swiss Light Source, Villigen, Switzerland). Images were taken with Δφ = 0.5°, exposure time of 0.5 s, 50% beam transmission, and detector distance of 640 mm and at 100 K. Agarose Gel Electrophoresis. Triangle strands were mixed at a stoichiometry of L/M/S = 3:1:3 in 25 mM Tris-HCl and 20 mM MgCl2, pH 8.5, with an M strand concentration of 6 μM. The mixtures were annealed at a thermal ramp decreasing from 65 to 20 °C with 1 min/°C. Annealed samples were diluted 10-fold and mixed with 6-fold gel loading dye (15% Ficoll 400, 5 mM Tris-HCl, 0.15% bromophenol blue, pH 8.5). Electrophoresis was carried out in 5% agarose gels containing electrophoresis buffer (1 mM EDTA, 44.5 mM Tris base, 44.5 mM boric acid, and 11 mM MgCl2, pH 8.4) and SybrGold (Thermofisher) at a dilution of 1 to 15 000. The samples were electrophoresed for 3 h at 90 V in an ice water cooled gel box filled with electrophoresis buffer. The gels typically contained a lowmolecular weight DNA ladder as migration standard (O’RangeRuler 10 bp DNA Ladder, Thermofisher). The agarose gels were scanned using a Typhoon 9500 FLA laser scanner (GE Healthcare) at a resolution of 50 μm/px to give 16-bit tif image files.

CONCLUSIONS The present study represents a stepping stone toward crystallizing larger multichain DNA objects by indicating several aspects that can guide efforts to crystallize DNA origami, whether via screening methods into unplanned lattices or via designed interactions into lattices with planned geometry. We have seen here that the concentration of the triangle particles to be crystallized matters more for the appearance of crystals within practical laboratory time scales than the additional presence of impurities of various kinds. Extrapolating from our tests, for crystallizing large multichain DNA objects, it appears desirable to focus on dense solutions where one particular species is present at concentrations in the >10 μM range. This simple insight puts a comparably high bar on the preparation of DNA origami solutions, since self-assembly of such objects is commonly accomplished in the dilute