Cryo-Electron Microscopy and Mass Analysis of Oligolysine-Coated DNA Nanostructures.

Cationic coatings can enhance the stability of synthetic DNA objects in low ionic strength environments such as physiological fluids. Here, we used single-particle cryo-electron microscopy (cryo-EM), pseudoatomic model fitting, and single-molecule mass photometry to study oligolysine and polyethylene glycol (PEG)-oligolysine-coated multilayer DNA origami objects. The coatings preserve coarse structural features well on a resolution of multiple nanometers but can also induce deformations such as twisting and bending. Higher-density coatings also led to internal structural deformations in the DNA origami test objects, in which a designed honeycomb-type helical lattice was deformed into a more square-lattice-like pattern. Under physiological ionic strength, where the uncoated objects disassembled, the coated objects remained intact but they shrunk in the helical direction and expanded in the direction perpendicular to the helical axis. Helical details like major/minor grooves and crossover locations were not discernible in cryo-EM maps that we determined of DNA origami coated with oligolysine and PEG-oligolysine, whereas these features were visible in cryo-EM maps determined from the uncoated reference objects. Blunt-ended double-helical interfaces remained accessible underneath the coating and may be used for the formation of multimeric DNA origami assemblies that rely on stacking interactions between blunt-ended helices. The ionic strength requirements for forming multimers from coated DNA origami differed from those needed for uncoated objects. Using single-molecule mass photometry, we found that the mass of coated DNA origami objects prior to and after incubation in low ionic strength physiological conditions remained unchanged. This finding indicated that the coating effectively prevented strand dissociation but also that the coating itself remained stable in place. Our results validate oligolysine coatings as a powerful stabilization method for DNA origami but also reveal several potential points of failure that experimenters should watch to avoid working with false premises.

The methods of DNA origami[1−5] enable building nanostructures with sub-nanometer precise features[6] and overall dimensions ranging from the nanometer to the micrometer scale,[3,7−17] covering molecular weights up to the gigadalton scale.[18,19] DNA origami objects may be site-specifically functionalized and modified with chemical groups and biomolecules[20−22] and can be designed to include mechanisms.[23−25] Custom DNA objects have been developed and successfully used in diverse applications in basic research, thereby delivering scientific insights and demonstrating the capacity of DNA nanotechnology to yield objects with utility. Examples range from structural biology,[26−28] biophysics,[29−34] photonics,[35−38] and plasmonics[39−43] to molecular electronics.[20,44−46] First steps have also been taken to design DNA objects as programmable agents in medical therapy.[25,47] In many of these applications, DNA nanostructures assume the role of a scaffold for controlling the type, number, relative position, and orientation of chemical groups or other non-DNA molecules such as proteins.[7,48]

To take advantage of the positioning opportunities in applications, it is necessary that the DNA nanostructures remain intact in the target environments, and that the user appreciates the shape that the DNA object will adopt in the target environment. Many DNA origami objects will by default not be stable at low ionic strengths, such as those present in physiological fluids, due to the presence of destabilizing repulsive electrostatic forces.[49−51] Insufficient screening of the electrostatic forces can cause dissociation of strands from the DNA nanostructures. Furthermore, DNA objects are prone to be digested by nucleases that are present in biological fluids such as blood or serum. To address the stability issues, researchers have developed various treatments for stabilizing DNA nanostructures post-assembly. The available methods include various types of chemical[52,53] or enzymatic strand ligation[54] and coatings with cofactors.[51,55] DNA nanostructures may also be further stabilized by the addition of chemicals such as 8-methoxypsoralen.[56]

In a method called UV point welding,[49] additional covalent bonds are introduced into DNA nanostructures in a sequence-programmable fashion by forming cyclobutane pyrimidine dimer bonds between colocalized thymidines upon exposure to 310 nm UV light. The UV-induced covalent bonds can be used to circularize DNA strands and to cross-link neighboring strands at DNA strand junctions by placing neighboring T’s at such sites during sequence design. By UV point welding, DNA nanostructures become topologically trapped and their disassembly would require breaking of covalent bonds. UV point-welded DNA nanostructures can be exposed to various environments, including distilled water and increased temperatures.

Another previously introduced stabilization method involves coating DNA nanostructures with cationic oligolysine molecules.[51] Ponnuswamy et al. demonstrated that DNA nanostructures coated by oligolysines are stable in low salt and up to 10-fold more resistant to digestion by DNA nuclease I (DNase I) than when uncoated. A tendency to aggregate when using oligolysine could be circumvented by coating with polyethylene glycol (PEG)-oligolysine copolymers. PEG-oligolysine coating is a simple and scalable approach that provides effective protection of DNA nanostructures for in vivo applications.[51] Anastassacos et al. recently reported that glutaraldehyde cross-linking of PEG-oligolysine-coated DNA extends survival by up to another ∼250-fold to >48 h during incubation with highly concentrated DNase I.[57]

The stabilization efficacy of UV point welding was previously investigated with respect to strand composition in gel electrophoretic shift assays. For example, the loss of DNA strands from UV point-welded objects after they were exposed to otherwise destabilizing conditions was quantified and used as a parameter to refine the treatment.[49] Furthermore, cryo-EM structures of UV point-welded objects were determined prior to and after treatment and after transfer to physiological buffer. Due to the strong electrostatic repulsion in physiological buffer, a swelling behavior was seen which altered the aspect ratio of the DNA origami test objects: the objects became shorter along the helical axis, whereas the lateral dimensions increased due to helical lattice expansion.[49]

Performing a compositional analysis using gel electrophoresis for oligolysine-coated objects is hampered by the fact that the objects are neutralized by the coating and thus no longer properly migrate in a gel matrix. Also, structural data beyond negative-staining TEM image data are so far missing for oligolysine-coated objects. For users of the oligolysine coating technique, it would be important to appreciate the thickness of the coating, whether the coating is smoothly distributed over the objects or whether it preferentially accumulates at certain hotspots, and whether the coating itself leads to structural deformations from the untreated shape. Furthermore, it would be interesting to learn whether the transfer of oligolysine-coated DNA origami objects to low ionic strength physiological buffers also leads to swelling and/or twisting as it has been seen for the UV-welded objects. Finally, functional features such as small chemical groups could potentially be occluded by the coating, which could impair the functionality of coated objects. Hence, it would be important to test whether reactive patches can still be accessed underneath the coating.

To address these questions, we constructed two test DNA origami objects which we coated with oligolysine and PEG-oligolysine, respectively. We determined cryo-EM maps of these objects in free-standing ice and constructed pseudoatomic models for a subset of the objects. We used single-molecule mass photometry to elucidate optimal coating conditions and to reveal changes in molecular mass upon exposure to physiological conditions. We studied how the coating alters the functionality of the test structures in the form of the ability to dimerize via base pair stacking interactions as a model for small reactive chemical patches.

Results and Discussion

Design and Preparation of Test Objects

We designed two test multilayer DNA origami objects, the A- and the B-bricks (Figure A). Both objects are designed in a honeycomb lattice helical packing. The A-brick is folded from a 7560 bases long scaffold single strand, whereas the smaller B-brick forms from a 2873[58] bases long custom scaffold strand (Figures S1 and S2). We designed the objects with a set of specific structural details to facilitate analysis by cryo-EM. Specifically, the A-brick features a number of protrusions and recesses which vary in size to test the effect of coating on differently sized design features. An asymmetric feature was placed on one of the helical interfaces of both bricks to facilitate particle alignment in 3D cryo-EM analysis. The B-brick features a protrusion on one of its sides that fits into one particular shape-complementary recess on the A-brick (Figure B). This feature enables dimerization of the A- and B-bricks via base pair stacking interactions that can engage upon precise fit of the protrusion into the recess.[12] We exploited this functionality to test the accessibility of reactive patches underneath the coating.

Figure 1

(A) Scheme of the A-brick (left) and B-brick (right). Cylinders represent DNA double helices. (B) Schematic representation of the dimerization process between A- and B-brick. (C) Agarose gel showing different PEG-oligolysine coating ratios (lanes 1−8, 0.25:1, 0.5:1, 0.75:1, 1:1, 2:1, 5:1, 10:1, 100:1 N/P ratios, respectively) for the monomeric A-brick (left half) and for the B-brick (right half). (D,E) Exemplary cryo-EM micrographs of the uncoated A- and B-bricks, respectively. Scale bar: 50 nm.

We used previously described folding procedures[59] to determine proper folding conditions for both A- and B-bricks (Figures S3 and S4) and purified the folded objects from excess oligonucleotides using ultrafiltration (Figure S5). We evaluated suitable coating ratios with oligolysine and PEG-oligolysine using gel electrophoresis. As described in previous work by Ponnuswamy et al.,[51] at ratios of 0.5:1 (N/P) (nitrogen in lysine to phosphorus in DNA) for oligolysine and at 0.75:1 (N/P) for PEG-oligolysine, the DNA objects were effectively neutralized based on the observation that the electrophoretic mobility was negligible at this ratio. Higher ratios of N/P led to inversion of the migration direction (Figure C and Figures S6–S8), suggesting that the objects were now cationic. After coating, we purified the objects from potential excess oligolysine and increased their concentration to approximately 100−400 nM using ultrafiltration to achieve conditions suitable for vitrification. We then acquired cryo-EM micrographs of uncoated (Figure D,E) and coated (Figures S9–S19, panels A) samples in free-standing vitrified ice. The micrographs of coated samples showed a substantially increased particle density compared to that of uncoated objects in the free-standing ice, even at order-of-magnitude lower concentrations for coated compared to uncoated samples (Figures S9–S19, panels A), which is an advantageous property. On the other hand, the coated samples also exhibited a more pronounced orientation bias in the free-standing ice. In spite of these sample-to-sample differences, we were able to determine reasonable 3D electron density maps for all samples.

Cryo-EM Structural Analysis

The resulting cryo-EM maps all coarsely matched the designed shape of the bricks (Figure ). The asymmetric features and the designed protrusions and recesses could be clearly identified in all of the maps. The cryo-EM maps that we obtained from the uncoated DNA origami objects (Figure A) had the best overall resolution (around 10–12 Å) and allowed discerning the major and minor grooves of individual helices and also the location of the majority of crossovers. The cryo-EM maps of the oligolysine-coated and PEG-oligolysine-coated nanostructures that we determined from data acquired at the same (high) ionic strength as that for the uncoated samples had lower resolution (Figure B) compared to the maps obtained from the uncoated samples, in spite of the fact that more or comparable particle numbers were used in each data set (see also Figures S9–S19 for angular distribution of views). Helical details such as grooves are no longer visible in the cryo-EM maps determined from the coated samples. Presumably, these features are now occluded by the coating, which is not expected to have specific order. The relative position and orientation of features such as protrusions and recesses were preserved by the coating, but the global shape was altered in several aspects as we will discuss below.

Figure 2

Cryo-EM maps (not to scale) of the test structures with different coating ratios and buffer conditions. (A) Different angular views of uncoated A-brick (top) and B-brick (bottom) in 5.5 mM MgCl2. (B) A-brick (top) and B-brick (bottom) in 5.5 mM MgCl2 coated with K10-oligolysine (0.5:1 N/P, light blue) and with PEG-oligolysine (0.75:1 N/P, orange). (C) A-brick (top) and B-brick (bottom) coated with PEG-oligolysine (0.75:1 N/P) in phosphate-buffered saline. (D) B-brick coated with PEG-oligolysine (0.2:1 N/P) in 5.5 mM MgCl2. (E) B-brick coated with PEG-oligolysine (100:1 N/P) in 5.5 mM MgCl2. (F) B-brick coated with PEG-oligolysine (100:1 N/P) in phosphate-buffered saline. (G) Uncoated B-brick (left) and B-brick coated with PEG-oligolysine (100:1 N/P) in 5.5 mM MgCl2 (right) showing the effect of coating with a high PEG-oligolysine ratio on the helix lattice packing.

To reveal the structural changes incurred upon transfer of the objects to physiological ionic strength conditions, we transferred the coated A- and B-brick samples to standard phosphate-buffered saline as a model for physiological ionic strength, after having coated these objects with oligolysine and with PEG-oligolysine, respectively. As a reminder to the reader, an uncoated control sample cannot be prepared at these conditions because, without the stabilizing coating, the multilayer DNA origami objects used herein will disintegrate. The cryo-EM maps that we determined from the samples dissolved in PBS (Figure C) allowed discerning individual helices but not much further detail. The global shape of the objects dissolved in PBS matched roughly but not in detail those of the coated objects in high ionic strength. A swelling perpendicular to the helical axis can be seen.

To study the structural consequences of coatings with different densities, we also prepared B-brick samples coated with a “thin” 0.2:1 N/P and a “thick” 100:1 N/P ratio of PEG-oligolysine molecules to DNA bases and determined cryo-EM maps of the objects dissolved at high ionic strength (5 mM MgCl2) and also dissolved in PBS (Figure D–F). The maps determined from the sample with the “thin” 0.2:1 coating are hardly distinguishable from those determined for the uncoated objects. By contrast, the 100:1 coating strongly affected the internal honeycomb lattice packing: it squeezed the lattice toward a more square-lattice-like appearance (Figure G).

To allow for a more detailed discussion of the structural changes, we constructed pseudoatomic models for the uncoated (Figure A and Supplementary Movie 1) and coated (Figure B–G) cryo-EM maps of the B-brick using ENRG-MD[60] initialized cascaded flexible fitting (“shrink wrap fitting”).[6,61] We note that the maps of the objects, in principle, do not have sufficient resolution to be interpreted with atomic models. Nonetheless, the models are useful for tracing features such as the helices and crossovers which can also be discerned in the maps by visual inspection. The atomic models relate these features to the design diagrams and enable a quantitative geometrical analysis. Comparison of uncoated and coated models shows a steady increase in root mean square displacement with a thicker coating of up to 16 Å (Supplementary Table 1).

Figure 3

Pseudoatomic models of the B-brick. (A) Uncoated B-brick in 5.5 mM MgCl2 (overlay with cryo-EM map, scaffold gray, staples colored). Left: View shows the opposite side of the structure than the cryo-EM maps from Figure . (B) Coated with PEG-oligolysine (0.2:1 N/P, gray) in 5.5 mM MgCl2. (C) Coated with PEG-oligolysine (0.75:1 N/P, orange) in 5.5 mM MgCl2. (D) Coated with PEG-oligolysine (100:1 N/P, light orange) in 5.5 mM MgCl2. (E) Coated with K10-oligolysine (0.5:1 N/P, light blue) in 5.5 mM MgCl2. (F) Coated with PEG-oligolysine (0.75:1 N/P, green) in phosphate-buffered saline. (G) Coated with PEG-oligolysine (100:1 N/P, light green) in phosphate-buffered saline. Model coloring matches the coloring of the respective cryo-EM maps from Figure .

To quantify the global twist present in the DNA objects along the helical axis in a model-free fashion, we aligned the cryo-EM map image stacks such that the helical axis was perpendicular to the image stacking direction. This way, each image in the stack represents a helical cross section slice. Each slice has a thickness of 4.4 Å, which corresponds to approximately 1.5 base pairs. We rotationally aligned the slices using cross-correlation to a reference cross section slice chosen from the center of each object. This procedure yields in silico twist-corrected variants of the cryo-EM maps (Figure A), and outputs the angles necessary to rotate the stack slices to align them with the reference slice (Figure B). As a result, we see that the uncoated A-brick features in total an approximately 35° twist deformation around the helical axis between slices 30 and 140 (Figure S20). The coatings affected the global twist in the A-brick in an inhomogeneous manner. The lower region of the A-brick was essentially twist-free after oligolysine coating with 0.5:1 N/P, whereas the upper part maintained approximately the twist density it had prior to coating. The global twist in the B-brick was less affected by the coating with a 0.5:1 oligolysine to DNA bases ratio. However, the high-density coating with 100:1 ratio practically straightened the B-brick.

Figure 4

Structural changes caused by coating. (A) Original (left) and in silico twist corrected (right) B-brick cryo-EM maps. (B) Plots of twist angle as a function of slices through the maps for B-brick. Colors indicate different coating and buffer conditions. (C) Helical projections of the original B-brick (top), after in silico twist correction (middle) and central slice (base 85–100) of the atomic model (bottom) with different coating and buffer conditions. Atomic model slices are colored by lattice column of the uncoated variant. (D) Lattice pattern for regular honeycomb lattice (left) and compressed honeycomb for structures coated with PEG-oligolysine 100:1 N/P ratio (right) for the B-brick. (E) Plots of structure dimension as a function of coating ratio. Directions x and y are calculated as an average of suitable helical pairs, while z is calculated as an average over all helical lengths. Helical ends and protrusions were excluded from dimension measurements. Lines are guides for the eye.

With the in silico twist-corrected cryo-EM maps, we also had a closer look at changes in the shapes of the cross sections of the objects. To this end, we computed the sum of the signals in the native and in silico twist-corrected cryo-EM maps back-projected along the helical directions and contrasted it with a central slice of the atomic models (Figure C). For the high-density coating with 100:1 oligolysine to DNA bases ratio, the cross-sectional projections reveal that the helices now appear packed on a more square-like lattice instead of on a honeycomb lattice. For the B-brick, we see that the coatings induce a bending deformation around the helical axis. The atomic models indicate that this effect is design-specific, as it is already present for the uncoated B-brick and likely caused by the asymmetric horizontal lattice pattern. This deformation is enhanced by the coating and may be considered as an intermediate step along the transformation into a square-lattice cross section (Figure D). We traced the midpoints of all DNA double helices to quantify the dimensions of the B-brick (Figure E). In the structures obtained from samples dissolved in PBS, we observe that the objects swell by approximately 8% in the y direction and shrink by approximately 10% in the helical (z) direction. Different degrees of expansion in x and y directions might be attributed to the lattice deformation.

From this analysis with two different objects, we see that oligolysine coatings do preserve coarse structural features well on a resolution of multiple nanometers, but the coatings also induce deformations such as twisting and bending, which should be taken into account in applications that require more accurate positioning.

Accessibility of Functional Features Underneath the Oligolysine Coating

The protrusions on the B-brick are designed to fit into the recesses on the A-brick (Figure A–C). The edges are lined with DNA blunt ends, which enables base pair stacking between the protrusions and recesses.[12] An increase in cation concentration in solution reduces electrostatic repulsive forces between the A- and B-bricks’ lateral surfaces and can stabilize short-ranged blunt-end stacking forces that can engage when the protrusions on the B-brick and the recesses of the A-brick come in close contact. This “docking” manifests as a shift from a monomeric to a dimeric state, as seen previously for many other objects with shape-complementary blunt-end stacking interfaces.[12,18,33,62] In order for this mechanism to work, the blunt-ended reaction patches must be accessible, but the oligolysine or PEG-oligolysine coating may potentially occlude these “sticky” patches. To investigate this issue, we used transmission electron microscopy (TEM) to compare the dimerization behavior of uncoated A-and B-brick samples with those of A- and B-brick samples that were coated with oligolysine or PEG-oligolysine, respectively (Figure D–F and Figures S21–S37). At 5 mM MgCl2, all samples remained in a monomeric state (Figure G). At 10 mM MgCl2, only the oligolysine-coated monomers started to dimerize. Further increasing the MgCl2 concentration to 15 mM led to aggregation of the oligolysine-coated objects. According to molecular simulations,[63] aggregation of oligolysine-coated structures can be explained by the mechanism with which oligolysine moieties bind to DNA origami helices: one oligolysine molecule usually connected two neighboring helices. This can also occur between helices of separate objects and hence promote aggregation.

Figure 5

Dimerization of the test structures. (A) Schematic representation of the dimerization mechanism of the bare A- and B-bricks. (B) Schematic representation of the dimerization mechanism of the A- and B-bricks coated with oligolysine. (C) Schematic representation of the dimerization mechanism of the A- and B-bricks coated with PEG-oligolysine. Zoom-ins of DNA in (A), (B), and (C) were created using BioRender. (D) Side (left) and top (right) views of negative-stain TEM class averages of the bare dimers. (E) Side (left) and top (right) views of negative-stain TEM class averages of the dimers coated with oligolysine (0.5:1 N/P ratio). (F) Side (left) and top (right) views of negative-stain TEM class averages of the dimers coated with PEG-oligolysine (0.75:1 N/P ratio). Scale bar 50 nm. (G) Types of structures (monomers, dimers, or aggregates) visible in negative-stain TEM images at different MgCl2 concentration conditions for the bare (gray), oligolysine-coated (cyan), and PEG-oligolysine-coated (yellow) structures.

The uncoated monomers dimerized in the presence of 40 mM MgCl2 and can be easily discerned as dimers in TEM data (Figure D). Notably, the uncoated monomers showed no aggregation whatsoever at any of the tested MgCl2 concentrations. The PEG-oligolysine-coated structures required further elevated MgCl2 concentrations (>140 mM MgCl2) for dimerization (Figure F), and aggregates appeared at 500 mM MgCl2 concentrations. We conclude that coating by itself does not occlude tiny reaction patches such as DNA blunt ends. When using oligolysine as a coating, the dimerization reaction requires much less additional screening agents in solution, which is expected because this role is now assumed by the oligolysine. When using PEG-oligolysine, however, dimerization presumably suffers from an additional entropic penalty caused by the randomly coiled PEG brushes, which have a tendency to exclude each other.

Mass Analysis of (PEG-) Oligolysine-Coated Nanostructures

Next, we addressed the question whether the composition of DNA nanostructures coated with oligolysine or PEG-oligolysine changes once placed into low ionic strength physiological environment, which would mean that DNA strands or parts of the coating dissociated from the DNA objects. Electrophoretic mobility analysis is not suited for addressing this question because the coating neutralizes the objects, and the electrophoretic mobility becomes negligible. To measure the mass, we resorted to single-molecule mass photometry.[64] Here, the particles under investigation diffuse freely in solution and can stochastically hit the surface of a coverslip placed into an interference scattering microscope. These “landing events” are detected in the form of transient diffraction-limited spots exhibiting a certain contrast over background (Figure A). The contrast of these spots can be related to the mass of the landing particles, given a calibration of the contrast scale relative to a reference sample with known mass, and provided that the scattering density and the refractive index of the specimen under study is comparable to that of the reference sample.[64] This approach has been exploited successfully in the past for measuring the mass of proteins in solution and for tracking multimerization dynamics.[65] Here, it is not clear a priori whether the coating with oligolysine changes the effective diffraction index or the density of the particles, but this must be clarified to address whether the interferential contrast measurements can be calibrated into an absolute mass scale.

Figure 6

B-brick mass photometry measurements and analysis. (A) Exemplary field of view of a mass spectroscopy measurement, where every black dot represents a landing event onto the microscope slide (left) and histograms showing the contrast of the samples depending on the type of coating: bare (gray), coated with oligolysine (cyan), and coated with PEG-oligolysine (yellow). (B) Contrast of the B-brick depending on the coating N/P ratio for 1 h incubation (left and continuous line) and for overnight incubation (dashed line). (C) Contrast of the B-brick in 5 mM MgCl2 buffer after coating with oligolysine (top left) or PEG-oligolysine (top right), after storing in PBS for 1 h (middle), and after filtration and buffer exchange in 5 mM MgCl2 (bottom).

The average scattering contrast per particle that we recorded for oligolysine-coated DNA origami B-brick samples (ratio 0.5:1 N/P) shifted toward higher contrast with respect to the bare object, suggesting an increase of the nanostructure’s mass caused by the coating (Figure A). The contrast of the PEG-oligolysine-coated sample (ratio 0.75:1 N/P) shows a larger shift to higher contrast, presumably indicating an even higher mass of the coated particles caused by the presence of the heavy PEG chains relative to the uncoated particles. We know from gel electrophoresis that at coating ratios around ∼0.5:1 the particles are effectively neutralized. Hence, assuming that each negative charge of the DNA backbone phosphates is screened by a positive charge of a nitrogen in an oligolysine chain, the PEG-oligolysine-coated DNA origami nanostructures should have a mass 3 times higher than that of the uncoated ones. Yet this expected mass increase is not reflected in a correspondingly massive scattering contrast increase: the scattering contrast between uncoated to coated particles differed only by ∼25%.

To investigate the origin of this discrepancy, we titrated the coating thickness 0.1:1 to 100:1 N/P and performed mass photometry for each condition. The thus measured average scattering contrast per particle as a function of the coating ratio follows a “check-mark”-shaped curve (Figure B and Figure S38). First, the measured scattering contrast decreases for very low N/P ratios (0.1:1–0.2:1). For N/P ratios bigger than 0.2:1, the contrast then increases again and saturates at a ratio of ∼5:1. The nonlinear behavior suggests that scattering density and index of refraction per particle change with different coating ratios. Given the nonlinear behavior of the observed contrast, the contrast axis thus cannot be easily calibrated into a mass axis by, for example, using the known mass and observed contrast of the uncoated objects as a point of reference.

However, in spite of this complex relationship, it is possible to use the photometric data to reveal whether the composition of coated DNA objects changes upon exposure to physiological ionic strength. To this end, we prepared a coated sample of the DNA origami test bricks and measured their scattering contrast profile at high salt conditions (5 mM MgCl2) (Figure C, top). Then, the samples where filtered and incubated for 1 h at 37 °C in PBS. Again, we recorded the scattering profile. Interestingly, the average contrast per particle shifted to lower values in PBS (Figure C, middle). However, upon restoring the initial high-salt conditions containing magnesium, the scattering contrast returned to the initial conditions (Figure C, bottom). The contrast histograms prior to and after exposure to PBS completely overlap for both DNA origami test objects, which indicates that the composition did not change. Hence, the coating prevented the dissociation of strands from the DNA origami. We speculate that the drop of the scattering contrast that we observed in PBS originates from a reduced mass density of the particles, consistent with the swelling that we saw in the cryo-EM structures recorded in the presence of PBS.

Conclusions

We analyzed two DNA origami test objects to gain insights into the structural and functional changes caused by coatings with oligolysine and PEG-oligolysine. The cryo-EM maps that we determined for the uncoated objects offered substantially more detail than those determined for the coated objects, even though the acquisition settings and the particle numbers used were comparable. Details such as the minor and major groove of DNA helices were no longer recognizable after coating. We attribute this blurring of details to the oligolysine chains surrounding the DNA origami helices. The cryo-EM maps that we obtained from objects dissolved in PBS with 150 mM NaCl were swollen compared to those obtained in the presence of MgCl2. Previous cryo-EM maps determined for UV-welded objects dissolved in PBS[49] also showed swelling. One possible explanation for the swelling may be the strong interhelical repulsion at low ionic strength, which will tend to increase the interhelical distances. This appeared to be a compelling explanation for the swelling previously seen in UV-welded objects. However, for the oligolysine-coated objects, internal repulsive electrostatic forces are presumably screened by the presence of the cationic oligolysine coating. Therefore, swelling by increased electrostatic repulsion in a low ionic strength environment such as PBS is not necessarily expected. We speculate that the swelling we observed in PBS with the oligolysine-coated samples may be caused by a superposition of multiple effects: residual low ionic strength related repulsion plus some conformational rearrangement of holiday junctions due to replacement of divalent MgCl2 with monovalent cations.

The global twist of one of our test structures was reduced after coating (similarly to the findings by simulations in ref (63)). The global twist remained comparable before and after coating for the second object, unless the excess of PEG-oligolysine molecules was high (100:1 N/P ratio). It is not clear to us why multilayer DNA origami designed in a honeycomb lattice exhibit a residual twist in the first place[6,18] and also why the two test objects reacted differently to the coating. Our findings indicate that the shape of structures stabilized with (PEG)-oligolysine coatings should be validated experimentally on a case-by-case basis.

In conclusion, we presented a detailed characterization of the structures and functionalities of DNA origami objects stabilized with (PEG-) oligolysine coating in a physiological ionic strength environment. Our results support oligolysine coatings as a powerful stabilization method for DNA origami but also reveal several potential points of failure that experimenters should watch to avoid working with false premises.

Materials and Methods

Design of Scaffolded DNA Origami Objects

The objects were designed using caDNAno v0.2.[4]

Folding of DNA Origami Nanostructures

The folding reaction mixtures for both the A- and B-bricks contained scaffold DNA[66] at a final concentration of 50 nM and oligonucleotide strands (IDT Integrated DNA Technologies) at 200 nM each. The folding reaction buffers contained 5 mM TRIS, 1 mM EDTA, 5 mM NaCl (pH 8), and 25 mM MgCl2 for the A-brick and 20 mM MgCl2 for the B-brick. The folding reaction mixtures were subjected to the same thermal annealing ramp using TETRAD (MJ Research, now Biorad) thermal cycling devices: 15 min at 65 °C, followed by 1 h intervals for each temperature, starting at 58 °C down to 47 °C, decreasing by 1 °C every step. Finally, the folding reaction mixtures were incubated at 20 °C before further sample preparation steps.

Purification of DNA Origami Nanostructures

Excess oligonucleotides were removed via ultrafiltration (Amicon Ultra 0.5 mL Ultracel filters, 50K) with FoB5 (5 mM TRIS, 1 mM EDTA, 5 mM NaCl, and 5 mM MgCl2) buffer.[56] All centrifugation steps were performed at 10000g for 3 min at 25 °C. The filters were first filled up with 0.5 mL of FoB5 buffer and centrifuged. Next, 0.1–0.2 mL of a folded object sample and 0.3–0.4 mL of FoB5 buffer were then added and centrifuged. Another three rounds of adding 0.45 mL of FoB5 buffer and subsequent centrifugation were performed before a final retrieving step, where the filter inset was turned upside down, placed into another tube, and centrifuged.

Coating of DNA Origami Nanostructures

Different volumes of purified DNA objects were mixed with oligolysine (K10) and PEG-oligolysine (K10-PEG5k), ranging from 10 to 500 μL in FoB5 buffer. The appropriate amounts of K10 and K10-PEG5k were determined based on the ratio of nitrogen in amines to the phosphates in DNA.[50] For the A-brick, the ratio was 0.5:1 for K10 and 0.75:1 for K10-PEG5k. For the B-brick, the ratio was 0.5:1 for K10 and 0.75:1 for K10-PEG5k.

Negative-Stain TEM

Five microliters of sample (10–20 nM DNA object concentrations, 20 mM MgCl2 concentration) was pipetted onto a carbon Formvar grid (Electron Microscopy Sciences). For bare DNA objects, the grids were plasma-treated (45 s, 35 mA) right before use. Coated DNA objects were pipetted onto the grids without plasma treatment. This was done to reduce positive staining effects. The sample droplets were incubated for 30–60 s on the grids and then blotted away with filter paper. A 5 μL droplet of 2% aqueous uranyl formate (UFO) solution containing 25 mM sodium hydroxide was added and immediately blotted away as a washing step. A 20 μL UFO droplet was then added, incubated for 30 s, and blotted away. The grids were then air-dried for 20 min before being imaged in an FEI Tecnai 120 and a JEOL JEM3200 FSC microscope. Automated particle picking was performed with crYOLO,[67] and 2D class averaging was performed with Relion.[68]

Cryo-EM: Sample Preparation, Data Acquisition, and Processing

For the bare structures, concentrations from 1 to 2.8 μM were used, whereas for the coated objects, a sufficient particle density was obtained with particle concentrations around 100 to 400 nM. Samples of 3–4 μL were applied to C-flat 1.2/1.3 or 2/1 grids (Protochips). A Vitrobot Mark V (Thermo Fischer Scientific) was used for plunge freezing at constant temperature (20 °C) and humidity (100%) with 0 s wait time, 2–4 s blot time, and −1 to 0 blot force.

Imaging was performed on a Titan Krios cryo-electron microscope operating at 300 kV, equipped with a Falcon III direct electron detector (Thermo Fisher Scientific) in fractioning mode. Movies composed by 7 frames were acquired at a magnification of 29000×, which corresponds to a pixel size of 2.319 Å, with 2.6 s exposure time and a total dose of 56 e–/Å2. A defocus value of −2 μm was used. All movies were motion corrected using MotionCor2[69] and CTF corrected with CTFFIND4,[70] both implemented in Relion.[68] Automated particle picking was performed with crYOLO.[67] All subsequent image-processing steps were carried out with Relion3.0. Several runs of 2D classifications were necessary to exclude bad particles, false positives, or particles lying on the carbon. After 3D classification, the classes which presented most features were further refined and postprocessed.

To further improve the resolution, the maps were cut in halves perpendicular to the helical axis with the erase tool in Chimera[71] and a MultiBody refinement job was performed in Relion. The bare and PEG-oligolysine-coated (in 5 mM MgCl2) A-bricks were further subjected to a local resolution estimation job to remove postprocessing artifacts.

Mass Spectrometry: Sample Preparation, Data Acquisition, and Processing

Mass spectrometry experiments were performed with a Refeyn OneMP (Refeyn Ltd.). Samples were evaluated with microscope coverslips (Art. No. H878.2, coverslips thickness = 1, 24 × 60 mm, Carl Roth). The coverslips were first cleaned in the following manner: (1) sonicate for 5 min in ddH2O; (2) sonicate for 5 min in ETOH; (3) sonicate for 5 min in ddH2O; (4) sonicate for 5 min in isopropyl alcohol; (5) sonicate for 5 min in ddH2O. A silicone template was placed on top of the coverslip to form reaction chambers. A sample of 8 μL was pipetted into the reaction chambers at a DNA object concentration of 5 nM in various 1× PBS or FoB5 buffer. The frame rate of the camera was set to 500 Hz. Images were pixel binned 2 × 2 and frame binned 5-fold. The exposure time was set to 1.5 ms. Image analysis was performed with the software provided by Refeyn Ltd., with the default settings provided by the manufacturer.

Pseudoatomic Model Construction: ENRG-MD Driven Cascaded Flexible Fitting

Pseudoatomic models were generated following a procedure based on a recent study on DNA origami called ENRG-MD driven cascaded flexible fitting.[6] Initial ENRG-MD[60] models and cryo-EM maps were globally aligned using VMD.[72] These models were relaxed using 4800 steps of steepest descent energy minimization and 7200 ENRG-MD steps to ensure a suitable geometry. For the cascade, six cryo-EM maps were generated from the original data, each low pass filtered using a Gaussian blur in 1 Å increments. As the purpose of the first cascade is to prealign the helices, it was sufficient to abort it after the first two cascade steps. This first cascade is followed by 12000 ENRG-MD steps to resolve strong deformations caused by the initial realignment. Interhelical bonds of the elastic network of ENRG-MD were removed after the second step of the second cascade. Both MDff[61,73] cascades were performed with a weight on the EM density of 0.3 kcal/mol. Each cascade step consists of 12000 MDff steps.

In contrast to the previous study,[6] the interhelical elastic network was not further reduced to base pair restraining bonds for all coated maps, as the presence of the coating and an overall resolution of 12 Å or worse does not allow fitting with the necessary level of accuracy. Fitting for uncoated maps was performed with extra 6000 MDff steps with only base pair restraining bonds. All fitting protocols were concluded with 12000 steps of energy minimization and a weight on the EM density of 1.0 kcal/mol.

NAMD 2.12_Linux[74] and the CHARMM36[75,76] nucleic acid force fields were used for all simulations. Cryo-EM map generation and fit evaluation were performed with the VMD packages mdff_0.5 and volutil_1.3. Selection of helical layers shown in Figure C was performed using the FitViewer Jupyter Notebook.[77] Dimensions of the B-brick were calculated using the helical trace defined by the C1′ center of each base pair. Helical pairs were selected in the context of the caDNAno strand diagram of the object’s core (helices 6–29, base position 44–126). The fitted model of the uncoated B-brick is available in the Protein Data Bank (PDB).[78] Models of coated maps were not uploaded to the PDB.