• By firing high energy electrons through our DNA boxes, we are able to create images that show what out structures look like

  • Boxes of the correct size and shape can easily be seen, so we know the boxes have formed (i.e., structural characterisation)!

  • A statistical analysis was performed for functional characterisation, but the inherent limitations of TEM have created discrepancies with other analysis techniques

If you’re TEMpted to read more, check the lab book here

TEM 1: Observation of Boxes


Transmission Electron Microscopy (TEM) is a key imaging technique used on DNA origami. TEM involves the transmission of high energy electrons through a thin sample. Just like how different materials allow different amounts of light to pass through them, a properly treated sample would let different amounts of electrons pass through them. If these electrons are detected on the other side of the sample, an image can be viewed.

We expect to see box-shaped structures, with dimensions of about 39.1 nm by 19.9 nm.


Why TEM?

TEM has a number of key uses:

  • High resolution imaging: The high energy of transmitting electrons allows them to have a small wavelength, and thus, create images with a much higher resolution than most other imaging techniques, while also attempting to minimise the sample damage. This allows for the formation of boxes to be visually confirmed and compared with one another.

  • Quantitative analysis of box shape and size: The images produced can be quantitatively analysed to gain information about the box’s physical structure. Their sizes and aspect ratios can be determined, for example, as well as their orientation, whether they’re open or closed, and whether or not they contain something inside their cavity.


To confirm the formation of box-shaped structures made using DNA origami.


A preliminary electrophoresis gel was made using unbraceed boxes for early confirmation of box formation, and to optimise for Mg2+ in synthesis (Figure 1, Table 1).


Figure 1: Unbraced boxes, with varying hinge torque and Mg2+ concentration.

Lane Identity Mg2+ Concentration (mM)
1 1kb Ladder -
2 m13 -
3 Staples -
4 Sealed 0
5 Untorqued 0
6 35° Torque 0
7 70° Torque 0
8 105° Torque 0
9 Sealed 12
10 Untorqued 12
11 35° Torque 12
12 70° Torque 12
13 105° Torque 12
14 Sealed 16
15 Untorqued 16
16 35° Torque 16
17 70° Torque 16
18 105° Torque 16
19 Sealed 20
20 Untorqued 20
21 35° Torque 20
22 70° Torque 20
23 105° Torque 20
24 1kb Ladder -

Table 1: Legend for Figure 1.

It appears that boxes successfully formed, due to the slower travel rate of the boxes compared to both the m13 scaffold and staple mixes. The bands also appear to be most obvious at 20mM Mg2+ concentration, and so, all future synthesis will be performed at this concentration. Aside from this, all boxes look identical, except for the sealed boxes in comparison to the non sealed boxes. The sealed boxes appear to move further than non-sealed boxes, suggesting they are smaller. This makes sense – if the sealing staples worked, the box will be stapled shut, and so would move more easily than the boxes whose halves move around.

For further confirmation, another gel was run with every possible box. PEG purification was also performed on many samples remove excess staples (Figures 2 to 3, Tables 2 to 3)


Figure 2: Gel run with every box combination, Part 1.

Lane Hinge Torque Brace Position Brace Angle Other
1 - - - 1kb Ladder
2 0 None - Sealed
3 0 None - -
4 35 None - -
5 70 None - -
6 105 None - -
7 0 1 45 -
8 35 1 45 -
9 70 1 45 -
10 105 1 45 -
11 0 2 45 -
12 35 2 45 -
13 70 2 45 -
14 105 2 45 -
15 0 1 90 -
16 35 1 90 -
17 70 1 90 -
18 105 1 90 -
19 0 2 90 -
20 35 2 90 -
21 70 2 90 -
22 105 2 90 -

Table 2: Legend for Figure 2.


Figure 2: Gel run with every box combination, Part 2.

Lane Hinge Torque Brace Position Brace Angle Other
23 0 1 135 -
24 35 1 135 -
25 70 1 135 -
26 105 1 135 -
27 0 2 135 -
28 35 2 135 -
29 70 2 135 -
30 105 2 135 -
31 - - - 1kb Ladder
32 0 None - Sealed, Unpurified
33 0 None - Unpurified
34 35 None - Unpurified
35 70 None - Unpurified
36 105 None - Unpurified
37 0 1 45 Unpurified
38 0 2 45 Unpurified
39 0 1 90 Unpurified
40 0 2 90 Unpurified
41 0 1 135 Unpurified
42 0 2 135 Unpurified
43 - - - Empty
44 - - - 1kb Ladder

Table 3: Legend for Figure 3.

The PEG purification appears to have been successful, with staples evidently removed and no other obvious difference between corresponding lanes. Again, the only significant difference between box varieties occur between the sealed box (lane 2) and all other boxes. As a result, it seems that all boxes have successfully formed, with the sealing staples functioning as they should.

To confirm the formation of DNA boxes, we imaged sealed boxes using TEM. A representative micrograph (Figure 4), shows many box-shaped structures. These boxes appear to contain a small, darkened cavity in their interior, which corresponds to the size and shape of the cavity that would hold the payload.


Figure 4: BM0044 (Sealed), 3/8/16. Pale boxes can be seen on a darker background. A small cavity, to fit the payload, can also be seen.

The dimensions of 52 of these boxes in the TEM micrographs were then measured to check if they were consistent with the theoretical predictions (Table 4, Figure 5).

  Length (nm) Height (nm) Aspect Ratio
Theoretical 19.9 39.1 1.96
Experimental 20.3 ± 1.81 39.2 ± 2.83 1.95 ± 0.21

Table 4: Sizes of 52 boxes, compared to their theoretical value.


Figure 5: The distribution of box sizes in BM0044, normalised to the theoretical value. Box sizes appear to be centred around the theoretical dimensions.

Materials and Methods

Boxes were produced, as described in DNA design. These structures could then be analysed by the following methods:

Gel Electrophoresis

To prepare a gel, 1.5% w/v agarose was dissolved in gel buffer (0.5x TBE and 11mM MgCl2) by heating and shaking until the opacity disappears. The solution was allowed to sit until warm to the touch. Redsafe stain and pour to let set. Once the gel was set, it was covered with Gel Buffer (stained with Redsafe) and placed in a tank, on ice. Gel loading buffer was added, and 6ul of sample is loaded into the wells. The samples are then run through the gel at 70V for 2 hours.

TEM of DNA Origami

2µL DNA origami structures with a concentration of 10 to 15nM were drop cast onto formvar carbon-coated grids. This was left to adsorb for 5 minutes at room temperature. Excess sample was wicked off, using Whatman filter paper.

Grids were then negatively stained using 2% uranyl acetate solution, by twice dipping grids onto separate droplets of uranyl acetate solution, and leaving it on a third droplet for 2 minutes. Excess stain was again wicked off, using Whatman filter paper.

The grids were then allowed to dry until imaging. This was performed using an FEI Tecnai G2 20 TEM, in bright field mode, at 200kV.

Determination of Box Sizes

In order to determine the sizes of the DNA origami boxes (Table 1), the position of each of the four corners of the DNA boxes were recorded (in pixels). This can then be converted to a length, with Pythagoras’ theorem, and this can be converted to nanometres by comparison with the scale bar.

The “length” of the boxes is the average distance of the two longer sides of the structure, and in Figure 2 is normalised with respect to the longest theoretical dimension of the boxes. The “height” of the boxes is the average distance of the two shorter sides. In Figure 2, this is normalised with respect to the shortest theoretical dimension of the boxes.


The observed structures strongly resemble the DNA boxes that we intended to create. Some small issues could be observed with the viewed structures, however.

In particular, the extreme conditions of TEM are likely to cause many issues with particle size and shape. Although less damaging than other techniques, like AFM, some damage and deformation may occur. Samples for TEM solutions must be dried and placed in a vacuum, for example, so behaviour in solutions can’t be modelled. Osmotic pressures of solution can also cause swelling or contraction of DNA structures, while the staining process can obscure data. As a result, absolute dimensions may not be a reliable value to compare.

In addition to this, the size of the boxes resemble TEM’s limit of resolution. Although we can take clear images of multiple boxes, images individual boxes are much less clear. This causes some measurement error. Additionally, the orientation of the boxes cannot be precisely determined; the boxes heights and widths are slightly different, and we are unable to determine which edge is exposed in the micrographs of the sealed boxes.

Despite this, however, the similarity of the theoretical and experimental aspect ratio, and the presence of the cavity in these boxes, strongly suggest that these observed structures are the intended DNA boxes.


Due to the visual appearance of the DNA origami structures (including their shape and the presence of the cavity), in addition to the similarity of their physical dimensions to the theoretical model, allow us to confidently conclude that our boxes have successfully formed and sealed.

TEM 2: Effectiveness of Hinges and Braces


The novelty of our DNA boxes relies on two main sections: the hinge torque, which causes the box to encapsulate a payload without the application of sealing staples; and the braces, which hold the box open until they are cut. Preliminary SAXS results suggest that brace position 2 is successful at increasing the held angle of the boxes.

It is hypothesised that the presence of braces should create a larger proportion of opened boxes, while the presence of hinges should create a larger proportion of closed boxes. Boxes that contain both braces and hinges, and neither, should have an intermediate proportion.

By observing and comparing the amounts of boxes that are considered to be opened and closed, with varying combinations of hinge torque and braces, the effectiveness of these staples can be determined.


To determine the effectiveness of hinge and brace structures on the state of our DNA origami boxes.


Images were taken of boxes with and without hinge torque, and with and without a brace (Figures 6 to 9).


Figure 6: BM0077 (No Hinge Torque, No Brace), 26/9/16.


Figure 7: BM0081 (70° Hinge Torque, No Brace), 27/9/16.


Figure 8: BM0089 (No Hinge Torque, 135° Brace in Position 2), 29/9/16.


Figure 9: BM0113 (70° Hinge Torque, 135° Brace in Position 2), 29/9/16.

Whether these boxes were opened or closed were also determined (Table 5). It is observed that regardless of box configuration, the proportion of boxes that are opened and closed are roughly 50%. Statistically significant differences (p < 0.5) can be seen between the application a hinge and/or a brace to a box (which all seem to open the box), but no other statistically significant changes can be seen.

  Untorqued 70° Torque
No Brace 44.12% 53.69%
135° Brace 55.53% 51.05%

Table 5: Percentage amount of opened boxes with various staples.

Materials and Methods

Sample synthesis and preparation occurred as described in TEM 1. For this experiment, however, glow discharging was performed on grids in order to improve image quality.

Glow Discharge

In order to produce ideal, negatively stained samples, without agglomeration, glow discharging was necessary. Grids were placed in a PELCO easiGlow™ Glow Discharge Cleaning System, and glow discharged in a reduced atmosphere of air, for approximately 1 minute. The grids were then removed, and the sample was immediately added.

Observation of boxes

Box status was manually and determined. Three micrographs of varying box compositions (Table 6) were randomly assorted, and an observer classified every box in each micrograph as either “opened” or “closed” (Figure 10). At least 600 boxes were sampled of each composition.

Name Hinge Torque Brace Angle
BM0077 0 None
BM0081 0 135
BM0089 70 None
BM0113 70 135

Table 6: List of boxes analysed to determine hinge and brace effectiveness.


Figure 10: Various types of boxes observed in BM0077. a) Open box, as two halves can be seen. The box is thought to be splayed fully open. b) Open box, with one edge appears particularly bright. The object is thought to be opened to approximately 90 degrees; the brighter edge is thought to be a half that rises towards the observer. c) Open box, as a small groove can be seen between two bright halves. The box is thought to be opened, with its hinge pointing upwards. d) Closed box, a bright figure with a darker cavity at its centre.


The collected results are surprising. From these results, unbraced, untorqued boxes appear to be the most closed boxes, and these results are apparently statistically significant. This is somewhat expected in the case of the braced boxes; the brace should serve to hold the box opened, even in the presence of a hinge torque. The torqued, unbraced boxes, however, appear to be more opened than their untorqued counterparts.

It is possible that the application of the torque could have damaged the structural stability of the hinge, leaving a higher proportion of boxes opened; however, it seems more likely that this result is an artefact of the experimental method. The classification of boxes into opened and closed conformations are subjective, and susceptible to the bias of the observer. Although steps were taken to inhibit these effects (classification by one observer, without knowing the origin), these controls are not perfect and human error would still occur.

Additionally, the flaws of TEM itself may cause issues. It is certainly possible that the drying or staining conditions of the imaging technique changed the conformation of the boxes, while imaging the boxes in different orientations creates ambiguity and variability in what could be defined as “opened” or “closed”. It is also a possibility that some boxes that are considered open could be simply be two boxes that have dimerised, or simply lie next to each other.

These flaws could introduce several errors into the results and subsequent analysis, and thus, it is difficult to say that, in reality, the application of hinges and braces do cause more boxes to be opened, compared to their absence.


The effectiveness of the braces and hinge torque are undetermined. Although the application of braces and hinges create statistically significant differences in boxes (p < 0.05), the methodology introduces a substantial amount of error, and fail to monitor the structure’s properties in solution. As a result, we are unable to state how effective these structures are, and further tests must be performed.

TEM 3: Binding of a Payload Sample


The ultimate purpose of our box is to capture a molecular payload, which could theoretically be transported and released elsewhere in the body. If this payload does lie within the box cavity, it should then be visible in TEM images.

GFP is a protein that should be large enough to be visible in TEM micrographs, and binding sites have previously been developed for FRET analysis. The use of these hinges, then, in combination with GFP, should therefore allow GFP to be visible within the cavities of our open boxes.


To generate micrographs of a model payload within our box.


GFP was added to sealed closed and braced open boxes that contained conjugation sites for GFP. Micrographs were then taken of these samples.


Figure 11: BM0124 (Sealed, Ni-NTA Hinge), 29/9/16.


Figure 12: BM0125 (90° Brace in Position 2, Ni-NTA Hinge), 17/10/16.

GFP can be seen in these images as white marks throughout the background. However, it seems to have developed an almost even coating along the sample Grid, while the GFP itself cannot be seen inside the boxes. It therefore appears as if the payload does not prefer to lie in the box.

Materials and Methods

Boxes were made under normal conditions, with the addition of Ni-NTA hinges. Preparation of GFP and NTA was performed in a mostly similar way as that described in FRET. Due to time constrains, however, GFP was only permitted approximately 20 minutes to incubate at room temperature; in anticipation of this, the GFP was added in 20x excess.


While FRET results clearly show a positive result to the binding of GFP to our Ni-NTA hinges, TEM now appears to show a negative result. Although strange, this discrepancy seems as if it could be justified by the limitations of the experimental method.

The key limitation appears to be the difference in GFP concentration and incubation time; although the GFP used in FRET was only at half the concentration to that used here, it was allowed to incubate for almost 10 times as long. This increased incubation time should make it more likely for the GFP to attach to the box, and thus, it seems quite possible that this deviation is the cause of the discrepancy.

In addition to this, the qualitative confirmation of GFP binding in TEM is more variable than the quantitative validation provided by FRET. In bulk FRET, a large amount of boxes are simultaneously analysed for the presence of the GFP in the boxes. Here, only a small portion of boxes can be imaged, and, even then, it remains possible that the presence of GFP is obscured by the stain, or the box itself.


Although FRET seems to validate the binding of a GFP payload to our boxes, TEM does not appear to corroborate with these results. Further testing should be performed with both FRET and TEM, ideally with identical, extended GFP incubation conditions, to more definitively state the ability of a payload to bind to our boxes.


Han, D., Pal, S., Nangreave, J., Deng, Z., Liu, Y. and Yan, H. (2011). DNA Origami with Complex Curvatures in Three-Dimensional Space. Science, 332(6027), pp.342-346.