Real-time imaging of sulfhydryl single-stranded DNA aggregation

The structure and functionality of biomacromolecules are often regulated by chemical bonds, however, the regulation process and underlying mechanisms have not been well understood. Here, by using in situ liquid-phase transmission electron microscopy (LP-TEM), we explored the function of disulfide bonds during the self-assembly and structural evolution of sulfhydryl single-stranded DNA (SH-ssDNA). Sulfhydryl groups could induce self-assembly of SH-ssDNA into circular DNA containing disulfide bonds (SS-cirDNA). In addition, the disulfide bond interaction triggered the aggregation of two SS-cirDNA macromolecules along with significant structural changes. This visualization strategy provided structure information at nanometer resolution in real time and space, which could benefit future biomacromolecules research.

The authors have done an excellent job imaging SS-DNA strand dynamics using in liquid EM. THey are to be congralated on this feat alone.
I do think this paper should be published, however, there are some issues with interpretation as a simple S-S bond formation or exchange. There are very few controls to isolate this bond and the spatial resolution is no way good enough to state that the chemical action is occurring solely at the S-H site as a unique bond activation process. THe authors do not seem to appreciate the harsh environment that the shower of probe electrons produce though the very scattering process exploited for imaging.
The authors should consider: -state explicitly the electron dose and give this key factor such a short shrift in their treatment. They need to educate themselves better on the electron pulse radiolysis processes going on in parallel. See work of F. Ross, Sneider on this essential process. Even if the authors want to insist "low dose" there will still be electron radical chemistry induced by the electron beam and the time window is well sufficient for even minute channels to lead to an effect.
-the authors have to consider other strand breaks, radical processes than just S-S bond processes and exchanges. Radicals, bond breaking with radical formation will ALSO form cross linkage to other DNA strands.
-the only distinction comparing SH-ssDNA is the video of ssDNA clustering dynamics. There is not much quantitative done with this one control and it is not convincing as the SH groups will change the surface adhesion and prospect for collision induced growth..with nothing to do with S-S exchange processes.
-THe authors state ssDNA forms a double helix. THis would be unusual and not sure this is what they meant and how they determined this point.
TO be concise, there resolution is no way good enough to uniquely assign the circulization of DNA and formation of clusters solely to S-S bond exchange. The authors need to consider the multitude of other pathways and properly cite other possible channels. The resolution should be significantly improved if they wish to make any claims that the chemistry is occurring at the SH points. Also they need to respond to how the could even observe the DNA circular clusters as simple estimates of translational diffusion and especially rotational diffusion should wash out details. It is clear that surface effects are retarding the DNA motion and therefore surface effects will also affect the chemistry. THe SH groups may be doing nothing more than increasing the surface adsorption and leading to better resolution of clusters that stick more. I hope the authors take the above comments as constructive. These are beautiful results. THey just need to be more open to other interpretations than a single point chemistry at SH points. Certainly better resolution and controls (post analysis) is needed to claim S-S, SH specific chemistry, nevermind trying to make the link to role of disulphide bond exchanges to protein structures. I am sure with some more circumspect discussion of possible mechanisms this paper should be published. It is an important step that follows up on previous work (refs 19-21) with better resolution.
Reviewer #2 (Remarks to the Author): The manuscript by Zeng claimed to report real time imaging of aggregation and rearrangement of single-stranded DNA as triggered by disulfide bond chemistry. Authors started to present images that matched up with the size of single-stranded DNA, whose size were observed to increase for a sample solution contains higher concentration of single-stranded DNA. The larger size objects were interpreted as aggregates modulated by disulfide bond chemistry. Authors identified some coalescence events and presented how smaller objects evolve towards larger objects. Lastly, authors picked some images as diverse forms of aggregates, from dimer to trimer, to hexamer, and some morphological changes of one particle as a function of time. Overall, not enough control experiments and quantitative analysis were performed to support the claim. Factors that are known to affect chemistry and motion of these molecular processes such as surface adsorption and electron beam induced effects were not mentioned and discussed. I would not recommend publication in communication chemistry at this stage, it may worth reconsideration after major revisions in addressing issues about reproducibility and over-interpretations of data.
Some general comments as follows.  Fig. 2C, authors claimed to resolve the helical structure, however, helix is not an expected structure for the base sequence of ssDNA that the authors used, usually it refers to dsDNA and it has a fixed pitch size; (3) there is no statistics about the reproducibility and representativeness for images and processes presented in Figure 4 and 5; (4) authors failed to explain how they ascribe images to different forms of aggregates, dimer, trimer, tetramer, pentamer, and hexamer, and how and why they should form in Figure 5.
3. Discussions of the known electron beam induced effects are lacking. Heating, charging1 and radiolysis2, these effects are in particular important for damage-prone biological samples, unlike nanoparticles. Although experiments from other groups have proven the possibility to image biomacromolecules, for example peptide assembly3, experimental conditions can be very different in experiments reported here. For example, graphene liquid cell is known to better protect fragile sample from radiation damage4 and the surface is more inert. Even for the similar silicon nitride window, as authors used thinner membrane, gap distance, electron dose, chips from different manufactures, more careful evaluation needs to be done. 4. Authors included SH-RNA as samples, however it does not seem to integrate with the main body of research.
5. Some interpretations of motion are not proper, such as the discussion of Brownian motion in lines of 173 to 183. Authors should have noted that surface mediated long jumps can occur for Brownian motion, which can lead to seemingly larger displacements for some steps. The surface effect could be very complicated in a silicon nitride liquid cell for sticking prone biomacromolecules and can be system specific, perhaps, authors should at least show that mean square displacements (MSD) with a slope 1 for a control ssDNA. With this verification, in order to understand whether the attraction indeed exceeds the range of Brownian motion, authors should then analyze the trajectory and MSD with some reproducible statistics of both SH-ssDNA and the control ssDNA. Some specific comments as follows. 1. Abbreviation should be explained when it first appears, like ssDNA. 2. In line 113, mean square displacement (MSD) of the ssDNA center of mass will better quantify the movement of ssDNA to support your claim of "Brownian motion was significantly restricted". 3. In line 116, it is not convincing to connect macromolecules size to the conclusion that a disulfide bond form by active dehydrogenation without any other experimental evidence. 4. From line 132 to 143, this paragraph is only a description of the experiment without corresponding analysis or discussion: In line 132, the authors claim that "Interactions between macromolecules drive spontaneous and continuous self-assembly of SH-ssDNA" without specifying weather the interaction they referred is related to disulfide bond. In line 136, the authors mention that "SH-ssDNA molecules presented different states at different concentrations" without any explanation or discussion.
5. In line 161, "no obvious disulfide bond opening was observed" is not supported by images presented at such a limited resolution. Thus, the conclusion that "SS-cirDNA rearranges through a disulfide-disulfide exchange path instead of a thiol-disulfide exchange path" is not well-supported unless further experimental evidence is provided.
6. In line 184, the title should be changed into "Disulfide bond induced HS-ssDNA morphology change" 7. In line 259, the total amount of data involved in the statistical process should be indicated.

Response to reviewers
Reviewer #1 (Remarks to the Author): The authors have done an excellent job imaging SS-DNA strand dynamics using in liquid EM. THey are to be congralated on this feat alone.
I do think this paper should be published, however, there are some issues with interpretation as a simple S-S bond formation or exchange.
There are very few controls to isolate this bond and the spatial resolution is no way good enough to state that the chemical action is occurring solely at the S-H site as a unique bond activation process. THe authors do not seem to appreciate the harsh environment that the shower of probe electrons produce though the very scattering process exploited for imaging. The authors should consider: -state explicitly the electron dose and give this key factor such a short shrift in their treatment. They need to educate themselves better on the electron pulse radiolysis processes going on in parallel. See work of F. Ross, Sneider on this essential process. Even if the authors want to insist "low dose" there will still be electron radical chemistry induced by the electron beam and the time window is well sufficient for even minute channels to lead to an effect.
-the authors have to consider other strand breaks, radical processes than just S-S bond processes and exchanges. Radicals, bond breaking with radical formation will ALSO form cross linkage to other DNA strands.
Response to reviewer: Thanks for the comment. The possible damage caused by the electron beam have to be considered during the in situ liquid phase TEM test. Water molecule radiolysis by the electrons would generate few species, including hydrated (solvated) electrons eh−, hydrogen radical, H•, hydroxyl radical OH•, and H2. In our experiment, the electron does is 60 e Å −2 s −1 . According to the reference [1], the concentration of the above species is around 10 -7 to 10 -5 mol/L which is a fairly low concentration and the influence on the reaction between SH-DNA is negligible. Parallel electron beams interact with DNA, which can disrupt DNA dynamics and cause structural damage. However, it has been reported that DNA dynamics was believed to be unaffected when the dose rate is not exceeded 110 e Å −2 s −1 . [2] On the other hand, the in situ TEM result highly matches with the DNA electrophoresis experiments that without interference of electron beam. In addition, the buffer ion pairs in solution might neutralize part of the byproducts of electron radiation water and reduce the impact on DNA molecules. -the only distinction comparing SH-ssDNA is the video of ssDNA clustering dynamics. There is not much quantitative done with this one control and it is not convincing as the SH groups will change the surface adhesion and prospect for collision induced growth with nothing to do with S-S exchange processes.
Response to reviewer ： We appreciate the reviewer's suggestion. Considering that in the liquid cell environment, SH groups may alter surface adhesion, and then affects molecular agglomeration. We performed aggregation experiments and electrophoresis in bulk solution, and we still found that SH-ssDNA had obvious aggregation compared with ssDNA (Fig 2c).
The SH group may change the surface adhesion, but we found significant movement of the molecules with 2.6 nm/s (Fig 4ab), and some movie shows the molecules move out of focus and cannot be seen in the field of view, which indicate they are not fully adsorbed on surface.
Our current resolution is indeed insufficient to observe detail S-S bond exchange with SH sites, so we deleted the content of S-S exchange in the manuscript. We will try to improve the resolution in the future study to obtain more clear and solid results.
-THe authors state ssDNA forms a double helix. This would be unusual and not sure this is what they meant and how they determined this point.
Response to reviewer：Thanks reviewer for pointing it out, due the hydrophobic force of ssDNA, these ssDNAs in solution will shrink to a dot, and the SH-ssDNA shrink to a string-like structure due to the SHwould aggregate to form S-S bond, the word helix indeed cannot present the shape changes, we change it according in the main text.
TO be concise, there resolution is no way good enough to uniquely assign the circulization of DNA and formation of clusters solely to S-S bond exchange. The authors need to consider the multitude of other pathways and properly cite other possible channels. The resolution should be significantly improved if they wish to make any claims that the chemistry is occurring at the SH points.
Response to reviewer：We appreciate the reviewer's suggestion. Our current resolution is indeed insufficient to observe detail of bonds, so we changed the description accordingly. We will improve our experiment with better imaging techniques and instruments, hope we could identify detail structure changes with higher resolution in the future.
Also they need to respond to how could even observe the DNA circular clusters as simple estimates of translational diffusion and especially rotational diffusion should wash out details.
Response to reviewer：Thanks reviewer for pointing it out. Figure A shows translational diffusion, we can see that the molecules did not move out focus plane, and the positions of the circular molecules do not change significantly on the Z-axis. If the molecules move up and down in solution, we could not image them all the time, some movie shows the molecules move out of focus and cannot continue to see them. There is a typical rotational diffusion in Figure 5c, where we can see that the white edge appears at 37s and disappears at 120s.
It is clear that surface effects are retarding the DNA motion and therefore surface effects will also affect the chemistry. The SH groups may be doing nothing more than increasing the surface adsorption and leading to better resolution of clusters that stick more.
Response to reviewer： Surface adsorption may affect the DNA motion. The adsorption on the surface might limit the movement and aggregation, however we still observed that lots of DNA molecule are not fully absorbed on the surface, which move out of focus and become invisible. According to literature [1], the viscosity of liquid layer close to surface will increase 2 orders of magnitude, thus caused much slower diffusion dynamics than that in bulk solution. The electrophoresis also clear proof that the SH-ssDNA would aggregate into larger molecules. Therefore, we think the molecule are close to surface but not fully adsorbed on the surface, and molecules are still be able to move freely and aggregate. I hope the authors take the above comments as constructive. These are beautiful results. THey just need to be more open to other interpretations than a single point chemistry at SH points. Certainly better resolution and controls (post analysis) is needed to claim S-S, SH specific chemistry, nevermind trying to make the link to role of disulphide bond exchanges to protein structures. I am sure with some more circumspect discussion of possible mechanisms this paper should be published. It is an important step that follows up on previous work (refs 19-21) with better resolution.
Response to reviewer：Thank you for your kind suggestion. Our analysis and understanding need to be strengthened. We will continue to study this topic in depth and improve imaging resolution, thus obtain solid results which could reveal the detail interaction mechanism of DNA molecules.

Reviewer #2 (Remarks to the Author):
The manuscript by Zeng claimed to report real time imaging of aggregation and rearrangement of singlestranded DNA as triggered by disulfide bond chemistry. Authors started to present images that matched up with the size of single-stranded DNA, whose size were observed to increase for a sample solution contains higher concentration of single-stranded DNA. The larger size objects were interpreted as aggregates modulated by disulfide bond chemistry. Authors identified some coalescence events and presented how smaller objects evolve towards larger objects. Lastly, authors picked some images as diverse forms of aggregates, from dimer to trimer, to hexamer, and some morphological changes of one particle as a function of time. Overall, not enough control experiments and quantitative analysis were performed to support the claim. Factors that are known to affect chemistry and motion of these molecular processes such as surface adsorption and electron beam induced effects were not mentioned and discussed. I would not recommend publication in communication chemistry at this stage, it may worth reconsideration after major revisions in addressing issues about reproducibility and over-interpretations of data.
Some general comments as follows. Response to reviewer：Thanks reviewer for pointing it out. Our current resolution is indeed insufficient to distinguish of closed and opened loop, so we deleted the content of close and open loop in the manuscript. We re-counted the particles and divided them into two groups: the circular group (with middle gaps) and the cluster group (no voids in the center), as Figure 3d.
(2) in line 99 and Fig. 2C, authors claimed to resolve the helical structure, however, helix is not an expected structure for the base sequence of ssDNA that the authors used, usually it refers to dsDNA and it has a fixed pitch size; Response to reviewer：Thanks reviewer for pointing it out. We have edited it. We use "string-like structure" to described the shape in Fig. 2a. (3) there is no statistics about the reproducibility and representativeness for images and processes presented in Figure 4 and 5; Response to reviewer: We appreciate the reviewer's suggestion. We have added 8 pairs quantitative analysis of macromolecules aggregation in the Supplementary Data for Figure 4, and quantitative analysis of diverse forms of aggregates in Figure 5b, which can provide data to verify repeatability. We have revised the content in the manuscript.
3. Discussions of the known electron beam induced effects are lacking. Heating, charging1 and radiolysis2, these effects are in particular important for damage-prone biological samples, unlike nanoparticles. Although experiments from other groups have proven the possibility to image biomacromolecules, for example peptide assembly3, experimental conditions can be very different in experiments reported here. For example, graphene liquid cell is known to better protect fragile sample from radiation damage4 and the surface is more inert. Even for the similar silicon nitride window, as authors used thinner membrane, gap distance, electron dose, chips from different manufactures, more careful evaluation needs to be done.
Response to reviewer：We appreciate the reviewer's comments. The possible damage caused by the electron beam have to be considered during the in-situ liquid phase TEM test. Water molecule radiolysis by the electrons would generate few species, including hydrated (solvated) electrons eh−, hydrogen radical, H•, hydroxyl radical OH•, and H2. In our experiment, the electron does is 60 e Å −2 s −1 . According to the reference [1], the concentration of the above species is around 10 -7 to 10 -5 mol/L, which is a fairly low concentration and the effect on the reaction between SH-DNA is negligible. The parallel electron beam interacts with DNA and this might disturb the dynamics of DNA and cause the structural damage. But it has been reported that DNA dynamics was believed to be unaffected when the dose rate is not exceeding 110 e Å −2 s −1 [2]. On the other hand, the in situ TEM result highly matches with the DNA electrophoresis experiments that without interference of electron beam. In addition, the buffer ion pairs in solution might neutralize part of the byproducts of electron radiation water and reduce the impact on DNA molecules.
Surface adsorption may affect the DNA motion. The adsorption on the surface might limit the movement and aggregation, however we still observed that lots of DNA molecule are not fully absorbed on the surface, which move out of focus and become invisible. According to literature [3], the viscosity of liquid layer close to surface will increase 2 order of magnitude, thus caused much more slower diffusion dynamics. The electrophoresis also clear proof that the SH-ssDNA would aggregate into larger molecules. Therefore, we think the molecule are close to surface but not fully adsorbed on the surface, and molecules are still be able to move freely and aggregate. These discussions are added to manuscript accordingly.