Detecting topological variations of DNA at single-molecule level

In addition to their use in DNA sequencing, ultrathin nanopore membranes have potential applications in detecting topological variations in deoxyribonucleic acid (DNA). This is due to the fact that when topologically edited DNA molecules, driven by electrophoretic forces, translocate through a narrow orifice, transient residings of edited segments inside the orifice modulate the ionic flow. Here we utilize two programmable barcoding methods based on base-pairing, namely forming a gap in dsDNA and creating protrusion sites in ssDNA for generating a hybrid DNA complex. We integrate a discriminative noise analysis for ds and ss DNA topologies into the threshold detection, resulting in improved multi-level signal detection and consequent extraction of reliable information about topological variations. Moreover, the positional information of the barcode along the template sequence can be determined unambiguously. All methods may be further modified to detect nicks in DNA, and thereby detect DNA damage and repair sites.

1.This is about the translocation events in SiNx nanopore. Authors used AFM to image the ds-ss-ds DNA hybrid molecule and reckoned that "the overall yield of forming ds-ss-ds complex is more than 80%". Then classify the spike-like signatures into '212', corresponding to complete translocation; '2' and '21', that was thought as partial translocation. And only 10% of the signatures were assigned to the '212'. Authors attributed this to a high entropic barrier for a complete translocation in such a small nanopore, but not explained clearly. How does the entropic barrier affect the translocation behavior? Why the complex translocates partially? And what about the rest part of the complex? Did the complex unzip under the interaction of nanopore? (C hem. C ommun., 2017,53, 3539-3542) Authors need to give a more detailed explanation. 2.Authors concluded that "only a small fraction of spike-like event can be attributed to translocation" in abstract. How did authors get this conclusion? C an this small fraction be estimated? 3. Table 1 are not clear. In addition, b should have been β? 4.In MoS2 experiment, author thought a salt gradient condition can "facilitate the capture of DNA segments and elongate the translocation time". How did authors get this conclusion without controllable experiments? If it's from the literature, authors should add a reference identifier. 5.Under the action of an external electric field, the electrons of the semiconductors, such as SiNx membrane, will move in a direction to form a current. For MoS2, there is also leakage current under applied voltage, which can be seen in Figure S4. C an authors explain the resource of the leakage current? 6.Authors need a uniform format, such as MoS2 should be MoS2 in page 14. 7.In Page 6，the author state that "the overall yield of forming ds-ss-ds complex is more than 80%" which is based on their proposed translocation mechanisms of ds-ss-ds complex. The author should use other characterization methods to determine the yield of ds-ss-ds complex. 8.The author ascribes the "212" type of signature to the real translocation the ds-ss-ds complex. However, the entrance of first ds-ss part of DNA follows by a bump out behavior could also contribute to "212" type of events. The author should design an asymmetric ds-ss-ds complex to confirm the real translocation signature. 9.Previous studies on biological nanopores have already show the "partial" translocation through the nanopore (Nat. Nanotechnol., 11, 713-716, 2016, C hem. C ommun. 48, 8784-8786, 2012. Please discuss the results presented in this manuscript with the previous studies on biological nanopores. 10.As shown in figure 2b, the "212" events own the distinguishable two level current with large current difference at nA level. However, the author further uses the nosie analysis to "discriminate" these "212" events. What's the advantages of frequency analysis for discriminating of "212" events? What can we learn from the frequency noise analysis? For example, is there any new knowledge could be gained for the translocation process of DNA complex? the interactions between pore and analyte? Since the nanopore current traces is non-stationary and non-linear, it is improper to use the PSD analysis to gain the detail frequency information, see Faraday Discussions, DOI: 10.1039/C 8FD00023A (2018) 11.The author demonstrates that they have achieve 1 nm spatial resolution. If it is the real 1 nm spatial resolution, this presented nanopore should discriminate 1 nm difference along the whole length of DNA not only the barcode part. On the other hand, the author should consider the conformational changes of the DNA complex induced by the barcode when they propose the "1 nm spatial resolution". 12.The previous nanopore study reported the 0.07 nm3 resolution for reading the primary structure of a protein (Nature Nanotechnology, 2016, 10.1023/NNANO.2016. The author should cite this paper and compare their spatial resolution to this paper.
Reviewer #2: Remarks to the Author: The manuscript by Liu et al describes the use of SiNx and MoS2 pores for distinguishing ssDNA from dsDNA in a passing DNA strand. While this is an important step in barcoding DNA, there are several issues that raise concerns and these should be addressed in full prior to consideration for publication. Some of these are minor and suggestive, and they are not organized in any particular order of importance, but all should be addressed if possible, because the remaining points question the reproducibility and robustness of what is shown in the manuscript.
-p-12: electrodes is mis-spelled -authors refer to distance between bases to be ~0.3nm, however, in ssDNA this distance is larger (~0.5nm), because ssDNA does not have a helical structure. -p-13: it is not clear why oxide protruding from MoS2 pore would slow down DNA. C an the authors develop this idea, or provide a reference that explains this? -on many occasions the authors refer to hemolysin as hemolysins, I wonder what the reason is? -the AFM image in S2 looks weird, the scan lines are at a strange angle, can the authors perhaps show a normal scan to coney their point? The 3D representation looks not convincing and artifactfull.
-on page 7, the authors say that the issue of entropic barrier in solid-state pores has rarely been discussed, however, there are several papers that discuss this with silicon nitride pores, including Wanunu 2008 Biophysical J. and van den Hout Biophysical J. 2010 to name a couple. Also, C arson et al (Biophysical J 2015) did a very systematic study of DNA going through pores in nitride pores of the same approximate geometry as in this paper.
-SI, p-11: "As a result, we are able to...?" -What is the yield of the DNA constructs? It would be valuable to see gels.
-the authors do not show any continuous traces, to allow one to see the quality of the pores.
-in the events shown in Figure 2, the open pore levels before and after most events are substantially different, as indicated by the raw (blue) trace as well as the fit (red). Why is that the case? -The barcode data with MoS2 pore is not too convincing, especially since very few datapoints are shown as traces, in contrast to the large number of points in the scatter plots. In 2 of the 7 events shown, the spike corresponding to dsDNA looks like noise, which goes both up and down for the events.
-The barcode current level of the histogram in the figure does not have a distinct characteristic peak, but rather looks like the tail of the ssDNA peak.
-Have any of these results been repeated for multiple pores? If so, what is the pore-to-pore variation?
Liu and co-worker's manuscript demonstrated that the solid-state nanopore can be used to detect topological variation in single DNA molecules. They applied ds-ss-ds DNA to explore the success rate and velocity profile of translocation events in SiNx nanopore. They claimed the high spatial resolution of 1 nm of the presented MoS2 nanopore. This is a very interesting paper and the experiments is well performed. However, the presented mechanism of DNA translocation seems plausible. The frequency analysis seems hardly supports the translocation mechanism. And the "1 nm spatial resolution" they claim should be confirmed by their experimental results. At this stage, I could not recommend this manuscript to be publish in Nature Communications without a major revision.
We thank the reviewer for acknowledging the importance of our work. Regarding the concerns about the translocation mechanism and the frequency analysis, we made detailed discussions in the revised version. Regarding the statement of "1 nm resolution", the aim of this paper is to provide a methodology to target specific short sequences instead of reading every single base in DNA. Due to the thin nature of MoS 2 membranes, we are able to detect 1 nm motif attached to ssDNA. As we state, in the abstract "In addition to their use in DNA sequencing, ultrathin nanopore membranes have potential applications in detecting topological variations in DNA", this paper gives a roadmap beyond sequencing.
1.This is about the translocation events in SiNx nanopore. Authors used AFM to image the dsss-ds DNA hybrid molecule and reckoned that "the overall yield of forming ds-ss-ds complex is more than 80%". Then classify the spike-like signatures into '212', corresponding to complete translocation; '2' and '21', that was thought as partial translocation. And only 10% of the signatures were assigned to the '212'. Authors attributed this to a high entropic barrier for a complete translocation in such a small nanopore, but not explained clearly. How does the entropic barrier affect the translocation behavior? Why the complex translocates partially? And what about the rest part of the complex? Did the complex unzip under the interaction of nanopore? (Chem. Commun., 2017,53, 3539-3542) Authors need to give a more detailed explanation.
We further implemented the "Kramer" barrier crossing model to elaborate the scenario, where the entropic barrier stemmed from the small nanopore used plays a vital role. The probability of a complete translocation has an exponential dependence on the barrier. We also used a similar barrier model for understanding molecular transport in nanopore in our previous publication (Nat. Nanotechnol., 10, 1070, 2015. The pore entrance can reject the rest of the complex. The unzipping can only happen when the pore diameter is smaller than dsDNA as reported in the above paper (Chem. Commun., 2017,53, 3539-3542). In our case, the pore diameter is larger than dsDNA. To clarify the "In contrast, for narrow nanopores (~ 3 nm), the entropic barrier plays a vital role. We used a barrier model for understanding molecular transport cross a nanopore in our previous publication.16 The probability of a complete translocation has an exponential dependence on the barrier energy." 2.Authors concluded that "only a small fraction of spike-like event can be attributed to translocation" in abstract. How did authors get this conclusion? Can this small fraction be estimated?
Based on the fact that, from AFM measurements, we got high percentage of "212" complex as input materials for translocation experiments. However, we only observed a rather low percentage (approximately 10%) of "212" events.
3. Table 1  We cited the relevant literature (Nat. Nanotechnol., 5, 160, 2010) in the revised revision, where a salt gradient was used to increase the capture rate.
5.Under the action of an external electric field, the electrons of the semiconductors, such as SiNx membrane, will move in a direction to form a current. For MoS2, there is also leakage current under applied voltage, which can be seen in Figure S4. Can authors explain the resource of the leakage current?
The low level of theleakage current can originate either from the defects on MoS 2 or the interface between SiN x and MoS 2 . The leakage current is usually a good indicator that electrochemical reaction (ECR)(REF Nano Letters)can happen for the lower transmembrane voltage range ( 0.8-1.5V). In the absence of the defects ECR threshold increases.

6.Authors need a uniform format, such as MoS2 should be MoS2 in page 14.
We unifromized abrevation for molybdenum disulfide it in the revised version. 7.In Page 6，the author state that "the overall yield of forming ds-ss-ds complex is more than 80%" which is based on their proposed translocation mechanisms of ds-ss-ds complex. The author should use other characterization methods to determine the yield of ds-ss-ds complex.
The overall yield is deduced from AFM measurements. AFM gives a direct statistic based on counting single molecules. We didn't perform bulk analysis, such as gel electrophoresis as the final concentration of our product was low (~ 1 nM).
Revised text is " The overall yield of forming ds-ss-ds complex is more than 80%, a result obtained after analyzing a number of such scans. AFM gives a direct statistic based on counting single molecules." 8.The author ascribes the "212" type of signature to the real translocation the ds-ss-ds complex. However, the entrance of first ds-ss part of DNA follows by a bump out behavior could also contribute to "212" type of events. The author should design an asymmetric ds-ss-ds complex to confirm the real translocation signature.
We used a "212121" design to prove the real translocation. As illustrated below, an asymmetric feature has been observed. We discussed the previous studies as mentioned by the referee and we have modified the revised version to include this suggestion by adding the following clarification.
Revised text is "Partial translocatons have also been observed in biological nanopores,24-26 as well in the context of narrow solid-state nanopores.27,28 However, most of the solid-state nanopores are relatively big, facilitating complete translocation." 10.As shown in figure 2b, the "212" events own the distinguishable two level current with large current difference at nA level. However, the author further uses the nosie analysis to "discriminate" these "212" events. What's the advantages of frequency analysis for discriminating of "212" events? What can we learn from the frequency noise analysis? For example, is there any new knowledge could be gained for the translocation process of DNA complex? the interactions between pore and analyte? Since the nanopore current traces is non-

stationary and non-linear, it is improper to use the PSD analysis to gain the detail frequency information, see Faraday Discussions, DOI: 10.1039/C8FD00023A (2018)
The aim of our frequency analysis is to provide a precise means to characterize nanopore noise at different current levels and for different DNA topologies. It is correct that some "212" events have visually distinguishable current differences at the nA level, but many of them do not. In order to automatically detect "212" events, we developed two new detection methods, operating in the time and frequency domain.
Regarding the comment "Since the nanopore currenpt traces is non-stationary and non-linear, it is improper to use the PSD analysis to gain the detail frequency information, see Faraday Discussions" we agree with the reviewer that there is reason to believe that the traces and nonstationary and nonlinear, but the reviewer should be aware that we are not trying to detect/estimate the bases, just changes in the topological structure --this makes the problem analytically more tractable. The Hilbert transform method outlined in the 2018 work suggested by the reviewer performs poorly on our datasets, and we provided a number of examples to show that this is the case. Frequency analysis (i.e., PSD analysis) is the most commonly used tool in the field, and unlike the Hilbert transform method, it has analytical performance guarantees.
We have added detailed discussions in the revised main text and SI (section S3.4).
"We conclude the analysis by pointing out that a recent trend in nanopore signal processing is to use the Hilbert-Huang Transform (HHT)33. HHT is a useful empirical tool for analyzing nanopore traces as it is designed to work for signals that are nonstationary and nonlinear. Still, the HHT method may not be useful for topological signal detection, as the information-bearing signal in this setting has significantly simpler statistical properties than the one encountered when performing DNA bases detection or estimation. To illustrate the point, we performed the HHT analysis, and one of the findings is shown in Figure SX. As may be seen, it is extremely hard to classify "2" and "1" fragments based on the HHT. A detailed discussion is relegated to SI." 11.The author demonstrates that they have achieve 1 nm spatial resolution. If it is the real 1 nm spatial resolution, this presented nanopore should discriminate 1 nm difference along the whole length of DNA not only the barcode part. On the other hand, the author should consider the conformational changes of the DNA complex induced by the barcode when they propose the "1 nm spatial resolution".
We demonstrated that 1 nm protruded feature along ssDNA can be distinguished by nanopore sensing. As we stated in the abstract, our aim is not to sequence the bases along the ssDNA, but to provide a barcode method to detect specific part along the strand.
12.The previous nanopore study reported the 0.07 nm3 resolution for reading the primary structure of a protein (Nature Nanotechnology, 2016, 10.1023/NNANO.2016.120). The author should cite this paper and compare their spatial resolution to this paper.
We would like to thank the reviewer for this valuable suggestion, however, given the focus of our paper on the barcoding, we don't find it fitting.
Reviewer #2 (Remarks to the Author): The manuscript by Liu et al describes the use of SiNx and MoS2 pores for distinguishing ssDNA from dsDNA in a passing DNA strand. While this is an important step in barcoding DNA, there are several issues that raise concerns and these should be addressed in full prior to consideration for publication. Some of these are minor and suggestive, and they are not organized in any particular order of importance, but all should be addressed if possible, because the remaining points question the reproducibility and robustness of what is shown in the manuscript.
We thank the reviewer for acknowledging the importance of our work. We addressed his useful comments below.
-p-12: electrodes is misspelled Thank you for noticing a spelling mistake, in the revised version we corrected it.
-authors refer to distance between bases to be ~0.3nm, however, in ssDNA this distance is larger (~0.5nm), because ssDNA does not have a helical structure.
We rephrased in the revised version to be more precise.
"equal to twice the distance between two bases in dsDNA" -p-13: it is not clear why oxide protruding from MoS2 pore would slow down DNA. Can the authors develop this idea, or provide a reference that explains this?
Here is the theoretical study that shows the edge chemistry can affect the translocation behavior (ACS Nano, 2014, 8, pp 7914-7922). MoS 2 nanopore benefits from a craftable pore architecture (combination of Mo and S atoms at the edge). Functionalization on the pore edge has also been proposed to slow down the transport of DNA.
Revised text is "Electrochemically etched MoS 2 nanopores have oxide protruding around the pore fringe, which has various possibilities of functionalization, and thus can further slow down the translocation." -on many occasions the authors refer to hemolysin as hemolysins, I wonder what the reason is?
We made the correction. It should be hemolysin.
-the AFM image in S2 looks weird, the scan lines are at a strange angle, can the authors perhaps show a normal scan to coney their point? The 3D representation looks not convincing and artifact-full.
We agree with the reviewer that the 3D representation doesn't give added information. We removed it for S2 to avoid misunderstanding.
-on page 7, the authors say that the issue of entropic barrier in solid-state pores has rarely been discussed, however, there are several papers that discuss this with silicon nitride pores, including Wanunu 2008 Biophysical J. and van den Hout Biophysical J. 2010 to name a couple. Also, Carson et al (Biophysical J 2015) did a very systematic study of DNA going through pores in nitride pores of the same approximate geometry as in this paper.
Thanks for pointing out the issue. We cited the relevant papers in the revised version.
As a result, we changed the phrasing to be: "The partial translocations have also been observed in biological nanopores,24-26 as well in the context of narrow solid-state nanopores.27,28 However, the phenomenon is less frequently observed in the most solid-state nanopores, as their sizes were relatively large comapred to biological pores." -SI, p-11: "As a result, we are able to...?" It should be "The fitted parameters are listed in Table 1." -What is the yield of the DNA constructs? It would be valuable to see gels.
AFM gives a direct statistic based on counting single molecules. The amount of the final complex is very low (~ 1 nM) to run a gel experiment.
-the authors do not show any continuous traces, to allow one to see the quality of the pores.
We added a continuous trace to show the pore condition. Figure R2. A 2.5 second continuous trace low-pass filtered at 10 kHz.
-in the events shown in Figure 2, the open pore levels before and after most events are substantially different, as indicated by the raw (blue) trace as well as the fit (red). Why is that the case?
This is due to the local environment of ions affected by the translocation. It is noted that the very few data points after one event show very large fluctuation, especially in small nanopores. As shown in this work (Science. 2010;327: 64-67.), this is probably due to the relaxation of ions inside the nanopore after translocation. At a longer timescale, for example, as shown in Figure 3a, the baseline is generally constant.
-The barcode data with MoS 2 pore is not too convincing, especially since very few datapoints are shown as traces, in contrast to the large number of points in the scatter plots. In 2 of the 7 events shown, the spike corresponding to dsDNA looks like noise, which goes both up and down for the events.
The complex with ds part has a significantly longer dwell time compared to ss-DNA. We speculate that the barcoded ds part has a specific interaction with the nanopore edge, which may slow down the speed. Moreover, as shown in Figure 4e, two entering signatures have been experimentally observed. This, indicates that the secondary-level blockage indeed originates from the topological change along the DNA strand. In the control experiments, we have translocated ssDNA and no 2 level events were observed. Regarding the noise, as we discussed for ds-ss-ds complex, a change in the topology can affect the noise features significantly.
-The barcode current level of the histogram in the figure does not have a distinct characteristic peak, but rather looks like the tail of the ssDNA peak.
All the ssDNA current values are lower than 2 nA. And dsDNA level has much fewer data points compared to ssDNA level and openpore level. However, three distinct levels are distinguishable in the histogram.
-Have any of these results been repeated for multiple pores? If so, what is the pore-to-pore variation?
Yes, figure 2 and figure 3 are from different pores, which have different current levels. "212" features have been observed in different pores. Different pores may have different entropic barriers, thus affecting the distribution of clusters.