Single molecule analysis reveals monomeric XPA bends DNA and undergoes episodic linear diffusion during damage search

Nucleotide excision repair (NER) removes a wide range of DNA lesions, including UV-induced photoproducts and bulky base adducts. XPA is an essential protein in eukaryotic NER, although reports about its stoichiometry and role in damage recognition are controversial. Here, by PeakForce Tapping atomic force microscopy, we show that human XPA binds and bends DNA by ∼60° as a monomer. Furthermore, we observe XPA specificity for the helix-distorting base adduct N-(2’-deoxyguanosin-8-yl)-2-acetylaminofluorene over non-damaged dsDNA. Moreover, single molecule fluorescence microscopy reveals that DNA-bound XPA exhibits multiple modes of linear diffusion between paused phases. The presence of DNA damage increases the frequency of pausing. Truncated XPA, lacking the intrinsically disordered N- and C-termini, loses specificity for DNA lesions and shows less pausing on damaged DNA. Our data are consistent with a working model in which monomeric XPA bends DNA, displays episodic phases of linear diffusion along DNA, and pauses in response to DNA damage.

The manuscript discusses the results a set of very insightful single molecule experiments, namely AFM and single molecule fluorescence microscopy, designed specifically to characterize the interaction between the NER protein XPA and DNA, with and without one single dG-C8-AFF lesion. It is well structured and well written and the results obtained are very important in clarifying exhaustively a few important aspects still very much in debate in regards to XPA binding to DNA and its potential role in the NER pathway, namely: a) XPA binds both damaged and undamaged DNA as a monomer b) XPA is a monomer when unbound in solution c) XPA causes severe bending in the DNA helix upon binding d) XPA has a higher affinity for damaged sites The latter point is well justified by the authors (although only qualitatively), in terms of the energetic cost of bending an already bent structure (because of the presence of the lesion), relative to the cost of bending a relatively straight one to the same degree. The higher affinity for bent sites can also explain the ability of XPA to locate preferentially at the damage site compared to the non-sequence specific binding observed for the ND538, and shown in Figure 1, panels D vs. E.
The only aspect that I do not find particularly convincing in this version of the manuscript is the suggested role of XPA in DNA damage detection/recognition. The authors themselves state in the introduction that early literature suggests that XPA is involved in the damage detection steps, but that more recent studies suggest that its primary role is as a scaffolding protein supporting the stability of the DNA repair bubble and coordinating the NER proteins/enzymes around it. Nevertheless, without any further discussion, the authors state that one of the main aims of the work is indeed to understand the role of XPA in damage detection, which seems a bit of a nonsequitur to me. In this regard, it would be important for the authors to present/discuss the evidence that they considered as a standpoint or that convinced them that XPA may indeed have a role in DNA damage detection.
Within this framework, it is not clear to me how once the XPA detected the damage the complex would evolve to support XPA's role as a scaffolding protein, migrating away from the damage site (pre-excision) and locating itself on one of the edges of the DNA repair bubble (see the comment Shell and Chazin, Structure (2012) 20(4):566 and refs therein). I believe that the authors should discuss their results within the context of the DNA excision step to clarify how their theory supports or clashes with the current view of the mechanism.
One point that may be of interest for such discussion is that XPA higher affinity for a "bent" DNA helix doesn't necessarily translate in a lesion recognition role. Indeed, within the current (and widely/generally accepted) view of the NER excision complex, XPA binds at one of the two junctions, either 3' or 5 ', not yet 100% proven which one, there are arguments/data supporting each. From a structural point of view looking at the two single strands opening in opposite directions, the junction is a bent DNA construct, not too different from a bent DNA double helix if the protein binds only one of the strands. Within this framework, it is possible that XPA has a higher affinity for bent DNA constructs, because it needs to locate at a junction and not because it is involved in the damage detection step. Maybe the authors should comment also on this point.
In summary, the topics of discussion I suggested above represent a minor revision that may require a slight change in this specific direction/aim. Nevertheless, I believe that such additional discussion would add to an already outstanding work and valuable contribution to the field. For this reason, with these minor revisions addressed, I strongly support the publication of the manuscript in Nature Communications.
Minor point: Reference from Kokic et al, bioRxiv page 3 line 75 should be listed with all other references for consistency.
Reviewer #2 (Remarks to the Author): This manuscript reports on the use of single molecule studies of XPA binding to damaged and undamaged DNA. The authors phrase their study in the context of nucleotide excision repair (NER), where XPA is known to function. However, this is a purely biophysical study of the protein in isolation, so the results only remotely relate to the role of XPA in NER. The data confirm previously published work of high affinity binding to an AAF lesion, that XPA binds DNA substrates as a monomer, and that DNA is bent by ~60° at sites where XPA is bound. Many different modes of linear diffusion of XPA on damaged and undamaged DNA, and the authors place these results in a model of XPA functioning in the search for distorted duplex DNA associated with damage.
The manuscript is written in clear language with a good logical flow. The experimental design, data acquisition, and analyses are technically sound and appear to be of high caliber. However, there is a fundamental flaw in this study when it comes to transfering the observations of the effects of isolated XPA to the broader landscape of the NER pathway. It is well accepted (and even summarized in the Introdcution) that XPA is not present at sites of DNA damage until the damage is recognized by other NER proteins and the TFIIH complex is recruited. TFIIH opens the damaged duplex and unwinds it around the site of damage, then XPA is recruited along with RPA. At this point, XPA is not freely diffusing in solution searching for damage in a duplex, hence it is completely unclear how the results pertain to XPA function in NER.
The results do show that the presence of DNA damage affects how isolated XPA binds and diffuses on the unencumbered substrate. But the absence of the many other components of the large NER machinery (over 30 different proteins) limits the scope and significance of their findings. Given these limitations, it is misleading to direct the Introduction to NER and to try and relate the data to NER, and then only in the last sentence state that the model presented is a 'starting point' and that XPA may well behave quite differently when bound to other proteins.
Overall, the results obtained are relevant to biophysical investigations of DNA repair, but only remotely to the role of XPA in NER. Potential functions of XPA outside of NER have been reported recently. Do one or more of these functional roles better relate to these data? As presented here, the significance and potential impact of these results make this manuscript better suited to a more specialized journal. Concerns 1. The Introduction (pg 3, lines 57-62) states: "XPA has been reported to display specificity for UV-irradiated DNA 6-4PPs, N-(2'-deoxyguanosin-8-yl)-2-acetylaminofluorene (dG-C8-AAF), cisplatin, benzo[a]pyrene, non-hybridized bases, and other artificially distorted substrates. Additional studies revealed that XPA also binds preferentially to partially single stranded DNA and forked substrates more closely resembling unwound NER intermediates. It has been previously proposed that XPA binding to damaged DNA induces conformational changes in the protein that are associated with different binding modes." The experimental design and presentation of results in this manuscript are not consistent with the well accepted role of XPA as a scaffold organizing the many partner proteins in NER complexes through protein-protein interactions. 4. Pg 5, line 119 -Of all the complexes observed, 33% were bound to the ends of the DNA substrate. This is quite a high rate of non-specific binding. Even 22% for the AAF538 substrate seems to be quite a high ratio of non-specific to specific binding. Why? Some comment on this is required in the text. The occurance of cases with two molecules bound at both ends also observed also need to be reported. 5. Pg 11, line 248 -Results show that most (60-70%) of the XPA protein is in a stationery mode on tightrope substrates. Is the remaining 30-40% fraction of motile protein sufficient to conclude that XPA actively screens for DNA-helix distortion? Does this ratio change when more than one damage site per DNA substrate is introduced? How do the authors explain that the motile fraction of the protein changes modes, including pausing, but the stationary fraction does not? 6. Is there any speculation on what the pausing positions on the tightropes are? 7. Pg 16, lines 391-399 -The authors speculate without any specific support that the disordered N-and C-termini may fold in and interrogate the DNA for lesions via DNA bending. Considering that XPA is known to function as the key scaffold for NER pre-incision complexes and the many studies showing that it interacts with many other NER proteins, it is illogical to speculate that the disordered N-and C-termini of XPA are important for DNA binding only. This speculation should be supported by data or completely deleted from the text.
Additional points. 1. Pg 3, line 68 -A zinc finger is a specific structural motif that is characterized by the coordination of one or more zinc ions in order to stabilize the fold; the term should be reserved for such motifs. XPA possesses a Zn-binding motif, but not a Zinc finger and the wording should be adjusted accordingly. 6. Pg 23, line 601 -XPA is known to bind a zinc ion, and the presence of Zn has an influence of the structure. Hence, use of EDTA in the buffer may affect the protein's fold, stability, and solubility. Were controls performed?
Reviewer #3 (Remarks to the Author): The manuscript by Beckwitt et al. presents single molecule atomic force microscopy (AFM) and fluorescence studies of the nucleotide excision repair (NER) factor Xeroderma pigmentosum group A (XPA). The authors aim at characterizing conformational and dynamic properties of XPA-DNA complexes during lesion search and at a target lesion to understand the mechanism of lesion recognition by XPA.
XPA is an essential protein and mutations in XPA severely affect human health. XPA functions in the NER pathway that repairs of a plethora of structurally and chemically diverse DNA lesions, including acetylaminofluorene (AAF), which is used in the study, and in particular and famously UV induced damages in the DNA. The exact role of XPA in the NER mechanism is still controversial. In addition, the full length protein in complex with DNA has not been accessible to X-ray crystallography due to the intrinsic disorder in parts of the protein. The questions addressed by the authors are therefore highly relevant and interesting. In their studies, the authors investigate XPA in isolation (without other NER components present), while in vivo XPA functions in NER in complex with the transcription factor IIH (TFIIH) helicase machinery. However, understanding the conformational dynamics of DNA interactions by the isolated XPA protein adds to our better appreciation of the role of XPA in NER lesion processing.
The authors strive to resolve the question of the oligomeric state of XPA, the DNA bending introduced by XPA in non-damaged DNA and at an AAF lesion, and the mode(s) of translocation on DNA in the absence and presence of lesions. Regarding the oligomeric state of XPA, contradictory findings had been reported in the past. Beckwitt and colleagues clearly establish the monomeric form of XPA to be the dominant species both in solution and when bound to DNA, using AFM and multi angle light scattering. AFM characterization of DNA bending by XPA revealed comparable bending at non-damaged and lesion sites, and the bending angle was similar to that observed in crystal structures (of the shortened construct). The authors argue with energetic cost of DNA bending as a means for XPA to distinguish between lesion and non-damaged DNA sites based on their different flexibilities, as has been proposed before by others for XPA and for other DNA repair proteins. Regarding XPA lesion search dynamics, the authors report several different modes of translocation on DNA and the frequent interchange between these different modes. They interpret the modes as XPA testing for damage (pausing), XPA carrying out one-dimensional diffusion on the DNA in its lesion search (short range motion), and XPA in a fast scanning mode (mid range or long range motion). In a model they summarize their conclusions that XPA lesion search involves three different states: XPA searches for lesions in a fast scanning complex with diffusion coefficient dependent on ionic strength of the buffer (which they interpret as an indication for hopping); XPA probes for lesions in a short range scanning complex with lower diffusion constants, which may be in part caused by DNA bending by the protein to probe for lesions; and XPA forms a strongly bent stable recognition complex at the lesion.
Overall, I find the topic and results of this study interesting and highly relevant and I mostly agree with their conclusions. I do however, have a few points of criticism listed below. After minor changes, I hence recommend the manuscript for publication in Nature Communications.
Specific points: 1. AAF binding specificity determination by AFM: The distribution of XPA positions on the DNA does not really resemble a > 600-fold preference for lesion binding (when compared, for example, to distributions in Yang et al. NAR 2005 as cited by the authors in this context). I agree with the authors that there is specificity for AAF by XPA, but I think the value is overestimated due to underestimation of non-specific background binding. Differently expressed: I think the Gaussian width in the fit to the distribution is too broad, artificially increasing the specific area that results in the high specificity value. A smaller binning interval might more clearly show whether the maximum at the lesion position is truly this broad or whether the Gaussian shaped peak is in fact much narrower, but would likely require more data points. In fact, the specificity that this distribution reflects likely fits nicely with that obtained by the authors from their EMSA analyses. However, based on the high flexibility at an AAF lesion, it actually seems surprising that the specificity of XPA for this lesion should not be larger than observed here, if energetic preference of enhanced flexibility sites determinesthe probability of occupancy. Furthermore, enhancement of TFIIH lesion recognition through XPA, as suggested previously (and also mentioned by the authors) is difficult to imagine with the low to moderate recognition specificity compared to -for instance, that shown for the XPD helicase of TFIIH (admittedly for different lesions). These points could be addressed.
2. XPA volume analyses by AFM: While I appreciate the thoroughness of their volume calibration for their peak force AFM, I think this section could mostly be moved to the Supplement. It is obviously indispensable, but does not add to the manuscript story. The authors analyzed the volumes of two proteins, APE1 and Pol(beta) bound to DNA, to test the validity of their volume calibration also for DNA bound protein volumes. In this context, I also think that the originally measured volumes for APE1 and Pol(beta) (before DNA volume subtraction) ought to be given in the text, in addition to the final volumes after DNA subtraction. After the thorough consideration of the contribution from DNA binding mode by APE1 and Pol(beta) to overall measured volume, it would recommend itself to also discuss this for the protein of interest in this study, XPA to support the approach of subtracting the DNA volume to determine the molecular mass of the DNA bound complex. Again, I think the originally measured volumes (prior to DNA volume subtraction) should also be given, as well as the DNA volumes.

DNA bend angle analyses by AFM:
In the bend angle distributions for DNA in the absence of protein ( Figures 4C and D), the first interval seems to be only half the size of all other intervals since it is centered at zero degrees bending. If this is the case, then at least in the histogram shown in (C), correction to full interval size for the first range of bend angles (0 to 5 or 10 or whatever degrees) would likely result in a distribution with exponential decay from 0 degrees (rather than the 5.6 degrees maximum reported here). For the distribution in (D), at the AAF position, the first maximum very likely remains at the second bin, ie. larger than 0 degrees. In addition, for a better comparison between the two distributions, ideally the interval (bin) size should be the same, which would then be limited by the sample with the smallest number of datapoints (also for Figures 4 G and H). Also, for smaller bin sizes, the distributions in Fig. 4 G and H may well reveal more than one bend angle species, as also suggested by the example images (but smaller binning will be difficult due to small samples sizes). Two different conformational states in lesion search and recognition by XPA would seem interesting also in the context of the final model proposed by the authors. However, none of this will drastically change the story. 4. DNA tightrope assay: a. The authors state that the proportion of immobile XPA is not affected by the DNA substrate. However, Figure 5B shows ~90% versus ~70% immobile XPA complexes on DNA (including dissociating and persistent molecules) for AAF versus UV or non-damaged substrate. The statistical non-significance level seems surprising! The authors state that the density of UV lesions in their substrate can be expected to be comparable to that in the AAF substrate (approximately one lesion per ~2,000 bp), yet the percentage of immobile complexes appears to differ between the UV and AAF substrates. Can the type of lesion explain the different behavior? b. The lack of medium/long range scanning species for AAF versus non-damaged or UV substrate is striking. Long/medium range motion is defined by the authors as movement over > 690 nm, the distance between two AAF lesions and approximate average distance between two UV lesions assuming the above stated density. Exclusively short range scanning by XPA is hence consistent with XPA not proceeding over an AAF lesion. Again, could this depend on the type of lesion, since for the UV substrate a significant population of medium/long range scanning molecules was observed? In contrast to AAF, CPD lesions, as one type of UV lesion, do not strongly distort DNA. c. Many of the sub-species do not seem to contain enough data points for statistical analyses. For example n=3 for long range diffusion coefficients of pooled UV and non-damaged substrates, or n=3 for maximum displacement at 1 M salt concentration. I appreciate that data collection with this type of single molecule experiment is time consuming, but would be careful to draw conclusions from such small sample sizes. d. I do not agree with the classification of long range scanning particles based on the example trace shown in Figure 5D. In this example, the "long range" trace seems to consist simply of two medium range movements interrupted by a clear pausing plateau. I would suggest to pool the data from medium range and long range modes. In the population distributions (e.g. Figure 6F) the long range species never seems to truly add any additional information. Also, the issue of not enough data points for some of the different species (point above) may automatically be removed when pooling medium and long range scanning particles. e. I also wonder if the salt concentration effect on the medium/long range but not on the short range moving species really supports hopping as the mode of translocation in the medium/long range species and one dimensional diffusion for the short range species. Regarding the short range species, the distribution of short range diffusion coefficients ( Figure 6E) appears to indicate one species consistent with one dimensional diffusion (below the limit line for diffusion) and a second with higher diffusion constants (above the limit line for diffusion). Regarding the medium/long range species, it seems to me that hopping is only one possible explanation of the effect observed at higher salt concentration. Could the salt concentration not instead simply affect stability of DNA interactions by the protein? These interactions seem to be majorly electrostatic when looking at published structural data. The example in Supplementary Figure 5c showing transition from paused to long range diffusion when increasing the salt concentration to 1M would be consistent with this. f. I wonder if DNA bending as seen in AFM applies for the tightrope assay or if the DNA is "fully" stretched between the anchoring beads? If XPA would not be able to bend the DNA as part of its lesion probing, this would affect the time spent probing for lesions (ie. bending the DNA, ie. in the paused state as well as in the short range linear diffusion state). The authors mention that other proteins previously tested in this assay did not show the changing between translocation modes seen for XPA. Were any of these proteins expected to bend the DNA?

Responses to each reviewer's comments
We greatly appreciate the care and depth of analysis that each reviewer has provided in their evaluation of our study.
In response to the reviewers' excellent points we have made major revisions to the manuscript (described in detail below, and marked in blue text in the manuscript). These included restructuring the results, refocusing the introduction and discussion per their concerns, doing extensive data re-analysis, and performing two new experiments (increased lesion density on the DNA tightropes and work-up of a truncated XPA variant), which have been incorporated into a new Figure 5. The study has been enhanced by the review process.
Please note all our responses

Reviewer #1
The manuscript discusses the results a set of very insightful single molecule experiments, namely AFM and single molecule fluorescence microscopy, designed specifically to characterize the interaction between the NER protein XPA and DNA, with and without one single dG-C8-AFF lesion. It is well structured and well written and the results obtained are very important in clarifying exhaustively a few important aspects still very much in debate in regards to XPA binding to DNA and its potential role in the NER pathway, namely: a) XPA binds both damaged and undamaged DNA as a monomer b) XPA is a monomer when unbound in solution c) XPA causes severe bending in the DNA helix upon binding d) XPA has a higher affinity for damaged sites The latter point is well justified by the authors (although only qualitatively), in terms of the energetic cost of bending an already bent structure (because of the presence of the lesion), relative to the cost of bending a relatively straight one to the same degree. The higher affinity for bent sites can also explain the ability of XPA to locate preferentially at the damage site compared to the non-sequence specific binding observed for the ND538, and shown in Figure 1, panels D vs. E.
The only aspect that I do not find particularly convincing in this version of the manuscript is the suggested role of XPA in DNA damage detection/recognition. The authors themselves state in the introduction that early literature suggests that XPA is involved in the damage detection steps, but that more recent studies suggest that its primary role is as a scaffolding protein supporting the stability of the DNA repair bubble and coordinating the NER proteins/enzymes around it. Nevertheless, without any further discussion, the authors state that one of the main aims of the work is indeed to understand the role of XPA in damage detection, which seems a bit of a non-sequitur to me. In this regard, it would be important for the authors to present/discuss the evidence that they considered as a standpoint or that convinced them that XPA may indeed have a role in DNA damage detection.
The reviewer (as well as Reviewer #2) brings the prevailing view in the literature, that XPA is involved in later steps in NER. It is interesting to note that reviewer 3 indicates: "The exact role Response to Reviewers of XPA in the NER mechanism is still controversial. In addition, the full length protein in complex with DNA has not been accessible to X-ray crystallography due to the intrinsic disorder in parts of the protein. The questions addressed by the authors are therefore highly relevant and interesting…understanding the conformational dynamics of DNA interactions by the isolated XPA protein adds to our better appreciation of the role of XPA in NER lesion processing." We also embrace this philosophy, and it was thus our intent to better understand XPA's interaction with DNA which we outline on the bottom of page 3 (lines 68-71). The introduction and discussion have been revised to emphasize the traditional role of XPA as a scaffold protein during NER. As attached at the end of the rebuttal (pg. 13) we have created a table showing that XPA displays specificity for specific DNA lesions and specific structures. We believe that the literature strongly indicates that XPA has specificity for bent structures, including largely distorting DNA lesions, which our new AFM data strongly support.
Within this framework, it is not clear to me how once the XPA detected the damage the complex would evolve to support XPA's role as a scaffolding protein, migrating away from the damage site (pre-excision) and locating itself on one of the edges of the DNA repair bubble (see the comment Shell and Chazin, Structure (2012) 20(4):566 and refs therein). I believe that the authors should discuss their results within the context of the DNA excision step to clarify how their theory supports or clashes with the current view of the mechanism.
We appreciate the reviewer raising this point. As we show in this study, XPA interacts with DNA in a dynamic manner going through episodic periods of pauses and then short range motion. Thus, one can easily envision initial binding to the damage site and then XPA's migration away from the damaged site facilitated by its interaction with other factors including XPB and XPD as part of TFIIH. This type of limited diffusion away from damage has been observed for Rad4 and PARP1 from our lab and most recently for XPC from the Scharer and Lee laboratories 1 . We have made this point in the discussion (lines 387-392).
One point that may be of interest for such discussion is that XPA higher affinity for a "bent" DNA helix doesn't necessarily translate in a lesion recognition role. Indeed, within the current (and widely/generally accepted) view of the NER excision complex, XPA binds at one of the two junctions, either 3' or 5', not yet 100% proven which one, there are arguments/data supporting each. From a structural point of view looking at the two single strands opening in opposite directions, the junction is a bent DNA construct, not too different from a bent DNA double helix if the protein binds only one of the strands. Within this framework, it is possible that XPA has a higher affinity for bent DNA constructs, because it needs to locate at a junction and not because it is involved in the damage detection step. Maybe the authors should comment also on this point.
We have updated the discussion (lines 329-331) to acknowledge that a bent DNA helix is a feature of both DNA lesions (such as AAF) and NER intermediates. With regard to a 5' or 3' junction, the idea that XPA is situated at the 5' junction of the lesion site is supported by recent Cryo-EM data from the Cramer laboratory showing a strong extended alpha helix interaction with XPB in TFIIH. Also, previous work showing XPA's interaction with the 5' nuclease ERCC1/XPF further supports the notion that XPA is interacting with a fork structure. Response to Reviewers In summary, the topics of discussion I suggested above represent a minor revision that may require a slight change in this specific direction/aim. Nevertheless, I believe that such additional discussion would add to an already outstanding work and valuable contribution to the field. For this reason, with these minor revisions addressed, I strongly support the publication of the manuscript in Nature Communications.
Minor point: Reference from Kokic et al, bioRxiv page 3 line 75 should be listed with all other references for consistency.
This paper was published after our initial submission and we have updated the citation 2 .

Reviewer #2
This manuscript reports on the use of single molecule studies of XPA binding to damaged and undamaged DNA. The authors phrase their study in the context of nucleotide excision repair (NER), where XPA is known to function. However, this is a purely biophysical study of the protein in isolation, so the results only remotely relate to the role of XPA in NER. The data confirm previously published work of high affinity binding to an AAF lesion, that XPA binds DNA substrates as a monomer, and that DNA is bent by ~60° at sites where XPA is bound. Many different modes of linear diffusion of XPA on damaged and undamaged DNA, and the authors place these results in a model of XPA functioning in the search for distorted duplex DNA associated with damage.
The manuscript is written in clear language with a good logical flow. The experimental design, data acquisition, and analyses are technically sound and appear to be of high caliber. However, there is a fundamental flaw in this study when it comes to transfering the observations of the effects of isolated XPA to the broader landscape of the NER pathway. It is well accepted (and even summarized in the Introdcution) that XPA is not present at sites of DNA damage until the damage is recognized by other NER proteins and the TFIIH complex is recruited. TFIIH opens the damaged duplex and unwinds it around the site of damage, then XPA is recruited along with RPA. At this point, XPA is not freely diffusing in solution searching for damage in a duplex, hence it is completely unclear how the results pertain to XPA function in NER.
The results do show that the presence of DNA damage affects how isolated XPA binds and diffuses on the unencumbered substrate. But the absence of the many other components of the large NER machinery (over 30 different proteins) limits the scope and significance of their findings. Given these limitations, it is misleading to direct the Introduction to NER and to try and relate the data to NER, and then only in the last sentence state that the model presented is a 'starting point' and that XPA may well behave quite differently when bound to other proteins.
As we discuss in our response to Reviewer 1, we do not claim to have defined the role of XPA in NER, or to have shown exactly how XPA finds its target in a living cell. We have emphasized in the discussion that "we present our model as an important starting point, illustrating XPA-DNA binding behavior and diffusive properties in isolation." 4 Response to Reviewers Overall, the results obtained are relevant to biophysical investigations of DNA repair, but only remotely to the role of XPA in NER. Potential functions of XPA outside of NER have been reported recently. Do one or more of these functional roles better relate to these data? As presented here, the significance and potential impact of these results make this manuscript better suited to a more specialized journal. Concerns 1. The Introduction (pg 3, lines 57-62) states: "XPA has been reported to display specificity for UV-irradiated DNA 6-4PPs, N-(2'-deoxyguanosin-8-yl)-2-acetylaminofluorene (dG-C8-AAF), cisplatin, benzo[a]pyrene, non-hybridized bases, and other artificially distorted substrates. Additional studies revealed that XPA also binds preferentially to partially single stranded DNA and forked substrates more closely resembling unwound NER intermediates. It has been previously proposed that XPA binding to damaged DNA induces conformational changes in the protein that are associated with different binding modes." The experimental design and presentation of results in this manuscript are not consistent with the well accepted role of XPA as a scaffold organizing the many partner proteins in NER complexes through protein-protein interactions.
Please see comments to Reviewer #1 with regard to XPA's role in NER. We have softened both the introduction and discussion. Briefly, we would like to emphasize that while we do not disregard the evidence for XPA acting as a scaffold protein, this is not mutually exclusive with its ability to recognize NER lesions in dsDNA. As noted by the reviewer, we have intentionally studied purified protein and purified DNA substrates in order to obtain a basic mechanistic understanding of XPA's interaction with damaged DNA. Our work does not preclude a role of XPA as a scaffolding protein in NER. We are simply trying to understand its stoichiometry, and DNA binding properties and whether XPA's proven specificity for DNA lesions (see our table starting on pg. 13 of this rebuttal) can be observed with long DNA substrates.
2. The findings on the stoichiometry of XPA binding to DNA are not new; it is broadly accepted that XPA does not dimerize. There are many studies supporting this, including the structure of truncated Rad14 protein, in which two molecules are bound to the damaged DNA duplex, but there are no direct protein-protein contacts between them.
We appreciate the fact that the Rad14 structure 3 contains no direct protein-protein contacts for the dimer. However, conclusions made in this study report that the protein does bind lesions as a dimer, and this paper has been cited as evidence for dimeric binding by XPA. Additionally, Yang et al 4 are also frequently cited in the recent literature as evidence of an XPA dimer. The new TFIIH structure 2 does show XPA as a monomer in the context of the verification bubble, which supports our conclusion in the context of XPA with other proteins on DNA, but we respectively disagree that it is "broadly accepted that XPA does not dimerize." Below are some other papers that report XPA dimers on DNA: Yang et al, 2006 5 Liu et al, 2005 6 Brown et al, 2010 7 Gilljam et al, 2012 8 Response to Reviewers 3. The experimental design is lacking one important control: a DNA substrate with a static bend but no DNA damage (introduced for example by the uneven lengths of strands). This would inform the role of XPA in DNA damage recognition. On pg 11 (line 268) the heading says Presence of UV damage slows the rate of mode switching, leading to longer phases. It is difficult to assess if the differences observed are caused truly by the helix-distorting DNA damage or by the DNA distortion on its own attracting XPA to the substrate. While the authors discuss this possibility on pg 15, line 350, and it is consistent with lack of binding to a static bent DNA reported previously by Yang and colleagues , an appropriate control is required. This is especially important in the context of NER, as XPA may preferentially bind DNA that has already been bent by other components of NER complexes.
Several binding studies on these types of substrates have been published and are cited in our manuscript (including papers by Missura et al 9 and Yang et al 5 ). These and other papers are further detailed in the table at the end of this document (pg. 13). We believe these studies do provide compelling support for the idea that XPA binds preferentially to bent DNA. However, we argue that these artificial substrates do not more closely resemble NER intermediates than they resemble NER substrates and, as such, do not necessarily resolve the debate of XPA's role in damage recognition/verification. Instead, they support the idea that XPA is able to recognize bent/flexible DNA, in agreement with our present results, and this propensity to bend DNA likely plays a role in both the search for DNA damage, as well as affinity for NER intermediates. We therefore feel strongly that additional experiments on bent DNA substrates are not warranted.
4. Pg 5, line 119 -Of all the complexes observed, 33% were bound to the ends of the DNA substrate. This is quite a high rate of non-specific binding. Even 22% for the AAF538 substrate seems to be quite a high ratio of non-specific to specific binding. Why? Some comment on this is required in the text. The occurance of cases with two molecules bound at both ends also observed also need to be reported.
Many proteins bind to the ends of DNA molecules (we have added citations) and we do not believe this is non-specific binding, but that XPA does, in fact, recognize specific types of DNA structures (like unpaired DNA ends) beyond damage site distortion. We have confirmed that we never see any cases in which a 538 bp DNA molecule is bound at both ends simultaneously. We have acknowledged this in the text (lines 114 and 312).
5. Pg 11, line 248 -Results show that most (60-70%) of the XPA protein is in a stationery mode on tightrope substrates. Is the remaining 30-40% fraction of motile protein sufficient to conclude that XPA actively screens for DNA-helix distortion? Does this ratio change when more than one damage site per DNA substrate is introduced? How do the authors explain that the motile fraction of the protein changes modes, including pausing, but the stationary fraction does not?
Due to the trade-off between observing and recording multiple binding events over the course of an experiment versus watching one particle for greater than 300 seconds, it is possible that some of these non-motile molecules would have diffused if we had observed for longer periods of time. It is also important to point out that the addition of increased salt helps to mobilize some of these non-motile molecules, so we do not believe these are necessarily dead end complexes. Furthermore, we have performed tightrope experiments on UV lambda DNA with an increased 6 Response to Reviewers UV dose (80 J/m 2 ), resulting in increased lesion density (~1 lesion every 550 bp). The experiments show that XPA pauses more frequently on this substrate. 6. Is there any speculation on what the pausing positions on the tightropes are?
We speculate in the manuscript (lines 237-240) about possible spontaneous damage or bent DNA sequences in the non-damaged substrate that may cause XPA to pause. To address this question further, we have performed tightrope experiments on λ DNA with an increased UV dose (80 J/m 2 ). We see a consistent trend of increased time spent in the paused mode and an increased number of pause sites when comparing XPA behavior on non-damaged λ DNA to its behavior on the λ DNA treated with either 20 J/m 2 or 80 J/m 2 UV, suggesting that these pause sites in the UV-irradiated DNA are indeed due to interrogation of UV-induced photoproducts, (Figure 5d-e).
7. Pg 16, lines 391-399 -The authors speculate without any specific support that the disordered N-and C-termini may fold in and interrogate the DNA for lesions via DNA bending. Considering that XPA is known to function as the key scaffold for NER pre-incision complexes and the many studies showing that it interacts with many other NER proteins, it is illogical to speculate that the disordered N-and C-termini of XPA are important for DNA binding only. This speculation should be supported by data or completely deleted from the text.
We greatly appreciate this suggestion by the reviewer, and show in Figure 5, panels d and e, that we have now created a truncation variant of XPA (M98 through T239) which removed residues 1-97 on the N-terminus and 240-273 on the C-terminus. In our EMSA studies, this variant displays identical non-specific DNA binding affinity as full length XPA and a 1.6 fold decreased specificity for AAF adducts (see new Supplementary Figure 1). Single molecule analysis revealed that this truncated XPA shows few pausing events on highly damaged λ DNA (80 J/m 2 UV-C), behaving like full length XPA protein on non-damaged DNA.
Additional points. 1. Pg 3, line 68 -A zinc finger is a specific structural motif that is characterized by the coordination of one or more zinc ions in order to stabilize the fold; the term should be reserved for such motifs. XPA possesses a Zn-binding motif, but not a Zinc finger and the wording should be adjusted accordingly.
While this is a relatively minor point with regard to the data presented in our study, we respectively disagree with the reviewer's distinction. XPA does in fact contain a C4 type zinc finger motif. We understand that C4 is not a standard/common zinc finger motif. UniProt classifies C4 zinc fingers as a family without sequence similarity beyond a zinc coordinated by 4 cysteines, but as a zinc finger nonetheless (https://www.uniprot.org/help/zn_fing). The term "zinc finger" has been used to describe this feature of XPA in several published papers, including the structure of Rad14 in complex with damaged DNA (Koch et al PNAS 2015). Thus, we did not change this nomenclature. Citation has been updated. 7 Response to Reviewers 3. Pg 4, lines 99-100 -How do the KD values compare to previously published data?
In line 96, we compare our results with previous KD's reported by Sancar and colleagues. We have assembled a table, beginning on pg. 13 of this document, of papers showing that XPA has affinity for specific DNA substrates and give the KD's when reported in the paper. However, it is not possible to directly compare our values to others, because substrates, protein preparations, buffer conditions, etc. differ. 4. Pg 6, line 150 -The expected molecular mass of XPA monomer is 32.6 kDa but on pg 3, line 66 XPA is described as a protein of molecular mass of 31.4 kDa. While the origin of this inconsistency (XPA-His vs XPA) is mentioned at the Figure 2 label, it should be also clarified in the main text.
We appreciate the reviewer finding this discrepancy. The manuscript has been updated to clarify size discrepancies of purified XPA. We now explicitly state the construct used for all experiments and its molecular weight for all stoichiometry experiments (His-flXPA is 32.6 kDa and His-flXPA-StrepII is 33.9 kDa).

Pg 7, line 164 -Wording should be changed for Because XPA is a small protein (…). For example, "XPA is a relatively small protein for precise AFM measurement (…)"
We appreciate the reviewer's suggestion, but line 164 (line 155 in the updated version) is not related to the precision of the AFM measurement, which is actually discussed in a previous section (lines 106-110). Instead, we are explaining that because the protein is small, the ratio of DNA/protein in the complex will be large enough to necessitate a method to subtract the DNA volume.
6. Pg 23, line 601 -XPA is known to bind a zinc ion, and the presence of Zn has an influence of the structure. Hence, use of EDTA in the buffer may affect the protein's fold, stability, and solubility. Were controls performed?
EDTA was only present in the tightrope buffer, which due to the light intensity and Qdots, can cause oxidation (not for EMSA or AFM). EDTA is routinely used in molecular biology assays to bind heavy metals and to avoid oxidation reactions, particularly at Cys residues, and has been

Reviewer #3 (Remarks to the Author):
The manuscript by Beckwitt et al. presents single molecule atomic force microscopy (AFM) and fluorescence studies of the nucleotide excision repair (NER) factor Xeroderma pigmentosum group A (XPA). The authors aim at characterizing conformational and dynamic properties of XPA-DNA complexes during lesion search and at a target lesion to understand the mechanism of lesion recognition by XPA.
XPA is an essential protein and mutations in XPA severely affect human health. XPA functions in the NER pathway that repairs of a plethora of structurally and chemically diverse DNA lesions, including acetylaminofluorene (AAF), which is used in the study, and in particular and famously UV induced damages in the DNA. The exact role of XPA in the NER mechanism is still controversial. In addition, the full length protein in complex with DNA has not been accessible to X-ray crystallography due to the intrinsic disorder in parts of the protein. The questions addressed by the authors are therefore highly relevant and interesting. In their studies, the authors investigate XPA in isolation (without other NER components present), while in vivo XPA functions in NER in complex with the transcription factor IIH (TFIIH) helicase machinery. However, understanding the conformational dynamics of DNA interactions by the isolated XPA protein adds to our better appreciation of the role of XPA in NER lesion processing.
The authors strive to resolve the question of the oligomeric state of XPA, the DNA bending introduced by XPA in non-damaged DNA and at an AAF lesion, and the mode(s) of translocation on DNA in the absence and presence of lesions. Regarding the oligomeric state of XPA, contradictory findings had been reported in the past. Beckwitt and colleagues clearly establish the monomeric form of XPA to be the dominant species both in solution and when bound to DNA, using AFM and multi angle light scattering. AFM characterization of DNA bending by XPA revealed comparable bending at non-damaged and lesion sites, and the bending angle was similar to that observed in crystal structures (of the shortened construct). The authors argue with energetic cost of DNA bending as a means for XPA to distinguish between lesion and non-damaged DNA sites based on their different flexibilities, as has been proposed before by others for XPA and for other DNA repair proteins. Regarding XPA lesion search dynamics, the authors report several different modes of translocation on DNA and the frequent interchange between these different modes. They interpret the modes as XPA testing for damage (pausing), XPA carrying out one-dimensional diffusion on the DNA in its lesion search (short range motion), and XPA in a fast scanning mode (mid range or long range motion). In a model they summarize their conclusions that XPA lesion search involves three different states: XPA searches for lesions in a fast scanning complex with diffusion coefficient dependent on ionic strength of the buffer (which they interpret as an indication for hopping); XPA probes for lesions in a short range scanning complex with lower diffusion constants, which may be in part caused by DNA bending by the protein to probe for lesions; and XPA forms a strongly bent stable recognition complex at the lesion.
Overall, I find the topic and results of this study interesting and highly relevant and I mostly agree with their conclusions. I do however, have a few points of criticism listed below. After minor changes, I hence recommend the manuscript for publication in Nature Communications. Response to Reviewers Specific points: 1. AAF binding specificity determination by AFM: The distribution of XPA positions on the DNA does not really resemble a > 600-fold preference for lesion binding (when compared, for example, to distributions in Yang et al. NAR 2005 as cited by the authors in this context). I agree with the authors that there is specificity for AAF by XPA, but I think the value is overestimated due to underestimation of non-specific background binding. Differently expressed: I think the Gaussian width in the fit to the distribution is too broad, artificially increasing the specific area that results in the high specificity value. A smaller binning interval might more clearly show whether the maximum at the lesion position is truly this broad or whether the Gaussian shaped peak is in fact much narrower, but would likely require more data points. In fact, the specificity that this distribution reflects likely fits nicely with that obtained by the authors from their EMSA analyses. However, based on the high flexibility at an AAF lesion, it actually seems surprising that the specificity of XPA for this lesion should not be larger than observed here, if energetic preference of enhanced flexibility sites determines the probability of occupancy. Furthermore, enhancement of TFIIH lesion recognition through XPA, as suggested previously (and also mentioned by the authors) is difficult to imagine with the low to moderate recognition specificity compared tofor instance, that shown for the XPD helicase of TFIIH (admittedly for different lesions). These points could be addressed.
We appreciate this reviewer raising this point, but we are constrained by the typical binning routine supported in the literature of using the number of bins equal to the square-root of the total number of observations 14 . The relatively broad distribution around the mean is typical for AFM studies of protein-DNA interactions as we and others have observed. We have recently reviewed 15 and we have specifically observed this type of data for Rad4 16 , PARP1 17 , APE1 and DNA polymerase β (unpublished). Thus, we do not feel that XPA is an outlier. The reviewer raises a good point and we strongly believe that XPA interaction with other proteins can help to increase specificity through protein-protein interactions and additional modification or the structure of the DNA. However, this work is outside the scope of this present body of data.

XPA volume analyses by AFM:
While I appreciate the thoroughness of their volume calibration for their peak force AFM, I think this section could mostly be moved to the Supplement. It is obviously indispensable, but does not add to the manuscript story. The authors analyzed the volumes of two proteins, APE1 and Pol(beta) bound to DNA, to test the validity of their volume calibration also for DNA bound protein volumes. In this context, I also think that the originally measured volumes for APE1 and Pol(beta) (before DNA volume subtraction) ought to be given in the text, in addition to the final volumes after DNA subtraction. After the thorough consideration of the contribution from DNA binding mode by APE1 and Pol(beta) to overall measured volume, it would recommend itself to also discuss this for the protein of interest in this study, XPA to support the approach of subtracting the DNA volume to determine the molecular mass of the DNA bound complex. Again, I think the originally measured volumes (prior to DNA volume subtraction) should also be given, as well as the DNA volumes.
We appreciate this suggestion and we have moved some of these data to the supplemental data section. We agree with showing both the volume prior to and after subtraction, and have added this data to Figure 3 and Supplementary Figure 4. Response to Reviewers 3. DNA bend angle analyses by AFM: In the bend angle distributions for DNA in the absence of protein ( Figures 4C and D), the first interval seems to be only half the size of all other intervals since it is centered at zero degrees bending. If this is the case, then at least in the histogram shown in (C), correction to full interval size for the first range of bend angles (0 to 5 or 10 or whatever degrees) would likely result in a distribution with exponential decay from 0 degrees (rather than the 5.6 degrees maximum reported here). For the distribution in (D), at the AAF position, the first maximum very likely remains at the second bin, ie. larger than 0 degrees. In addition, for a better comparison between the two distributions, ideally the interval (bin) size should be the same, which would then be limited by the sample with the smallest number of datapoints (also for Figures 4 G and  H). Also, for smaller bin sizes, the distributions in Fig. 4 G and H may well reveal more than one bend angle species, as also suggested by the example images (but smaller binning will be difficult due to small samples sizes). Two different conformational states in lesion search and recognition by XPA would seem interesting also in the context of the final model proposed by the authors. However, none of this will drastically change the story.
As suggested by this reviewer, we have replotted all DNA bend angles in Figure 4, and we appreciate the reviewer suggesting this change. The non-damaged DNA bend angle is centered on zero degrees. 4. DNA tightrope assay: a. The authors state that the proportion of immobile XPA is not affected by the DNA substrate. However, Figure 5B shows ~90% versus ~70% immobile XPA complexes on DNA (including dissociating and persistent molecules) for AAF versus UV or non-damaged substrate. The statistical non-significance level seems surprising! The authors state that the density of UV lesions in their substrate can be expected to be comparable to that in the AAF substrate (approximately one lesion per ~2,000 bp), yet the percentage of immobile complexes appears to differ between the UV and AAF substrates. Can the type of lesion explain the different behavior?
With regard to the statistical question, while the proportions of stationary/motile particles appear different between UV and AAF tightropes, the χ 2 test, which is the appropriate statistic for analysis of all groups, shows no significant difference between groups. While the 6-4PP is more helix-distorting than the CPD, we have shown that Rad14 (XPA homolog in yeast) has higher affinity for an AAF than for either photoproduct 3 and this may, in part, explain the behavior observed in the current study. It is possible that an AAF lesion can more readily induce a conformational change in XPA to promote pausing and stationary complexes. While we agree with the reviewer, we have erred on the side of caution when drawing conclusions from these comparisons. It is interesting to note in Figure 5d, where the motile fraction of XPA molecules is investigated in more detail, that lesion density alters the type of motion in a statistically significant way. Please see also point 5 to Reviewer 2.
b. The lack of medium/long range scanning species for AAF versus non-damaged or UV substrate is striking. Long/medium range motion is defined by the authors as movement over > 690 nm, the distance between two AAF lesions and approximate average distance between two UV lesions assuming the above stated density. Exclusively short range scanning by XPA is hence consistent with XPA not proceeding over an AAF lesion. Again, could this depend on the type of lesion, since for the UV substrate a significant population of medium/long range scanning molecules was observed? In contrast to AAF, CPD lesions, as one type of UV lesion, do not strongly distort DNA.
The reviewer is correct to point out the lack of long-distance motion on the AAF arrays. In order to look at this in more detail, we have done an additional experiment with λ tightropes treated with 80 J/m 2 UV-C. While there is a slight decrease in long-range diffusion under these conditions, the most significant response to lesion density and type of lesion is time spent in the paused mode (Fig. 5d). With increasing UV, where the 6-4PPs are at higher density (approximately 1 6-4PP per 2.2 kbp for 80 J/m 2 treated tightropes), we see an increased amount of pausing, comparable to the AAF array (1 AAF per 2 kbp).
c. Many of the sub-species do not seem to contain enough data points for statistical analyses. For example n=3 for long range diffusion coefficients of pooled UV and non-damaged substrates, or n=3 for maximum displacement at 1 M salt concentration. I appreciate that data collection with this type of single molecule experiment is time consuming, but would be careful to draw conclusions from such small sample sizes.
We understand this limitation, and have tried to be cautious in drawing conclusions. Combining mid-range and long-range modes into one category (see point below) has improved the sample sizes in several cases. d. I do not agree with the classification of long range scanning particles based on the example trace shown in Figure 5D. In this example, the "long range" trace seems to consist simply of two medium range movements interrupted by a clear pausing plateau. I would suggest to pool the data from medium range and long range modes. In the population distributions (e.g. Figure 6F) the long range species never seems to truly add any additional information. Also, the issue of not enough data points for some of the different species (point above) may automatically be removed when pooling medium and long range scanning particles.
We appreciate this comment and have now combined mid/long-range diffusion into one mode (called long-range).
e. I also wonder if the salt concentration effect on the medium/long range but not on the short range moving species really supports hopping as the mode of translocation in the medium/long range species and one dimensional diffusion for the short range species. Regarding the short range species, the distribution of short range diffusion coefficients ( Figure 6E) appears to indicate one species consistent with one dimensional diffusion (below the limit line for diffusion) and a second with higher diffusion constants (above the limit line for diffusion). Regarding the medium/long range species, it seems to me that hopping is only one possible explanation of the effect observed at higher salt concentration. Could the salt concentration not instead simply affect stability of DNA interactions by the protein? These interactions seem to be majorly electrostatic when looking at published structural data. The example in Supplementary Figure  5c showing transition from paused to long range diffusion when increasing the salt concentration to 1M would be consistent with this. 12

Response to Reviewers
The reviewer is correct in that increased ionic strength will affect electrostatic interactions. The reason that increasing ionic strength increases rates of diffusion during hopping is that, hops during micro-dissociation of XPA from the DNA get longer due to shielding of the electrostatic potential of the negatively charged phosphates. It is generally accepted in the single molecule field that if increasing ionic strength increases diffusion rates, then the protein is displaying hopping 18,19 . f. I wonder if DNA bending as seen in AFM applies for the tightrope assay or if the DNA is "fully" stretched between the anchoring beads? If XPA would not be able to bend the DNA as part of its lesion probing, this would affect the time spent probing for lesions (ie. bending the DNA, ie. in the paused state as well as in the short range linear diffusion state). The authors mention that other proteins previously tested in this assay did not show the changing between translocation modes seen for XPA. Were any of these proteins expected to bend the DNA?
We have clarified in the text that DNA is not fully stretched in the DNA tightropes, but only elongated to 90% of its contour length. We and others have used this flow cell set-up successfully to study other DNA-bending proteins, including Rad4-Rad23 16 . Table 1. Published XPA-DNA Interactions demonstrating DNA damage and/or structure-specific specificity. Unless otherwise noted: studies were performed using full-length wild-type human protein (recombinant, purified), fold specificities are reported as overall binding between substrates, and DNA substrates were prepared using short oligonucleotides (less than 60 bp or nt).
Filter binding b,c UV-irradiated dsDNA (0 -9 kJ/m 2 ) Similarly, in interpreting their data in the Discussion section, the authors need to acknowledge that the prevailing belief in the field is that TFIIH opens the damaged duplex and unwinds it around the site of damage, then XPA is recruited along with RPA, which means XPA is not freely diffusing in solution searching for damage in a duplex. Stating this, and then providing the alternate view that XPA may participate in the search for regions of the DNA containing lesions, strengthens their argument, by first providing a balanced perspective for discussing the relevance of the results, which would then set the stage for presenting the model in Figure 7. As noted previously, the authors could also add further support for the relevance of this study by noting that there is increasing evidence of XPA functions outside of NER for which the DNA binding properties characterized here could well be critical.
Another very critical concern is that there are a number of phrases in the manuscript that are imprecise and must be clarified if the manuscript is to be deemed suitable for publication. Line 257-given the differences in the data, the word "sharply" should be deleted.

Page 9
Line 266-267 This sentence is highly speculative with no evidence from this study to support it; it should be deleted.
Line 325-… DNA helix(13,25) RELATIVE TO UNDAMAGED DNA, and … Line 328-There are no data supporting protein folding, so 'fold into' should be replaced by form.

Page 11
Line 354-the inclusion of the term "initial damage recognition" assumes XPA is involved in this step, but as noted above, this issue is controversial not proven. Since the authors provide no direct evidence that XPA is involved in initial damage recognition, this sentence needs to be rephrased.

Page 12
Line 381-… specificity for DNA CONTAINING an AAF adduct … and … for DNA CONTAINING a CPD lesion … Line 383-replace "motion" with search.
Reviewer #3 (Remarks to the Author): The 3) In Figure 4, I was surprised that the authors suggest that the 10.5° ± 7° bend angle state in (d) for AAF-DNA is comparable to the 0° state in (c) for non-damaged DNA. These data look like they would easily pass any significance test. This is surprising since these are DNA substrates of identical sequence (apart from the presence or absence of the AAF lesion). The authors also specifically state in the text that they did measure DNA bending also at the 30% positions for the non-damaged DNA substrate (rather than at random positions along the DNA), which eliminates the possibility of sequence effects. Assuming that an experimenter bias towards larger bending at AAF-lesions can be excluded, I think these significantly different DNA bend angles are difficult to explain, but may not want to be completely ignored. Again, as in my previous comments, this does not change the story of the manuscript, which rather focuses on the larger bend angle of approximately 35° at the AAF lesion in the absence of XPA (which is clearly not present in the nondamaged DNA) compared to 60° in the XPA-DNA complex. Response to reviewers that the prevailing belief in the field is that TFIIH opens the damaged duplex and unwinds it around the site of damage, then XPA is recruited along with RPA, which means XPA is not freely diffusing in solution searching for damage in a duplex. Stating this, and then providing the alternate view that XPA may participate in the search for regions of the DNA containing lesions, strengthens their argument, by first providing a balanced perspective for discussing the relevance of the results, which would then set the stage for presenting the model in Figure 7. As noted previously, the authors could also add further support for the relevance of this study by noting that there is increasing evidence of XPA functions outside of NER for which the DNA binding properties characterized here could well be critical.
We appreciate this concern and have added this point to the discussion (lines 342-345). We have intentionally presented Figure 7 as a working model of XPA behavior on DNA, based on what we have learned in this study. We have purposely not placed our model in the context of the greater NER pathway. The main purpose of the figure is to show (1) the different modes that XPA exhibits on DNA and (2) the possible role of the disordered domains in its motion.
Another very critical concern is that there are a number of phrases in the manuscript that are imprecise and must be clarified if the manuscript is to be deemed suitable for publication.
1. It is important to state at the outset and incorporate terminology throughout the text that clarifies XPA recognizes aberration in the DNA duplex and not the lesion itself. Adding the phrase 'DNA containing' (e.g. 'XPA recognizes DNA containing damage') is a simple way to address this issue. This is central to how NER can repair such a diverse range of lesions.
Additional clarification has been added to the manuscript in response to the specific issues listed below (point 3). We wish to clarify our terms here, as we agree with the reviewer at a fundamental level, but respectfully disagree on the semantics. With regard to this semantic difference between seeing the actual modified base or the altered structure induced by the modified base, we believe it is unfair to distinguish a helical distortion caused by a lesion from the lesion itself. For example, XPA recognition of helical aberrations may be what allows the protein to recognize and bind at an AAF adduct, as we have shown by AFM. We believe XPC and even UV-DDB probably work in a similar manner.
2. The issue of 'specificity' in binding DNA also needs greater clarification. A simple way to remedy this problem would be to include the phrase 'relative to undamaged DNA'. The table provided in the response letter nicely shows how difficult it is to draw conclusions from data in the literature. For example, Ref 16 reports that XPA ss-ds DNA junctions with much higher affinity than DNA containing lesions and the Kd value reported in that study is two-fold lower than the Kd value reported here for AAF containing DNA. But the methods used are very different. Addressing the specificity for damaged versus ss-ds junctions would be highly informative, especially in the context of defining the role of XPA in NER. The authors focused only on the binding of undamaged versus damaged DNA, so this should be made evident in presenting data and discussing the results.
We agree with this point and have made the suggested changes. While we are limited by space, we do state that XPA has specificity for ss-ds DNA junction in the introduction (lines 53-54). As the reviewer mentions, due to variation between methods and experimental conditions between papers, it is very difficult to draw conclusions and determine fold-differences by comparing separate studies. Ref 16 (Yang et al, 2006 1 ) reports that XPA binds to a forked Y-shaped DNA substrate with higher affinity than to dsDNA with a G[8,5-Me]T intrastrand crosslink.