Recruitment and positioning determine the specific role of the XPF‐ERCC1 endonuclease in interstrand crosslink repair

Abstract XPF‐ERCC1 is a structure‐specific endonuclease pivotal for several DNA repair pathways and, when mutated, can cause multiple diseases. Although the disease‐specific mutations are thought to affect different DNA repair pathways, the molecular basis for this is unknown. Here we examine the function of XPF‐ERCC1 in DNA interstrand crosslink (ICL) repair. We used Xenopus egg extracts to measure both ICL and nucleotide excision repair, and we identified mutations that are specifically defective in ICL repair. One of these separation‐of‐function mutations resides in the helicase‐like domain of XPF and disrupts binding to SLX4 and recruitment to the ICL. A small deletion in the same domain supports recruitment of XPF to the ICL, but inhibited the unhooking incisions most likely by disrupting a second, transient interaction with SLX4. Finally, mutation of residues in the nuclease domain did not affect localization of XPF‐ERCC1 to the ICL but did prevent incisions on the ICL substrate. Our data support a model in which the ICL repair‐specific function of XPF‐ERCC1 is dependent on recruitment, positioning and substrate recognition.

Thank you again for submitting your manuscript on the characterization of XPF-ERCC1 separationof-function mutants to The EMBO Journal. I would like to apologize for the delay in its evaluation, which was due to limited referee availability and the need for extended review times during the summer vacation season. We have now received the complete set of referee reports on your study, which I am enclosing below for your information.
As you will see, all referees acknowledge the technical quality and potential interest of the presented data. However, referee 2 raise some concerns regarding the overall advance conveyed by your new results, and referee 4 criticizes aspects of analysis and presentation in both figures and text.
Faced with these mixed recommendations, I would like to give you an opportunity to address the referees' concerns via a revised manuscript. In this respect, I realize that novelty concerns may be in parts already alleviated by a crisper and more concise presentation and focus on the particularly novel insights and implications; however, I feel it will also be important to extend the mutant characterization to obtain some deeper understanding into the molecular basis of ICL/NER functional separation of some of the mutants, as requested by referee 2 (and to some extent also by the other two reviewers).
Should you be able to adequately improve these two key aspects, we would be happy to consider this study further for publication. However, please remember that it is our policy to allow only a single round of major revision, making it important to carefully respond to all points raised during this round. As usual, any related/competing work published during the revision period will have no negative impact on our final assessment of your revised study. Further information regarding The data in this paper are interesting and of a high technical quality. The main concern in terms of suitability specifically for EMBO is that the conclusions are not particularly surprising or novel. For example, it has already been demonstrated that the ICL repair role of Xpf can be separated genetically from the NER role. This isn't a novel finding. Also, cells deficient in Xpf or Slx4 show ICL repair defects; Xpf-deficient cells are NER-defective but Slx4-deficient cells are not. So mutations in Xpf that prevent binding to Slx4 would be expected to inhibit ICL repair without affecting NER. This is not surprising or unanticipated. Similarly, the Knipscheer lab has already shown the role of Xpf in ICL repair is at the unhooking stage, so it's not surprising that the ICL repair-deficient mutants in Xpf are deficient in unhooking. That's the step Xpf catalyzes. What would be novel would be some insight into why it is that the ICL repair-defective mutants do not affect NER. Getting some handle on the difference between Xpf in the two different contexts would represent an important advance. But this paper doesn't go there, and so in its current form its suitable for a more specialist journal (JBC?).
One other point: two mutants which do not bind to Slx4 are used -G314E and deltaNGWS. The former is proficient for ICL repair while the latter is not. The explanation given is that G314E doesn't disrupt the Slx4 interaction sufficiently to inhibit ICL repair. But there's no data on the difference between the two mutants in terms of Slx4 binding -they may be equally defective. If so, it's possible that the NGWS affects an aspect of Xpf function other than Slx4 interaction which explains the ICL repair defect. Its true that the L219R mutation inhibits Slx4 interaction and ICL repair but its simply correlation. Having an Xpf mutant that doesn't interact with Slx4 and that doesn't affect ICL repair argues the interaction with Slx4 is not important for ICL repair and this would need to be addressed.
Referee #3: In a previous paper in Mol Cell in 2014, the authors showed that ERCC1-XPF is responsible for the unhooking step in interstrand crosslink (ICL) repair and that this activity depends on an interaction with SLX4. In the present paper, they analyze how a number of mutant alleles of XPF, isolated from FA patients and in screens in drosophila affect ICL repair and nucleotide excision repair (NER -XPF was first described as an NER protein).
The authors use their xenopus oocyte system to monitor replication-dependent ICL repair and expressed and purified six mutant XPF proteins. Consistent with the patient and cellular phenotypes of these proteins, they are proficient in NER, while displaying a defect in ICL repair. Importantly, the authors can pinpoint the ICL repair defect to the unhooking step, answering a long-standing question in the field, whether the key role of XPF in ICL repair is in the unhooking or a later step in recombination. Interestingly, with the exception of the L219P mutant all of the proteins are recruited to sites of ICL repair and interact with SLX4, showing that their defect in ICL repair results from a defect in positioning at the incision site, either through a defect in the interaction with DNA around the nuclease active site (R670S, S767F), or a defect due positioning in the ICL repair complex likely by protein-protein interaction (C225R, ∆NSGW). By contrast L219P does not bind SLX4 and is not recruited to ICL repair sites.
This work is carried out to a high technical standard and the great care that was taken to ensure that none of the defects observed was due to protein aggregation, a common problem with XPF mutant proteins, is particularly impressive. The work significantly adds to our understanding of the mechanisms by which XPF contributes to ICL repair. 5) P14. line 10 -An alternative explanation for why XPF-∆NSGW is defective in ICL repair is that it may well have an interaction defect with SLX4, but that this defect does not affect recruitment to sites of ICLs, but rather the position of the two proteins at the sites of ICL incision site. It is entirely possible that there are multiple sites of interaction between SLX4 and XPF. This point should be considered here, on p.15 (line 10) and in the discussion on p.16/17.

Referee #4:
In this paper Knipscheer explores the role of the excision nuclease Xpf-Ercc1 in replication coupled DNA crosslink repair. Building on her own pioneering work with J Walter where together they established how a DNA crosslink is repaired during DNA replication using a Xenopus egg extract cell free system. In later follow up work Knipscheer established how the key excision or unhooking step occurs through the interplay between the FA pathway and Slx4 with Xpf-Ercc1 (Molecular Cell 2014). Now in this paper they carry out a more detailed analysis of the nuclease complex with Slx4, more specifically they exploit single mutations in the Xpf/Ercc1 complex that co-segregate with a human syndrome with overlap with FA and Cockaynes syndrome. These mutations cause DNA crosslinker sensitivity but appear not to impact on the role of this complex in nucleotide excision repair.
The main conclusions of this current study is that mutants of Xpf-Ercc1 complex can de divided into two groups -those that impact both NER and CX repair (nuclease active site mutants), those that work in NER but do not unhook DNA crosslinks. This latter group falls into two further groupsthose that bind Slx4 and are then not recruited to the crosslink, and those that bind Slx4 are recruited to the Cx but fail to unhook it. They correctly conclude from this work that Slx4/ Xpf-Ercc1 complex not only works to recruit the nuclease to the damage site but also must somehow position the nuclease once it is there so that it cuts this substrate.
These are novel and important insights into DNA CX repair and therefore merits very strong consideration for publication in EMBO. However despite my obvious enthusiasm for the work I would like the authors to address my following concerns.
1. Generally the manuscript is written rather poorly, the introduction is way too long and utterly boring to read. After several expresso cups I did manage to wade through this tome akin to reading War and Peace in one setting for a romantically disinclined individual. It should be 1/3 the size and can be much more succinct. 2. Generally the figures are just appallingly laid out. I use the test that one should be able to "read" the figures without having to look at the text. I found this impossible in this case. The figures in her and Walters papers are really excellent -she should consider emulating these or at the very least matching their quality and clarity.
3. The last figure should have a model that sums up what they are saying in the paper. Without such a visual encapsulation of the message its impact will be lost to many. 4. Fig 1 is excellent then all this goes rather downhill. Fig 2 WB should have a loading control. Perhaps space should be devoted to the Walter Knipsheer model and how this is assayed in their graphs. 5. Figure 3 : Awful ! Have they shown that depleting other components of NER abrogates UDS in this assay ? 2B why is the UDS analysis only provided for one of the mutants and not all of them shown in the panel below. 2C there is no stats here and I am concerned about the range of intact excision repair, for instance some the mutants are even better than the wild type ( the last one for instance). I would also like to see this data corroborated by showing by slot blot that CPD dimers are removed ( there are good antibodies against these lesions that track their removal. Getting this figure right matters since the rest of the paper builds on the mutants that work in NER but not in CX repair. 6. Figure 4 again difficult to follow-work in improving clarity here . Excessively cropped WB are not really acceptable these days, also again no loading control. 7. Figure 6 no loading control. We thank the reviewers for their feedback on our work. Our response to their comments is shown in blue italics.

Referee #2:
Knipscheer and colleagues present an interesting study, which reports isolation of separation of function Xpf mutations. These mutations affect repair of ICL repair without affecting repair of UV damage. The mutations were chosen for analysis were from taken reports in the literature of Xpf mutations in different disease syndromes. Two were designed based on reports of mutations that disrupt interaction with Slx4. The equivalent mutations were made in Xenopus Xpf and introduced into Xe extracts depleted of Xpf-Ercc1-Slx4 and rescue experiments to analyse DNA repair were carried out. The bottom line is isolation of three mutations that are deficient in the unhooking step of ICL repair but which don't affect ICL repair. One of the mutations blocks XPF recruitment by blocking the binding of XPF to SLX4.
The data in this paper are interesting and of a high technical quality. The main concern in terms of suitability specifically for EMBO is that the conclusions are not particularly surprising or novel. For example, it has already been demonstrated that the ICL repair role of Xpf can be separated genetically from the NER role. This isn't a novel finding. What would be novel would be some insight into why it is that the ICL repair-defective mutants do not affect NER.

In this study we have chosen to study why these mutants affect ICL repair, not why they do not affect NER. We felt this was the more interesting question to answer.
Getting some handle on the difference between Xpf in the two different contexts would represent an important advance. But this paper doesn't go there, and so in its current form its suitable for a more specialist journal (JBC?).
One other point: two mutants which do not bind to Slx4 are used -G314E and deltaNGWS. The former is proficient for ICL repair while the latter is not. The explanation given is that G314E doesn't disrupt the Slx4 interaction sufficiently to inhibit ICL repair. But there's no data on the difference between the two mutants in terms of Slx4 binding -they may be equally defective.

Using a pull down experiment after overexpression of SLX4 with XPF mutants in Sf9 insect cells we show that the XPF ΔNSGW mutant still interacts normally with SLX4. We have now added data to figure 6B showing that this is also the case for the XPF G314E mutant. However, based on previous reports (Andersen et al. Mol Cell 2009, Guervilly et al. Mol Cell 2015) and our finding that the XPF ΔNSGW mutant is defective in ICL unhooking and repair, there is likely an important interaction between this region of XPF and SLX4. Mutating this interaction site does not prevent XPF -SLX4 interaction because we show there is a second, high affinity, interaction site involving the known SLX4 MLR domain and the leucine 219 of XPF.
If so, it's possible that the NGWS affects an aspect of Xpf function other than Slx4 interaction which explains the ICL repair defect.
This is a possibility we can not completely discard. However, since the XPF ΔNSGW mutant complex has a mutation in the same residue that was previously shown to be required for the interaction with the BTB domain of SLX4 this is the most likely explanation. We have now addressed the possibility that it could also affect another aspect of XPF function in the discussion on page 16.
Its true that the L219R mutation inhibits Slx4 interaction and ICL repair but its simply correlation.
Having an Xpf mutant that doesn't interact with Slx4 and that doesn't affect ICL repair argues the interaction with Slx4 is not important for ICL repair and this would need to be addressed. figure 6A. Therefore, it is highly unlikely that a mutant exists that does not interact with SLX4, but is competent in ICL repair.

We do not describe a mutant that does not interact with SLX4 and does not affect ICL repair. The only mutant that shows reduced SLX4 interaction is the L219R mutant and that mutant is defective in ICL repair. SLX4 is absolutely required for the recruitment of XPF to the ICL, this is shown in
Referee #3: In a previous paper in Mol Cell in 2014, the authors showed that ERCC1-XPF is responsible for the unhooking step in interstrand crosslink (ICL) repair and that this activity depends on an interaction with SLX4. In the present paper, they analyze how a number of mutant alleles of XPF, isolated from FA patients and in screens in drosophila affect ICL repair and nucleotide excision repair (NER -XPF was first described as an NER protein).
The authors use their xenopus oocyte system to monitor replication-dependent ICL repair and expressed and purified six mutant XPF proteins. Consistent with the patient and cellular phenotypes of these proteins, they are proficient in NER, while displaying a defect in ICL repair. Importantly, the authors can pinpoint the ICL repair defect to the unhooking step, answering a long-standing question in the field, whether the key role of XPF in ICL repair is in the unhooking or a later step in recombination. Interestingly, with the exception of the L219P mutant all of the proteins are recruited to sites of ICL repair and interact with SLX4, showing that their defect in ICL repair results from a defect in positioning at the incision site, either through a defect in the interaction with DNA around the nuclease active site (R670S, S767F), or a defect due positioning in the ICL repair complex likely by protein-protein interaction (C225R, ∆NSGW). By contrast L219P does not bind SLX4 and is not recruited to ICL repair sites.
This work is carried out to a high technical standard and the great care that was taken to ensure that none of the defects observed was due to protein aggregation, a common problem with XPF mutant proteins, is particularly impressive. The work significantly adds to our understanding of the mechanisms by which XPF contributes to ICL repair.
A few minor concerns should be addressed prior to publication: We thank the referee for the suggestion and have added the reference to page 11. 5) P14. line 10 -An alternative explanation for why XPF-∆NSGW is defective in ICL repair is that it may well have an interaction defect with SLX4, but that this defect does not affect recruitment to sites of ICLs, but rather the position of the two proteins at the sites of ICL incision site. It is entirely possible that there are multiple sites of interaction between SLX4 and XPF. This point should be considered here, on p.15 (line 10) and in the discussion on p. 16 Referee #4: In this paper Knipscheer explores the role of the excision nuclease Xpf-Ercc1 in replication coupled DNA crosslink repair. Building on her own pioneering work with J Walter where together they established how a DNA crosslink is repaired during DNA replication using a Xenopus egg extract cell free system. In later follow up work Knipscheer established how the key excision or unhooking step occurs through the interplay between the FA pathway and Slx4 with Xpf-Ercc1 (Molecular Cell 2014). Now in this paper they carry out a more detailed analysis of the nuclease complex with Slx4, more specifically they exploit single mutations in the Xpf/Ercc1 complex that co-segregate with a human syndrome with overlap with FA and Cockaynes syndrome. These mutations cause DNA crosslinker sensitivity but appear not to impact on the role of this complex in nucleotide excision repair.
The main conclusions of this current study is that mutants of Xpf-Ercc1 complex can de divided into two groups -those that impact both NER and CX repair (nuclease active site mutants), those that work in NER but do not unhook DNA crosslinks. This latter group falls into two further groupsthose that bind Slx4 and are then not recruited to the crosslink, and those that bind Slx4 are recruited to the Cx but fail to unhook it. They correctly conclude from this work that Slx4/ Xpf-Ercc1 complex not only works to recruit the nuclease to the damage site but also must somehow position the nuclease once it is there so that it cuts this substrate.
These are novel and important insights into DNA CX repair and therefore merits very strong consideration for publication in EMBO. However despite my obvious enthusiasm for the work I would like the authors to address my following concerns.
1. Generally the manuscript is written rather poorly, the introduction is way too long and utterly boring to read. After several expresso cups I did manage to wade through this tome akin to reading War and Peace in one setting for a romantically disinclined individual. It should be 1/3 the size and can be much more succinct.
We regret to hear the referee required a high dose of caffeine to get through our introduction. We have now drastically shortened it and hope he/she can now get through on a cup of tea.
2. Generally the figures are just appallingly laid out. I use the test that one should be able to "read" the figures without having to look at the text. I found this impossible in this case. The figures in her and Walters papers are really excellent -she should consider emulating these or at the very least matching their quality and clarity.
We have made many changes to improve the layout of the figures and added better cartoons for clarification.
3. The last figure should have a model that sums up what they are saying in the paper. Without such a visual encapsulation of the message its impact will be lost to many.
This is an excellent suggestion and we have now added a model to Fig 7. 4. Fig 1 is  All western blots on figure 2 contain samples taken straight from extract without any manipulations. We have never added loading controls to these blots in all previous papers (including the ones from Johannes Walter's lab). In these particular cases a loading control would not give extra information because we are merely showing the depletion (that we also confirm functionally) and the level of protein we add back.
The Walter/Knipscheer ICL repair model is now added to Figure EV1 and can be easily pulled up for clarity. In addition, we have clarified the cartoons explaining the assays in the main figures.
5. Figure 3 : Awful ! Have they shown that depleting other components of NER abrogates UDS in this assay ?
We thank the reviewer for this good suggestion. We have now depleted PCNA and XPA and show that this inhibits UDS in this assay. This data is added to Fig EV4. 2B why is the UDS analysis only provided for one of the mutants and not all of them shown in the panel below.
We have now added the data for all mutants to Fig 3. 2C there is no stats here and I am concerned about the range of intact excision repair, for instance some the mutants are even better than the wild type ( the last one for instance).
We have statistically analyzed our data and show that the mutants are not statistically different from the wild type protein, except for the catalytically inactive XE D668A mutant. This has been added to the manuscript.
I would also like to see this data corroborated by showing by slot blot that CPD dimers are removed ( there are good antibodies against these lesions that track their removal. Getting this figure right matters since the rest of the paper builds on the mutants that work in NER but not in CX repair. We have analyzed our plasmid with the antibody against CPD's. We did not use slot blots because they suffered from high nonspecific background signal but we used the same antibody in an ELISA assay (Cell Biolabs, USA). With a low UV dose we could clearly see that the CPD damage is repaired in Xenopus egg extract (Fig EV4C). For our UDS assay a higher UV dose was required to obtain enough signal to measure UDS unambiguously. At this high dose of UV we could not observe the decrease in CPDs, most likely because only a minor fraction is repaired. This data is added to Fig EV4C. 6. Figure 4 again difficult to follow-work in improving clarity here . Excessively cropped WB are not really acceptable these days, also again no loading control.
We have made changes to this figure to improve clarity and have reduced cropping of the blots.