The concerted roles of FANCM and Rad52 in the protection of common fragile sites

Common fragile sites (CFSs) are prone to chromosomal breakage and are hotspots for chromosomal rearrangements in cancer cells. We uncovered a novel function of Fanconi anemia (FA) protein FANCM in the protection of CFSs that is independent of the FA core complex and the FANCI–FANCD2 complex. FANCM, along with its binding partners FAAP24 and MHF1/2, is recruited to CFS-derived structure-prone AT-rich sequences, where it suppresses DNA double-strand break (DSB) formation and mitotic recombination in a manner dependent on FANCM translocase activity. Interestingly, we also identified an indispensable function of Rad52 in the repair of DSBs at CFS-derived AT-rich sequences, despite its nonessential function in general homologous recombination (HR) in mammalian cells. Suppression of Rad52 expression in combination with FANCM knockout drastically reduces cell and tumor growth, suggesting a synthetic lethality interaction between these two genes, which offers a potential targeted treatment strategy for FANCM-deficient tumors with Rad52 inhibition.

1. The part of Introduction describing the conclusion is too long and overlaps with Discussion. It should be shortened. 2. Figure 1c and 1d are out of the order in citation in the text. This should be corrected. 3. Page 7, 2nd paragraph, the authors should make clear to readers that the DNA binding activity of FAAP24 is required for recruitment of FAAP24 to C FS, whereas the FANC M interacting motif is dispensable for the same recruitment. As written, it is unclear to reader whether the authors are describing recruitment of FAAP24 or FANC M. 4. RAD52 and FANC M are synthetic lethal, suggesting that the two proteins work in two parallele pathways to promote cell proliferation. This synthetic lethal interaction is different from their genetic interactions in C FS protection. The authors should make this clear to readers in the discussion. 5. Page 10, first paragraph, the authors stated that "these studies reveal an important function of RAD52 in mammalian cells in a special context when DNA ends are blocked by DNA secondary structure". I did not see evidence for this statement. Figure 5d shows that the luciferase contruct has the same effect as the Flex report. Do the authors think that the luciferase reporter can also form secondary DNA structures? If this statement is just speculation, please move this part to the Discussion.
Reviewer #3: Remarks to the Author: This is an interesting paper describing the role of FANC M in common fragile site protection and FANC M-Rad52 synthetic lethality. However, there are several concerns which the authors should address.
Major points: 1) Just one shRNA for each gene (FANC A, FANC D2, FAAP24, MHF1, RAD51, RAD52) was used for Figure 1b, 1c, Supplemental Figure 1c, Figure 5a,b,d. The authors should perform an "add-back experiment" by knocking down with shRNA followed by transfecting an expression plasmid to rescue the phenotype or use at least two shRNAs/gene. (For FANC M, the authors have performed the add-back experiments.) 2) Usage of just one cancer cell line (U2OS or HC T116) for most of the experiments is a concern, since whether the conclusion can be generalized is questionable. The authors should use multiple cell lines for some key experiments. 3) Figure 3e : the authors should test the effect of FANC M-MM1 mutant or depletion of other FA proteins (FANC A, FANC D2, etc.) on fragile site breakage. 4) Some statistical tests should be done for any quantitative data analyses.
Minor points: 1) Page 3 Line 21: "the FANC D2 and FANC I heterodimeric complex (ID2), which further recruits 1 Reviewer #1: "Overall the results in this manuscript provide many details of how FANCM functions at the Flex1 reporter. However, where these results can be generalized to various types of endogenous CFSs still needs further investigations." Response: We acknowledge that multiple mechanisms underlie CFS instability, but fork stalling at unusual DNA sequences such as structure-forming AT-rich sequences is believed to be one of the general causes to induce CFS instability in addition to paucity of replication origin, large-sized open reading frames and late replication timing. The following points support the idea that FANCMdependent protection of AT-rich structure-forming DNA sequences at CFSs can be regarded as a general mechanism for the maintenance of CFS stability.
2. In this revision, we cloned another two AT-rich sequences from FRA16C and one from FRA3B, and showed that these sequences also induce mitotic recombination in a manner similar as Flex1 (new data in Supplementary Fig.1), suggesting that these sequences also cause genome instability as Flex1. Thus, induction of genome instability by CFS-derived ATrich structure-forming DNA sequences is not limited to a specific AT-rich sequence Flex1, but likely a general mechanism applicable to other CFSs.
3. In normal cells, these AT-rich structure-forming DNA sequences are protected by the mechanisms such as FANCM-mediated fork reversal as we described, which minimizes the effect of DNA secondary structure-induced genome instability at CFSs. However, in the absence of such protection mechanisms, for example in FANCM deficient cells, more DSBs are generated at these AT-rich sequences, and consequently forming DNA secondary structure becomes a prominent cause to induced genome instability and CFS breakage. We added in new data and showed that depletion of FANCM increases instability not only at Flex1, but also at the AT rich sequences from FRA16C and FRA3B (new data in Fig.1d and data not shown). These data suggest that FANCM has a general role in protecting AT-rich sequences at CFSs. Response: CFS expression (breakage) can be induced by various mechanisms. The HR-Flex reporter is designed to specifically monitor features associated with two important mechanisms ( Figure 1 in this text). (1). One is the protection mechanism at AT-rich structure-forming DNA sequences. If a defect in protection results in DSB formation at Flex1 without influencing HR activity, more green signals would be generated from our reporter ( Figure 1 in this text, right top box). For instance, FANCM deficiency results in more DSB formation but does not significantly influence HR, so we saw more green signals (Fig.1b, 1f and 2a). (2). The other is HR mechanism that is used to repair DSBs at Flex1. If HR is defective, green signals would be reduced even if DSBs are formed at Flex1 ( Figure 1 in this text, right bottom box). For instance, BRCA1 deficiency leads to impaired HR, thereby resulting in reduction of green cell accumulation ( Supplementary Fig. 2a). Thus, our reporter assay suggests that FANCM plays a role in preventing DSB formation at Flex1, while BRCA1 is important for HR repair at Flex1. Both of these functions are important for maintaining CFS stability. However, CFS expression can be caused by multiple mechanisms and those defects related to replication origin firing and transcription regulation would not be detected by our reporter as expected. FANCD2 is important for CFS protection including FRA16D, but the underlying mechanisms involve inhibition of dormant origin firing and prevention of DNA/RNA hybrid formation 8 . These features would not be detected in our HR-Flex reporter. Thus, although FANCD2, FANCA and possible other factors do not have a direct role in protecting Flex1 as revealed by our HR-Flex reporter, they are still important for CFS stability via other mechanisms. On the other hand, although HR-Flex reporter cannot score all defects causing CFS instability, it allows us to perform screening to identify new factors such as FANCM which are critical for maintaining CFS stability and to illustrate the underlying mechanisms.
2. In contrast to FRA16D, the fragility of FRA3B is thought to arise from low origin firing instead of DNA secondary structures and fork stalling (Letessier et al. 2011 Nature). It is very surprising that FANCM knockdown also increased the expression of FRA3B in Fig. 3.
Response: As described above, multiple mechanisms including paucity of replication origin, largesized genes, fork stalling at unusual DNA sequences and late replication timing underlie CFS instability. The contribution of each factor to the fragility of a specific CFS may vary depending on the context of CFSs. For instance, certain CFSs may have fewer replication origins and also fewer structure-forming DNA sequences than others, and for these CFSs, low origin firing may be the predominant cause to induce CFS instability. On the other hand, some other CFSs may contain more AT rich sequences and thus formation of DNA secondary structures to stall DNA replication becomes a more important contributor for inducing instability of these CFSs.
We agree that a major mechanism to induce FRA3B instability under normal condition is low origin firing. However, FRA3B does contain AT-rich sequences prone to forming secondary structures. We cloned one such sequence and showed that this AT-rich sequence also induces mitotic recombination, and FANCM is important for the protection of this sequence (new data in Fig.1d and Supplementary Fig. 1c). Under normal conditions, the AT rich sequences at FRA3B are protected by FANCM-mediated mechanism, while lack of origin may be the predominant mechanism causing FRA3B breakage. However, when FANCM is deficient, these AT rich sequences are not protected leading to significant more DSB accumulation and increased expression of FRA3B.
3. In Fig. 4, how Ras induces recombination at FRA16D and the Flex1 reporter is not explained at all. Oncogene activation could induce replication stress in many different ways, and most of these are very indirect. Without an understanding of how Ras induces CFS expression, it is difficult to appreciate the function of FANCM in this context. Response: We provided new data to show that Ras expression induces replication stress as revealed by ATR checkpoint activation (Chk1 and RPA2 phosphorylation) and DSB formation (γH2AX phosphorylation) (new data in Supplementary Fig. 9). We also performed ChIP analysis and showed that DSBs are accumulated at Flex1 upon Ras overexpression (new data in Fig. 4e). These data suggest that Ras induces replication stress which directly causes DSB formation at Flex1, thereby inducing recombination at Flex1 and causing CFS expression. Fig. 5a and 5b can be interpreted differently. Rad52 may be required for forming secondary structures, fork stalling, or DSB formation at the Flex1 reporter.

The results in
Response: In our model, we propose that when FANCM is deficient, DSBs accumulated at Flex1 would rely on Rad52 to repair. We provide new data to show that when Rad52 is depleted in FANCM deficient cells, DSB formation at Flex1 is increased as revealed by ChIP of γH2AX (new data in Fig. 5d). This suggests that Rad52 is not required for DSB formation, but rather for suppression of DSB formation at Flex1 when FANCM is deficient. Thus, these data support our model that Rad52 is needed for repairing DSBs at Flex1 when FANCM function is impaired. Fig. 5d, the authors proposed that Rad52 is important for HR when DSB ends are blocked by secondary structures. However, in the case of HR-Flex, DSB ends are blocked by both secondary structures and a non-homologous sequence. In the case of HR-Luc, DSB ends are blocked by a non-homologous sequence only. In neither case, DSB ends are blocked by secondary structures only. Therefore, the authors' hypothesis is not supported by experimental evidence.

Based on the results in
Response: We performed new experiments by generating new reporters (HR-Flex/D-Flex and HR-Luc/D-Luc, Fig. 5f), which contain Flex1 or Luc sequences in the donor templates. In this way, the I-SceI cleavable EGFP receipt cassettes (EGFP::Flex1/I-SceI or EGFP::Luc/I-SceI) contain perfect homology to the donor templates (D-Flex or D-Luc), while Flex1 but not Luc would form secondary structures after end resection. Due to insertion of Flex1 or Luc in the donor templates, we cannot use green signals for detection of HR products. Instead, we used a PCR-based analysis. We introduced BamHI and EcoRI Sites to the donor templates, and if HR is used, the BamHI and EcoRI sites would be transferred to the receipt cassettes. Thus, BamHI/EcoRI cleavable PCR products among all (uncleavable bands are the products of imperfect end joining) would reflect HR efficiency (see more detailed description in the manuscript). As shown in Fig. 5f, inactivation of Rad52 significantly reduces HR products from Flex1-containing reporter, but not Luc-containing reporter. These results support the hypothesis that even when perfect homology is present at the donor templates, Rad52 is required for promoting HR at the DSB ends that are blocked by DNA secondary structures. 4 6. The interpretation of the results in Fig. 6 may not be correct. FANCM is important for the response to replication stress in various situations, and its function is clearly not limited to CFSs. Even if FANCM deficient cells are dependent on Rad52 for proliferation, it does not necessarily mean that the proliferation defect of the double mutant arises from problems at CFSs. The dependency of FANCM deficient cells on Rad52 is virtually the same as the previously reported Rad52 dependency of BRCA2 defective cells (Zeng et al. 2010 PNAS). This dependency could be explained by the redundant HR functions of FA/BRCA proteins and Rad52.
Response: We agree that the synthetic lethality phenotype of FANCM knockout and Rad52 knockdown is not necessarily caused only by the functions that we described at CFSs. We have a discussion on that in the "Discussion" section. In addition to CFSs, there are many other types of DNA secondary structures such as G quatraduplexes (G4s). FANCM may play an important role to protect those sites in addition to CFSs, and in the absence of FANCM, Rad52 would also be needed for repair of DSBs at those sites as well. Thus, the proliferation defects of the double mutants could arise from problems more than CFSs.
However, the synthetic lethality interaction of FANCM and Rad52 is distinct from that of BRCA2 and Rad52 as reported previously, where it was proposed that Rad52 serves as an alternative HR mechanism when HR is impaired in BRCA2 deficient cells 9 . FANCM is not essential for HR and its deficiency only causes mild impairment in HR, which is in sharp contrast to a strong HR defect observed in BRCA2 deficient cells. BRCA1 and BRCA2 knockout mice are embryonic lethal due to a HR defect, while FANCM mice grow normally. Loss of FANCM results in a substantial increase of HR-mediated mitotic recombination at Flex1, whereas inactivation of BRCA1 or BRCA2 leads to a reduction of HR at Flex1. Thus, the dependency of Rad52 and FANCM for cell survival is different from the redundant HR function of Rad52 for BRCA2.
Reviewer #2: 1. The part of Introduction describing the conclusion is too long and overlaps with Discussion. It should be shortened.
Response: We have revised the Introduction and simplified the conclusion part.
2. Figure 1c and 1d are out of the order in citation in the text. This should be corrected.
Response: We have corrected the order of Figures. 3. Page 7, 2nd paragraph, the authors should make clear to readers that the DNA binding activity of FAAP24 is required for recruitment of FAAP24 to CFS, whereas the FANCM interacting motif is dispensable for the same recruitment. As written, it is unclear to reader whether the authors are describing recruitment of FAAP24 or FANCM.
Response: We revised that part in the manuscript to make it clear that DNA binding of FAAP24 but not its interaction with FANCM is important for FAAP24 recruitment.
4. RAD52 and FANCM are synthetic lethal, suggesting that the two proteins work in two parallel pathways to promote cell proliferation. This synthetic lethal interaction is different from their genetic interactions in CFS protection. The authors should make this clear to readers in the discussion.
Response: We propose that in the normal situation, FANCM pathway is used preferentially to protect Flex1 to prevent DNA secondary structure formation and subsequent DSB formation at Flex1, and thus Rad52 is minimally needed. However, when FANCM is deficient, massive DSBs are generated at Flex1, which then rely on Rad52 for repair. Thus, cells would die when Rad52 activity is inhibited in FANCM deficient cells. The synthetic lethality interaction of these two proteins is more likely due to their concerted and distinct roles, but not necessarily by paralleled activities. We added in more discussion to clarify our model based on identified mechanisms and synthetic lethality interaction of these two proteins.
5. Page 10, first paragraph, the authors stated that "these studies reveal an important function of RAD52 in mammalian cells in a special context when DNA ends are blocked by DNA secondary structure". I did not see evidence for this statement. Figure 5d shows that the luciferase contruct has the same effect as the Flex report. Do the authors think that the luciferase reporter can also form secondary DNA structures? If this statement is just speculation, please move this part to the Discussion.
Response: We have performed new experiments (Fig. 5f) and showed that even when perfect homology is present at the donor templates, Rad52 is still required for repairing DSBs at Flex1 but not Luc. This supports the model that Rad52 is required for HR when DSB ends are blocked by DNA secondary structures. Also see Response to Reviewer #1, comment 5.

Reviewer #3:
Major points: 1) Just one shRNA for each gene (FANCA, FANCD2, FAAP24, MHF1, RAD51, RAD52) was used for Figure 1b, 1c, Supplemental Figure 1c, Figure 5a,b,d. The authors should perform an "add-back experiment" by knocking down with shRNA followed by transfecting an expression plasmid to rescue the phenotype or use at least two shRNAs/gene. (For FANCM, the authors have performed the add-back experiments.) Response: We have repeated several key experiments using two different shRNAs for each gene. New data have been added in Fig.1b, Supplementary Fig. 2, Supplementary Fig. 4, Supplementary Fig. 6 and Supplementary Fig.10a, 10b, 10c. 2) Usage of just one cancer cell line (U2OS or HCT116) for most of the experiments is a concern, since whether the conclusion can be generalized is questionable. The authors should use multiple cell lines for some key experiments.
Response: We have repeated key experiments using more cell lines. We showed that Flex1 also induces mitotic recombination in Hela, MCF7 and T98G cells in addition to U2OS cells and inactivation of FANCM further increases genome instability at Flex1 as revealed by increased mitotic recombination (Fig. 1c and Supplementary Fig. 3a). In addition to U2OS cells, we also showed that Flex1 and other CFS-derived AT-rich sequences induce more mitotic recombination in FANCM KO HCT116 cells compared to wild type cells (new data in Fig. 1d and Supplementary Fig. 3b).
3) Figure 3e: the authors should test the effect of FANCM-MM1 mutant or depletion of other 6 FA proteins (FANCA, FANCD2, etc.) on fragile site breakage.
Response: The expression of CFSs in FANCA and FANCD2 deficient cells have been published 10 , which shows that loss of FANCA and FANCD2 leads to an increase of CFS breakage. We tested fragile site breakage at FRA16D in the FANCM-MM1 mutant and added in new data ( Supplementary  Fig. 8).
FANCM-MM1 mutant (defective in binding to the FA core complex) does not show significant defect in suppressing mitotic recombination at Flex1 (Fig. 1e), suggesting that the interaction of FANCM with the FA core complex is not important for Flex1 protection. However, FANCM-MM1 mutant shows an increased CFS expression (Supplementary Fig. 8), although not as significant as FANCM translocase mutant FANCM-K117R. This suggests that the interaction of FANCM with the FA core complex is still important for protection of CFSs, but mechanistically it is not through protecting the Flex1 site. In this regard, FANCA and FANCD2 are also important for CFS maintenance, but through other mechanisms. Consistently, loss of FANCA and FANCD2 activity does not cause hyper mitotic recombination at Flex1 (Fig. 1b, also see Response to Reviewer #1, specific comment 1).
4) Some statistical tests should be done for any quantitative data analyses.
Response: Thanks to the reviewer for pointing out this. We have revised the text accordingly.
2) In reference 14 (Howlett, N.G.,et al. Hum Mol Genet 14, 693-701 (2005)), it is reported that FANCD2 depletion in HCT116 leads to elevated fragile site instability following exposure to APH. The authors should comment on the interpretation of the discrepancy between ref 14 and the current paper.
Response: FANCD2 is important for CFS maintenance, but using mechanisms such as inhibition of dormant origin firing and prevention of DNA/RNA hybrid formation 8 , which are different from protection of AT-rich sequences at CFSs as described in this manuscript. Our HR reporter is specifically designed to detect HR-mediated repair of DSBs generated at Flex1 (see Figure 1 in this text), and it is expected that our reporter would not detect these mechanisms. Also see Response to Reviewer #1, specific comment 1.

Reviewers' C omments:
Reviewer #1: Remarks to the Author: The authors have performed a number of new experiments to improve the paper. However, some of the issues that I raised are not completely addressed.
General: I agree that many C FSs contain AT-rich structure-forming DNA sequences, but whether these sequences are sufficient for conferring fragility at endogenous C FSs is not always clear. Moreover, although the report assay developed by the authors is useful for characterizing specific AT-rich sequences derived from C FSs, whether the fragility observed in the reporter is the main source of fragility at endogenous C FSs is still unclear. For example, even if the AT-rich sequences from FRA3B are fragile in the reporter, are these sequences a significant source of fragility at endogenous FRA3B in cells? If the fragility of FRA3B comes from multiple sources, how significant is the contribution of these AT-rich sequences? It is also worth noting that all the AT-rich sequences tested in the reporter are from C FSs. What if some AT-rich sequences from non-fragile sites are tested in the reporter? It is clear that the reporter is a nice system to study the properties of AT-rich sequences, but whether the mechanistic details from this reporter are significantly relevant at endogenous C FSs is less clear.
1. I agree with the authors' explanation of why BRC A1 and FANC D2 did not display the same effects as FANC M in the reporter assay. 5. I am confusing by the authors' interpretation. In Fig. 5d, HR-Luc is Rad52 dependent. In Fig. 5f, HR-Luc/D-Luc is Rad52 independent. If I combine these results, I would conclude that Rad52 is important for dealing with a non-homologous sequence at DSBs. If we compare HR-Flex/D-Flex and HR-Luc/D-Luc in Fig. 5f, Rad52 has a specific role in HR-Flex/D-Flex. However, it is not clear whether Rad52 is required at DNA end, donor, or both. At any rate, the data do not specifically support that Rad52 functions on secondary DNA structures at DNA ends.
6. I don't agree with the authors' argument. Although BRC A1 and BRC A2 deficient cells display stronger HR defects than FANC M mutant, it is impossible to exclude that FANC M Rad52 double mutant is not severely defective for HR. The authors should test this directly.
Reviewer #3: Remarks to the Author: The authors have addressed all the concerns I raised. The manuscript has improved.
Minor points: Figure 5f The picture of one of the gels (the right one) is mislabeled; "HR-Flex/D-Flex" must be "HR-Luc/D-Luc".

Response to Reviewers' Comments
We thank the reviewers for the comments. In response to Reviewer 1's comments, we performed new experiments and have included additional new data in the manuscript (Fig. 3e and Supplementary Fig.  2b). We also corrected Figure 5 according to Reviewer 3's comment. Specific explanations are outlined below. The changes in the manuscript are marked by lines on the left side of the text.

Reviewer 1
General: I agree that many CFSs contain AT-rich structure-forming DNA sequences, but whether these sequences are sufficient for conferring fragility at endogenous CFSs is not always clear. Moreover, although the report assay developed by the authors is useful for characterizing specific AT-rich sequences derived from CFSs, whether the fragility observed in the reporter is the main source of fragility at endogenous CFSs is still unclear. For example, even if the AT-rich sequences from FRA3B are fragile in the reporter, are these sequences a significant source of fragility at endogenous FRA3B in cells? If the fragility of FRA3B comes from multiple sources, how significant is the contribution of these AT-rich sequences? It is also worth noting that all the AT-rich sequences tested in the reporter are from CFSs. What if some AT-rich sequences from non-fragile sites are tested in the reporter? It is clear that the reporter is a nice system to study the properties of AT-rich sequences, but whether the mechanistic details from this reporter are significantly relevant at endogenous CFSs is less clear.
Response: It has been shown that replication stalls at AT-rich sequences at endogenous CFSs 1 , suggesting that AT-rich structure-forming DNA sequences contribute to fragility at endogenous CFSs.
Our studies further showed that the expression of FRA16D and FRA3B is significantly increased when AT-rich sequences are not properly protected in FANCM deficient cells (Fig.3d). This suggests that AT-rich sequences at endogenous CFSs are indeed vulnerable sites for CFS breakage especially when the protection mechanism is defective.
As we discussed in the previous response, the presence of AT-rich sequences is one of the multiple mechanisms underlie CFS instability and the contribution of AT-rich sequences to fragility of a specific CFS may vary depending on the context of CFSs. AT-rich sequences at CFSs are normally protected by the FANCM pathway. The impact of these AT-rich sequences on CFS fragility becomes apparent when the protection mechanism is defective. For instance, we showed that endogenous FRA3B expression is increased in FANCM deficient cells (Fig.3d). We also presented new data that when FANCM is depleted, DSB formation is indeed increased around AT-rich sequences at the endogenous FRA3B locus (new data in Fig. 3e).
It is true that our reporter can also test the effect of AT-rich sequence from non-fragile sites as well as other structure-forming DNA sequences. In this aspect, it worth to note that many key regulators for keeping CFS stability also have general roles in protecting global genome stability. For instance, ATR has a general role in checkpoint activation and fork protection but identifying its specific role in CFS protection is still of great importance 2 . Conversely, finding the role of certain proteins in CFS protection (e.g. maintaining stability of AT-rich sequences in CFSs) and then elucidating their more general roles (e.g. also important for maintaining stability of other structure prone DNA sequences) is also important. FANCM protects CFSs, but this does not exclude its function in protecting other places in the genome containing DNA secondary structures. We discussed the possible role of FANCM in protecting other structure-forming DNA sequences in the genome (p18 top). The relevance of the findings from our repair reporters to the mechanism of CFS maintenance has been addressed by examining endogenous CFS expression (Fig. 3c, 3d and 3f) and DSB formation at endogenous CFSs including FRA3B (new data Fig. 3e).
1. I agree with the authors' explanation of why BRCA1 and FANCD2 did not display the same effects as FANCM in the reporter assay.
2. Are DSBs formed at the AT-rich sequences in FRA3B in FANCM knockdown cells? FANCM may have other functions in stabilizing forks in long genes.
Response: We performed new experiments using ChIP analysis of H2AX at AT-rich sequences in endogenous FRA3B locus and show that H2AX signals are significantly increased when FANCM is knocked down (new data in Fig.3e). This suggests that DSBs are accumulated around the AT-rich sequences at endogenous FRA3B in FANCM deficient cells and FANCM plays an important role in preventing DSB formation at the vicinity of AT rich sequences of FRA3B.
3. How does Ras induce replication stress specifically at Flex1 (or CFSs)? If Ras induces replication stress globally, why are AT-rich sequences affected more than other regions of the genome?
Response: Ras induces global replication stress, similar to HU or APH treatment. Due to replication stress, single stranded DNA (ssDNA) is accumulated at replication forks, which allows structure-prone DNA sequences (when they are in the ssDNA form), such as Flex1 and other AT-rich sequences, to form secondary structures at replication fork (see Figure 7, left and top, and see text). Thus, AT-rich sequences are affected more because they form DNA secondary structures upon replication stress, while other regions of the genome do not form secondary structures and thus would be affected less.