The interplay of RNA:DNA hybrid structure and G-quadruplexes determines the outcome of R-loop-replisome collisions

R-loops are a major source of genome instability associated with transcription-induced replication stress. However, how R-loops inherently impact replication fork progression is not understood. Here, we characterize R-loop-replisome collisions using a fully reconstituted eukaryotic DNA replication system. We find that RNA:DNA hybrids and G-quadruplexes at both co-directional and head-on R-loops can impact fork progression by inducing fork stalling, uncoupling of leading strand synthesis from replisome progression, and nascent strand gaps. RNase H1 and Pif1 suppress replication defects by resolving RNA:DNA hybrids and G-quadruplexes, respectively. We also identify an intrinsic capacity of replisomes to maintain fork progression at certain R-loops by unwinding RNA:DNA hybrids, repriming leading strand synthesis downstream of G-quadruplexes, or utilizing R-loop transcripts to prime leading strand restart during co-directional R-loop-replisome collisions. Collectively, the data demonstrates that the outcome of R-loop-replisome collisions is modulated by R-loop structure, providing a mechanistic basis for the distinction of deleterious from non-deleterious R-loops.


INTRODUCTION
1 Genome maintenance is dependent on the complete and accurate replication of the chromosomal DNA 2 prior to cell division. However, chromosomes present diverse obstacles to normal replication fork progression, 3 including protein-DNA complexes, DNA damage, and non-B-form DNA secondary structures. Among these, R-4 loops have emerged as critical determinants of transcription-replication conflict (TRC) linked to genome 5 instability in human developmental disorders and disease 1,2 . How R-loops inherently impact fork progression is 6 not understood.
corresponding to the RNA:DNA duplex. As in the analysis with T4 gp32, and consistent with the gel analysis in 1 Figure 1E, R-loops observed in the presence of formamide exhibit a range of sizes, demonstrating that the 2 size variations are not due to the sample preparation method. This is further confirmed by similar observations 3 using atomic force microscopy (AFM) (Figure S1D). In conclusion, the data demonstrates efficient formation of 4 canonical R-loop structures of variable size and position at the Airn sequence in vitro.

6
Both CD and HO R-loops perturb normal fork progression 7 To test the impact of R-loops on fork progression, DNA templates are linearized with NsiI prior to origin 8 firing. Replication products are digested with ClaI prior to gel analysis to reduce heterogeneity in their gel-9 mobility caused by the distributive initiation of strand synthesis at the origin ( Figure 2A+B) 49 . In the absence of 10 R-loops, the Airn sequence does not present a notable obstacle to fork progression in either orientation. In 11 contrast, CD or HO R-loops induce a striking loss of full-length replication products and the appearance of new 12 replication intermediates. Native gel analysis reveals reduced levels of full-length linear daughter molecules in 13 the presence of either CD or HO R-loops, as well as the appearance of stalled forks and partially replicated 14 daughter molecules resulting from the uncoupling of DNA unwinding from leading strand synthesis ( Figure   15 2C+D). Consistent with the native gel data, nascent strand analysis on denaturing gels demonstrates that CD 16 and HO R-loops cause a decrease in full-length rightward leading strands by ~ 75 % and ~ 55 %, respectively 17 ( Figure 2C). The levels of leftward leading strands are not affected, confirming that the loss of full-length 18 rightward leading strands is not due to reduced origin activity. Instead, the loss of full-length rightward leading 19 strands correlates with the generation of prominent ~2.3 kb or ~2.8 kb stall products at CD and HO R-loops, 20 respectively. The heterogeneity of the stalled leading strand products is reduced after ClaI digest, confirming 21 that they originate near ARS305 ( Figure S2A). Accordingly, leading strand stalling coincides with the Airn 22 sequence, which spans the region 1.8 kb -3.2 kb downstream of the origin (Figure 2A). The difference in the 23 position of the leading strand stall sites at CD and HO R-loops is, therefore, attributable to the asymmetric 24 distribution of R-loops in the Airn sequence ( Figure 1G). The heterogeneity in stall products remaining after 25 ClaI digest is likely a consequence of variations in R-loop sizes and positions (Figure 1).

26
Two-dimensional gel analysis demonstrates that leading strand stalling at CD and HO R-loops results 27 from both fork stalling and helicase uncoupling ( Figure 2E). Time course analyses demonstrate that leading 28 strands stall within minutes of origin firing and remain stable for at least two hours ( Figure S2B+C). Similarly, 1 native gel analysis reveals that a fraction of forks remains stably stalled specifically in the presence of R-loops 2 ( Figure S2B+C). Fork uncoupling occurs after leading strand stalling, as expected due to the reduced rate of 3 DNA unwinding by CMG upon uncoupling from leading strand synthesis 42,50 .

4
In addition to stalled leading strands, denaturing gel analysis reveals the formation of novel ~5.5 kb and 5 ~5.0 kb products indicative of leading strand restart downstream of CD and HO R-loops, respectively ( Figure   6 2C+D). The identity of these products is confirmed by multiple lines of evidence: First, these products are 7 sensitive to cleavage by the leading strand-specific nicking enzyme Nb.BbvCI, but not lagging strand-specific 8 Nt.BbvCI, demonstrating that they are nascent leading strand products ( Figure S2D). Second, these leading 9 strand products are not sensitive to ClaI cleavage, indicating that they do not originate near ARS305 ( Figure   10 S2A). Third, in regular time course experiments, these products accumulate with slower kinetics than stalled 11 leading strands, as expected for restart following leading strand stalling ( Figure S2B). Fourth, in experiments 12 in which nascent strands are pulse-labeled with [ 32 P]-dATP for 2.5 minutes following origin activation before 13 being chased with excess cold dATP restart products are not detectable, as expected if they form after origin 14 activation ( Figure S2C). Moreover, below we will show that this leading strand restart occurs at G4s in the 15 leading strand template.

16
We conclude that both CD and HO R-loops can perturb replication fork progression, which is consistent 17 with observations in human cells 14 . Replication abnormalities in either orientation include fork stalling, 18 uncoupling of leading strand synthesis from fork progression, and discontinuous leading strand synthesis 19 involving leading strand restart. A fraction of R-loops in either orientation is also bypassed by replisomes 20 without disruption. This diversity in outcomes is explained by the heterogeneity of R-loops in our templates 21 (Figure 1). Because the replicative DNA helicase, CMG, can efficiently bypass steric blocks on the displaced 22 lagging strand 51,52 , fork stalling is likely induced by obstacles on the leading strand. This raises questions 23 about the fork stalling mechanism as forks will encounter either RNA:DNA hybrids or the displaced non-24 template strand on the leading strand during CD and HO R-loop-replisome collisions, respectively. We, 25 therefore, investigated the molecular basis for replication aberrations at R-loops.

27
Fork stalling at R-loops is not dependent on Tof1 28 Csm3, Tof1, and Mrc1 (CTM) form a fork protection complex (FPC) that associates with replisomes to 1 promote normal fork progression, prevent the uncoupling of replisomes from DNA synthesis after nucleotide 2 depletion by hydroxyurea (HU), and maintain fork integrity during the replication of structure-forming 3 trinucleotide repeats [53][54][55][56][57][58] . In addition, Tof1 mediates fork pausing at protein-DNA complexes, such as the rDNA 4 replication fork barrier (RFB), tRNA genes, and centromeres 57,59,60 . We find that fork stalling, uncoupling, and 5 restart at CD and HO R-loops are not affected by CTM, demonstrating that the mechanism of fork stalling at R-6 loops is distinct from that at protein-DNA barriers ( Figure S2E). 7 8 PCNA suppresses helicase-polymerase uncoupling at Airn sequence 9 A previous study found that RNA at 5' primer-template junctions promotes strand-displacement by Pol  10 61 . We, therefore, tested if Pol  may promote fork progression through R-loops ( Figure 2F). On R-loop-11 containing templates, replication intermediates obtained in the absence of Pol  or its processivity factor PCNA 12 (together with its loader RFC) were indistinguishable from those obtained with complete replisomes. Thus, 13 strand-displacement by Pol  does not promote fork progression at R-loops. The data also demonstrates that 14 Pol  is not required for leading strand restart under these conditions. Intriguingly, however, we note that the 15 absence of PCNA increases leading strand stalling and fork uncoupling at the Airn sequence even in the 16 absence of R-loops ( Figure 2F, lanes 3+9), indicating a role for PCNA in maintaining the coupling of 17 replisomes to leading strand synthesis at G4 sequences. We speculate that PCNA mediates this function 18 through stabilization of the leading strand polymerase, Pol , on the template.

20
RNase H1 promotes fork passage specifically at CD R-loops 21 RNase H1 overexpression is commonly used to assess the contribution of R-loops to TRC. We, 22 therefore, tested how purified yeast RNase H1 ( Figure 3A) affects fork progression at R-loops in vitro. At CD 23 R-loops, RNase H1 decreased the levels of stalled and uncoupled replication intermediates while increasing 24 the formation of full-length replication products, suggesting that RNA:DNA hybrids on the leading strand can 25 impede fork progression ( Figure 3B+C, lanes 1-4; Figure S3A). In contrast, leading strand restart at CD R-26 loops was not affected by RNase H1, indicating that restart is not a direct consequence of RNA:DNA hybrids.

27
Instead, this suggests a role for G4s in leading strand restart at CD R-loops (Figure S3B), which is also 1 supported by data below. While not being dependent on RNA:DNA hybrid persistence, we note that formation 2 of the restart-inducing structures is dependent on transcription ( Figure 3B, compare lanes 1+2 to 3+4).

3
In contrast, fork stalling and uncoupling were largely unaffected by RNase H1 at HO R-loops, indicating 4 that RNA:DNA hybrids on the lagging strand are not the direct cause for these events (Figure 3B, lanes 5-8).

5
Instead, this data again suggests that G4s on the leading strand template, formed by the G-rich displaced 6 strand in this orientation, cause fork stalling or uncoupling. Moreover, the insensitivity of the fork block at HO 7 R-loops to RNase H1 demonstrates that G4s on the leading strand template, while dependent on transcription 8 for formation, are not dependent on the persistence of RNA:DNA hybrids on the lagging strand template 9 ( Figure S3C). In contrast, leading strand restart at HO R-loops is sensitive to RNase H1, indicating that the

20
In contrast to RNase H1, we find that Pif1 does not promote fork progression through CD R-loops, 21 supporting the notion that its limited processivity is insufficient to resolve long RNA:DNA hybrids ( Figure 3C, 22 lanes 1-3) 71 . However, RNase H1 and Pif1 together effectively eliminate fork stalling, uncoupling, and restart 23 events CD R-loops ( Figure 3C, lane 4). This suggests that RNase H1 and Pif1 coordinately promote fork 24 progression at CD R-loops by eliminating both RNA:DNA hybrids and G4s from the leading strand template.

25
Conversely, at HO R-loops Pif1 was markedly more efficient than RNase H1 in promoting fork progression 26 ( Figure 3C, lanes 5-7), which is consistent with the notion that G4s on the leading strand template cause fork 27 stalling at HO R-loops. Accordingly, addition of RNase H1 did not further promote fork progression at HO R-28 loops in the presence of Pif1 ( Figure 3C, lane 8). Together, the data suggests that fork stalling at R-loops can 1 be induced by RNA:DNA hybrids and G4s on the leading strand template, while neither structure presents a 2 fork block when formed on the lagging strand. To directly test the impact of G4s on fork progression we performed reactions in the presence of the 6 G4-stabilizer, pyridostatin (PDS). On DNA templates that lack the Airn sequence PDS does not impede DNA 7 replication ( Figure 3D, lanes 1+2). In contrast, PDS induces a strong block to fork progression specifically at 8 the Airn sequence in the HO orientation, i.e. when the G-rich strand forms the leading strand template, even in 9 the absence of R-loops ( Figure 3D, lanes 3-6). Thus, G4s forming on the G-rich strand of the Airn sequence in 10 the absence of transcription impede replication forks only when stabilized by PDS ( Figure 3D, lanes 3+5),

11
demonstrating that G4 stability determines the fork block potential of G4s. In contrast, fork stalling is induced at 12 HO R-loops even in the absence of PDS, and this effect is enhanced by PDS ( Figure 3D, lanes 7-10). Thus, 13 R-loop formation modulates the G4 composition on the displaced strand, which is consistent with studies 14 demonstrating a cooperative relationship between R-loops and G4s 26,27,72 . Since RNase H1 is unable to 15 prevent fork stalling at HO R-loops, R-loop formation induces the formation of fork-stalling G4s on the 16 displaced strand but is not required for their maintenance.

17
In the CD orientation, PDS also induces fork stalling at the Airn sequence in the absence of R-loops, 18 but the extent of stalling is less pronounced than in the HO orientation, consistent with the lower G4 potential 19 on the C-rich strand ( Figure 3E, lanes 1-4; Figure S1A). In addition, in the absence of R-loops, PDS induces 20 leading strand restart at the Airn sequence, demonstrating that replisomes can pass some PDS-stabilized G4s

27
Replisome uncoupling from leading strand synthesis at PDS-stabilized G4s indicates that some G4s 1 pose a block to the leading strand polymerase, Pol , but not the CMG helicase ( Figure 3E, native gel, lanes 2 2+4). Importantly, while a fraction of uncoupling events at G4s leads to persistent unwinding in the absence of 3 DNA synthesis, we also observe significant leading strand restart downstream of G4s. We note that the leading 4 strand restart efficiency observed at G4s appears to be greater than that observed previously at leading strand 5 DNA damage 74 , which we will discuss further below. Strikingly, unlike PDS-induced fork stall events in the CD 6 orientation, CD R-loop-induced fork stall events are sensitive to RNase H1, occur at a position closer to the 7 origin-proximal side of the Airn sequence, and are not exacerbated by PDS ( Figure 3E, compare lanes 5+6 to 8 1+2). This indicates that fork stalling at R-loops in the CD orientation is caused by RNA:DNA hybrids on the 9 leading strand template. Accordingly, stall sites are shifted downstream at CD R-loops in the presence of both 10 RNase H1 and PDS ( Figure 3E, lanes 6+8).

11
In summary, the data demonstrates that both RNA:DNA hybrids and G4s on the leading strand can        substrates. Strikingly, the wildtype G4 sequence, but not its mutant derivative, strongly attenuates DNA 25 unwinding by CMG when placed in the template strand ( Figure 4D). In contrast, substrate unwinding 26 efficiencies were equivalent in the presence of either wildtype or mutant G4 sequences on the non-template 27 strand ( Figure 4E). Thus, G4s on the leading strand template have the potential to block DNA unwinding by 28 1 the steric exclusion model for DNA unwinding by CMG 52 , G4s on the lagging strand do not pose an obstacle to 2 fork progression, enabling RNase H1 to promote efficient fork progression at CD R-loops.

11
In the CD orientation and in the absence of R-loops, replication products obtained in the presence of 12 Fen1/Cdc9 correspond primarily to full-length replication products, indicating that the G-rich strand of the Airn 13 sequence does not pose an intrinsic obstacle to lagging strand synthesis ( Figure 5A, lanes 1-4). In the 14 presence of CD R-loops, leading strand stall and restart products are detected as before in the absence of    This data is consistent with the notion that R-loop formation induces G4s on the displaced non-template 1 strand, blocking the progression of the lagging strand polymerase during CD replisome collisions, thereby 2 causing a gap in the nascent lagging strand. We, therefore, tested the ability of Pif1 to overcome the lagging 3 strand block at CD R-loops. Indeed, addition of Pif1 greatly reduced the formation of lagging strand gaps at CD 4 R-loops in the presence of RNase H1 ( Figure 5D, lanes 2+4). Thus, Pif1 promotes lagging strand synthesis at 5 G4s in vitro, which is supported by observations in vivo 32 .

22
Consistent with the observations above, we find that Pif1 promotes lagging strand synthesis at G4 810-828 23 in the absence of R-loops ( Figure 6C). This Pif1 function is specifically dependent on the Pif1 helicase activity, 24 as mutation of the Pif1 Walker B motif, which has been shown to disrupt the helicase but not DNA binding 25 activity of the human Pif1 orthologue 81 , abrogates the ability of Pif1 to promote lagging strand synthesis at 26 G4 810-828 ( Figure 6C). In contrast, Pif1 alone was unable to promote lagging strand synthesis at RNA:DNA 27 hybrids ( Figure 6D, lanes 1+3) but promoted completion of lagging strand synthesis in conjunction with RNase 28 H1 ( Figure 6D, lanes 2+4). We conclude that both RNA:DNA hybrids and G4s can inhibit lagging strand 1 synthesis at R-loops, and this inhibition is reversed by RNase H1 and Pif1, respectively.

R-loop transcripts can prime leading strand restart after CD R-loop-replisome collisions 4
Studies in E. coli have demonstrated that transcripts can prime leading strand synthesis after CD 5 collisions of replisomes with RNAP 82 . Whether a similar mechanism exists in eukaryotes, is unknown. This 6 question is particularly intriguing as CMG tracks along the leading strand template, while the bacterial 7 replicative DNA helicase, DnaB, tracks along the lagging strand template. Our purification protocol for R-loop 8 templates involves digestion of free RNA with RNase A. However, RNase A cleavage leaves a 3' phosphate 9 (3'-P), which prevents extension by DNA polymerase. We, therefore, converted RNA 3'-P ends into 3'-hydroxyl 10 (3'-OH) ends using T4 polynucleotide kinase (T4 PNK). Strikingly, T4 PNK treatment significantly increased the 11 levels of leading strand restart products at CD R-loops ( Figure 7A, lanes 1+3). Unlike leading strand restart 12 products obtained at CD R-loops in the absence of T4 PNK treatment, these novel restart products are 13 sensitive to RNase H1 treatment ( Figure 7A). Moreover, formation of these products is dependent on DNA 14 synthesis at replisomes ( Figure 7B) and is not observed at HO R-loops, in which the RNA:DNA hybrid is on 15 the lagging strand ( Figure 7C). Together, this data indicates the repriming of leading strand synthesis at the 16 RNA 3'-OH of RNA:DNA hybrids ( Figure 7D).

17
We find that under the conditions used here Pol , but not Pol  or Pol  is capable of efficiently 18 extending R-loop-associated RNA with DNA ( Figure S5A). Extension by Pol  is limited to ~ 200 -700 nt of 19 DNA and is further increased to up to ~ 2.5 kb in the presence of all three polymerases, well below the length 20 of complete restart products observed in the context of replisomes. This suggests that leading strand restart at 21 R-loop RNA during CD R-loop-replisome collisions is initiated by Pol  and extended by Pol  and/or Pol .

22
Although Pol  is not required for leading strand restart at R-loop RNA (Figure S5B), we do not rule out its 23 involvement, as discussed below.

24
How may replisomes utilize R-loop transcripts to prime leading strand restart? It is possible that CMG 25 slides over RNA:DNA duplexes at CD R-loops harboring nicks in the RNA, analogous to replisome encounters 26 with nicks in the lagging strand template 78 , which is supported by our demonstration that CMG can translocate 27 on RNA:DNA duplexes (Figure 4). To test this model, we treated CD R-loops with sub-saturating 28 concentrations of RNase H1 to introduce nicks in the RNA of CD R-loops. While the formation of full-length 1 leading strand products increases with the concentration of RNase H1 due to the resolution of the RNA:DNA 2 hybrids, leading strand restart correlates inversely with the concentration of RNase H1 ( Figure 7E).

3
Concomitantly, RNase H1 treatment reduces the formation of uncoupled products in the native gel analysis.

4
Similarly, we find that a sub-saturating concentration of purified RNase H2 promotes leading strand restart 5 specifically at CD R-loops ( Figure S5C). We note that RNase H generates a 3'-OH at the cleavage site, 6 eliminating the requirement for T4 PNK to convert RNA 3'-P ends into 3'-OH ends for restart here 83 . We 7 conclude that R-loop-associated RNAs can prime leading strand restart when CMG translocates over

18
We uncover several ways in which replisomes can continue progression at R-loops. 5' RNA flaps 19 promote the unwinding of RNA:DNA hybrids by CMG on the leading strand, allowing unhindered fork 20 progression. This mechanism is consistent with the observation that CD transcription-replication collisions 21 reduce R-loop levels in cells 14 . In the HO orientation forks may also progress continuously, provided the 22 displaced R-loop strand, which forms the leading strand template for the replisome, is devoid of inhibitory 23 secondary DNA structure. We demonstrate here that G4s can inhibit DNA unwinding by CMG, but it is 24 conceivable that other DNA secondary structures forming on the displaced non-template strand similarly 25 impede fork progression 84 . The ability of replisomes to pass through G4s is likely modulated by G4 stability, 26 which has been shown to influence the impact of G4s on genome stability 73 .

27
In addition, we demonstrate that forks can progress through R-loops harboring RNA:DNA hybrids or 1 G4s with polymerase-stalling potential on the leading strand. This results in the uncoupling of leading strand 2 synthesis from replisome progression. Persistent uncoupling would be expected to induce the S phase 3 checkpoint in cells, which may facilitate the clearance of polymerase blocks before S phase completion 85,86 .
4 Importantly, we also identify distinct mechanisms by which leading strand synthesis can be restarted at G4s 5 and RNA:DNA hybrids.

6
In the case of G4s, leading strand synthesis is reprimed downstream of the block, causing a gap in the 7 nascent leading strand. Transient uncoupling of leading strand synthesis from fork progression at G4s has 8 been suggested to cause epigenetic instability in vertebrate cells, where repriming is promoted by PrimPol 87 .

9
Notably, the leading strand restart observed at G4s here appears to be significantly more efficient than that 10 reported previously at DNA damage sites 74 . We hypothesize that transient stalling of CMG, which may be less 11 pronounced at DNA damage that does not pose a physical obstacle to the CMG, may promote leading strand 12 repriming at G4s.

13
Our data demonstrates that leading strand restart at RNA:DNA hybrids occurs by an 'on-the-fly' 14 mechanism in which Pol  extends the RNA 3' end with DNA before handover to the leading strand 15 polymerase at the replisome (Figure 7F). To a degree this restart mechanism is analogous to that at DNA duplexes. This mechanism requires the RNA 5' end of RNA:DNA hybrids to be annealed to the template strand 24 in order to prevent unwinding by the replisome. The structure of R-loops in vivo is not known and is likely 25 variable. However, we note that R-loop formation may be promoted by invasion of the DNA duplex by the RNA 26 5' end in order to limit topological clashes during the winding of the RNA around the template strand 91 .

27
Alternatively, flush RNA 5' ends may be encountered by the replisome at nicks in the RNA, analogous to 28 lagging strand nicks 78 . Such nicks may be generated by RNase H1 or RNase H2, whose activities are spatially 1 and temporally limited, respectively 92,93 .

2
We find that both RNA:DNA hybrids and G4s can also induce gaps in nascent lagging strands. This is 3 consistent with lagging strand G4s impeding DNA replication in vivo and with G4s posing a block to Pol  in 4 primer extension assays 32,80 . However, the block to lagging strand synthesis by RNA:DNA hybrids is surprising 5 given the proficiency of the lagging strand machinery in removing RNA:DNA hybrids at Okazaki fragment 5' 6 ends 94 . The reasons for this are currently unknown, but it is possible that long RNA 5' flaps or G4s in the RNA 7 strand impinge on the ability of Pol  and Fen1 to process RNA:DNA hybrids.

8
We show that RNA:DNA hybrid-and G4-induced blocks to fork progression, as well as leading and 9 lagging strand synthesis, are mitigated by RNase H1 and Pif1, respectively. This data is consistent with in vivo 10 studies demonstrating that both RNase H1 and Pif1 suppress genome instability at R-loops 21,22,95 and promote 11 replication fork progression at R-loops and G4 sequences, respectively 8,68,96 . We note that the coincidence of 12 RNA:DNA hybrids and G4s at R-loops may limit the efficiency of RNase H1 overexpression to resolve R-loops, 13 which is commonly used to assess the contribution of R-loops to TRC. 14 R-loops in both the CD and HO orientation can induce persistent fork stalling, which requires the 15 RNA:DNA hybrid-and/or G4-resolving activities of RNase H1 and Pif1 for continued fork progression.

16
However, we also find that replisomes have an intrinsic capacity to progress through R-loops by bypassing or 17 unwinding G4s or RNA:DNA hybrids. Despite the potential to induce nascent strand gaps, continued fork 18 progression may be beneficial in order to establish homology-directed repair-competent chromatin on sister 19 chromatids 97,98 . Nascent strand discontinuities at R-loops may also be processed post-replicatively 99 , which 20 could allow the timely progression of S phase. Alternatively, replisome bypass of R-loops may promote the 21 resolution of R-loops analogous to the repair of DNA-protein crosslinks 100 .

22
The molecular characterization of R-loop-replisome collision presented here enhances our 23 understanding of the mechanism(s) by which G4 ligands can affect the growth of cancer cells 27 . Moreover, we 24 expect that the system developed here will aid future studies directed at characterizing the molecular functions

24
Calculation of G/C skew:

25
GC skew at every nucleotide of the 1374 bp Airn sequence was calculated using a 100-nucleotide 1 sliding window as described 20 . Briefly, the number of Gs and Cs was counted in each given window. The G/C 2 Skew was calculated as (G-C) / (G+C).

4
Templates for helicase assays 5 Individual oligonucleotides for helicase template preparation (Table 3) were gel-isolated using the crush 6 and soak method: Oligonucleotides were electrophoresed on 8 % denaturing polyacrylamide gels at 120 V for

17
The reactions were stopped by adding 20 mM EDTA and 0.12 % SDS. Reaction products were fractionated by 18 10 % native PAGE in a Biorad Mini-PROTEAN system at 75 V for 2 hours in 0.5x TAE buffer. Gels were dried 19 on a Whatman paper, and exposed to phosphorimager screen, and scanned on a Typhoon 7000 imager (GE 20 Healthcare). The images were quantified using ImageJ and plotted using GraphPad Prism software.

21
For assays in Figure 4B