Direct visualization of replication and R-loop collision using single-molecule imaging

Abstract R-loops are three-stranded nucleic acid structures that can cause replication stress by blocking replication fork progression. However, the detailed mechanism underlying the collision of DNA replication forks and R-loops remains elusive. To investigate how R-loops induce replication stress, we use single-molecule fluorescence imaging to directly visualize the collision of replicating Phi29 DNA polymerase (Phi29 DNAp), the simplest replication system, and R-loops. We demonstrate that a single R-loop can block replication, and the blockage is more pronounced when an RNA–DNA hybrid is on the non-template strand. We show that this asymmetry results from secondary structure formation on the non-template strand, which impedes the progression of Phi29 DNAp. We also show that G-quadruplex formation on the displaced single-stranded DNA in an R-loop enhances the replication stalling. Moreover, we observe the collision between Phi29 DNAp and RNA transcripts synthesized by T7 RNA polymerase (T7 RNAp). RNA transcripts cause more stalling because of the presence of T7 RNAp. Our work provides insights into how R-loops impede DNA replication at single-molecule resolution.


Introduction
R-loops are triple-stranded nucleic acid structures consisting of an RNA-DNA hybrid and a displaced single-stranded DNA ( ssDNA ) ( 1 , 2 ) .R -loops are generated when RNA transcripts anneal back into the template DNA strand ( 3 ) .R-loops play important roles in diverse cellular activities including gene expression, chromosome segregation, immunoglobulin class switching, telomere regulation and immune activation (4)(5)(6)(7)(8)(9) .However, the misregulated accumulation of R-loops results in imperfect transcription and transcription-replication conflicts ( TRCs ) , leading to genomic instability ( 10 , 11 ) .Rloops hinder the progression of the replication fork, resulting in replication stress ( 10 ,12-14 ) , and the displaced ssDNA be-comes susceptible to endonucleases ( 15 ,16 ) .Biochemical studies using a purified Esc heric hia coli ( E. coli ) DNA replication system demonstrated that R-loops on the leading strand cause transient replication stalling while they do not have a significant impact on the lagging strand.The E. coli replisome can synthesize DNA by bypassing R-loops ( 17 ) .Using a yeast replisome system, it was reported that R-loops perturb fork progression in both head-on ( HO ) and co-directional ( CD ) collisions ( 18 ) .A recent native electron microscopy ( EM ) study directly visualized TRCs mediated by R-loops, showing the bypass of R-loops by the replication fork ( 19 ) .However, the molecular mechanism by which R-loops cause replication stalling remains elusive.Moreover, our understanding of the precise interactions between replication forks and R-loops remains limited.In this study, we conducted single-molecule imaging using the simple DNA replication system provided by the Phi29 DNA polymerase ( Phi29 DNAp ) to investigate the molecular mechanisms underlying replication fork stalling at R-loops.
Phi29 DNAp is a single-subunit DNA polymerase of Bacillus subtilis bacteriophage Phi29 and belongs to B-family polymerases ( Figure 1 A ) ( 20 ) .Phi29 DNAp has high fidelity due to an exonuclease domain for proofreading ( 21 ,22 ) .Structurally, the polymerase is comprised of a unique TPR2 domain, which encloses a primer-template junction and accommodates only ssDNA by forming a narrow channel together with thumb, palm and exonuclease domains ( 23 ,24 ) .This configuration enables Phi29 DNAp to unwind duplex DNA without a separate helicase.In addition, the channel also bestows high processivity on Phi29 DNAp, bypassing the need for any auxiliary proteins such as a sliding clamp ( 23 ,24 ) .Therefore, Phi29 DNAp is one of the simplest DNA replication systems.Because of its high fidelity and processivity, the polymerase has been extensively utilized for molecular biology approaches such as single-molecule sequencing and strand displacement amplification (25)(26)(27)(28) .
In this study, we directly visualized the DNA replication catalyzed by Phi29 DNAp using DNA curtain, a high-throughput single-molecule imaging technique that combines lipid fluidity, nano-fabrication, microfluidics and total internal reflection fluorescence microscopy ( TIRFM ) ( 29 ,30 ) .We first characterized the biophysical properties of Phi29 DNAp at the single-molecule level.We then investigated the collision between replicating Phi29 and a single artificially inserted Rloop.We demonstrated that the single R-loop interrupts the Phi29 DNAp-dependent replication.When an RNA-DNA hybrid was formed on the non-template strand, replication was more severely stalled compared with the template strand.This result is consistent with previous TRC results showing that head-on TRCs are more deleterious than co-directional collisions.We revealed that secondary structures such as RNA-DNA hybrids and G-quadruplexes on the non-template strand impede the replication of Phi29 DNAp.Moreover, we observed the collision between Phi29 DNAp and RNA transcripts generated by T7 RNA polymerase ( T7 RNAp ) .The RNA transcripts cause more stalling than the artificial R-loop because of the presence of RNAp.Our results provide insights into the principles of how replication stress is generated by Rloops that are also relevant for higher organisms.

Sample preparation
Details of protein purification and DNA preparation are described in the Supplementary Information.The activity of purified Phi29 DNAp was compared with a commercial preparation ( M0269, NEB ) using rolling circle replication ( RCR ) .The activity of purified replication protein A labeled with enhanced green fluorescent protein ( RPA-eGFP ) was tested by electrophoretic mobility shift assay with 30 nt ssDNA ( 31 ) .Synthesized oligomers are listed in Supplementary Table S1.
To measure the melting temperature between RNA-DNA hybrid and DNA duplex, temperature-dependent circular dichroism was performed in Phi29 DNAp buffer.The temperature of the spectropolarimeter was changed from 25 to 98 • C at a rate of 2 • C / min, and the spectra between 220 and 300 nm were measured every 5 • C.

DNA curtain assay
Details for DNA curtain assays are described in the Supplementary Information.Briefly, flowcells with nano-trench patterns were prepared by following the previous protocol ( 32 ) .A concentration of 2 nM of engineered lambda DNA ( λ-I3 ) with biotin and a primer was pre-incubated with Phi29 DNAp at 30 • C in Phi29 DNAp buffer [50 mM Tris-HCl ( pH 7.5 ) , 10 mM MgCl 2 , 10 mM ( NH 4 ) 2 SO 4 and 4 mM DTT] supplemented with 0.1 mM dNTPs except for dGTP to attach Phi29 DNAp to the primer ( 33 ) .Then the λ-I3 was anchored on the lipid bilayer via biotin-streptavidin linkage ( Figure 1 B ) .All DNA curtain experiments were conducted at room temperature ( 25 • C ) .The DNA curtains were formed in imaging buffer [50 mM Tris-HCl ( pH 8.8 ) , 10 mM ( NH 4 ) 2 SO 4 , 10 mM MgCl 2 , 4 mM DTT, 3.2% glucose and 0.1 × gloxy].Replication in the curtain was started by switching from 'imaging buffer' to 'replication buffer' ( imaging buffer supplemented with 3 nM RPA-eGFP and 0.1 mM dNTP ) at a flow rate of 0.5 ml / min.For R-loop-containing GQ structures, 50 mM KCl was further supplemented to the replication buffer.The fluorescence signal was imaged by NIS-Element software ( Nikon ) for 6 min with 0.5 s / frame.For the replication resumption experiments at R-loops, 83 U / ml of RNase H ( M0297L, NEB ) was injected into the flowcell in replication buffer supplemented with 75 mM KCl to remove RNA.

Image analyses
Imaging data were exported into TIFF format and analyzed by ImageJ software ( NIH ) .Kymographs for each replication were made.Speed was analyzed by measuring the slope of the replication initiation point to end point.Processivity was analyzed by measuring the vertical distance between initiation and end points.Based on the kymographs, line and area species were counted, and then the relative population of area species was calculated by dividing the number of area species by the sum of area and line species.

Real-time visualization of DNA replication conferred by Phi29 DNAp
We purified Phi29 DNAp and compared its DNA synthesis activity with that of a commercial preparation using RCR, and found equivalent synthesis activity ( Supplementary Figure S1A, B ) .The replication behavior of Phi29 DNAp was examined using DNA curtains, comprised of engineered lambda DNA ( λ-I3 ) ( Figure 1 B ) ( 29 ,33 ) .One end of λ-I3 was tagged with biotin for anchoring on the lipid bilayer, while the other end is open with the terminal single-stranded region bound to a primer for replication ( Figure 1 B ) .During DNA synthesis, Phi29 DNAp unwound duplex DNA and generated a displaced ssDNA, which was visualized by binding of RPA-eGFP under TIRFM ( Figure 1 B; Supplementary Figure S1C,  D ) .DNA replication conferred by Phi29 DNAp in the presence of all four nucleotides was observed in real time by monitoring the extension of fluorescent lines, indicating RPA-eGFP binding to the displaced ssDNA ( Supplementary Movie S1 ) .About 77% of the λ-I3 molecules in the DNA curtains were replicated ( Figure 1 C ) .Given that replication as monitored by 'line formation' did not occur when dGTP was omitted, our assay reflects DNA replication conferred by Phi29 DNAp ( Figure 1 C; Supplementary Figure S1E ) .The starting position of the fluorescent lines was at ∼47 kbp, which is consistent with the primer position ( ∼47.5 kbp ) in λ-I3 ( Supplementary Figure S1F ) .We also tested a Phi29 DNAp deletion mutant ( TPR2 ) , in which the TPR2 domain ( amino acids 395-426 ) was deleted ( Supplementary Figure S1A ) .The TPR2 domain is essential for helicase activity and high processivity of Phi29 DNAp ( 23 ,24 ) .TPR2 did not show any replication activity in our system as evidenced by the absence of any growing fluorescent lines ( Supplementary Figure S1G ) .These data support that the fluorescence lines represent the replication of Phi29 DNAp.Unexpectedly, we observed two distinct types of replication ( Figure 1 D, E ) .One is a triangle-shaped kymograph ( area profile ) , which results from the covering of RPA-eGFP of the whole displaced ssDNA as expected ( Figure 1 D ) .The other one is a line-shaped kymograph ( line profile ) ( Figure 1 E ) .We assumed that the line-shaped kymographs would result from incomplete RPA binding, which may be due to an inability to completely unfold secondary structures of the displaced ssDNA.This notion was tested by increasing the RPA-eGFP concentration.With increasing RPA-eGFP concentration, the area profile became more populated ( Figure 1 F ) .In addition, we observed that the line profile changed to an area profile during the replication because the unwinding of secondary structures by RPA-eGFP molecules appears to occur at different time points ( Figure 1 G ) .

Biophysical properties of Phi29 DNAp
For both area and line profiles, Phi29 DNAp displayed linear movement with time until it stopped at a certain distance ( Figure 1 D, E ) .We did not observe transient pausing and backtracking, which were shown in the previous optical tweezers assays, because the spatiotemporal resolution of the DNA curtain is not sufficient ( ∼1 kbp / pixel and 0.5 s / frame ) (34)(35)(36) .From the kymographs, the replication speed and processivity of Phi29 DNAp were estimated at the single-molecule level ( Figure 2 A, B ) .At 2 nM RPA, where area and line profiles occurred in equal abundance, the replication speed at 0.1 mM dNTPs was ∼130 bp / s regardless of line and area profiles ( Figure 2 A ) .The speed did not change proportionally with RPA concentrations, indicating that RPA binding to the displaced ssDNA does not influence replication ( Figure 2 B ) .The replication speed we observed was slightly higher than in previous reports, probably because we used higher dNTP concentrations ( 34 ,35 ) .Rapid replication also ensures that the hydrodynamic force ( < 5 pN ) by buffer flow does not disturb the replication, consistent with the previous result that DNA polymerization by Phi29 DNAp is not influenced by tension to DNA lower than 14 pN ( 35 ) .The processivity was ∼30 kbp, which was independent of line and area species ( Figure 2 C, D ) .The processivity of Phi29 DNAp was smaller than the value ( ∼70 kbp ) estimated from the previous bulk experiment ( 37 ) .However, the bulk experiment did not completely exclude multiple turnovers of Phi29 DNAp and hence allowed for further elongation of DNA synthesis.Instead, taking into consideration the fact that the bacteriophage Phi29 genome is ∼19 kbp, the processivity we observed allows single Phi29 DNAp to replicate the entire Phi29 genome in most cases ( 38 ) .We examined the speed and processivity in relation to dNTP concentrations.The dNTP-dependent speed followed Michaelis-Menten kinetics with a V max of 150 bp / s and a K M of 36 μM, compatible with previous measurements ( Figure 2 E ) ( 36 ) .On the other hand, the processivity reached ∼30 kbp at 30 μM dNTP and remained constant at higher dNTP concentrations ( Figure 2 F ) .This indicates that the processivity of Phi29 DNAp is not affected by the dNTP concentration.At dNTP concentrations < 50 μM, Phi29 DNAp slowly replicated DNA and was unable to synthesize DNA up to 30 kbp within the given time window ( Figure 2 E ) .Hence, the processivity increased as the dNTP concentration increased.
We also tested bacterial ( E. coli ) ssDNA-binding protein ( EcSSB ) tagged with eGFP instead of RPA because Phi29 DNAp acts in bacteria ( Supplementary Figure S2A ) .EcSSB displayed line and area profiles similar to those formed by RPA ( Supplementary Figure S2B ) .The speed and the processivity in the presence of EcSSB were identical to those in the presence of RPA, indicating that the replication is not differentially affected by the ssDNA-binding proteins from different species ( Supplementary Figure S2C, D ) .However, the intensity of the fluorescence signal of EcSSB was lower than that of RPA.

Replication stalling at an R-loop with an RNA-DNA hybrid on the non-template strand
We next examined the collision of replicating Phi29 DNAp with an R-loop ( Figure 3 A ) .A single R-loop was artificially inserted into a specific location of λ-I3, which has seven nickase sites between base pairs 33 514 and 33 638 ( Supplementary Figure S3A ) ( 29 , 33 )    treatment, leading to RNA degradation, ensuring the R-loop formation ( Supplementary Figure S3D ) .Prior to the collision with the R-loop, we tested a bubble structure without RNA-DNA hybrid ( Figure 3 B ) .As expected, 73% of Phi29 DNAp molecules passed the bubble ( Figure 3 B; Supplementary Figure S4A ) ; 27% of stalling would result from any defects in bubble formation such as improper hybridization or ligation during λ-I3 preparation.Indeed, when the gap formed on λ-I3 was annealed with a complementary DNA oligomer, the stalling fraction was 24%, equivalent to that for the bubble, showing that, as expected, a mismatched sequence did not provide a major obstacle ( Supplementary Figure S4B ) .In addition, when we removed the RNA strand from the R-loop, the stalling fraction was 27%, which was comparable with that for the bubble ( Supplementary Figure S4C ) .RPA molecules bound to ssDNA of R-loops or bubbles ( Supplementary Figure S4D-F ) .Even in the presence of RPA, the passing fraction was high for a bubble, suggesting that RPA binding to the ssDNA either in bubbles or of R-loops does not hinder the replication progression.
We then investigated the collision between replicating Phi29 DNAp and the artificial single R-loop.The RNA strand in the R-loop was placed on either the template or the nontemplate strand ( Figure 3 A ) .When Phi29 DNAp encountered the R-loop, Phi29 DNAp passed or stalled at the R-loop.We also observed a transient pause at the R-loop followed by replication progression ( Supplementary Figure S4G ) .The transient pausing events were rare ( ∼7%, 6 out of 85 ) and the pause time was 36 ± 20 s.We scored these transient pauses as pass events because Phi29 DNAp actually passed the R-loop.Interestingly, the extent of stalling was dependent on which strand the RNA was annealed.When the RNA-DNA hybrid was formed on the non-template strand ( RD_NT ) , ∼62 ± 5% of Phi29 DNAp exhibited stalling, whereas ∼38 ± 5% of Phi29 DNAp showed stalling for the RNA-DNA hybrid on the template strand ( RD_T ) ( Figure 3 C-F ) .The stalling fraction for RD_NT was higher than the background stalling ( ∼27 ± 3% ) , and the difference was statistically effective ( Pvalue < 0.001 ) .In contrast, the stalling fraction for RD_T was not statistically different from the background stalling ( P -value ∼ 0.31 ) .These results clearly showed that the presence of a single R-loop can impede the Phi29 DNAp replication.However, the positioning of the RNA-DNA hybrid on the non-template strand ( RD_NT ) represented a significant barrier to replication, while the presence of the hybrid on the template strand ( RD_T ) does not exert the same inhibitory effect.Furthermore, we tested the collision between Phi29 DNAp and an R-loop at a high RPA concentration ( 20 nM ) , where line species were suppressed.Even at this elevated RPA concentration, we observed no significant alternation in the stalling fraction.This result strongly suggested that RPA concentration does not affect Phi29 DNAp behavior at the R-loop ( Supplementary Figure S4H, I ) .

RNA remains on the R-loop when the replication stalls
We next examined whether the R-loop remained intact following the collision with Phi29 DNAp or if the RNA strand was displaced.As RNA was labeled with Cy5, a 637 nm laser was illuminated to monitor the RNA after the collision ( Figure 3 G, H ) .In the absence of Phi29 DNAp, ∼10% of RNA disappeared, leaving 90% remaining, presumably due to Cy5 photobleaching and / or spontaneous RNA dissociation.In the case of RD_NT with 39 nt RNA, when Phi29 DNAp was stalled, 82% of the RNA remained at the R-loop.When Phi29 DNAp passed through the R-loop, 68% of the RNA still remained.These findings indicated that Phi29 DNAp does not disrupt the RNA in the RD_NT.For RD_T, when Phi29 DNAp was stalled at the R-loop, a substantial portion of the RNA ( 70% ) survived.Taken together with the RD_NT case, our results demonstrated that the stalled Phi29 DNAp does not lead to RNA displacement at R-loops.In stark contrast, when Phi29 DNAp passed the RD_T R-loop, the remaining RNA fraction was dramatically reduced ( 28% ) , indicating that Phi29 DNAp displaces RNA or unwinds an RNA-DNA hybrid on the template strand when it passes the R-loop.This is consistent with our result from an RNA-DNA hybrid experiment in Figure 4 A. On the other hand, there was still 28% RNA left even in the case of Phi29 DNAp passage for RD_T, implying that replication bypasses through R-loops as shown in previous studies ( 17 , 19 , 39 ) .For the R -loop with 20 nt RNA, we had similar results ( Supplementary Figure S5A, B ) .We also examined if the stalled replication resumes when the RNA is removed by RNase H ( Figure 3 I ) .RNase H eliminated almost all RNA molecules from R-loops within 3 min ( Supplementary Figure S3D ) .Upon injecting RNase H after replication initiation, we observed the resumption of stalled Phi29 DNAp, indicating that the R-loop structure indeed acts as a blockage to the progression of replication ( Figure 3 J, K ) .The relatively small fraction of resumption could be attributed to the dissociation of the stalled Phi29 DNAp.

The R-loop structurally hinders the replication progression
We explored how a single R-loop blocks the replication of Phi29 DNAp.First, we tested if Phi29 DNAp can synthesize DNA when passing through an RNA-DNA hybrid ( Figure 4 A ) .The fraction of passing polymerases through an RNA-DNA hybrid ( 73% ) was as high as that for a bubble, indicating that RNA-DNA hybrids do not impede the replication of Phi29 DNAp.This accounts for the high passing fraction for RD_T.We next reduced the size of RNA.For 10 nt RNA, the stalling fraction was reduced to 40% whereas 20 nt RNA showed a similar stalling fraction to 39 nt RNA ( Figure 4 B; Supplementary Figure S6A, B ) , suggesting that RNA size affects the replication stalling but has little effect beyond a certain size for the artificial single R-loop.We then tested an R-loop in which RNA at the 3 end was partially mismatched to DNA ( Figure 4 C; Supplementary Figure S6C ) .The mismatched 3 overhang did not change the stalling fraction ( 59% for the 39 nt RNA-DNA hybrid with a 19 nt mismatch and 64% for the 20 nt RNA-DNA hybrid with a 19 nt mismatch ) .To check if the stalling for RD_NT is specific to the RNA-DNA hybrid, we examined a D-loop, in which DNA was annealed to either the template ( DD_T ) or non-template strand ( DD_NT ) ( Figure 4 D ) .Similar to the Rloop, DD_NT showed a higher stalling fraction ( ∼56% ) compared with DD_T ( ∼31% ) ( Figure 4 E-G; Supplementary Figure S6D ) , indicating that RNA or DNA placed on the nontemplate strand has a tendency to impede replication.On the other hand, the overall stalling fraction for the D-loop was 6-7% lower than that for the R-loop.This difference could be potentially attributed to the difference of thermodynamic stability between a RNA-DNA hybrid and double-stranded DNA.The thermodynamic stability of RNA-DNA and DNA-DNA depends on the sequences ( 40 ,41 ) .To check the thermodynamic stability of the RNA-DNA hybrid and a DNA duplex, we measured the temperature-dependent cicular dichroism for each substrate ( Supplementary Figure S7A, B ) .As the temperature increased, the cicular dichroism spectra of both 39 bp RNA-DNA hybrid and DNA duplex gradually changed and abruptly changed at high temperature.When the change of peak at 265 nm was monitored, the melting temperatures of the RNA-DNA hybrid and the DNA duplex were 82.7 and 83.8 • C, respectively, indicating that there is no thermodynamic difference between 39 bp DNA duplex and 39 bp RNA-DNA hybrid ( Supplementary Figure S7C ) .These results suggest that the difference of stalling fractions for the R-loop and D-loop is not attributed to the thermodynamic stability of the RNA-DNA hybrid or DNA duplex in the loop structures.
In the future, further study is required to understand why the D-loop showed a lower stalling fraction than the R-loop.Based on the above results, we assumed that a secondary structure formed on the non-template strand leads to the stalling of Phi29 DNAp during replication.To prove this, we tested hairpin structures with a 4 nt loop and a stem with different stem lengths ( Figure 4 H ) .The pass fraction of a hairpin with a 5 bp stem is similar to that of a bubble ( 0 bp stem ) , showing that a 5 bp stem hairpin does not impede the Phi29 DNAp replication.However, with an increase of the stem length, the pass fraction decreased.These results support our assumption that the presence of secondary structures on the displaced non-template strand hinders the Phi29 DNAp replication.In addition, it is worth noting that there is a noticeable reduction of the pass fraction between 5 bp and 10 bp stems, suggesting that a hairpin with a 10 bp stem, at minimum, can block the Phi29 replication.

G-quadruplex formation at the displaced ssDNA enhances replication stalling
When the displaced ssDNA is guanine rich, it can form GQ structures, which facilitate R-loop formation and stabilize Rloops (42)(43)(44)(45)(46) .We examined the effect of GQ formation at the displaced ssDNA for RD_T on the Phi29 DNAp replication because RD_T displayed a reduced stalling fraction compared with RD_NT.To this end, when a bubble was formed during the λ-DNA preparation, oligomers that contained 1 × hT el_GQ, 2 × hT el_GQ or the cMyc promoter sequence were used ( Supplementary Table S1 ) .Hence, the displaced ssDNA of RD_T contained GQ structures such as 1 × hTel_GQ, 2 × hTel_GQ or the cMyc promoter ( Figure 5 A, B ) .The formation of GQs in Phi29 DNAp buffer was verified by cicular dichroism measurements ( Supplementary Figure S8A, B ) .Interestingly, even in the absence of K + , human telomeric sequences and the cMyc promoter formed GQ structures in the replication buffer because the replication buffer contained 10 mM ammonium ions, which can stabilize GQ formation even at low concentration ( Supplementary Figure S8A, B ) ( 47 ) .For 1 × hTel_GQ, the dichroism spectra did not display typical hybrid conformation in Phi29 DNAp buffer even in the presence of 50 mM K + because Phi29 DNAp buffer contained 10 mM MgCl 2 and 10 mM ( NH 4 ) 2 SO 4 , which can induce other GQ conformations (48)(49)(50) .The 2 × hTel_GQ showed the hybrid conformation, which can form double stacks of GQs in the presence of 50 mM K + ( Supplementary Figure S8A ) ( 51 ,52 ) .On the other hand, the cMyc promoter sequence, which is also known to form a GQ structure, displayed a typical dichroism spectrum in Phi29 DNAp buffer regardless of K + ( Supplementary Figure S8B ) (53)(54)(55)(56) .RPA-eGFP did not bind 1 × hTel_GQ even at 30 nM, as reported previously, and hence GQs in our experiments were not disrupted by 2 nM RPA-eGFP ( Supplementary Figure S8C ) ( 57 ) .We first examined the effect of GQs on replication without the RNA-DNA hybrid.When GQ was formed on the displaced ssDNA, the stalling fraction was increased, showing that GQ by itself has an inhibitory effect on the replication ( Figure 5 A, C; Supplementary Figure S8D ) .This result is consistent with previous studies showing that GQs block the replication in vitro and in vivo ( 58 ,59 ) .When GQs were formed on the displaced ssDNA of RD_T, 1 × hTel_GQ and 2 × hTel_GQ increased the stalling fraction by 1.4 and 1.8 times, respectively ( Figure 5 B, C; Supplementary Figure S8E ) .In particular, 2 × hTel_GQ increased the stalling fraction up to 63%, which was comparable with that of RD_NT.This increase of stalling for RD_T with 2 × hTel_GQ was due to the synergistic effect of R-loop and GQ formation.Even in the absence of K + , the human telomeric sequence still blocked the progression of Phi29 DNAp at the bubble because it could form GQ structure in the replication buffer ( Supplementary Figure S8A, F ) .We next tested the promoter of the cMyc gene, which was also inserted into λ-I3.cMyc GQ alone blocked the replication of Phi29 DNAp in up to 57% of all cases ( Figure 5 D ) .When cMyc GQ was formed on the displaced ssDNA of RD_T, the stalling fraction increased up to 73%, which was higher than 2 × hTel_GQ ( Figure 5 E ) .We believe that this dramatic increase results from higher thermodynamic stability of cMyc GQ compared with hTel_GQs.cMyc GQ has a melting temperature ( T m ) of 84 • C and a free energy difference ( G ) between unfolded and folded conformations at 100 mM K + of 7.6 kcal / mol, whereas for 1 × hTel_GQ these values are 66 • C and 4.7 kcal / mol ( 55 , 60 , 61 ) .Taken together, our results demonstrated that GQ formation at the displaced ssDNA of an R-loop enhances the stalling of the progression of replication.Moreover, the thermodynamic stability of GQ influences the stalling.These results strongly support our claim that secondary structures of the non-template strand block the replication of Phi29 DNAp.

Collision of Phi29 DNAp with transcription of T7 RNAp
We also investigated how transcribing RNAp affects replication using T7 RNAp.T7 RNAp was used to generate fluorescent RNA transcripts with Cy5-labeled UTP ( Figure 6 A; Supplementary Figure S9A ) .To produce RNA transcripts by T7 RNAp on λ-DNA, we inserted the T7 promoter sequence into λ-DNA ( Figure 6 B ) .RNA transcripts produced by T7 RNAp in the λ-DNA were fluorescently visualized in DNA curtains ( Supplementary Figure S9B ) .We then examined the collision between Phi29 DNAp and T7 RNAp transcripts.Depending on the positions of biotin and the primer, HO and CD collisions between Phi29 DNAp and T7 RNAp were chosen ( Figure 6 B, C ) .In the case that T7 RNAp was not removed, Phi29 DNAp mostly stalled at RNA transcripts ( Figure 6 D, E ) .The stalling fractions were 87% and 72% for HO and CD collisions, respectively.HO collision induced more stalling events than CD collision ( Figure 6 E ) .We then examined whether or not RNA transcripts remained at the site of the collision.When Phi29 DNAp was stalled at the transcripts, RNA transcripts mostly remained ( 86% for HO and 73% for CD ) , indicating that stalled Phi29 DNAp does not remove RNA transcripts ( Figure 6 F ) .In addition, the stalling fractions at the transcripts were higher than those at the artificial R-loop.Such an increase of stalling fractions could result from T7 RNAp remaining at the transcripts.To examine whether or not the remaining T7 RNAp affects the stalling, we heat-inactivated T7 RNAp or deproteinized the reaction using thermo-labile proteinase K, which is inactive when shifted to 55 • C. In either case, we did not observe any Cy5-labeled RNA transcripts in DNA curtains after removing T7 RNAp ( Supplementary Figure S9C, D ) .We thus speculated that RNA transcripts were also dissociated when T7 RNAp was removed.Instead, we examined the collision with T7 RNAp bound at the T7 promoter in the absence of rNTPs.Excessive T7 RNAp was incubated with λ-DNA containing a T7 promoter to ensure that the T7 promoter site was mostly occupied by T7 RNAp ( Supplementary Figure S9E ) .T7 RNAp alone could impede the replication ( 49% stall ) ( Figure 6 G ) .However, the effect of T7 RNAp alone is less than that of transcribing RNAp containing RNA transcripts, suggesting that the high stalling rate for transcribing RNAp is attributed to the presence of RNAp in RNA transcripts.

Speed and processivity difference from previous results
The replication speed of Phi29 DNAp from our singlemolecule experiments is ∼130 bp / s, which is higher than that of previous bulk experiments ( < 33 bp / s ) ( 62 ,63 ) .This discrepancy primarily results from the low dNTP concentration of the previous bulk experiments, in which the dNTP concentration was 20 μM.At 20 μM, our replication rate is ∼50 bp / s, which is comparable with the previous bulk results ( Figure 2 E ) .In addition, in the previous assays, the binding and DNA polymerization of Phi29 DNAp and Phi29 terminal protein ( TP ) were not separated.In contrast, our single-molecule experiments allowed us to differentiate between polymerase binding and DNA synthesis, providing a more accurate measurement of the replication speed.The replication speed of Phi29 DNAp was also assessed using optical tweezers at the single-molecule level ( 34 ,35 ) .Under low tension to DNA ( < 10 pN ) at 50 μM dNTP, the replication speed was ∼80 bp / s, which is comparable with that in our study ( Figure 2 E ) .On the other hand, our observed processivity ( ∼30 kbp ) is 2.3 times lower than the estimation obtained from the previous bulk replication assays ( ∼ 70 kbp ) ( Figure 2 F ) ( 37 ) .The earlier experiments adopted RCR with primed M13 closed-circular ssDNA, which did not represent a singleturnover reaction.In contrast, our single-molecule measurements captured the DNA synthesis of the polymerase without free polymerases, ensuring a single-turnover replication.Hence, our determination of processivity is more reliable.Despite the observed decrease in the processivity, our measured processivity is still sufficient to encompass the entire genome of Phi29 DNA ( ∼19 285 bp ) , indicating that a single round of replication can replicate the Phi29 genome in the vast majority of cases ( 38 ) .

Stalling efficiency at the collision with an R-loop depends on the position of RNA-DNA hybrid
We tested the collision between Phi29 DNAp and a single Rloop without RNAp.We found that the R-loop with RD_NT inhibited the Phi29 replication more severely compared with RD_T.To investigate why RD_NT blocks the replication more strongly than RD_T, we checked the effect of RNA-DNA hybrids, which are structurally and thermodynamically different from DNA homoduplexes.Phi29 DNAp demonstrated the ability to synthesize DNA traversing an RNA-DNA hybrid, indicating that Phi29 DNAp can displace the RNA-DNA hybrids during replication ( Figure 4 A ) .To reduce the thermodynamic stability of the RNA-DNA hybrid, we shortened the length of the RNA within the R-loop ( Figure 4 B ) .When the RNA-DNA hybrid was reduced to 20 bp, the stalling fraction did not change.Instead, with a further reduction to a 10 bp RNA-DNA hybrid, the stalling fraction decreased to ∼40%, which was not as dramatic as anticipated.Similarly, the 3 -mismatched RNA also blocked the replication at the same rate as RD_NT.These findings suggest that the thermodynamic stability of the RNA-DNA hybrid does not play a critical role in the replication stalling.Subsequently, we tested a D-loop configuration.Similar to the R-loop configuration, DD_NT exhibited a stronger hindrance on Phi29 DNAp than DD_T.The stalling fraction was slightly lower for the D-loop than for the R-loop even though the thermodynamic stability of DNA duplex was similar to that of the RNA-DNA hybrid ( Figure 4 F, G; Supplementary Figure S7C ) .These results demonstrate that the replication stalling observed for RD_NT is not solely attributed to the presence of an RNA-DNA hybrid.
We then hypothesized that a secondary structure in the displaced ssDNA or the non-template strand hinders the repli-cation of Phi29 DNAp.To verify the hypothesis, another secondary structure, a hairpin configuration on the non-template strand, was tested.The hairpin also inhibited the progression of replication, demonstrating that formation of a secondary structure on the non-template strand such as an RNA-DNA hybrid and DNA duplex stalls Phi29 DNAp-mediated replication.Our claim is also supported by our GQ experiments.The addition of GQ structures to the non-template strand increased the stalling efficiency of RD_T by 1.7 times, implying that the nucleic acid structures on the non-template strand would interfere with Phi29 DNAp.Why does a secondary structure interfere with the Phi29 DNAp-mediated replication?Structurally, the strand displacement activity comes from the sharp bend ( ∼90 • ) of the template strand inside the Phi29 DNAp and the steric exclusion of the non-template strand at the narrow channel that is formed by the TPR2, palm and finger domains, together with the exonuclease subdomains ( 20 ) .A single-molecule optical tweezers assay reported that when the sharp template bending was inhibited under tension, the strand displacement was impeded, and Phi29 DNAp underwent a long pause ( 35 ) .When a secondary structure such as an RNA-DNA hybrid, DNA-DNA annealing, GQ or a hairpin is formed at the non-template ssDNA, it reduces the flexibility of displaced ssDNA at the junction and prevents the template strand from bending sharply, and hence Phi29 DNAp may stall.Many helicases including replicative ones require the sharp bending of the template strand to unwind duplex DNA ( 64 ,65 ) .Besides the direct impact of a secondary structure on the non-template strand, there is a possibility that the secondary structure indirectly influences the template strand.Such an indirect effect could play a role in shaping the behavior of Phi29 DNAp.On the other hand, we cannot exclude the possibility that Phi29 DNAp can have two conformations: one for a stationary mode and another for a progressive mode needed for DNA synthesis.When the Phi29 DNAp encounters a R-loop or D-loop, it might change the conformation from a progressive mode to a stationary mode.The speculations can be verified by other biophysical studies such as single-molecule FRET.

Collision of Phi29 DNAp with RNA transcripts of T7 RNAp
We examined the collision between Phi29 DNAp and RNA transcripts generated through ongoing T7 RNAp, which was more biologically relevant.The collision with the RNA transcripts led to higher stalling fractions compared with the artificial R-loop ( Figure 6 E ) .Such an increase can be attributed to the remaining T7 RNAp.We attempted to remove solely T7 RNAp.However, when T7 RNAp was removed, RNA transcripts were equally dissociated.Structurally, the 8 bp RNA-DNA hybrid formed as part of the transcription bubble is thermodynamically unstable ( Figure 6 A ) .Hence, the removal of T7 RNAp may induce the dissociation of RNA transcripts.Instead, we tested T7 RNAp bound to the T7 promoter, which also blocked the replication.Our results suggest that the remaining T7 RNAp at an RNA transcript plays an important role in the replication stalling.This is consistent with the finding that RNA mostly remained when Phi29 DNAp was stalled because T7 RNAp was not displaced by Phi29 DNAp ( Figure 6 F ) .
In addition, HO collision exhibited a higher stalling fraction ( ∼80% ) than CD collision ( ∼70% ) .The HO and CD configurations were identical to RD_NT and RD_T, respectively.Therefore, the difference between HO and CD appears to come from the location of the RNA-DNA hybrid.Consistent with our results, many previous studies about TRCs reported that HO collision is more deleterious than CD collision ( 66 ,67 ) .On the other hand, in a previous study using an E. coli replication system, CD collision caused more stalling than HO collision, suggesting that R-loops in the lagging strand cause little stalling ( 17 ) .Because E. coli replicative helicase moves along the lagging strand, the CD collision is similar to the configuration of RD_T in our experiments, giving less stalling.

G-quadruplex formation at a R-loop enhances the replication stalling
We demonstrated that GQ structures formed at the displaced ssDNA of a single R-loop enhance replication stalling.Moreover, multiple GQs more strongly block the replication.It was reported that GQs and R-loops are formed in human telomeres consisting of tandem repeats of d ( TTAGGG ) (68)(69)(70) .Human telomeric GQ can be stacked by forming hybrid structures in the presence of K + .On the other hand, telomeric R-loops derived from non-coding RNA TERRA ( telomeric repeat-containing RNA ) are involved in telomere maintenance and alternative lengthening of telomeres ( ALT ) ( 69 ,71 ) .Our results imply that GQs at R-loops synergistically cause replication stress at telomeres, which is known as a hallmark of ALT cells ( 70 ) .On the other hand, we tested the promoter sequence of the cMyc gene, which is also known to form a GQ structure and is involved in gene regulation ( 55 ,56 ) .For cMyc GQ, the stalling fraction is higher than for hTel_GQs.The cMyc GQ ( G : 6-8 kcal / mol ) is more stable than 1 × hTel_GQs ( ∼3 kcal / mol ) ( 55 , 60 , 72 ) .The increased stalling conferred by cMyc GQ results from the high thermodynamic stability of the GQ structure.This is consistent with our data which revealed that 2 × hTel_GQ showed more stalling than 1 × hTel_GQ.Our results show that the thermodynamic stability of GQs influences the stalling of replication at R-loops.Furthermore, our results imply that the stable secondary structures at the displaced ssDNA enhance the R-loop stability ( 43 ) .

Figure 1 .
Figure 1.Replication of Phi29 DNAp.( A ) Str uct ure ( PDB: 2PZS ) and domains of Phi29 DNAp.( B ) Schematic of a DNA curtain for Phi29 DNAp replication.Engineered lambda DNA ( λ-I3 ) is modified with biotin and a primer for Phi29 DNAp replication at either end.In the DNA curtain, Phi29 DNAp synthesizes DNA from the primer unwinding duplex DNA.RPA-eGFP binds to the displaced non-template strand.RPA-eGFP is fluorescently imaged by TIRFM.( C ) Phi29 DNAp replication efficiency.In the presence of all dNTPs, 77.3 ± 5.0% of DNA molecules display replication.In the absence of dGTP, no replication is shown.( D ) A representative kymograph for ( left ) the area profile and ( right ) its schematic model.The black bar and arrow at the left indicate the barrier and flow direction, respectively.( E ) A representative kymograph for ( left ) the line profile and ( right ) a schematic model.The black bar and arrow at the left indicate the barrier and flow direction, respectively.( F ) Relative population of area species according to RPA concentration.Relative population is calculated by dividing the number of area species by the sum of area and line species.The error bars are obtained from the standard deviation ( SD ) in triplicate.( G ) A kymograph for transition from line profile to area profile.The black bar and arrow at the left indicate the barrier and flow direction, respectively.

Figure 2 .Figure 3 .
Figure 2. Speed and processivity of Phi29 DNAp.( A ) Speed distributions for area ( magenta ) and line ( green ) profiles of Phi29 DNAp replication at 2 nM RPA.Each distribution is fitted with a single Gaussian function.The peak centers for area and line profiles are 127 ± 2 and 131 ± 3 bp / s, respectively, indicating little difference in the replication speed between area and line profiles.( B ) Replication speed of Phi29 DNAp according to RPA-eGFP concentration.Errors are obtained from the width of a single Gaussian fit.( C ) Processivity distributions for area ( magenta ) and line ( green ) profiles of Phi29 DNAp replication at 2 nM RPA.Each distribution is fitted with a single Gaussian function.The peak centers for area and line profiles are 28 ± 8 and 27 ± 10 kbp, respectively, indicating little difference in the processivity between area and line profiles.( D ) Phi29 DNAp processivity according to RPA-eGFP concentration.The data and error bars are obtained from the single Gaussian fitting of ( C ) .( E ) Replication speed as a function of dNTP concentration.Data are fitted by the Michaelis-Menten equation.V max ( maximum replication speed ) and K M ( Michaelis constant ) were estimated as 150.6 ± 12.2 bp / s and 36.2 ± 3.8 μM, respectively.( F ) Processivity as a function of dNTP concentration.

Figure 4 .
Figure 4. Replication stalling mechanism at an R-loop.( A ) ( Top ) Schematic of an RNA-DNA hybrid inserted into the non-template strand.( Bottom ) Pass and stall fractions of Phi29 DNAp in the presence of the RNA-DNA hybrid at the non-template strand.The total number of molecules analyzed ( N ) is 133.Error bars represent the SD in triplicate.( B ) Stall fraction of Phi29 DNAp according to the size of RNA at the RD_NT R-loop.( C ) Collision of Phi29 DNAp with the RD_NT R-loop having a 3 overhang of RNA, which formed a 39 bp RNA-DNA hybrid with 19 nt mismatches at the 3 end.The total number of molecules analyzed ( N ) is 88.Error bars represent the SD in triplicate.( D ) Schematic of collision of Phi29 DNAp with a D-loop, which had 39 nt DNA on either the template strand ( DD_T ) or the non-template strand ( DD_NT ) .Cy5 is labeled at the 3 end to a v oid un w anted replication from the D-loop DNA.( E ) Kymographs for the collision events between Phi29 DNAp and a D-loop with a DNA-DNA hybrid on the non-template strand ( DD_NT ) .Phi29 DNAp is either stalled at the D-loop ( top ) or passes the D-loop ( bottom ) .The magenta arrowhead and yellow dashed line indicate the D-loop position.( F ) Pass and stall fractions for a D-loop with DD_T.The total number of molecules analyzed ( N ) is 158.Error bars represent the SD in triplicate.( G ) Pass and stall fractions for the D-loop with DD_NT.The total number of molecules analyzed ( N ) is 102.Error bars represent the SD in triplicate.( H ) ( Lef t ) Sc hematic of collision of Phi29 DNAp with a hairpin on the non-template strand.The hairpin consists of a 4 nt loop and a stem with different lengths ( 5, 10, 15 and 21 bp ) .( Right ) Pass fraction as a function of stem sizes of hairpin str uct ures.The data point at 0 stem is adopted from the pass fraction of a bubble str uct ure in Figure 3 B. The number of molecules analyzed for each data point is > 100.The error bar represents the SD in triplicate.

Figure 5 .
Figure 5. G-quadruplex ( GQ ) effect on replication stalling at an R-loop.( A ) Pass and stall fractions for 2 × hTel_GQ in a bubble without RNA-DNA hybrid at 50 mM K + .The schematic is shown above.The total number of molecules analyzed ( N ) is 115.Error bars represent the SD in triplicate.( B ) Pass and stall fractions for 2 × hTel_GQ for the RD_T R-loop at 50 mM K + .The schematic is shown above.The total number of molecules analyzed ( N ) is 135.Error bars represent the SD in triplicate.( C ) Stall fractions for hTel_GQs in a bubble and an R-loop ( RD_T ) at 50 mM K + .* P -value ≤ 0.05, ** P -value ≤ 0.01 and *** P -value ≤ 0.001.( D ) Pass and stall fractions for 1 × cMyc GQ in a bubble without RNA-DNA hybrid at 50 mM K + .The total number of molecules analyzed ( N ) is 127.Error bars represent the SD in triplicate.( E ) Pass and stall fractions for 1 × cMyc GQ for the RD_T R-loop at 50 mM K + .The total number of molecules analyzed ( N ) is 76.Error bars represent the SD in triplicate.

Figure 6 .
Figure 6.Collision of Phi29 DNAp with RNA transcripts.( A ) Str uct ure of T7 RNAp ( PDB: 1H38 ) .( B ) Schematic of λ-DNA for the collision between Phi29 DNAp and RNA transcripts of T7 RNAp.The T7 promoter is inserted between base pairs 13 829 and 13 847 of λ-DNA.Depending on which end of the λ-DNA has biotin and the primer, ( top ) HO and ( bottom ) CD collisions are determined.( C ) Schematic of the DNA curtain assay for ( top ) HO collision and ( bottom ) CD collision of Phi29 DNAp and RNA transcripts generated by T7 RNAp.( D ) Kymographs for the collision between Phi29 DNAp and RNA transcripts of T7 RNAp: ( top ) HO collision and ( bottom ) CD collision.Cy5-labeled RNA transcript and RPA-eGFP for replication are denoted as red and green arrows above, respectively.The black bar and black arrow on the left represent the barrier and flow direction, respectively.The yellow dashed line and the orange arrowhead on the right indicate the position and orientation of the T7 promoter that is inserted into the λ-DNA, respectively.( E ) Pass and stall fractions for the collisions of Phi29 DNAp and RNA transcripts without removal of T7 RNAp.N represents the number of analyzed molecules.The error bars are SDs in triplicate.( F ) RNA remaining fractions for the stalled Phi29 DNAp.N represents the number of analyzed molecules.Error bars are obtained from the SD of binomial distribution.** P -value ≤ 0.01.( G ) Pass and stall fractions for the collision of Phi29 DNAp and T7 RNAp bound at a T7 promoter site in the absence of rNTPs.N represents the number of analyzed molecules.The error bars are the SD in triplicate.