Improved short nascent strand sequencing (iSNS-seq) enhances DNA replication origin detection and reduces non-origin biases

  1. Miiko Sokka
  2. John M. Urban
  3. Nicola Neretti
  4. Susan A. Gerbi

Reviewed by

Read the full article See related articles

Discuss this preprint

Start a discussion What are Sciety discussions?

Listed in

Log in to save this article

Abstract

Identifying DNA replication origins in human and other metazoan genomes has been challenging, as highlighted by the fact that various methods for mapping them have produced conflicting results. A popular method, short nascent strand sequencing (SNS-seq), enriches newly replicated short single-stranded DNA by size selection and λ -exonuclease ( λ -exo) digestion of parental DNA. Surprisingly, SNS-seq has never been validated in Saccharomyces cerevisiae where origins have been well characterized genome-wide. We improved the SNS-seq protocol through biochemical optimization and benchmarked its origin-mapping sensitivity and precision with traditional SNS-seq in asynchronous populations of S. cerevisiae . The improved SNS-seq protocol significantly enhanced the enrichment of origin-derived DNA. Strikingly, the traditional SNS-seq failed to detect known origins and instead enriched non-origin DNA, likely arising from RNA:DNA hybrids. These findings have important implications for the interpretation of previously published datasets that rely on λ -exo for origin mapping. Overall, our biochemical and genomic analyses help unravel the mystery of the inconsistencies between SNS-seq and other techniques used to map DNA replication origins genome-wide.

Article activity feed

  1. Review Commons

    Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

    Learn more at Review Commons

    Reply to the reviewers

    Reply to the reviewers:

    We thank the three reviewers for their thoughtful and constructive comments, which have helped us to better explain our arguments and strengthen the manuscript. In this revision, we have clarified several points raised in the initial review and added new evidence that we believe further supports our conclusions.

    Point-by-point responses:

    Reviewer #1:

    ____Additional in vitro experiment using artificial R-loop structures as substrate for lambda-exonuclease could prove the efficacy of the nuclease to digest through R-loop structures.____

    We agree this experiment would be the cleanest way to demonstrate λ-exo's behavior on an R-loop substrate under defined biochemical conditions, and we would have liked to include it. However, this work originated in a laboratory that has since been closed due retirement, and further wet-lab experimentation is not feasible for this revision.

    Nonetheless, the manuscript already provides strong convergent evidence for RNA:DNA hybrid-mediated obstruction, short of this direct reconstitution. First, in vivo colocalization analysis shows that tSNS-seq (but not iSNS-seq) signal is enriched at S1-DRIP-seq and S9.6 ChIP-seq sites in an RNase H-sensitive manner, directly implicating RNA:DNA hybrids at the genomic loci where tSNS-seq peaks form. The reciprocal is also true, RNA:DNA hybrid signal is enriched at tSNS-seq sites, but not iSNS-seq sites (nor at known origins), in an RNase H-sensitive matter. Second, this obstruction signature is not a yeast-specific artifact of our analysis. The same asymmetric, tRNA/snoRNA-anchored peak shape is present in independently generated tSNS-seq datasets from Drosophila and C. elegans. In fact, C. elegans shows this artifact is in the non-replicating control, and is eliminated in the RNase control. Furthermore, a recent publication using (strand-specific) tSNS-seq in Trypanosoma bruceireports that large majority (90%) of its peaks map to R-loops (Stanojcic *et al. *2026), although they report it as mapping true replication origins, not as a demonstration of the artifact. Finally, we added an analysis comparing R-loops mapped by RIAN-seq in mouse (Li et al. 2025) and two independently generated mouse tSNS-seq datasets (Cayrou et al. 2015 and Pratto et al.2021) that show strong correlation. Taken together, this is not one experiment pointing to hybrid-mediated blockage, but a repeated, multiple-species in vivo-validated pattern observed in tSNS-seq by multiple independent laboratories.

    What our data cannot do is isolate the enzyme's behavior on a defined R-loop substrate in vitro, separate from all other cellular context. That is a distinct and narrower question from whether hybrid-mediated obstruction explains the peaks we observe. The latter is what the convergent evidence above addresses. Future work can nail down the exact enzymology and conditions needed to replicate this in vitro.

    In Fig. 1C, there seems to be a minor increase of lambda-exo activity on R10D30 substrate in the Tris-HCI buffer compared to Glycine-KOH buffer, maybe a quantification using the loading control (D20) would help.

    As suggested by the reviewer, we have quantified band intensities for the RNA-DNA chimera (R10D30), the G4 oligo (D10G427D23), and the all-DNA control (D25) shown in Fig. 1C, normalized to the D20 loading control at each timepoint (new Supplementary Figure S5). As this gel-based densitometry is semi-quantitative at best, the values below should be interpreted as indicative rather than precise measurements; nonetheless, they confirm the reviewer's observation: the RNA-DNA chimera was digested to a somewhat greater extent in Tris-HCl (35% signal remaining at 16 h) than in Glycine-KOH (65% remaining). We have revised the Results text to report these values explicitly and to clarify our claim. The relevant passage in the Results now reads: "However, the entire oligo (D10G427D23) was digested in Tris-HCl buffer very efficiently, with barely visible bands after as early as 1 h of λ-exo digestion (1% remaining at 16 h, comparable to the D25 control). The RNA-DNA chimera (R10D30) was digested slowly in both buffers over the 16 h time course, though to a somewhat greater extent in Tris-HCl (35% remaining) than in Glycine-KOH (65% remaining). This shows that RNA-primers offer partial, but not full, protection from λ-exo digestion in either buffer, distinct from the buffer-dependent protection conferred by G-rich DNA in glycine-KOH buffer. Thus, conditions for better digestion through G4-containing DNA in Tris-HCl buffer may be more suitable for mapping origins with SNS-seq."

    We further note that this observation does not bear on the final iSNS-seq protocol, since iSNS-seq relies on 5'-OH rather than 5' RNA to protect nascent strands from λ-exo digestion. The buffer-dependent difference in RNA-primer protection described here is therefore relevant only to the mechanistic characterization of λ-exo behavior in Fig. 1C, and does not affect the enrichment logic or performance of iSNS-seq itself.

    In Fig. 2D, lower panel, the digestion time of RecJf at different units were not indicated.

    We thank the reviewer for catching this point. We have revised the Figure 2D legend to remove this ambiguity, clarifying that the top panel shows a RecJf timecourse, while the bottom panel shows a titration of RecJf units (0–300 U) at a fixed 16 h digestion time.

    In Fig. 3B, since SNS-seq detect fired origins, an overlap of iSNS rep 1 and iSNS rep 2 peaks with confirmed origins from OriDB that are also early and efficient origins may yield a higher percentage of overlap.

    We thank the reviewer for this suggestion. Rather than stratifying discrete peak-overlap percentages, which are sensitive to peak-calling thresholds and tend to understate genuine concordance, we assessed this using continuous signal enrichment at OriDB-confirmed origins, that were stratified into quartiles by annotated origin efficiency (Hawkins et al., 2013) (see modified Fig. 3C). Both iSNS-seq replicates showed a clear trend of increasing median signal from the least to the most efficient origin quartile. This is consistent with the expected behavior of SNS-seq, which detects origins in proportion to their firing frequency in the asynchronous population, and we now state this explicitly in the manuscript.

    In Fig. 3D, in addition to showing the tSNS and iSNS data on ChrIV with indicated known origins, showing at the same time the corresponding FORK-seq, OK-seq, and ORC ChIP signal will give a more comprehensive view of the performance of different origin mapping methods.

    We agree with the reviewer that including complementary origin-mapping datasets alongside tSNS-seq and iSNS-seq would provide a more comprehensive view of method performance. We have added FORK-seq initiation site midpoints, ORC ChIP-seq signal, and OK-seq origin efficiency metric (OEM) signal to Fig. 3D, alongside the tSNS-seq and iSNS-seq tracks and annotated known origins on Chr IV. In addition, we have added close-up views of the tSNS-seq and iSNS-seq tracks around three selected origins in Fig. 3D, which highlight that the prominent tSNS-seq peaks do not align with known origins, in contrast to iSNS-seq.

    In Fig. 5F, in supporting the model, a comparison of tSNS-seq peak with known budding yeast DRIP-seq datasets would help to see if the tSNS method indeed enrich for RNA:DNA hybrid signals.

    We agree with the reviewer that a direct comparison to known R-loop maps would substantially strengthen the proposed model. We have added this analysis, described briefly below, in a new Results subsection and Figure 6.

    We obtained raw, publicly deposited budding yeast S1-DRIP-seq and S9.6 ChIP-seq data, including RNase H-treated and RNase H-deficient (RNase HΔ) conditions, and called peaks from these datasets ourselves using our own pipeline, so that all datasets were processed identically. A new heatmap colocalization figure (Fig. 6A) shows that tSNS-seq (but not iSNS-seq) signal is clearly elevated at these R-loop peaks, most strongly in the RNase HΔ condition and abolished in RNase-treated controls, consistent with genuine RNA:DNA hybrid dependence. The reciprocal analysis confirms this specificity: R-loop signal is concentrated at tSNS-seq peaks, but shows no enrichment at iSNS-seq, OK-seq, or FORK-seq peaks, or at confirmed origins. R-loop signal is also enriched at tRNA, snoRNA, and snRNA genes, mirroring the tSNS-seq enrichment pattern at these same loci. Together, these results directly demonstrate that tSNS-seq peaks colocalize with bona fide R-loops rather than replication origins, supporting the model in Fig. 5F.

    Finally, to test whether this observation generalizes beyond yeast, we correlated published mouse RIAN-seq signal — an antibody-free, nuclease-based R-loop mapping method developed independently of SNS-seq — with two independent mouse tSNS-seq-type datasets (Cayrou et al. 2015; Pratto et al. 2021). tSNS-seq enrichment correlated significantly with RIAN-seq R-loop signal across replicates and genomic scales, indicating that the same hybrid-mediated obstruction mechanism likely operates in a mammalian system as well. Moreover, analyses of tSNS-seq signal in Drosophila and C. elegans confirms that the enrichment bias at tRNA sites also exists in those systems. The C. elegans data shows it is not dependent on DNA replication, and is dependent on RNA, just like in yeast.

    ____Reviewer #2:____

    The authors' sugestion that tSNS miscalling of origins may be due to the presence of RNA:DNA hybrids is largely persuasive in theoretical terms but is, surpisingly, untested. The first issue that is unexamined in whether or not RNA:DNA hybrids would survive the two rounds of 95 degree denaturation that are central to both froms of SNS; can the auhtors comment or provide evidence?

    Second, it is important that the authors test their proposal (Fig.5) of RNA:DNA hybrids being the cause of at least some non-origin tSNS signal by mapping the correspondance of their SNS-seq data with the locations of RNA:DNA hybrids, since there are several forms of such mapping available for S. cervisiae (eg using S9.6 antiserum DRIP-seq, or RNaseH1 ChIP-seq).

    We thank the reviewer for raising these two important points, both of which push us to be more precise about what our model does and does not claim.

    On denaturation and hybrid survival: We agree this is a critical point to clarify, and we think it reflects a distinction we had not made explicit enough in the original submission. We are not proposing that in vivo-formed R-loops survive the 95°C denaturation steps intact. That would indeed be difficult to reconcile with the protocol. Rather, our data suggest that the highly abundant RNA species themselves (tRNAs, snoRNAs) survive denaturation as free single-stranded RNA and re-anneal with their complementary genomic DNA strand at, or before, the λ-exo digestion step, which is carried out at 37°C over an extended overnight incubation. This reannealed RNA:DNA hybrid is what we propose obstructs λ-exo in tSNS-seq. Indeed, the preservation of this RNA throughout the protocol until the λ-exo step is a built-in feature of the traditional protocol, since the RNA is only hydrolyzed after the λ-exo digestion is complete. We have added the following sentence to the Discussion to make this explicit: “Importantly, we do not propose that in vivo R-loops survive the denaturation steps of the iSNS-seq protocol. Rather, we propose that RNA:DNA hybrids reform in vitro after denaturation at genomic loci that are prone to R-loop formation. The likelihood of hybrid reformation would be expected to increase with local RNA abundance. These reformed hybrids would then selectively block λ-exo digestion, producing the characteristic asymmetric enrichments observed in tSNS-seq.”

    On testing correspondence with mapped RNA:DNA hybrids: We agree, and we have added a new Results subsection and Figure 6 to directly test this. We generated a new heatmap colocalization figure (Fig. 6A) comparing tSNS-seq and iSNS-seq signal against published budding yeast S1-DRIP-seq and S9.6 ChIP-seq maps, including RNase-treated controls. tSNS-seq signal is clearly elevated over R-loop sites identified by these orthogonal methods, but not peaks called from RNase-treated controls; iSNS-seq signal is flat over both R-loop and control peak sites. We also performed the reciprocal analysis and see that R-loop signal is highly concentrated at tSNS-seq peaks, but not iSNS-seq peaks (Fig. 6B). Importantly, this RNA:DNA hybrid enrichment signal at tSNS-seq peaks is abolished in the RNase H-treated controls and enhanced in RNase deficient mutants that accumulate R-loops, both consistent with genuine RNA:DNA hybrid dependence. Finally, we demonstrate that, like tSNS-seq, the R-loop signal is also highly enriched at tRNA, snoRNA, and snRNA genes (Fig 6C). Analyses of tSNS-seq signal in Drosophila and C. elegans shows this SNS enrichment bias at tRNA sites also exists in those systems. Importantly, the C. elegans data shows it is not dependent on DNA replication, and is dependent on RNA, and is thereby consistent with the RNA:DNA enrichment artifact we discovered in yeast.

    As a further, independent test of generalizability, we correlated tSNS-seq signal with RIAN-seq (which maps R-loops genome-wide) using two published mouse datasets: the original tSNS-seq data from Cayrou et al. 2015 and the strand-specific tSNS-seq data from Pratto et al. 2021. We found a significant positive correlation with RIAN-seq R-loop signal in both, indicating the same mechanism operates in a mammalian system (Figure 6D, E). We also note in the Discussion that a recent report (preprint at the time of this review, Stanojcic et al. 2026) similarly found that 90% of tSNS-seq peaks in Trypanosoma bruceioverlap with mapped R-loops, providing independent, multiple-species support for this interpretation.

    Fig.S7. Can the authors comments on the poor reproducibility between SNS-seq replicates? Though there isa four-fold increase in concordance between iSNS experments compared with tSNS, 80% of potetial origins are missed; can this be explained? In this regard, is the statment 'Both of the iSNS-seq samples showed signal enrichment around most of the confirmed origins' correct: while Fig.3C gives this impression, Fig.3B appears to diagree.____

    We agree the original Fig. 3B (Euler diagram) understated the reproducibility of iSNS-seq and, on reflection, we do not think it was the appropriate metric to lead with. Euler/peak-set overlap is a binary, threshold-dependent measure. Peaks that fall just above the calling threshold in one replicate and just below it in the other are counted as fully discordant even when the underlying signal is well correlated. We have moved the Euler diagram to the supplement (Fig. S9) and replaced Fig. 3B with a cross-replicate F1 curve, and cross-replicate signal heatmaps. Together these analyses show that concordance between replicates is graded and substantially above what the fixed-threshold Euler comparison implied. We have revised the text accordingly, and noted that the aggregate enrichment is not fully captured by peak-overlap statistics for the reasons above.

    The same failure mode of overlap analysis applies to overlap of iSNS (or tSNS) peaks with known origins. While the SNS signal may be fully correlated with origin positions and even origin efficiency, parameter choices in peak calling and using binary overlap statistics can mask the underlying concordance of the data with known replication origins. Moreover, the list of known replication origins, even confirmed origins, is not a list of origins most likely to be active. SNS methods can only detect active origins. Therefore, we assessed concordance of the SNS methods with origin efficiency using signal enrichment at OriDB-confirmed origins that were stratified into quartiles by annotated origin efficiency (Hawkins et al., 2013) (see modified Fig. 3C). Both iSNS-seq replicates showed a clear trend of increasing enrichment signal from the least to the most efficient origin quartile. This is consistent with the expected behavior of SNS-seq, which detect origins in proportion to their firing frequency in the asynchronous population, and we now state this explicitly in the manuscript.

    Given the very nice data in Figs.1 and 2 showing the confounding effect of G4s on tSNS, can the authors comment on why G4s show little overlap with either SNS-seq mapping in Fig.4, and why they saw no increase in tSNS-seq or iSNS-seq signal over G4 motifs in Supplementary Figure S9B? Might this indicate that the in vitro work does not translate well to in vivo mapping?

    We agree this warranted comment and have expanded the Discussion accordingly. We were ourselves somewhat surprised not to see a higher genome-wide G4 signal, though not entirely so. This is not evidence that the in vitro work fails to translate in vivo: the same λ-exo bias has been shown to manifest genome-wide in human cells (Foulk et al. 2015). Rather, we suggest the limited yeast signal likely reflects yeast-specific properties, which has a genome with more uniform local GC content than human (new Supplementary Figure S22), far fewer G4 motifs overall (Wu et al. 2021), and possibly less stable G4 folding in vivo (Tran et al. 2011). Moreover, the G4s in the human genome are non-randomly clustered in the GC-rich isochores, which compounds the G4 bias further with the known GC bias of Lambda exonuclease. These points are now made in the revised manuscript.

    The known G4 problem was the basis for searching for better buffer conditions in this study. While we found conditions that eliminate the G4 bias in vitro, the yeast genome did not provide ample opportunity to test this. However, the yeast genome ultimately allowed us to discover a possibly more dominant systematic bias, which is the correlation with tRNA and other high copy number RNA species that likely form RNA:DNA hybrids in vitro. Moreover, the lack of G4s allows a clean separation of G4s and R-loops, features that are highly correlated in the human genome.

    A requirement to validate the interesting suggestion that RNA:DNA hybrids are a cause of tSNS-seq miscalling of origins; this should be a combination of in vitro tests and colocalisation of tSNS-seq signal and RNA:DNA signal in vivo using available datasets (eg DRIP-seq).

    We agree that a direct in vitro biochemical demonstration that λ-exo digestion is specifically obstructed by an RNA:DNA hybrid substrate would further strengthen this model. However, as the laboratory where the wet-lab work for this study was performed has since been closed due to retirement, additional experiments of this kind are not feasible for this revision. We have instead addressed this request through a combination of new in vivo colocalization analyses, and an additional line of convergent evidence from the literature, which together we believe make a compelling case for the model.

    First, we have now added direct in vivo colocalization evidence: a new heatmap figure (Fig. 6A) comparing tSNS-seq and iSNS-seq signal to published budding yeast S1-DRIP-seq and S9.6 ChIP-seq maps, with RNase H-treated controls. tSNS-seq signal is clearly elevated over R-loop sites identified by these orthogonal methods, but not peaks called from RNase-treated controls; iSNS-seq signal is flat over both R-loop and control peak sites. Importantly, this RNA:DNA hybrid enrichment signal at tSNS-seq peaks is abolished in the RNase H-treated controls and enhanced in RNase deficient mutants that accumulate R-loops, both consistent with genuine RNA:DNA hybrid dependence. Finally, we demonstrate that, like tSNS-seq, the R-loop signal is also highly enriched at tRNA, snoRNA, and snRNA genes (Fig. 6C). Analyses of tSNS-seq signal in Drosophila and C. elegans shows this SNS enrichment bias at tRNA sites also exists in those systems. Importantly, the C. elegans data shows it is not dependent on DNA replication, and is dependent on RNA, and is thereby consistent with the RNA:DNA enrichment artifact we discovered in yeast.

    Second, we further demonstate that R-loop and tSNS-seq correlation extends to mouse as well. A novel R-loop mapping method RIAN-seq (Li et al. 2025) uses λ-exo, together with nuclease P1 and T5 exonuclease, to selectively degrade single-stranded RNA, single-stranded DNA, and double-stranded DNA from digested genomic DNA, while RNA:DNA hybrids resist this treatment and are selectively recovered. That λ-exo digestion is used as one of the core enzymatic steps to enrich for RNA:DNA hybrid-containing DNA is itself a genome-wide demonstration that λ-exo activity is impeded by RNA:DNA hybrid structures. Building on this, we correlated tSNS-seq signal against RIAN-seq signal using two independent published mouse tSNS-seq datasets (Cayrou et al. 2015 and Pratto et al. 2021) and found significant positive correlation in both, directly linking λ-exo obstruction by RNA:DNA hybrids to tSNS-seq peak formation in an independent system (Fig. 6D, E).

    Finally, a recent study (Stanojcic et al. 2026) using tSNS-seq on trypanosomes noted the majority (90%) of their peaks were associated with R-loop structures. Thus, we can defensibly conclude that this correlation is seen in yeast, C. elegans, Drosophila, mouse, and trypanosome datasets.

    An explanation and/or comment on the low reproducibility and low signal-to-noise ratio of iSNS-seq; specifically, although it improves on tSNS, does it provide accurate origin prediction?

    We do not believe reproducibility is in fact poor; the discrete peak-overlap metric we originally used was overly conservative (although still significantly above random). On signal-to-noise: we have added FRiP-vs-cumulative-ranked-peaks curve (Fig. 3E) showing that both iSNS-seq and tSNS-seq achieves higher than random raw FRiP. While tSNS-seq achieved higher FRiP, the iSNS-seq shows greater specific enrichment (SN/PPV against OriDB) at real origins. This is because the majority of tSNS-seq reads are in tRNA-associated non-origin peaks. In other words, tSNS-seq's apparently favorable S:N is driven in large part by non-origin enrichment (e.g., R-loop-forming loci), consistent with our RNA:DNA hybrid-mediated bias hypothesis. In contrast, iSNS-seq's lower overall S:N reflects the removal of much of that strong non-origin signal, leaving a smaller but more origin-specific signal. Whereas the non-origin biases are strongly present in a sample, nascent strands from a replication origin are very rare in the sample in comparison; present only in the fraction of S-phase cells in an asynchronous population where the nascent origin-proximal DNA is Comment on the lack of in vivo evidence, from available mapping data, for the confounding effect of G4s seen in the in vitro experiments.

    In vivo evidence for this λ-exo bias does exist in human cells (Foulk et al. 2015). Its absence in our yeast mapping data is better explained by yeast-specific genomic features than by the effect being artifactual, or in vitro-only. We have revised explicit comment on this in the Discussion:

    “Given the clear G4-mediated λ-exo obstruction we observed in traditional conditions in vitro, as well as previously demonstrated genome-wide blockage at G4s in human cells and preference of λ-exo for AT-rich over GC-rich DNA (Foulk et al 2015), we were somewhat surprised not to observe a higher proportion of G4-overlapping fragments in tSNS-seq compared to iSNS-seq. However, this is not entirely unexpected for several reasons. First, the budding yeast genome is markedly more uniform than the human genome: local GC content (100 bp bins) spans a 10th–90th percentile range of 30–46% in yeast versus 26–55% in human (Supplementary Figure S22). In fact, the yeast genome has roughly half the variation of GC content found in the human genome as measured by standard deviation and median absolute deviation (MAD) (Supplementary Figure S18). Budding yeast also has substantially fewer G4 motifs than humans, both in absolute number and as a proportion of the genome (Wu et al. 2021), limiting the opportunity for a genome-wide G4 signal to emerge regardless of any per-motif blocking effect. Finally, we cannot rule out that budding yeast G4 motifs simply form less stable quadruplexes, as has been shown for yeast telomeric G4s specifically (Tran et al. 2011).”.

    Reviewer #3:

    While the method developed (iSNS-seq) appears to improve aspects of the previously used method (tSNS-seq), poor signal-to-noise ratios and reproducibility are major concerns. The signal-to-noise ratio of iSNS-seq, as shown in Figure 3D, appears low. The authors should discuss what practical consequences this has for the applicability of the method, especially in species with less well-defined/efficient origins, and how it affects the reliability of peak calling, and overlap with known origins. Consistently, the overlap of called peaks between biological replicates of iSNS-seq is relatively low (Fig3B). This raises a concern for the usability of the method, the interpretation of the genome-wide data and the quality metrics provided. The authors should discuss these issues.

    We have added a Discussion paragraph addressing this directly. Because the intrinsic signal-to-noise ceiling of SNS-seq-type methods scales with the size and firing efficiency/synchrony of the origin population (as detailed in our existing calculation based on Cadoret et al. (2008)), we now state explicitly that applying iSNS-seq to organisms or cell populations with less efficient, less synchronized, or less well-defined origins than budding yeast should be expected to yield correspondingly lower signal-to-noise and reduced peak-calling reliability, with a likely higher false-negative rate for genuine origins. We now acknowledge the need for further optimization to improve the underlying SNS:background ratio before extending the method to other organisms.

    On reproducibility: the original Fig. 3B (Euler diagram) understated the reproducibility of iSNS-seq and, on reflection, we do not think it was the appropriate metric to lead with. Euler/peak-set overlap is a binary, threshold-dependent measure. Peaks that fall just above the calling threshold in one replicate and just below it in the other are counted as fully discordant even when the underlying signal is well correlated. We have moved the Euler diagram to the supplement (Fig. S9) and replaced Fig. 3B with a cross-replicate F1 curve, and cross-replicate signal heatmaps. Together these analyses show that concordance between replicates is graded and substantially above what the fixed-threshold Euler comparison implied.

    We have replaced the Euler-diagram comparison with a graded, cross-replicate F1 analysis (new Fig. 3B) and supplementary correlation/Jaccard metrics, which show reproducibility is substantially above random for both methods and is in fact tighter for iSNS-seq than tSNS-seq at the level of peak-height correlation.

    Related to the above, please explain/discuss the rationale behind retaining all peaks, even the ones observed in only one of the biological replicates, when comparing called peaks to confirmed ORIs. In Figure 3B, clarify what are the percentages depicted.

    We agree with the reviewer that this point deserves fuller explanation, and we address the two parts in turn.

    Rationale for retaining single-replicate peaks. Origin firing is stochastic across an asynchronous population, and any single replicate is a depth-limited sample of the true origin population rather than an exhaustive one. Requiring a peak to be called in both replicates before comparing it to OriDB would systematically discard true origins that happen to be under-sampled in one replicate, artificially inflating apparent PPV at the direct cost of sensitivity. Since our goal in this analysis was to characterize each method's raw concordance with known origins, we compared each replicate's full peak set to OriDB independently.

    What the percentages in the diagram represent. We have moved this Euler diagram comparison from the main text to Supplementary Figure S9. We have clarified the legend of Supplementary Figure S9 (formerly Figure 3B) to read: "Percentages represent the fraction of the total combined peaks across all three sets being compared (i.e., the union of iSNS-seq rep 1, iSNS-seq rep 2, and confirmed ORIs for panel A; tSNS-seq rep 1, tSNS-seq rep 2, and confirmed ORIs for panel B) that fall into each region of the diagram."

    Why this analysis was moved to the supplement. Binary overlap statistics of this kind are informative but limited: they collapse a continuous relationship (signal strength versus origin identity) into a single threshold-dependent yes/no call, and are sensitive to peak-calling parameters. Moreover, the list of OriDB-confirmed origins is not itself a list of the origins most likely to be active in a given asynchronous population: most methods can only ever detect origins that actually fired, so a "true" origin that failed to overlap a peak may simply reflect biological non-firing rather than a false negative of the method.

    For these reasons, we chose to complement the overlap-based comparison with an analysis that queries origin correspondence directly against the experimental signal rather than against a binarized peak call. Specifically, we stratified OriDB-confirmed origins into quartiles by annotated origin efficiency (Hawkins et al., 2013) and examined SNS signal enrichment across quartiles (modified Fig. 3C). Both iSNS-seq replicates show a clear, monotonic increase in enrichment signal from the lowest to the highest efficiency quartile. This is the expected behavior for a method that detects origins in proportion to their firing frequency in an asynchronous population. We now state this rationale explicitly in the manuscript text: “To assess the reproducibility of iSNS-seq and tSNS-seq more rigorously than a fixed-threshold peak-overlap comparison allows […] Because this approach evaluates concordance at every possible rank cutoff rather than a single arbitrary threshold, it is not subject to the same sensitivity to near-threshold peaks that limits discrete overlap statistics such as Euler diagrams (Supplementary Figure S9).”

    The title and abstract do not accurately convey the limitations of the improved method and must be rephrased.

    We thank the reviewer for this comment and have revised the abstract accordingly. Specifically, we have (1) clarified that our benchmarking was conducted in S. cerevisiae precisely because it offers a well-defined set of confirmed origins for direct comparison, (2) made explicit that our enrichment claims are relative to traditional SNS-seq rather than absolute, and (3) added language framing this work as a proof-of-concept benchmark, with extension to metazoan systems as a next step rather than a claim already established here. We believe these changes address the concern while accurately reflecting what our data show.

    We have retained the title, as we believe "enhances DNA replication origin detection and reduces non-origin biases" is a comparative claim relative to traditional SNS-seq, which is directly supported by our benchmarking data, and does not imply resolution of broader field-wide inconsistencies in origin mapping.

    One of the manuscript's central mechanistic claims is that residual cellular RNA reanneals to genomic DNA, creating RNA:DNA hybrids that obstruct λ-exonuclease digestion. While the presented data from cells are consistent with this model, the manuscript lacks a direct biochemical demonstration of hybrid-mediated obstruction in a controlled system. The manuscript would benefit from in vitro studies corroborating this conclusion, using the in vitro system established.

    We agree this would be a valuable addition, and in principle the in vitro system established in this study (Fig. 1, 2) is well suited to such a test. However, the laboratory where this wet-lab work was performed has since been closed due to retirement, and we are unable to carry out new biochemical experiments of this kind for this revision.

    We believe the manuscript already provides convergent evidence for RNA:DNA hybrid-mediated obstruction, short of this direct reconstitution. First, our in vivo colocalization analysis shows that tSNS-seq (but not iSNS-seq) signal is enriched at S1-DRIP-seq and S9.6 ChIP-seq sites in an RNase H-sensitive manner, directly implicating RNA:DNA hybrids at the genomic loci where tSNS-seq peaks form. Second, this obstruction signature is not a yeast-specific artifact of our analysis. The same asymmetric, tRNA/snoRNA-anchored peak shape is present in independently generated tSNS-seq datasets from Drosophilaand C. elegans. Furthermore, a recent publication using stranded tSNS-seq in Trypanosoma brucei reports that a large majority (90%) of its peaks map to R-loops (Stanojcic *et al. *2026). Finally, we added an analysis comparing R-loops mapped by RIAN-seq in mouse (Li et al. 2025) and two independently generated mouse tSNS-seq datasets (Cayrou et al. 2015 and Pratto et al. 2021) that show strong correlation. Taken together, this is not one experiment pointing to hybrid-mediated blockage, but a repeated, cross-species in vivo-validated pattern observed in tSNS-seq by multiple independent laboratories. The distinct contribution of our yeast experiments is that a well-defined set of confirmed origins allowed us, for the first time, to directly distinguish genuine replication signal from this R-loop-associated background.

    What our data cannot do is isolate the enzyme's behavior on a defined R-loop substrate in vitro, separate from all other cellular context. That is a distinct and narrower question from whether hybrid-mediated obstruction explains the peaks we observe. The latter is what the convergent evidence above addresses. Future work can nail down the exact enzymology and conditions needed to replicate this in vitro.

    Currently, the RNA-dependence of the tRNA/snoRNA-associated obstruction signal is inferred from indirect evidence (UMAP clustering, feature overlap, peak shape) plus a re-analysis of an RNase-treated control from a previously published C. elegans dataset, not from a control generated within the authors' own yeast system, where the paper's primary genome-wide claims are made. Including such controls would be required to provide direct evidence for this central mechanistic claim.

    We appreciate the opportunity to clarify this, as we believe we already have exactly this control within our own yeast dataset. iSNS-seq itself functions as this RNA-dependence control for tSNS-seq: unlike tSNS-seq, the iSNS-seq protocol includes an RNA hydrolysis step (NaOH) performed before λ-exo digestion, removing the RNA that would otherwise be available to reanneal and form obstructing RNA:DNA hybrids. Both tSNS-seq and iSNS-seq are generated from the same size-selected input material from the same yeast cultures, differing specifically in whether RNA is retained (tSNS-seq) or hydrolyzed (iSNS-seq) prior to the λ-exo step (see Fig 3A, which is now improved to clarify this point). The result is a matched, within-system, RNA-present versus RNA-removed comparison, generated entirely in our own hands in budding yeast: tSNS-seq (RNA intact) shows strong signal and peak enrichment at tRNA/snoRNA loci, while iSNS-seq (RNA hydrolyzed) shows essentially none (Fig. 5). This is, in effect, an RNase-equivalent control built directly into our primary experimental design, rather than one requiring a separate treatment arm.

    We have clarified this point explicitly in the Discussion by noting that the tSNS-seq vs. iSNS-seq comparison itself constitutes the within-system RNA-dependence control, complementing the new yeast S1-DRIP-seq/S9.6 ChIP-seq RNase H colocalization analysis (Fig. 6A-C), which independently corroborates this at the level of orthogonal R-loop mapping methods.

    A substantial portion of the in vitro biochemical work (Figure 1B-C, E, Figure 2B-C with BKGmimic) is dedicated to establishing that G4 motifs cause significant λ-exo obstruction in glycine-KOH buffer, and that the Tris-HCl buffer switch resolves this. However, the genome-wide finding that G4 motif overlap is nearly identical between tSNS-seq and iSNS-seq (4-5% in both), with no significant differences in signal over G4 motifs in either dataset, requires clarification/discussion.

    We have clarified this in the expanded Discussion. We note it was somewhat surprising to us as well not to observe more obstruction in tSNS-seq samples. However, our data is consistent with prior human work showing the same λ-exo bias (Foulk et al. 2015), combined with the yeast genome's comparative GC uniformity (new Supplementary Figure S22) and its markedly lower G4 motif density (Wu et al. 2021), and possibly less stable yeast G4 folding (Tran et al. 2011). We believe these factors sufficiently explain why a robust in vitro effect does not produce a differential genome-wide signal in this organism.

    The documented G4 bias was a motivation to find new buffer conditions in the first place, and our further study of this problem highlighted something important for the field even if it doesn’t matter for the yeast genome: RNA primers are less effective at protecting downstream DNA than G4 motifs. Moreover, that observation led us to inventing a way where nascent strands were more strongly protected than G4s by using the 5’-OH after purposeful RNA hydrolysis. Finally, RNA hydrolysis led to the discovery that there is a major RNA-dependent bias in tSNS-seq. Thus, the paper documents the journey and key insights that led to the iSNS-seq protocol, and major findings in this study regarding the tSNS-seq protocol.

    None of the gel-based panels in Figure 1 (B, C, E) or Figure 2D state whether the result shown is from a single experiment or is representative of multiple independent experiments. Please state explicitly, for each gel-based panel, whether it represents a single experiment or is representative of repeated independent experiments (and if the latter, how many).

    We agree with the reviewer that this information was missing and should be stated explicitly. The gel panels in Figures 1B, 1C, 1E, and 2D arise from an extensive experimental search (for buffer conditions and additives) that would increase digestion through G4 motifs while preserving the RNA-DNA chimera, during which these conditions were tested many times over the course of the study. Across this large number of experiments testing different buffers and additives, the same core observations consistently emerged: (1) -exo digests through the RNA-DNA chimera, (2) -exo has difficulty digesting through G4 motifs in Glycine-KOH buffer, but not in Tris-HCl, (3) a 5'-OH end protects DNA from -exo digestion much more effectively than an RNA primer, and (4) RecJf does not digest through a 5' RNA end (or through G4 motifs), but can reduce background DNA levels. The panels shown are representative of these consistently observed outcomes, which are also consistent with the overall findings presented throughout the manuscript. We have added a statement to this effect in the Methods and to each relevant figure legend.

    In the Introduction section, the authors state: "...traditional SNS-seq enriches non-origin DNA that appears to originate from RNA:DNA hybrids due to contaminating RNA, while failing to enrich known replication origins above random expectation." Replace the term "contaminating RNA" with the term "residual cellular RNA", accurately conveying that this is endogenous RNA persisting through the protocol.

    We thank the reviewer raising this point. We corrected the sentence accordingly, which now reads: “[…]traditional SNS-seq enriches non-origin DNA that appears to originate from RNA:DNA hybrids due to residual cellular RNA, while failing to enrich known replication origins above random expectation.”

    The sentence describing that "the RNA-DNA chimera was not digested when 5' phosphorylation by PNK was performed prior to RNA degradation, rather than after the removal of the RNA primer by NaOH" introduces the PNK/NaOH order-of-operations logic before this concept has been explained to the reader, the significance of treatment order is not established until two sections later. A minimal fix would be adding a forward-referencing clause or relocating this result to a later paragraph.

    We agree that the result appeared to be out of place. We have relocated a slightly modified version of this sentence to a later paragraph, just before explaining the Figure 1C experimental logic. The relocated part now reads: “When characterizing λ-exo properties in vitro, we found that the RNA-DNA chimera was not digested when 5' phosphorylation by PNK was performed prior to RNA degradation, rather than after the removal of the RNA primer by NaOH (Supplementary Figure S4D); also confirming that a 5' phosphate is required for λ-exo activity. This result prompted us to further assess how hydrolyzing RNA primers to obtain 5'-OH DNA protects short nascent strand (SNS) DNA from λ-exo digestion compared to 5'-phosphorylated DNA, and to devise an experiment that varied the order of RNA hydrolysis (NaOH) and phosphorylation (T4 polynucleotide kinase, PNK) (Figure 1D).”

    In Figure 2D. (top panel), please state the amount of λ-exo used to perform this experiment in the corresponding figure legend.

    We believe there may be a small mix-up here: Figure 2D shows RecJf digestion throughout (top and bottom panels) and λ-exo was not used in either. However, we agree with the underlying point: the amount of RecJf used in the top panel (timecourse) was not stated in the legend. We have corrected this omission and, as also noted in a related minor comment from Reviewer 1, revised the Figure 2D legend overall for clarity, now stating explicitly that the top panel shows a timecourse of RecJf digestion (with hours indicated) and the bottom panel shows a titration of RecJf units (0–300 U) over a fixed 16 h digestion.

    In Figure 2D. where sonicated, heat-denatured yeast genomic DNA is used as a substrate, please explain how this DNA is prepared and if residual cellular RNA reannealing to complementary genomic sequences would/would not be expected in this prep. Discuss with respect to the RecJf experiments shown.

    The sonicated genomic DNA substrate in Figure 2D (bottom panel) was RNase-treated prior to RecJf digestion, which removes cellular RNA and precludes RNA reannealing to complementary genomic sequences in this specific preparation. We acknowledge that RNase treatment means this experiment does not fully replicate the RNA content of genomic DNA in an actual iSNS-seq reaction, where residual cellular RNA is present. However, we note that in the iSNS-seq workflow, RecJf digestion is used solely as an additional bulk genomic DNA clean-up step prior to λ-exonuclease digestion. Any genomic DNA fragments that resist RecJf digestion due to reannealed residual RNA would still be subject to the downstream PNK, NaOH, and λ-exonuclease steps that carry out the primary enrichment for origin-proximal SNS molecules. We have modifed the Results section to clarify this point: “[…] RecJf significantly reduced the levels of yeast genomic DNA that had been sonicated to the average SNS size range (0.5--2.0 kb), RNase-treated, and denatured by heat (Figure 2D bottom panel). Note that in the iSNS-seq workflow, RecJf digestion serves as an additional bulk gDNA clean-up step upstream of λ-exo digestion. Any genomic DNA species that escape RecJf digestion, including those potentially protected by reannealed residual RNA, remain subject to the subsequent PNK, NaOH (that hydrolyzes residual RNA), and λ-exo steps that establish specificity for SNS molecules.”

  2. Review Commons

    Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

    Learn more at Review Commons

    Referee #3

    Evidence, reproducibility and clarity

    Summary

    The authors of the paper "Improved short nascent strand sequencing (iSNS-seq) enhances DNA replication origin detection and reduces non-origin biases" investigate the biochemical basis of short nascent strand sequencing (SNS-seq), a widely used method for mapping DNA replication origins genome-wide. SNS-seq relies on λ-exonuclease digestion of genomic DNA, while retaining short nascent strands, which are protected by their 5' RNA primer. Despite its widespread use, SNS-seq has not been rigorously validated in a system with well-characterized replication origins and has often shown limited concordance with alternative origin-mapping approaches. Using a series of in vitro biochemical assays, the authors demonstrate that the key assumptions underlying SNS-seq are not well supported, G4 quadruplexes show strong protection against λ-exo digestion, while 5' RNA primers provide only partial protection. Based on these findings, the authors have developed an improved protocol (iSNS-seq) that achieves noticeably higher in vitro enrichment of nascent DNA than the traditional protocol. The authors then benchmark traditional SNS-seq (tSNS-seq) against iSNS-seq genome-wide in budding yeast (S. cerevisiae), a system with well-curated origin annotations. iSNS-seq appears to outperform tSNS-seq in sensitivity and precision against confirmed origins. The study also shows that tSNS-seq has a very poor performance for origin mapping in budding yeast and identifies a major source of false-positive signals in tSNS-seq, λ-exonuclease obstruction by RNA: DNA hybrids formed between residual cellular RNA and complementary genomic DNA during denaturation. The manuscript convincingly shows that tSNS-seq has important limitations for origin mapping and that iSNS-seq represents a meaningful improvement over the traditional protocol and resolves several of its biases. None-the-less, the signal-to-noise ratio of the improved method appears low and reproducible calling of peaks problematic. Furthermore, the data provided show that iSNS-seq under-performs relative to available alternative methods for nascent strand mapping.

    Major comments

    1. While the method developed (iSNS-seq) appears to improve aspects of the previously used method (tSNS-seq), poor signal-to-noise ratios and reproducibility are major concerns. The signal-to-noise ratio of iSNS-seq, as shown in Figure 3D, appears low. The authors should discuss what practical consequences this has for the applicability of the method, especially in species with less well-defined/efficient origins, and how it affects the reliability of peak calling, and overlap with known origins. Consistently, the overlap of called peaks between biological replicates of iSNS-seq is relatively low (Fig3B). This raises a concern for the usability of the method, the interpretation of the genome-wide data and the quality metrics provided. The authors should discuss these issues.
    2. Related to the above, please explain/discuss the rationale behind retaining all peaks, even the ones observed in only one of the biological replicates, when comparing called peaks to confirmed ORIs. In Figure 3B, clarify what are the percentages depicted.
    3. The title and abstract do not accurately convey the limitations of the improved method and must be rephrased.
    4. One of the manuscript's central mechanistic claims is that residual cellular RNA reanneals to genomic DNA, creating RNA:DNA hybrids that obstruct λ-exonuclease digestion. While the presented data from cells are consistent with this model, the manuscript lacks a direct biochemical demonstration of hybrid-mediated obstruction in a controlled system. The manuscript would benefit from in vitro studies corroborating this conclusion, using the in vitro system established.
    5. Currently, the RNA-dependence of the tRNA/snoRNA-associated obstruction signal is inferred from indirect evidence (UMAP clustering, feature overlap, peak shape) plus a re-analysis of an RNase-treated control from a previously published C. elegans dataset, not from a control generated within the authors' own yeast system, where the paper's primary genome-wide claims are made. Including such controls would be required to provide direct evidence for this central mechanistic claim.
    6. A substantial portion of the in vitro biochemical work (Figure 1B-C, E, Figure 2B-C with BKGmimic) is dedicated to establishing that G4 motifs cause significant λ-exo obstruction in glycine-KOH buffer, and that the Tris-HCl buffer switch resolves this. However, the genome-wide finding that G4 motif overlap is nearly identical between tSNS-seq and iSNS-seq (4-5% in both), with no significant differences in signal over G4 motifs in either dataset, requires clarification/discussion.
    7. None of the gel-based panels in Figure 1 (B, C, E) or Figure 2D state whether the result shown is from a single experiment or is representative of multiple independent experiments. Please state explicitly, for each gel-based panel, whether it represents a single experiment or is representative of repeated independent experiments (and if the latter, how many).

    Minor comments

    1. In the Introduction section, the authors state: "...traditional SNS-seq enriches non-origin DNA that appears to originate from RNA:DNA hybrids due to contaminating RNA, while failing to enrich known replication origins above random expectation." Replace the term "contaminating RNA" with the term "residual cellular RNA", accurately conveying that this is endogenous RNA persisting through the protocol.
    2. The sentence describing that "the RNA-DNA chimera was not digested when 5' phosphorylation by PNK was performed prior to RNA degradation, rather than after the removal of the RNA primer by NaOH" introduces the PNK/NaOH order-of-operations logic before this concept has been explained to the reader, the significance of treatment order is not established until two sections later. A minimal fix would be adding a forward-referencing clause or relocating this result to a later paragraph.
    3. In Figure 2D. (top panel), please state the amount of λ-exo used to perform this experiment in the corresponding figure legend.
    4. In Figure 2D. where sonicated, heat-denatured yeast genomic DNA is used as a substrate, please explain how this DNA is prepared and if residual cellular RNA reannealing to complementary genomic sequences would/would not be expected in this prep. Discuss with respect to the RecJf experiments shown.

    Significance

    This is a carefully executed and methodologically interesting study. Its principal strength is the use of budding yeast as a bench marking system for direct validation of SNS-seq peaks, the demonstration that there is a very small overlap between origins mapped by the traditional method and known origins in this well characterized system and the discovery of a specific, mechanistically coherent, and cross-species-reproducible artifact (RNA:DNA hybrid-mediated λ-exo obstruction) that provides an explanation for the observed low overlap of replication origin mapping by tSNSseq and alternative techniques. These findings are important for interpreting origin mapping studies in different organisms, including metazoa. In addition, the manuscript uses a well-controlled, quantitative in vitro biochemical assays to improve SNS-seq, and analysis in budding yeast shows that indeed the ability of this method to identify origins is improved over the traditional protocol. However, the low signal to noise ratio of the new method and rather low reproducibility between biological replicates, in a system (budding yeast) with well-defined and efficient origins, limits the applicability of the improved method for origin mapping. The improved method appears to be considerably underperforming against other available methods for mapping origins. Overall, this is a carefully executed study that would be of interest to an audience interested in origin mapping techniques in different species.

    Advance

    To our knowledge, this is the first study to benchmark SNS-seq against the well-characterized set of replication origins of yeast, despite the method being used for many years in different species. This is an important and overdue technical advance. Beyond the benchmarking itself, the identification of RNA: DNA hybrid-mediated obstruction as a systematic bias is a mechanistic finding that is notable. The identification of such biases is a conceptual contribution with implications reaching beyond this paper's own experiments, since it calls into question a standard control used throughout the published literature. The advance is primarily methodological and mechanistic, with secondary conceptual implications for ongoing debates over metazoan replication origins, since the paper raises legitimate doubt on whether SNS-seq-derived signals represent genuine origins.

  3. Review Commons

    Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

    Learn more at Review Commons

    Referee #2

    Evidence, reproducibility and clarity

    This is very nice paper, which combines meticulous in vitro and in vivo experiments to provide two new insights into a widely used approach - SNS-seq - to map DNA replication origins. The first insight is to detail a key limitation in traditional SNS-seq (tSNS), which leads to very substantial off-target mapping, resulting in predictions of origins that are not origins and, very likely, explaining the poor correspodance of this approach with orthogonal means of origin mapping in metazoans. Second, through carefully executed in vitro analysis, the authors provide a updated form of SNS-seq (iSNS) that more accurately predicts origins in yeast, where mutiple lines of evidence have provided a carefully curated list of origins; as result, the authors provide a new approach to origin mapping that should be widely applicable. I have one major issue and two minor issues with the work.

    Major issue.

    The authors' sugestion that tSNS miscalling of origins may be due to the presence of RNA:DNA hybrids is largely persuasive in theoretical terms but is, surpisingly, untested. The first issue that is unexamined in whether or not RNA:DNA hybrids would survive the two rounds of 95 degree denaturation that are central to both froms of SNS; can the auhtors comment or provide evidence? Second, it is important that the authors test their proposal (Fig.5) of RNA:DNA hybrids being the cause of at least some non-origin tSNS signal by mapping the correspondance of their SNS-seq data with the locations of RNA:DNA hybrids, since there are several forms of such mapping available for S. cervisiae (eg using S9.6 antiserum DRIP-seq, or RNaseH1 ChIP-seq).

    Minor comments.

    1. Fig.S7. Can the authors comments on the poor reproducibility between SNS-seq replicates? Though there isa four-fold increase in concordance between iSNS experments compared with tSNS, 80% of potetial origins are missed; can this be explained? In this regard, is the statment 'Both of the iSNS-seq samples showed signal enrichment around most of the confirmed origins' correct: while Fig.3C gives this impression, Fig.3B appears to diagree.
    2. Given the very nice data in Figs.1 and 2 showing the confounding effect of G4s on tSNS, can the authors comment on why G4s show little overlap with either SNS-seq mapping in Fig.4, and why they saw no increase in tSNS-seq or iSNS-seq signal over G4 motifs in Supplementary Figure S9B? Might this indicate that the in vitro work does not translate well to in vivo mapping?

    Referees cross-commenting

    From my reading of the three reviews, there are three overlapping issues that are brought up and should be addressed:

    1. A requirement to validate the interesting suggestion that RNA:DNA hybrids are a cause of tSNS-seq miscalling of origins; this should be a combination of in vitro tests and colocalisation of tSNS-seq signal and RNA:DNA signal in vivo using available datasets (eg DRIP-seq).
    2. An explanation and/or comment on the low reproducibility and low signal-to-noise ratio of iSNS-seq; specifically, although it improves on tSNS, does it provide accurate origin prediction?
    3. Comment on the lack of in vivo evidence, from available mapping data, for the confounding effect of G4s seen in the in vitro experiments.

    Significance

    This is a meticulously performed study with only some relatively minor issues needing resolved before publication, merely requiring new analysis and not new experiments. The work provides an important clarification on the shortcomings of a widely used strategy to map sites of DNA replication and, moreover, provides an updated and improved approach that should be widely adopted. This work will be of interest to researchers in the broad and fundamental field of DNA replication.

  4. Review Commons

    Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

    Learn more at Review Commons

    Referee #1

    Evidence, reproducibility and clarity

    Previous studies have shown that different origin mapping techniques could produce diverging results, introducing debate over whether metazoan origins are discrete or dispersed. Methods like SNS-seq and INI-seq suggested discrete initiation sites, whereas OK-seq, bubble-seq, and EdU/BrdU incorporated origin mapping suggested broad initiation zones with more dispersed initiation sites. Attempts to reconcile the diverged results presuppose that the discrete signals detected by SNS-seq represent bona fide replication intermediates. In this study, the authors set out to investigate the sensitivity and precision of traditional SNS-seq method, using budding yeast as the model organism, given that its replication origins are well known and defined with confidence. They addressed major known issues of lambda-exonuclease digestion in the traditional SNS-seq protocol, such as decreased efficiency of digesting through G4 DNA structure, and limited protection of RNA-primed DNA sequence, and showed that traditional SNS-seq failed to detect known origins. In addition, the authors test improved conditions (iSNS) suggested by the in vitro experiments and demonstrated an enhanced detection rate of known origins in yeast. Finally, the data show that the false positive signal derived from tSNS-Seq are enriched for tRNAs, snRNAs and highly transcribed regions forming RNA:DNA hybrids that overall likely represent regions with secondary structure formation potential that are protected from traditional lambda exonuclease digestion conditions.
    Overall, the study is timely and the data presented are convincing and should be published given that traditional SNS-Seq has been used extensively in the past to map the origin landscape of multiple organisms. Therefore, it will be important to reconcile these results in the light of the findings of this paper.

    Major comments:

    • Are the claims and the conclusions supported by the data or do they require additional experiments or analyses to support them?

    The claims and the conclusion are supported by the data.

    • Please request additional experiments only if they are essential for the conclusions. Alternatively, ask the authors to qualify their claims as preliminary or speculative, or to remove them altogether.

    There are no additional experiments requested.

    • If you have constructive further reaching suggestions that could significantly improve the study but would open new lines of investigations, please label them as "OPTIONAL".

    There are a few further reaching suggestions listed in the minor comments section.

    • Are the suggested experiments realistic in terms of time and resources? It would help if you could add an estimated time investment for substantial experiments.
    • Are the data and the methods presented in such a way that they can be reproduced?

    Yes. They are.

    • Are the experiments adequately replicated and statistical analysis adequate?

    Yes. They are.

    Minor comments:

    • Specific experimental issues that are easily addressable.

    Additional in vitro experiment using artificial R-loop structures as substrate for lambda-exonuclease could prove the efficacy of the nuclease to digest through R-loop structures.

    • Are prior studies referenced appropriately?

    Yes.

    • Are the text and figures clear and accurate?

    Yes.

    • Do you have suggestions that would help the authors improve the presentation of their data and conclusions?

    Yes. They are the followings:

    In Fig. 1C, there seems to be a minor increase of lambda-exo activity on R10D30 substrate in the Tris-HCI buffer compared to Glycine-KOH buffer, maybe a quantification using the loading control (D20) would help. In Fig. 2D, lower panel, the digestion time of RecJf at different units were not indicated. In Fig. 3B, since SNS-seq detect fired origins, an overlap of iSNS rep 1 and iSNS rep 2 peaks with confirmed origins from OriDB that are also early and efficient origins may yield a higher percentage of overlap. In Fig. 3D, in addition to showing the tSNS and iSNS data on ChrIV with indicated known origins, showing at the same time the corresponding FORK-seq, OK-seq, and ORC ChIP signal will give a more comprehensive view of the performance of different origin mapping methods. In Fig. 5F, in supporting the model, a comparison of tSNS-seq peak with known budding yeast DRIP-seq datasets would help to see if the tSNS method indeed enrich for RNA:DNA hybrid signals.

    Significance

    This study investigates the sensitivity and precision of traditional SNS-seq methods, questioning the interpretation of existing SNS-seq datasets, which help to reconcile discrepancies among different origin mapping datasets. It also provides a first improved SNS-seq methodology that is tested against bona fide replication origins identified in budding yeast, allowing subsequent improvement of the SNS-seq methodology.

    The following aspects are important:

    • General assessment: provide a summary of the strengths and limitations of the study. What are the strongest and most important aspects? What aspects of the study should be improved or could be developed? The strengths in this study are the followings: It provided the first, and much improved SNS-seq dataset in budding yeast. It provided the first-ever assessment of traditional SNS-seq method using bona fide replication origins identified in budding yeast as quality control check, while at the same time established an improved SNS-seq method, again verified by quality control check using budding yeast origins. It demonstrated that traditional SNS-seq dataset may enrich a much lower percentage of replication origins, while having more false positive origins which are potential RNA:DNA hybrids. The aspect that could be improved or developed: More secondary DNA structures could be tested for lambda-exonuclease activity on them.
    • Advance: compare the study to the closest related results in the literature or highlight results reported for the first time to your knowledge; does the study extend the knowledge in the field and in which way? Describe the nature of the advance and the resulting insights (for example: conceptual, technical, clinical, mechanistic, functional,...). The study provides important insight on how to interpret the existing SNS-seq datasets and reconcile discrepancies among them. It provides immediate technical advance in SNS-seq sample preparation which could translate further to conceptual and mechanistic advance.
    • Audience: describe the type of audience ("specialized", "broad", "basic research", "translational/clinical", etc...) that will be interested or influenced by this research; how will this research be used by others; will it be of interest beyond the specific field? The audience will be basic researchers who work on DNA replication, particularly origin mapping.
    • Please define your field of expertise with a few keywords to help the authors contextualize your point of view. Indicate if there are any parts of the paper that you do not have sufficient expertise to evaluate. Replication origin regulation, RNA:DNA hybrids and Transcription-Replication Conflicts. I feel confident in evaluating all aspects of the paper.
  5. Version published to 10.64898/2025.12.22.696044 on bioRxiv