FORK-seq: replication landscape of the Saccharomyces cerevisiae genome by nanopore sequencing

Magali Hennion, Jean-Michel Arbona, Laurent Lacroix, Corinne Cruaud, Bertrand Theulot, Benoît Le Tallec, Florence Proux, Xia Wu, Elizaveta Novikova, Stefan Engelen, Arnaud Lemainque, Benjamin Audit, Olivier Hyrien

Preprint posted on 10 April 2020

Article now published in Genome Biology at

Following the FORKseq’s in the road. Paving the way for high-resolution mapping of DNA replication.

Selected by Jennifer Ann Black

Categories: genomics


To duplicate their genome on time, many Eukaryotes initiate DNA replication at multiple different sites on their chromosomes, known as origins of replication. When replication is initiated, the movement of the machinery required for replication forms structures known as a replication forks, which move bi-directionally from the origin. When they converge together, DNA replication can terminate (1). Most genome-wide approaches to study DNA replication are performed on cell populations. This means we lose resolution and important information on DNA replication events within individual cells. Here, Hennion and colleagues have developed a new protocol for examining single molecule DNA replication using the recently developed MinION nanopore sequencing technology from Oxford Nanopore Technologies (2). This technology can allow the sequencing of native single DNA strand in real time. When each DNA base enters and travels through the ‘nanopore’ channel in the flow cell, changes in the surrounding electrical field occur and corresponding electrical signature can be matched to a DNA base. Using the yeast Saccharomyces cerevisiae (S. cerevisiae) as a model, cells were labelled with bromodeoxyuridine (BrdU), an analogue of the DNA base thymidine (3), to track DNA replication on single strands of DNA by analysing location of BrdU incorporation and orientation of BrdU-abundance gradients. They term this approach ‘FORKseq’. Their data reveals a high-resolution and genome-wide picture of DNA replication events in yeast.

FORKseq takes a similar approach to the recently published D-NAscent method (4) which allows the examination of single molecule DNA replication events using BrdU incorporation and nanopore sequencing. The differences between the two techniques lie predominantly in their computational pipelines and their BrdU labelling strategies.


Figure 1. (1) The ‘Nanopore’ embedded in the membrane (blue). An ion current is flowed across the membrane. When the DNA library is applied to the membrane containing the Nanopore, the adaptors can bind to ‘tethers’ on the membrane surface. (2) A motor protein (green) helps to guide the DNA molecule through the pore. (c) As bases enter the Nanopore, the disrupted the electrical current can be read as patterns unique to each base allowing base calling to occur. Adapted from Leggett and Clark, 2017.

Key Findings

Nanopore sequencing technology can distinguishing between thymidine and its base analogue BrdU in a native single strand of DNA.

By sequencing primer extension products containing the presence or absence of BrdU incorporation, a difference between the electrical charge of BrdU and thymidine could be detected by the MinION sequencer using custom python scripts (Fig. 1). Next, by implementing two different machine learning algorithms (CNN and TM), the authors assessed abundance of BrdU incorporation within 100 bp windows, with their bioinformatics approach largely agreeing with average BrdU abundance detected by their mass spectrometry analysis confirming their analysis approach is viable.

FORKseq can determine replication fork direction

To look for replication fork progression, yeast cells were cultivated (pulsed) in BrdU and thymidine conditions for 2 mins then ‘chased’ using thymidine. When the resulting DNA was sequenced, the authors observed regions in which sharp transitions from low BrdU abundance to high BrdU abundance occurred (Fig.2).

Using their machine learning approach, they found these transitions revealed the direction of the replication fork, which they confirmed using a previously published approach known as OK-seq (4,5). They also show they can resolve these sites to within ~ 200 bps of the start and initiation site giving a high-resolution picture of replication events.

 FORKseq can map individual replication initiation and termination sites in the yeast genome

Replication initiation and termination sites have been previously mapped throughout the yeast genome. Using FORKseq data, the authors largely confirm the positionings of these sites and additionally identified regions of initiation (9 %) and termination (18 %) that were likely missed previously due to a lack of sensitivity for sites of infrequent usage. Here, they reveal that these additional initiation sites are new sites of DNA replication initiation out with the known origin of replication supporting the model that yeast DNA replication can commence from canonical origins (91%) and non-canonical origin sites (9%). The additional termination sites were located within regions previously only associated with DNA replication initiation (Fig. 4-7).

FORKseq is adaptable for use in other Eukaryotic systems

The FORKseq approach relies on BrdU incorporation as a measure of DNA synthesis. BrdU has been successfully used widely in other Eukaryotic organisms to study DNA replication (3) and established published protocols are available for reference in the usage of this base analogue. Though the authors do acknowledge that organisms with larger genome could likely prove challenging to analyse due to restrictions on the throughput achieved by the MinION sequencer.


What I liked about this preprint:

DNA replication often studied in a population of cells as examining single molecules of DNA can be challenging meaning we are likely more familiar with a lower resolution representative of this process. I really enjoyed reading this article by Hennion and colleagues because of their approach to tackle this lack of resolution. They present their findings clearly, acknowledge their protocol in the context of the field and other techniques developed for similar purposes and employ clever strategies to maximise their datasets.


Questions for the authors:

  1. Your approach could be used to address how replication dynamics are affected by the presence of exogenous replication stress, such as the addition of chemicals like Hydroxyurea. Do you foresee any challenges or limitations for FORKseq in the analysis of such data?
  2. As mentioned, nanopore sequencing can be challenging to ensure reproducibility between experiments. What did you find most challenging about the sample preparation and/or the data analysis to minimise variation between experiments?


(1) Fragkos, M., Ganier, O., Coulombe, P & Mechali, M. DNA replication origin activation in space and time. Nature Reviews Molecular Cell Biology, 16 (2015)


(3) Cavanagh, B.L., Walker, T., Norazit, A. & Meedeniva, A.C. Thymidine analogues for tracking DNA synthesis. Molecules, 16(9) (2011).

(4) Muller, C.A., Boemo, M.A., Spingardi, P., Kessler, B.M., Kriaucionis, S., Simpson, J.T. & Nieduszynski, C.A. Capturing the dynamics of genome replication on individual ultra-long nanopore sequence reads. Nature Methods, 16 (2019).

(5) McGuffee, S.R., Smith, D.J. & Whitehouse, I. Quantitative, genome-wide analysis of Eukaryotic replication initiation and termination. Molecular Cell, 50(1) (2013).

(6) Petryk N, Kahli M, d’Aubenton-Carafa Y, Jaszczyszyn Y, Shen Y, Sylvain M, Thermes C, Chen CL, Hyrien O. Replication landscape of the human genome. Nature Comm., 7, 10208 (2016).

Figure reference: 

Leggett, R.M & Clark, M.D. A world of opportunities with nanopore sequencing. J Exp. Botany., 68, 20 (2017).

Tags: dna replication, nanopore, yeast

Posted on: 27 April 2020


Read preprint (No Ratings Yet)

Author's response

The author team shared

Response to question 1:

We might encounter different types of challenges depending on the nature and extent of inflicted replication stress. For example, a low dose of hydroxyurea may simply slow down the forks while higher doses may stop fork progression.
Although fork speed analysis was not addressed in the preprint, we are able to extract single fork speeds from FORK-seq signals and this should be directly applicable to moderate fork slowing conditions.
Our pulse-chase labelling protocol ensures that forks move away from sites of BrdU incorporation so that during ligation of sequencing adapters, covalent bonds can be made between the BrdU-labelled strand and the adaptor. In the absence of thymidine chase, ligation of “bubble-shaped” fragments would result in adapter ligation to the two parental strands would predominantly occur so the BrdU strands in a replication bubble configuration would be excluded from the nanopore during sequencing. Therefore, to study treatments that prevent fork progression away from BrdU incorporation tracts, we may need to modify the preparation of sequencing libraries.
Finally, forks that have arrested progression before the BrdU pulse may not incorporate BrdU at all and may therefore escape direct detection. Consecutive pulses of two distinct thymidine analogs may address some of these issues. Multiple analogs can be detected, but resolving one from another (e.g. iodo- from chlorodeoxyuridine) on each molecule poses additional challenges.

Response to question 2:

Sample preparation is easily reproducible as it consists of standard pulse-chase experiments followed by DNA extraction and sequencing adapter ligation. The number of detected replication events is highly dependent on the length of the DNA molecules, so it is important to use protocols that minimize DNA breakage. Nowadays many labs are able to routinely obtain very long nanopore reads. What we found the most challenging was to implement robust signal analysis tools to detect the variable signal modifications at incorporated BrdU bases and obtain precise measurements of BrdU content over short consecutive sequence windows. We found that it was much more robust for machine learning strategy to estimate the proportion of BrdU incorporation within 100bp than to call BrdU incorporation at every T site. This might be a generally good strategy when assessing deviations from canonical DNA (methylation, non-standard nucleotides, lesions, … ) using nanopore sequencing. We also found that analysing electic current increments at translocation steps rather than absolute current values was important to eliminate low frequencies in signal drift.

Have your say

Your email address will not be published.

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Sign up to customise the site to your preferences and to receive alerts

Register here