Transcription promotes the restoration of chromatin following DNA replication

  1. Susanne Bandau
  2. Vanesa Alvarez
  3. Hao Jiang
  4. Sarah Graff
  5. Ramasubramanian Sundaramoorthy
  6. Matt Toman
  7. Tom Owen-Hughes
  8. Simone Sidoli
  9. Angus Lamond
  10. Constance Alabert

Reviewed by

Read the full article See related articles

Listed in

Log in to save this article

Abstract

DNA replication results in the transient eviction of nucleosomes, RNAPII and transcription regulators. How chromatin organization is duplicated on the two daughter strands is a central question in epigenetics. In mammals, transcription restarts on newly replicated DNA within a couple of hours, promoting chromatin accessibility. However, the role of transcription in the restoration of other chromatin determinants following DNA replication remains unclear. Here we have monitored protein re-association to newly replicated DNA upon inhibition of transcription using iPOND coupled to quantitative mass spectrometry. We show that nucleosome assembly and the re-establishment of most histone modifications are uncoupled from transcription restart. However, upon transcription inhibition, the re-association of many proteins was altered, including ATP-dependent remodellers, transcription regulators, the histone variant H2A.Z, histone modifiers as well as the restoration of H3.3K36me2. Finally, transcription also provoked the recruitment of several DNA repair proteins, revealing that transcription promotes chromatin reestablishment post-replication but is also a potential source of genotoxic stress.

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

    'The authors do not wish to provide a response at this time.'

    Our point-by-point response contains figures that we could not manage to upload in this text box.

  5. Review Commons

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

    Learn more at Review Commons

    Reply to the reviewers

    The authors do not wish to provide a response at this time.

  6. 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

    In this study, the authors used metabolic labeling of newly-replicated or nascent chromatin followed by quantitative Mass spectrometry (iPOND-MS) to characterize protein composition of nascent chromatin at time points after DNA replication: immediately after a short pulse of EdU labeling (nascent), and after 1 and 2 hours of Thymidine chase (maturing chromatin). The iPOND method was established before but in the current manuscript the authors combined this with inhibiting RNA Pol II transcription at distinct stages to determine the effects on transcription and RNA Pol II cycle on chromatin protein dynamics at the wake of DNA replication. The inhibitors they used are Triptolide, which blocks transcription initiation and induces a proteasomal degradation of all chromatin bound RNA PolII, and DRB, which blocks transcription elongation causing an enrichment of paused RNA PolII. The authors compared the relative enrichment of ~1200 proteins on nascent and maturing chromatin and the effects on transcription inhibition on these proteins.

    The authors found that RNA PolII does not affect the loading or retention of most histones on nascent chromatin except for the histone variant H2A.Z, which requires RNA PolII loading. However, DRB treatment (no elongation) resulted in stabilization of all histones (which the authors do not seem to catch on). Interesting, unlike the histone, both replication-coupled and -independent histone chaperons seem to be enriched immediately behind the fork and are affected by RNA PolII to different extents. They next look at ATP-dependent remodelers and find that most remodeler families are facilitated by RNA PolII loading, while elongation affects some remodeler families and not others. They see the same trend looking at a wide variety of transcription factors. Interestingly, while RNA PolII loading is required for the establishment of some histone post translational modifications (H3K36me3), some others such as H3K9me3 and H4K20me2 are negatively affected. Finally, the authors find that RNA PolII elongation promotes binding of several DNA repair proteins, and speculate that this is because of DNA damage from replication-transcription conflict.

    My main concern about this manuscript is that the relative enrichment of most factors show variability across the time points, which make the interpretation of the data difficult. This becomes more concerning when we look at protein complexes such as the ATP-dependent remodelers. Subunits of the same complex which are expected to bind together show different patterns of enrichment. This raises the concern as to how data was normalized. Furthermore, how do the replicates compare to each other? The others selected ~1200 proteins which were enriched in all three replicates, but how does their relative enrichment compare in the replicates? The authors need to show some kind of comparison across replicates to confirm that the differential relative enrichments are real and biologically meaningful.

    Also, the TF data is very descriptive. Insightful analysis of similarities/differences between types of TFs would be interesting.

    Minor comment: The formaldehyde cross linking used in iPOND makes it difficult to interpret/distinguish what is actually chromatin bound versus what is enriched due to protein-protein cross linking. The authors should highlight that in the limitations section.

    Referees cross-commenting

    I agree with most of Reviewer 1's comments about the lack of proper controls and normalization, which make the interpretations difficult. Particularly all of the controls mentioned under point 1 should not be difficult to perform, and if included, would strengthen the study and the manuscript.
    Reviewer 1 makes an important point about normalization, which I totally agree with. Ideally, a spike-in approach would help obtain a much more quantitative and reliable understanding of differential protein enrichment. However, repeating all iPOND experiments with spike-in might be a big ask. What the authors could do at minimum is show how replicates compare with each other. It looks like they pooled all three replicates for analysis, but comparing relative enrichment of all 1257 proteins across replicates would help. The point about delayed histone occupancy is a critical one and difficult to rationalize. To note, histone chaperons are enriched on nascent, but histones are not. Besides, in the current way that the data is analyzed and presented, there are a lot of fluctuations in protein enrichment across the 1-2 hour timepoints of chromatin maturation, which would be very interesting if real. For e.g., Fig. 1I, Triptolide treatment, most of the cluster I and cluster II proteins show medium-high enrichment on nascent, depleted in 1h, but recover in 2h. If the binding/recruitment of these proteins on newly-replicated chromatin is RNA Pol II dependent, why would they come back after 1h? If this real, this would be very interesting. There are several additional examples of problems with quantification/normalization. As for SWI/SNF subunits, both SMARCA4 and SMARCC1 are core subunits and based on several thorough biochemical studies, cannot be expected to bind separately. However, they show different kinetics in DMSO as well as TPL and DRB.

    Another problem of the assay is that it shows genome-wide average. As Reviewer 1 rightly pointed out, transcription inhibition could disproportionally affect chromatin maturation kinetics in different genomic regions. Perhaps it would be interesting to analyze sets of genomic regions separately, such as highly transcribed and lowly transcribed genes. This might be achieved by adding a purification step using pools of DNA sequence probes before or after the streptavidin enrichment.

    Additional comment: The formaldehyde cross linking used in iPOND makes it difficult to interpret/distinguish what is actually chromatin bound versus what is enriched due to protein-protein cross linking. The authors should highlight that in the limitations section.

    On a positive note, it is a very important and timely study and the manuscript has a lot to consider. Addition of proper controls and normalization/analysis of replicates will make it stronger

    Significance

    Overall, it is a very important and timely study, and the manuscript has a lot to consider. There are several recent papers on the kinetics of chromatin maturation behind the replication fork, and this study adds a very important dimension to this ongoing investigation, and will be of interest to a broad readership in the chromatin and transcription field.

  7. 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

    In this manuscript, the authors characterized the re-establishment of chromatin after DNA replication in fibroblasts using iPOND-MS. By using a short pulse of EdU, followed by different length of thymidine-chase, the authors compare the proteome at nascent DNA (just after the EdU pulse) with the proteome on re-established chromatin (1h and 2h post EdU pulse). Moreover, by using two different transcription inhibitors, they investigate the implication of active transcription elongation and of RNAPII binding itself on the reestablishment of chromatin. They show that different transcription factors bind to newly replicated DNA with different kinetics and are affected differentially by transcription inhibition. They also show that upon transcription inhibition by DRB, certain DNA damage repair proteins are depleted, implicating transcription in the recruitment of these factors at nascent DNA. Chromatin remodelers were shown to be enriched on nascent DNA, but triptolide-transcription inhibition reduced their enrichment, implicating RNAPII in the reestablishment of chromatin structure and of steady-state chromatin accessibility. Lastly, the authors show that histone incorporation and histone modification restoration on nascent DNA is mostly uncoupled from transcription with the exceptions of H3.3K36me2 (transcription inhibition by triptolide or DRB drastically reduces restoration) and H4K16ac (DRB treatment increases its incorporation in nascent DNA).

    Overall, the results and the analysis of the datasets appear robust and well executed. Nevertheless, the work provided by the authors feels mainly descriptive and does not provide further mechanistic insights beyond the current state of the art. Some follow-up experiments to study the functional impact on the different enrichment patterns on nascent DNA or the function of the dependency on RNAPII for the reestablishment of steady-state enrichment on chromatin of some factors would have greatly increased the scientific impact of the manuscript. Nevertheless, the proteome of nascent DNA, its kinetic, and the effect of transcription inhibitors will provide interesting information and a useful resource for research groups in the DNA replication, chromatin, epigenetic, and DNA damage repair fields. Thus, in conclusion, I would recommend this manuscript to be published in its current state in a lower tier journal such as MBoC or PLOS ONE journals. If the authors can provide additional mechanistic insights by addressing at least a few of the specific points listed below, I think it would become a stronger candidate for a journal with higher impact.

    Major comments:

    1. At p.7, the authors state: "Altogether, this analysis further confirms that RNAPII's binding and elongation on newly replicated chromatin are a source of genotoxic stress, and identifies dedicated repair factors handling transcription replication conflicts.". I don't think that depleted DNA repair proteins from nascent chromatin upon DRB treatment is enough to claim that the analysis confirms that transcription on nascent DNA is a source of stress. Another possibility could be that transcription helps handling prior DNA damage on nascent DNA without causing the damage. A useful experiment to clarify this point would be the direct quantification of DNA damage markers on nascent chromatin (e.g yH2AX-EdU colocalization quantification by immunofluorescence). Has the yH2AX variant been detected in the iPOND MS dataset? Another possible follow-up experiment could be to detect direct physical DNA damage on nascent DNA for example by using a TUNEL assay or similar DSB mapping method. Can the DNA damage be prevented by DRB or TRP addition?
    2. Figure 1B-E: Can the authors also show quantifications of EU, RNAPII and EdU at the 1h and 2h timepoints after the chase?
    3. The authors state in p.7 that "The other proteins are DNA repair proteins involved in fork quality control and HR as well as transcription replication conflicts (Berti et al., 2020).". I think this gives rise to the question if the effect of DRB treatments on the enrichment of certain proteins at nascent DNA is due to the inhibition of transcription elongation inhibition on nascent DNA or in front of replication forks, affecting the enrichment of proteins implicated in handling transcription-replication conflicts in front of the fork and not on nascent DNA itself. The authors should address the possibility that some of the proteins enriched in the iPOND-MS datasets could be there because they are enriched in front of the replisome instead of on the nascent DNA.
    4. On this topic, transcription inhibition is performed for two hours prior to the EdU pulse and iPOND-MS procedures. For the DRB treatment, I would expect RNAPII to be paused/stalled prior to the passage of the replication fork that will replicate the analyzed EdU-labelled nascent DNA. This would mean that replication forks during the EdU pulse will encounter paused/stalled RNAPII, generating potential problems. Such interference would most probably lead to chromatin removal of RNAPII from the chromatin. Surprisingly, the authors show enrichment of RNAPII at nascent DNA. How can the author differentiate from accumulation of RNAPII in front of the fork, leading to purification by iPOND, and RNAPII on nascent DNA. Also, if the accumulation of RNAPII is on the nascent DNA, do the authors suggests that RNAPII gets loaded more on nascent DNA while under DRB inhibition or that stalled RNAPII are mainly by-passed by replication forks, leading to their enrichment on nascent DNA?
    5. At p.14, the authors state: "Because they share the same DNA template, transcription is known to challenge replisome progression at high frequency, from RNAPII constituting a roadblock to progressing replisomes, to generate RNA:DNA hybrids (Berti et al., 2020). It is therefore remarkable that behind replisomes, only a handful of DNA repair factors appear to be involved in response to RNAPII binding and elongation.". How does the fact that transcription represents a roadblock in front of the forks makes it remarkable that only a handful of DNA damage repair pathways are involved behind the fork (where they are not a roadblock to any replisome)?
    6. At p.11, the authors states: "As di and tri-methylations require several hours to be re-established following DNA replication (Alabert et al., 2015; Reveron-Gomez et al., 2018), 11 minutes after the passage of the fork, such increase most probably reflects an increase of H4K20me2 and H3K9me3 on recycled parental histones.". Can the authors extend their interpretation of this result? Do the authors think that DRB treatments increase methylation of histones in G1, prior to replication, or specifically in front of the fork (due to conflicts? DNA damage?), and that those methylated histones get recycled on nascent DNA?
    7. Figure 4: The authors perform the histone PTM analysis under 0h (nascent chromatin) versus 2h (re-established chromatin) timepoints. It would have been insightful to also include a 1h timepoint in this experiment. There appear to be some trends/changes but they do not show statistical significance (e.g. H4K5K12ac or H3K14ac). It might be useful to increase the number of biological replicates (including the 1h timepoint) here, which could improve the confidence in the results and/or discover additional transcription-dependent changes of histone PTM restoration.

    Minor comments:

    1. Fig3I: It would be nice to show a TF from the "Restored within 11 min" category as a comparison point.
    2. In page 14, th authors state: "However, we did not detect significant signs of DNA damage in DRB treated cells (Fig. 2A, 2B).". Which signs the authors looked at?
    3. In the iPOND experiment, which size of DNA fragments is achieved?
    4. At p.14, the authors state: "Because they share the same DNA template, transcription is known to challenge replisome progression at high frequency, from RNAPII constituting a roadblock to progressing replisomes, to generate RNA:DNA hybrids (Berti et al., 2020)." The paper has not addressed the role of RNA:DNA hybrids in these processes.
    5. Fig3D: Is there enough datapoints to state a conclusion?
    6. S1A: mistakes in the x axis labels ("no EU" in a EdU quantification graph, "no EdU" in a EU quantification graph).
    7. S1F is not sufficiently described in the legend. It took me some time and additional efforts to understand what the right panel of S1F was showing.
    8. S2E-F: are the axis wrong? Is it supposed to be Nascent when its comparing total extracts?
    9. A lot of graphs have non-precise axis labels that needs reading of the manuscript and/or legends to understand. For example: 1K-L (distribution, %), 2L (% of the max), 3B-C-D log2(Nascent/2h), 3G IP/Input, 4C (Inhibitor treatment/DMSO (%)), S2E-F (TPL/DRB Nascent/ DMSO Nascent), S3A (IP/Input), S4A (No Y axis label).
    10. FigS4: Assignment of colors in bar graphs of C-J to treatments is not shown. Heatmaps in H and K do not show if these are 0 or 2h. The heatmap in H shows H3 modification and the heatmap in K shows modification in H3.3 but the labels of the modification in K (except the first one) are the names of the modifications of H3, not H3.3. In the legend, GAPDH is written GABDH.

    Significance

    In this manuscript, the authors characterized the re-establishment of chromatin after DNA replication in fibroblasts using iPOND-MS. As mentioned above, the work provided by the authors feels mainly descriptive and incremental and therefore does not provide further mechanistic insights beyond the current state of the art. Some follow-up experiments to study the functional impact on the different enrichment patterns on nascent DNA or the function of the dependency on RNAPII for the reestablishment of steady-state enrichment on chromatin of some factors would have greatly increased the scientific impact of the manuscript. Nevertheless, the proteome of nascent DNA, its kinetic, and the effect of transcription inhibitors will provide interesting information and a useful resource for research groups in the DNA replication, chromatin, epigenetic, and DNA damage repair fields. Thus, in conclusion, I would recommend this manuscript to be published in its current state in a lower tier journal. If the authors can provide additional mechanistic insights by addressing at least a few of the specific points, I think it would become a stronger candidate for a journal with higher impact.

  8. 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

    The goal was to characterize the changes in the composition of the proteome associated with replicated DNA in conditions of genome-wide inhibition of transcription initiation or transcription elongation. They use iPOND, a MS based technique that identifies proteins specifically associated with replicated DNA labeled with EdU. They use non-synchronized foetal human lung fibroblasts and examine time points immediately after replication (after the 11 min EdU pulse) and 1 and 2 hrs after the Thymidine "chase" later when chromatin has "matured", to assess how the inhibition of transcription influences chromatin maturation and the binding of chromatin associated proteins to replicated DNA.

    The question is pertinent and is in line with the long-standing interest of the group in chromatin replication dynamics. They conclude that 1. RNAPII loading is necessary for the binding of some TFs, chromatin remodelers and DNA repair factors and 2. RNAPII elongation is needed for H2A.Z incorporation , H3K36me2 restoration and DNA repair factor binding. Transcription is on the other hand not needed for nucleosome assembly, histone acetylation and H3K9me3, H3K27me3 and H4K20me2 incorporation or restoration.

    There are two main issues that make the interpretation of their results very difficult and make me question their conclusions:

    1. They don't provide sufficient evidence that the treatments with TPL and DRB do not interfere with replication. The distributions of EdU intensity per EDU+ cell after treatment in Figures 1D-E and S1A are not sufficient. It is not clear why EdU incorporation is so heterogeneous in the cell population (the range of intensities goes from near 0 to 50000!), which makes me wonder if the DMSO treatment also has an effect on replication. I don't think this heterogeneity can simply be explained by the fact that the the cell population is asynchronous. They need to show a -DMSO control as well. Besides since they are only using a positive EdU signal as their criteria for replicating cells, they cannot rule out that some of the EdU signal is coming from DNA repair after replication and depending on how deleterious DMSO/TPL/DRB are to replication the fraction of cells that undergo DNA repair might be significant. More importantly, they need to show that the various treatments don't interfere with the replication program, especially since replication is coupled with new nucleosome assembly and the transcription of replication dependent histone variants is induced during S-phase. Transcription inhibition could disproportionally affect the replication of some parts of the genome more than others and since there is no evidence to the contrary the differences that they observe between the TPL/DRM treated and DMSO treated proteomes bound to replicated DNA could just be because they were isolated from different genomic loci. I am also not convinced that they are able to stop EdU incorporation after 11min with the addition of only equimolar amounts of Thymidine (20µM EDU and 20µM Thymidine). Equimolar amounts of Thymidine are not sufficient to stop EdU incorporation rapidly. They need to show the kinetics of EdU incorporation in synchronized cells +/- Thymidine.
      Without these controls it is impossible to draw any meaningful conclusions from the iPOND data.
    2. The normalization of iPOND and total protein MS data is problematic. It seems that each time point from each treatment was first normalized internally to the median of all protein levels in each dataset and then the relative abundances of each protein were normalized to 100% over all treatments and time points. Internal normalization makes it impossible to directly compare time points and treatments between each other. If the enrichment of a protein goes down from one time point to the next it doesn't mean that there is less of that protein on replicated DNA in absolute terms, it just means that there is less of that protein relative to the median of the whole set of proteins at that time point. Their claim that they are comparing iPOND enrichments to total protein abundance is misleading since the data from total protein extracts was also internally normalized so they are comparing relative enrichments in iPOND data to relative enrichments in total cell extracts, which unsurprisingly do not correlate. It is impossible to make any meaningful conclusions about proteome dynamics using this kind of analysis. They should have used external normalization with a "spiked in" protein to be able to directly compare time points and treatments.
      Such as it is right now, their analysis produces some puzzling conclusions that I suspect will turn out to be artefacts of their normalization procedure. It is not clear for example why the appearance of histones on replicated DNA would be delayed as they claim: in yeast nucleosomes (new and old recycled ones) are assembled on replicated DNA within minutes of the passage of the replication fork, I don't see why this would not be the case in human cells since the replication machinery is essentially the same in humans and yeast. It is also puzzling why RNAP2 is enriched in the nascent and 1hr time points but then becomes depleted in the 2hr time point in the DRB treatment since global RNAPII levels don't change in the DRB treatment compared to DMSO (Figure 1C). All the conclusions for PTM restoration/incorporation are essentially meaningless: internal normalization makes it impossible to detect whether PTM levels double at the 2hr time point compared to the Nascent time point in the DMSO treatment, as would be expected for all examined PTMs except for H4K5K12Kac which are marks of new histones. Right now, relative PTM levels are all over the place: only histone acetylations seem to increase, while H3K9me3 and H3K27me3 don't change even though they should also double since heterochromatin should also be restored on both sister chromatids. They will only be able to accurately assess the impact of transcription inhibition on PTM restoration when they are able to reliably measure the rate of increase of PTM levels during chromatin maturation.

    Referees cross-commenting

    On reviewer's 2 comment on significance:
    I think a thorough descriptive analysis of a biological process is extremely valuable and unlike my colleague, I think these types of studies need to be published in high impact journals with a broad readership. Biological processes need to be described first as completely as possible before we can propose meaningful models on how they function and identify the molecular mechanisms that execute and regulate them. As my colleague is surely aware, thorough descriptive studies of any poorly characterized biological process take years (i.e. at least one grant cycle) and comprehensive follow up mechanistic studies can take even longer than the initial descriptive study and can only be done during the following grant cycle, if the authors were lucky enough to obtain funding. Funding agencies however are more likely to award grants to perform these follow up mechanistic studies if the authors (especially if they are junior PIs) have published in higher impact journals in their previous grant cycle. The kind of thinking exhibited by reviewer 2 disproportionately disadvantages junior PIs that work on understudied biological processes. It is a disservice to scientific progress to dismiss excellent descriptive studies and "downgrade" them to lower impact journals where they will be unfairly labeled as a "work of lesser importance". This kind of thinking is also a disservice to the lower impact journals that often publish works whose quality is comparable to articles published in high impact journals. I value more any comprehensive description of a biological process over what most of the time passes for mechanistic insight that is deemed worthy of publication in a high impact journal i.e. a hastily analyzed phenotype of, more often than not, one single mutant tacked on at the end of a descriptive study. This one mutant phenotype then forms the basis of a somewhat "slapdash" model that is often proven wrong by subsequent publications and that the authors would have probably dismissed themselves had they been given more time to develop and test their model in a follow up publication.
    I do not think the main issue with the present study is its descriptive nature. As I said in my review, the main issues are technical: the lack of external normalization of MS data and insufficient evidence of the impact of transcription inhibitors on replication dynamics. The study should not be published in any journal (high or low impact) before those issues are resolved.

    on reviewer 2's remarque 4. in major comments:
    iPOND identifies proteins bound to 100-300bp fragments labeled with EdU (i.e. after replication or DNA repair). It is by definition identifying proteins bound to chromatin behind the fork, so I don't think that the isolation of RNAPolII bound in from of the fork is a major issue

    Significance

    I am not convinced by their conclusions and I cannot recommend that the the study be published at this stage due to normalization issues and insufficient evidence that transcription inhibition does not perturb the replication program (see above). They would need to redo all the iPOND experiments using external "spike in" normalization and monitor replication genome-wide before they can make any meaningful conclusions about the transcription dependent composition of the proteome associated with replicated DNA.

    Expertise keywords: Chromatin, Genomics (assay development and bioinformatics analysis) , Replication, Transcription

  9. Version published to 10.1101/2023.04.19.537523v1 on bioRxiv