Single molecule microscopy reveals key physical features of repair foci in living cells

  1. Judith Miné-Hattab
  2. Mathias Heltberg
  3. Marie Villemeur
  4. Chloé Guedj
  5. Thierry Mora
  6. Aleksandra M Walczak
  7. Maxime Dahan
  8. Angela Taddei

Reviewed by

Read the full article

Listed in

Log in to save this article

Abstract

In response to double strand breaks (DSB), repair proteins accumulate at damaged sites, forming membrane-less sub-compartments or foci. Here we explored the physical nature of these foci, using single molecule microscopy in living cells. Rad52, the functional homolog of BRCA2 in yeast, accumulates at DSB sites and diffuses ~6 times faster within repair foci than the focus itself, exhibiting confined motion. The Rad52 confinement radius coincides with the focus size: foci resulting from 2 DSBs are twice larger in volume that the ones induced by a unique DSB and the Rad52 confinement radius scales accordingly. In contrast, molecules of the single strand binding protein Rfa1 follow anomalous diffusion similar to the focus itself or damaged chromatin. We conclude that while most Rfa1 molecules are bound to the ssDNA, Rad52 molecules are free to explore the entire focus reflecting the existence of a liquid droplet around damaged DNA.

Article activity feed

  1. Version published to 10.7554/elife.60577 on eLife
  2. eLife

    ###Reviewer #3:

    In the manuscript by Miné-Hattab et al the authors revisit a phenomenon that has been extensively studied for over 10 years: the subdiffusive and diffusive properties of DNA damage binding factors in repair foci (inside and outside of foci). The work is carefully done and brings a few observations of interest, but the novel insights are extremely limited. The most original aspect is that they characterize the movement of repair molecules within the focus with movement of the focus itself (the movement of foci has been done by many and turnover of factors has also been done by many). That they compare the two with one set of measurements is the key contribution of the paper, and they do find differences in diffusion coefficients. It is likely that this was not done previously. It is difficult to judge, as key papers that showed similar conclusions or datasets are not cited.

    Here are a few key examples:

    1. In the last year the Haber lab published a very similar study in Plos Genetics (Live cell monitoring of double strand breaks in S. cerevisiae, Waterman et al 2019 https://doi.org/10.1371/journal.pgen.1008001 ). Although they tracked Ddc2 and Rad51, they also looked at the behavior of separate foci and this paper is not even cited. The data should be compared at the very least.

    2. The characteristics of 53BP1 foci have been extensively studied by many labs including those of Altmeyer, Scherthan, DeLange and others, with very similar findings as Miné-Hattab reports for Rad52 (for example, Phase separation of 53BP1 determines liquid‐like behavior of DNA repair compartments, Kilic et al., EMBO J. 2019 38(16): e101379; Live Dynamics of 53BP1 Foci Following Simultaneous Induction of Clustered and Dispersed DNA Damage in U2OS Cells Alice Sollazzo et al., Int. J. Mol. Sci. 2018, 19, 519 as well as the single molecule work of the lab of Eric Greene). Moreover both rad52 and PCNA foci were studied by Essers et al. (Kanaar and Vermeulen) MCB 2005. 25(21): 9350-9359 and EMBO J. 2002 Apr 15. Comparisons with these studies needs to be made.

    3. A number of earlier studies followed Rad52 foci in budding yeast on induced double strand breaks (even using the I-Sce1-cut system used here) that are not taken into consideration. The diffusion coefficients presented here have to be compared with these earlier studies and differences should be resolved by comparing techniques and conditions of imaging. For instance, Dion et al., Nature Cell Biology 2012).

    In brief, while the execution and analysis of the data shown here is very good, without direct comparison with other data sets, it is difficult to see exactly where this paper goes beyond published studies. This is especially crucial as the paper as written makes no effort to compare their data with existing datasets. Most specifically a comparison with LLPS as defined for other chromatin-foci forming proteins in the nucleus needs to be done - particularly addressing studies in mammalian cells concerning 53BP1 and other repair factors. This, plus a careful comparison with data from induced Ise1-break movement, must both be included. Finally, insufficient data are provided to draw conclusions about whether or not the authors' observations are reflective of phase separation. Additional mobility studies in conditions that disrupt LLPS are needed, both for the individual protein and for the foci. In conclusion, serious revision is needed and an effort must be made to show to the reader that this data is comparable (or not) with other data in the literature.

  3. eLife

    ###Reviewer #2:

    Miné-Hattab et al. conduct a study focusing on the behaviour of the DNA repair protein Rad52 at sites of DNA damage in budding yeast. Several DNA repair proteins, including yeast Rad52, have been previously observed to phase separate at sites of DNA damage in a number of organisms. However, the authors here aimed to more accurately consider the potential phase separation behaviour of Rad52 by using single particle tracking (SPT) and Photo-activatable Localization Microscopy (PALM). Overall, the findings are consistent with previous studies and provide additional evidence supporting the concept that Rad52, but not the ssDNA-binding protein RPA, phase separates at the site(s) of DNA damage. The data shown also support the long-appreciated concept that different DSB sites cluster within the nucleus, albeit this study presents higher resolution data. The study falls within an important area of investigation.

    1. The study does not present a novel conceptual advance.

    2. What is the evidence that the biophysical properties observed are of direct relevance to DNA repair? For example, is the mobility of Rad52 within the repair focus important for repair? Is the difference in diffusion kinetics within and outside of the repair focus important for genome stability? What could the authors do to alter that diffusion profile and what would be the consequence on repair? Also, addressing this point implies the need to use a more physiologically relevant system with repairable DSBs, and not the irreparable DSB system used here. The authors describe the work of many in the field as "extremely phenomenological", yet it is not clear what the authors did to go beyond such a statement.

    3. Overall, the statistical significance of most of the presented data is either lacking or unclear. This needs to be carefully addressed.

    4. It is unclear if the 'absence of DNA damage' condition discussed in the first section of the results is the non-induced version of the system described in the second section of the results. Also regarding these sections, it seems that the 'absence of DNA damage' control conditions were not conducted as part of the same experiments with the I-SceI DSB.

    5. Several statements made are not supported by the data and without clearly stating that the statements represent speculations. E.g. page 4, longer tail is due to Rad52 molecules diffusing slowly inside the focus; page 8, observing the 2 populations also in G1 does not necessarily mean that the 2 populations in S/G2 do not reflect replication forks at all. The authors need to carefully revise their claims/statements and consider alternative explanations. Also, the writing is often unclear or confusing and the authors should consider substantially revising it to clarify their claims, clearly indicate speculations that are not supported by the data, and make the text as accessible as possible to non-specialists.

    6. How do the authors reconcile previous findings indicating that recombinant DNA repair proteins phase separate in vitro with their claim that "Rad52 acts as a client of the LLPS but does not drive its formation" on page 11?

    7. How was the cell cycle stage determined?

    8. Fig S1 data appear to show the existence of a partial loss of Rad52 function in the Rad52-Halo cells. This should be clearly expressed in the results and consequent limitations/caveats discussed. Also, please clarify whether Fig S1 shows the viability of Rad52-Halo cells in the presence or absence of JF646.

    9. Regarding the possible categories of traces evaluated, one category is not included in the study. The surface tension that defines LLPS-dependent bodies is known to both help maintain focus integrity and partly counter LLPS body fusions. So if the foci represent true phase-separated bodies, have the authors then observed traces where Rad52 molecules interact with yet fail to enter the larger Rad52 foci?

    10. The authors present no direct evidence for an "attractive potential" that drives molecules towards the centre of the focus. For example, what if the 'attractive potential' is simply the focus' boundary surface tension creating a barrier against which some of the molecules inside the focus bounce back towards the centre of the focus?

    11. Consider revising the discussion to shorten it while making it more focused on conceptual advances and higher level interpretations, without re-describing the results in detail.

    12. Can the authors visualize the fusion of the Rad52 foci/DSBs in live cells within their experimental systems?

    13. The authors state on page 10 that "Here, we found that upon different levels of Rad52 over-expression, the background concentration increases (Figure S8) suggesting that Rad52 might not be the driving molecule responsible for the LLPS formed at the damaged site." Can the authors explain the logical transition here more clearly, it was unclear.

  4. eLife

    ###Reviewer #1:

    In this manuscript by Mine-Hattab and colleagues, the authors use single-molecule tracking in yeast to dissect the formation of the double-stranded break response in living cells. Specifically, they try to determine the nature of Rad52 clustering at the DSB focus. The sequential recruitment pathway is well-studied in yeast (RPA --> Rad52 -->Rad51), and the inducible I-SceI break offers a controlled system for DNA damage. Moreover, yeast could be an excellent model system to elucidate if there is any conservation or function for such compartments. Overall, I found the data and the subsequent analysis to be both rigorous and nuanced. Ultimately, one is trying to distinguish whether the focus is due to a clustering of binding sites or liquid-liquid phase separation, or perhaps some combination of the two. I feel the story falls short of providing a definitive answer, as do many in this field, but the authors conclude that the preponderance of evidence points to a LLPS model for Rad52 clustering.

    1. How is it possible to distinguish a cluster of binding sites from liquid-liquid phase separation? To this referee, that is the question that needs answering. In the absence of breaks, there are two Rad52 diffusion populations (D= 1.2 and 0.3 um2/s), which the authors attribute to monomers and multimers. They don't verify these multimers by alternative approaches (say number and brightness analysis), but it seems like a reasonable possibility. After a break, a third component - slower than the previous two --becomes evident. This slow population coincides with the break. In the vicinity of the break, there is now only 1 component diffusion (D=0.03 um2/s). Also, the motion is now more confined, but not absolutely so. Also, Rad52 diffuses faster than Rfa1, which is bound to ssDNA. At this point, there is no data to distinguish between two possibilities: slow diffusion or diffusion + binding. Except, if it were diffusion + binding, one might perhaps expect to still see the free diffusion component. However, I can imagine lots of different scenarios and a range of binding affinities and multimer states that would make that analysis an unholy mess.

    The authors then turn to diffusion at the boundary (Fig. 5), which I agree can be a more informative measure. Here, they see changes in the diffusion estimator for trajectories which cross the boundary, using displacement which they argue is more robust for slow diffusion. The problem is that the 'boundary' is determined by the very thing they are trying to measure, not some independent marker of the compartment. In other words, Rad52 defines the compartment, unless I missed something fundamental in the experimental design. Ideally, the way such an experiment would be done to test the hypothesis that Rad52 is forming a LLPS compartment is to look at the diffusion of an inert tracer as it comes in and out of the compartment. As designed, I frankly do not see how the observation of different diffusivities in and out of the compartment distinguishes between a cluster of binding sites and an LLPS. If you accept that DNA-binding is in no way biasing the kinetics, then the authors' interpretation seems like the most sensible one. But the fact that Rad52 is involved in DNA repair makes that a hard assumption to swallow.

    Furthermore, I'm not sure I entirely grasp the significance of Fig. 6. Since Rad52 can easily escape one focus and enter another, regardless of whether it is a cluster of binding sites or a phase, I don't see how the radius of confinement measurement distinguishes between these two alternatives. The observation that the foci are 2x larger in diploids but at similar density is compelling, although recent data from the Brangwynne lab point out that conserved density need not be the case (PMID: 32405004).

    1. In the syntax of this paper, Rad52 is a client in the LLPS, leaving the question of the scaffold unaddressed. After all, the Rad52 focus ultimately disappears, meaning that something caused this phase to be dispersed. So is RPA the scaffold? It might be possible to address both points 1 and 2 by knowing what is responsible for forming the LLPS in the first place.

    In summary, I found the paper to be balanced and rigorous when exploring possible interpretations of the data. Although the authors may feel the preponderance of their data is consistent with LLPS, I don't feel they have nailed it. It's hard to identify a smoking gun. Of their four observations in the discussion only the second is direct, and that observation may have other explanations. However, I am not sure what experiment to recommend which would be definitive. Such is the nature of this field.

  5. eLife

    ##Preprint Review

    This preprint was reviewed using eLife’s Preprint Review service, which provides public peer reviews of manuscripts posted on bioRxiv for the benefit of the authors, readers, potential readers, and others interested in our assessment of the work. This review applies only to version 1 of the manuscript.

    This manuscript is in revision at eLife

    ###Summary:

    In this manuscript by Miné-Hattab and colleagues, the authors use single-molecule imaging approaches to investigate local dynamics of Rad52 foci at DSBs in budding yeast, which is an important area of investigation. They show that the dynamics of Rad52 molecules inside foci are consistent with protein movement within LLPS domains, while Rfa1 dynamics are not. Their data also provide supporting evidence to previous observations that repair sites cluster within the nuclei, and suggest that clustered foci behave as larger phase separated structures. While the idea that Rad52 and other repair proteins form phase separated domains is not novel, this study presents higher resolution data in support of this model. The reviewers generally agree that the study is interesting and well conducted, but the conceptual advancement is limited. Specifically, more convincing experiments demonstrating that the observed Rad52 dynamics reflect LLPS are required. Evidence that the dynamics are relevant for DNA repair and genome stability should also be provided. Additionally, the study should be better integrated with previous studies, statistical analyses need to be more rigorous/better presented, and the text should include a clearer separation between observations and speculations.

  7. Version published to 10.1101/2020.06.18.160085v1 on bioRxiv