Condensin DC loads and spreads from recruitment sites to create loop-anchored TADs in C. elegans

  1. Jun Kim
  2. David S Jimenez
  3. Bhavana Ragipani
  4. Bo Zhang
  5. Lena A Street
  6. Maxwell Kramer
  7. Sarah E Albritton
  8. Lara H Winterkorn
  9. Ana K Morao
  10. Sevinc Ercan

  • Curated by eLife

    eLife logo

    Evaluation Summary:

    This paper is potentially of broad interest to researchers in the chromosome biology field. With specific loading sequences identified, the condensin DC complex studied here provides an elegant system to investigate the in vivo activities of SMC complexes. Combining Hi-C, ChIP-seq and RNA-seq, the authors have a comprehensive suite of assays to probe their questions. However, not all of their major conclusions are currently supported by the data.

    (This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. The reviewers remained anonymous to the authors.)

Reviewed by

Read the full article

Listed in

Log in to save this article

Abstract

Condensins are molecular motors that compact DNA via linear translocation. In Caenorhabditis elegans , the X-chromosome harbors a specialized condensin that participates in dosage compensation (DC). Condensin DC is recruited to and spreads from a small number of r ecruitment e lements on the X -chromosome ( rex ) and is required for the formation of topologically associating domains (TADs). We take advantage of autosomes that are largely devoid of condensin DC and TADs to address how rex sites and condensin DC give rise to the formation of TADs. When an autosome and X-chromosome are physically fused, despite the spreading of condensin DC into the autosome, no TAD was created. Insertion of a strong rex on the X-chromosome results in the TAD boundary formation regardless of sequence orientation. When the same rex is inserted on an autosome, despite condensin DC recruitment, there was no spreading or features of a TAD. On the other hand, when a ‘ super rex ’ composed of six rex sites or three separate rex sites are inserted on an autosome, recruitment and spreading of condensin DC led to the formation of TADs. Therefore, recruitment to and spreading from rex sites are necessary and sufficient for recapitulating loop-anchored TADs observed on the X-chromosome. Together our data suggest a model in which rex sites are both loading sites and bidirectional barriers for condensin DC, a one-sided loop-extruder with movable inactive anchor.

Article activity feed

  1. Version published to 10.7554/elife.68745 on eLife
  2. eLife

    Author Response

    Reviewer #1 (Public Review):

    The study by Jimenez et al. investigates the molecular mechanism by which dosage compensating (DC) condensins spread along the X chromosomes of C. Elegans worms. It has been previously known that DC condensins are loaded onto X chromosomes at specific sites called rex, that are distributed along the whole length of the chromosome. Here, Jimenez et al showed that an insertion of one or multiple rex sites into an autosome is sufficient for DC condensin recruitment and spreading. Using ChIP-seq, they show that DC condensins spread for hundreds of kilobases on the both sides of the rex site, with occasional sites of accumulation. The authors used Hi-C to study the effect of rex insertion on the chromosome conformation. They found that individual rex sites form boundaries that insulate spatial contacts regardless of their orientation, while two adjacent insertion sites can form loop-anchored contact domains. These findings support the model, in which DC condensins spread along the chromosome via the process of loop extrusion. In addition, the authors fused the X chromosome with the chromosome V and demonstrated that condensins can spread for multiple megabases across the fusion site and induce local compaction of the affected region. Finally, the targeted dCas9-Suntag complex to multiple adjacent copies of a repeat on chrX to demonstrate that condensins can accumulate at "bulky" obstacles.

    Overall, I find the experiments in this study are sufficient to support the key statements. My only comment is minor. In the discussion, the authors seem to imply that their data supports bi-directional loop extrusion by DC condesins (p.11 line 16). Yet, their data is consistent with a model, where condensins are loaded in a random orientation, but then extrude loops only into one fixed direction. Along these lines, ref. [20] (Terekawa et al) is mentioned as supporting bi-directional extrusion, while this paper in fact demonstrated that, once loaded onto DNA, condensins keep moving into a single direction with barely any observed inversions.

    Data is consistent with a model, where condensins are loaded in a random orientation, but then extrude loops only into one fixed direction:

    We agree with this interpretation and is explicitly stated and incorporated into the model (see section on A model to explain X-specific recruitment of condensin DC and formation of loop-anchored TADs by rex sites)

    Reviewer #2 (Public Review):

    SMC complexes play critical roles in chromosome organization from bacteria to humans. Recently in vitro studies found that SMC complexes function by extrude DNA loops. In vivo evidence for the loop extrusion model is less direct. The study by Jimenez et al investigated the mechanism of a specialized SMC complex called Condensin DC that mediates dosage compensation in C. elegans. This is an excellent experimental system to study SMC action in vivo because the specific loading sequence (rex) for Condensin DC was identified. The authors inserted the sites ectopically into autosomes and found that Condensin DC was recruited to ectopic sites and spreads to long distances. In a strain with a fusion chromosome (X;V), the complex spread beyond ChX to ChV. Finally, the authors generated a dCas9 mediated protein roadblock to test whether a large protein barrier prevents Condensin DC from spreading.

    Strengths:

    The authors have an elegant experimental system to investigate SMC action in vivo. They have a comprehensive set of tools including ectopic loading sites, fusion chromosomes, dCas9 block, Hi-C, ChIP-seq and RNA-seq.

    Weaknesses:

    While the experimental system has great potential, some specific choices of insertion sites did not yield clear results and caused confusions. If they modify the location of rex site or the dCas9 binding sites, they might be able to bring more insights. I detail them below.

    1. The authors inserted rex sites to autosomes and observed recruitment of Condensin DC to the ectopic sites. The engineering of rex sites to ectopic locations was done before, so was the observation that these ectopic sites recruit Condensin DC and generate TAD border (Albritton 2018; Anderson 2019). The current study has 3 rex sites on ChII and has the potential to bring new insights on how multiple rex sites act cooperatively and how they create TAD borders. However, the results presented were not clear because the author used rex sites with different strengths. The middle site did not form TAD loops with the other two sites. It is unclear whether the strength of the rex site matter or whether the distance between the sites matter. If they used only the two strong sites, or used all 3 sites of the same strength, the authors could have clarified this point.

    Reviewer comment on the middle rex in the three rex insertion did not form TAD loops:

    Upon repeating Hi-C experiments in L3 (where Hi-C features are more clear compared to mixed developmental stage embryos) and additional analysis/visualization of the data (log2ratio to the wildtype condition), we show that the middle rex (rex-1) also forms TAD loops with both flanking stronger rex-8. The loop involving rex-1 (weaker rex) is clearly weaker than the TAD loop between the two flanking strong rex-8.

    Reviewer question on “does the strength of rex sites matter”:

    It is likely that the strength of the rex contributes to the strength of TAD loops based on the observation that flanking rex-8 inserts form a TAD loop stronger than one between rex-1 and rex-8. We agree with the reviewer that to address the contribution of the strength of rex sites, we would need to insert the same rex with increasing rex strength in pairwise fashion of equal distance. However, insertion of single rex-1 did recruit condensin DC (Supplemental Figure 5A). Therefore, double rex-1 insertion is unlikely to work and necessitates the presence of a strong rex like rex-8 nearby. The use of a strong “super rex” clarifies that the rex sites need to both recruit and act as a boundary and both their function correlate with their strength (see section under Condensin DC is loaded at rex sites and spreads in either direction)

    Reviewer question on “does the distance of rex sites matter”:

    In Anderson 2009 (Figure S4) the authors inserted three strong rex sites greater than 1MB apart from each other on chromosome-I, and observed no changes in Hi-C matrix. On the other hand, our three rex sites are inserted within a 100kb region, and showed loop-anchored TAD. This suggests that distance matters, and distribution of rex sites contribute to cooperativity (see Discussion on The cooperativity of rex sites contributes to the X-specific recruitment and spreading of condensin DC)

    1. The authors used the dCas9 system to test the loop extrusion model. They found that DPY-27 is enriched at the dCas9 array. They concluded that the dCas9 array blocked Condensin DC spreading and this result supported the loop extrusion model. However, this interpretation is not supported by the DPY-27 enrichment profile or the HiC profile. If the authors were correct that Condensin DC, loaded on rex sites on either side of the array, extruded DNA loops and got blocked by the dCas9 array, we would expect DPY-27 enrichment to build up highest at the periphery of the array and lowest at the center of the array; we would expect a domain border to form at the array because of the lack of interactions between regions outside of the array. Yet, the DPY-27 ChIP profile is flat and there is no change in HiC profile. The near-identical shape of the dCas9 and DPY-27 ChIP-seq peaks is reminiscent of a technical bias of ChIP-seq, that is open chromatin is more "ChIP-able" (Teytelman PNAS 2013). It is possible that dCas9+sgRNA unwinding the DNA caused artifact in ChIP-seq. It is possible that a freely diffusing nuclear-localized protein will show the same ChIP profile at the dCas9 site with no biological relevance. Since this result is a major conclusion of the paper, it is necessary for the authors to perform a ChIP-seq control using a freely diffusing nuclear protein.

    We thank the reviewer for their recognition of an potentially artifactual pattern and urge to perform additional controls. Please see our detailed response to Essential Revision point (3) and new section under (A dCas9-based block failed to recapitulate rex-like boundary on the X-chromosome).

    1. If the authors targeted dCas9 to a different site, they might be able to clearly show whether Condensin DC spreading is blocked by such road block. For instance, if they use the X-V fusion, and target dCas9 to a region on ChV but close to the junction, they could test their hypothesis by DPY-27 ChIP-seq.

    This is an excellent idea and one we had hoped to initially do years ago. However, this turned out to be a difficult experiment as there is no unique repetitive sequence near the fusion site on ChrX;V and dCas9 resulting in ChIP artifacts in our existing system as demonstrated in Supplemental Figure 4-1.

    1. The model (Fig 6) is confusing. The authors are trying to support the loop extrusion model in the text but their drawing is not loop extrusion (Banigan and Mirny 2020). The author should clarify what they mean. For instance, after recruitment at rex site (red bar, with two arrows pointing left and right), Condensin DC was drawn to encircle a single piece of DNA as it moves to the left. It is not clear how the blue ring on the right can capture another piece of DNA and extrude a DNA loop and then later reversed to encircling a single piece of DNA before approaching the green protein block.

    We updated the model figure (Figure 7) with clearly stated properties of the model in the legend and with figures more in line with the depictions of loop extrusion in previous work.

    Reviewer #3 (Public Review):

    1. The insertion locations of new rex sites is clear in the top panels of Figures 1A and S1A, but not in the bottom panels of these two figures. My interpretation of these figures is that the lines with pink and grey boxes are shown to help the reader understand how many rex sites are inserted in each line but the location of these boxes does not coincide with the actual location of the insertion sites. In the top panel of Figure 1A, it appears that the two "pink" sites correspond to two very large peaks of Dp727, whereas in the bottom they appear to be present in a region devoid of Dpy27. Authors should fix this because it is very confusing, since the bottom panel suggest that condensin is recruited to rex sites and then spreads to other sites in the genome without any condensin remaining at the rex sites.

    We have updated the figure with proper legend. The insertion sites are computationally annotated with more intuitive cartoons shown on top (Figure 5,6).

    1. The idea that rex sites recruit condensin would require the existence of a sequence-specific DNA binding protein that binds to rex and then interacts with condensin. This protein would then release condensin, which would extrude away and stop at TSSs. Has this been actually shown previously or is this an interpretation of observations such as those shown in Figure 1A? If not, the results shown in Figure 1 would equally agree with a model suggesting that condensin loads randomly in both autosomes and the X chromosome, and extrusion is stopped by large protein complexes bound to rex sites, which explains the accumulation at these sites. TSSs contain large transcription complexes that are not sufficient to stop condensin on their own but are able to if the second anchor contains 1-2 rex sites. This would make more sense in the context of what is known about cohesin in mammals. If someone has unequivocally shown that this is not the case, authors should discuss this in the Introduction because most non-worm readers will be thinking in these terms.

    It has not been unequivocally shown that rex sites are the loading sites. However, all prior information better supports this conclusion. We did add previous work demonstrating that rex sites autonomously recruit the DCC on extrachromosomal arrays to help readers outside the field. We directly address the “loading at all chromosomes” model and more extensive comparison to the cohesin system in the result section (Condensin DC is loaded at rex sites and spreads in either direction) and the discussion (Previous models of condensin DC binding on the X chromosomes). We conclude that rex sites functioning as mere barrier elements are insufficient to explain the system and likely also function as loading sites for condensin DC.

    1. Figure 1A. Authors should not ignore the large Dpy27 peak in the worms with one rex insertion. What is at this site, where one can also observe a Dpy27 peak in wt worms? Are there similar sites in other regions of the autosomes of wt worms? If so, if these sites do not contain rex motifs, they may indicate alternative regions of the genome that can either recruit or stop extrusion of condensin.

    Binding of the the DCC to the X and autosomes have been extensively analyzed in previous work using immunofluorescence (Csankovzski et al 2003 Science and other Meyer lab papers, FRAP and DPY-27 Halo localization in Breimann-Morao et al JCS 2022) ChIP-chip (Ercan et al Nature Genetics 2007, Jans et al Genes and Development 2009) and ChIP-seq (Albritton et al Elife 2017). Autosomal localization is not noticeable in IF data. In ChIP-chip and ChIP-seq, DCC is largely specific to the X-chromosome but there is some background signal on autosomes with a slight enrichment at active promoters, which are also shown to be more easily ChIPped in many studies. There is also some binding to repetitive regions e.g. histone genes, but no specific features or strong autosomal binding sites emerged in previous analyses by our lab and others. In summary, like any ChIP-seq data, there is a non-zero number of peaks (based on MACS peak caller) on autosomes, which may be true off-target binding events or technical artifacts.

    The peak that was present in the original figure is not prominent upon new read mapping and normalization of the data and use of equal y-scales between the insertion region and a comparable X-chromosomal region (See same site appearing as a blimp just past 9 Mb in Supplemental Figure 5-1). We notice such mapping and technical variability occur at some regions that were not originally designated as “blacklisted”by modENCODE, but give spurious peaks.

    In summary, autosomal binding is far less than that of X (we provide ectopic versus X chromosome DCC binding in Figure 5 and cite prior literature using IF, FRAP, ChIP-chip, ChIP-seq data). In addition, we are aware of the shortcomings of ChIP-seq technique, and only make relative conclusions. We show that when rex sites are inserted, binding near rex sites is higher than regions that are farther away (‘spreading’) (Figure 5A, Supplemental Figure 5-1). We also show that relative to other autosomes, the mean signal on chromosome-II increases when rex sites are inserted (Figure 5B). This relative comparison allows us to say that insertion of rex sites increases the binding frequency on chromosome-II without knowing the true off-target/autosomal binding frequency prior to rex insertion.

    1. Please include a supplementary table describing all the Hi-C data used in the manuscript, including numbers of replicates, total number of sequenced read pairs, mapped reads, inter-and intra-chromosomal contacts, and number of contacts >20 kb and <20 kb.

    This is provided in the corresponding tabs of the Supplemental File 1.

    1. Page 8, lines 24-38. Based on this discussion, it is difficult to visualize what is happening. First, the authors suggest that condensin is recruited to the ectopic rex sites and "spreads" bidirectionally away from these sites to stop at various sites in the genome. Now, in this discussion, the authors suggest that rex sites containing condensin make loops. Does this happen without extrusion, just by the rex sites coming together in the 3D space? Are the loops formed through interactions between two condensin rings? When the authors say that condensin "spreads", does this take place by extrusion or a different mechanism? As mentioned in #2 above, everything would make better sense if the accumulation of condensin at rex sites is not a consequence of initial recruitment but rather a consequence of random loading followed by extrusion and retention at rex sites.

    Please see our new discussion sections (Previous models of condensin DC binding on the X-chromosomes and A model to explain X-specific recruitment of condensin DC and formation of loop-anchored TADs by rex sites) and Figure 7 for better explanation of why “recruitment everywhere” is insufficient to explain all aspects of the system and the data, as well as a more clear explanation of our model.

    1. Figure 2C. Were the interactions highlighted in this figure determined to be the only statistically significantly different between control and rex insertions or were they defined visually? The interaction between the center rex bait and the right rex pink site appears to be the same as in control. However, there seem to be some significantly visually different interactions between the center and right baits and other regions in the genome. Authors should test whether these interactions are statistically significant and, if so, what is located at these non-rex sites.

    We report the differences between the wild type and insertion strains in the revised manuscript by using log2ratio of the Hi-C matrices (Figure 6). Here, the stripes and the TAD insulation effects are more clear and indicative of barrier function of rex sites (also shown as insulation score between insertion/wild type shown below the matrices). The DPY-27 binding sites near the inserted rex sites tend to be actively transcribed genes. This is consistent with previously observed positive correlation between DPY-27 and Pol II ChIP-seq data at non-rex DCC sites on X-chromosomes and at the autosomal spreading region in the X;V fusion chromosome (Ercan 2009 Current Biology, and Street et al 2019 Genetics).

    1. Figure 3A. The fact that rex sites can contain more than one motif, presumably a binding site for an unknown protein, complicates data interpretation. It would be helpful if the authors indicate at the top of Figure 3A the number of motifs and their orientation for each rex site currently shown. In the bottom panels of this figure, it appears that not all rex sites indicated at the top are able to "recruit" condensin. Authors should comment on this, and if there are differences in the number of motifs at these sites or the sequence of the motifs. Also, the newly inserted sites appear to "recruit" less condensin than some of the existing ones. Do the sites with the taller Dpy27 peaks have more motifs?

    To make the interpretation simple in the revised Figure 3, we provide the Hi-C matrix comparison between the two insertion strains where rex-8 was inserted in two opposite directions. Hi-C interactions are similar in two strains, thus rex-8 direction does not matter.

    Reviewer comment on “Rex sites containing more than one motif complicates data interpretation”:

    It is important to highlight that while rex-8 contains multiple motifs, as is the case for many strong rex sites (Albritton 2017, Figure 5A), they are all oriented in the same direction for rex-8. This is the main rationale for using rex-8 as opposed to other rex sites. The number and the orientation of motifs are indicated in the cartoon in the revised figure.

    Reviewer comment on “It appears that not all rex sites “recruit” condensin. Do sites with taller DPY27 peaks have more motifs?”

    This question has been addressed in our prior work, Albritton et al. 2017. Like shown for the binding motifs corresponding to transcription factors, not all 12-bp rex motifs are bound by the DCC, and the strength and the number of motifs correlate with binding strength but not absolutely (Albritton 2017, Figure 5A), and the nucleotide perturbation of the motif abolishes binding (McDonel 2006, Figure 2). Similarly, the strength of binding also correlates with insulation strength (Anderson 2019, Figure S3B).

    1. It is unclear from the experiments described in Figure 1 how the formation of new loops would affect transcription. In Figure 3A, it appears that some of the Hi-C heatmaps show signal that could correspond to compartmental interactions. I wonder if the authors have tested whether the formation of new loops disrupts these interactions, which may contribute to the stabilization of promoter contacts and affect transcription. It may be informative to look at subtraction heatmaps between the new insertion data and control, although the Hi-C data in the center panel appears to have lower quality.

    Linking 3D interaction and transcription, while an ongoing project in our lab, is difficult. C. elegans has high gene density (~5kb/gene) with small gene length (average ~2kb). This makes analyzing E-P or P-P interactions as a result of rex insertion difficult. However, we agree that compartment analysis is a good starting point. We thus provide compartment analysis for the endogenous X (Figure 1) and discuss implications in repression using the X;V fusion experiment (Figure 2, see section Spreading of condensin DC entails loop extrusion but cannot sufficiently form TADs without rex sites). Future work involving higher resolution techniques such Micro-C could better address the relationship between condensin DC mediated loops and promoter contacts related to transcription.

    1. Figure 4 and page 9 lines 16-36. It is not completely clear from the discussion of Figure 4 whether the Hi-C data from wildtype was obtained with fixed embryos whereas the data from X;V was obtained with unfixed embryos. If this is the case, it may not be appropriate to directly compare the two samples. When the authors say "the autosomal spreading region showed an increase in DNA contacts measured by Hi-C", is this within the region or between the region and other sites in the genome? Since the two datasets have been normalized to the same number of contacts, an increase in interactions within the chromosome V region adjacent to the X chromosome in the X;V sample could be explained if this region interacts less with the adjacent X chromosome. Authors should discuss in more detail how this analysis was performed and perhaps use subtraction heatmaps to illustrate the point.

    We repeated experiments in L3 with controls all performed under the same crosslinking conditions. We also adopted the use of log-derivatives to infer loop size, which has been shown to better capture the chromatin state than the P(s) (Polovnikov 2022). Schematics for sub-regions of the XV chromosomes, for which the log-derivative of P(s) is computed, are also provided for comparison, thus the readers can compare the region of condensin DC spreading (proximal V) to chromosome V that is unbound by condensin DC (distal V). We reason that the log-derivative of P(s), which uses the relative change in P(s), is better equipped to deal with the reviewer’s concerns regarding the relative nature of the normalized contact frequency matrix.

    1. Figure S4A. If there is an increase in condensin (Dpy27) in chromosome V and an increase in interactions in this region, would this imply that the "spreading" of condensin takes place by loop extrusion? Otherwise, the "spreading" of condensin as suggested in the model of Figure 6 would not create new interactions.

    Our data showing increased interactions specifically in the proximal V along with decreased compartmentalization at this region (Figure 2) indicate that at least some spreading involves loop extrusion. This is reflected in our discussion of the model (Figure 7).

    1. Figure 5 and page 10, lines 20-21. It is clear from Figure 5B that the presence of the block leads to an accumulation of condensin, although the bottom panels of Figure 5C suggest that this accumulation is lower than at the flanking rex-33 and rex-14 sites. However, contrary to the author's conclusion that this in vivo evidence for loop extrusion, the result may suggest the opposite. If condensin was extruding loops and stopped at the dCas9 site it should have formed a loop. Were the same cells used for the ChIP-seq and Hi-C experiments? If not, one trivial explanation is that dCas9 failed to work in the cells used for Hi-C. Authors should comment on the fact that the rex-23 and rex-34 sites do not seem to be located at TAD boundaries, whereas TAD boundaries in the left region of the figure seem to lack rex sites.

    We agree with the author’s prediction that stopping at the dCas9 site should have formed a loop. See our response to Essential Revision 2.

  3. eLife

    Evaluation Summary:

    This paper is potentially of broad interest to researchers in the chromosome biology field. With specific loading sequences identified, the condensin DC complex studied here provides an elegant system to investigate the in vivo activities of SMC complexes. Combining Hi-C, ChIP-seq and RNA-seq, the authors have a comprehensive suite of assays to probe their questions. However, not all of their major conclusions are currently supported by the data.

    (This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. The reviewers remained anonymous to the authors.)

  4. eLife

    Reviewer #1 (Public Review):

    The study by Jimenez et al. investigates the molecular mechanism by which dosage compensating (DC) condensins spread along the X chromosomes of C. Elegans worms. It has been previously known that DC condensins are loaded onto X chromosomes at specific sites called rex, that are distributed along the whole length of the chromosome. Here, Jimenez et al showed that an insertion of one or multiple rex sites into an autosome is sufficient for DC condensin recruitment and spreading. Using ChIP-seq, they show that DC condensins spread for hundreds of kilobases on the both sides of the rex site, with occasional sites of accumulation. The authors used Hi-C to study the effect of rex insertion on the chromosome conformation. They found that individual rex sites form boundaries that insulate spatial contacts regardless of their orientation, while two adjacent insertion sites can form loop-anchored contact domains. These findings support the model, in which DC condensins spread along the chromosome via the process of loop extrusion. In addition, the authors fused the X chromosome with the chromosome V and demonstrated that condensins can spread for multiple megabases across the fusion site and induce local compaction of the affected region. Finally, the targeted dCas9-Suntag complex to multiple adjacent copies of a repeat on chrX to demonstrate that condensins can accumulate at "bulky" obstacles.

    Overall, I find the experiments in this study are sufficient to support the key statements. My only comment is minor. In the discussion, the authors seem to imply that their data supports bi-directional loop extrusion by DC condesins (p.11 line 16). Yet, their data is consistent with a model, where condensins are loaded in a random orientation, but then extrude loops only into one fixed direction. Along these lines, ref. [20] (Terekawa et al) is mentioned as supporting bi-directional extrusion, while this paper in fact demonstrated that, once loaded onto DNA, condensins keep moving into a single direction with barely any observed inversions.

  5. eLife

    Reviewer #2 (Public Review):

    SMC complexes play critical roles in chromosome organization from bacteria to humans. Recently in vitro studies found that SMC complexes function by extrude DNA loops. In vivo evidence for the loop extrusion model is less direct. The study by Jimenez et al investigated the mechanism of a specialized SMC complex called Condensin DC that mediates dosage compensation in C. elegans. This is an excellent experimental system to study SMC action in vivo because the specific loading sequence (rex) for Condensin DC was identified. The authors inserted the sites ectopically into autosomes and found that Condensin DC was recruited to ectopic sites and spreads to long distances. In a strain with a fusion chromosome (X;V), the complex spread beyond ChX to ChV. Finally, the authors generated a dCas9 mediated protein roadblock to test whether a large protein barrier prevents Condensin DC from spreading.

    Strengths:

    The authors have an elegant experimental system to investigate SMC action in vivo. They have a comprehensive set of tools including ectopic loading sites, fusion chromosomes, dCas9 block, Hi-C, ChIP-seq and RNA-seq.

    Weaknesses:

    While the experimental system has great potential, some specific choices of insertion sites did not yield clear results and caused confusions. If they modify the location of rex site or the dCas9 binding sites, they might be able to bring more insights. I detail them below.

    1. The authors inserted rex sites to autosomes and observed recruitment of Condensin DC to the ectopic sites. The engineering of rex sites to ectopic locations was done before, so was the observation that these ectopic sites recruit Condensin DC and generate TAD border (Albritton 2018; Anderson 2019). The current study has 3 rex sites on ChII and has the potential to bring new insights on how multiple rex sites act cooperatively and how they create TAD borders. However, the results presented were not clear because the author used rex sites with different strengths. The middle site did not form TAD loops with the other two sites. It is unclear whether the strength of the rex site matter or whether the distance between the sites matter. If they used only the two strong sites, or used all 3 sites of the same strength, the authors could have clarified this point.

    2. The authors used the dCas9 system to test the loop extrusion model. They found that DPY-27 is enriched at the dCas9 array. They concluded that the dCas9 array blocked Condensin DC spreading and this result supported the loop extrusion model. However, this interpretation is not supported by the DPY-27 enrichment profile or the HiC profile. If the authors were correct that Condensin DC, loaded on rex sites on either side of the array, extruded DNA loops and got blocked by the dCas9 array, we would expect DPY-27 enrichment to build up highest at the periphery of the array and lowest at the center of the array; we would expect a domain border to form at the array because of the lack of interactions between regions outside of the array. Yet, the DPY-27 ChIP profile is flat and there is no change in HiC profile. The near-identical shape of the dCas9 and DPY-27 ChIP-seq peaks is reminiscent of a technical bias of ChIP-seq, that is open chromatin is more "ChIP-able" (Teytelman PNAS 2013). It is possible that dCas9+sgRNA unwinding the DNA caused artifact in ChIP-seq. It is possible that a freely diffusing nuclear-localized protein will show the same ChIP profile at the dCas9 site with no biological relevance. Since this result is a major conclusion of the paper, it is necessary for the authors to perform a ChIP-seq control using a freely diffusing nuclear protein.

    3. If the authors targeted dCas9 to a different site, they might be able to clearly show whether Condensin DC spreading is blocked by such road block. For instance, if they use the X-V fusion, and target dCas9 to a region on ChV but close to the junction, they could test their hypothesis by DPY-27 ChIP-seq.

    4. The model (Fig 6) is confusing. The authors are trying to support the loop extrusion model in the text but their drawing is not loop extrusion (Banigan and Mirny 2020). The author should clarify what they mean. For instance, after recruitment at rex site (red bar, with two arrows pointing left and right), Condensin DC was drawn to encircle a single piece of DNA as it moves to the left. It is not clear how the blue ring on the right can capture another piece of DNA and extrude a DNA loop and then later reversed to encircling a single piece of DNA before approaching the green protein block.

  6. eLife

    Reviewer #3 (Public Review):

    1. The insertion locations of new rex sites is clear in the top panels of Figures 1A and S1A, but not in the bottom panels of these two figures. My interpretation of these figures is that the lines with pink and grey boxes are shown to help the reader understand how many rex sites are inserted in each line but the location of these boxes does not coincide with the actual location of the insertion sites. In the top panel of Figure 1A, it appears that the two "pink" sites correspond to two very large peaks of Dp727, whereas in the bottom they appear to be present in a region devoid of Dpy27. Authors should fix this because it is very confusing, since the bottom panel suggest that condensin is recruited to rex sites and then spreads to other sites in the genome without any condensin remaining at the rex sites.

    2. The idea that rex sites recruit condensin would require the existence of a sequence-specific DNA binding protein that binds to rex and then interacts with condensin. This protein would then release condensin, which would extrude away and stop at TSSs. Has this been actually shown previously or is this an interpretation of observations such as those shown in Figure 1A? If not, the results shown in Figure 1 would equally agree with a model suggesting that condensin loads randomly in both autosomes and the X chromosome, and extrusion is stopped by large protein complexes bound to rex sites, which explains the accumulation at these sites. TSSs contain large transcription complexes that are not sufficient to stop condensin on their own but are able to if the second anchor contains 1-2 rex sites. This would make more sense in the context of what is known about cohesin in mammals. If someone has unequivocally shown that this is not the case, authors should discuss this in the Introduction because most non-worm readers will be thinking in these terms.

    3. Figure 1A. Authors should not ignore the large Dpy27 peak in the worms with one rex insertion. What is at this site, where one can also observe a Dpy27 peak in wt worms? Are there similar sites in other regions of the autosomes of wt worms? If so, if these sites do not contain rex motifs, they may indicate alternative regions of the genome that can either recruit or stop extrusion of condensin.

    4. Please include a supplementary table describing all the Hi-C data used in the manuscript, including numbers of replicates, total number of sequenced read pairs, mapped reads, inter-and intra-chromosomal contacts, and number of contacts >20 kb and <20 kb.

    5. Page 8, lines 24-38. Based on this discussion, it is difficult to visualize what is happening. First, the authors suggest that condensin is recruited to the ectopic rex sites and "spreads" bidirectionally away from these sites to stop at various sites in the genome. Now, in this discussion, the authors suggest that rex sites containing condensin make loops. Does this happen without extrusion, just by the rex sites coming together in the 3D space? Are the loops formed through interactions between two condensin rings? When the authors say that condensin "spreads", does this take place by extrusion or a different mechanism? As mentioned in #2 above, everything would make better sense if the accumulation of condensin at rex sites is not a consequence of initial recruitment but rather a consequence of random loading followed by extrusion and retention at rex sites.

    6. Figure 2C. Were the interactions highlighted in this figure determined to be the only statistically significantly different between control and rex insertions or were they defined visually? The interaction between the center rex bait and the right rex pink site appears to be the same as in control. However, there seem to be some significantly visually different interactions between the center and right baits and other regions in the genome. Authors should test whether these interactions are statistically significant and, if so, what is located at these non-rex sites.

    7. Figure 3A. The fact that rex sites can contain more than one motif, presumably a binding site for an unknown protein, complicates data interpretation. It would be helpful if the authors indicate at the top of Figure 3A the number of motifs and their orientation for each rex site currently shown. In the bottom panels of this figure, it appears that not all rex sites indicated at the top are able to "recruit" condensin. Authors should comment on this, and if there are differences in the number of motifs at these sites or the sequence of the motifs. Also, the newly inserted sites appear to "recruit" less condensin than some of the existing ones. Do the sites with the taller Dpy27 peaks have more motifs?

    8. It is unclear from the experiments described in Figure 1 how the formation of new loops would affect transcription. In Figure 3A, it appears that some of the Hi-C heatmaps show signal that could correspond to compartmental interactions. I wonder if the authors have tested whether the formation of new loops disrupts these interactions, which may contribute to the stabilization of promoter contacts and affect transcription. It may be informative to look at subtraction heatmaps between the new insertion data and control, although the Hi-C data in the center panel appears to have lower quality.

    9. Figure 4 and page 9 lines 16-36. It is not completely clear from the discussion of Figure 4 whether the Hi-C data from wildtype was obtained with fixed embryos whereas the data from X;V was obtained with unfixed embryos. If this is the case, it may not be appropriate to directly compare the two samples. When the authors say "the autosomal spreading region showed an increase in DNA contacts measured by Hi-C", is this within the region or between the region and other sites in the genome? Since the two datasets have been normalized to the same number of contacts, an increase in interactions within the chromosome V region adjacent to the X chromosome in the X;V sample could be explained if this region interacts less with the adjacent X chromosome. Authors should discuss in more detail how this analysis was performed and perhaps use subtraction heatmaps to illustrate the point.

    10. Figure S4A. If there is an increase in condensin (Dpy27) in chromosome V and an increase in interactions in this region, would this imply that the "spreading" of condensin takes place by loop extrusion? Otherwise, the "spreading" of condensin as suggested in the model of Figure 6 would not create new interactions.

    11. Figure 5 and page 10, lines 20-21. It is clear from Figure 5B that the presence of the block leads to an accumulation of condensin, although the bottom panels of Figure 5C suggest that this accumulation is lower than at the flanking rex-33 and rex-14 sites. However, contrary to the author's conclusion that this in vivo evidence for loop extrusion, the result may suggest the opposite. If condensin was extruding loops and stopped at the dCas9 site it should have formed a loop. Were the same cells used for the ChIP-seq and Hi-C experiments? If not, one trivial explanation is that dCas9 failed to work in the cells used for Hi-C. Authors should comment on the fact that the rex-23 and rex-34 sites do not seem to be located at TAD boundaries, whereas TAD boundaries in the left region of the figure seem to lack rex sites.

  7. Version published to 10.1101/2021.03.23.436694v2 on bioRxiv
  8. Version published to 10.1101/2021.03.23.436694v1 on bioRxiv