Transcription decouples estrogen-dependent changes in enhancer-promoter contact frequencies and spatial proximity

  1. Luciana I. Gómez Acuña
  2. Ilya Flyamer
  3. Shelagh Boyle
  4. Elias Friman
  5. Wendy A. Bickmore

Reviewed by

Read the full article

Listed in

Log in to save this article

Abstract

How enhancers regulate their target genes in the context of 3D chromatin organization is extensively studied and models which do not require direct enhancer-promoter contact have recently emerged. Here, we use the activation of estrogen receptor-dependent enhancers in a breast cancer cell line to study enhancer-promoter communication. This allows high temporal resolution tracking of molecular events from hormone stimulation to efficient gene activation. We examine how both enhancer-promoter spatial proximity assayed by DNA fluorescence in situ hybridization, and contact frequencies resulting from chromatin in situ fragmentation and proximity ligation by Capture-C, change dynamically during enhancer-driven gene activation. These orthogonal methods produce seemingly paradoxical results: upon enhancer activation enhancer-promoter contact frequencies increase while spatial proximity decreases. We explore this apparent discrepancy using different estrogen receptor ligands and transcription inhibitors. Our data demonstrate that enhancer-promoter contact frequencies are transcription independent but are influenced by enhancer-bound protein complexes whereas altered enhancer-promoter proximity depends on transcription. Our results emphasize that the relationship between contact frequencies and physical distance in the nucleus, especially over short genomic distances, is not always a simple one.

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

    We thank the reviewers for their critiques of our manuscript and for recognizing the importance of the questions about 3D genome organisation that it addresses. We plan to address most of their comments in our revised manuscript.

    Reviewer #1

    1. The aneuploid karyotype of the MCF-7 cells used is a concern. GREB1 is present in four copies, with two on abnormal chromosomes which may not be regulated in the same way as primary cells. The authors should include caveats to this effect in the text to account for this.

    We indicated (pg 5) that there are 4 copies of GREB1, 2 of which are on re-arranged chromosomes. RNA FISH (Figure 1C) suggests all 4 of these alleles are induced by estrogen. On each allele, the GREB1 enhancer and promoter remain closely apposed by imaging (Figure 2, DNA FISH) indicating no gross chromosomal rearrangements around the GREB1 locus. This is confirmed by our Hi-C data (Figure 2A), where any genomic rearrangements at the GREB1 locus would be detectable when the sequencing data were aligned to the reference genome. In the revised manuscript we highlight these points in the respective results sections (pgs 5 and 6). Our data suggest that all 4 alleles of GREB1 in MCF7 cells are regulated in the same way.

    *2. The authors should also include more information on the generation and verification of the enhancer deletion cell lines. An illustration of the PCR primers used for screening, as well as an illustration of the sequenced product traces aligned with the reference genome (as opposed to just showing the deleted regions) should be included in Fig. S1D. This would give the reader more confidence that the designed knockout has occurred in the same way on all alleles. Furthermore, long-range PCRs and sequencing should be considered to confirm that no larger deletions have *occurred (e.g. Owens et al., 2019 PMID: 31127293).

    We have replaced FigS1D with a new Figure Supplement (Figure S1.2A) that incorporates a more comprehensive diagram of the strategy used for the generation and screening of the enhancer deletion cell lines. This also includes the sequencing traces aligned to the reference genome for each of the clones used in this work. Additionally, in the revised manuscript, we will check the deletions using the C-TALE sequencing data obtained from the enhancer-deleted clones.

    1. The changes in the measured E-P interaction frequency following gene activation are __weak __at best and make visual interpretation of the results difficult. Showing the reciprocal virtual 4C plots from the promoter would help to reassure the reader that the observed effect is real.

    We thank the reviewer for this suggestion, and we will now include virtual 4C plots from the GREB1 and NRIP1promoters in our revised manuscript. These will be in figures 2B, 2E, 3C, 4C and in the supplementary figures 2B, 4B, 5C and 6C.

    4. Furthermore, the precise 3C method used is not clear. The authors repeatedly refer to "Capture-C" (a commonly used 3C-based approach using biotinylated oligos to pull down targets of interest) but the citation used (Golov et al. 2019) refers to a conceptually similar method called "C-TALE". This should be clarified in the text.

    We thank the reviewer for pointing out this potential confusion. We replace the term Capture-C with C-TALE throughout the revised manuscript.

    5. As for the changes in contact frequency, the observed changes in distance measurements between conditions are very small (although statistically significant). We acknowledge that this is likely due to the relatively small linear distances between enhancers and promoters in this study. However, it would be helpful to see the effects of the induction/treatments on a one or more control loci which is not affected by oestrogen signalling given that global changes in nuclear shape/volume and/or cell cycle effects could occur within this time (e.g. effects of tamoxifen treatment on MCF-7 cell cycle distribution, (Osborne et al. 1983 PMID: 6861130), which could impact nuclear volume.

    Data from DNA FISH control probes are already included in Supplementary Figure S3 showing no change in intra-nuclear distances and thus no general effects on chromatin compaction due to nuclear volume or cell cycle. Virtual 4C data for the entire captured regions around GREB1and NRIP1 are show in Fig S2C, also showing no general effect on the wider capture windows. We will include similar data from the viewpoint of the gene promoters in the revised manuscript. Hi-C and imaging data from the enhancer deletion cell lines (Fig S4) also supports that we are looking at an ER-specific effect, not a global one. With the regard to the comment on the effects of tamoxifen treatment on MCF-7 cell cycle distribution, we see no effects of tamoxifen on 3D genome organisation at *GREB1 *and NRIP1 by Hi-C or by imaging.

    6. The authors discuss previous studies demonstrating that E2 and 4OH recruit different sets of proteins to their target genes. Given that this is central to the conclusion that the ER ligand (and its recruited co-factors) determines the E-P interaction frequency and 3D distances observed, it would be important to demonstrate this at the GREB1/NRIP1 loci specifically. ChIP data of the co-activators/repressors recruited by E2 and 4OH, respectively, would greatly strengthen this claim.

    We acknowledge that investigating co-activator and co-repressor recruitment to the studied loci will strengthen our interpretation our conclusions. In the revision we will perform and include ChIP-qPCR at NRIP1, GREB1 and control loci assaying for PolII, co-activators such as p300, mediator and SRC-3 and the co-repressors N-CoR in control, estradiol and tamoxifen treated cells. We will also perform ChIP-qPCR of PolII and co-activators in cell treated with flavopiridol and triptolide.

    1. The observed uncoupling of E-P contact frequency and 3D distance upon transcriptional inhibition is interesting and offers clues to the molecular details underlying E-P interactions. However, the use of flavopiridol and triptolide, while common in the field, should be carefully qualified given the potential for their indirect effects on transcription. This is particularly important for flavopiridol given its ability to target multiple cyclin-dependent kinases beyond CDK9 and its role in transcription initiation.

    In the revised manuscript we indicate that “Flavopiridol inhibits several CDKs, including CDK9/PTEF-b”

    Minor comments:

    i. In the introduction and beginning of discussion, it would be helpful to detail previous studies where FISH-based analyses have shown more proximal E-P positioning upon activation, to make it clear that differences in E-P proximity appear to be gene-specific. Some examples include Williamson et al. (2016; PMID 27402708) and Chen et al. (2018; PMID 30038397). Speculation as to why some genes behave in this way while others do not, would also be worthwhile.

    We have followed the reviewer’s suggestion and noted these two studies in the Introduction of a revised manuscript. Given that the focus of this current manuscript is to explore discrepancies between Hi-C and DNA FISH, we do not think that this is the right forum for a wider discussion of why there might be differences in E-P proximity between different biological systems.

    ii. On page 6, the authors state that after deletion of the NRIP1 enhancer there is "almost total loss of NRIP1 induction in response to E2". This does not seem to match the data where in 3 out of 4 replicates (2 for each clone) there is a statistically significant increase in number of RNA FISH foci upon E2 stimulation in the NRIP1 enhancer KOs. This suggests that, as for GREB1, the regulation of these genes is not solely controlled by the deleted enhancers. This should be clarified in the text.

    The reviewer is referring to the data on NRIP1 expression in two NRIP1 enhancer deletion clones in Fig 1D and the replicate data in Supplementary Fig S1 (upper-right panel). These data show almost no induction of NRIP12 by E2 compared to wild-type cells. We stand by our statement.

    iii. The labelling of the FISH probes in Supp. Fig. S2 could be improved as it is currently very difficult to read these.

    We will try to improve this in a revised Figure S2.

    iv *Given that the authors have referenced a distance of 200 nm as potentially being an important threshold for gene activation, it would be useful to include the fraction of alleles which are below this *distance alongside the cumulative frequency plots in Figure 2D and elsewhere in the paper as the cumulative frequency plots can be hard to read in some cases (e.g. Supp. Fig. S3B e-p). This would also allow the authors to show consistency across replicates.

    We thank the reviewer for this suggestion to make the data easier to interpret. In a revised manuscript, we will incorporate the fraction of alleles below and above 200 nm for the DNA-FISH experiments in Figure 2D and Figure S4A-B.

    v. For clarity, it would be helpful to include the difference map between the vehicle-treated unstimulated/stimulated conditions for the 3C plots in Fig. 4. This would help contextualise the resulting differences observed with the drug treatments. Same for Supp. Fig. S6.

    We will include the difference heatmap between the vehicle- and estradiol treated samples for vehicle, flavopiridol and triptolide treated samples.

    vi. Statistical comparisons are not shown for all 3D FISH-based distance measurements (e.g. Supp. Figs. S3A, S4C, D, S6E). If this is because the tests were done and the results were non-significant this should be indicated.

    We had omitted all non-significant p values (>0.05) from the graphs to stop them getting too cluttered. All p values are documented in the supplementary tables. However, following the reviewer’s comment, we will indicate all non-significant statistical comparisons on the graphs.

    vii. On page 13, the authors state that increased E-P separation occurs "before nascent transcription of the gene is detected by either TT-seq or RNA FISH". This does not appear to be correct given that baseline levels of transcription are observed in the absence of ER stimulation by both methods (Fig. 1). This should be clarified in the text.

    We have amended this statement to now indicate that “This is before an induction of nascent transcription of the gene….”

    Reviewer #2

    1. The authors make strong claims and although these are generally reasonably well supported by the data, it is important to acknowledge that they are based on two loci. This manuscript would be stronger if the authors could include additional loci in their study design. If this is not possible, it would be good to acknowledge that the conclusions are preliminary/speculative at this stage.

    The reviewer makes a fair point, and we emphasized throughout the text – including at the end of the Discussion - that we are examining just two gene loci. In a revised manuscript we will include DNA-FISH data for a third locus comprising the CCND1 gene, for which we have preliminary data.

    *2. It would be helpful if the authors could clarify the strategy they used for their FISH probe design. The enhancer and promoter fosmid probes (which are used for the majority of the experiments) are not centered on the active elements and do not even seem to overlap in the case of the GREB1 enhancer fosmid probe. The 10 kb enhancer probe seems better placed for the GREB1 locus, but the 10 kb enhancer probe does not seem to overlap with the enhancer in the NRIP1 locus. It is conceivable that the exact location of the probes has a big impact on the measurements and it would therefore be helpful if the authors could comment on the location of the probes and add additional probes if required to strengthen their conclusions. In addition, the fosmid probes are very large (40 kb). Although the authors acknowledge this, it would be helpful if they could comment on how overlap between 40 kb probes should be interpreted in relation to a potential rather focal contact between (proteins bound to) regions of In the case of GREB1, the fosmid probes were chosen to maximize the distance between them as the promoter and the enhancer of the gene are genomically relatively close to each other. This was not an issue in the case of the NRIP1 locus where fosmid probes could be placed centered on the TSS and the enhancer region. In the case of the 10 kb probes, these were designed to be centered on the regions where higher E2-induced C-TALE contact frequencies were detected. Virtual 4C plots using the TSS regions as viewpoints (incorporated into the revised manuscript) clearly show that, in the case of NRIP1, the contact frequency peak does not fall on the main ER peak.

    1. It is not clear to me why the authors would choose to work with a locus that is present in 4 copies in their cell line. Is the entire regulatory region (incl. enhancers) preserved for the two additional copies of the gene? Can the authors comment on how this may impact on their measurements?

    See response to Reviewer 1, point 1. Our Hi-C data would have revealed if there were genomic rearrangements in the 600kb window surrounding GREB1.

    4. Figure 2D shows an increase in E-P separation for the NRIP1 locus across all timepoints, with cumulative frequency plots shown for the 10 min timepoint. However, the data for the second replicate shown in Figure S2D are a lot less robust and not significant for the 10 min timepoint. It is important that the authors either provide additional data to support the robustness of this experiment or acknowledge that the results are not fully reproducible.

    We acknowledge this, but we would like to note that there is an increase in the median distance for all time points, although this difference is not significant in some of the timepoints. Additionally, DNA-FISH data obtained using the 10 kb probes confirm these observations.

    5. The data presented in Figure 2F for clone 2 of the GREB1 enhancer deletion still show increased E-P distance upon activation. How do the authors explain this?

    This increase in distance is not statistically significant (p-0.33 – see Table S2) and is not seen for the replicate data in Fig. S4.

    Minor comments:

    i. Could the authors comment on the observation that the NRIP1 promoter is not bound by ERa or p300 upon estrogen activation? Are there ATAC-seq or H3K27ac ChIP-seq data available for these conditions?

    We included ATAC-seq tracks in Figure 1A where a peak on the NRIP1 promoter is clearly seen.

    ii. It is not obvious which timepoint is shown in Figure 1D.

    Pre-mRNA FISH in enhancer deleted clones was done in cells treated with vehicle or E2 for 60 minutes. This will be made clearer in the figure legend.

    iii. Why did the authors choose e-i and p-i instead of e-c and p-c in Supplementary Figure 3B?

    We apologize as it was an oversight not to include the e-c data for this experiment. This is now included in Supplementary figure S4B.

    iv. "We treated hormone starved MCF-7 cells with flavopiridol or triptolide for 5 min before adding E2 for 30 min (Fig. 4A)." Does this mean that the FLV/TRP treatment lasted for 35 min or did the authors wash it out before adding E2? Please clarify.

    This observation is correct, and it was made clear in Figure 4A and in the figure legend.

    v. The authors refer to their Capture-C data as "high-resolution". However, the methods section mentions that the data for the GREB1 and NRIP1 locus are 5 kb and 10 kb resolution, respectively. This is not particularly high for a targeted approach, certainly not in light of the MNase-based approaches that have recently been developed. I therefore think that the "high-resolution" claims should be removed from the paper.

    In line with the reviewer’s suggestion, we have removed the term high-resolution when referring from our own data.

  2. Review Commons

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

    Learn more at Review Commons

    Referee #2

    Evidence, reproducibility and clarity

    In this manuscript, Gómez Acuña and colleagues have investigated changes in enhancer-promoter (E-P) interactions with both 3C and DNA FISH. As a model system, they have used the activation of estrogen receptor-dependent enhancers, which allows for examination of changes in E-P interactions at relatively high temporal resolution. Surprisingly, they find that gene activation is associated with increased E-P interactions as measured by 3C but reduced spatial proximity as measured by DNA FISH. The authors show that both these measurements are dependent on the presence of the enhancer. In contrast, blocking transcription with inhibitors does not have a strong effect on the 3C measurements, but abolishes the increased spatial E-P separation as measured by DNA FISH following estrogen induction.

    Overall, this is an interesting and thought-provoking study. However, the strong conclusions are not fully supported by the data, as explained in further detail below.

    Major comments:

    • The authors make strong claims and although these are generally reasonably well supported by the data, it is important to acknowledge that they are based on two loci. This manuscript would be stronger if the authors could include additional loci in their study design. If this is not possible, it would be good to acknowledge that the conclusions are preliminary/speculative at this stage.
    • It would be helpful if the authors could clarify the strategy they used for their FISH probe design. The enhancer and promoter fosmid probes (which are used for the majority of the experiments) are not centered on the active elements and do not even seem to overlap in the case of the GREB1 enhancer fosmid probe. The 10 kb enhancer probe seems better placed for the GREB1 locus, but the 10 kb enhancer probe does not seem to overlap with the enhancer in the NRIP1 locus. It is conceivable that the exact location of the probes has a big impact on the measurements and it would therefore be helpful if the authors could comment on the location of the probes and add additional probes if required to strengthen their conclusions. In addition, the fosmid probes are very large (40 kb). Although the authors acknowledge this, it would be helpful if they could comment on how overlap between 40 kb probes should be interpreted in relation to a potential rather focal contact between (proteins bound to) regions of <1 kb.
    • It is not clear to me why the authors would choose to work with a locus that is present in 4 copies in their cell line. Is the entire regulatory region (incl. enhancers) preserved for the two additional copies of the gene? Can the authors comment on how this may impact on their measurements?
    • Figure 2D shows an increase in E-P separation for the NRIP1 locus across all timepoints, with cumulative frequency plots shown for the 10 min timepoint. However, the data for the second replicate shown in Figure S2D are a lot less robust and not significant for the 10 min timepoint. It is important that the authors either provide additional data to support the robustness of this experiment or acknowledge that the results are not fully reproducible.
    • The data presented in Figure 2F for clone 2 of the GREB1 enhancer deletion still show increased E-P distance upon activation. How do the authors explain this?

    Minor comments:

    • Could the authors comment on the observation that the NRIP1 promoter is not bound by ERa or p300 upon estrogen activation? Are there ATAC-seq or H3K27ac ChIP-seq data available for these conditions?
    • It is not obvious which timepoint is shown in Figure 1D.
    • Why did the authors choose e-i and p-i instead of e-c and p-c in Supplementary Figure 3B?
    • "We treated hormone starved MCF-7 cells with flavopiridol or triptolide for 5 min before adding E2 for 30 min (Fig. 4A)." Does this mean that the FLV/TRP treatment lasted for 35 min or did the authors wash it out before adding E2? Please clarify.
    • The authors refer to their Capture-C data as "high-resolution". However, the methods section mentions that the data for the GREB1 and NRIP1 locus are 5 kb and 10 kb resolution, respectively. This is not particularly high for a targeted approach, certainly not in light of the MNase-based approaches that have recently been developed. I therefore think that the "high-resolution" claims should be removed from the paper.

    Referees cross-commenting I agree with the comments raised by Reviewer 1

    Significance

    Since 3C and DNA FISH are widely used, the discrepancy between these measurements that is described here is of potential broad interest to the field. Since these claims are rather strong and have potential far-reaching implications, it would be helpful if the authors could strengthen their conclusions further, by improving the robustness of the data and including additional loci and additional probes to show that the measurements are not specific for these two loci or dependent on the location of the probes. I think that the paper is in principle also publishable without these additional experiments, but in that case, it would be very important to explicitly acknowledge the limitations of the data throughout the manuscript and clarify that the conclusions are preliminary/speculative at this stage.

    Expertise: 3D genome organization.

  3. Review Commons

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

    Learn more at Review Commons

    Referee #1

    Evidence, reproducibility and clarity

    Gómez Acuña et al.

    Transcription decouples estrogen-dependent changes in enhancer-promoter contact frequencies and spatial proximity

    Summary

    This study by Gómez Acuña et al. addresses a highly relevant question in the field of gene regulation: what are the mechanisms by which enhancers and promoters communicate to control expression of a target gene? This work tackles an apparent dichotomy as to whether enhancers (E) and promoters (P) need to come into close physical proximity in order to activate gene expression. While 3C-based methods have consistently suggested that the frequency of 'contacts' between enhancers and promoters increases during gene activation, imaging approaches have, in some instances, shown that these elements move further away from each other during this process. Gómez Acuña et al. have used 3C- and imaging-based methods in parallel at two inducible genes in the MCF-7 breast cancer cell line to confront this issue.

    The study focusses on two oestrogen-responsive genes, GREB1 and NRIP1, which are fully activated within 1 h of oestradiol (E2) treatment of MCF-7 cells in in vitro culture. Using a capture-based 3C method, the authors show small increases in contacts between putative enhancers and promoters during gene induction. In contrast, for both genes they also observe increased distances between these elements by DNA FISH. The authors used CRISPR-Cas9 to delete the putative nearby enhancers and showed that these phenomena are likely dependent on these elements.

    Next, the authors questioned whether recruitment of the oestrogen receptor (ER) itself or co-activators recruited by ER upon stimulation are responsible for the observed effects. By comparing E2 with tamoxifen (4OH), which recruit co-activators and -repressors, respectively, they only observed the effects on E-P contact and proximity in the presence of the activating stimulus (E2), suggesting recruitment of activating protein machinery is required.

    Finally, they tested the effects of small molecules known to inhibit different stages of the transcription cycle. Both flavopiridol and triptolide abrogated the effect on E-P proximity observed by DNA FISH but did not affect the increase in E-P contacts seen by C-TALE. This suggested that these phenomena can be uncoupled, and that the act of transcription itself (and perhaps the associated production of RNA) is required for the physical separation of enhancers and promoters but not the increase in E-P contact frequency.

    Major comments

    Given the significance of the question posed in this study (see Significance section), it is important that the claims highlighted in this paper are appropriately supported by the experimental evidence. There are a number of issues with the design of the experiments in the current manuscript which are as follows:

    1. The aneuploid karyotype of the MCF-7 cells used is a concern. GREB1 is present in four copies, with two on abnormal chromosomes which may not be regulated in the same way as primary cells. The authors should include caveats to this effect in the text to account for this.
    2. The authors should also include more information on the generation and verification of the enhancer deletion cell lines. An illustration of the PCR primers used for screening, as well as an illustration of the sequenced product traces aligned with the reference genome (as opposed to just showing the deleted regions) should be included in Fig. S1D. This would give the reader more confidence that the designed knockout has occurred in the same way on all alleles. Furthermore, long-range PCRs and sequencing should be considered to confirm that no larger deletions have occurred (e.g. Owens et al., 2019 PMID: 31127293).
    3. The changes in the measured E-P interaction frequency following gene activation are weak at best and make visual interpretation of the results difficult. Showing the reciprocal virtual 4C plots from the promoter would help to reassure the reader that the observed effect is real. Furthermore, the precise 3C method used is not clear. The authors repeatedly refer to "Capture-C" (a commonly used 3C-based approach using biotinylated oligos to pull down targets of interest) but the citation used (Golov et al. 2019) refers to a conceptually similar method called "C-TALE". This should be clarified in the text.
    4. As for the changes in contact frequency, the observed changes in distance measurements between conditions are very small (although statistically significant). We acknowledge that this is likely due to the relatively small linear distances between enhancers and promoters in this study. However, it would be helpful to see the effects of the induction/treatments on a one or more control loci which is not affected by oestrogen signalling given that global changes in nuclear shape/volume and/or cell cycle effects could occur within this time (e.g. effects of tamoxifen treatment on MCF-7 cell cycle distribution, (Osborne et al. 1983 PMID: 6861130), which could impact nuclear volume.
    5. The authors discuss previous studies demonstrating that E2 and 4OH recruit different sets of proteins to their target genes. Given that this is central to the conclusion that the ER ligand (and its recruited co-factors) determines the E-P interaction frequency and 3D distances observed, it would be important to demonstrate this at the GREB1/NRIP1 loci specifically. ChIP data of the co-activators/repressors recruited by E2 and 4OH, respectively, would greatly strengthen this claim.
    6. The observed uncoupling of E-P contact frequency and 3D distance upon transcriptional inhibition is interesting and offers clues to the molecular details underlying E-P interactions. However, the use of flavopiridol and triptolide, while common in the field, should be carefully qualified given the potential for their indirect effects on transcription. This is particularly important for flavopiridol given its ability to target multiple cyclin-dependent kinases beyond CDK9 and its role in transcription initiation.

    Minor comments

    1. In the introduction and beginning of discussion, it would be helpful to detail previous studies where FISH-based analyses have shown more proximal E-P positioning upon activation, to make it clear that differences in E-P proximity appear to be gene-specific. Some examples include Williamson et al. (2016; PMID 27402708) and Chen et al. (2018; PMID 30038397). Speculation as to why some genes behave in this way while others do not, would also be worthwhile.
    2. On page 6, the authors state that after deletion of the NRIP1 enhancer there is "almost total loss of NRIP1 induction in response to E2". This does not seem to match the data where in 3 out of 4 replicates (2 for each clone) there is a statistically significant increase in number of RNA FISH foci upon E2 stimulation in the NRIP1 enhancer KOs. This suggests that, as for GREB1, the regulation of these genes is not solely controlled by the deleted enhancers. This should be clarified in the text.
    3. The labelling of the FISH probes in Supp. Fig. S2 could be improved as it is currently very difficult to read these.
    4. Given that the authors have referenced a distance of 200 nm as potentially being an important threshold for gene activation, it would be useful to include the fraction of alleles which are below this distance alongside the cumulative frequency plots in Figure 2D and elsewhere in the paper as the cumulative frequency plots can be hard to read in some cases (e.g. Supp. Fig. S3B e-p). This would also allow the authors to show consistency across replicates.
    5. For clarity, it would be helpful to include the difference map between the vehicle-treated unstimulated/stimulated conditions for the 3C plots in Fig. 4. This would help contextualise the resulting differences observed with the drug treatments. Same for Supp. Fig. S6.
    6. Statistical comparisons are not shown for all 3D FISH-based distance measurements (e.g. Supp. Figs. S3A, S4C, D, S6E). If this is because the tests were done and the results were non-significant this should be indicated.
    7. On page 13, the authors state that increased E-P separation occurs "before nascent transcription of the gene is detected by either TT-seq or RNA FISH". This does not appear to be correct given that baseline levels of transcription are observed in the absence of ER stimulation by both methods (Fig. 1). This should be clarified in the text.

    Referees cross-commenting

    Reviewer #2 is in good agreement with our review. The study is interesting but there are several caveats that need addressing as pointed out by both reviewers. We agree that it could take a lot of work to sort these points out in full experimentally but without this the authors should be careful not to overinterpret their data and comment on the shortcomings as it stands.

    Significance

    The question the authors are addressing here is of importance to the field. For several years, researchers have been attempting to uncover the mechanisms by which enhancers deliver information to their target promoters and this study highlights some potentially fundamental issues with the way in which these problems are typically addressed. It is generally accepted that in most cases, the frequency of 'contacts' between enhancers and promoters as measured by 3C-based methods increases as a gene becomes active. It is still unclear what exactly these 3C methods are measuring given that the radius of crosslinking and the extent to which protein-protein linkages between two DNA helices contribute to the observed contact frequency are unknown. In short, despite its widespread use, it is not clear whether contact frequency as measured by 3C is proportional to three-dimensional distance or something more nebulous. The issues raised in this paper are therefore critical for the accurate conception of mechanistic models of gene regulation.

    The use of a well-studied inducible system to enable sampling of gene activation events at high temporal resolution means that the authors can ensure a level of control over the experimental conditions. The experiments appear to have been carried out to a high standard, with necessary controls generally included where appropriate. The analyses are well-documented and the results are consistent over the various reported experiments within the manuscript. The conclusions are thought-provoking and will challenge the field to be clearer about what 3C-based methods can and cannot show. This study will also surely lead to follow-up work to more specifically address the possible molecular mechanisms explaining the above observations. The main limitations of this study are the cell line and genes studied, with GREB1 being present in four copies in the aneuploid genome of MCF-7 cells. As stated above, there is some concern that using cells with such a chaotic genome as a starting point for investigating E-P interactions might prevent the authors from drawing clear and unambiguous conclusions. Perhaps related to this, the effect size of the E-P interactions upon stimulation as measured by C-TALE/Capture-C are underwhelming to say the least. Particularly for NRIP1, it was sometimes hard to tell where to look to see the supposed increase in E-P interactions, even with arrows as guidance.

    Despite these limitations, this work raises an important issue that needs further clarification. It will require the field to reconsider what 3C-based analyses actually tell us with respect to E-P contact/proximity, and as a result models of E-P communication may need revising. This work is of broad interest to those work in the fields of nuclear biology and more specifically gene regulation. This will include both basic research scientists as well as those attempting to exploit enhancer-promoter dependencies for therapeutic purposes.

    Expertise

    Gene regulation, genome organisation and 3D structure.

  4. Version published to 10.1101/2023.03.29.534720v2 on bioRxiv
  5. Version published to 10.1101/2023.03.29.534720v1 on bioRxiv