The anti-cancer drugs curaxins target spatial genome organization

Recently we characterized a class of anti-cancer agents (curaxins) that disturbs DNA/histone interactions within nucleosomes. Here, using a combination of genomic and in vitro approaches, we demonstrate that curaxins strongly affect spatial genome organization and compromise enhancer-promoter communication, which is necessary for the expression of several oncogenes, including MYC. We further show that curaxins selectively inhibit enhancer-regulated transcription of chromatinized templates in cell-free conditions. Genomic studies also suggest that curaxins induce partial depletion of CTCF from its binding sites, which contributes to the observed changes in genome topology. Thus, curaxins can be classified as epigenetic drugs that target the 3D genome organization.

important because cells are incubated in curaxin for 6 hours, which could result in the depletion of many RNAs and proteins, leading to the observed changes in 3D organization. Authors should demonstrate that the observed effects are not an indirect consequence of transcription changes causing depletion of proteins involved in 3D organization, rather than a direct effect of curaxin on this process. 8. Page 6. To analyze the effect of curaxin on enhancer-promoter interactions, authors use PSYCHIC, which is a recently published program whose name does not inspire a lot of confidence when one is looking for statistical significance in the outcome. Does PSYCHIC determine significant interactions in the Hi-C data before finding interactions between enhancers and promoters? Since the authors only obtained approximately 150 million contacts after merging replicates, which is quite low for a mammalian genome, it is possible that the resolution in determining enhancerpromoter interactions is very low. This is not mentioned in the description. Are the authors able to assign each interaction to a single enhancer-promoter pair? Are interactions between enhancers or between promoters also affected, or were all these considered together? A more conventional way of doing this analysis would be to identify significant interactions using FitHi-C, for example, and then examine the effect of curaxin on these interactions after appropriate normalization.
Reviewer #2 (Remarks to the Author): The authors studied the mechanism of a previous found drug (CBL0137) on supressing cancer cells. Through in vitro and in vivo studies, they proposed a model where the drug disrupts the interaction between super-enhancers (SE) and promotoers, and causes gloabl 3D genome alterations at the TAD, loop and compartment levels. The specific effect of the drug on downregulating MYC may contribute to its anti-cancer effect.
Specific comments: 1. The manuscript is short and formatted as a "Brief communications" type, rather than a typical NC style. The authors should consider to rewrite and add more details to comply to the NC format.
2. The drug causes a lot of changes in the 3D genome. How are the TAD and compartment changes correlate with gene expression changes?
3. The bromodomain inhibitors also disrupt SE functions and repress the expression of nearby genes. How does CBL0137 compare to these drugs in terms of drug efficacy, mechanism and the effect on the 3D genome?
Reviewer #3 (Remarks to the Author): Kantidze et al. report that the drug curaxin CBL0137 affects genome organization by reducing intra-TAD interactions, which in turn promotes decreased expression of enhancer-regulated genes. CBL0137 is able to destabilize nucleosomes and stabilize Z-DNA formation, as shown by the same authors in a previous work, which might cause the reduced enhancer-promoter communications observed both in the chromatinized system and in cells by Hi-C. Based on these findings the authors present CBL0137 as a novel type of epigenetic drugs affecting genome architecture.
Overall, the manuscript provides interesting insights about the drug CBL0137 and its effect on chromatin. If the authors adequately address the comments outlined below, I recommend consideration of the manuscript for publication.
Major comments: 1) Kantidze et al. claim that CBL0137 suppresses enhancer-dependent transcription of many genes. By RNA-seq and WB the authors show that MYC expression is strongly decreased after drug treatment. In 2012, the groups of R. Young and D. Levens demonstrated that MYC is a "transcription amplifier". It binds promoters and enhancers of active genes stimulating their expression exponentially. If MYC expression is strongly affected by CBL0137, all "MYC targets" (i.e. genes with open promoters bound by RNA polymerase) will be affected, which would explain the decreased expression of the high output promoters. Also, the group of R. Casellas has recently shown that reduction of MYC levels affects genome organization and promoter-enhancers interactions (Kieffer-Kwon KR, et al. Molecular Cell 2017). If the main point of the manuscript is to demonstrate the ability of CBL0137 to directly change the structure of the genome (as stated in the abstract) the authors need to rule out the possibility that this is a secondary effect of reduced MYC levels (for example, by testing the effect of CBL0137 on MYC independent genes).
2) Results in Figure 1C are a bit unclear. The expression of the reporter gene regulated by NF-Kβ responsive element is affected by CBL0137 but not the one carrying the MYC promoter. The authors claim that this because of a "different genomic context". I could not find information about the features of the MYC minimal promoter (in this paper or in the Young's paper cited). Does the MYC minimal promoter contain the FUSE element? FUSE regulates the MYC promoter (Lui J, et al. EMBO 2006) and it is susceptible to negative supercoiling. Since CBL0137 preferentially binds underwound DNA, it can affect FUSE melting. Therefore the absence of the FUSE sequence might explain why the expression of the reporter gene is not affected. This possibility need to be investigated.
3) Kantidze et al. show that CBL0137 changes the structure and the properties of the linker DNA increasing the "fraction of nucleosomes with a larger distance between the linkers". This finding supports a previous observation where the same authors demonstrated that CBL0137 destabilizes nucleosomes. Nucleosome destabilization should lead to a more open and flexible chromatin conformation and thus facilitate Enhancer-Promoter interactions. Follow this logic it is a bit unclear how the drug can promote decreased gene expression by favoring nucleosome destabilization. Is the chromatinized template characterized by Z-DNA forming sequences? Z-DNA could potentially form when transcription is turned ON and this might interfere with the EPC.
4) The presentation of the Hi-C data could be improved. For example, it would be nice to see a graph showing the effect of CBL0137 on segregation of A/B compartments as a function of gene density or gene expression (using the RNA-seq data already available). These analyses will support the overall message of the paper.
Minor points: 1) Introduction: Can the authors explain what are the "MYC family of genes"? Are these MYCtarget genes? The authors might want to clarify the sentence.
2) Page 3 (2nd, 3rd line): According to Figure 1E or 1F (cited in the text), I think the comparison should be between super-enhancers and typical enhancers or presence vs. absence of enhancers. Please change the sentence. Figure 2B: Based on the micrococcal nuclease assay, the authors claim CBL0137 does not affect the nucleosome structure. I am not fully convinced by the figure. The overall quantity of DNA appears to be different in individual samples (see for instance the background of lanes 4 and 6). Thus, the question is: how do the authors check for equal loading among samples in the gel? This is particular important as differences are modest. Also, the authors mighty want to clarify which is the concentration that "strongly affects EPC".

Reviewer #1 (Remarks to the Author):
The manuscript explores the effects of curaxin, a presumed anticancer drug, on gene expression.  The majority of plasmids preserved their supercoiled conformation in course of this transfectionextraction experiment». The photo of the gel is presented below for evaluation by the reviewer. (Fig. 1c) -See the answer to the previous comment. We also would like to stress the attention that in the sentence cited by the reviewer we use the term "suggests" that acknowledges a possibility of other interpretations. Figure 2f and  The pdf file is attached to allow in-depth evaluation by the reviewers.

4.
Moreover, the distances between the linkers have been studied by others using FRET previously (Toth et al., 2001; PMID:11389607). According to this work, inter-linker distance (end-to end distance) varies from 60Å for a 150-bp DNA template to 75Å for 170-bp DNA template organized into a nucleosome. Figure 3 could be due to differences in library quality. Authors should show information on quality control steps for the different biological replicates in a supplemental table. In particular, it would help to know if the number of intra-and inter-chromosomal interactions as wells as short range (less than 20 kb) versus long range (more than 20 kb) is the same for each replicate of each sample.

Figure 3 and page Differences in the Hi-C maps observed in various panels of
-The requested information is presented in the Supplementary Table 2   induces partial depletion of CTCF from its binding sites") of the Results section: "CBL0137 induces partial depletion of CTCF from its binding sites. In vertebrates, CTCF, cohesin, and condensin almost exclusively maintain spatial genome organization5. In an attempt to uncover mechanisms underlying effects of CBL0137 on chromatin structure, we have analyzed whether it affects abovementioned architectural factors. We have shown that treatment of HT1080 cells with CBL0137 for 6 hours did not alter protein levels of CTCF, and subunits of cohesin and condensin complexes (Rad21 and SMC2, respectively) (Fig. 6a). Next, we analyzed the distribution of these proteins in different chromatin fractions obtained from control and CBL0137treated HT1080 cells and found that CBL0137 treatment led to a redistribution of CTCF, but not cohesin and condensin, from the fraction of proteins strongly associated with chromatin (Fig. 6b). This might reflect the CTCF dissociation from its binding sites upon curaxin treatment. To test this assumption directly, we have analyzed the genomic distribution of CTCF in control and CBL0137-treated cells using chromatin immunoprecipitation-sequencing assay (ChIP-seq). In control HT1080 cells, about 45 000 CTCF-enriched peaks were mapped using PePr computational approach 43 that is in a good agreement with the previously published data 44. Moreover, the positions of the peaks are almost the same to those available from ENCODE consortium (see an example of CTCF distribution in HT1080 (our data) versus HeLa S3 (ENCODE) cells on Supplementary Fig. 8). Being a chromatin loop-organizing factor of crucial importance, in control HT1080 cells CTCF is strongly enriched at loop anchor regions (Fig. 6c). Upon CBL0137 treatment, some portion of the CTCF peaks present in control cells disappears (Fig. 6c). Genome-wide, CTCF depletes from as many as ~40% of initially found peaks (Fig. 6d, e). The results suggest that curaxins-induced partial dissociation of CTCF from its binding sites may underlie the changes in 3D genome organization observed." 7. Figures 5f and 5g,  These results are shown in Figure 6 of the revised MS. Following the reasoning of the reviewer we also considered a possibility that suppression of transcription per se can affect the 3D genome organization. This possibility was excluded by demonstration that the effect of CBL0137 on 3D genome organization was equally pronounced in transcribed areas and in gene deserts (Fig. 5 of the revised MS).

Page 6. To analyze the effect of curaxin on enhancer-promoter interactions, authors use
PSYCHIC, which is a recently published program whose name does not inspire a lot of confidence when one is looking for statistical significance in the outcome. Does PSYCHIC determine significant interactions in the Hi-C data before finding interactions between enhancers and promoters? Since the authors only obtained approximately 150 million contacts after merging replicates, which is quite low for a mammalian genome, it is possible that the resolution in determining enhancer-promoter interactions is very low. This is not mentioned in the description. Are the authors able to assign each interaction to a single enhancer-promoter pair? Are interactions between enhancers or between promoters also affected, or were all these considered together? A more conventional way of doing this analysis would be to identify significant interactions using FitHi-C, for example, and then examine the effect of curaxin on these interactions after appropriate normalization.
-There are indeed various algorithms for annotating spatial contacts between distant genomic elements. We preferred PSYCHIC because this algorithm is focusing on identifying intra-TAD contacts. We explained this in the revised MS: "In contrast to other related techniques like HiCCUPS 4 and Fit-Hi-C 40 , PSYCHIC-mediated annotation of promoter-enhancer interactions is TAD-specific39. Thus, the data obtained by PSYCHIC are generally more accurate and is not skewed by TAD boundary elements 39 ".
We were not able to assign each interaction to a single enhancer-promoter pair and did not aim to discriminate enhancer-promoter, enhancer-enhancer and promoter-promoter interactions. This may be a subject of another study. Here our aim was rather to show that CBL0137 exert a strong general effect on the 3D genome.

Reviewer #2 (Remarks to the Author):
The authors studied the mechanism of a previous found drug (CBL0137) on supressing cancer

cells. Through in vitro and in vivo studies, they proposed a model where the drug disrupts the interaction between super-enhancers (SE) and promotoers, and causes gloabl 3D genome alterations at the TAD, loop and compartment levels. The specific effect of the drug on downregulating MYC may contribute to its anti-cancer effect.
Specific comments: 1. The manuscript is short and formatted as a "Brief communications" type, rather than a typical NC style. The authors should consider to rewrite and add more details to comply to the NC format.
-The revised MS is written according to Nature Communication format.

The drug causes a lot of changes in the 3D genome. How are the TAD and compartment changes correlate with gene expression changes?
-CBL0137 strongly suppress transcription, especially transcription of genes controlled by enhancers and super-enhancers (Fig. 1d-e of the revised MS). It is, however, difficult to discriminate the effect caused by changes in 3D genome from other effects related, for example, to trapping of chromatin remodeler FACT. On the other hand, our data strongly suggest that effect of CBL0137 on 3D genome is direct as (i) it can be observed in both transcribed and nontranscribed areas (Fig. 5 of the revised MS) and (ii) it is not due to the depletion of major architectural proteins that might be caused by transcription suppression (Fig. 6 of  -In Fig. 5 of the revised MS we present the evidence that CBL0137 strongly affects the 3D genome organization even in non-transcribed areas (gene deserts). Additionally, we demonstrate that exposure of cells to CBL0137 does not cause a depletion of major architectural proteins known to be essential for the establishing and maintaining of 3D genome (Fig. 6 a,b of the revised MS). These two lines of evidence disagree with a supposition that changes of 3D genome organization in cells exposed to CBL0137 originate as a consequence of transcription suppression and strongly support a conclusion that effect of CBL0137 on 3D genome is direct. Figure 1C  In living cells, CBL0137 exerts numerous effects on chromatin described in references 27-28. As mentioned in Discussion section all these effects can influence the long-range configuration of a chromatin fiber and thus compromise enhancer-promoter communication:

2) Results in
"Modulation of chromatin fiber flexibility may be sufficient to modify the 3D organization of extended genomic segments and thus affect the EPC 48 ". However, in the revised MS we also present the evidence that direct effect of CBL0137 on 3D genome organization may be due to displacement of CTCF from some of the binding sites (see Fig. 6 and related text in the Results section).

4) The presentation of the Hi-C data could be improved. For example, it would be nice to see a graph showing the effect of CBL0137 on segregation of A/B compartments as a function of gene
density or gene expression (using the RNA-seq data already available). These analyses will support the overall message of the paper.
-Although it is clear that exposure of cells to CBL0137 decrease the segregation of A/B compartments it is not possible to present the graph proposed by the reviewer because there is no way to take into account the size of uninterrupted compartments. Of course, it is possible to present individual bins. However this will be counterintuitive because the neighboring bins should influence interactions at the compartment level. -In the first paragraph of the RESULTS section the phrase mentioning MYC gene family was modified as follows: "Expression of MYC family genes (c-MYC, NMYC, and LMYC 29 ) at both mRNA and protein levels is highly suppressed by CBL0137 in various cell lines, independently on whether they contained wild-type MYC gene locus or translocated MYC (Fig. 1a-b)" -The Figure 1e was corrected (the comparison was made between genes controlled by enhancers (or super-enhancers in case of MM1.S) and genes lacking enhancers. The text of the first paragraph of results section was also corrected as follows: "Moreover, all genes regulated by enhancers (HT1080) or SEs (MM1.S) in these cells were inhibited by CBL0137 stronger and at lower concentrations than genes lacking remote enhancers (Fig. 1e-f)." Figure 2B: Based on the micrococcal nuclease assay, the authors claim CBL0137 does not affect the nucleosome structure. I am not fully convinced by the figure. The overall quantity of DNA appears to be different in individual samples (see for instance the background of lanes 4 and 6). Thus, the question is: how do the authors check for equal loading among samples in the gel? This is particular important as differences are modest. Also, the authors mighty want to clarify which is the concentration that "strongly affects EPC".

3) Supplementary
-The apparent differences in DNA loading in Fig. S2B are primarily due to different labeling of MNase-digested DNA by the protein kinase. We carefully controlled the amount of material used in the experiment by measuring the absorbance at 260 nm (before digestion with MNase and controlled the extent of digestion by monitoring the ratios between the different bands in the gel. The ratios between the bands directly reflect the extent of the digestion. Quantitative analysis of the data in Fig. S2B is shown below and is included in the revised version of the figure: The gel shown in A (corresponds to the original Fig. S2B) was quantified using a PhosphorImager. Amount of label present in each band was calculated as % of label present in all four bands (1-nucleosome etc.). The distribution of the label between the bands is very similar in all lanes, suggesting that the chromatin was digested by MNase to a similar extent.
Curaxins strongly affects EPC in vitro when present at concentrations of 1 or 2.5 uM (Fig. 2d). The concentrations are now described in the text. The corresponding phrase is modified as follows: "We found that addition of CBL0137 (1-2.5 µM) causes a strong decrease in the yield of the transcript on the both chromatinized and free DNA model construct (Fig. 2c-d)."