The Mediator complex regulates enhancer-promoter interactions

Enhancer-mediated gene activation generally requires physical proximity between enhancers and their target gene promoters. However, the molecular mechanisms by which interactions between enhancers and promoters are formed are not well understood. Here, we investigate the function of the Mediator complex in the regulation of enhancer-promoter interactions, by combining rapid protein depletion and high-resolution MNase-based chromosome conformation capture approaches. We show that depletion of Mediator leads to reduced enhancer-promoter interaction frequencies, which are associated with a strong decrease in gene expression. In addition, we find increased interactions between CTCF-binding sites upon Mediator depletion. These changes in chromatin architecture are associated with a redistribution of the Cohesin complex on chromatin and a reduction in Cohesin occupancy at enhancers. Together, our results indicate that the Mediator and Cohesin complexes contribute to enhancer-promoter interactions and provide insights into the molecular mechanisms by which communication between enhancers and promoters is regulated.

Article https://doi.org/10.1038/s41594-023-01027-2 resolution [24][25][26][27][28][29] . It is therefore possible that changes in fine-scale genome architecture, including enhancer-promoter interactions, could not be reliably identified. For a better understanding of the function of the Mediator complex in genome regulation, it is important to examine the impact of acute Mediator perturbations on chromatin architecture with high resolution and sensitivity.
Here, we overcome limitations of current studies and investigate the function of the Mediator complex by combining rapid protein depletion and high-resolution analysis of genome architecture using both conventional and MNase-based 3C approaches. We find that depletion of Mediator leads to a significant reduction of enhancer-promoter interactions. Interestingly, we also find that Mediator depletion causes increased interactions between CTCF-binding elements. We show that these changes in interaction patterns are associated with a redistribution of the Cohesin complex on chromatin and a loss of Cohesin occupancy at enhancers upon Mediator depletion. These results suggest that enhancer-promoter interactions are dependent on both Mediator and Cohesin and provide support for a model in which the Cohesin complex bridges and stabilizes interactions between enhancers and promoters bound by Mediator.

Mediator depletion causes changes in chromatin interactions
Because the MED14 subunit acts as a central backbone that connects the Mediator head, middle and tail modules 19 , its degradation disrupts the integrity of the Mediator complex 27,28 . We have therefore used an HCT-116 MED14-dTAG cell line 28 to study the function of the Mediator complex in genome regulation. Using immunoblotting (Extended Data Fig. 1a,b) and chromatin immunoprecipitation and sequencing (ChIP-seq) (Extended Data Fig. 1c-i), we have confirmed efficient MED14 depletion within 2 h of treatment with a dTAG ligand.
Previous work has shown that Mediator depletion leads to strong downregulation of cell-type-specific genes that are associated with super-enhancers 28 (Extended Data Fig. 1j). Super-enhancers are stretches of clustered enhancers with high levels of Mediator that are thought to have a central role in driving high expression levels of key cell identity genes 30 . Previous studies could not detect changes in interactions between promoters and (super-)enhancers upon Mediator depletion [27][28][29] . However, these studies relied on genome-wide 3C approaches, such as Hi-C and Hi-ChIP, with relatively low resolution (4-5 kb). It is therefore possible that small-scale changes in enhancer-promoter interactions could not be reliably detected.
To investigate changes in genome architecture upon Mediator depletion in more detail, we used targeted 3C approaches, which are not limited by sequencing depth and can detect changes in genome structure at high resolution and with high sensitivity. We focused our analyses on 20 genes (Extended Data Fig. 1j), which we selected on the basis of the following criteria: (1) robust gene activity in HCT-116 cells; (2) significant downregulation of gene expression upon Mediator depletion; (3) high Mediator occupancy at the gene promoter; and (4) association with a super-enhancer. We initially used Capture-C 31,32 , a targeted 3C method based on DpnII digestion, to evaluate changes in chromatin interactions with the promoters of these genes. Capture-C interaction profiles display interaction frequencies with selected viewpoints per DpnII restriction fragment and therefore have an average resolution of ~250 bp. By comparing Capture-C data generated in HCT-116 MED14-dTAG cells treated with DMSO or dTAG ligand, we find that Mediator depletion leads to subtle changes in the interaction patterns of the selected gene promoters ( Fig. 1 and Extended Data Fig. 2a-d). Unexpectedly, we observe patterns of both decreased and increased interactions.
For example, in the MTAP locus, the Capture-C data show reduced interactions in the upstream region, in which two super-enhancers are located, and a trend towards increased interactions in the regions further upstream and downstream (Fig. 1a). In the region containing tail module of the Mediator complex interacts with the activation domains of transcription factors bound at enhancers and the head and middle modules interact with the pre-initiation complex (PIC) at gene promoters [17][18][19] , it has been proposed that Mediator acts as a bridge between enhancers and promoters (reviewed in refs. [20][21][22][23]. Initial studies based on knockdown of Mediator subunits over the course of several days provided evidence for this hypothesis [24][25][26] . However, since the Mediator complex has a central function in RNA polymerase II (Pol II)-mediated transcription, its long-term perturbation causes secondary, confounding effects, which complicate the interpretation of these early studies.
To overcome these limitations, more recent studies have used rapid protein-depletion strategies to investigate the function of the Mediator complex in gene regulation and genome organization 27,28 . These studies did not detect changes in chromatin architecture and enhancer-promoter interactions upon Mediator depletion, despite strongly reduced expression levels of enhancer-dependent genes. On the basis of these findings, it has been concluded that Mediator is dispensable for enhancer-promoter interactions and acts as a functional rather than an architectural bridge between enhancers and promoters 27,28 .
A caveat of current studies of the role of Mediator in genome architecture is that enhancer-promoter interactions have been assessed with chromosome conformation capture (3C) methods at relatively low

Fig. 1 | Changes in chromatin interactions upon Mediator depletion.
a, Capture-C interaction profiles from the viewpoint of the MTAP promoter in HCT-116 MED14-dTAG cells treated with DMSO (dark blue; n = 3 biologically independent samples) or dTAG ligand (light blue; n = 3 biologically independent samples). Gene annotation, DNase-hypersensitive sites (DHS) and ChIP-seq data for CTCF and MED26 are shown above and a differential profile (Δ DMSO − dTAG) is shown below. Super-enhancers are highlighted in green below the MED26 profiles and orientations of CTCF motifs are indicated with arrowheads (forward orientation in red; reverse orientation in blue). The gray box highlights a broad reduction in interactions in the region containing super-enhancers in dTAG-treated cells. The axes of the DHS and ChIP-seq profiles are scaled to signal; the axes of the Capture-C profiles are fixed (ranges indicated in brackets). Coordinates (hg38): chr9:21,096,000-22,491,000. b, Data are as described in a, but for the HMGA2 locus. Coordinates (hg38): chr12:65,260,000-66,115,000.
Article https://doi.org/10.1038/s41594-023-01027-2 the HMGA2 oncogene, there are fewer interactions in the upstream region, which contains two super-enhancers, whereas interactions in the downstream region are increased (Fig. 1b). We observe similar patterns in other loci we investigated (Extended Data Fig. 2a-d). However, in regions containing genes that are not highly expressed in HCT-116 cells and are not sensitive to Mediator depletion, we do not see clear changes in interaction patterns (Extended Data Fig. 2e,f).

Depletion of Mediator reduces enhancer-promoter interactions
To examine the broad changes in the Capture-C interaction profiles in further detail, we performed Micro-Capture-C (MCC) experiments 33 in DMSO-and dTAG-treated HCT-116 MED14-dTAG cells, using viewpoints targeting the same set of gene promoters. Compared with Capture-C, MCC has an advantage in that it uses MNase instead of DpnII for chromatin digestion. The resolution of MCC is therefore not limited by the distribution of DpnII cut sites across the genome, enabling analysis at base-pair resolution 33 . The MCC data resolve the broad interaction patterns in the Capture-C data and clearly show that Mediator depletion leads to reduced interactions between gene promoters and Mediator-bound enhancer regions in the MTAP and HMGA2 loci (Fig. 2a,b).
We find that depletion of Mediator leads to a decrease in the frequency of enhancer-promoter interactions in the 20 regions that we focused on (Extended Data Fig. 2g-j). Quantification of the MCC interactions between gene promoters and clusters of Mediator-bound enhancers indicates an average reduction of 22% across these regions (Fig. 2c). The reduction in interaction frequency between the promoters and a narrow region covering the largest Mediator peak within these broad clusters is, on average, 34% (Fig. 2d). These changes are associated with an average decrease in gene expression of 7.5-fold (Extended Data Fig. 1j). Of note, the Capture-C data also detect a reduction in enhancer-promoter interactions in most regions of interest, with an average decrease in interaction frequency of 9% (Fig. 2e). Although it is in accordance with the MCC data, this comparison highlights the need for analyses with sufficient resolution and sensitivity to robustly detect changes in enhancer-promoter interactions.

CTCF-dependent interactions increase upon Mediator depletion
The MCC data do not only identify specific reductions in interactions with enhancers, but also uncover very precise increased interactions following depletion of Mediator. Strikingly, these increased interactions all overlap with CTCF-binding sites. For example, in the CTCF-dense MTAP locus, we see strong increases in interactions formed with CTCF-binding sites in the region upstream of the super-enhancers and downstream of the gene promoter (Fig. 2a). Notably, the interacting CTCF-binding sites upstream are all in a forward orientation, whereas the interacting CTCF-binding sites downstream are all in a reverse orientation.
We observe a similar pattern of increased interactions with convergently orientated CTCF-binding sites in the MYC locus after Mediator depletion (Extended Data Fig. 2g). In the HMGA2, ITPRID2, ERRFI1 and KRT19 loci, which contain fewer CTCF-binding sites, the patterns are a bit more subtle, but also clearly present ( Fig. 2b and Extended Data Fig. 2h-j).
It has been suggested that MNase-based 3C data could be biased by varying chromatin accessibility and MNase digestion efficiency across regions or conditions. However, the fact that we detect a significant decrease in enhancer-promoter interactions in both the MCC and the Capture-C data, which are generated with restriction enzyme digestion, indicates that reduced enhancer-promoter interactions after depletion of Mediator are unlikely to reflect underlying changes in chromatin accessibility. In addition, the observation of both decreased and increased interactions following depletion of Mediator, with increased interactions specifically overlapping with CTCF-binding sites in a convergent orientation, indicates that it is improbable that the MCC data are skewed by nucleosome positioning. To further demonstrate that the   Fig. 3). These experiments show that Mediator depletion does not lead to strong changes in chromatin accessibility in our regions of interest. Together, these observations indicate that the changes in enhancer-promoter interactions detected by MCC reflect bona fide changes in chromatin architecture.

Mediator depletion causes changes in intra-TAD interactions
The MCC data show clear and precise changes in chromatin interactions upon depletion of the Mediator complex. However, since the MCC viewpoints are very narrow and focused on gene promoters, it remains unclear how large-scale 3D genome architecture is changed, and how interactions between other cis-regulatory elements are impacted by Mediator depletion. We therefore used the Tiled-MCC approach, in which MCC library preparation is combined with an enrichment strategy based on capture oligonucleotides tiled across large genomic regions of interest 14 , to investigate changes in genome architecture in DMSO-and dTAG-treated HCT-116 MED14-dTAG cells in a broader context. We focused on the MYC (3. In line with previous studies that have used Hi-C or Hi-ChIP to examine changes in genome architecture 27-29 , we do not detect drastic changes in large-scale genome organization after Mediator depletion. We find that TAD organization is preserved, without any shifts in the location of boundaries. However, we find subtle changes in interaction  Article https://doi.org/10.1038/s41594-023-01027-2 patterns within TADs. In line with the Capture-C and MCC data, we observe that enhancer-promoter interactions are reduced after Mediator depletion. In addition, we detect strengthening of interactions anchored at CTCF-binding sites. As a result, we see subtle increases in 'looping' between the CTCF-bound anchors of TADs and sub-TADs.

Cohesin binding patterns are altered upon Mediator depletion
It has been shown that CTCF and Cohesin co-localize and that interactions between CTCF-binding sites are formed via loop extrusion by the Cohesin complex [35][36][37][38] . Notably, Cohesin also co-localizes with Mediator, and co-immunoprecipitation experiments have suggested that these complexes interact 24,25,39 . However, a functional link between Mediator and Cohesin has not been identified.
Because our data show that depletion of Mediator causes a decrease in enhancer-promoter interactions and an increase in CTCF-mediated interactions, we hypothesized that these altered interaction patterns could be explained by changes in the distribution of the Cohesin complex on chromatin. To test this, we mapped Cohesin occupancy using cleavage under targets and tagmentation (CUT&Tag 40 ) in DMSO-and dTAG-treated HCT-116 MED14-dTAG cells ( Fig. 4 and Extended Data Fig. 7). These data show clear changes in Cohesin occupancy upon Mediator depletion. For example, in the MTAP and HMGA2 loci, we observe a significant reduction in Cohesin levels at the super-enhancers and other Mediator-bound elements ( Fig. 4a,b). By contrast, Cohesin occupancy at CTCF-binding sites in these regions is not grossly affected by Mediator depletion. We find similar patterns in the MYC, ITPRID2, ERRFI1 and KRT19 loci (Extended Data Fig. 7). Genome-wide quantification of Cohesin occupancy at Mediator-bound enhancers and CTCF-binding sites shows a significant reduction in Cohesin levels at enhancers and stable occupancy at CTCF-binding sites after depletion of Mediator (Fig. 4c,d). These results show that the distribution of Cohesin is altered when Mediator is depleted and suggest that Mediator contributes to the stabilization of Cohesin at enhancer elements.

Mediator depletion causes changes in nano-scale interactions
To further analyze the impact of Mediator depletion on chromatin architecture, we leveraged the ability of Tiled-MCC to directly identify ligation junctions and resolve localized nano-scale interaction patterns 14 . We focused our analyses on ligation junctions in regions containing super-enhancers, genes and boundary elements in the MYC, MTAP, HMGA2 and ITPRID2 loci ( Fig. 5 and Extended Data Fig. 8).
Within the MYC super-enhancer, we observe enriched interactions between the individual elements of the super-enhancer (Fig. 5, left matrix). After depletion of Mediator, the frequency of these interactions is decreased. We observe similar patterns in the MTAP, HMGA2 and ITPRID2 loci (Extended Data Fig. 8). In the MTAP locus, we could also resolve the interactions between the gene promoter and a nearby enhancer. After Mediator depletion, there are fewer of these interactions (Extended Data Fig. 8a). These results show that interactions between active enhancer and promoter elements across very small distances are dependent on Mediator.
It has previously been shown that regions containing CTCF-binding sites form characteristic architectural patterns, in which phased nucleosomes surrounding the CTCF motif form a grid-like structure, which is associated with strong insulation between the regions upstream and downstream of the CTCF-binding site 14,41 . We observe these patterns at the intergenic CTCF-binding sites in the loci we investigated and do not see any changes upon depletion of Mediator (Fig. 5, right matrix, and Extended Data Fig. 8).
At the level of individual genes, we observe domain-like structures extending across the gene body (Fig. 5, middle-left matrix, and Extended Data Fig. 8). Interestingly, we observe that depletion of Mediator results in the appearance of specific structures within the MYC gene, which are centered around hypersensitive and CTCF-bound elements (Fig. 5, middle-left matrix). Zooming in on this region at higher resolution ( Fig. 5, middle-right matrix) resolves a structure that is reminiscent of intergenic CTCF-binding sites at the CTCF-bound region within the MYC gene body when Mediator is depleted. This suggests that high transcriptional activity in the presence of Mediator leads to a disruption of the specific nucleosome structures that are normally formed around CTCF-binding sites. We observe similar patterns at the CTCF-binding sites contained within the MTAP and ITPRID2 gene bodies upon depletion of Mediator (Extended Data Fig. 8a,c).  Article https://doi.org/10.1038/s41594-023-01027-2

Comparison of Mediator loss and transcription inhibition
The reduction in enhancer-promoter interactions that we observe after Mediator depletion is associated with a strong decrease in gene expression. A plausible explanation for these observations is that weakening of enhancer-promoter interactions leads to lower levels of gene activity. However, it is also possible that reduced transcriptional activity leads to weakening of enhancer-promoter interactions. To get more insight into the cause-consequence relationship between regulatory interactions and transcription, we performed MCC experiments in cells treated with triptolide, which inhibits initiation of transcription (Extended Data Fig. 9). Comparison of the MCC data from DMSO-treated cells with those from triptolide-treated cells shows that chemical inhibition of transcription does not lead to a reduction of enhancer-promoter interactions. By contrast, we find that enhancer-promoter interactions are significantly weaker in cells in which Mediator is depleted than in cells in which transcription is inhibited. This indicates that enhancer-promoter interactions are dependent on Mediator and not on the process of transcription.
Although chemical inhibition of transcription does not result in reduced enhancer-promoter interactions, we observe increased interactions with CTCF-binding sites following triptolide treatment. This indicates that it is possible that the increased CTCF-mediated interactions, which we detect after Mediator depletion, result from reduced transcription in the locus.

BET proteins do not compensate for depletion of Mediator
Our data show that both short-and long-range interactions between enhancers and promoters are dependent on Mediator. However, we find that enhancer-promoter interactions are not completely abolished when Mediator is depleted. This indicates that other factors are involved in mediating enhancer-promoter interactions and possibly compensate for the loss of Mediator. It has recently been suggested that BRD4 plays a role in genome organization and stabilizes Cohesin on chromatin 42 . Although it has been shown that inhibition of BET proteins alone does not lead to changes in enhancer-promoter interactions (despite having a strong impact on transcription) 43   whether Mediator and BET proteins might have (partly) redundant roles in enhancer-promoter interactions. We therefore investigated the impact of combined Mediator depletion and chemical BET inhibition on enhancer-promoter interactions with Capture-C (Extended Data Fig. 10). However, we do not find consistent additional effects on enhancer-promoter interactions after combined Mediator depletion and BET inhibition, compared with depletion of Mediator alone. This suggests that enhancer-promoter interactions result from a more complex interplay between many regulatory factors.

Discussion
In this study, we have investigated the function of the Mediator complex in the regulation of chromatin architecture and enhancer-promoter interactions (Fig. 6). To overcome limitations of existing studies 24-29 , we have combined rapid depletion of Mediator using dTAG technology and analysis of genome architecture at very high resolution with targeted MNase-based 3C approaches. This strategy has enabled us to demonstrate that depletion of Mediator leads to a significant reduction in enhancer-promoter interactions.
We have focused our analyses on 20 gene loci containing strong super-enhancers and found an average decrease in interaction strength of ~34% between promoters and Mediator-bound enhancer elements in these regions. This reduction in enhancer-promoter interactions is associated with an average downregulation of expression of ~7.5-fold for the genes we investigated. The relatively small effect on interaction frequency in comparison with gene activity is in agreement with recent studies that have shown that the relationship between enhancer-promoter interaction frequency and transcriptional output is not linear and that small changes in genome architecture can have a large impact on gene activity levels 44,45 .
In the context of Mediator depletion, there are several possible explanations for these observations. We have focused our analyses on genes regulated by super-enhancers, which are composed of many individual elements. For example, the MTAP gene is regulated by two super-enhancers, which contain more than twenty individual active elements. The additive and potentially synergistic impact of reduced interactions of each of these elements could cumulatively cause large changes in gene expression levels. In addition, the Mediator complex plays a central role in the regulation of gene expression and is thought to act at several stages of the transcription cycle. It is therefore likely that the large decrease in transcriptional output upon Mediator depletion is related not only to weaker enhancer-promoter interactions, but also to the loss of the general function of Mediator in the regulation of initiation (for example, PIC assembly and activation), re-initiation, elongation and transcriptional bursting 23,46 . Moreover, it is thought that the function of the Mediator complex in gene regulation is (partly) dependent on the formation of nuclear condensates [47][48][49][50][51] . In agreement with this model, it has been shown that MED14 depletion leads to dissolved Pol II clusters 28 . It is possible that the reduced interactions between enhancers and promoters after Mediator loss are not sufficient to establish the required concentrations of transcription factors, coactivators and Pol II for the formation of nuclear condensates in which transcription can be efficiently initiated. Finally, it is important to note that enhancer-promoter interactions are thought to be transient and vary from cell to cell 2 . It has been shown that enhancer-promoter proximity does not necessarily co-occur with transcriptional burst 52 ; the precise mechanisms by which interactions between enhancers and their target gene promoter relate to transcriptional activation therefore require further investigation.
Our data indicate that Mediator's role in enhancer-promoter interactions is (partly) dependent on Cohesin. Although it has previously been shown that Mediator co-localizes with Cohesin 24,25 , the functional relationship between these complexes has thus far been unclear. Our data show that Cohesin levels at enhancers are reduced when Mediator is depleted. A possible explanation for this observation is that Mediator stabilizes Cohesin on chromatin. Although further investigation of the interaction between Mediator and Cohesin is required, this suggests that Cohesin and Mediator cooperate in the formation of enhancer-promoter interactions and provides support for a model in which extruding Cohesin molecules are stalled at Mediator-bound enhancers and promoters and thereby bridge interactions between these elements. These findings indicate that Cohesin extrusion trajectories are dependent on multiple regulatory proteins and that these factors cooperate in the formation of specific 3D chromatin structures in which gene expression is regulated 53 .
The high resolution of our data has enabled us to visualize the effects of Mediator depletion on nano-scale genome organization. We find that interactions between the individual elements within super-enhancers and interactions between enhancers and promoters across very small distances are dependent on Mediator. Of note, we have previously shown that Cohesin depletion leads to a reduction of enhancer-promoter interactions across medium and large genomic distances (>~10 kb), but that Cohesin is not involved in regulating short-range enhancer-promoter interactions or interactions within enhancer clusters 14 . This suggests that Cohesin has a role in facilitating longer-range enhancer-promoter interactions and that Mediator can function independently on smaller scales. At the level of nano-scale genome organization, we also detect changes in chromatin structure at CTCF-binding sites. Since we only observe these changes at CTCF-binding sites within gene bodies and not at intergenic CTCF-binding sites, we think that these changes are related to the reduced transcription levels following Mediator depletion. This implies that specific higher-order nucleosome structures within genes can form only in the absence of high transcriptional activity, which is consistent with experiments in yeast that have shown that the transcriptional machinery disrupts regular nucleosome spacing 54 . Although the changes in chromatin structure at intragenic CTCF-binding sites are likely related to lower transcription levels that result from Mediator depletion, it is important to note that we do not observe a reduction in enhancer-promoter interactions following treatment with triptolide to chemically inhibit transcription. These observations are consistent with several recent reports in which the impact of acute inhibition of transcription was analyzed with high-resolution Micro-C approaches 55,56 . This indicates that enhancer-promoter interactions depend on the Mediator complex and not on the process of transcription. However, it is of interest that we observe increased interactions with CTCF-binding sites following transcription inhibition. This suggests that the increased CTCF-mediated interactions that we detect after Mediator depletion could be related to the reduced levels of transcription that are associated with loss of Mediator. A possible explanation for these observations is that transcribing Pol II molecules form an obstacle to extruding Cohesin molecules; CTCF loops might therefore form more efficiently when transcription levels are reduced. This model fits with previous work that has shown that the distribution of Cohesin is dependent on transcription 57,58 and with two recent reports indicating that Pol II can form barriers to loop extrusion 59,60 .
With the exception of a subtle increase in the strength of TAD and sub-TAD boundaries, we do not observe large-scale changes in genome architecture upon Mediator depletion. This is consistent with previous reports, in which the impact of Mediator depletion has been investigated with lower resolution approaches, such as Hi-C and Hi-ChIP [27][28][29] . On the basis of knockout of the Mediator-CDK module, it has recently been suggested that the Mediator complex is involved in the regulation of heterochromatin domains and genome compartmentalization 61 . We do not observe clear changes in compartmentalization after 2 h of Mediator depletion, but it is likely that changes in compartmentalization would require more time to manifest [62][63][64] .
Although our data clearly show that enhancer-promoter interactions are dependent on Mediator, we do not observe a complete loss of interactions when Mediator is depleted. This suggests that additional proteins and mechanisms play a role in mediating enhancer-promoter interactions. We find that the interactions that remain following depletion of Mediator are not dependent on BET proteins. However, many other regulatory factors, such as tissue-specific transcription factors 65-67 and more widely expressed transcription factors, such as LDB1 (refs. 68-71) and YY1 (refs. 72,73), have been implicated in enhancer-promoter interactions. It is likely that the regulation of enhancer-promoter interactions is dependent on a complex interplay between multiple regulatory proteins, which might act in a (partly) redundant manner to ensure the formation of robust enhancer-promoter interactions. In line with biochemical and structural evidence 20,21 , our data show that the Mediator complex is one of the factors with an important role in regulating enhancer-promoter communication and gene expression, by acting as both a functional and an architectural bridge between enhancers and promoters.

ChIP-seq
Calibrated MNase ChIP-seq was performed as described previously 75 , with some modifications for three biological replicates per experimental condition. Fresh protease (Roche, 11873580001) and phosphatase inhibitors (Roche, 4906837001) were added to all buffers. Briefly, 6 × 10 7 cells were crosslinked with 1% formaldehyde for 8 min at room temperature, followed by quenching with 125 mM glycine for 5 min. The fixed cells were scraped from the plates, washed twice with ice-cold PBS and centrifuged. The cell pellets were resuspended in Farnham lysis buffer (5 mM PIPES pH 8, 85 mM KCl, 0.5% NP-40) and incubated on ice for 10 min. After centrifugation, the nuclei pellets were resuspended in 1% SDS lysis buffer (50 mM Tris-HCl pH 8, 10 mM EDTA, 1% SDS). Following incubation at room temperature for 10 min, IP buffer (20 mM Tris-HCl pH 8, 1 mM EDTA, 150 mM NaCl, 1% Triton X-100) supplemented with 5 mM CaCl 2 was added to quench the reaction and to further dilute the SDS (0.1% final concentration). The samples were then digested with 20,000 U of MNase (NEB, M0247S) at 37 °C for 20 min, followed by the addition of 20 mM EDTA and 10 mM EGTA to quench the MNase digestion. The digested samples were sonicated, and the chromatin supernatants were collected afterwards. For each IP, 45 μg of sample chromatin and 200 ng of Drosophila S2 MNase-digested chromatin were used. The samples were pre-cleared with Dynabeads Protein G (Thermo Fisher Scientific, 10009D) for 30 min at 4 °C. Pre-cleared samples were incubated with 1.32 μg of rabbit anti-HA-Tag (C29F4) antibody (Cell Signaling Technology, 3724) and 1 μg of Drosophila spike-in antibody (Active Motif, 61686) and incubated overnight with gentle rotation. Following incubation, inputs were collected and stored for each sample. The samples were further incubated with Dynabeads Protein G at 4 °C for 3 h. Bead washes were performed at 4 °C for 5 min in the following order: 1× with Buffer 1 (20 mM Tris-HCl pH 8, 2 mM EDTA, 150 mM NaCl, 1% Triton X-100, 0.1% SDS), 4× with Buffer 2 (20 mM Tris-HCl pH 8, 2 mM EDTA, 500 mM NaCl, 1% Triton X-100, 0.1% SDS), 1× with Buffer 3 (10 mM Tris-HCl pH 8, 1 mM EDTA, 250 mM LiCl, 1% NP-40, 1% sodium-deoxycholate), and 3× with TE buffer (10 mM Tris-HCl pH 8, 1 mM EDTA, 50 mM NaCl). The beads were subsequently eluted in elution buffer (0.1 M NaHCO 3 , 160 mM NaCl, 1% SDS). The samples were de-crosslinked, and DNA extraction was performed. Library preparations were performed using the NEBNext Ultra II DNA Library Prep Kit for Illumina (NEB, E7645S) with a modified thermocycler program for the End Prep reaction (20 °C for 30 min, 50 °C for 1 h; heated lit set to 60 °C). The amplified libraries were size selected with double-sided (1.0-1.2x) SPRI bead purification. The final libraries were assessed on a fragment analyzer and sequenced using the NextSeq550 Illumina platform (43-bp paired-end reads). Paired-end reads were processed for adapter removal and mapped to the hg38 reference genome using Bowtie2 76 . Duplicates were filtered and removed using SAMtools 77 .
https://doi.org/10.1038/s41594-023-01027-2 Spike-ins from Drosophila chromatin were used for normalization. Normalized bigwig files were generated using Deeptools 78 . Peak calling was performed with MACS2 (ref. 79) in DMSO samples using input files for thresholding. Box plots were generated with R using default settings.

Capture-C
Capture-C was performed as described previously 80,81 for three biological replicates per experimental condition. Briefly, 10 × 10 6 cells per biological replicate were crosslinked, followed by cell lysis. 3 C libraries were generated by DpnII digestion and subsequent proximity ligation. After decrosslinking and DNA extraction, the resulting 3 C libraries were sonicated to a fragment size of ~200 bp and indexed with Illumina sequencing adapters, using Herculase II polymerase (Agilent, 600677) for library amplification. To boost library complexity, indexing was performed in two parallel reactions for each sample. Biotinylated oligonucleotides (70 nt) were designed using a python-based oligo tool 82 (https://oligo.readthedocs.io/en/latest/) and used for enrichment of the libraries in two consecutive rounds of hybridization, biotin-streptavidin bead pulldown (Invitrogen, 65306), bead washes and PCR amplification (KAPA HyperCapture Reagent Kit, Roche, 09075828001). The final libraries were assessed on a fragment analyzer and sequenced using the NextSeq550 Illumina platform (75-bp paired-end reads). Data analysis was performed using the CapCruncher pipeline 80 (https://github.com/sims-lab/CapCruncher).

Tiled Micro-Capture-C
Tiled-MCC was performed using the generated MCC libraries, following a tiled enrichment procedure as described previously 14 , using the Twist Hybridization and Wash Kit (Twist Bioscience, 101025). Briefly, indexed MCC libraries were pooled and dried completely in a vacuum concentrator at 45 °C. Dried DNA was resuspended in blocker solution and pooled with the hybridization solution containing a custom panel of biotinylated oligonucleotides (70 nt; designed using a python-based oligo tool 82 (https://oligo.readthedocs.io/en/latest/) and incubated at 70 °C overnight. Streptavidin bead pulldown and bead washes were performed with Twist Wash Buffers according to the manufacturer's instructions (Twist Target Enrichment Protocol). Subsequently, post-hybridization PCR was performed with 11 cycles of amplification. PCR-amplified libraries were purified using pre-equilibrated Twist DNA Purification Beads. The final libraries were assessed on a fragment analyzer and sequenced using the NextSeq550 Illumina platform (150-bp paired-end reads). Data analysis was performed using the MCC pipeline 33 (https://github.com/jojdavies/Micro-Capture-C) and HiC-Pro pipeline 83 (https://github.com/nservant/HiC-Pro) as described previously 14 . All contact matrices were balanced using ICE-normalization 84 . The large-scale contact matrices have a resolution of 500 bp -2 kb (depending on the size of the region); the resolution of the nano-scale matrices is indicated in the figures.

ATAC-seq
Assay for Transposase-Accessible Chromatin using sequencing (ATAC-seq) was performed as described previously 34,85 with some modifications. Three biological replicates per experimental condition were used for the experiment. Briefly, 1.5 × 10 5 washed cells were split over two tubes, followed by centrifugation. Cell pellets were resuspended in fresh cold lysis buffer (10 mM Tris-HCl pH 7.5, 10 mM NaCl, 3 mM MgCl 2 , 0.1% Igepal CA-630) and incubated on ice for 3 min. The lysates were washed once with cold PBS, and the resulting nuclear pellets were resuspended in the tagmentation mix (Illumina, 20034198). The tagmentation reactions were performed at 37 °C and 1,000 r.p.m. for 30 min, followed by DNA purification using MinElute PCR purification kit (Qiagen, 28004). The indexed samples were amplified using Nextera indexing primers and NEBNext High-Fidelity PCR Master Mix (NEB, M0541), with an initial 5-min extension step at 72 °C. A real-time PCR library amplification kit (KAPA, KK2701) was used to calculate the required number of PCR cycles (11 cycles) in order to minimize library amplification bias. Size selection was performed with double-sided SPRI bead purification to remove primer dimers and larger fragments (>700 bp). The final libraries were assessed on a fragment analyzer and sequenced using the NextSeq550 Illumina platform (75-bp paired-end reads). The data from each replicate were down-sampled to the library with the lowest read depth and analyzed using the NGseqBasic pipeline 86 .

CUT&Tag
CUT&Tag 40 was performed for three biological replicates (for a total of five technical replicates) per experimental condition using the CUT&Tag-IT Assay Kit (Anti-Rabbit) (Active Motif, 53160), according to the manufacturer's instructions with some modifications. Briefly, 0.5 × 10 6 cells were mildly crosslinked with 0.3% paraformaldehyde (Science Services, E15710), followed by quenching with 125 mM cold glycine. Meanwhile, concanavalin A beads were prepared, following the manufacturer's instructions. The fixed cells were washed, resuspended in wash buffer and incubated with concanavalin A beads for 10 min on a rotator at room temperature. The samples were placed on a magnetic stand to clear the liquid, and the samples were resuspended with ice-cold antibody buffer supplemented with protease inhibitor cocktail and digitonin. Then, 1 μg rabbit anti-SMC1A antibody (1:50, Abcam, ab9262) or 1 μg rabbit IgG isotype control antibody (1:50, Cell Signaling Technology, 2729S) was added to each sample, and the samples were incubated overnight at 4 °C on a rotator in 0.2-mL PCR tubes. The next day, the samples were incubated with guinea pig anti-rabbit secondary antibody (1:100, Active Motif, 53160) for 1 h at room temperature on a rotator, followed by washes with dig-wash buffer. The samples were placed on a magnetic stand to clear the liquid, and the beads were resuspended with CUT&Tag-IT Assembled pA-Tn5 Transposons. The reactions were subsequently incubated at room temperature on a rotator, followed by washes with Dig-300 buffer. After clearing the liquid Article https://doi.org/10.1038/s41594-023-01027-2 on a magnetic stand, the beads were resuspended with tagmentation buffer. The tagmentation reactions were subsequently incubated at 37 °C for 60 min. The samples were de-crosslinked, and DNA extraction was performed according to the manufacturer's instructions. The libraries were amplified by PCR, and size selection was performed with two rounds of SPRI bead purification to remove primer dimers. The final libraries were assessed on a fragment analyzer and sequenced using the NextSeq550 Illumina platform (75-bp paired-end reads). The data were analyzed using the NGseqBasic pipeline 86 . Peak calling was performed with MACS2 (ref. 79) using IgG controls for thresholding. Normalized bigwig files and meta peak profiles were generated using Deeptools 78 and LOESS regression was applied for smoothening of the data. Box plots were generated with R using default settings. Differential binding analysis was performed in R using the DiffBind package. An adjusted P value of 0.05 (Benjamini-Hochberg method) was used to identify differentially bound SMC1A peaks after Mediator depletion (Supplementary Table 1).

Public data analysis
DNase-I hypersensitivity data 87 (ENCSR000ENM) and ChIP-Seq data for CTCF 87 (ENCSR000BSE) and MED26 (ref. 27) in HCT-116 cells were analyzed using the NGseqBasic pipeline 86 . TT-seq data files for HCT-116 MED14-dTAG cells 28 were shared by the authors, and differential expression analysis was performed in R using the DESeq2 package 88 .

Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability
The raw sequencing and processed data are available from the Gene Expression Omnibus (GEO) as a SuperSeries under accession number GSE205984. DNase-I hypersensitivity data 87 and ChIP-seq data for CTCF 87   Gene annotation, DNase hypersensitive sites (DHS) and ChIP-seq data for CTCF and MED26 are shown at the bottom. Super-enhancers are highlighted in green below the MED26 profiles and orientations of CTCF motifs are indicated with arrowheads (forward orientation in red; reverse orientation in blue). The dashed black ovals in the dTAG and differential contact matrices highlight decreased enhancer-promoter interactions, whereas the solid ovals indicate increased CTCF interactions. Coordinates (hg38): chr9:21,000,000-22,550,000. Gene annotation, DNase hypersensitive sites (DHS) and ChIP-seq data for CTCF and MED26 are shown at the bottom. Super-enhancers are highlighted in green below the MED26 profiles and orientations of CTCF motifs are indicated with arrowheads (forward orientation in red; reverse orientation in blue). The dashed black ovals in the dTAG and differential contact matrices highlight decreased enhancer-promoter interactions, whereas the solid ovals indicate increased CTCF interactions. Coordinates (hg38): chr2:181,380,000-182,280,000. Fig. 8 | Micro-topology analysis of the MTAP, HMGA2 and ITPRID2 loci in HCT-116 MED14-dTAG cells. a. Tiled-MCC ligation junctions in the MTAP locus in HCT-116 MED14-dTAG cells treated with DMSO (top; n = 3 biologically independent samples) or dTAG ligand (bottom; n = 3 biologically independent samples), displayed in localized contact matrices at high resolution. Annotation as described in Fig. 5. The regions covered in the contact matrices show an intergenic CTCF-binding site, an interacting enhancer and promoter, a promoter and CTCF-binding sites and intragenic CTCF-binding sites, at the indicated resolution. The squares in the left matrices highlight regular nucleosome interactions surrounding an intergenic CTCF-binding site, which do not change upon Mediator depletion. The ovals in the middle-left matrices highlight very short-range enhancer-promoter interactions, which are significantly reduced following loss of Mediator (p = 0.030713; two-sided unpaired t-test). The arrows in the right two matrices highlight the appearance of CTCF-mediated insulation stripes and CTCF-mediated short-range interactions within the gene body after Mediator depletion. These patterns are difficult to appreciate at the lower resolution displayed in the middle-right matrix, but more clearly visible in the right matrix. b. Data as described in a for the HMGA2 locus.

Extended Data
The regions covered in the contact matrices show intergenic CTCF-binding sites, a super-enhancer and a gene, at the indicated resolution. The squares in the left matrices highlight regular nucleosome interactions surrounding intergenic CTCF-binding sites, which do not change upon Mediator depletion. The ovals in the middle matrices highlight interactions between the constitutive elements of the super-enhancer, which are significantly reduced following loss of Mediator (p = 0.000246; two-sided unpaired t-test). The arrows in the right two matrices highlight the appearance of specific interaction patterns within the gene body after Mediator depletion. c. Data as described in a for the ITPRID2 locus. The regions covered in the contact matrices show a super-enhancer, a gene and an intragenic CTCF-binding site, at the indicated resolution. The circles in the left matrices highlight interactions between the constitutive elements of the super-enhancer, which are significantly reduced upon Mediator depletion (p = 0.027392; two-sided unpaired t-test). The arrows and squares in the middle and right matrices highlight the appearance of regular nucleosome patterning surrounding a CTCF-binding site within the gene body following loss of Mediator. These patterns are difficult to appreciate at the lower resolution displayed in the middle matrix, but more clearly visible in the right matrix.