Cohesin is required for long-range enhancer action at the Shh locus

The regulatory landscapes of developmental genes in mammals can be complex, with enhancers spread over many hundreds of kilobases. It has been suggested that three-dimensional genome organization, particularly topologically associating domains formed by cohesin-mediated loop extrusion, is important for enhancers to act over such large genomic distances. By coupling acute protein degradation with synthetic activation by targeted transcription factor recruitment, here we show that cohesin, but not CTCF, is required for activation of the target gene Shh by distant enhancers in mouse embryonic stem cells. Cohesin is not required for activation directly at the promoter or by an enhancer located closer to the Shh gene. Our findings support the hypothesis that chromatin compaction via cohesin-mediated loop extrusion allows for genes to be activated by enhancers that are located many hundreds of kilobases away in the linear genome and suggests that cohesin is dispensable for enhancers located more proximally. Kane et al. show that cohesin is required for long-range, but not shorter-range, enhancer function at the Shh locus, implicating loop extrusion as a mechanism for keeping enhancer and target genes sufficiently close together for effective enhancer–promoter communication.


Introduction
The mammalian genome is organised into topologically associating domains (TADs) that are formed through the process of cohesin-driven loop extrusion [1][2][3] and whose extent is constrained at TAD boundaries by orientation-dependent CTCF binding [4][5][6][7] .The large regulatory landscapes of developmental genes frequently correspond to TADs, leading to the hypothesis that TADs and/or loop extrusion are important for enhancers to act on their cognate gene 8,9 .
However, it has proven hard to interpret the consequences of experimental disruption of TADs or loop-extrusion on gene regulation 3,6,10 , in part because of the difficulty in distinguishing direct from indirect effects on enhancer-driven gene expression.CTCF null mice show early embryonic lethality 11 and conditional knockout of CTCF in the developing mouse limb results in extensive cell death 12 .Cohesin is also essential for cell proliferation, limiting study in vivo 13 .In vitro removal of cohesin does not seem to have very substantial effects on specific gene regulation 3 , but it is required for inducible gene regulation in primary haematopoietic cells 14 .Conversely, conditional removal of cohesin in post-mitotic neurons has been reported to perturb gene expression but not to affect enhancer-driven inducible immediate early gene activation 15 .Depletion of the cohesin loading factor NIPBL in nondividing liver cells in vivo could not attribute transcriptional effects to systematically altered enhancer function 16 .
Here, we exploit synthetic transcriptional activators, coupled with the acute degradation of CTCF or cohesin, to investigate mechanisms of enhancer action at a distance, using the Shh locus as a paradigm.In mouse embryonic stem cells (mESCs) we show that cohesin, but not CTCF, is required for activation of Shh by enhancers located many hundreds of kilobases upstream, but is dispensable for an enhancer located closer to the target gene.

Synthetic enhancer activation at long distance
Shh acts as a concentration-dependent morphogen during vertebrate embryonic development and the complex Shh regulatory domain is a paradigm for long-range enhancer regulation.Many of the tissue-specific enhancers of Shh operate over large genomic distances with the regulatory landscape extending over approximately 1 Mb (Fig. 1a).The limits of this regulatory landscape, defined using transposon-based regulatory sensors 9,17 , correspond with a TAD which contains Shh and all of its enhancers that have been defined so far.One of the TAD boundaries lies in an intergenic region 3' of Shh whereas the other is near the Lmbr1 promoter.The murine Shh regulatory landscape contains at least five CTCF binding sites (Fig. 1a), including two strongly interacting convergent sites which may form the Shh TAD boundaries and block loop extrusion 18 .
ZRS, located 849 kb upstream of Shh in intron 5 of the widely expressed Lmbr1, is the most distal Shh enhancer.ZRS is both necessary and sufficient for Shh expression in the zone of polarising activity (ZPA) in distal posterior mesenchymal cells of the developing limb bud 19,20 .Increased Shh-ZRS colocalization, observed in the ZPA, may be consistent with a gene-enhancer interaction 21 .Large inversions that encompass the Shh TAD boundaries disrupt Shh-ZRS interactions and Shh regulation in limb buds 9 but small deletions of CTCF sites at the Shh TAD boundaries, though disrupting TAD structure and reducing Shh -ZRS colocalization, do not alter the developmental pattern of Shh expression or cause a mutant phenotype 18 .
Enhancers are activated by binding of the appropriate transcription factors (TFs) which can be mimicked by targeting of artificial TFs.Previously, we demonstrated synthetic activation of Shh in mESCs using transcription activator-like (TAL) effectors (TALEs) fused to multimers of VP16 22 .Shh expression could be induced by activator binding at the Shh promoter (tShh) and at the neural enhancers SBE6 (100 kb upstream) and SBE2 (410 kb upstream) (tSBE2-VP64 and tSBE2-VP64, respectively).Local peaks of H3K27 acetylation were also induced at the site of TALE binding and at the activated Shh in both cases 22 .To determine if Shh transcription could also be triggered by synthetic activator binding at the far end of the TAD, we designed a TALE for ZRS (tZRS) (Fig. 1a, b).
Previously we used qRT-PCR to assay the steady state level of Shh mRNA induced by synthetic activators averaged across the transfected cell population 22 .To detect Shh nascent transcripts at a single cell/allele level, here we used RNA FISH in mESCs 48hrs after transfections with tShh-VP64, tSBE2-VP64 and tZRS-VP64 (Fig. 1c, d, e) and with control constructs lacking the activation domain (-Δ) (Extended Data Fig. 1a,b).A probe set detecting Lmbr1 nascent transcripts was used as a positive control as TALE binding was not thought to be able to affect this broadly expressed gene.
Consistent with our previous demonstration of Shh activation by TALE-Vp64s, Shh nascent RNA FISH signals were detected in mESCs transfected with tShh-VP64 (9-11% of Shh alleles) or tSBE2-VP64 (4-5% of alleles)(Fig.1f, Extended Data Fig. 1b,c).Cells transfected with TALE-Δ, and non-transfected cells (ntc) showed a very low signal levels.tZRS-VP64 also activated Shh (4-6% of alleles detected) indicating that Shh can be expressed following activator binding 850kb away.Induced Shh expression levels from all TALE constructs fused to VP64 were significantly greater than the equivalent TALE-Δ transfected cells (Extended Data Fig. 1d).Lmbr1 transcripts were detected at approximately 60% of alleles and these levels were similar in cells transfected with either tShh, tSBE2 or tZRS with or without fusion to Vp16 (Fig. 1f; Extended Data Fig. 1c).

Synthetic activation in the absence of CTCF
The Shh TAD contains a number of CTCF binding sites (Fig. 1a) important for TAD structure but that individually are not necessary for Shh regulation in vivo 18 .Combinatorial deletion suggests that loss of more than one CTCF site within the Shh TAD may have a more marked effect on expression 23 .Genome-wide depletion of CTCF in mESCs dramatically alters TAD insulation with rather minimal effects on ongoing gene expression 6 .However, those studies did not address where the complete loss of CTCF affects the induction of gene activation, and particularly via enhancers.To investigate whether synthetic activation of Shh was dependent on CTCF we used mESCs in which the degradation of CTCF can be induced via an auxin-inducible degron (AID) (CTCF-AID) 6 .FACS (for GFP) indicated that CTCF depletion occurred as early as 6 hours after auxin addition and persisted for up to 48hrs of auxin treatment (Fig. 2a).CTCF-AID auxin-treated cells appeared to divide for at least 1-2 cell cycles, maintained a normal colony morphology and did not show significant levels of cell death up to 48 hours of auxin treatment (Extended Data Fig. 1e).Immunofluorescence indicated a very small proportion of GFP positive cells in CTCF-AID cells treated for 6 hours and so 24 hour auxin treatment, when GFP +ve cells were completely absent, was used for subsequent experiments using CTCF-AID cells (Fig. 2b).To ensure that auxin addition did not impact TALE activity per se, wild type mESCs were transfected with TALE-VP64/-Δ targeting the Shh promoter, SBE2 and ZRS and auxin added to the media on the day after transfection for 24 hours.Targeting the Shh promoter or distal enhancers (SBE2/ZRS) with TALE-VP64, but not with TALE-Δ, led to activation of Shh expression both in the absence and presence of auxin with no significant differences in the proportion of expressing alleles caused by the addition of auxin (Fig. 1f and Extended Data Fig. 1c,d).Lmbr1 expression was also consistent across all conditions and unaffected by the TALEs.
It has been reported that CTCF-AID molecules bound at different CTCF sites have different susceptibilities to auxin-dependent degradation 24 , and ChIP for CTCF following auxin treatment of CTCF-AID mESCs 6 shows a complete absence of CTCF at sites 2 and 3 within the Shh TAD and a substantial reduction, but not complete absence, of CTCF sites 1, 4 and 5 at the Shh TAD boundaries (Fig. 2c).However, analysis of Hi-C data 6 from auxin-treated CTCF-AID cells shows that insulation at the Shh TAD boundaries, and particularly that at the Lmbr1 end, are weakened (Fig. 2c), and more inter-TAD interactions between the Shh and neighbouring TADs are detected in the absence of CTCF.Intra-TAD interactions were also affected by CTCF depletion, confirmed by virtual 4C display of the Hi-C data (Extended Data Fig. 2a).Using the Shh promoter as a viewpoint, proteolytic degradation of CTCF leads to decreased interactions of Shh with sequences within its own TAD and increased interactions with sequences in the adjacent En2 containing TAD.Of note, our previous 5C analysis of the Shh TAD did not detect the formation of specific enhancer-promoter loops as a consequence of TALE-Vp64 enhancer activation 22 .
Given this altered 3D chromatin landscape, we tested whether CTCF depletion affected the ability to synthetically activate Shh, including from distal enhancers.CTCF-AID cells were transfected with tShh-VP64, tSBE2-VP64 and tZRS-VP64 and with the corresponding TALE-Δs controls.Auxin was added the day after transfection for 24 hours.Shh expression could still be induced in CTCF-depleted cells when targeting either the Shh promoter or the enhancers with TALE-VP64 (Fig. 3a, Extended Data Fig. 3a, c).These data suggest that activation of Shh expression by targeting its distal enhancers does not require CTCF.
Consistent both with the Hi-C/ virtual 4C data (Fig. 2c, Extended Data Fig. 2a) from auxintreated CTCF-AID cells, and with our previous analysis deleting specific CTCF sites at the Shh locus 18 , DNA FISH on CTCF-AID cells transfected with tSBE2-VP64 or tZRS-VP64 revealed some decompaction within the Shh TAD caused by CTCF loss (Fig. 3b,c and Extended Data Fig. 3e).Notably, after CTCF loss in tSBE2-VP64 transfected cells we saw no alleles where Shh and SBE2 were spatially co-localised (within 200nm) (Fig. 3c) despite no effect of CTCF loss on Shh nascent transcription by tSBE2-VP64 (Fig. 3a).This is consistent with previous observations that enhancer-gene co-localisation is not required for enhancerdriven gene-activation 22 , though we cannot exclude extremely transient colocalization events that are undetectable by FISH.

Synthetic activation of Shh from a distance is cohesin dependant
To examine effects of cohesin loss on synthetic Shh activation we used auxin to acutely deplete SCC1 (RAD21) from mESCs 25 .Cohesin is required for sister chromatid cohesion during mitosis 13 and in its absence SCC1-AID cells fail to divide and die.FACS and immunofluorescence indicated that SCC1 depletion occurs as early as 6 hours after auxin addition (Fig. 2a, b) and we detected substantial cell death following 24 hours of auxin treatment of SCC1-AID cells (Extended Data Fig. 1d).Therefore, 6hrs of auxin treatment were used for subsequent experiments.
Genome-wide depletion of cohesin by auxin treatment of SCC1-AID is reported to erase TADs with minimal effects on steady-state gene expression 3 .Hi-C reveals a pronounced effect of SCC1 depletion on Shh TAD structure 25 (Fig. 2d).Both Shh TAD boundaries were abrogated, and intra-TAD interactions severely depleted.Virtual 4C analysis revealed the profound loss of long-range interactions of Shh both within its own TAD but also with the adjacent En2 TAD (Extended Data Fig. 2b).
To assess if distal enhancers can still activate Shh expression in the absence of cohesin, we transfected SCC1-AID cells with TALE-VP64/-Δ proteins targeting the Shh promoter, SBE2 and ZRS.Similar to results for CTCF-AID, Shh was activated in auxin-treated SCC1-AID cells by tShh-VP64 targeting the Shh promoter (Fig. 3d, Extended Data Fig. 3b).Shh was also activated from distal sites using tSBE2-VP64 and tZRS-VP64 in SCC1-AID cells in the absence of auxin.However, synthetic Shh activation from these two distal sites was drastically curtailed in auxin-treated SCC1-AID cells, to the extent that there was no significant difference between the TALE-VP64 and TALE-Δ transfected cells (Fig. 3d, Extended Data Fig. 3b, d).Lmbr1 expression was unaffected by the depletion of SCC1.These data suggest that SCC1/cohesin is necessary for distal activation of Shh from its enhancers.
As expected, given the Hi-C and virtual 4C data, in the absence of cohesin-mediated loop extrusion (SCC1 degradation) DNA FISH confirmed significant decompaction across the Shh TAD that was more dramatic than that seen after CTCF depletion.Significantly increased physical distances were measured between Shh and the distal SBE2 and ZRS enhancers (Fig.

Cohesin is not required for activation of Shh from a close enhancer
These data might indicate that cohesin is required for activation from all enhancers or may reflect a requirement for activation from large genomic distances.We previously demonstrated synthetic activation of Shh in mESCs by a TALE-VP64 targeting SBE6 (tSBE6-VP64) 22 , a Shh enhancer active in the developing brain and neural tube in vivo and neuronal progenitor cells ex vivo, and located only 100kb 5' of Shh (Fig. 1a) 26 .
Cohesin degradation in SCC1-AID ESCs following auxin treatment did not impact on the ability of tSBE6-VP64 to activate expression from Shh (Fig. 4a and Extended Data Fig. 4a,   b).Therefore, cohesin is not essential for enhancer-driven Shh activation per se, but it may be required for the function of enhancers located at relatively large genomic distances (>100kb) from their target promoter.
As we have previously reported 22 , targeting of an activator to SBE6 (in the absence of auxin) led to increased spatial separation between Shh and this enhancer (Fig. 4b, c).Cohesin depletion also led to decompaction between Shh and SBE6 in the absence of activator (tSBE6-Δ).Notably, in the presence of an activator (tSBE6-VP64) cohesin depletion had no further effect on Shh-SBE6 distances (Fig. 4b,c and Extended Data Fig. 4c).

Discussion
Our data support the observations that TAD boundaries formed by CTCF sites are not absolutely essential for an enhancer to activate its target gene located within the same TAD 18,23,27 .We find no role for either CTCF or cohesin in gene activation driven directly from the endogenous Shh promoter, but our data do indicate that cohesin-mediated loop extrusion is essential for activation from enhancers located at large genomic distances (>400kb) from their target gene.However, we show that cohesin is dispensable for activation from an enhancer (SBE6) that is located closer (100kb) to Shh.We note that this is consistent with a recent report showing that immediate early neuronal genes are still inducible after conditional ablation of cohesin in mouse neurons 15 .The enhancers in these cases appear to be close (~100kb) to their target gene promoters, whereas neuronal genes with longer chromatin loops had compromised expression.
One caveat of our study described here is that we are activating enhancers using a single synthetic transcription factor (TALE-VP64) and achieve levels of activation (% alleles expressing) that are far lower than those seen in some Shh-expressing tissues in vivo 18 where activation is driven by endogenous enhancers that likely recruit a much more complex cocktail of transcription factors and co-activators.However, we have shown that TALE-VP64s induce robust H3K27 acetylation at the enhancer they are targeted to and at the Shh gene 22 .Moreover, our recruitment of the same activator domain at different position within the Shh TAD may minimise some of the intrinsic differences between the enhancers and the differing cocktail of transcription factors they would each normally recruit during development.
Our data suggest that the process of cohesin-mediated loop-extrusion per se is not required for enhancer function -at least at the Shh locus.Rather, DNA-FISH suggests that it is the chromatin compaction brought about by loop-extrusion 28 that may be the important factor to consider.Whereas median inter-probe distances measured at ~400kb intervals across the Shh TAD were modestly affected by CTCF degradation (increases of between 10 to 140nm, Fig. 3c, Extended Data Fig. 3c), cohesin (SCC1)-loss leads to more extensive decompaction (median distance increases in the range of 130 -330nm); Figure 3f, Extended Data Fig. 4b).In contrast, cohesin loss has no effect on chromatin decompaction between Shh and SBE6 (100kb away) when SBE6 is targeted by an activator (median distances 420nm with and without cohesin).At this stage, we cannot exclude that the differential requirement for cohesin we observe is a consequence of the inevitable cell cycle change that results from cohesin depletion in mESCs, with cells accumulating in late G2 25 , though it is difficult to understand why this would affect long-range but not a relatively more closely located enhancer.
In contrast to a recent study examining the effect on reporter expression of genomic distance between promoters and enhancers inserted ectopically in mouse ESCs 29 , here we find no decrease in the efficiency of nascent transcription (RNA-FISH) from an endogenous promoter driven by targeted activator (VP64) binding to different endogenous enhancer sites 100, 450 or 850kb away.One significant difference in the present study is that the activation signal has to overcome the repressive local and higher-order polycomb-mediated chromatin environment of the Shh locus in ESCs 30 .
The molecular mechanisms by which activating signals seeded at an enhancer transmit triggers for transcriptional activation at a distant promoter remain unclear.They might involve direct translocation of regulatory information along the chromatin fibre driven by the forces of cohesin-mediated loop extrusion, but our finding that activation from an enhancer located 100kb away from a promoter is cohesin independent argues against this model.Rather we suggest that cohesin-mediated loop extrusion acts to maintain the entire regulatory domain in a compact conformation 28 .This could then enable random close encounters between enhancers and promoters to initiate molecular interactions between them, or could facilitate both loci engage, for example, with the same transcriptional hub 31 .Our hypothesis is also compatible with the transcription factor activity gradient model in which enhancers act as nucleation sites to create diffusion gradients of activating signals that decay rapidly with physical distance 32 .The size of an enhancer's sphere of influence remains to be determined but our data examining the loss of enhancer-proximity caused by cohesin loss and the ability of targeted enhancers to activate transcription suggest that this may be <500nm, compatible with the observed distances seen between active enhancers and genes in vivo 21 .eGFP were analysed for RNA FISH.Measured transfection efficiencies for the TAL-VP64 constructs in Figure 1 were: tShh-VP64 92%, tSBE2-VPp64 69%, tZRS-VP64 71%.For the data in Figs 3 and 4 they were; tShh-VP64 76%, tSBE6-VP64 82%, tSBE2VPp64 78%.Auxin-inducible degron induction Indole-3-acetic acid (auxin) (MP Biomedicals 102037) was added to the medium either 6 (SCC1-AID) or 24 (wild type or CTCF-AID) hours prior to cell collection.500 µM of auxin (1000X stock diluted in DMSO) was used for all experiments and stored at 4°C for up to a month or at -20°C for long-term storage.TALE Design and Assembly TALEs targeting the Shh promoter, SBE2 and SBE6 had previously been designed and assembled 22 .TALE protein specific to the limb enhancer ZRS was designed using TAL Effector Nucleotide Targeter 2.0 software (https://tale-nt.cac.cornell.edu)and assembled by golden-gate assembly using a modified protocol 22,34 .In brief, a DNA binding domain specific for a 15 nucleotide sequence was generated by the modular assembly of 4 pre-assembled multimeric TALE repeat modules (three 4-mer and one 3-mer) into a modified TALE backbone vector containing VP64-2A-eGFP.The backbone vector used for assembly of the ZRS TALE was modified to replace the ampicillin resistance cassette with spectinomycin resistance.TALE modules were picked from glycerol stocks of module library plates (Addgene 1000000034), incubated overnight at 37°C in L-broth containing 50 ng/μL ampicillin and DNA isolated using the QIAprep Spin Miniprep kit (Qiagen 27104) according to the manufacturer's instructions.Miniprep DNA was quantified using the Quibit dsDNA broad range assay with the Quibit 4 fluorometer.TALE modules were assembled into backbone vector by setting up a 20 μL onepot golden-gate reaction as follows: vector (100 ng), TALE modules (200ng each), 10X Tango buffer (ThermoFisher ER0451), 20 Units Esp3I (ThermoFisher ER0451), 10 Units T4 DNA ligase (New England Biosciences M0202M), 1mM ATP in ddH2O.Golden-gate reaction was performed on a thermal cycler ((37 o C 10 mins, 16 o C 10 mins x12) 36 o C 15 mins, 80 o C 5 mins, 4 o C).Competent E. coli (Invitrogen LS18263012) were transformed with 5 μL of reaction.Colonies were screened by PCR for fully assembled TALEs by setting up a 30 µL reaction as follows: single colony, 2X DreamTaq Green PCR Master mix (Thermo Scientific K1082) and 0.5 µM forward (5'GGCCAGTTGCTGAAGATCG3') and reverse (5'CGCTACAAGATGATCATTAGTG3') primers in ddH2O.Colony PCR was performed on a thermal cycler (95 o C 3 mins, (95 o C 30s, 55 o C 30s, 72 o C 120s) x30), Reaction products were run on a 1.2% agarose gel to identify positive colonies and these colonies were confirmed by Sanger sequencing.TALE-Δ constructs were made by removing the BamHI-Bg1II fragment containing VP64 from the fully assembled TALE-VP64 plasmid by restriction digest.All TALE-VP64 and TALE-Δ plasmids were either miniprepped using the QIAprep Spin Miniprep kit (Qiagen 27104) or maxi-prepped.Plasmid DNAs were quantified using the Quibit dsDNA broad range assay with the Quibit 4 fluorometer and then stored at -20°C prior to transfection.RNA FISH Custom Stellaris® RNA FISH Probes were designed against Shh and Lmbr1 nascent mRNAs (pool of 48 unique 22-mer probes) using the Stellaris® RNA FISH Probe Designer (www.biosearchtech.com/stellarisdesigner(version 4.2)).Following permeabilization, slides were incubated in wash buffer (2X SSC, 10% deionised formamide) for 5 mins at room temperature.Slides were hybridized with the Shh and Lmbr1 Stellaris FISH Probe set labelled with Quasar 670 and 570, respectively (Biosearch Technologies, Inc.), following the manufacturer's instructions (www.biosearchtech.com/stellarisprotocols).RNA FISH probes were warmed to room temperature, diluted to 125 nM in Stellaris RNA FISH hybridisation buffer (#SMF-HB1-10) containing 10% formamide and hybridised to slides overnight in a humidified chamber at 37°C.Slides were washed twice for 30 minutes in wash buffer at 37°C and rinsed in PBS.Slides were stained with 0.5 µg/mL DAPI and mounted using Vectashield.PBS and ddH2O used during RNA FISH were treated with DEPC and autoclaved to inactivate RNase enzymes.RNase free consumables were used throughout and glassware treated with RNaseZAP (Invitrogen AM9780).DNA FISH Following RNA FISH, slides were re-probed for DNA FISH.Following the removal of coverslips, slides were briefly washed in PBS and then for 10 mins in 2xSSC at 85°C followed by denaturation in 70% formamide/2xSSC at 85°C for 50 minutes before a series of alcohol washes (70% (ice-cold), 90% and 100%).160-240 ng of biotin-and Green496-dUTP-labeled (Enzo Life Sciences) (2-colour) or biotinand digoxigenin-and red-dUTP-labeled (Alexa Fluor™ 594-5-dUTP, Invitrogen) (4-colour) fosmid probes (Table S1) were used per slide, with 16-24 µg of mouse Cot1 DNA (Invitrogen) and 10 µg salmon sperm DNA.EtOH was added and the probe air dried.Hybridisation mix containing deionised formamide, 20 x SSC, 50% dextran sulphate and Tween 20 was added to the probes for ~1h at room temperature.The hybridisation mix containing the probes was added to the slides and incubated overnight at 37°C.Following a series of washes in 2X SSC (45°C) and 0.1X SSC (60°C) slides were blocked in blocking buffer (4 x SSC, 5% Marvel) for 5 min.The following antibody dilutions were made: fluorescein anti-dig FAB fragments (Roche cat.no.11207741910) 1:20, fluorescein anti-sheep 1:100 (Vector, cat.no.FI-6000)/ streptavadin Cy5 1:10 (Amersham, cat.no.PA45001, lot 17037668), biotinylated anti-avidin (Vector, cat.no.BA-0300, lot ZF-0415) 1:100, and streptavidin Cy5 1:10 for 3-colour detection; Texas Red avidin (Vector, cat.no.A2016) 1:500, biotinylated anti-avidin (Vector) 1:100 for 2-colour detection.Slides were incubated with antibody in a humidified chamber at 37°C for 30-60 min in the following order with 4X SSC/0.1% Tween 20 washes in between: fluorescein anti-dig, fluorescein anti-sheep, biotinylated anti-avidin, streptavidin Cy5 for 3-colour; Texas Red avidin, biotinylated antiavidin, Texas Red avidin for 2-colour detection.Slides were treated with 1:1000 dilution of DAPI (stock 50ug/ml) for 5min before mounting in Vectashield.Image acquisition and deconvolution Slides from RNA and DNA FISH were imaged using a Photometrics Coolsnap HQ2 CCD camera and a Zeiss AxioImager A1 fluorescence microscope with a Plan Apochromat 100x 1.4NA objective, a Nikon Intensilight Mercury based light source (Nikon UK Ltd, Kingstonon-Thames, UK) and Chroma #89014ET (3 colour) or #89000ET (4 colour) single excitation and emission filters (Chroma Technology Corp., Rockingham, VT) with the excitation and emission filters installed in Prior motorised filter wheels.A piezoelectrically driven objective mount (PIFOC model P-721, Physik Instrumente GmbH & Co, Karlsruhe) was used to control movement in the z dimension.Step size for z stacks was set to 0.2 µm.Hardware control and image capture were performed using Nikon Nis-Elements software (Nikon UK Ltd, Kingstonon-Thames, UK).Images were deconvolved using a calculated PSF with the constrained iterative algorithm in Volocity (PerkinElmer Inc, Waltham MA).RNA FISH signal quantification was carried out using the quantitation module of Volocity (PerkinElmer Inc, Waltham MA).Expressing alleles were calculated by segmenting the hybridisation signals and scoring each nuclei as containing 0, 1 or 2 RNA signals.Only cells expressing TALE constructs (cytoplasmic eGFP) were scored.DNA FISH measurements were carried out using the quantitation module of Volocity (PerkinElmer Inc, Waltham MA).For DNA FISH, only alleles with single probe signals were analysed, to eliminate the possibility of measuring sister chromatids.

Figure legends
Hi-C data analysis and generation of virtual 4C profiles Published data from ref. 6 (NCBI GEO: GSE98671 and ref. 25 (ArrayExpress: E-MTAB-7816) were re-analysed and ref. 22 was re-analysed using the distiller pipeline (https://github.com/open2c/distiller-nf). Balanced matrices at 10 kbp were used to extract the interaction profiles of the bin containing the Shh promoter with the rest of the genome in all conditions.Then these profiles, and log2-ratio of treatment over control, were saved as bigWig files and visualised using HiGlass.

Statistics & Reproducibility
The proportion of activated alleles in wild type, CTCF-AID, and SCC1-AID ESCs transfected with the various TALE constructs, as measured by RNA FISH, in the presence or absence of auxin were compared using two-sided Fisher's Exact Test, a categorical test that provides exact P-values and is suitable for small sample sizes.A biological replicate of each experiment was performed and these data are presented in Extended Data Figures.
DNA FISH interprobe distance data sets were compared using the two-tailed Mann-Whitney U test, a nonparametric test that compares two unpaired groups.Statistics and graphs were performed using Prism software (Graphpad).
No statistical method was used to predetermine sample size.No data were excluded from the analyses.Experiments were not randomized.Investigators were not blinded to allocation during experiments and outcome assessment.

Figure 1 .
Figure legendsFigure 1. Synthetic Shh activation (a) Hi-C heatmap of the Shh TAD from wild type mESCs at 16kb resolution.Data are from ref 33 and were created using HiGlass.Genes, positions of TALE target sequences and the CTCF ChIP-seq track -including CTCF motif orientation -are shown below.Genome coordinates: mm9 assembly of the mouse genome.(b) Schematic of TALE-VP64 constructs used to target the Shh promoter (tShh-VP64), SBE6 (tSBE6-VP64) SBE2 (tSBE2-VP64) or ZRS (tZRS-VP64) enhancers.NLS: nuclear localisation sequence; 2A: self-cleaving 2A peptide.Repeat variable diresidue (RVD) code is displayed below using the one letter amino acid abbreviations.Equivalent TALE-Δ constructs lack the Vp64 module (c-e) Representative images of nuclei from mESCs transfected with (c) tShh-VP64, (d) tSBE2-VP64 and (e) tZRS-VP64 showing RNA FISH signals for Shh (white) and Lmbr1 (red).Scale bars = 5 μm.(f) Timecourse of TALE transfection and auxin treatment is shown above.The percent of Shh (left) and Lmbr1 (right) expressing alleles in wild-type mESCs transfected with TALE-Vp64or TALE-Δ constructs assayed by RNA FISH in the absence or presence of auxin.The data were compared using a two-sided Fisher's Exact Test, n = numer of alleles scored, ns -not significant (p>0.05).Source Data Fig1.The biological replicate for these data, and the full statistical evaluation of all comparisons for Shh are in Extended Data Fig.1c.and d.

Figure 3 .
Figure 3. Depletion of cohesin, but not of CTCF, inhibits distal enhancer driven gene activation

a 1 r
b m L ) % ( s e l e l l a g n i s s e r p x e -Shh-expressing alleles (%)