CTCF is a DNA-tension-dependent barrier to cohesin-mediated loop extrusion – Nature

DNA constructs for use as substrates in the cohesin diffusion assay

DNA fragments containing a single HighOc1 CTCF-binding site⁵¹ (TCAGAGTGGCGGCCAGCAGGGGGCGCCCTTGCCAGA) were generated by PCR using Phusion Hot Start DNA polymerase (NEB, M0535S) and inserted into the plasmid pPlat (25,754 bp) at the FspAI (Thermo Fisher Scientific, ER1661) restriction site in either forward or reverse complement orientation using Gibson assembly⁵². The constructs were then linearized using the restriction enzyme SpeI (New England Biolabs, R3133S) and biotinylated as previously described⁴⁴.

DNA constructs for use as substrates in the loop-extrusion assay

We prepared two constructs of 31.8 kb length containing a CTCF site placed asymmetrically ~9.7 kb from one end, which enables discrimination of the orientation of the DNA construct on the basis of the binding position of CTCF. One construct was oriented such that the N terminus of CTCF points towards the longer end of the DNA (plasmid 121; used for N-terminal encounters) and the motif direction of the other construct was reversed (plasmid 128; used for C-terminal encounters). Plasmid 121 was generated using plasmids 64, 66, 67, 69, 118 and 71 (see Supplementary Table 1 for a complete list of the intermediate vectors and primers used). Plasmid 128 was generated using plasmids 64, 66, 124, 69, 118 and 71 (Supplementary Table 1). Plasmids 121 and 128 were constructed using Golden Gate cloning, using BsaI-HFv2 as the type-2 restriction enzyme (NEB, E1602). Intermediate vectors (64, 66, 67, 124, 69, 118 and 71) were generated using Gibson assembly and traditional (restriction enzyme based) cloning techniques (Supplementary Table 1) (NEB, E2621 Gibson mix; NEB, M0515 Q5 polymerase).

Biotin-containing handles were generated by a PCR reaction with primers JT337 (biotin-GACCGAGATAGGGTTGAGTG, IDT) and JT338 (biotin-CAGGGTCGGAACAGGAGAGC, IDT) on plasmid 18 pBluescript SK+ (Stratagene), using GoTaq 2 (Promega, M7845). This results in a 1,238 bp PCR fragment, which was cleaned up using Promega Wizard SV Gel and PCR Cleanup System (Promega, A9282). Fresh plasmids 121 and 128 were purified using the Qiafilter plasmid midi kit (Qiagen, 12243). After purification, the plasmids were cut with both XhoI and NotI-HF and biotin handles were cut with either XhoI or NotI-HF. The digested products were mixed together with around a 10× molar excess of the biotin handle over the linearized plasmid. Ligation was performed using T4 DNA ligase (NEB, M0202L) overnight at 16 °C and heat-inactivated the next morning for 20 min at 65 °C. The resulting 31.8 kb DNA construct was cleaned up using the ÄKTA pure system, with a homemade gel-filtration column containing approximately 46 ml of Sephacryl S-1000 SF gel filtration medium (Cytiva) in TE + 150 mM NaCl₂. The sample was run at 0.2 ml min⁻¹ and fractions of 0.5 ml were collected.

DNA constructs for use as substrates in magnetic-tweezer assays

DNA constructs for magnetic-tweezer experiments of 1.5 kb length were synthesized as described previously⁴⁸.

DNA constructs for protein expression

Human NIPBL with N-terminal Flag and Halo tags and a C-terminal 10×His tag as a tandem construct with untagged human MAU2 in pLib was described previously⁹. 6×His-Halo-EcoRI^E111Q and 6×His-tetR-Halo in pLib were described previously⁴⁴. 10×His-CTCF-Halo-Flag was inserted into pLib by combining the human CTCF ORF and the Halo-tag ORF using Gibson assembly. A C-terminal Flag-tag sequence was introduced as a 5′ overhang in the reverse primer used for Halo-tag ORF amplification. To generate 10×His-CTCF-Halo-Avi-Flag, the 10×His-CTCF-Halo-Flag vector backbone was amplified around the end of the Halo-tag sequence, at which position an Avi-tag was introduced using Gibson assembly.

Generation of a radioactively labelled dsDNA probe for EMSA

dsDNA fragments (100 bp) containing WT or scrambled versions of the HighOc1 CTCF-binding site⁵¹ (WT, TCAGAGTGGCGGCCAGCAGGGGGCGCCCTTGCCAGA) were prepared by overlap-extension PCR: two ssDNA oligos with partially overlapping sequences were used in a PCR reaction catalysed by Phusion Hot Start DNA Polymerase (NEB, M0535S) and purified using the PureLink PCR Purification Kit (Invitrogen, K3110002). A total of 1 pmol of dsDNA probe was subsequently incubated with 0.5 µl [γ-³²P]ATP (3,000 Ci mmol⁻¹, 10 mCi ml⁻¹; Hartmann Analytic, SCP-301) and T4 polynucleotide kinase (NEB, M0201S) in a 20 µl reaction at 37 °C for 1 h. T4 polynucleotide kinase was subsequently heat-inactivated by incubating the reaction at 65 °C for 10 min.

Generation of a methylated dsDNA probe for EMSA

A 100 bp dsDNA fragment containing the HighOc1 CTCF-binding site described above⁵¹ was methylated in vitro using M.SssI CpG methyltransferase (NEB, M0226S) according to the manufacturer’s protocol. To increase methylation efficiency, four rounds of methylation, each followed by DNA purification using the PureLink PCR Purification Kit (Invitrogen, K3110002), were performed. The methylation efficiency was assessed by incubating 300 ng of purified methylated DNA with 1 µl of the methylation-sensitive restriction enzyme EaeI (NEB, R0508S) in a 20 μl reaction containing 1× CutSmart buffer (NEB) at 37 °C for 1 h. The reaction products were resolved by electrophoresis on a 0.8% agarose gel and ethidium bromide staining was detected using the BioRad ChemiDoc Imaging System. The final dsDNA fragment used as unlabelled, methylated competitor in Fig. 1b was methylated with about 80% efficiency.

Generation of CTCF–Halo–Flag HeLa Kyoto cell line

HeLa Kyoto cells (RRID: CVCL_1922), a gift from S. Narumiya, were cultured as described previously³. HeLa Kyoto cells were authenticated by STR fingerprinting and tested negative for mycoplasma contamination. The CTCF-Halo-Flag HeLa Kyoto cell line was generated by homology-directed repair using CRISPR Cas9 (D10A) paired nickase⁵³. A donor plasmid comprising CTCF homology arms (719 bp and 459 bp on either side of the coding sequence stop site) and Halo-Flag were cloned into plasmid pJet1.2. Cas9 guide RNA sequences were identified using an online tool (https://crispr.mit.edu; gRNA1: CACCGCAGCATGATGGACCGGTGA; gRNA2: CACCGGAGGATCATCTCGGGCGTG) and inserted into plasmid pX335 (a gift from F. Zhang, Addgene, 42335). HeLa Kyoto cells were transfected with donor Cas9 nickase plasmids using Lipofectamine 2000 (Invitrogen, 11668019). Then, 7 days later, cells were labelled with Halotag TMR ligand (Promega, G8251) and sorted by flow cytometry (Supplementary Fig. 2). The clonal cell line was selected after verification of homozygous Halo-Flag insertion by PCR amplification of genomic DNA, immunoblotting and inspection by microscopy.

Protein expression and purification

Baculoviruses for protein expression in Sf9 insect cells (Thermo Fisher Scientific) were generated as described previously⁵⁴. Expression cultures were incubated at 27 °C for 48–60 h after infection. Cells were centrifuged, washed in PBS, frozen in liquid nitrogen and stored at −80 °C.

Purification of recombinant CTCF protein

Baculovirus-infected cell pellets from cultures supplemented with 0.1 mM ZnCl₂ were lysed by Dounce homogenization and resuspended in CTCF lysis buffer (35 mM NaH₂PO₄/Na₂HPO₄ pH 7.4, 350 mM NaCl, 0.1 mM ZnCl₂, 5% glycerol, 0.05% Tween-20 and 5 mM imidazole) supplemented with 1 mM PMSF, EDTA-free cOmplete tablet (1 per 50 ml) (Roche, 11873580001), 1 mM DTT and 0.001 U µl⁻¹ benzonase. The lysate was cleared by centrifugation at 18,000g for 1 h at 4 °C. The soluble fraction was incubated with NiNTA agarose (Qiagen, 30230) for 1 h at 4 °C and washed with CTCF buffer (35 mM NaH₂PO₄/Na₂HPO₄ pH 7.4, 150 mM NaCl, 0.1 mM ZnCl₂, 5% glycerol) supplemented with 1 mM DTT and 35 mM imidazole. For the final wash step, DTT was omitted from the wash buffer. Protein was eluted with CTCF buffer supplemented with 300 mM imidazole. The eluate was subsequently concentrated approximately twofold using a Sartorius Vivaspin 50 kDa MWCO concentrator (Sartorius, VS2031) and incubated with Anti-FLAG M2 Affinity Gel (Sigma-Aldrich, A2220) for 90 min at 4 °C. The resin was washed with CTCF buffer and incubated with Halotag TMR ligand (Promega, G8252) or Halotag Alexa660 ligand (Promega, G8472) for 15 min at room temperature. After extensive washing with CTCF buffer, the labelled protein was eluted in CTCF buffer supplemented with 0.5 mg ml⁻¹ 3×Flag peptide. The eluate was supplemented with 1 mM DTT, concentrated two- to fourfold using the Sartorius Vivaspin 50 kDa MWCO concentrator, flash-frozen and stored at −80 °C.

HeLa CTCF–Halo–Flag purification

HeLa CTCF–Halo–Flag protein was purified as described for SCC1–Halo–Flag⁹, except 20 mM Tris pH 7.5 was used in all of the CTCF purification buffers instead of 25 mM NaH₂PO₄/Na₂HPO₄ pH 7.5, and 0.1 mM ZnCl₂ was included in all of the purification buffers except for the Flag elution buffer. HeLa CTCF was labelled with JF646-HaloTag ligand. JF646-HaloTag ligand was prepared as described previously⁹.

Recombinant cohesin, HeLa cohesin, NIPBL–MAU2 and EcoRI(E111Q) protein purification

Recombinant cohesin, HeLa SCC1–Halo–Flag cohesin and recombinant NIPBL–MAU2 were purified as described previously⁹. EcoRI(E111Q)–Halo and TetR–Halo were purified as described previously⁴⁴.

EMSA

For the competition EMSA assay, 60 fmol of recombinant CTCF was mixed with 1 µg poly(dI-dC) (Thermo Fisher Scientific, 20148E) in a 20 µl reaction containing 35 mM Tris pH 7.9, 50 mM KCl, 50 mM NaCl, 5 mM MgCl₂, 0.1 mM ZnCl₂, 5% glycerol, 1 mM DTT and 50 ng µl⁻¹ BSA at room temperature for 20 min. Subsequently, 21 fmol of [γ-³²P]ATP-labelled (Hartmann Analytic, SCP-501) dsDNA probe was added in the presence of 100× unlabelled competitors (dI-dC; WT; scrambled or methylated CTCF oligo), and the reaction was incubated at room temperature for an additional 10 min. The binding reactions were loaded onto prerun (1 h, 100 V, 10 mA, ice-cold water bath, 0.5× TBE running buffer) 4% non-denaturing acrylamide gel and the samples were resolved for 1 h under the same conditions as the prerun. The gel was exposed to a storage phosphor screen overnight and analysed using a Typhoon Scanner (GE Healthcare). Images shown are representative of two independent experiments.

Recombinant CTCF single-molecule imaging characterization

CTCF flow-in, washing and imaging

Flow cells were incubated with Avidin DN (Vector Laboratories, A3100) and DNA as described previously⁹, except that pPlat containing a single HighOc1 CTCF-binding site was used instead of λ-DNA. Flow cells were washed with 400 µl WB buffer (20 mM Tris pH  7.5, 50 mM KCl, 5 mM EDTA) supplemented with 0.1 mg ml⁻¹ BSA and 10 nM Sytox Green (Thermo Fisher Scientific, S7020) or Sytox Orange (Thermo Fisher Scientific, S11368) at 50 µl min⁻¹. A total of 100 µl recombinant CTCF–Halo (labelled with TMR in experiments shown in Fig. 1 and in Extended Data Figs. 1a–c,f,g,j and 2; or labelled with Alexa 660 in experiments shown in Extended Data Fig. 1d,e) was then introduced into the flow chamber at 2.5 nM final concentration in CL100 buffer (35 mM Tris pH 7.5, 100 mM KCl, 5 mM MgCl₂, 5% glycerol, 0.005% Tween-20, 0.1 mg ml⁻¹ BSA, 1 mM TCEP) at 30 µl min⁻¹ and subsequently incubated for 4 min without buffer flow. Flow cells were then washed with CL150 buffer (CL100 buffer supplemented with 50 mM KCl) at a rate of 50 µl min⁻¹ to remove non-specifically bound CTCF molecules.

To determine the orientation of DNA molecules after image acquisition, TMR labelled EcoRI(E111Q)–Halo or TetR–Halo was flowed into the flow cells at 2 nM or 5 nM final concentration, respectively, in EcoRI buffer (20 mM Tris pH 7.5, 150 mM KCl, 0.1 mg ml⁻¹ BSA) supplemented with 10 nM Sytox Green at 30 µl min⁻¹, incubated for 4 min and washed with 200 µl of EcoRI buffer.

All recombinant CTCF single-molecule imaging characterization and cohesin diffusion assay experiments were performed at room temperature. Unless stated otherwise, time-lapse microscopy images were acquired at 4 s intervals using the Zeiss TIRF 3 Axio Observer set-up and 488 nm, 561 nm and 639 nm lasers⁴⁴. A protocatechuic acid/protocatechuate-3,4-dioxygenase/trolox oxygen scavenger system (final concentration 10 nM protocatechuate-3,4-dioxygenase, 2.5 mM protocatechuic acid and 2 mM trolox); was added to all buffers used during data acquisition.

Imaging the kinetics of recombinant CTCF association with DNA

To image the kinetics of CTCF association with DNA (Fig. 1c,d and Extended Data Fig. 1a), 0.5 nM TMR-labelled CTCF–Halo was introduced into flow cells in CL100 buffer at 30 µl min⁻¹. For the experiments shown in Fig. 1d and Extended Data Fig. 1a, images were acquired at 3.12 s intervals. For measurements of CTCF residence time on DNA (Fig. 1c and Extended Data Fig. 1f,g) images were acquired at 10.15 s intervals.

Positional analysis of recombinant CTCF on DNA

The position of recombinant CTCF on DNA was analysed in Fiji. EcoRI or TetR mediated end-labelling was used to unambiguously assign the orientation of DNA strands tethered to the surface. The distance between the centre of the mass of fluorescence intensity signal marking the DNA end and the fluorescence signal of protein was measured, and the ratio between the measured distance and the total length of the DNA molecule was calculated as a position along the DNA in bp. Single-molecule tracking of the CTCF position was performed using the custom Fiji macro KymoAnalysis_2.1.ijm.

CTCF diffusion coefficient analysis

Single-molecule tracking of the CTCF position was performed using the custom Fiji macro KymoAnalysis_2.1.ijm. Spatial positions along the DNA molecule versus time for individual molecules were converted to base pairs by multiplying the positions in micrometres by the average number of base pairs per micrometre, that is, with the factor (26,123 bp)/R, where R denotes the end-to-end length of the DNA molecule containing 26,123 bp. The MSD was calculated for individual traces and a linear regression in the form MSD(τ) = Dτ + o  was applied to the first ten timepoints (corresponding to a maximum time lag of 31.2 s). Here, D denotes the diffusion coefficient, τ is the time lag and o is an offset to correct for a finite localization uncertainty. Larger time lags were not considered for the regression to exclude artificial flattening of the MSD curves by reaching the DNA ends.

Recombinant CTCF photobleaching analysis

To quantify the number of recombinant Alexa 660 (A660)-labelled CTCF molecules bound at a CTCF DNA-binding site, A660 signals on DNA were identified in laser-profile-corrected images, subtracted from the local background, averaged over ten frames and plotted in Extended Data Fig. 1e.

Determining the residence time of recombinant CTCF on DNA

To control for fluorophore bleaching in the CTCF in vitro residence-time experiments, the dwell time of ‘on-DNA’ CTCF–HaloTMR molecules (n = 140) and ‘on-glass’ CTCF–HaloTMR–Avi–biotin molecules (n = 142) (the latter coupled to the biotin-PEGylated glass surface through Avidin DN) was determined by imaging populations of these molecules in the same microfluidic flow cell. We then performed a regression of the fluorescence lifetime distribution to an exponential function on the on-glass population to compute the photobleaching half-life, which was determined to be T_{1/2_on-glass} = 77.3 min. The ‘on-DNA’ dataset was best described by a two-exponential decay fit with a fixed percentage of events (97 out of 140, 69%) that displayed rapid unbinding, which were attributed to non-specific DNA-binding events based on their position along the DNA molecule. This resulted in residence times of T_{1/2_fast_on-DNA} = 1.2 min and T_{1/2_slow_on-DNA} = 29.2 min, corresponding to non-specific and CTCF site-specific DNA-binding events.

Neither single-exponential nor two- or three-exponential fits in which one of the components was fixed to T_{1/2_on-glass} was suitable to describe the observed data. On the basis of this and the finding that T_{1/2_slow_on-DNA} was ~2.7× shorter than T_{1/2_on-glass} (29.2 min and 77.3 min, respectively), we concluded that the off-rate of CTCF on-DNA was significantly faster than the fluorophore bleaching rate and therefore the observed on-DNA dwell time of CTCF was not significantly limited by fluorophore bleaching.

HeLa CTCF single-molecule imaging characterization

CTCF flow-in, washing and imaging

Flow cells⁴⁴ were incubated with 1 mg ml⁻¹ Avidin DN (Vector Laboratories) for 15 min and washed extensively with DNA buffer (20 mM Tris pH 7.5, 150 mM NaCl, 0.25 mg ml⁻¹ BSA (Thermo Fisher Scientific, AM2616)). A total of 150 µl of 31.8 kb DNA containing a single CTCF site and biotinylated ends was introduced into flow cells at around 20 pM final concentration at 50 µl min⁻¹ in DNA buffer supplemented with 20 nM Sytox Orange (Thermo Fisher Scientific, S11368). Flow cells were washed with 400 µl of wash buffer 2 (50 mM Tris pH 7.5, 50 mM NaCl, 2.5 mM MgCl₂, 0.25 mg ml⁻¹ BSA, 0.05% Tween-20, 20 nM Sytox Orange) at 100 µl min⁻¹, followed by 100 µl of imaging buffer (50 mM Tris pH 7.5, 50 mM NaCl, 2.5 mM MgCl₂, 0.25 mg ml⁻¹ BSA, 0.05% Tween-20, 0.2 mg ml⁻¹ glucose oxidase (Sigma-Aldrich, G2133), 35 mg ml⁻¹ catalase (Sigma-Aldrich, C-40), 9 mg ml⁻¹ b–d-glucose, 2 mM trolox (Cayman Chemical, 10011659)) and 5 mM ATP (Jena Biosciences, NU- 1010-SOL)) supplemented with 20 nM Sytox Orange at 100 µl min⁻¹. Stock solutions of glucose oxidase (20 mg ml⁻¹), catalase (3.5 mg ml⁻¹) and glucose (450 mg ml⁻¹) were prepared as described previously⁵⁵. JF646-labelled HeLa CTCF was then introduced into the flow chamber at a final concentration of 0.5 nM in 100 µl imaging buffer supplemented with 20 nM Sytox Orange at 30 µl min⁻¹. Non-specifically bound CTCF was removed by washing three times with 100 µl imaging buffer supplemented with 220 nM Sytox Orange at 100 µl min⁻¹.

All HeLa CTCF single-molecule characterization and loop-extrusion experiments were performed at 37 °C. Time-lapse microscopy images were acquired using the Zeiss Elyra 7 with Lattice SIM² equipped with 561 nm and 639 nm lasers, two PCO Edge 4.2 sCMOS cameras and a ×63/1.46 NA Alpha Plan-Apochromat oil objective. Images with an exposure time of 100 ms were acquired sequentially for each channel at 0.4 s intervals in HILO mode.

HeLa CTCF photobleaching analysis

To quantify the number of HeLa JF646-labelled CTCF molecules bound at a CTCF DNA-binding site, JF646 signals on DNA were identified in laser-profile-corrected images, subtracted from the local background and averaged over all frames before a bleaching event and plotted in Extended Data Fig. 3e. The number of bleaching steps per molecule was determined manually and indicated on Extended Data Fig. 3e. The fluorescence intensity of molecules bound at a CTCF DNA-binding site that bleached in a single step was 2.2 ± 0.6 (mean ± s.d.).

HeLa CTCF positional analysis

The position of HeLa CTCF on DNA was analysed as described in the ‘Determination of DNA loop size and position of single molecules’ section (Supplementary Note).

Cohesin diffusion assay and image analysis

Cohesin diffusion assays were performed essentially as described previously⁴⁴. CTCF was introduced into flow cells at 2 nM final concentration and incubated for 4 min as described in the ‘Recombinant CTCF single-molecule imaging characterization’ section above. Flow cells were then washed with CL150* buffer (35 mM Tris pH 7.5, 75 mM NaCl, 75 mM KCl, 1 mM MgCl₂, 10% glycerol, 0.003% Tween-20 and 0.1 mg ml⁻¹ BSA). Cohesin and NIPBL–MAU2 were introduced into flow cells at 0.8–2 nM and 2 nM, respectively, in 100 µl of CL100* buffer (35 mM Tris, pH 7.5, 50 mM NaCl, 50 mM KCl, 1 mM MgCl₂, 10% glycerol, 0.003% Tween-20 and 0.1 mg ml⁻¹ BSA) at 30 µl min⁻¹. Flow cells were incubated for a further 4 min without buffer flow and then washed with CL250* buffer (35 mM Tris pH 7.5, 125 mM NaCl, 125 mM KCl, 1 mM MgCl₂, 10% glycerol, 0.003% Tween-20 and 0.1 mg ml⁻¹ BSA). Cohesin and CTCF imaging was then performed in the absence of buffer flow for 160 s at 4 s per frame intervals. Image acquisition was repeated for 3–5 fields of view. DNA orientation was determined by flowing in Sytox Green and EcoRI(E111Q)–Halo or TetR–Halo as described in the ‘Recombinant CTCF single-molecule imaging characterization’ section above. Biotin-conjugated quantum dots QD705 (Invitrogen, Q101163MP) or CTCF–Halo–Avi–biotin were used as fiducial markers.

CTCF–cohesin channels were aligned with TetR/EcoRI(E111Q)–DNA channels using the custom-written Fiji macro Movement_analysis_macro_Kymo_10c_3Ch.ijm. Each DNA molecule containing diffusing cohesin was manually examined for the presence of a single CTCF signal positioned at the regions in which the CTCF-binding site was introduced. DNA molecules containing multiple or non-specifically bound CTCF molecules were excluded from the analysis. The number of diffusing cohesin foci on the selected DNA molecules was determined and DNA molecules containing more than four mobile cohesin foci were excluded from the analysis. Cohesin behaviour on DNA was then analysed and classified as follows. (1) Cohesin diffusion blocked: (i) cohesin diffuses freely along the DNA and reaches CTCF roadblock, bounces back but does not go past the roadblock during the time of imaging; (ii) cohesin diffuses freely along the DNA, reaches CTCF and becomes immobilized; (iii) two or more cohesin molecules blocked by CTCF. (2) Cohesin passes CTCF in one direction: cohesin passes CTCF during imaging and diffuses back towards CTCF but does not pass back to the other side. (3) Cohesin passes CTCF multiple times.

DNAs with the following events were also excluded from analysis: (1) cohesin diffusing or co-localizing with CTCF. (2) Cohesin failing to encounter CTCF. (3) Cohesin blocked by a high fluorescence intensity CTCF signal, presumably a multimer. (4) Cohesin or CTCF bleaches during image acquisition.

Loop-extrusion assay

Perpendicular flow loop-extrusion assays were performed essentially as described previously^9,55. Flow cells were incubated with 1 mg ml⁻¹ Avidin DN (Vector Laboratories) for 15 min and washed extensively with DNA buffer (20 mM Tris pH 7.5, 150 mM NaCl, 0.25 mg ml⁻¹ BSA (Thermo Fisher Scientific, AM2616)). A total of 40 µl of 31.8 kb DNA containing a single CTCF site and biotinylated ends was introduced into flow cells at about 3 pM final concentration at 15 µl min⁻¹ in DNA buffer supplemented with 20 nM Sytox Orange (Thermo Fisher Scientific, S11368). The flow cells were washed with 20 µl of wash buffer 1 (50 mM Tris pH 7.5, 200 mM NaCl, 1 mM MgCl₂, 5% glycerol, 1 mM DTT, 0.25 mg ml⁻¹ BSA, 20 nM Sytox Orange) at 5 µl min⁻¹. Flow was then switched to perpendicular mode and a further 350 µl of wash buffer 1 was introduced at 100 µl min⁻¹. A total of 400 µl of wash buffer 2 (50 mM Tris pH 7.5, 50 mM NaCl, 2.5 mM MgCl₂, 0.25 mg ml⁻¹ BSA, 0.05% Tween-20, 20 nM Sytox Orange) was then introduced at 100 µl min⁻¹, followed by 100 µl of imaging buffer (50 mM Tris pH 7.5, 50 mM NaCl, 2.5 mM MgCl₂, 0.25 mg ml⁻¹ BSA, 0.05% Tween-20, 0.2 mg ml⁻¹ glucose oxidase (Sigma-Aldrich, G2133), 35 mg ml⁻¹ catalase (Sigma-Aldrich, C-40), 9 mg ml⁻¹ b–d-glucose, 2 mM trolox (Cayman Chemical, 10011659)) and 5 mM ATP (Jena Biosciences, NU-1010-SOL)) supplemented with 20 nM Sytox Orange at 100 µl min⁻¹. JF646-labelled CTCF was then introduced into the flow chamber at 0.5 nM final concentration in 100 µl imaging buffer supplemented with 20 nM Sytox Orange at 30 µl min⁻¹. Non-specifically bound CTCF was removed by washing three times with 100 µl imaging buffer supplemented with 220 nM Sytox Orange at 100 µl min⁻¹. HeLa cohesin and recombinant NIPBL–MAU2 were then introduced into the flow chamber at 0.5 nM and 3.54 nM, respectively, in 250 µl imaging buffer supplemented with 220 nM Sytox Orange at 30 µl min⁻¹.

For loop-extrusion assays in the absence of buffer flow, flow cells were incubated with Avidin DN and washed with DNA buffer as described above. DNA was introduced at 15–25 µl min⁻¹ to vary the DNA tension. Flow cells were then washed and incubated as above without switching to perpendicular mode.

dCas9 binding to DNA

crRNA sequences were chosen at around one-third of the DNA length and, at each end, two sequences were used for efficient binding of the dCas9–gRNA complex per DNA. If located at the same ends, crRNA sequences were spaced at least 2 kb apart to allow discrimination (additionally to bleaching curves) of occasional binding of two dCas9–gRNA complexes per DNA end. Binding sequences were chosen using CRISPOR (http://crispor.tefor.net/crispor.py; PAM indicated in bold): seq7932, ACTGGACTGCGACCGGGCAGGGG; seq11802, CGCGGTGGAGGCAGACGTGGCGG; seq18967, CTGGTTATGCAGGTCGTAGTGGG; and seq21005, GGCATACAAATATTCCATGAAGG.

gRNA was obtained by annealing a mixture of universal 67-mer Alt-R CRISPR–Cas9 ATTO550-labelled tracrRNA and crRNA (IDT) matching the binding sites at 95 °C for 2.5 min and slow cooling to 5 °C in steps of 5 °C for 2.5 min each. To couple gRNA to dCas9, 200 nM dCas9 (NEB, NEBM0652T) was mixed with 2 µM gRNA on ice in NEBuffer3.1, incubated at 37 °C for 10 min and placed on ice again.

To bind the dCas9–gRNA complex to DNA, DNA constructs of 31.8 kb length were used to facilitate measurements at a similar end-to-end length and force regime as for the CTCF experiments. DNA was bound to the pegylated glass surface and unbound DNA was washed off with 100 µl imaging buffer. Then, 1 nM dCas9–gRNA was flushed into the flow cell and incubated for 5 min. Non-specifically bound dCas9–gRNA was removed by flushing with 100 µl imaging buffer supplemented with 1 mg ml⁻¹ heparin. Heparin was removed by washing with 100 µl imaging buffer. This typically left one to two dCas9–gRNA complexes per DNA. Loop-extrusion experiments were then performed as described above with 30 pM cohesin and 75 pM NIPBL–MAU2. DNA was visualized by staining with 25 nM Sytox Green and exciting with a 488 nm laser. gRNA–ATTO550 was excited by 561 nm laser light in an alternating excitation scheme using a ×60 oil-immersion, 1.49 NA CFI APO TIRF (Nikon) objective. Emission was collected on a Photometrics Prime BSI sCMOS camera using continuous imaging and an exposure time of 100 ms per frame.

Magnetic-tweezer experiments

The magnetic-tweezer instrument and experiments were conducted essentially as described previously⁴⁸ with minor modifications. The instrument consisted of a pair of vertically aligned (1 mm apart) permanent neodymium-iron-boron magnets (Webcraft) that were was used to generate the magnetic field⁵⁶. The magnet pair was placed on a motorized stage (translation: Physik Instrumente, M-126.PD2; rotation: Physik Instrumente, C-150.PD) and the light of a red LED (λ = 630 nm) was allowed to pass the magnet pair gap to illuminate the sample. Transmission was collected by a ×50 oil-immersion objective (CFI Plan 50XH, Achromat; NA = 0.9, Nikon), and the bead diffraction patterns were recorded with a four-megapixel CMOS camera (Falcon, 4M60; Teledyne Dalsa) at 50 Hz. The real-time tracking of the magnetic bead movement in all three dimensions was conducted using LabView 2011-based (National Instruments) control software described and published previously^57,58. Surface-adhered 1.5 µm polystyrene reference beads (PolySciences) were used as a reference to correct for instrumental drift occurring during measurements. In total, 100–200 beads could be tracked simultaneously in one field of view with a spatial resolution of around 2 nm for the 1.5-kb-long dsDNA tethers⁴⁸.

The flow cell and DNA tethering were prepared as described previously⁴⁸. In brief, the reference beads were diluted 1:1,500 in PBS buffer (pH 7.4; Sigma-Aldrich) and then adhered (~5 min) to the cover glass surface of the flow cell. After removal of non-adhered beads by washing with PBS, sheep digoxigenin antibodies (Roche) at a concentration of 0.1 mg ml⁻¹ were incubated in the flow cell for 1 h, after a 500 µl wash with PBS and 2 h incubation with 10 mg ml⁻¹ BSA (New England Biolabs, UK) diluted in PBS (pH 7.4) buffer. After washing with 500 µl PBS buffer, 1 pM of the 1.5 kb linear dsDNA construct was incubated in PBS buffer for 20 min in the flow cell. After washing again with 500 µl PBS buffer, Streptavidin-coated superparamagnetic beads (DynaBeads MyOne, LifeTechnologies; diluted 1:400 in PBS) with a diameter of 1 µm were added resulting in the attachment of the beads to the surface-tethered dsDNA constructs after around 5 min; unbound beads were washed out afterwards with PBS.

Before the cohesin loop-extrusion experiments, the quality of tethered dsDNA constructs was assessed by applying a combination of zero and high force (8 pN), and 30 rotations in each direction at high force. Only tethers with singly bound dsDNA and correct DNA end-to-end lengths were used for the subsequent single-molecule experiments. After washing the flow cell with cohesin reaction buffer (40 mM Tris-HCl pH 7.5, 50 mM NaCl, 2.5 mM MgCl₂, 1 mM DTT, 0.25 mg ml⁻¹ BSA, 0.05% Tween-20), 0.1 nM cohesin and 0.25 nM NIPBL–MAU2 were introduced in cohesin buffer supplemented with 2 mM ATP to stretched dsDNA tethers at high force (8 pN). For force-titration experiments (Extended Data Fig. 7), the force was lowered in individual experiments to 1, 0.8, 0.6, 0.4, 0.3, 0.2 and 0.1 pN, and maintained for 10 min. All magnetic-tweezer experiments were performed at room temperature.

The Z-bead position over time was extracted using custom-written scripts in IGOR Pro (v.6.37, Wavemetrics), as previously described^48,59 and a custom-written automated step detection algorithm (MATLAB, MathWorks) was applied to the individual traces as described previously^48,60 to extract individual loop-extrusion step sizes. Step sizes measured under the same conditions from different traces and experiments were pooled and converted into base pairs⁴⁸ to construct the distribution of cohesin step sizes in dependence of force (Extended Data Fig. 7c,d).

Simulating the encounter probability of cohesin and CTCF, given force-dependent cohesin step sizes

A 10 kb stretch of DNA was simulated on which CTCF was assumed to be positioned 7 kb from one end. The cohesin-binding site was uniformly sampled along the DNA length. For each force value, step sizes were sampled from the empirically obtained distribution as measured by magnetic-tweezer experiments. The simulations were repeated 500 times for every force value and events in which cohesin came within 50 bp of CTCF were counted as encounters, which constitutes a conservative threshold for the interaction distance between cohesin and CTCF.

Statistical analysis and reproducibility

Statistical analysis was performed using GraphPad Prism (v.9.4.1) or Python (v.3.7.7) using scipy (v.1.5.2)⁶¹, numpy (v.1.21.6), trackpy (v.0.4.2)⁶² and statsmodels (v.0.12.2). No statistical methods were used to determine sample size. Experiments were not randomized and the investigators were not blinded to allocation. Figures were assembled using Adobe Illustrator 27.2. All of the experiments were performed at least twice with consistent results. The experiments shown in Fig. 1a,b and Extended Data Figs. 1h,i and 3a,f were performed twice with consistent results. The number of replicates for the experiments shown in Figs. 1g,h, 2e,g and 3c–e and Extended Data Figs. 5b,g,h, 7c, 9a,b,g–l and 10 is listed in the respective figure legends.

Reporting summary

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