Cohesin and condensin extrude loops in a cell-cycle dependent manner

Chromatin undergoes a dramatic reorganization during the cell cycle1–3. In interphase, chromatin is organized into compartments and topological-associating domains (TADs) that are cell-type specific4–7, whereas in metaphase, chromosomes undergo large-scale compaction, leading to the loss of specific boundaries and the shutdown of transcription8–12. Loop extrusion by structural maintenance of chromosomes complexes (SMCs) has been proposed as a mechanism to organize chromatin in interphase and metaphase13–19. However, the requirements for chromatin organization in these cell phases are very different, and it is unknown whether loop extrusion dynamics and the complexes that extrude them also differ. Here, we used Xenopus egg extracts to reconstitute and image loop extrusion of single DNA molecules during the cell cycle. We show that loops form in both metaphase and interphase, but with distinct dynamic properties. Condensin extrudes asymmetric loops in metaphase, whereas cohesin extrudes symmetric loops in interphase. Our data show that loop extrusion is a general mechanism for the organization of DNA, with dynamic and structural properties that are molecularly regulated during the cell cycle.

Finally, the amount of DNA to the left and right of the loop corresponds to the integrated intensity of the DNA strands outside the loop region (see Methods). This assay allows us to observe loop extrusion in extract, to quantify the partitioning of DNA between the looped and outer regions, and to examine the symmetry of the reeling-in of DNA. When applied to nucleosome-depleted extract arrested in metaphase, this assay showed that loops in metaphase are extruded at 2.3 ± 0.5 kb/s (mean ± SEM) and appeared to be extruded from one side of the loop (Fig. 2Bi, 2C, and Fig. S2).
To further characterize the symmetry of extrusion in metaphase, we quantified the total decrease in DNA from the left and right regions of the loop between the onset of loop formation and the final steady-state size of the loop (Fig. 2Bii). We used these quantities to define a symmetry score as the relative difference between the decrease of these two regions and the total amount of extruded DNA. The histogram of symmetry scores peaked at 1, which corresponds to asymmetric loops (Fig. 2Biii).
Our assay allows the final steady state amount of DNA in the loop to be determined as a function of the end-to-end DNA-binding distance. These measurements show that the loop-extrusion process stops when the relative extension of DNA outside of the loop reaches ~75%, with a corresponding stall force of 0.41 pN ± 0.01 pN (mean ± SEM), (Fig. 2D). Thus, our analysis demonstrates that loop extrusion in metaphase is preferentially one-sided, with extrusion speeds and stall forces similar to those measured in vitro 24,27,28 .
Next, we used nucleosome-depleted extract in interphase to investigate whether the dynamics of loop extrusion share similar properties during the cell cycle ( Fig. 2Civ-vi).
Loop extrusion in interphase displayed a similar distribution of extrusion rates of 2.1 ± 0.5 kb/s, and stall forces, 0.23 pN ± 0.01 pN, (Fig. 2C, 2D). However, the distribution of symmetry scores of these loops peaked towards zero, suggesting that these loops are preferentially extruded symmetrically. Thus, the mechanisms of loop extrusion differ between interphase and metaphase.
The different dynamic properties of loop formation we observe in interphase and metaphase suggest that different molecular activities may be responsible for loop formation during the cell cycle 15,29 . Recent work has suggested that cohesin could extrude loops in interphase, though this activity has not been directly visualized in cellular contexts [30][31][32] . Thus, cell-cycle-dependent activities of condensin and cohesin could account for the transition between symmetric and non-symmetric loop extrusion 15 . To assess the role of cohesin and condensin during loop extrusion in interphase and metaphase, we selectively depleted these proteins in egg extract using antibodies (Fig. S3). We then tested for loop extrusion activity in each depleted condition. We found that, in metaphase, the occurrence of loop extrusion was significantly (p<0.01) reduced upon depletion of condensin I and II, but was unaffected by cohesin depletion (Fig. 3A). In contrast, there was a significant (p<0.01) decrease in loop extrusion following cohesin depletion in interphase, but was unaffected by condensin depletion (Fig. 3A). We confirmed these depletions with immunostainings that showed colocalization of cohesin and condensin with the loops observed in interphase and metaphase, respectively (Fig. 3B). Thus, cohesin extrudes symmetric loops during interphase, whereas condensin extrudes asymmetric loops in metaphase.
Our findings provide the first evidence that loop extrusion is a general mechanism of DNA organization in a cellular context, and furthermore, that it is differentially regulated during the cell cycle. This regulation is achieved by the distinct activities of cohesin 30,31 and condensin 33,34 during interphase and metaphase, and may control different levels of DNA organization during the cell cycle: from chromatin that is mostly decondensed and spatially organized into TAD structures during interphase to highly compacted chromosomes in metaphase. Symmetric loop extrusion by cohesin in interphase may ensure the formation of specific TAD boundaries by bringing together neighboring CTCF sites [35][36][37] . In metaphase, reorganization of chromosomes into compact chromatids requires the loss of boundaries and shut down of transcription 3,8,12 , which may be achieved by condensin spooling activity. However, our findings highlight the need to revise our understanding of how loop extrusion accounts for the different levels of chromatin organization in interphase and metaphase, and why different extrusion symmetries are required during the cell cycle.
Our assay will allow dissection of the molecular components that regulate the dynamic properties of loop formation in physiological contexts, and make it possible to reconstitute more complex processes such as the formation of boundary elements and the interplay between transcription, replication, and loop extrusion.   (A) Loop extrusion probability in metaphase and interphase under different depletion conditions. In metaphase, co-depleting condensin I, condensin II, and H3-H4 (using anti-XCAP-C/E and anti-H4K12Ac) significantly (p<0.01, Binomial test) reduced loop extrusion probability compared to H3-H4-depleted extract, but had no effect in interphase. In contrast, depleting cohesin and H3-H4 (using anti-XRAD21/XSMC1 and anti-H4K12Ac) significantly (p<0.01) decreased loop extrusion probability in interphase but had no effect in metaphase. (B) Snapshots of antibody stainings of representative loops in metaphase and interphase.

Xenopus laevis egg extract preparation and immunodepletion
Cytostatic factor (CSF)-arrested Xenopus laevis egg extract was prepared as described previously 38 . In brief, unfertilized oocytes were dejellied and crushed by centrifugation, generating an extract that was arrested in meiosis II. We added and monoclonal mouse antibodies to detect tubulin using anti-DM1a (1:10000, MPI-CBG antibody facility). We detected primary antibodies using LI-COR IRDye secondary antibodies and imaged the western blots on an Odyssey Infrared Imaging System. We analyzed the blots using FIJI.

Antibody production and labeling
We raised rabbit polyclonal antibodies for immunodepletion against peptides SDIVATPGPRFHTV and DLTKYPDANPNPND corresponding to antibodies that target cohesin's XRAD21 and XSMC1 subunits. We also raised rabbit polyclonal antibodies against peptides AAKGLAEMQSVG and SKTKERRNRMEVDK corresponding to antibodies that target XCAP-C and XCAP-E for both condensin I and II for immunodepletion 34 . We added a cysteine residue on the peptide's Nterminus for sulfhydryl coupling, and subsequent keyhole limpet hemocyanin conjugation and affinity purification was performed by MPI-CBG antibody facility. We labeled antibodies with fluorophores for localization following the small-scale on-resin labeling technique from 40 . Briefly, we prepared a 200-μl pipette tip to act as our resin bed. We then loaded 40 μl of rProtein A Sepharose (GE Healthcare) resin into the tip, washing three times with 10 mM K-HEPES (pH=7.7), 150 mM NaCl. We labeled both the antibody targeting the cohesin subunit XRad21 and the antibody targeting condensin I and II's subunit XCAP-C. We flowed 70 μg antibody 5 times consecutively through the packed resin bed in order to bind the antibody to the resin.
The resin was then washed three times with 200 mM K-HEPES (pH=7.7). We then added 0.5 μl 50 mM NHS-Ester-Alexa488 (Alexa Fluor TM NHS Ester, A20000, Thermo Fischer) to 25 μl 200 mM K-HEPES (pH=7.7), and immediately added it to the resin, incubating the resin, antibody, and dye for 10-60 minutes at room temperature. To remove the unbound dye, the resin bed was washed 5 times with 10 mM K-HEPES (pH=7.7), 150 mM NaCl. We eluted the labelled antibody with 5x15 μl of 200 mM acetic acid. We neutralized each eluate immediately with 5 μl of 1 M Tris-HCl, pH=9, and cooled to 0 °C. The labelled antibody is stable for months kept at 4 °C.

DNA functionalization
To biotinylate DNA purified from λ-phage (λ-DNA) 41 , we combined 10 μg of λ-DNA (NEB, N3011S) and 5 μl of a 10X polymerase buffer (50 mM Tris-HCl, pH=7.2, 10 mM MgSO4, 100 μM DTT) to a total reaction volume of 50 μl. We then heated the mixture up to 65 °C for 7 minutes to break apart the λ-DNA's sticky ends. After heat treatment, we added 100x molar excess of biotinylated dATP, biotinylated dUTP, and dGTP, and dCTP. We then added 1 unit (~1 μl) of Klenow enzyme, mixed well, and incubated overnight at room temperature. We purified the biotinylated λ-DNA using ethanol precipitation and stored at -20 °C.

PEGylation of cover slips and DNA micro-channel preparation
We functionalized glass cover slips with mPEG and PEG-Biotin. Briefly, we sonicated coverslips first in acetone for 15 minutes followed by 5 rinses with MilliQ water, and then another sonication step in 5 M KOH for 40 minutes. After rinsing the coverslips 3 times with water and then 3 times with methanol, we dried the coverslips with N2. We silanized the coverslips combining 250 ml methanol, 12.5 ml acetic acid, and 2.5 ml (3-aminopropyl)-trimethoxysilane, incubating the coverslips in this mixture for 10 minutes at room temperature, sonicating for 1 minute, and then incubating the coverslips for an additional 10 minutes. Next we rinsed the coverslips once with methanol, once with water, and once again methanol, and dried with N2. Then we mixed 100 mg mPEG and ~1.5 mg Biotin-PEG with 450 μl PEGylation buffer (0.1M Sodium Bicarbonate, pH=8.5), and spun the reaction at 10000 RPM for 1 minute. We pipetted 25 μl of the PEG mixture onto a dried, silanized coverslip and put another coverslip on top, generating a coverslip sandwich. We incubated these sandwiches over night in distilled water-filled pipette tip-boxes in the dark. After incubation, we carefully disassembled the coverslips, rinsed with MilliQ water, and dried with N2. To generate a channel for imaging, we first drilled holes through a cleaned cover slide-these holes acted as channel inlets and outlets. We placed custom-designed, laser-cut doublesided tape onto the coverslip, defining the channel geometry. We then placed a functionalized PEG-biotinylated coverslip on top of the double-sided tape, sealing the channel on either end with Valap. We filled the channel with ~10-15 μl of 0.1 mg/ml free streptavidin, incubating the channel with streptavidin for 1 minute. To remove the free, unbound streptavidin, we flushed ~100 μl channel washing buffer (40 mM Tris-HCl, pH=8.0, 20 mM NaCl, 0.4 mM EDTA) through the channel, using the drilled holes as channel inlets and outlets. We added 20 μl of 1:1000 biotinylated λ-DNA (~5 pM), incubating it for ~1 min and then washed the channel with 3x100 μl of channel washing buffer.

Imaging
For live imaging of looping events, we fluorescently stained immobilized DNA strands with 50nM Sytox Orange (S11368, ThermoFisher), a DNA intercalating dye, in imaging buffer (50mM Tris-HCl pH 7.7, 50mM KCl, 2.5mM MgCl2, 2mM ATP) similar to 24 or Xenopus Buffer (XB: 100mM KCl, 1mM MgCl2, 0.1mM CaCl2, 2mM ATP). We excited Sytox Orange-labelled DNA using a 561nm laser, and imaged the strands using a Nikon Eclipse microscope stand with a Nikon 100x/NA 1.49 oil SR Apo TIRF and an Andor iXon3 EMCCD camera using a frame-rate of 100 -300ms. A highly inclined and laminated optical sheet (HILO) microscopy mode was established using a Nikon Ti-TIRF-E unit mounted onto the microscope stand to improve signal-to-noise ratio by cutting off background fluorescence signal from unbound DNA dye in the buffer. To trigger the formation of DNA loops, we flowed about 2ul of nucleosome depleted extract into the channel (total channel volume ~10ul) and let the extract diffuse further down the channel. We then imaged looping events at the moving front of the diffusing extract.

Hydrodynamic stretching of loops
To visualize DNA loop topology which cannot be observed in the normal mode of data acquisition, we hydrodynamically stretched DNA strands that exhibited looping events using a flow-controlled syringe pump (Pro Sense B.V., NE-501), see also Movies 9-12. The flow direction was set to be perpendicular to the strand orientation by a crossshaped channel design. Depending on the width of the channel, we used flow rates between 100 μl/min and 500 μl/min to extend DNA loops. Specifically, we introduced nucleosome depleted extract into the channel as described above and, upon loop formation, stretched DNA strands by flowing imaging buffer from the opposite side.

Loop extrusion analysis
DNA traces were analyzed using custom-written Python scripts motivated by 24 . We converted movies of fluorescent DNA molecules into one-dimensional intensity profiles by summing the intensity values along the direction perpendicular to the DNA strand in each frame. We removed the background signal using a median filter. From the summed intensity profile for each frame we built kymographs by concatenating all time points (Fig. 2 and Fig. S1) Fig. 2A).
We calculated the relative sizes of the three regions in kilo-base pairs (kb) for each time frame by multiplying the 48.5 kb total length of lambda DNA with the ratio of each summed intensity value and the total summed intensity of the strand for every time point. From these values we calculated the relative change in each region over time by subtracting the averaged ten last data points from the averaged ten first data points in each region. We used the resulting values a and b for the region left and right of the loop to assign a symmetry score for each looping event by calculating = ( , ) − ( , ) + This procedure orders the extrusion from region a and b such that the symmetry score is always positive and ranges from 0 to 1. Our symmetry score intends to quantify the amount of DNA extruded into the loop from the outer regions. A positive relative change from one side implies that the no DNA from that side has been extruded into the loop, and thus we set that change to 0 (if a >0: a=0; if b >0: b=0).
We extracted the initial loop extrusion rates from the first derivative at time point zero of a single exponential fit to the values of the loop growth over time (Fig. 2B-C). The To quantify the effect of cohesin and condensin depletion, we determined the probability of loop extrusion by counting the number of observable loop extrusion events in all data taken for one condition and dividing it by the total number of DNA strands with sufficient slack (< 70% relative extension) to support the formation of a loop for that condition.