Cohesin reshapes replication fork contacts to aid fork slowing and reversal

Replication of organized genomes is spatiotemporally regulated, in which topologically associating domains (TADs) containing marks of open chromatin and active transcription replicate earlier than compact, transcriptionally repressed TADs^11,12. Initiation of DNA replication takes place at DNA replication origins that spread along the genome^13,14, which are enriched in initiation zones, defined by population studies¹⁵. Recent replication-specific Hi-C analyses revealed nascent DNA contacts within the same initiation zone, possibly representing sister forks of the same replicon and/or converging forks from neighbouring replicons, which are physically coupled during DNA synthesis⁵. However, how nascent DNA organization is affected by conditions interfering with fork progression (that is, replication stress (RS)) remains unknown. RS triggers changes in the composition of the replication apparatus and in the architecture of replication forks. This ‘fork plasticity’ assists cells tolerating DNA damage and replication interference^1,11. One such mechanism is fork reversal, the active and reversible remodelling of the replication fork into a four-way junction through annealing of the two nascent DNA strands. This requires RAD51 and the DNA translocases SMARCAL1, HLTF and ZRANB3. An alternative mechanism sustaining DNA synthesis under stress is repriming by the DNA primase and polymerase PRIMPOL to promote discontinuous DNA synthesis¹. How these mechanisms are used locally and globally remains unknown and may reflect chromatin architecture and the accessibility of nascent DNA to different stress-tolerance factors.

The cohesin complex (SMC1, SMC3, RAD21 and either STAG1 or STAG2) organizes the genome into TADs and chromatin loops because of its ability to extrude DNA^16,17,18. This function requires its association with the cohesin chromatin loader NIPBL–MAU2 (refs. ^2,19). Loop-extruding cohesin is stopped at certain locations by protein barriers, such as CTCF, or the MCM complex belonging to the replicative CMG helicase (CDC45–MCM–GINS)^20,21. Cohesin also interacts with replicative pre-initiation complexes and aids localizing replication origins and initiation zones at early replicating domains^3,4. During DNA replication, different mechanisms to embrace both sister chromatids by cohesin (cohesion) have been proposed and possibly coexist. These include replication fork passage through the ring, transient ring opening or cohesin conversion at replication termination events^22,23. Binding of Sororin to cohesin complexes promotes cohesion by preventing cohesin unloading by WAPL²⁴. In yeast, the MRX complex promotes cohesin recruitment to stalled replication forks aiding their stability and restart^25,26,27. However, mechanisms of fork stability differ greatly between yeast and human cells. Although fork reversal in yeast is mainly found in checkpoint-deficient mutants, in human cells it evolved as one of the most frequent physiological responses at stressed replication forks, assisting genome stability¹. Moreover, yeast cells do not possess PRIMPOL and mainly use Polα primase for this purpose^28,29. In human cells, the cohesin complex rapidly accumulates at stalled replication forks³⁰, and excessive accumulation due to unloading failure impairs replication fork progression^31,32. However, whether physiological levels of cohesin actively restrain fork progression upon RS remained unknown. In this study, we characterized the molecular functions and dependencies of cohesin stabilization at stalled replication forks. We describe how cohesin reshapes replicating domains and drives fork plasticity, mediating the cellular response to genotoxic agents.

Using ‘isolation of proteins on nascent DNA’ coupled to mass spectrometry³³ (iPOND–MS; Fig. 1a), we characterized the proteome landscape of replication forks when cells experience two different treatments with the ribonucleotide reductase inhibitor hydroxyurea (HU): 2 h at 2 mM, which stalls all ongoing replication forks by severely depleting the cellular nucleotide pool³⁴; and 1 h at 100 µM, which induces fork slowdown³⁵ through mild reduction of cellular nucleotides and by promoting TIMELESS dissociation at forks through reactive oxygen species (ROS) and disruption of oligomerized PRDX2 (ref. ³⁶). A thymidine chase control to enrich for mature, replicated chromatin-associated proteins (Thy) and a negative control, in which azide-biotin is substituted by DMSO, were also included (Extended Data Fig. 1a). As expected, mature, replicated chromatin is enriched in core histone proteins and lamins, whereas the untreated replication fork (UNT) is enriched in replisome core components (for example, POLE, POLD1, RFC1-5; ref. ³⁰) (Extended Data Fig. 1b). After 2 h in 2 mM HU, we observed the enrichment of proteins involved in the RS response (for example, FANCD2/I, BLM, RAD51), in line with previous studies³⁰ (Fig. 1b). Expectedly, fewer proteins were recruited to nascent DNA after 1 h at 100 µM HU (Extended Data Fig. 1c). Gene ontology analysis showed a high enrichment of factors involved in chromosome organization and other expected biological functions, such as DNA damage or cell cycle regulation factors (Extended Data Fig. 1d,e). Among those factors involved in chromosome organization, we found several subunits of the cohesin complex, especially enriched after 2 h at 2 mM HU (Fig. 1b). Importantly, following 1 h 100 µM and 2 h 2 mM HU treatments, STAG2, NIPBL, SMC3 and Sororin (CDCA5) were significantly enriched (Extended Data Fig. 1f). To support these results with an orthogonal assay, we used ‘in situ protein interactions with nascent DNA replication forks’ (SIRF), a proximity ligation-based assay (PLA) between cohesin and EdU-labelled DNA³⁷ (Fig. 1c). We observed increased presence of SMC1 and RAD21 at nascent DNA after treatment with HU or with other fork stalling agents, for example, DNA polymerase α and δ inhibition by Aphidicolin (Aph) and a synthesis-permissive treatment with the DNA crosslinker Mitomycin C (MMC) (Fig. 1c,d and Extended Data Fig. 1g). Cohesin recruitment to nascent DNA was validated by proximity of SMC1 and the replisome component MCM2 (Extended Data Fig. 1h). We ruled out the possibility that double-strand breaks (DSBs) were responsible for cohesin accumulation at forks by analysing γH2AX (replication stress or DSB maker) and pATM foci (DSB signalling marker) in S-phase cells treated with 2 mM HU or γ-irradiation. Although γH2AX was strongly induced following 2 mM HU, we observed a mild, non-significant increase for pATM. By contrast, after γ-irradiation, both markers were significantly increased (Extended Data Fig. 1i). This is in line with previous reports showing fork collapse only at prolonged times of high HU treatment³⁸.

Next, we asked which cohesin function drives its accumulation at stalled forks. To interrogate loop extrusion, we first confirmed by SIRF that NIPBL levels increase at stalled forks after 2 mM HU (Fig. 1e), in agreement with our iPOND data (Fig. 1b) and others³⁰. Next, we generated HCT116 OsTIR1 (F74G)-expressing cell lines, in which NIPBL was endogenously and homozygously fused to a mini auxin-inducible degron (AID) and mini-Clover (mAID-mClover) protein tag, as a rapid and efficient inducer of protein degradation⁶ (Extended Data Fig. 2a). We confirmed a marked reduction of NIPBL cellular levels after addition of auxin (aux, 5-pH-IAA) in two independent clones (3 and 13) (Fig. 1f). Cell cycle progression and EdU incorporation in S-phase are not visibly affected by aux treatment (Extended Data Fig. 2b). Micro-C experiments show impaired TAD formation and insulation, decreased loop formation and increased compartmentalization (long-range interactions), as expected from a defect in loop extrusion³⁹ (Extended Data Fig. 2c–f). Finally, after 2 mM HU treatment, NIPBL depletion abolished cohesin SMC1 and RAD21 accumulation at stalled forks (Fig. 1g,h and Extended Data Fig. 2h) with a marginal, non-significant increase in cohesin at stalled forks after HU treatment in NIPBL-depleted cells, probably reflecting incomplete NIPBL depletion in our degron system. This effect cannot be explained by impaired cohesin loading in the absence of NIPBL, because cohesin levels on chromatin are about 75–90% of those in the presence of NIPBL (Extended Data Fig. 2g,i), consistent with previous reports^40,41.

As Sororin (CDCA5) was enriched at stalled forks in the iPOND–MS data (Fig. 1b), we tested whether the Sororin-mediated establishment of cohesion contributes to the observed accumulation of cohesin at forks. Using a degron strategy, validated by the loss of cohesion upon Sororin depletion (Extended Data Fig. 3a–c), we discarded this possibility for RAD21 and SMC1 subunits (Extended Data Fig. 3d). Likewise, we could not find any effect of WAPL depletion on cohesin recruitment at stalled forks, even if WAPL depletion proved to shift genomic contacts to a longer range, decreasing TAD insulation and increasing loop length (Extended Data Fig. 3e–j). These observations indicate that NIPBL is the only cohesin-associated factor required for its accumulation at stalled forks.

To address whether the lack of cohesin accumulation at stalled forks after NIPBL depletion is a consequence of deficient loop extrusion, disrupted genome organization or both, we analysed cohesin recruitment at HU-stalled replication forks after CTCF elimination using a previously generated degron cell line⁶. As expected, CTCF depletion impaired TAD formation and weakened insulation scores⁴². We observed a non-significant trend towards decreased cohesin enrichment near forks in the absence of RS but no effect on cohesin recruitment to stalled forks (Extended Data Fig. 3k–p). This result rules out an indirect effect of disrupted TAD organization in the absence of NIPBL and rather indicates that cohesin loading and/or loop extrusion are essential for cohesin accumulation at stalled forks.

If cohesin is recruited to stalled forks by NIPBL-dependent loop extrusion², it may be required to reorganize chromatin contacts at replicating regions in response to RS. To test this hypothesis, we developed Repli-C, a Micro-C-based methodology to look at genomic contacts established at replicating regions by ligating DNA contacts after pulling down EdU-labelled nascent DNA (Fig. 2a). In short, cells were labelled with EdU for 15 min to focus our observations on the first kilobases of nascent DNA. EdU was linked to digoxigenin (Dig) through a click reaction. Cells were then exposed to controlled MNase digestion and chromatin contacts containing Dig–EdU were isolated. Finally, contacts were ligated, and libraries were prepared and sequenced. We obtained high-resolution maps of replication-associated genomic DNA contacts and identified ‘fountain-like’ structures at replication origins, previously described with Repli-Hi-C, which probably correspond to coupled sister forks⁵. These origins were identified as sites of replication fork divergence by analysing inflection points in replication fork directionality (RFD) profiles derived from polymerase usage sequencing (Pu-seq) data⁴³. Fountain structures appear specifically in Repli-C maps, and not in parallel Micro-C analyses, confirming that they are replication-associated structures (Fig. 2b). To test the effect of cohesin depletion on these structures, we generated SMC1-mClover-mAID-expressing HCT116 cells, in which SMC1 can be depleted to less than 5% of the endogenous levels after a 5 h aux treatment (Extended Data Fig. 4a). We performed Repli-C in cells with or without cohesin and treated or not with 2 mM HU (Extended Data Fig. 4b, lanes 1–4). Visual inspection of Repli-C maps from all four conditions showed that fountains at origins get extended when cohesin is absent (Fig. 2b,c). This suggests that these replicative structures are largely cohesin-independent; their extended detection in the absence of cohesin may reflect the disruption of alternative contacts that replicating regions establish with other regions in a cohesin-dependent manner. This is in line with a recent report in yeast, showing cohesin-independent fountain structures at early-firing origins in Hi-C maps⁴⁴. Quantification of the number and length of fountains by fountain detection analysis⁵ showed that SMC1 depletion generates many new fountains. Moreover, SMC1 depletion increased the fountain length, whereas HU consistently reduced it independently of cohesin (Fig. 2c). Hence, cohesin restricts fountain extension in unchallenged conditions, whereas HU promotes fountain shortening, suggesting that alternative cohesin-dependent and independent contacts may limit sister-fork proximity under stress.

Next, we plotted Repli-C contacts around all replication origins defined by Pu-seq⁴³. We observed that HU shortened the fountain-like structure and concentrated new interactions close to the diagonal, that is, close to the origins of replication (intra-replicon contacts). This effect was not observed in the absence of cohesin, despite better detection of fountain-like structures (Fig. 2d, quantification in Fig. 2e). Importantly, no enrichment was found on Micro-C maps (Extended Data Fig. 4c), confirming the specificity of these contacts for replicating DNA. Similar observations were made when plotting Repli-C interactions at initiation zones identified by high-resolution Repli-seq¹⁵ (Extended Data Fig. 4d). Thus, after HU treatment, cohesin function is required to stabilize contacts at chromosomal regions proximal to origins.

In yeast Hi-C experiments, stalled replication forks were shown to act as barriers for cohesin-mediated loop extrusion⁴⁵. Moreover, we noticed that HU markedly increased ‘stripes’ (Fig. 2f)—typically representing anchored loop extrusion in Hi-C and Micro-C maps—and confirmed that HU-induced stripes are dependent on cohesin (Extended Data Fig. 4e). We reasoned that increased loop extrusion on HU could be connecting replication forks from neighbouring replicons. To explore this possibility, we plotted Repli-C contacts at pairs of replication origins separated by a distance of 200–400 kb (in line with the average size of a replicon being 200 kb; ref. ¹²). In these Repli-C metaplots, clear contacts are detected between yet distant replication forks of neighbouring replicons—with each origin showing fountain-like structures (Fig. 2g)—while control Micro-C maps display spread contacts between the origins, similar to intra-TAD contacts, possibly owing to the frequent positioning of origins at TAD borders¹¹ (Extended Data Fig. 4f). In both cases, inter-fork contacts disappear after SMC1 depletion. HU treatment markedly increased the frequency of these inter-replicon fork contacts, and this effect was entirely dependent on cohesin (Fig. 2f–h). Hence, cohesin establishes physical contacts between distant forks, which are stabilized under replication stress.

To further address the mechanism by which cohesin assists the response to replication stress, we set out to test recently described cohesin mutants^7,8,9,46. In the SMC1^3D mutant, three lysine residues in the hinge domain are replaced by aspartic acid, and the resulting complex can sustain loop extrusion in vitro and shows a mild loop extrusion defect in vivo but is completely deficient in cohesion and cell proliferation⁸ (Fig. 2i). In the SMC1^4E mutant, four K/R residues critical for DNA binding are replaced by alanine, and the complex is unable to perform loop extrusion in vitro⁷, albeit partially retaining this function in vivo^9,46 (Fig. 2i).

Flag-tagged SMC1^WT, SMC1^3D or SMC1^4E were ectopically expressed under the control of doxycycline (dox) in SMC1-mClover-mAID cells. We assessed the impact of these genetic perturbations on long-term cellular proliferation, by adding dox for 48 h before inducing endogenous SMC1 degradation for 7 days (Extended Data Fig. 5a). As expected, depletion of SMC1 had a major impact on cell proliferation that could be rescued by expression of ectopic SMC1^WT and SMC1^4E, but not SMC1^3D. This result confirms that the SMC1^3D, but not the SMC1^4E mutation, impairs the essential function of cohesin in cohesion establishment (Extended Data Fig. 5a,b). Shortly after SMC1 replacement (48 h dox followed by 4 h aux), total cellular levels of all SMC1 variants were lower than endogenous SMC1 (Extended Data Fig. 5c). Nonetheless, they were recruited to chromatin as part of cohesin (Extended Data Fig. 5c,d), and complexes containing SMC1^4E but not SMC1^3D became acetylated and associated with Sororin, consistent with their ability and inability to perform cohesion, respectively (Extended Data Fig. 5d). Importantly, in these short-term experiments (48 h dox, 4 h aux), depletion of SMC1 and re-expression of the SMC1 variants did not have a major impact on the cell cycle and EdU incorporation (Extended Data Fig. 5e). However, in SIRF assays with cohesin RAD21 subunit, SMC1^WT and SMC1^3D cohesin complexes significantly accumulate at forks following HU treatment, whereas SMC1^4E cohesin does not (Fig. 2j). The intermediate phenotype observed for SMC1^3D is probably due to its incomplete loop extrusion activity in vivo because chromatin-bound levels in RAD21+/EdU+ cells are comparable to SMC1^WT and SMC1^4E (Extended Data Fig. 5f). These data confirm that cohesion is not required for cohesin recruitment at stalled forks.

Next, we substituted endogenous SMC1 with SMC1^WT or SMC1^4E for 5 days in culture. This long-term substitution resulted in chromatin-bound cohesin levels similar to those in the parental cell line, as well as similar EdU incorporation (Extended Data Fig. 5g,h). In these conditions, we first interrogated genome organization by performing Micro-C. As expected, SMC1-depletion abolished TADs (Extended Data Fig. 6a), reduced contacts in the cohesin-mediated loop range (Extended Data Fig. 6b) and reduced the number of loops (Extended Data Fig. 6c), as well as TAD insulation (Extended Data Fig. 6d). Expression of SMC1^WT rescued all these phenotypes, whereas expression of SMC1^4E could only partially do so (Extended Data Fig. 6a–d), consistent with recent reports⁴⁶. Having observed a defect in loop extrusion in SMC1^4E even after prolonged substitution, we tested its ability to be recruited at stalled forks by RAD21 SIRF. SMC1^4E could not support cohesin accumulation at HU-stalled replication forks (Fig. 2k), which is not merely a consequence of lower chromatin association of cohesin (Extended Data Fig. 7a), confirming that cohesin stabilization at stalled forks specifically requires loop extrusion.

Next, we used these controlled conditions of selective loop-extrusion impairment (SMC1^4E) to assess by Repli-C the functional relevance of loop extrusion for the increased contacts induced by HU between sister forks and between forks belonging to adjacent replicons (Extended Data Fig. 4b, lanes 5–8). SMC1^4E exhibited a slight increase in fountain formation and length, and the latter was reduced after HU treatment (Extended Data Fig. 7b,c), consistent with sister-fork coupling being reduced upon RS, independently of loop extrusion. Similar to our previous experiments in the SMC1-degron cell line (Fig. 2d,e), we observed that HU treatment induces intra-replicon contacts close to the diagonal although the level was lower for SMC1^4E cells (Fig. 2l and Extended Data Fig. 7d), indicating that loop extrusion contributes to intra-replicon re-arrangement. However, a higher impact of loop extrusion was observed at inter-replicon fork contacts; SMC1^4E cells showed a striking decrease in these contacts in unchallenged conditions, that were only marginally increased after HU treatment (Fig. 2m and Extended Data Fig. 7e). Altogether, these results indicate that efficient loop extrusion stabilizes cohesin in proximity of stalled replication forks, inducing new three-dimensional (3D) contacts of replicating DNA that extend to neighbouring replicons.

Next, we tested the functional relevance of these cohesin-mediated intra- and inter-replicon fork contacts in the cellular response to RS. Cohesin accumulation due to defective unloading or regulation leads to replication fork slowdown^31,32. To test the impact of acute cohesin inactivation on RS, we used a previously described mAID-mClover-RAD21 HCT116 cell line⁶, in which efficient RAD21 degradation has no major impact on the cell cycle and EdU incorporation profiles (Extended Data Fig. 8a,b). Single-molecule analysis of replication fork progression in the presence of 50 µM HU (which induces fork-reversal-dependent slowdown by ROS, independently of nucleotide depletion^36,47), 200 nM MMC or 100 nM Aph, showed that these treatments reduced fork rate only in the presence of RAD21 (Fig. 3a,b). Depletion of NIPBL also led to unrestrained fork progression following HU addition (Fig. 3c), whereas Sororin, WAPL and CTCF depletions did not (Extended Data Fig. 8c–e). These results identify loop extrusion as the key cohesin function mediating the slowdown of stressed replication forks. Accordingly, ectopic expression of SMC1^WT and partly SMC1^3D, but not SMC1^4E, restores fork slowdown after HU treatment when endogenous SMC1 is depleted (Fig. 3d). Unrestrained fork progression in SMC1^4E-expressing cells was also observed after long-term substitutions (Fig. 3e).

Fork slowdown following HU treatment is commonly associated with fork reversal. BRCA2 protects nascent DNA at stalled forks from nucleolytic activities by stabilizing RAD51 filaments, in a process that requires fork reversal. Hence, fork degradation in BRCA2-defective cells is often used as an indirect readout of proficient fork reversal¹. After 3 h of 4 mM HU treatment, BRCA2 knockdown in mAID-mClover-RAD21 resulted in nascent DNA degradation (which impedes any DNA synthesis), as observed by the lower IdU/CldU ratio of single replication tracks (Extended Data Fig. 8f). However, co-depletion of RAD21 or NIPBL prevented this degradation (Extended Data Fig. 8f,g). Moreover, SMC1^WT and SMC1^3D, but not SMC1^4E, restore fork degradation in BRCA2/SMC1 co-depleted cells (Extended Data Fig. 8h). SMC1^4E could not restore fork degradation even in long-term substitutions (Extended Data Fig. 8i).

These results indicate that cohesin promotes fork reversal in an NIPBL-dependent manner, through its loop extrusion activity. Consistent with this possibility, fork reversal frequency assessed by electron microscopy⁴⁸ was severely reduced in NIPBL-depleted cells and cells depleted of SMC1 or expressing SMC1^4E following HU treatment (Fig. 3f and Extended Data Fig. 8j). Furthermore, we observed that cohesin accumulation at stalled forks is not altered when fork reversal is genetically inactivated by RAD51 depletion, indicating that this accumulation precedes fork reversal (Extended Data Fig. 8k).

To further investigate the consequences of the defective fork reversal and slowdown in the absence of cohesin, we interrogated the nature of the unrestrained DNA synthesis and the factors involved in RS tolerance. First, we combined DNA fibre assays with incubation with S1 nuclease, which attacks newly synthesized DNA only if it presents single-stranded DNA (ssDNA) gaps⁴⁹. IdU tracks were shortened by S1 treatment in HU-treated RAD21-depleted cells, pointing towards discontinuous DNA synthesis (Extended Data Fig. 9a). The main contributor to discontinuous DNA synthesis following fork reversal defects is the PRIMPOL primase⁴⁷. To test the possible role of PRIMPOL, we carried out PRIMPOL SIRF assays and showed that PRIMPOL increases at HU-stalled replication forks, an effect that is reversed by PRIMPOL downregulation (Extended Data Fig. 9b). Depletion of SMC1 or NIBPL, but not Sororin, further increased PRIMPOL levels at stalled replication forks (Fig. 4a and Extended Data Fig. 9c,d). To test the functional relevance of this increase, we downregulated PRIMPOL and observed that it prevented the unrestrained DNA synthesis of RAD21-depleted cells (Fig. 4b and Extended Data Fig. 9e), NIPBL-depleted cells (Fig. 4c and Extended Data Fig. 9f) and long-term substituted SMC1^4E-expressing cells (Fig. 4d and Extended Data Fig. 9g), linking defective loop extrusion to deregulated PRIMPOL action. An alternative pathway possibly driving unrestrained fork progression requires the action of RECQ1 helicase^50,51. However, depletion of RECQ1 (Extended Data Fig. 9e) had no effect on the unrestrained fork progression of RAD21-depleted cells treated with HU (Fig. 4b), indicating that reversed forks are unlikely to be formed in the first place in the absence of cohesin.

Next, we asked whether PRIMPOL depletion restores the ability of stalled forks to undergo fork reversal. To answer this question, we reverted to fork degradation assays in the absence of BRCA2. As shown before, RAD21 depletion prevented nascent DNA degradation in BRCA2-deficient cells (Extended Data Fig. 8f), but fork degradation was readily restored by PRIMPOL depletion (Fig. 4e and Extended Data Fig. 9h). Hence, defective loop extrusion impairs fork reversal due to PRIMPOL deregulation, although different from other genetic perturbations that drive PRIMPOL-dependent unrestrained fork progression (RAD51 depletion) in which fork reversal is not recovered after PRIMPOL depletion⁵² (Fig. 4e and Extended Data Fig. 9h).

Finally, we tested whether cohesin loop extrusion is important to maintain cellular fitness after long-term treatments with low doses of RS-inducing agents. For this purpose, we used NIPBL-depleted and SMC1^4E-expressing cells, which can proliferate for long periods. Loss of NIPBL or SMC1^4E substitution had a main impact on the growth of single colonies after genotoxic treatments (Fig. 4f,g). This sensitivity is comparable to that observed after fork reversal inactivation, induced by acute depletion of endogenously tagged Halo-SMARCAL1 after 24-h treatment with Halo-PROTAC3 (HP3), as validated in DNA fibre assays (Extended Data Fig. 9j–n). These results further confirm that cohesin-mediated loop extrusion helps cells cope with RS during prolonged treatments. This sensitivity to genotoxic agents may partly reflect unleashed PRIMPOL action and toxic accumulation of ssDNA gaps. However, cohesin-mediated loop extrusion also plays a part in the repair of DSBs, which are likely to arise during the prolonged treatments used for these experiments⁵³.

Cell culture

HCT116 cells and derivative cell lines (Supplementary Table 1) were grown in DMEM (Life Technologies, 41966-029) supplemented with 10% fetal bovine serum (FBS, GIBCO), 100 U ml⁻¹ penicillin and 100 mg ml⁻¹ streptomycin in an atmosphere containing 6% CO₂ at 37 °C. HCT116 OsTIR (F74G), mAID-mClover-RAD21 HCT116 OsTIR (F74G)⁶, WAPL-mClover-mAID HCT116 OsTIR (F74G)⁴ and CTCF-mClover-mAID HCT116 OsTIR (F74G)⁶ were provided by M. Kanemaki. None of the cell lines were authenticated in house. All cell lines used in this study were subjected to regular mycoplasma testing in house, and all of them consistently reported negative results. To degrade mAID-tagged proteins, cells were treated with 2 µM 5-Ph-IAA (aux; BioAcademia). To degrade Halo-SMARCAL1, cells were treated with Halo-PROTAC3 at 1:1,000 for the indicated times.

Cell line generation

Cell lines were generated following published protocols^56,57. HCT116 OsTIR (F74G) cells were transfected with CRISPR-Cas9 and donor plasmids using FuGENE HD Transfection Reagent (Promega) in a 12-well plate following the instructions of the manufacturer. Two days after transfection, cells were plated in 10 cm² dishes and selected with antibiotics. Selected clones were isolated and confirmed for protein expression of the modified alleles. Oligonucleotides to generate sgRNA or amplify homology arms to build donor plasmids are provided in Supplementary Table 1.

RNA interference

For RNA interference, cells were transfected with the indicated siRNAs (Supplementary Table 1) for the indicated times and concentrations. Transfections were carried out using RNAiMax (Thermo Fisher) according to the instructions of the manufacturer.

Lentiviral transduction

For cell lines expressing SMC1^WT, SMC1^3D or SMC1^4E, SMC1-3×Flag variant genes were cloned into LT3GEPIR (generated by GeneScript) and transfected into to HEK293T cells together with the lentiviral packaging vectors. After 48 h, lentiviral culture medium was harvested, filtered and added to SMC1-mClover-mAID HCT116 OsTIR (F74G) cells. After infection for 24 h, puromycin (2 μg ml⁻¹, InvivoGen) resistant single cells were isolated. To induce the expression of the SMC1 variants, cells were treated with 1 µg ml⁻¹ dox (Sigma-Aldrich).

iPOND–MS

iPOND was performed as described^33,58 with minor modifications. HCT116 cells were labelled with 10 µM EdU for 10 min and treated with the different drugs as indicated in Extended Data Fig. 1a. Cells were crosslinked with 1% formaldehyde for 20 min at room temperature (RT), quenched with 0.125 M glycine for 5 min, and washed three times with cold PBS. EdU was linked to biotin, after permeabilization with 0.25% Triton X-100/PBS for 30 min, by incubating in click reaction buffer (10 mM sodium-l-ascorbate, 20 µM biotin azide (Vanderbilt University) and 2 mM CuSO₄) at RT for 1 h on a rotator. Cells were washed twice with PBS, resuspended in lysis buffer (50 mM Tris-HCl, pH 8.0 and 1% SDS) supplemented with protease inhibitors, and chromatin was solubilized by sonication in a Bioruptor Pico (Diagenode) at 4 °C 10 min (30 s on and 30 s off cycles). After centrifugation for 10 min at 16,000g, supernatants were diluted with 1:1 PBS (vol/vol) containing protease inhibitors and incubated overnight with myOne streptavidin C1 dynabeads (Thermo Fisher). Beads were washed once with lysis buffer, once with 1 M NaCl, twice with lysis buffer and once with PBS. Captured proteins were digested on beads using 500 ng of sequencing grade modified trypsin (Promega, V5111), including reduction and alkylation of cysteines with tris(2-carboxyethyl)phosphine hydrochloride (TCEP) and 2-chloroacetamide (ClAA) addition, respectively. Resulting peptides were cleaned using the Phoenix kit (Preomics) according to instructions in the kit. Liquid chromatography-tandem mass spectrometry analysis of peptide mixture was conducted on an Orbitrap Exploris 480 mass spectrometer (Thermo Fisher) directly coupled to an ACQUITY UPLC M-Class System (Waters) configured for 75-μm scale single-pump trapping. Peptides were separated on a nanoEase HSS C18 T3, 100 A, 75 μm × 250 mm analytical column (Waters, PN: 186008818) at a constant flow rate of 300 nl min⁻¹ applying a peace-wise linear gradient from 5% to 33% solvent B in 45 min (solvent A: water, including 0.1% formic acid; solvent B: acetonitrile, including 0.1% formic acid). MS data acquisition was conducted in data-independent mode. DIA scans were acquired in the Orbitrap mass analyser at 15,000 Resolution (normalized AGC target: 3,000%, maxIT: 23 ms) covering the m/z range from 350 to 1,050 in 70 non-overlapping isolation windows. Precursors were quadrupole isolated at 10m/z and HCD fragmented at an NCE of 28 (calculated for the middle of the isolation window and charge 2).

Data analysis was performed using the PROSTAR package (v.1.34.6) in R (v.4.3.3). Protein intensity values were log₂-transformed, and missing values for partially observed values were imputed using the SLSA (Structured Least Squares Algorithm). Data normalization was carried out first within groups and subsequently globally (excluding no biotin and thymidine chase controls) using the LOESS method. The normalized dataset was further processed in Perseus (v.1.6.10.50), in which remaining missing values were imputed from a normal distribution. For each comparison, proteins were retained only if at least two non-imputed valid values were present in at least one group. The resulting dataset was re-imported into PROSTAR, and differential abundance analysis was performed using the LIMMA test with a log₂ fold-change threshold of 0.5. Benjamini–Hochberg correction was used to calculate P-values and FDR. Volcano plots were generated using GraphPad Prism (v.10).

PLA and SIRF

Cells were asynchronously grown on poly-lysine-coated coverslips. Cells were labelled with 25 µM EdU (Thermo Fisher) for 10 min. After the indicated treatments, cells were washed with PBS and pre-extracted for 5 min using CSK buffer (10 mM HEPES, 50 mM NaCl, 0.3 M sucrose, 3 mM MgCl₂, 1 mM EDTA or 1 mM EGTA and 0.5% Triton X-100) at 4 °C, followed by fixation in 4% formaldehyde at RT for 15 min. Following fixation, cells were washed three times with PBS. In SIRF experiments, EdU was linked to biotin by incubating in click reaction (0.1 M Tris, pH 8.5, 0.1 M sodium-l-ascorbate, 2 mM CuSO₄ and 0.1 mM biotin azide) for 1 h at 37 °C in a humidified chamber. After three washes with PBS, the cells were incubated in blocking buffer (5% BSA in PBS) at 37 °C for 1 h and incubated overnight at 4 °C with the indicated antibodies (Supplementary Table 1). After washing twice for 5 min with TBST (0.05% Tween-20 in TBS), cells were incubated with PLA probes (Merck) for 1 h at 37 °C, ligated for 30 min at 37 °C and subjected to a polymerase reaction for 100 min at 37 °C according to the instructions of the manufacturer. After washing twice for 5 min with TBS, cells were incubated at 37 °C for 30 min with fluorescent-labelled secondary antibodies (Supplementary Table 1), against the individual targets of the PLA, in TBS containing DAPI (0.5 µg ml⁻¹). Following two 5-min washes in TBS and letting coverslips air-dry in the dark, coverslips were mounted with Prolong Gold antifade reagent. Microscopy imaging was performed using a Leica DM6 B microscope (HCX PL APO 63× objective). SIRF/PLA and individual channel quantification was performed using Cell Profiler, and plotted and statistically assessed using GraphPad Prism 10. At least 100 cells were measured per condition.

Biochemical fractionation, co-immunoprecipitation assays and Western blotting

Biochemical fractionation of cells was performed as previously described^33,58. Co-immunoprecipitation experiments were performed by lysing cells in a buffer consisting of 20 mM Tris (pH 7.5), 150 mM NaCl, 5 mM MgCl₂, 2 mM NaF, 10% glycerol, 0.2% NP40, 20 mM β-glycerophosphate, 0.5 mM DTT, protease inhibitor cocktail and Benzonase (Millipore), and rotated at 4 °C for 4 h. Lysates were centrifuged at 16,000g for 30 min at 4 °C. Supernatants were incubated with anti-Flag M2 affinity beads (Merck) overnight at 4 °C. Beads were washed three times with the lysis buffer. Biochemical fractions, co-immunoprecipitation beads and total cell extracts were prepared in Laemmli sample buffer (4% SDS, 20% glycerol and 120 mM Tris-HCl, pH 6.8) and loaded onto 4–20% Mini-PROTEAN TGX Precast Protein Gels (BioRad). For BRCA2 detection, extracts were run in 3–8% NuPAG Tris-Acetate Midi Protein Gels (Thermo Fisher). Proteins from all gels were separated by electrophoresis at 180 V followed by protein transfer to a nitrocellulose membrane in transfer buffer (25 mM Tris and 192 mM glycine) containing 20% methanol. Before addition of primary antibodies, membranes were blocked in 5% milk in 0.1% TBST (1× TBS supplemented with 0.1% Tween-20) for 1 h and incubated in 3% BSA with primary antibodies overnight at 4 °C (Supplementary Table 1). Secondary antibodies were added for 1 h at RT (in blocking solution; Supplementary Table 1). Membranes were washed three times with 0.1% TBST, 10 min each, after primary and secondary antibody incubations and detected with ECL detection reagent (GE Healthcare).

Flow cytometry

Cell lines were labelled with 25 µM EdU for 30 min, harvested by standard trypsinization, washed with PBS once and fixed for 12 min with 4% formaldehyde in PBS. Cells were permeabilized with 0.25% Triton X in PBS, washed twice with PBS, and EdU was labelled by incubating in click reaction (0.1 M Tris pH 8.5, 0.1 M sodium-l-ascorbate, 2 mM CuSO₄ and 0.1 mM Alexa Fluor-linked azide (Thermo Fisher)) for 1 h at RT. Cells were washed twice with PBS and DNA was stained with 1 μg ml⁻¹ DAPI in PBS for 1 h. Samples were run on Attune NxT Flow Cytometer (Thermo Fisher) and analysed using FlowJo software v.10.0.8 (FlowJo). At least 10,000 cells were measured per condition. For gating strategy, see Supplementary Fig. 2.

Metaphase spreads

Asynchronous mAID-mClover-Sororin cells were treated with 2 μM aux and/or nocodazole for a total of 4 h. Mitotic cells were collected by shake-off, incubated in hypotonic buffer (0.03 M sodium citrate), fixed in Carnoy’s solution, dropped on acid-washed slides and left overnight at RT. After staining with 1 μg ml⁻¹ DAPI, slides were mounted with ProLong antifade and left to dry overnight before imaging in a Leica DM6000 microscope with LAS AF software.

DNA fibre assays

All cell lines subjected to this analysis were grown asynchronously and labelled with 30 μM of the thymidine analogue 5-chloro-2′-deoxyuridine (CldU; Sigma-Aldrich) for 30 min, washed three times with warm PBS and exposed to 250 μM of 5-iodo-2′-deoxyuridine (IdU) for 30 min. For fork progression and S1 DNA fibre assessments, IdU was added alone or in combination with the indicated genotoxic treatments and concentrations. For fork degradation assays, IdU was added alone for 30 min, cells were washed three times with PBS and media containing 4 mM HU was added for 3 h. All cells were collected by trypsinization. For S1 assays, cells were incubated with S1 and prepared for spreading as described in ref. ⁴⁹. All cells or nuclei were resuspended in cold PBS at 5 × 10⁵ cells ml⁻¹ or 2 × 10⁶ nuclei ml⁻¹. A total of 2 μl of this cell suspension was mixed with 7 μl of lysis buffer (200 mM Tris-HCl, pH 7.5, 50 mM EDTA and 0.5% SDS) on a glass slide. After incubating for 6 min at RT, the slides were tilted to a 45° angle to stretch the DNA fibres onto the slide. The resulting DNA spreads were air-dried and fixed in 3:1 methanol:acetic acid. The DNA fibres were denatured by incubating them in 2.5 M HCl for 1 h at RT, washed three times with PBS and blocked with 1% BSA in PBS-T (0.05% Tween-20 in PBS) for 1 h at RT. CldU and IdU tracks were stained for 2 h at RT using two different anti-BrdU antibodies recognizing CldU and IdU, respectively (Supplementary Table 1). After washing three times with PBS, slides were stained with secondary antibodies (Supplementary Table 1) for 1 h at RT in the dark. The slides were mounted in 30 μl Prolong Gold antifade reagent (Invitrogen). Microscopy was done using a Leica DM6 B microscope (HCX PL APO 63× objective). At least 100 fibres per sample were measured using ImageJ. Values were plotted and statistically assessed using GraphPad Prism 10.

Electron microscopy

Cells were asynchronously grown, treated with 1 µg ml⁻¹ dox for 48 h, 2 µM aux for 5 h and 50 µM HU where indicated. Cells were collected, resuspended in ice-cold PBS and crosslinked with 10 μg ml⁻¹ 4,5′,8-trimethylpsoralen and pulses of 365 nm ultraviolet (UV) irradiation with monochromatic light (UV Stratalinker 1800; Agilent Technologies). DNA was extracted as previously described⁴⁸. Briefly, cells were lysed (1.28 M sucrose, 40 mM Tris-HCl [pH 7.5], 20 mM MgCl₂, and 4% Triton X-100; Qiagen) and digested (800 mM guanidine-HCl, 30 mM Tris-HCl pH 8.0, 30 mM EDTA pH 8.0, 5% Tween-20 and 0.5% Triton X-100) at 50 °C for 2 h in presence of 1 mg ml⁻¹ proteinase K. The DNA was purified using chloroform:isoamylalcohol (24:1) and precipitated in one volume of isopropanol. DNA was washed with 70% EtOH and resuspended in 200 μl TE (Tris-EDTA) buffer. A total of 6 μg of was incubated with 120 U of PvuII HF (New England Biolabs) for 5 h at 37 °C. RNase A (Sigma-Aldrich, R5503) was added to a final concentration of 250 µg ml⁻¹ for the last 2 h of this incubation. Digested DNA was purified using a Silica Bead Gel Extraction kit (Thermo Fisher Scientific) according to the instructions of the manufacturer. DNA was spread on carbon-coated 400-mesh nickel grids (G2400N, Plano, using benzyl-dimethyl-alkyl-ammonium chloride). DNA was then coated with platinum using a High Vacuum Evaporator (EM BAF060, Leica) as previously described⁴⁸. The grids were imaged automatically at 28,000× using a Talos 120 transmission electron microscope (FEI; LaB6 filament, high tension ≤120 kV) with a bottom-mounted CMOS camera BM-Ceta (4,096 × 4,096 pixels) and the MAPS 3 software (Thermo Fisher Scientific). For analysis, the samples were annotated for replication intermediates using the MAPS offline viewer (v.3.28, Thermo Fisher Scientific), and the corresponding images were extracted. The replication intermediates were scored blind to the experimental condition using Fiji⁵⁹. For each experimental condition, at least 65 replication fork molecules were analysed in two distinct biological replicates. The values were plotted and statistically assessed using GraphPad Prism 10.

Proliferation and clonogenic assay

Cells were treated with 1 µg ml⁻¹ of dox for 48 h (for WT and 4E-SMC1 substituted cells) and 2 µM aux (1 day before treatment in NIPBL-cells and 3 days before treatment in SMC1-substituted cells). In each well of a 6-well plate, 1,000 cells were seeded. The cells were then optionally treated with 90 nM MMC (Sigma-Aldrich) for 24 h, washed, and incubated in normal media, or treated with 150 µM HU (Sigma-Aldrich) or 50 nM Aph (Sigma-Aldrich) for 9 days to allow formation of colonies. Subsequently, colonies were stained with 0.5% (wt/vol) crystal violet in 20% ethanol. Plates were imaged using a plate reader, and the number of colonies was quantified using ImageJ. Values were normalized to the UNT control, and then plotted and statistically assessed using GraphPad Prism 10.

Micro-C experiments

The Micro-C library was prepared using the Dovetail Micro-C Kit according to the protocol of the manufacturer. Briefly, cells were treated with 1 µg ml⁻¹ dox for 48 h and 2 µM aux for 5 h where indicated. The chromatin was fixed with disuccinimidyl glutarate (DSG) and formaldehyde in the nucleus. The crosslinked chromatin was then digested in situ with micrococcal nuclease (MNase). After digestion, the cells were lysed with SDS to extract chromatin fragments, which were then bound to Chromatin Capture Beads. Next, the chromatin ends were repaired and ligated to a biotinylated bridge adapter, followed by proximity ligation of adapter-containing ends. After proximity ligation, the crosslinks were reversed, the associated proteins were degraded, and the DNA was purified and then converted into a sequencing library using Illumina-compatible adaptors. Biotin-containing fragments were isolated using streptavidin beads before PCR amplification.

Repli-C experiments

Cells were treated with 1 µg ml⁻¹ dox for 48 h, 2 µM aux for 5 h where indicated. A total of 10 µM EdU was added for 15 min, either before collection or before washing and treating cells with 2 mM HU for 2 h. Repli-C was performed using the Dovetail HiChIP MNase Kit with modifications. The cells were collected by trypsinization, the chromatin was fixed with DSG and formaldehyde and permeabilized with 0.25% Triton X in PBS. EdU was clicked to digoxigenin by incubating in a click reaction (0.1 M Tris, pH 8.5, 0.1 M sodium-l-ascorbate, 2 mM CuSO₄ and 0.3 mM digoxigenin-azide (ATTBio)). Then, crosslinked chromatin was digested in situ with MNase and extracted after cell lysis using a combination of RIPA lysis and gentle sonication using Bioruptor Pico in RIPA buffer. The chromatin fragments were incubated with anti-digoxigenin antibody pre-coupled to A/G-coated beads, overnight rotating at 4 °C. Next, the chromatin ends were repaired and ligated to a biotinylated bridge adapter followed by proximity ligation of adapter-containing ends. After proximity ligation, the crosslinks were reversed, the associated proteins were degraded, and the DNA was purified and converted into a sequencing library using Illumina-compatible adaptors. Biotin-containing fragments were isolated using streptavidin beads before PCR amplification.

Micro-C and Repli-C analyses

Micro-C and Repli-C data were processed using a standardized Hi-C analysis pipeline. Raw paired-end sequencing reads were quality-trimmed using Trim Galore (quality threshold = 20, minimum length = 30 bp) and aligned to the hg38 human reference genome using BWA-MEM⁶⁰) with Hi-C-specific parameters (−5SP −T0). Aligned reads were processed with the pairtools suite: contacts were parsed with a minimum mapping quality of 40 (–min-mapq 40), allowing walks of up to five unique alignments (–walks-policy 5unique) with a maximum inter-alignment gap of 30 bp. Parsed pairs were sorted, and PCR duplicates were removed while retaining duplicate marking statistics. Valid contact pairs were converted to ‘.hic’ format using Juicer Tools (v.1.22.01) and to multi-resolution cool format (.mcool) using HiCExplorer with Knight–Ruiz (KR) matrix balancing normalization. For comparative analyses between conditions, biological replicates were merged using pairtools merge, and datasets were downsampled to equalize sequencing depth across conditions using pairtools sample. Downsampling fractions were calculated to match the condition with the lowest sequencing depth, ensuring unbiased comparisons (Supplementary Tables 2–6). Contact probability as a function of genomic distance was calculated as an intra-chromosomal contact frequency distribution, using logarithmically increasing genomic distance bins. The insulation score was computed using a custom Python script following the methodology described in ref. ⁶¹, extracting 10-kb resolution raw matrices and using a sliding window of 100 kb × 100 kb. Chromatin loops were detected using Chromosight⁶² on balanced contact matrices at 25 kb resolution, with a scanning distance range of 50 kb to 1 Mb. Moreover, architectural stripes (left and right patterns) were identified at 25 kb resolution.

Replication fountains, characterized by enriched long-range chromatin interactions radiating from replication origins, were identified using the FUN (Fountains Using Neighborhoods) algorithm⁵ on balanced contact matrices at 10 kb and 25 kb resolutions. Signal-over-noise (SoN) scores were calculated using a sliding window approach with an extension length of 500 kb, offset of 50 kb and padding widths of three bins (10 kb) or 1 bin (25 kb). Fountain structures were detected by extending outwards from SoN summits (extension pixels: 6–100 bins, step size 2 at 10 kb; 3–35 bins, step size 1 at 25 kb) and evaluated for statistical significance (P-value < 0.05, signal-to-background ratio >1.3).

Metaplots were created using coolpuppy⁶³ on Knight–Ruiz balanced matrices. In the inter-replication fork metaplots, the –scale option was added to scale all the regions to the same size. Intra-replicon was quantified from metaplot diagonal decay profiles by calculating the log₂ fold-change ratio between short-range (≤50 kb) and mid-range (50–100 kb) normalized contact frequencies. For each condition, diagonal decay was computed by extracting contact values at increasing distances from the diagonal using a sliding window approach (three bins each side of the centre point), with background normalization calculated from an extended window (10 bins). Values were log₂-transformed and smoothed using a three-point moving average kernel. The intra-replicon contact index was defined as log₂(mean contacts ≤50 kb/mean contacts 50–100 kb), in which higher values indicate increased short-range contacts. Values were plotted and statistically assessed using GraphPad Prism 10. Inter-replication fork interaction strength was quantified by extracting the sum of log₂ contact frequencies from a 6 × 6 pixel region centred in the origin–origin interaction, representing the enrichment of contacts between paired replication origins.

Identification of replication origins

Replication fork directionality (RFD) profiles were obtained from published Pu-seq data⁴³, which quantifies the differential usage of leading and lagging strand DNA polymerases during replication. To identify replication origins, we analysed the RFD signal across the genome at 10-kb resolution. For each chromosome (excluding chrM and chrY), binned mean values were extracted and smoothed using a moving average filter (window size = 5) followed by Savitzky–Golay filtering (window = 15, polynomial order = 3) to preserve signal features while reducing noise. Replication origins and termination zones were identified by detecting inflection points in the smoothed RFD profile through second-derivative analysis. The second derivative captures the curvature of the signal, identifying the precise genomic positions at which replication fork dynamics fundamentally change. Replication origins were called as downward inflection points (second-derivative changes from positive to negative), corresponding to bidirectional fork movement away from the origin. For each putative origin, replication strength was calculated as the difference between the maximum RFD value in the downstream region (between the origin and the next termination zone) and the minimum RFD value in the upstream region (between the previous termination zone and the origin). Origins with replication strength below a threshold of 0.05 were considered weak and excluded from further analysis. When weak origins were removed, adjacent termination zones were merged by averaging their genomic positions. The remaining strong origins were reassigned to their flanking termination zones to ensure consistency between origin and termination zone annotations. Origins lacking flanking termination zones on either side were discarded^64,65,66,67.

Reporting summary

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

1. Berti, M., Cortez, D. & Lopes, M. The plasticity of DNA replication forks in response to clinically relevant genotoxic stress. Nat. Rev. Mol. Cell Biol. 21, 633–651 (2020).

   Article  CAS  PubMed  Google Scholar 

2. Davidson, I. F. & Peters, J.-M. Genome folding through loop extrusion by SMC complexes. Nat. Rev. Mol. Cell Biol. 22, 445–464 (2021).

   Article  CAS  PubMed  Google Scholar 

3. Guillou, E. et al. Cohesin organizes chromatin loops at DNA replication factories. Genes Dev. 24, 2812–2822 (2010).

   Article  CAS  PubMed  PubMed Central  Google Scholar 

4. Emerson, D. J. et al. Cohesin-mediated loop anchors confine the locations of human replication origins. Nature 606, 812–819 (2022).

   Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

5. Liu, Y. et al. Fork coupling directs DNA replication elongation and termination. Science 383, 1215–1222 (2024).

   Article  ADS  CAS  PubMed  Google Scholar 

6. Yesbolatova, A. et al. The auxin-inducible degron 2 technology provides sharp degradation control in yeast, mammalian cells, and mice. Nat. Commun. 11, 5701 (2020).

   Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

7. Bauer, B. W. et al. Cohesin mediates DNA loop extrusion by a “swing and clamp” mechanism. Cell 184, 5448–5464.e22 (2021).

   Article  CAS  PubMed  PubMed Central  Google Scholar 

8. Nagasaka, K. et al. Cohesin mediates DNA loop extrusion and sister chromatid cohesion by distinct mechanisms. Mol. Cell 83, 3049–3063.e6 (2023).

   Article  CAS  PubMed  Google Scholar 

9. Guérin, T. M., Barrington, C., Pobegalov, G., Molodtsov, M. I. & Uhlmann, F. An extrinsic motor directs chromatin loop formation by cohesin. EMBO J. 43, 3 (2024).

   Article  PubMed  PubMed Central  Google Scholar 

10. De Koninck, M. & Losada, A. Cohesin mutations in cancer. Cold Spring Harb. Perspect. Med. 6, a026476 (2016).

    Article  PubMed  PubMed Central  Google Scholar 

11. González-Acosta, D. & Lopes, M. DNA replication and replication stress response in the context of nuclear architecture. Chromosoma 133, 57–75 (2024).

    Article  PubMed  Google Scholar 

12. Vouzas, A. E. & Gilbert, D. M. Mammalian DNA replication timing. Cold Spring Harb. Perspect. Biol. 13, a040162 (2021).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

13. Wang, W. et al. Genome-wide mapping of human DNA replication by optical replication mapping supports a stochastic model of eukaryotic replication. Mol. Cell 81, 2975–2988.e6 (2021).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

14. Carrington, J. T. et al. Most human DNA replication initiation is dispersed throughout the genome with only a minority within previously identified initiation zones. Genome Biol. 26, 122 (2025).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

15. Zhao, P. A., Sasaki, T. & Gilbert, D. M. High-resolution Repli-Seq defines the temporal choreography of initiation, elongation and termination of replication in mammalian cells. Genome Biol. 21, 76 (2020).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

16. Cuadrado, A. & Losada, A. Specialized functions of cohesins STAG1 and STAG2 in 3D genome architecture. Curr. Opin. Genet. Dev. 61, 9–16 (2020).

    Article  CAS  PubMed  Google Scholar 

17. Davidson, I. F. et al. DNA loop extrusion by human cohesin. Science 366, 1338–1345 (2019).

    Article  ADS  CAS  PubMed  Google Scholar 

18. Kim, Y., Shi, Z., Zhang, H., Finkelstein, I. J. & Yu, H. Human cohesin compacts DNA by loop extrusion. Science 366, 1345–1349 (2019).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

19. Alonso-Gil, D. & Losada, A. NIPBL and cohesin: new take on a classic tale. Trends Cell Biol. 33, 860–871 https://doi.org/10.1016/j.tcb.2023.03.006 (2023).

    Article  CAS  PubMed  Google Scholar 

20. Dequeker, B. J. H. et al. MCM complexes are barriers that restrict cohesin-mediated loop extrusion. Nature 606, 197–203 (2022).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

21. Li, Y. et al. The structural basis for cohesin–CTCF-anchored loops. Nature 578, 472–476 (2020).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

22. Minamino, M., Diffley, J. F. X. & Uhlmann, F. Biochemical reconstitution of sister chromatid cohesion establishment during DNA replication. Mol. Cell 85, 3576–3587.e6 (2025).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

23. Cameron, G. et al. Sister chromatid cohesion establishment during DNA replication termination. Science 384, 119–124 (2024).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

24. Nishiyama, T. et al. Sororin mediates sister chromatid cohesion by antagonizing Wapl. Cell 143, 737–749 (2010).

    Article  CAS  PubMed  Google Scholar 

25. Frattini, C. et al. Cohesin ubiquitylation and mobilization facilitate stalled replication fork dynamics. Mol. Cell 68, 758–772.e4 (2017).

    Article  CAS  PubMed  Google Scholar 

26. Tittel-Elmer, M. et al. Cohesin association to replication sites depends on Rad50 and promotes fork restart. Mol. Cell 48, 98–108 (2012).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

27. Delamarre, A. et al. MRX increases chromatin accessibility at stalled replication forks to promote nascent DNA resection and cohesin loading. Mol. Cell 77, 395–410.e3 (2020).

    Article  CAS  PubMed  Google Scholar 

28. Fumasoni, M., Zwicky, K., Vanoli, F., Lopes, M. & Branzei, D. Error-free DNA damage tolerance and sister chromatid proximity during DNA replication rely on the Polα/Primase/Ctf4 complex. Mol. Cell 57, 812–823 (2015).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

29. Mourón, S. et al. Repriming of DNA synthesis at stalled replication forks by human PrimPol. Nat. Struct. Mol. Biol. 20, 1383–1389 (2013).

    Article  PubMed  Google Scholar 

30. Dungrawala, H. et al. The replication checkpoint prevents two types of fork collapse without regulating replisome stability. Mol. Cell 59, 998–1010 (2015).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

31. Morales, C. et al. PDS5 proteins are required for proper cohesin dynamics and participate in replication fork protection. J. Biol. Chem. 295, 146–157 (2020).

    Article  CAS  PubMed  Google Scholar 

32. Carvajal-Maldonado, D. et al. Perturbing cohesin dynamics drives MRE11 nuclease-dependent replication fork slowing. Nucleic Acids Res. 47, 1294–1310 (2019).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

33. Sirbu, B. M., Couch, F. B. & Cortez, D. Monitoring the spatiotemporal dynamics of proteins at replication forks and in assembled chromatin using isolation of proteins on nascent DNA. Nat. Protoc. 7, 594–605 (2012).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

34. Guarino, E., Salguero, I. & Kearsey, S. E. Cellular regulation of ribonucleotide reductase in eukaryotes. Semin. Cell Dev. Biol. 30, 97–103 (2014).

    Article  CAS  PubMed  Google Scholar 

35. Wilhelm, T. et al. Spontaneous slow replication fork progression elicits mitosis alterations in homologous recombination-deficient mammalian cells. Proc. Natl Acad. Sci. USA 111, 763–768 (2014).

    Article  ADS  CAS  PubMed  Google Scholar 

36. Somyajit, K. et al. Redox-sensitive alteration of replisome architecture safeguards genome integrity. Science 358, 797–802 (2017).

    Article  ADS  CAS  PubMed  Google Scholar 

37. Roy, S., Luzwick, J. W. & Schlacher, K. SIRF: quantitative in situ analysis of protein interactions at DNA replication forks. J. Cell Biol. 217, 1521–1536 (2018).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

38. Petermann, E., Orta, M. L., Issaeva, N., Schultz, N. & Helleday, T. Hydroxyurea-stalled replication forks become progressively inactivated and require two different RAD51-mediated pathways for restart and repair. Mol. Cell 37, 492–502 (2010).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

39. Schwarzer, W. et al. Two independent modes of chromatin organization revealed by cohesin removal. Nature 551, 51–56 (2017).

    Article  ADS  PubMed  PubMed Central  Google Scholar 

40. Alonso-Gil, D., Cuadrado, A., Giménez-Llorente, D., Rodríguez-Corsino, M. & Losada, A. Different NIPBL requirements of cohesin-STAG1 and cohesin-STAG2. Nat. Commun. 14, 1326 (2023).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

41. Shah, R. et al. Cohesin cofactor dosage sets the rate of loop extrusion, rendering genome folding tunable yet vulnerable to genetic disruption. Mol. Cell 86, 1635–1652.e8 https://doi.org/10.1016/j.molcel.2026.03.037 (2026).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

42. Nora, E. P. et al. Targeted degradation of CTCF decouples local insulation of chromosome domains from genomic compartmentalization. Cell 169, 930–944.e22 (2017).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

43. Koyanagi, E. et al. Global landscape of replicative DNA polymerase usage in the human genome. Nat. Commun. 13, 7221 (2022).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

44. D’Asaro, D. et al. Genome-wide modeling of DNA replication in space and time confirms the emergence of replication specific patterns in vivo in eukaryotes. Genome Biol. 26, 431 (2025).

    Article  PubMed  PubMed Central  Google Scholar 

45. Jeppsson, K. et al. Cohesin-dependent chromosome loop extrusion is limited by transcription and stalled replication forks. Sci. Adv. 8, eabn7063 (2022).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

46. Davidson, I. F. et al. Cohesin supercoils DNA during loop extrusion. Cell Rep. 44, 115856 (2025).

    Article  CAS  PubMed  Google Scholar 

47. Bai, G. et al. HLTF promotes fork reversal, limiting replication stress resistance and preventing multiple mechanisms of unrestrained DNA synthesis. Mol. Cell 78, 1237–1251.e7 (2020).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

48. Zellweger, R. & Lopes, M. Dynamic architecture of eukaryotic DNA replication forks in vivo, visualized by electron microscopy. Methods Mol. Biol. 1672, 261–294 (2018).

    Article  CAS  PubMed  Google Scholar 

49. Quinet, A. et al. PRIMPOL-mediated adaptive response suppresses replication fork reversal in BRCA-deficient cells. Mol. Cell 77, 461–474.e9 (2020).

    Article  CAS  PubMed  Google Scholar 

50. Berti, M. et al. Human RECQ1 promotes restart of replication forks reversed by DNA topoisomerase I inhibition. Nat. Struct. Mol. Biol. 20, 347–354 (2013).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

51. Cherdyntseva, V. et al. Nucleoplasmic Lamin A/C controls replication fork restart upon stress by modulating local H3K9me3 and ADP-ribosylation levels. Nat. Commun. 16, 11239 (2025).

    Article  ADS  PubMed  PubMed Central  Google Scholar 

52. Liu, W. et al. RAD51 bypasses the CMG helicase to promote replication fork reversal. Science 380, 382–387 (2023).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

53. Marin-Gonzalez, A. & Ha, T. Folding a broken genome: the versatile roles of cohesin in genome maintenance. Nat. Rev. Genet. https://doi.org/10.1038/s41576-026-00988-6 (2026).

    Article  PubMed  Google Scholar 

54. Martínez-Jiménez, M. I., Lahera, A. & Blanco, L. Human PrimPol activity is enhanced by RPA. Sci. Rep. 7, 783 (2017).

    Article  ADS  PubMed  PubMed Central  Google Scholar 

55. Gaggioli, V. et al. Replication-stress-induced chromatin loops protect fork stability. Nature 656, 493–505 (2026).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

56. Yesbolatova, A., Natsume, T., Hayashi, K. & Kanemaki, M. T. Generation of conditional auxin-inducible degron (AID) cells and tight control of degron-fused proteins using the degradation inhibitor auxinole. Methods 164–165, 73–80 (2019).

    Article  PubMed  Google Scholar 

57. Ran, F. A. et al. Genome engineering using the CRISPR-Cas9 system. Nat. Protoc. 8, 2281–2308 (2013).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

58. Sirbu, B. M. et al. Analysis of protein dynamics at active, stalled, and collapsed replication forks. Gene Dev. 25, 1320–1327 (2011).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

59. Schindelin, J. et al. Fiji: an open-source platform for biological-image analysis. Nat. Methods 9, 676–682 (2012).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

60. Li, H. Aligning sequence reads, clone sequences and assembly contigs with BWA-MEM. Preprint at arxiv.org/abs/1303.3997 (2013).

61. Crane, E. et al. Condensin-driven remodelling of X chromosome topology during dosage compensation. Nature 523, 240–244 (2015).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

62. Matthey-Doret, C. et al. Computer vision for pattern detection in chromosome contact maps. Nat. Commun. 11, 5795 (2020).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

63. Flyamer, I. M., Illingworth, R. S. & Bickmore, W. A. Coolpup.py: versatile pile-up analysis of Hi–C data. Bioinformatics 36, 2980–2985 (2020).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

64. Remeseiro, S. et al. Cohesin-SA1 deficiency drives aneuploidy and tumourigenesis in mice due to impaired replication of telomeres. EMBO J. 31, 2076–2089 (2012).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

65. González-Acosta, D. et al. PrimPol-mediated repriming facilitates replication traverse of DNA interstrand crosslinks. EMBO J. 40, e106355 (2021).

    Article  PubMed  PubMed Central  Google Scholar 

66. Rao, S. S. P. et al. Cohesin loss eliminates all loop domains. Cell 171, 305–320.e24 (2017).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

67. Porter, H. et al. Cohesin-independent STAG proteins interact with RNA and R-loops and promote complex loading. eLife 12, e79386 (2023).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

Download references

Author notes

1. These authors contributed equally: Daniel González-Acosta, Daniel Giménez-Llorente

Authors and Affiliations

1. Institute of Molecular Cancer Research (IMCR), University of Zurich, Zurich, Switzerland

   Daniel González-Acosta, Melani Rodrigues, Moses Aouami & Massimo Lopes

2. Chromosome Dynamics Group, Spanish National Cancer Research Centre (CNIO), Madrid, Spain

   Daniel Giménez-Llorente, Ana Cuadrado & Ana Losada

3. Cell Signaling and Clinical Proteomics Group, Biobizkaia Health Research Institute, Barakaldo, Spain

   Orhi Barroso-Gomila & Javier Muñoz

4. Ikerbasque, the Basque Foundation for Science, Bilbao, Spain

   Javier Muñoz

Contributions

M.L., D.G.-A. and A.L. conceptualized and designed the study. D.G.-A. designed and performed most of the experiments in the manuscript. D.G.-L. analysed all Repli-C and genomic data. M.R. performed and analysed all SIRF, DNA fibre and clonogenic survival experiments following NIPBL depletion, and assisted with several other experiments. M.A. processed, annotated and scored EM experiments, assisted by D.G.-A. O.B.-G. and J.M. analysed iPOND–MS data. A.C. performed and analysed metaphase spreads following Sororin depletion. D.G.-A., A.L. and M.L. acquired funding. A.C., A.L. and M.L. supervised the study. D.G.-A. and M.L. wrote the original manuscript, which was reviewed and edited by D.G.-L., M.R., A.C. and A.L.

Corresponding authors

Correspondence to Daniel González-Acosta, Ana Losada or Massimo Lopes.

Competing interests

The authors declare no competing interests.

Peer review information

Nature thanks Philippe Pasero and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Peer reviewer reports are available.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions