Sororin locks the DNA-exit gate of cohesin to preserve sister-chromatid cohesion

Abstract

The cohesin complex mediates sister-chromatid cohesion by topologically entrapping DNA within an SMC1-SMC3-RAD21 ring, yet how Sororin preserves cohesion beyond its known role of antagonizing Pds5 binding to the release factor Wapl has remained unclear. Here, we show that the extreme C-terminal region (CTR) of Sororin functions as a direct structural lock for cohesin’s DNA-exit gate by engaging the RAD21-SMC3 interface. Centromere-tethered Sororin-CTR fully restores cohesion after Sororin depletion, whereas constitutive chromatin tethering prevents cohesin removal, recapitulating Wapl-loss phenotypes, including impaired mitotic chromosome condensation, decatenation and segregation. Through biochemical reconstitution, AlphaFold3-guided modeling, and targeted mutagenesis, we define conserved hydrophobic and electrostatic contacts between Sororin-CTR and the RAD21-SMC3 gate, the disruption of which abolishes cohesion in a Wapl-dependent manner. Furthermore, mitotic phosphorylation of Sororin selectively disrupts Pds5 binding while leaving gate engagement intact, providing a regulated molecular switch for cohesin release. Together, these findings redefine Sororin as a dual-function regulator that both antagonizes Wapl-Pds5 and directly locks the RAD21-SMC3 exit gate to stabilize sister-chromatid cohesion while permitting its timely dissolution.

Chen et al. identify that the extreme C-terminal region of Sororin directly locks the cohesin DNA-exit gate to stabilize sister-chromatid cohesion, a mechanism distinct from its established role in antagonizing Wapl binding to Pds5.

Introduction

Accurate chromosome segregation is fundamental to genomic stability and cellular viability¹. This process relies on the cohesin complex, a conserved ring-shaped molecular machine that topologically entraps DNA to perform two essential functions: mediating sister-chromatid cohesion in trans and organizing chromatin architecture through DNA loop extrusion in cis². Cohesive cohesin maintains the physical connection between sister chromatids from S‑phase until anaphase to ensure their faithful separation^3–6, whereas extrusive cohesin shapes the three‑dimensional genome organization^7–12. Given these central roles, dysregulation of cohesin contributes to developmental disorders and tumorigenesis, underscoring the need for a detailed mechanistic understanding of cohesin regulation^13–17.

Structurally, the cohesin ring is formed by two SMC proteins, SMC1 and SMC3, bridged by the kleisin subunit RAD21 (also called Scc1/Mcd1); RAD21 further associates with Pds5 and SA1/SA2 to recruit regulatory factors (Fig. 1A)¹⁸. SMC1 and SMC3 dimerize at their hinge domains, while their ATPase head domains are connected by RAD21 at the opposite ends¹⁹. DNA is thought to enter the ring through the SMC1‑SMC3 hinge and exit via a transiently opened interface between RAD21 and SMC3^20–23. This DNA-exit gate is regulated by the release factor Wapl, which, together with Pds5, promotes gate opening and cohesin turnover^24–34.

Fig. 1. The extreme C‑terminus of Sororin is necessary and sufficient to promote cohesion.

A Schematic of cohesin-mediated sister-chromatid cohesion, illustrating Sororin’s interaction with Pds5 to antagonize Wapl. B Schematic representation of human Sororin variants, highlighting the CTR. C HeLa cells were transfected with an empty vector, H2B-GFP, or the indicated Sororin-GFP variants. Twenty-four hours post-transfection, cell lysates were immunoblotted for GFP and α-Tubulin (n = 1). α-Tubulin from a parallel gel served as a loading control. Irrelevant lanes were excised from the gel image (indicated by a vertical line). Uncropped blots are available in the Source Data file. D Cells were transfected as in C. Twenty-four hours post-transfection, cells were treated with nocodazole for 3 h. Mitotic chromosome spreads were prepared and stained with DAPI and antibodies against GFP and ACA. The percentage of mitotic cells exhibiting predominantly arm-resolved or unresolved sister chromatids was quantified in at least 200 cells per condition across four independent experiments (n = 4), except for parental cells and H2B-GFP-expressing cells, for which at least 50 cells per condition were analyzed in one experiment. Statistical significance was calculated with two-tailed, unpaired Student’s t-test. Error bars represent standard deviation (SD). Raw data are available in the Source Data file. E Cells were treated and analyzed as in D. Representative images are shown (n = 1). Scale bar = 10 μm. F HeLa cells were transfected with H2B-GFP or the indicated H2B-Sororin-GFP variants. Twenty-four hours post-transfection, cell lysates were immunoblotted for GFP and GAPDH. GAPDH from a parallel gel served as a loading control. Blots are representative of three independent experiments. Uncropped blots are available in the Source Data file. G Cells were transfected as in F. Twenty-four hours post-transfection, cells were treated with nocodazole for 3 h. Mitotic chromosome spreads were stained with DAPI and antibodies against CENP-C. The percentage of mitotic cells with predominantly arm-resolved or unresolved sister chromatids was quantified from at least 300 cells per condition across three independent experiments (n = 3). Statistical significance was calculated with two-tailed, unpaired Student’s t-test. Error bars represent SD. Raw data are available in the Source Data file. H Cells were treated and analyzed as in G. Representative images from three independent experiments are shown. Scale bar = 10 μm.

During S‑phase, acetylation of SMC3 by Esco1 and Esco2 promotes recruitment of Sororin, which stabilizes cohesion primarily by competing with Wapl for binding to Pds5^{26,27,35–45}. In early mitosis, phosphorylation of Sororin disrupts its association with Pds5, allowing Wapl to rebind and trigger cohesin release from chromosome arms^{24–27,46–50}. In contrast, centromeric cohesin is protected by Sgo1, which associates with PP2A to counteract Sororin phosphorylation, until Separase-mediated cleavage of RAD21 initiates anaphase^51–56.

Despite its essential role in cohesion maintenance^{27,36,44,57,58}, Sororin’s mechanistic contribution beyond Wapl displacement remains unresolved^{26,27,37,49,54,59–63}. A central unanswered question is whether Sororin stabilizes cohesion solely through indirect competition with Wapl for Pds5 binding or whether it also directly engages cohesin to reinforce closure of the DNA-exit gate^58,64. Although the C-terminal region of Sororin has been implicated in cohesion and proposed to interact with the SA2 subunit^65,66, the molecular basis and functional relevance of this interaction remain unclear.

Here, we show that the extreme C-terminal region (CTR) of Sororin directly binds the RAD21-SMC3 interface, functioning as a molecular lock for cohesin’s DNA-exit gate. Through integrated cell-biological, biochemical, and structural analyses, we demonstrate that this gate-locking activity is both necessary and sufficient for cohesion maintenance, operates independently of Pds5 competition, and is tightly regulated during the cell cycle. Our findings redefine Sororin as a dual-function regulator that both antagonizes Wapl binding to Pds5 and directly stabilizes cohesin ring closure to ensure faithful chromosome segregation.

Results

The extreme C‑terminus of Sororin is necessary and sufficient to promote cohesion

To define the minimal Sororin domain required for sister-chromatid cohesion, we generated GFP‑tagged Sororin truncation mutants (Fig. 1B), and assessed their ability to promote cohesion in mitotic HeLa cells arrested with the spindle destabilizer nocodazole (Fig. 1C–E). Chromosome spreads from control cells displayed the characteristic X‑shaped morphology, reflecting normal chromosome-arm resolution. As expected, expression of full‑length Sororin‑GFP, but not H2B‑GFP, markedly impaired arm resolution, consistent with Sororin’s established role. Strikingly, the C‑terminal fragments Sororin (150-252)‑GFP and Sororin (223–252)‑GFP also strongly inhibited arm resolution, whereas the N‑terminal fragments Sororin (1-149)‑GFP and Sororin (1-222)‑GFP did not.

Consistent with previous reports^27,44,67, all Sororin‑GFP variants localized to interphase nuclei but were largely excluded from mitotic chromosomes (Fig. 1E and Supplementary Fig. 1). To exclude the possibility that the inactivity of the N-terminal fragments resulted from defective chromatin association, we forcibly tethered them to chromatin by fusion to histone H2B (Fig. 1F–H). While H2B‑Sororin‑GFP potently blocked arm resolution, H2B‑Sororin (1-149)‑GFP and H2B‑Sororin (1-222)‑GFP remained ineffective.

Together, these results demonstrate that the extreme C‑terminal region (residues 223–252), hereafter termed the CTR, is both necessary and sufficient for Sororin to promote sister-chromatid cohesion.

Centromere-tethered Sororin-CTR bypasses the requirement for Sororin but not Esco1/2 in cohesion protection

Given that Sororin accumulates at mitotic centromeres^27,54,59,62, we asked whether targeted localization of the CTR to centromeres could functionally replace full-length Sororin. To this end, we generated HeLa cell lines stably expressing either Sororin-CTR-GFP fused to the centromere-targeting domain of CENP-B (CB-Sororin-CTR-GFP) or a CB-GFP control (Fig. 2A), then depleted endogenous Sororin by siRNA and examined its effect on mitotic index (Fig. 2B–D). As expected, Sororin knockdown in parental HeLa cells induced profound mitotic arrest, increasing the mitotic index from 4.5% to 39.8%, consistent with spindle-assembly-checkpoint activation following cohesion loss⁴⁴. Comparable mitotic arrest was observed upon Sororin depletion in CB-GFP-expressing cells. In contrast, Sororin depletion had little effect on the mitotic index of CB-Sororin-CTR-GFP cells, indicating that centromeric Sororin-CTR effectively rescued this defect. Consistent with this, chromosome-spread analysis revealed severe cohesion loss in ~85% of Sororin-depleted parental or CB-GFP cells, compared with only 5.9% of CB-Sororin-CTR-GFP cells (Fig. 2E, F). Thus, centromeric tethering of Sororin-CTR is sufficient to maintain sister-chromatid cohesion in the absence of full-length Sororin.

Fig. 2. Centromere-tethered Sororin-CTR bypasses the requirement for Sororin but not Esco1/2 in cohesion protection.

A Parental HeLa cells and stable cell lines expressing CB-GFP or CB-Sororin-CTR-GFP were analyzed by immunoblotting for GFP and α-Tubulin. α-Tubulin from a parallel gel served as a loading control. Blots are representative of two independent experiments. Uncropped blots are available in the Source Data file. B The indicated cells were transfected with control or Sororin siRNA. Forty-eight hours post-transfection, cell lysates were immunoblotted for Sororin and α-Tubulin. α-Tubulin from a parallel gel served as a loading control. Blots are representative of two independent experiments. S., short exposure; L., long exposure. Uncropped blots are available in the Source Data file. C Cells were transfected as in B. Forty-eight hours post-transfection, the mitotic index was determined from at least 4000 cells per condition across four independent experiments (n = 4), except for CB-GFP-expressing cells, for which at least 1000 cells per condition were analyzed in one experiment. Statistical significance was calculated with two-tailed, unpaired Student’s t-test. Error bars represent SD. Raw data are available in the Source Data file. D Cells were treated and analyzed as in C. Representative phase-contrast images are shown (n = 1); round cells were scored as mitotic. Scale bar = 100 μm. E Cells were transfected as in B, and 48 h later, treated with nocodazole for 3 h. Mitotic chromosome spreads were stained for DAPI and antibodies against GFP and ACA. The percentage of mitotic cells with predominantly separated or unseparated sister chromatids was quantified from at least 400 cells per condition across four independent experiments (n = 4), except for CB-GFP-expressing cells, for which at least 100 cells per condition were analyzed in one experiment. Statistical significance was calculated with two-tailed, unpaired Student’s t-test. Error bars represent SD. Raw data are available in the Source Data file. F Cells were treated and analyzed as in E. Representative images are shown (n = 1). Scale bar = 10 μm. G Cells were transfected with control siRNA or co-transfected with siRNAs targeting Esco1 and Esco2. After 48 h, cells were treated with nocodazole for 3 h, and mitotic chromosome spreads were stained with DAPI and antibodies against GFP. The percentage of mitotic cells with predominantly separated or unseparated sister chromatids was quantified from at least 300 cells per condition across three independent experiments (n = 3). Statistical significance was calculated with two-tailed, unpaired Student’s t-test. Error bars represent SD. Raw data are available in the Source Data file. H Cells were treated and analyzed as in G. Representative images from three independent experiments are shown. Scale bar = 10 μm. I Cells were transfected as in G. After 48 h, cell lysates were immunoblotted for Esco1 and GAPDH. GAPDH from a parallel gel served as a loading control. Blots are representative of three independent experiments. Uncropped blots are available in the Source Data file. J Cells were transfected as in G. After 48 h, total RNA was analyzed by quantitative RT-PCR using Esco2-specific primers (n = 3). Statistical significance was calculated with two-tailed, unpaired Student’s t-test. Error bars represent SD. Raw data are available in the Source Data file.

We next examined whether centromeric Sororin-CTR could compensate for the loss of Sgo1, which promotes Sororin-dependent cohesion protection at mitotic centromeres^54,62. Depletion of Sgo1 caused cohesion loss in 84% of CB-GFP cells, a phenotype comparable to Sororin knockdown (Supplementary Fig. 2A–C). This defect was partially but significantly rescued in CB-Sororin-CTR-GFP cells, where only 32.9% showed cohesion loss. This partial rescue suggests that while centromeric Sororin-CTR can mitigate Sgo1 depletion, it cannot fully substitute for Sgo1, which also protects cohesin through PP2A-mediated dephosphorylation⁵⁴.

During S phase, Esco1 and Esco2 acetylate two conserved lysine residues (K105 and K106) in the SMC3 ATPase head, thereby promoting Sororin recruitment^{27,43,68–70}. We therefore asked if these two acetyltransferases remain essential when Sororin-CTR is constitutively localized to centromeres. Co-depletion of Esco1/2 by siRNA induced similar levels of cohesion loss in CB-GFP and CB-Sororin-CTR-GFP cells (~23.3% and ~20.5%, respectively) (Fig. 2G–J). Thus, enforced centromeric localization of Sororin-CTR cannot bypass the requirement for Esco1/2.

In summary, constitutive centromeric targeting of Sororin-CTR is sufficient to replace full-length Sororin, but not Sgo1 or Esco1/2, in protecting sister-chromatid cohesion

Chromatin-tethered Sororin‑CTR impairs mitotic chromosome condensation and decatenation

The cohesin release factor Wapl regulates cohesin’s dynamic association with chromatin^25,47. Depletion of Wapl prevents cohesin release from mitotic chromosomes, leading to defective sister-chromatid resolution, impaired chromosome condensation, and segregation errors^48,50. Because Sororin antagonizes Wapl, we asked whether constitutive retention of Sororin-protected cohesin on chromatin could phenocopy Wapl depletion.

To address this, we generated HeLa cell lines stably expressing H2B-Sororin-CTR-GFP, which tethers the CTR constitutively to chromatin, or an H2B-GFP control (Fig. 3A). In these cells, we first examined the release of cohesin from chromosome arms in nocodazole-arrested mitotic cells. Following endogenous Sororin depletion, the cohesin subunit SMC1 remained largely chromosome-associated in H2B-Sororin‑CTR-GFP cells, but was efficiently removed in H2B-GFP cells (Fig. 3B), confirming that chromatin-tethered Sororin-CTR blocks cohesin release. Consistently, expression of H2B-Sororin-CTR-GFP but not H2B-GFP potently inhibited chromosome arm resolution (Fig. 3C). Moreover, mitotic chromosomes in H2B-Sororin‑CTR-GFP cells were significantly longer than those in control H2B-GFP cells (2.65 µm vs. 1.77 µm) (Fig. 3D), indicating defective chromosome condensation.

Fig. 3. Chromatin-tethered Sororin‑CTR impairs mitotic chromosome condensation and decatenation.

A Parental HeLa cells and stable cell lines expressing H2B-GFP or H2B-Sororin-CTR-GFP were analyzed by immunoblotting for GFP and α-Tubulin. α-Tubulin from a parallel gel served as a loading control. Blots are representative of two independent experiments. Uncropped blots are available in the Source Data file. B The indicated stable cell lines were transfected with Sororin siRNA and, 48 h later, treated with nocodazole for 3 h. Mitotic chromosome spreads were stained with DAPI and antibodies against GFP, SMC1, and CENP-C. Representative images from two independent experiments are shown. Scale bar = 10 μm. C The indicated stable cell lines were treated with nocodazole for 3 h and mitotic chromosome spreads were stained with DAPI and antibodies against ACA. Representative images are shown (n = 1). Scale bar = 10 μm. D The indicated stable cell lines were treated as in C. Chromosome lengths were measured from 1061 chromosomes across 20 cells per condition (n = 1061 chromosomes). Statistical significance was calculated with two-tailed, unpaired Student’s t-test. Error bars represent SD. Raw data are available in the Source Data file. E The indicated stable cell lines were synchronized in early S phase by thymidine block. Seven hours after release, live-cell imaging was performed to monitor mitotic progression. The time from nuclear envelope breakdown (NEB) to anaphase onset was measured (n = 45 for H2B-GFP; n = 60 for H2B-Sororin-CTR-GFP). Statistical significance was calculated with two-tailed, unpaired Student’s t-test. Error bars represent SD. Raw data are available in the Source Data file. F The indicated stable cell lines were treated and imaged as in E. Representative frames are shown, with arrows indicating chromosome bridges. Time is shown in h:min. Scale bar = 10 μm. See also Supplementary Movies 1 and 2. G The indicated stable cell lines were treated and imaged as in E. The percentage of anaphase with chromosome bridges was quantified (n = 1). Raw data are available in the Source Data file. H The indicated stable cell lines were stained with DAPI and antibodies against ACA. The percentage of anaphases with lagging chromosomes or chromosome bridges was determined from at least 200 cells per condition (n = 2 independent experiments). Raw data are available in the Source Data file. Representative images with arrows indicating lagging chromosomes or bridges are shown. Scale bar = 10 μm. I The indicated stable cell lines were stained with DAPI and antibodies against PICH. The percentage of PICH-positive anaphases was quantified from at least 200 cells per condition (n = 2 independent experiments). Raw data are available in the Source Data file. Representative images with arrows indicating ultra-fine bridges are shown. Scale bar = 10 μm.

We next assessed the requirement for Sgo1 and Esco1/2 under these conditions. Sgo1 knockdown caused severe cohesion loss in H2B-GFP cells, but this defect was largely suppressed in H2B-Sororin-CTR-GFP cells (Supplementary Fig. 3A, B). This indicates that chromatin-tethered Sororin-CTR can largely bypass the requirement for Sgo1, supporting the model that Sgo1 protects cohesin both at centromeres and on chromosome arms^71–75. In contrast, co-depletion of Esco1/2 resulted in comparable cohesion defects in both H2B-GFP and H2B-Sororin-CTR-GFP cells (Supplementary Fig. 3C–F), demonstrating that Esco1/2 perform non-redundant functions that cannot be bypassed by Sororin.

To determine the consequences of impaired cohesin removal, we analyzed mitosis by live-cell and fixed-cell imaging. Live-cell imaging revealed comparable mitotic durations in cells expressing H2B-GFP and H2B-Sororin‑CTR-GFP (38.5 min vs. 36.7 min; Fig. 3E, F and Supplementary Movie 1, 2). However, anaphase chromosome-segregation errors increased dramatically, from 4.4% in control cells to 38.3% in H2B-Sororin-CTR-GFP cells (Fig. 3G). Analysis of paraformaldehyde (PFA)-fixed cells showed a specific increase in anaphase chromosome bridges from 6.7% to 32.9%, whereas lagging‑chromosome rates remained similar (3.8% vs. 4.5%) (Fig. 3H). Consistently, centromeric accumulation of Aurora B, which is important for correcting erroneous kinetochore-microtubule attachments^76–81, was comparable in cells expressing H2B-GFP and H2B-Sororin-CTR-GFP (Supplementary Fig. 3G, H), indicating that the increased chromosome bridges do not result from Aurora B mislocalization. Furthermore, H2B-Sororin-CTR-GFP cells exhibited a ~ 3-fold increase in ultrafine DNA bridges marked by PICH (37.9% vs. 12.4%; Fig. 3I), indicative of defective decatenation. A similar increase in anaphase bridges was observed upon overexpression of untethered Sororin-CTR-GFP (Supplementary Fig. 3I–K), linking these defects directly to the CTR’s cohesion-stabilizing activity.

In summary, constitutive chromatin tethering of Sororin-CTR prevents mitotic cohesin removal and recapitulates key features of Wapl loss, including defective chromosome condensation, impaired decatenation, and erroneous chromosome segregation.

Sororin-CTR directly binds the closed RAD21-SMC3 interface

To elucidate the molecular basis of Sororin-CTR-mediated cohesion stabilization, we examined its interaction with the cohesin complex. Bacterially expressed GST-Sororin fragments were used in pull-down assays with HeLa cell lysates. Full-length GST-Sororin and GST-Sororin-CTR, but not GST-Sororin (1-222), bound endogenous SMC3, SMC1, RAD21, and SA2 (Fig. 4A, B). While full-length Sororin and the N-terminal fragment interacted with Pds5B, GST-Sororin-CTR did not, indicating that the CTR associates with the cohesin core independently of Pds5.

Fig. 4. Sororin-CTR directly binds the closed RAD21-SMC3 interface.

A HeLa cells were synchronized in late G2 phase by treatment with RO-3306 for 18 h. Cell lysates were subjected to pull-down assays with GST or GST-Sororin (WT or residues 1-222), followed by immunoblotting for SMC1, SMC3, RAD21, SA2, and Pds5B, and Coomassie Brilliant Blue (CBB) staining. Blots are representative of three independent experiments. Uncropped blots are available in the Source Data file. B HeLa cells were synchronized in late G2 phase as described in A and subjected to pull-down with GST or GST-Sororin (residues 1-222 or 223-252), followed by immunoblotting and CBB staining. Blots represent three independent experiments. Uncropped blots are available in the Source Data file. C SA2-KO cells were synchronized in late G2 as described in A and subjected to pull-down with the indicated proteins, followed by immunoblotting and CBB staining. Blots represent three independent experiments. Uncropped blots are available in the Source Data file. D HeLa and SA2-KO HeLa cells were analyzed by immunoblotting for SA2 and α-Tubulin. Blots represent three independent experiments. Uncropped blots are available in the Source Data file. E HeLa cells were transfected with control siRNA or RAD21-targeting siRNA (#2). After 48 h, cells were subjected to pull-down with GST or GST-Sororin-CTR, followed by immunoblotting for SMC1, SMC3, and RAD21, and CBB staining. Blots are representative of three independent experiments. Uncropped blots are available in the Source Data file. F Schematic diagram of the domain organization of human RAD21. G HEK-293T cells expressing SMC1-GFP, SMC3-GFP and/or RAD21-Myc were subjected to pull-down with GST or GST-Sororin-CTR, followed by immunoblotting for GFP and the Myc tag, and CBB staining, representative of three independent experiments. Uncropped blots are available in the Source Data file. H HEK-293T cells expressing SMC3-HD-Flag and/or RAD21-N103-Myc were subjected to pull-down with GST or GST-Sororin-CTR, followed by immunoblotting for the Flag and Myc tags, and CBB staining. Blots are representative of three independent experiments. Uncropped blots are available in the Source Data file. I Recombinant SMC3-HD-Flag and RAD21-N103-2xStrep were subjected to pull-down with GST or GST-Sororin-CTR, followed by immunoblotting for the Flag and Strep tags, and CBB staining. Blots are representative of three independent experiments. Uncropped blots are available in the Source Data file. J GST-Sororin were subjected to pull-down with recombinant SMC3-HD-Flag and RAD21-N103-2xStrep, followed by immunoblotting for GST and CBB staining. Blots are representative of three independent experiments. Uncropped blots are available in the Source Data file. K HeLa cells synchronized at the G1/S boundary (0 h after double-thymidine release) or in G2 phase (8 h after release) were subjected to pull-down with GST or GST-Sororin-CTR, followed by immunoblotting for SMC1, SMC3, and RAD21, and CBB staining. Blots are representative of two independent experiments. Uncropped blots are available in the Source Data file. L HeLa cells synchronized in late G2 phase (RO-3306 treatment for 16 h) or in mitosis (nocodazole treatment for 16 h) were subjected to pull-down with GST or GST-Sororin-CTR, followed by immunoblotting for SMC1, SMC3, and RAD21, and CBB staining. Blots are representative of two independent experiments. Uncropped blots are available in the Source Data file. M HeLa cells stably expressing Sororin-GFP were synchronized in mitosis by nocodazole treatment for 16 h and subjected to chromatin fractionation. A portion of the whole cell lysate (WCL), supernatant, and chromatin pellet were analyzed by immunoblotting for histone H3, GAPDH, GFP, RAD21, SMC3, and Pds5B. Blots are representative of three independent experiments. Uncropped blots are available in the Source Data file. N Cells were treated as in M. The remaining chromatin pellet and 5% of the supernatant were subjected to immunoprecipitation with anti-GFP beads, followed by immunoblotting for GFP, RAD21, SMC3, and Pds5B. Blots are representative of three independent experiments. S., short exposure; L., long exposure. Uncropped blots are available in the Source Data file.

A previous study suggested that the C‑terminus of Sororin interacts with the SA2 subunit⁶⁵. To test whether SA2 is required for cohesin’s interaction with Sororin‑CTR, we performed pull-down assays in SA2-knockout HeLa cells generated by CRISPR/Cas9. GST-Sororin-CTR retained binding to SMC3, SMC1, RAD21, and SA1 in the absence of SA2 (Fig. 4C, D and Supplementary Fig. 4A, B). In contrast, siRNA-mediated depletion of RAD21 abolished the interaction between GST-Sororin-CTR and the SMC subunits (Fig. 4E), identifying RAD21 as the critical anchor. Consistent with the previous observation that the D793K mutation prevented SA2 from binding RAD21⁸², this mutation eliminated the co-immunoprecipitation of Myc-tagged SA2 with GFP-fused RAD21 in HEK-293T cells (Supplementary Fig. 4C). Notably, this mutation also prevented Myc-SA2 from binding GST-Sororin-CTR (Supplementary Fig. 4D), indicating that SA2 binding to Sororin-CTR depends on its interaction with RAD21.

Within the cohesin ring, RAD21 connects to the head domains of SMC3 and SMC1 via its N-terminal helical domain (NHD) and C-terminal winged-helix domain (WHD), respectively (Fig. 4F)^34,83. To determine which protein is recognized by Sororin-CTR, we co-expressed Myc-tagged RAD21 with GFP-tagged SMC3 or SMC1 in HEK-293T cells. GST‑Sororin‑CTR selectively pulled down RAD21‑Myc together with SMC3‑GFP, but not with SMC1‑GFP (Fig. 4G), demonstrating preferential binding to the RAD21-SMC3 interface.

Because the RAD21-SMC3 interaction involves RAD21’s NHD and the head-proximal coiled-coil of SMC3 (referred to as SMC3‑HD)^34,83, we tested whether these regions were sufficient for Sororin-CTR binding. Pulldown assays showed that GST-Sororin-CTR bound Flag-tagged SMC3-HD only when co-expressed with Myc-tagged RAD21 fragments comprising residues 1-281 (referred to as N281) or residues 1-103 (referred to as N103), and that deletion of RAD21’s Pds5‑interaction region (residues 102-122) did not affect this interaction (Fig. 4H and Supplementary Fig. 4E).

We further generated a recombinant protein complex formed between 2xStrep-tagged RAD21-N103 and SMC3-HD-Flag, which were co-expressed in E. coli and purified with the streptavidin beads. Pull-down assays showed that this minimal RAD21-N103/SMC3-HD complex directly bound GST-Sororin-CTR and GST-Sororin (Fig. 4I, J), establishing a direct, specific interaction.

We next examined the cell-cycle regulation of this interaction. We synchronized HeLa cells in early-S phase and G2 phase by a double-thymidine block-and-release protocol and performed pull-down assays. GST-Sororin-CTR bound endogenous SMC3, SMC1, and RAD21 more efficiently in G2-phase cell lysates than in early S-phase lysates (Fig. 4K). Furthermore, this interaction persisted in late-G2 phases (arrested by the Cdk1 inhibitor RO-3306) but was markedly reduced in mitotic cells (arrested by nocodazole) (Fig. 4L). Thus, Sororin-CTR binding to cohesin peaks during the cohesion-maintenance phase of the cell cycle.

Finally, to distinguish chromatin-bound from soluble cohesin pools, we performed chromatin fractionation followed by co-immunoprecipitation. Sororin-GFP stably expressed in HeLa cells co-precipitated RAD21, SMC3, and Pds5B specifically from the chromatin-bound fraction of mitotic cells, but not from the soluble fraction (Fig. 4M, N), confirming that Sororin interacts with functional, chromatin-associated cohesin.

Collectively, these results demonstrate that Sororin’s CTR directly and specifically binds the RAD21-SMC3 interface in a cell-cycle-regulated manner on chromatin, consistent with a model in which Sororin stabilizes cohesion by locking the closed DNA exit gate of cohesin.

Mitotic phosphorylation of Sororin selectively disrupts Pds5B binding while preserving interaction with the RAD21-SMC3 interface

Sororin phosphorylation in mitosis promotes cohesin release from chromosome arms^27,46,54,62. To examine whether this modification also affects Sororin’s interaction with the RAD21-SMC3 interface, we arrested HeLa cells co-expressing Sororin-GFP, RAD21-N103-Myc, and SMC3-HD-Flag either in G2 (using RO-3306) or in mitosis (with nocodazole), followed by immunoprecipitation of Sororin-GFP. In G2-arrested cells, Sororin-GFP efficiently co-immunoprecipitated RAD21-N103-Myc, SMC3-HD-Flag, and endogenous Pds5B (Fig. 5A). In mitotic cells, however, the Sororin-Pds5B interaction was abolished, whereas Sororin’s binding to RAD21-N103 and SMC3-HD was retained. Moreover, λ-phosphatase treatment of lysates from nocodazole-treated HEK-293T cells markedly enhanced the Sororin-Pds5B co-immunoprecipitation, without affecting interactions with RAD21-N103 or SMC3-HD (Fig. 5B). Similarly, inhibition of CDK1 (Roscovitine), Aurora B (Hesperadin), or Plk1 (BI 2536) in nocodazole-arrested mitotic HeLa cells restored Sororin-Pds5B interaction, but had no effect on Sororin association with RAD21-N103 or SMC3-HD (Fig. 5C). These results indicate that mitotic phosphorylation selectively disrupts Sororin’s interaction with Pds5B while leaving its binding to the RAD21-SMC3 interface intact.

Fig. 5. Mitotic phosphorylation of Sororin selectively disrupts Pds5B binding while preserving interaction with the RAD21-SMC3 interface.

A HeLa cells co-expressing RAD21-N103-Myc, SMC3-HD-Flag, and Sororin-GFP were synchronized in late G2 by RO-3306 treatment (16 h) or in mitosis by nocodazole treatment (16 h). Cells were subjected to anti-GFP immunoprecipitation followed by immunoblotting for GFP, Pds5B, and the Myc and Flag tags. Blots are representative of three independent experiments. S., short exposure; L., long exposure. Uncropped blots are available in the Source Data file. B HEK-293T cells co-expressing RAD21-N103-Myc, SMC3-HD-Flag, and Sororin-GFP were treated with nocodazole for 18 h. Cell lysates were incubated with or without λ-PPase and subjected to anti-GFP immunoprecipitation followed by immunoblotting for GFP, Pds5B, and the Myc and Flag tags. Blots are representative of three independent experiments. Uncropped blots are available in the Source Data file. C HeLa cells co-expressing RAD21-N103-Myc, SMC3-HD-Flag, and Sororin-GFP were synchronized in mitosis with nocodazole for 16 h, followed by treatment with MG132 and the indicated inhibitors (or DMSO) for 2 h. Cells were then subjected to anti-GFP immunoprecipitation followed by immunoblotting for GFP, Pds5B, and the Myc and Flag tags. Blots are representative of three independent experiments. Noc, nocodazole; Ros, Roscovitine; BI, BI 2536; Hes, Hesperadin. S., short exposure; L., long exposure. Uncropped blots are available in the Source Data file. D HEK-293T cells expressing Sororin-GFP (WT or ASE) were subjected to anti-GFP immunoprecipitation, followed by immunoblotting for GFP, Pds5B, RAD21, SA2, SMC1, and SMC3. Blots are representative of three independent experiments. S., short exposure; L., long exposure. Uncropped blots are available in the Source Data file. E HeLa cells were subjected to pull-down with GST or GST-Sororin (WT or ASE), followed by immunoblotting for Pds5B, RAD21, SMC3, Pds5A, GAPDH, and CBB staining. Blots are representative of three independent experiments. Uncropped blots are available in the Source Data file. F Sequence alignment of vertebrate Sororin around the YSR motif from the indicated vertebrate species. The conserved serine at position 145 is marked with an asterisk. G Asynchronous or nocodazole-arrested mitotic HeLa cells were subjected to immunoprecipitation with control IgG or Sororin antibody, followed by immunoblotting for Sororin-pS145 and Sororin. Blots are representative of two independent experiments. Uncropped blots are available in the Source Data file. H HeLa cells stably expressing Flag-Sororin were synchronized in mitosis with nocodazole and then treated with MG132 and Hesperadin (or DMSO) for 2 h. Cells were subjected to anti-Flag immunoprecipitation followed by immunoblotting for Sororin-pS145 and the Flag tag. Blots are representative of two independent experiments. Uncropped blots are available in the Source Data file. I HEK-293T cells expressing Flag-Sororin (WT or S145A) were immunoprecipitated with anti-Flag beads and subjected to in vitro phosphorylation by GST-Aurora B/IN-box, followed by immunoblotting for GST, Sororin-pS145, and the Flag tag. Blots are representative of two independent experiments. Uncropped blots are available in the Source Data file. J MBP-Pds5B-N300 was subjected to pull-down by Sororin (WT or pS145) peptides immobilized on beads, followed by immunoblotting for MBP and streptavidin-HRP. Blots are representative of two independent experiments. Uncropped blots are available in the Source Data file. K Sororin peptide was subjected to in vitro phosphorylation by GST-Aurora B/IN-box. After stringent salt washes, the peptides were used to pull down MBP-Pds5B-N300, followed by immunoblotting for MBP and streptavidin-HRP. Blots are representative of two independent experiments. Uncropped blots are available in the Source Data file.

Sororin contains a conserved YSR motif that is required for Pds5 binding²⁶. Consistently, co-immunoprecipitation assays showed that endogenous Pds5B associated with Sororin-GFP, but not with the Sororin-ASE mutant in which the YSR motif (residues 146-148) was replaced by ASE (Fig. 5D). GST pull-down assays further demonstrated that GST-Sororin, but not GST-Sororin-ASE, bound endogenous Pds5B (but not its paralog Pds5A) from HeLa cell lysates (Fig. 5E). Notably, the Sororin-ASE mutant exhibited reduced binding to cohesin core subunits, indicating that Pds5B contributes to the overall association of Sororin with cohesin.

To identify the phosphorylation site responsible for disrupting Pds5B binding, Sororin-GFP immunoprecipitated from HeLa cells was analyzed by mass spectrometry, revealing 18 phosphorylation sites (Supplementary Fig. 5A and Supplementary Data 1). Among these, serine 145 (S145), which lies adjacent to the YSR motif and is conserved across vertebrates, emerged as a strong candidate (Fig. 5F). Using a phospho-specific antibody raised against Sororin phosphorylated at S145 (Sororin-pS145), we found that S145 phosphorylation was detected exclusively in mitotic cells (Fig. 5G). Antibody specificity was confirmed using Flag-tagged Sororin purified from mitotic cells, as the signal was absent in the S145A mutant and abolished by λ-phosphatase treatment (Supplementary Fig. 5B, C).

S145 resides within a consensus Aurora kinase phosphorylation motif (R/K-x-S/T-Φ). Accordingly, treatment of nocodazole-arrested mitotic cells with Hesperadin eliminated the pS145 signal on Flag-Sororin (Fig. 5H), indicating that Sororin-S145 phosphorylation in cells requires Aurora B kinase activity. To determine whether Aurora B can phosphorylate Sororin at S145, we performed in vitro kinase reactions using bacterially produced recombinant GST-Aurora B in complex with the IN-box of INCENP (GST-Aurora B/IN-box)⁸⁴. In the presence of ATP, this complex phosphorylated Flag-Sororin at S145 (Fig. 5I). These results indicate that Aurora B is the kinase responsible for Sororin-S145 phosphorylation.

We next assessed the effect of S145 phosphorylation on the Sororin-Pds5B interaction. Pull-down assays using a synthesized Sororin peptide (residues 136-158) and a recombinant MBP-tagged Pds5B fragment (residues 1-300) showed that the non-phosphorylated peptide bound Pds5B, whereas the S145-phosphorylated peptide did not (Fig. 5J). Moreover, in vitro phosphorylation of the Sororin peptide by GST-Aurora B/IN-box similarly abolished Pds5B binding (Fig. 5K), demonstrating that S145 phosphorylation directly disrupts this interaction.

Consistent with these in vitro findings, co-immunoprecipitation assays in G2-synchronized HeLa cells revealed that Myc-tagged Pds5B associated with wild-type Sororin-GFP, but not with the phospho-mimetic Sororin-S145E-GFP mutant (Supplementary Fig. 5D). Moreover, Aurora B inhibition in mitotic cells enhanced Pds5B binding to wild-type Sororin-GFP, but not to Sororin-S145E-GFP (Supplementary Fig. 5E), confirming that Aurora B-dependent phosphorylation at S145 disrupts the Sororin-Pds5B interaction in vivo.

We further tested the effect of a phospho-deficient mutation, S145N, on Sororin interactions. In mitotic cells, Sororin-S145N exhibited enhanced association with Pds5B, SMC3, and RAD21 compared with wild-type Sororin, whereas no difference was observed in G2-phase cells (Supplementary Fig. 5F, G).

Finally, because Sgo1-associated PP2A protects centromeric cohesion by dephosphorylating Sororin^51–54, we examined whether the S145N mutant could bypass Sgo1 depletion. Although the S145N mutant partially rescued the cohesion defect caused by Sgo1 knockdown, it did not fully restore cohesion (Supplementary Fig. 5H–J). This indicates that preservation of the Sororin-Pds5 interaction via S145 dephosphorylation represents a major, but not exclusive, function of Sgo1-PP2A in cohesion protection.

In summary, Aurora B-mediated phosphorylation of Sororin at S145 acts as a mitotic switch that selectively disrupts its interaction with Pds5B, thereby promoting cohesin release from chromosome arms, while leaving its gate-locking interaction with the RAD21-SMC3 interface intact.

Identification of key residues mediating the interaction between Sororin and the RAD21-SMC3 interface

The 30-residue CTR of Sororin (residues 223-252) is enriched in hydrophobic (17) and acidic (9) residues. AlphaFold predictions indicate that the segment spanning residues K227 to E245 likely adopts an α-helical conformation (AlphaFoldDB: Q96FF9). To elucidate how Sororin interacts with the RAD21-SMC3 interface, we used AlphaFold3 to model the binding of a Sororin-CTR peptide to the interface formed by RAD21-NHD and SMC3-HD (Fig. 6A). The model predicts multiple specific contacts: Sororin W233 with RAD21 Y3; Sororin F241 with RAD21 K50 and SMC3 L1031; Sororin A243 and F247 with SMC3 K1038; Sororin E245 with SMC3 K1034; and Sororin E252 with SMC3 R1099. Sequence alignment revealed that these residues are evolutionarily conserved among vertebrates, with the exception of Sororin W233, which is replaced by a similarly hydrophobic leucine in zebrafish (Fig. 6B and Supplementary Fig. 6A, B).

Fig. 6. Identification of key residues mediating the interaction between Sororin and the RAD21-SMC3 interface.

A AlphaFold3 model of the RAD21-SMC3-Sororin-CTR complex. RAD21, SMC3, and Sororin are shown in cyan, green, and magenta, respectively. The binding interface between Sororin-CTR and RAD21-SMC3 is shown. B Multiple sequence alignment of Sororin-CTR from the indicated vertebrate species. Residues required for binding to the RAD21-SMC3 interface are indicated by asterisks. C Recombinant SMC3-HD-Flag and RAD21-N103-2xStrep were subjected to pull-down with GST or GST-Sororin-CTR (WT, W233A/E252K, or F241A/F247A), followed by immunoblotting for the Flag and Strep tags, and CBB staining. Blots are representative of three independent experiments. Uncropped blots are available in the Source Data file. D HeLa cells were subjected to pull-down with GST or GST-Sororin-CTR (WT, W233A/E252K, or F241A/F247A), followed by immunoblotting for SMC3, SMC1, RAD21, and SA2, and CBB staining. Blots are representative of three independent experiments. Uncropped blots in the Source Data file. E HEK-293T cells co-expressing SMC3-HD-Flag and RAD21-N103-Myc (WT or Y3A/K50E) were subjected to pull-down with GST or GST-Sororin-CTR, followed by immunoblotting for the Flag and Myc tags, and CBB staining. Blots are representative of three independent experiments. Irrelevant lanes were excised from the gel image (indicated by a vertical line). Uncropped blots are available in the Source Data file. F HEK-293T cells co-expressing SMC3-HD-Flag (WT or K1034E/K1038E) and RAD21-N103-Myc were subjected to pull-down with GST or GST-Sororin-CTR, followed by immunoblotting for the Flag and Myc tags, and CBB staining. Blots are representative of three independent experiments. Uncropped blots are available in the Source Data file. G HEK-293T cells co-expressing SMC3-HD-Flag (WT or R1099E) and RAD21-N103-Myc (WT or Y3A) were subjected to pull-down with GST or GST-Sororin-CTR, followed by immunoblotting for the Flag and Myc tags. Blots are representative of three independent experiments. Uncropped blots are available in the Source Data file. H Recombinant GST-Sororin-CTR proteins were subjected to pull-down using recombinant SMC3-HD-Flag (WT or K1034E/K1038E) and RAD21-N103-2xStrep (WT or Y3A/K50E), followed by immunoblotting with GST and the Strep tag, and CBB staining. Blots are representative of three independent experiments. Uncropped blots are available in the Source Data file. I HEK-293T cells expressing Myc-Sororin were subjected to pull-down using recombinant SMC3-HD-Flag (WT or K1034E/K1038E) and RAD21-N103-2xStrep (WT or Y3A/K50E), followed by immunoblotting for the Myc and Strep tags, and CBB staining. Blots are representative of three independent experiments. Uncropped blots are available in the Source Data file.

To experimentally test these predictions, we performed binding assays using bacterially expressed recombinant proteins. A complex consisting of RAD21-N103-2xStrep and SMC3-HD-Flag was efficiently pulled down by GST-Sororin-CTR, but not by the double mutant GST-Sororin-CTR-W233A/E252K (hereafter WA/EK), in which W233 and E252 were mutated to alanine and lysine, respectively (Fig. 6C). A comparable loss of binding was observed for the GST-Sororin-CTR-F241A/F247A mutant (hereafter FA/FA), where both F241 and F247 were replaced by alanine. Consistently, pull-down assays in HeLa cells showed that endogenous SMC3, SMC1, RAD21, and SA2 interacted with GST-Sororin-CTR, but not with the WA/EK or FA/FA mutants (Fig. 6D). Moreover, a cancer-associated Sororin-E252Q mutation reduced co-immunoprecipitation of SMC3 and RAD21 with Sororin-Flag in HEK-293T cells (Supplementary Fig. 6C), suggesting that this mutation partially compromises Sororin’s gate-locking function and may contribute to genomic instability.

We next examined the contribution of residues within RAD21-NHD and SMC3-HD to Sororin binding. In HEK-293T cells, simultaneous mutation of RAD21 Y3 to alanine and K50 to glutamic acid (hereafter YA/KE) abolished the ability of co-expressed RAD21-N103-Myc and SMC3-HD-Flag to bind GST-Sororin-CTR (Fig. 6E). Co-immunoprecipitation assays indicated that the YA/KE mutations did not affect the interaction between RAD21-N103-Myc and SMC3-HD-Flag (Supplementary Fig. 6D). Similarly, mutation of SMC3 K1034 and K1038 to glutamic acid (hereafter KE/KE) strongly reduced pull-down of co-expressed SMC3-HD-Flag and RAD21-N103-Myc by GST-Sororin-CTR (Fig. 6F), without affecting RAD21-SMC3 association (Supplementary Fig. 6E). Additional binding assays performed in HEK-293T cells showed that GST-Sororin-CTR failed to pull down the co-expressed mutants RAD21-N103-Y3A and SMC3-HD-R1099E-Flag, in which residues RAD21 Y3 and SMC3 R1099 were mutated to alanine and glutamic acid, respectively (Fig. 6G). We confirmed that these two mutations did not perturb the binding between RAD21 and SMC3 (Supplementary Fig. 6F).

To further validate these interactions using a fully recombinant system, we co-expressed RAD21-N103-2xStrep and SMC3-HD-Flag in E. coli and immobilized this complex on streptavidin beads. The wild-type RAD21-N103-2xStrep/SMC3-HD-Flag complex efficiently pulled down GST-Sororin-CTR, whereas complexes containing either RAD21-N103-YA/KE or SMC3-HD-KE/KE failed to do so (Fig. 6H). Consistent results were obtained when these complexes were used to pull down Myc-Sororin expressed in HEK-293T cells (Fig. 6I).

Esco1/2-mediated acetylation of SMC3 at K105/K106 promotes Sororin recruitment to chromatin in S phase^{26,27,35,39–45}. Notably, neither acetylation-deficient (K105A/K106A) nor acetylation-mimetic (K105Q/K106Q) mutations in SMC3 affected co-immunoprecipitation of Sororin-GFP with the SMC3-HD-Flag/RAD21-N103-Myc complex (Supplementary Fig. 6G). Consistently, AlphaFold3 modeling indicates that K105 and K106 are spatially distant from the Sororin-binding residues (K1034, K1038, R1099) on SMC3 (Supplementary Fig. 6H; AlphaFoldDB: Q9UQE7). Thus, SMC3 acetylation licenses Sororin recruitment without directly regulating its binding to the RAD21-SMC3 interface.

Collectively, these data identify critical residues in Sororin-CTR, RAD21-NHD, and SMC3-HD that together mediate Sororin binding to the RAD21-SMC3 interface.

Sororin-CTR residues that bind the RAD21-SMC3 interface are required for sister-chromatid cohesion

We next assessed whether the Sororin-CTR residues that mediate binding to the RAD21-SMC3 interface are required for sister-chromatid cohesion. To this end, we analyzed chromosome-arm resolution in HeLa cells transiently expressing H2B-fused Sororin-CTR-GFP, either wild type or carrying the WA/EK or FA/FA mutations (Fig. 7A). Chromosome spreads from nocodazole-arrested mitotic cells showed that expression of wild-type H2B-Sororin-CTR-GFP, but not the WA/EK or FA/FA mutants, prevented arm resolution and markedly impaired chromosome condensation (Fig. 7B, C and Supplementary Fig. 7A). Similar phenotypes were observed in non-transformed RPE-1 cells expressing these H2B fusion proteins (Supplementary Fig. 7B-D). Consistently, expression of H2B-fused full-length Sororin-GFP, but not the corresponding WA/EK or FA/FA mutants, inhibited mitotic arm resolution in both HeLa and RPE-1 cells (Supplementary Fig. 7E–J).

Fig. 7. Sororin-CTR residues that bind the RAD21-SMC3 interface are required for sister-chromatid cohesion.

A HeLa cells were transfected with H2B-GFP or H2B-Sororin-CTR-GFP (WT, W233A/E252K, or F241A/F247A). Twenty-four hours post-transfection, cell lysates were immunoblotted for GFP and α-Tubulin. α-Tubulin from a parallel gel served as a loading control. Blots are representative of three independent experiments. Uncropped blots are available in the Source Data file. B Cells were transfected as in A. Twenty-four hours post-transfection, cells were treated with nocodazole for 3 h. Mitotic chromosome spreads were stained with DAPI. The percentage of mitotic cells exhibiting predominantly resolved or unresolved arm cohesion was quantified from at least 300 cells per condition across three independent experiments (n = 3). Statistical significance was calculated with two-tailed, unpaired Student’s t-test. Error bars represent SD. Raw data are available in the Source Data file. C HeLa cells were treated as in B. Representative images from three independent experiments are shown. Scale bars = 10 μm. D HeLa cells stably expressing Sororin-GFP (WT, W233A/E252K, or F241A/F247A) were analyzed by immunoblotting for GFP and GAPDH. Blots are representative of three independent experiments. Uncropped blots are available in the Source Data file. E HeLa cells stably expressing Sororin-GFP (WT, W233A/E252K, or F241A/F247A) were synchronized in G2 phase by treatment with RO-3306 for 18 h. Cell lysates were subjected to immunoprecipitation using anti-GFP beads, followed by immunoblotting for SMC3, SMC1, RAD21, and GFP. Blots are representative of three independent experiments. Uncropped blots are available in the Source Data file. F Parental HeLa cells and stable cell lines expressing Sororin-GFP (WT, W233A/E252K, or F241A/F247A) were transfected with control or Sororin siRNA. Forty-eight hours post-transfection, cell lysates were immunoblotted for GFP, Sororin and GAPDH. GAPDH from a parallel gel served as a loading control. Blots are representative of three independent experiments. Uncropped blots are available in the Source Data file. G Cells were transfected as in F. Forty-eight hours post-transfection, cells were analyzed for mitotic index. At least 3000 cells were scored per condition across three independent experiments (n = 3). Statistical significance was calculated with two-tailed, unpaired Student’s t-test. Error bars represent SD. Raw data are available in the Source Data file. H Cells were treated and analyzed as in G. Phase-contrast images, representative of three independent experiments, are shown; round cells scored as mitotic. Scale bars = 100 μm. I The indicated cells were transfected with control or Sororin siRNA. Forty-eight hours post-transfection, cells were treated with nocodazole for 3 h. Mitotic chromosome spreads were stained with DAPI and antibodies against GFP and ACA. The percentage of mitotic cells exhibiting predominantly separated or unseparated sister chromatids was quantified from at least 300 cells per condition across three independent experiments (n = 3). Statistical significance was calculated with two-tailed, unpaired Student’s t-test. Error bars represent SD. Raw data are available in the Source Data file. J Cells were treated and analyzed as in I. Representative images from three independent experiments are shown. Scale bars = 10 μm.

We next tested whether the WA/EK and FA/FA mutants could support Sororin’s function in protecting sister-chromatid cohesion. To this end, we generated HeLa cell lines stably expressing wild-type Sororin-GFP or the WA/EK and FA/FA mutants (Fig. 7D). Co-immunoprecipitation assays confirmed that endogenous RAD21, SMC3, and SMC1 associated with wild-type Sororin-GFP, but not with the WA/EK or FA/FA mutants (Fig. 7E). Following siRNA-mediated depletion of endogenous Sororin, cells expressing the WA/EK or FA/FA mutants displayed a pronounced increase in mitotic index, whereas cells expressing wild-type Sororin-GFP did not (Fig. 7F–H). Consistently, chromosome-spread analyses showed that the premature cohesion loss induced by Sororin depletion was rescued by exogenous expression of wild-type Sororin-GFP or Sororin-SFB, but not by the WA/EK or FA/FA mutants, in both HeLa and RPE-1 cells (Fig. 7I, J and Supplementary Fig. 7K–M).

Collectively, these results demonstrate that Sororin-CTR residues required for binding to the RAD21-SMC3 interface are essential for Sororin-mediated protection of sister-chromatid cohesion.

RAD21 and SMC3 residues that bind Sororin-CTR are essential for sister-chromatid cohesion

We next investigated the functional significance of the RAD21 and SMC3 residues that mediate binding to Sororin-CTR. We first examined sister-chromatid cohesion in cells in which endogenous RAD21 was replaced with the exogenous RAD21-YA/KE mutant defective in Sororin binding. To this end, we established HeLa-derived cell lines stably expressing siRNA-resistant RAD21 (wild-type or the YA/KE mutant) fused to SFB, which is a triple-tag consisting of an S tag, a Flag tag, and a biotin-binding peptide (Fig. 8A). Co-immunoprecipitation assays confirmed that wild-type RAD21-SFB interacted with endogenous Sororin, whereas the RAD21-YA/KE mutant did not; importantly, the YA/KE mutant retained its interaction with endogenous SMC3 (Fig. 8B). Mitotic chromosome-spread analysis showed that, following depletion of endogenous RAD21, 44.2% of HeLa cells exhibited premature loss of cohesion (Fig. 8C–E). This defect was efficiently rescued by expression of wild-type RAD21-SFB, but not by the YA/KE mutant, demonstrating that RAD21 residues required for Sororin-CTR binding are critical for sister-chromatid cohesion.

Fig. 8. RAD21 and SMC3 residues that bind Sororin-CTR are essential for sister-chromatid cohesion.

A Parental HeLa cells and stable cell lines expressing inducible RAD21-SFB (WT or Y3A/K50E) were treat with 1 μg/ml doxycycline (Dox) for 24 h, followed by immunoblotting for RAD21 and GAPDH. GAPDH from a parallel gel served as a loading control. Blots are representative of three independent experiments. Uncropped blots are available in the Source Data file. B The indicated stable cell lines treated as in A were subjected to immunoprecipitation using anti-Flag beads, followed by immunoblotting for the Flag tag, SMC3, and Sororin. Irrelevant lanes were excised from the gel image (indicated by a vertical line). Blots are representative of three independent experiments. Uncropped blots are available in the Source Data file. C The indicated stable cell lines were treated with Dox and transfected with control or RAD21 siRNA. Sixty hours post-transfection, cell lysates were immunoblotted for RAD21 and GAPDH. GAPDH from a parallel gel served as a loading control. Blots are representative of three independent experiments. Uncropped blots are available in the Source Data file. D The indicated stable cell lines were treated as in C. Sixty hours post-transfection, cells were treated with nocodazole for 3 h. Mitotic chromosome spreads were stained with DAPI and antibodies against the Flag tag and CENP-C. The percentage of mitotic cells exhibiting predominantly separated or unseparated sister chromatids was quantified from at least 300 cells per condition across three independent experiments (n = 3). Statistical significance was calculated with two-tailed, unpaired Student’s t-test. Error bars represent SD. Raw data are available in the Source Data file. E The indicated stable cell lines were treated and analyzed as in D. Representative images from three independent experiments are shown. Scale bars = 10 μm. F HeLa cells stably expressing SMC3-GFP (WT or K1034E/K1038E) were analyzed by immunoblotting for GFP and GAPDH. GAPDH from a parallel gel served as a loading control. Blots are representative of three independent experiments. Uncropped blots are available in the Source Data file. G The indicated stable cell lines were immunoprecipitated using anti-GFP beads, followed by immunoblotting for GFP, RAD21, and Sororin. Blots are representative of two independent experiments. S., short exposure; L., long exposure. Uncropped blots are available in the Source Data file. H The indicated stable cell lines were transfected with control or SMC3 siRNA. Seventy-two hours post-transfection, cell lysates were immunoblotted for GFP, SMC3, and α-Tubulin. α-Tubulin from a parallel gel served as a loading control. Blots are representative of three independent experiments. Uncropped blots are available in the Source Data file. I The indicated stable cell lines were transfected as in H. Seventy-two hours post-transfection, cells were treated with nocodazole for 3 h. Mitotic chromosome spreads were stained with DAPI and antibodies against GFP and ACA. The percentage of mitotic cells exhibiting fully separated, partially separated, or unseparated sister chromatids was quantified from at least 300 cells per condition across three independent experiments (n = 3). Statistical significance was calculated with two-tailed, unpaired Student’s t-test. Error bars represent SD. Raw data are available in the Source Data file. J The indicated stable cell lines were treated and analyzed as in I. Representative images from three independent experiments are shown. Scale bars = 10 μm.

We next determined the contribution of SMC3 residues involved in Sororin binding. We generated HeLa cell lines stably expressing siRNA-resistant SMC3-GFP, either wild type or carrying the Sororin-binding-defective KE/KE mutations (Fig. 8F). Co-immunoprecipitation assays showed that the SMC3-KE/KE mutant failed to interact with Sororin but retained binding to RAD21 (Fig. 8G). Depletion of endogenous SMC3 in these cells caused a cohesion defect in control cells (Fig. 8H–J), albeit milder than that observed following RAD21 depletion, likely due to reduced cell viability upon SMC3 loss^85,86. Notably, this defect was rescued by expression of wild-type SMC3-GFP but not by the SMC3-KE/KE-GFP mutant, indicating that SMC3 residues mediating Sororin-CTR binding are required for proper cohesion.

Collectively, these results establish that specific residues in both RAD21 and SMC3 that mediate interaction with Sororin-CTR are essential for maintaining sister-chromatid cohesion.

Interaction between Sororin and the RAD21-SMC3 interface is dispensable for cohesion in the absence of Wapl

Previous studies have shown that Sororin is required for sister-chromatid cohesion only in the presence of Wapl^27,60, suggesting that Sororin preserves cohesion primarily by antagonizing Wapl activity. We therefore asked whether depletion of Wapl could suppress the cohesion defects caused by Sororin mutants defective in binding the RAD21-SMC3 interface. HeLa cells stably expressing Sororin-GFP mutants were subjected to siRNA-mediated depletion of Sororin and Wapl, either individually or in combination (Fig. 9A). Mitotic chromosome-spread analysis revealed that the premature cohesion loss observed upon replacement of endogenous Sororin with Sororin-WA/EK-GFP or Sororin-FA/FA-GFP was effectively suppressed by Wapl depletion (Fig. 9B, C).

Fig. 9. Interaction between Sororin and the RAD21-SMC3 interface is dispensable for cohesion in the absence of Wapl.

A HeLa cells stably expressing Sororin-GFP (W233A/E252K or F241A/F247A) were transfected with control siRNA, Sororin siRNA and/or Wapl siRNA. Forty-eight hours post-transfection, cell lysates were immunoblotted for Wapl, GFP, Sororin, and GAPDH. GAPDH from a parallel gel served as a loading control. Blots are representative of three independent experiments. Uncropped blots are available in the Source Data file. B The indicated stable cell lines were transfected as in A. Forty-eight hours post-transfection, cells were treated with nocodazole for 3 h. Mitotic chromosome spreads were stained with DAPI and antibodies against GFP and ACA. The percentages of mitotic cells exhibiting predominantly separated or unseparated sister chromatids were quantified from at least 300 cells per condition across three independent experiments (n = 3). Statistical significance was calculated with two-tailed, unpaired Student’s t-test. Error bars represent SD. Raw data are available in the Source Data file. C Cells were treated and analyzed as in B. Representative images from three independent experiments are shown. Scale bars = 10 μm. D HeLa cells stably expressing inducible RAD21-SFB (Y3A/K50E) were treated with Dox and transfected with control siRNA, RAD21 siRNA, and/or Wapl siRNA. Sixty hours post-transfection, cell lysates were immunoblotted for Wapl, RAD21, and α-Tubulin. α-Tubulin from a parallel gel served as a loading control. Blots are representative of three independent experiments. Uncropped blots are available in the Source Data file. E Cells were treated as in D. Sixty hours post-transfection, cells were treated with nocodazole for 3 h. Mitotic chromosome spreads were stained with DAPI and antibodies against ACA. The percentage of mitotic cells exhibiting predominantly separated or unseparated sister chromatids was quantified from at least 300 cells per condition across three independent experiments (n = 3). Statistical significance was calculated with two-tailed, unpaired Student’s t-test. Error bars represent SD. Raw data are available in the Source Data file. F Cells were treated and analyzed as in E. Representative images from three independent experiments are shown. Scale bars = 10 μm. G HeLa cells stably expressing SMC3-GFP (K1034E/1038E) were transfected with control siRNA, SMC3 siRNA, and/or Wapl siRNA. Seventy-two hours post-transfection, cell lysates were immunoblotted for GFP, SMC3, Wapl, and GAPDH. GAPDH from a parallel gel served as a loading control. Blots are representative of three independent experiments. Uncropped blots are available in the Source Data file. H Cells were transfected as in G. Seventy-two hours post-transfection, cells were treated with nocodazole for 3 h. Mitotic chromosome spreads were stained with DAPI and antibodies against ACA. The percentages of mitotic cells exhibiting fully separated, partially separated, or unseparated sister chromatids were quantified from at least 300 cells per condition across three independent experiments (n = 3). Statistical significance was calculated with two-tailed, unpaired Student’s t-test. Error bars represent SD. Raw data are available in the Source Data file. I Cells were treated and analyzed as in H. Representative images from three independent experiments are shown. Scale bars = 10 μm. J Working model illustrating the function of Sororin’s CTR as a direct structural lock for cohesin’s DNA-exit gate by engaging the RAD21-SMC3 interface. Mutations in Sororin-CTR that disrupt binding to the RAD21-SMC3 interface cause Wapl-dependent opening of the DNA-exit gate and loss of sister-chromatid cohesion.

We next examined whether Wapl depletion could similarly rescue cohesion defects caused by Sororin-binding-deficient mutations in cohesin subunits RAD21 and SMC3. Notably, Wapl knockdown prevented premature sister-chromatid separation in cells in which endogenous RAD21 and SMC3 were replaced by exogenous RAD21-YA/KE-SFB (Fig. 9D–F) and SMC3-KE/KE-GFP (Fig. 9G–I), respectively.

Together, these results demonstrate that the cohesion defects resulting from disruption of Sororin binding to the RAD21-SMC3 interface are dependent on Wapl activity, and that this interaction becomes dispensable for cohesion when Wapl is absent.

Discussion

The establishment and maintenance of sister-chromatid cohesion are fundamental for genome stability and depend on precise regulation of the cohesin complex²¹. Sororin has long been recognized as a key cohesion factor, with its function primarily attributed to antagonizing the release factor Wapl through competitive binding to Pds5^{26,27,37,54,60–62,64}. In this study, we redefine Sororin’s role by demonstrating that its extreme C-terminal region (CTR) acts as a direct structural lock for cohesin’s DNA-exit gate (Fig. 9J). This gate-locking activity operates synergistically with, yet independently of, Pds5 competition, providing a unified mechanism that stabilizes sister-chromatid cohesion while permitting its timely dissolution during mitosis.

We show that the 30-amino acid CTR of Sororin is both necessary and sufficient to promote cohesion. Remarkably, constitutive tethering of this minimal domain to centromeres fully restores cohesion in Sororin-depleted cells, demonstrating that centromere-localized gate engagement can bypass the requirement for full-length Sororin. By contrast, genome-wide tethering of Sororin-CTR phenocopies the effects of Wapl loss^48,50,87, causing defective chromosome condensation, impaired decatenation, and segregation errors. These findings underscore the struct need for spatial and temporal control of Sororin activity: persistent gate locking is required at centromeres to protect cohesion until anaphase, whereas timely removal of Sororin and cohesin from chromosome arms is essential for proper mitotic progression.

At the molecular level, we identify a conserved network of hydrophobic and electrostatic interactions through which Sororin-CTR directly engages the RAD21-SMC3 interface. AlphaFold3-guided structural modeling, supported by mutagenesis and binding assays, pinpoints key Sororin residues (W233, F241, F247, and E252) that contact RAD21 and SMC3. Disruption of these interactions abolishes Sororin binding to cohesin and results in cohesion loss. Notably, these defects are fully suppressed by Wapl co-depletion, demonstrating that Sororin’s essential function is to protect cohesin’s the exit gate from Wapl-mediated opening rather than to recruit accessory factors or alter cohesin ring conformation. This genetic epistasis firmly establishes Sororin as a direct antagonist of Wapl at the cohesin gate.

Our data further elucidate how Sororin activity is regulated across the cell cycle. Sororin-CTR binding to cohesin peaks during G2 phase, when cohesion is maintained, and is markedly reduced during mitosis, consistent with cohesion establishment following DNA replication and its dissolution before chromosome segregation. We additionally identify a phosphorylation-dependent regulatory switch that differentially modulates Sororin interactions. Aurora B-mediated phosphorylation at S145 selectively disrupts Sororin binding to Pds5 while preserving its engagement with the RAD21-SMC3 gate. This distinction provides a mechanism for cohesin release in early mitosis: on chromosome arms, phosphorylation permits Wapl rebinding and cohesin removal, whereas at centromeres, PP2A-mediated dephosphorylation primarily preserves Sororin-Pds5 interaction and sustained gate locking, thereby ensuring continued cohesion protection.

Our findings also clarify the relationship between Sororin and cohesin subunits SA2 and SMC3. Although previous work proposed a direct Sororin-SA2 interaction⁶⁵, we show that SA2 is dispensable for Sororin-CTR binding to the cohesin core and instead associates indirectly via RAD21. Furthermore, we demonstrate that Esco1/2-mediated acetylation of SMC3, a critical licensing step for Sororin recruitment^27,35–38, does not directly influence Sororin affinity for the RAD21-SMC3 interface. Additionally, we find that chromatin-tethered Sororin-CTR cannot bypass the requirement for Esco1/2 in safeguarding cohesion. These results support a sequential model in which SMC3 acetylation enables Sororin recruitment during S phase, followed by CTR-mediated gate locking to stabilize the cohesin ring. Separation of recruitment and locking functions likely provides a robust, multilayered mechanism for cohesion establishment.

The implications of these findings extend beyond sister-chromatid cohesion. Cohesin also plays a central role in organizing interphase genome architecture through DNA loop extrusion¹⁹. Persistent gate-locked cohesin on chromatin, as observed upon ectopic Sororin-CTR tethering, interferes with mitotic chromosome compaction and decatenation. This raises important questions regarding the functional state of gate-locked cohesin: does it retain loop extrusion activity, and how does its unscheduled persistence during mitosis disrupt higher-order chromatin organization? Future studies integrating live-cell imaging of cohesin dynamics with genome-wide chromatin conformation analyses will be essential to link cohesion maintenance to broader principles of genome organization^88,89.

Clinically, mutations in cohesin components and their regulators underlie cohesinopathies and are frequently observed in cancer¹⁵. The Sororin-CTR interaction surface on RAD21 and SMC3 identified here represents a potential hotspot for pathogenic variants that compromise gate locking and promote genomic instability. Consistent with this idea, we show that a cancer-associated Sororin-E252Q mutation weakens its interaction with cohesin. Systematic interrogation of mutations affecting this interface may reveal novel disease mechanisms and inform therapeutic strategies aimed at modulating cohesin dynamics, potentially sensitizing cancer cells to replication or mitotic stress.

In summary, we propose a revised dual-function model for Sororin in cohesion maintenance (Fig. 9J). Sororin acts not only as a competitive inhibitor of Wapl via Pds5 but also as a structural lock that directly occludes the RAD21-SMC3 exit gate. This gate-locking function is spatially restricted to centromeres and temporally regulated by phosphorylation, ensuring that cohesion is maintained precisely where and when it is required. By defining this direct molecular mechanism, our work resolves long-standing questions in cohesin biology and provides a refined framework for understanding, and potentially targeting, cohesin regulation in development and disease.

Methods

Cell lines, plasmids, siRNA, transfection, and drug treatments

HEK-293T cells, HeLa cells, and their derivatives were cultured in Dulbecco’s Modified Eagle Medium (DMEM), while RPE-1 cells and their derivatives were cultured in DMEM/F12. Both media were supplemented with 10% fetal bovine serum (Gibco) and 1% penicillin-streptomycin. All cells were maintained at 37 °C in a humidified atmosphere containing 5% CO₂. HeLa cells stably expressing Sororin-GFP (WT, WA/EK, or FA/FA), CENP-B-GFP, CENP-B-Sororin-CTR-GFP, H2B-GFP, H2B-Sororin-CTR-GFP, and SMC3-GFP (WT or KE/KE) were cultured in medium supplemented with 1.5 µg/ml blasticidin (Sigma). HeLa cells stably expressing Tet-on inducible RAD21-SFB (WT or YA/KE) were maintained in medium containing 0.5 mg/ml G418 (Calbiochem). Pools of RPE-1 cells expressing Tet-on inducible Sororin-SFB (WT, WA/EK, or FA/FA) were generated by lentiviral infection and selected in medium containing 0.5 mg/ml G418.

The construct for pGEX-6P-Aurora B (60-361)-rbs-INCENP (790-856) was a gift from Dr. Andrea Musacchio (Max Planck Institute of Molecular Physiology). The Myc-Pds5B construct was kindly provided by Dr. Hongtao Yu (Westlake University). To generate the Flag-Sororin, Lenti-Sororin-SFB, and Lenti-RAD21-SFB constructs, the corresponding cDNAs in the entry vectors were transferred into Gateway-compatible destination vectors harboring an N-terminal Flag-tag and a C-terminal SFB-tag, respectively. Myc-SA2 was constructed similarly using a vector harboring an N-terminal Myc-tag. To generate Myc-Sororin, the cDNA encoding Sororin was amplified by PCR and subcloned into an expression vector containing a Myc epitope tag. The pBos-H2B-Sororin-GFP plasmid series was constructed by subcloning PCR-amplified fragments encoding full-length Sororin or truncated variants (1-149, 1-222, 223-252) into pBos-H2B-GFP (BD PharMingen) using the BamHI restriction site. The pBos-CENP-B-GFP plasmid was constructed by replacing the H2B coding sequence in pBos-H2B-GFP with KpnI/BamHI-digested PCR fragments encoding the centromere-targeting domain of CENP-B (residues 1-163). The pBos-CB-Sororin (223-252)-GFP construct was generated by subcloning the corresponding Sororin fragment into pBos-CENP-B-GFP. The pBos-Sororin-GFP plasmid series was constructed by replacing the H2B coding sequence in pBos-H2B-GFP with PCR-amplified sequences encoding full-length Sororin or its truncated variants (1-149, 150-252, 1-222, 223-252). The pBos-SMC1-GFP, pBos-SMC3-GFP, and pBos-RAD21-GFP constructs were generated by replacing the H2B coding sequence with PCR-amplified coding sequences of SMC1, SMC3, and RAD21, respectively. Plasmids encoding GST-fused Sororin (Full-length or truncations) were generated by subcloning the corresponding PCR fragments into pGEX-4T1 (GE Healthcare). The MBP-Pds5B (1-300) construct was generated by cloning the PCR products into pMAL-C2E (New England Biolabs). For the pEF6-RAD21-Myc construct series, PCR-amplified fragments encoding full-length RAD21 or its variants (1-281, 1-101 + 123-281, 1-103) were cloned into the pEF6-Myc vector. The pEF-Sororin-Flag and pEF-SMC3-HD-Flag constructs were generated by cloning PCR-amplified Sororin and SMC3-HD fragments (residues 1-253 and 959-1217 connected by a GGGGS linker) into pEF-Flag, respectively. To generate the pET45-RAD21-N103-2xStrep-rbs-SMC3-HD-Flag construct, the PCR-amplified RAD21-N103-2xStrep fragment, rbs fragment, and SMC3-HD-Flag fragment were subcloned into pET45 using the MultiF Seamless Assembly Mix Kit (ABclonal Biotechnology).

All point mutations were introduced using either the QuikChange II XL Site-Directed Mutagenesis Kit (Agilent Technologies) or the MultiF Seamless Assembly Mix Kit (ABclonal Biotechnology). All constructs were verified by DNA sequencing to confirm the presence of the desired mutations and the absence of unwanted sequence alterations.

siRNA duplexes targeting Sororin, Esco1, Esco2, Sgo1, Wapl, RAD21, and SMC3 were purchased from Integrated DNA Technologies (IDT) or RiboBio: siSororin targeting the 3’UTR (5’-CUCGGAAAGUUUCCUCGCGUAdTdT-3’), siEsco1 (5’-AGAAGGUCGUUACGCAGUAdTdT-3’), siEsco2 (5’-GAAAGAACGUGUAGUAGCAdTdT-3’), siSgo1 (5’-GAGGGGACCCUUUUACAGAdTdT-3’), siWapl (5’-CGGACUACCCUUAGCACAAdTdT-3’), siRAD21 (5’-AUACCUUCUUGCAGACUGUdTdT-3’), siRAD21#2 (5’-GCACUACUACUUCUAACCUdTdT-3’), siSMC3 (5’-GGAGGGCAGUCAGUCUCAAGAUGAAdTdT-3’). siRNA transfections were carried out twice at a 24-h intervals using Oligofectamine (Invitrogen) or Lipofectamine RNAiMAX (Thermo Fisher Scientific). For SMC3 siRNA transfection, cells at approximately 40% confluence were transfected once using Lipofectamine RNAiMAX. Plasmid transfections were performed using FuGENE 6 (Promega) or Lipofectamine 2000 (Thermo Fisher Scientific).

Cell-cycle synchronization was performed as follows: cells were synchronized at the G1/S boundary using a double-thymidine (2 mM, Sigma) block; in G2 phase using either a single-thymidine block or a double-thymidine block followed by an 8-h release; in late G2 phase using RO-3306 (9 μM; Selleckchem) for 18 h; and in prometaphase using nocodazole (100 ~ 300 ng/ml; Selleckchem) for 15–18 h or STLC (5 μM, Tocris Bioscience) for 3 h. Mitotic cells were collected by mechanical detachment (“shake-off”). The kinase inhibitors used in this study were as follows: CDK1 (Roscovitine, Selleckchem; 100 μM), Aurora B kinase (Hesperadin, Selleckchem; 25–100 nM), and Plk1 (BI 2536, Selleckchem; 100 nM). The proteasome inhibitor MG132 (10 μM; Selleckchem) was also used.

Antibodies

The following rabbit polyclonal antibodies were used: GFP (A11122, Invitrogen), GAPDH (14C10, Cell Signaling Technology/CST), RAD21 (ab992, Abcam), SMC1 (A300-055A, Bethyl Laboratories), SMC3 (A300-060A, Bethyl Laboratories), Pds5A (A300-089A, Bethyl Laboratories), Pds5B (A300-537A and A300-538A, Bethyl Laboratories), SA1 (HPA035015, Sigma), SA2 (#5882, CST), Wapl (A300-268A, Bethyl Laboratories), Esco1 (29821-1-AP, Proteintech), Sororin (ab192237, Abcam), Sororin-pS145 (generated by immunization with the synthetic peptide LEMSKKVRRpSYSRLETLGSASTS, where pS indicates phosphoserine), and Flag-tag (A01868, GenScript). Mouse monoclonal antibodies included: Aurora B (611083, BD Transduction Laboratories), α-Tubulin (AC012, ABclonal), Myc-tag (clone 4A6, 05-724, Millipore), Flag-tag (clone M2, F-3165, Sigma), SA2 (sc-81852, Santa Cruz Biotechnology), GFP (M20004, Abmart), PICH (H00054821-B01P, Abnova), Wapl (M221-3, MBL), Strep II-Tag (AE066, ABclonal), MBP (AE016, ABclonal), and GST-tag (M20007, Abmart). The anti-centromere antibody (ACA) was obtained from ImmunoVision (HCT-0100). Guinea pig polyclonal antibodies against CENP-C were purchased from MBL (PD030).

Secondary antibodies for immunoblotting included Streptavidin-HRP (434323, Invitrogen), goat anti-rabbit IgG-HRP (7074, CST), and horse anti-mouse IgG-HRP (7076, CST). Secondary antibodies for immunofluorescence staining included: donkey anti-rabbit IgG conjugated to Alexa Fluor 488 or Cy3 (711-545-152 or 711-166-152, Jackson ImmunoResearch); anti-mouse IgG conjugated to Alexa Fluor 488, 546 (A-21202 or A-11030, Invitrogen) or Cy5 (715-175-150, Jackson ImmunoResearch); goat anti-guinea pig IgG conjugated to Alexa Fluor 647 (A-21450, Invitrogen); and anti-human IgG conjugated to Alexa Fluor 647 (709-605-149, Jackson ImmunoResearch).

Lentivirus packaging and infection

The Tet-On inducible SFB-tagged lentiviral vector and packaging plasmids (pMD2G and pSPAX2) were provided by Dr. Jun Huang (Life Sciences Institute, Zhejiang University). Virus-containing supernatant was collected 48 h after the co-transfection of the lentiviral vector and packaging plasmids (pMD2G and pSPAX2) into HEK-293T cells. HeLa cells were infected with the viral supernatants in the presence of 8 μg/ml polybrene (Sigma), and stable pools were selected using medium containing 0.5 mg/ml G418. For the experiments described in this study, expression of the indicated genes in the stable pools was induced by the addition of 1 μg/ml doxycycline (Selleckchem) for 24 h.

CRISPR/Cas9-mediated editing of the SA2 gene STAG2 in human cells

Single-guide RNA (sgRNA) for human STAG2 (encoding SA2) was designed, ordered as complementary oligonucleotides, annealed, and cloned into the dual Cas9 and sgRNA expression vector pX330 (Addgene, #42230) using BbsI restriction sites. The resulting plasmid was transfected into HeLa cells using FuGENE 6 according to the manufacturer’s protocol. After 48 h incubation, cells were seeded at low density to establish clonal cell lines following brief selection with 1 μg/ml puromycin for 2–3 days. To obtain SA2-KO cells, clones showing undetectable SA2, as confirmed by immunostaining and immunoblotting, were isolated. SA2 KO cells were generated using an sgRNA targeting the sequence 5’-GGCCCAGCAGAAAAGGGCAA-3’.

Fluorescence microscopy, chromosome spreads, and chromosome length measurement

HeLa cells were typically fixed in 2% paraformaldehyde (PFA) in phosphate-buffered saline (PBS) for 10 min at room temperature, followed by permeabilization with 0.5% Triton X-100 in PBS for 5 min. Cells were subsequently blocked with 3% bovine serum albumin (BSA) in PBS and incubated with primary antibodies for 1–2 h, followed by secondary antibody incubation for 1 h, all incubations were performed at room temperature in 3% BSA/PBS. Nuclear DNA was counterstained with DAPI for 5 min.

For chromosome spread preparation, HeLa cells were treated with 100-300 ng/ml nocodazole for 3 h. Mitotic cells were collected by selective detachment and incubated in hypotonic buffer (75 mM KCl or 0.25×PBS) for 10–15 min at room temperature. Cells were then cytocentrifuged onto glass coverslips using a Cytospin at 260 × g for 5 min. Following attachment, chromosome spreads were fixed in 2% PFA in PBS for 15–20 min, permeabilized with 0.5% Triton X-100 in PBS for 10 min, blocked with 3% BSA in PBS, and incubated with primary antibodies for 2 h at room temperature. For SMC1 immunofluorescence staining, cells were pre-extracted with 0.5% Triton in PHEM buffer (60 mM Pipes, 25 mM HEPES, 10 mM EGTA, and 2 mM MgCl₂, pH 6.9) for 5 min, then fixed with 4% PFA in PHEM for 20 minutes.

Fluorescence microscopy was performed at room temperature using a Nikon ECLIPSE Ni microscope equipped with a Plan Apo Fluor 60×oil immersion objective (NA 1.4) and a Clara CCD camera (Andor Technology). To measure the length of mitotic chromosomes, images were visualized using NIS-Elements BR software (Nikon). Chromosome lengths were measured along the outline of the center of the chromosomes using a point-by-point tracing tool. Cells were randomly selected for each condition, and all chromosomes within the selected cells were measured.

Time-lapse live cell imaging

Time-lapse live cell imaging was carried out using the GE DV Elite Applied Precision DeltaVision system (GE Healthcare) equipped with Olympus oil objectives of 60×(NA 1.42) Plan Apo N or 40×(NA 1.35) UApo/340 oil objective, an API Custom Scientific complementary metal-oxide semiconductor camera, and Resolve3D softWoRx imaging software. Cells were plated in four-chamber glass-bottomed 35-mm dishes (Cellvis) coated with poly-D-lysine, and imaged in a climate-controlled and humidified environment (37 °C and 5% CO₂). Chromosomes were labeled with H2B-GFP. Images were captured every 3 minutes and processed using Adobe Photoshop and Adobe Illustrator.

Protein expression and purification

Plasmids encoding GST-, Strep-, or MBP-fused proteins were transformed into E. coli BL21 (DE3) competent cells. Transformed cells were cultured in LB medium supplemented with the appropriate antibiotic at 37 °C until the optical density at 600 nm (OD₆₀₀) reached 0.6–0.8. Protein expression was induced by adding 0.4 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) and incubating at 16 °C for 18 h.

For GST fusion proteins, cells were lysed by sonication in Buffer A (50 mM Tris-HCl, pH 8.0, 300 mM NaCl, 1 mM EDTA, 0.5% Triton X-100, 1 mM DTT, 1 mM phenylmethylsulfonyl fluoride/PMSF)). For MBP fusion proteins, lysis was performed in Buffer B (50 mM Tris-HCl, pH 7.5, 300 mM NaCl, 1 mM EDTA, 0.5% Triton X-100). Lysates were clarified by centrifugation and incubated with Glutathione Sepharose 4B beads (GE Healthcare) for GST fusion proteins or Amylose Resin (New England Biolabs) for MBP fusion proteins. After washing with the respective lysis buffer, proteins were eluted using 100 mM reduced glutathione for GST fusions or 10 mM maltose for MBP fusions.

For Strep-fused protein purification, cells were resuspended in lysis buffer (25 mM Tris-HCl, pH 7.5, 300 mM NaCl, 1 mM DTT, 1% Triton X-100, 1 mM EDTA, and 1 mM PMSF). Following sonication and a freeze-thaw cycle, the lysate was clarified by centrifugation, and the supernatant was applied to a Strep-Tactin Superflow affinity column (IBA Lifesciences). The resin was washed four times with lysis buffer, and bound proteins were eluted with lysis buffer supplemented with 10 mM desthiobiotin.

In vitro kinase reaction

For the phosphorylation of Flag-Sororin (Fig. 5I), asynchronous HEK-293T cells transiently expressing Flag-Sororin (WT or S145A) were lysed in buffer containing 50 mM Tris-HCl, pH 7.5, 500 mM NaCl, 0.1% NP-40, and 10 mM MgCl₂, supplemented with protease and phosphatase inhibitors as in P150 buffer. The anti-Flag immunoprecipitates were then subjected to phosphorylation by 0.2 μg GST-Aurora B/IN-box. For the phosphorylation of Sororin peptide (Fig. 5K), 0.5 μg of substrates was incubated with 0.2 μg of recombinant GST-Aurora B/IN-box. The kinase reactions were conducted in 50 μl kinase reaction buffer (50 mM Tris-HCl pH 8.0, 150 mM NaCl, 10 mM MgCl₂) with 0.2 mM ATP for 20 min at 30 °C. The reactions were terminated by adding SDS sample buffer, boiled, and analyzed by immunoblotting.

Chromatin fractionation

Mitotic cells collected after 18 h of nocodazole treatment were lysed in fractionation buffer (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 5 mM MgCl₂, 10% glycerol, 0.2% NP-40) supplemented with 1 mM DTT, protease inhibitor cocktail (P8340, Sigma), 10 mM NaF, and 20 mM β-glycerophosphate. After centrifugation at 2000 g for 5 min, the supernatant was collected and used for immunoblotting and immunoprecipitation. The chromatin pellets were washed three times with the same buffer and subsequently analyzed by immunoblotting and immunoprecipitation.

Pull-down, immunoprecipitation, and immunoblotting

For GST- or Strep-fused protein pull-down assays with cell lysates, HeLa cells were lysed in P150 buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1% Triton X-100, 10 mM MgCl₂, 5 mM EDTA) supplemented with 1 mM dithiothreitol (DTT), protease inhibitor cocktail, 1 mM PMSF, 0.1 μM okadaic acid (Calbiochem), 10 mM sodium fluoride, 20 mM β-glycerophosphate, and Benzonase (GenScript). Insoluble material was removed by high-speed centrifugation, and the lysates were precleared with either glutathione Sepharose 4B beads (GE Healthcare) or StrepTactin 4FF beads (Smart Life Sciences). The precleared lysates were then incubated with bead-immobilized GST- or Strep-fused proteins for 4 h at 4 °C. Beads were washed 4 times with lysis buffer, boiled in SDS sample buffer, and subjected to immunoblotting.

For Sororin (136-158) peptide pull-down of recombinant MBP-Pds5B-N300, 1 μg of peptides were incubated with 4 μg of MBP-Pds5B-N300 in 1 ml of binding buffer (50 mM Tris, pH 7.5, 300 mM NaCl, 0.5% Triton X-100) for 3 h, followed by 1-h incubation with NeutrAvidin Plus Ultralink Resin (Thermo Fisher Scientific). The resin was then washed 4 times with binding buffer. The Sororin peptide of residues, corresponding to residues 136-158 (either wild-type or the pS145 form) and followed by a Lys-biotin moiety, were synthesized by GL Biochem (Shanghai).

For co-immunoprecipitation, cells were lysed in P150 buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1% Triton X-100, 10 mM MgCl₂, 5 mM EDTA) supplemented with the same additives as above. Insoluble material was removed by high-speed centrifugation, and the lysates were pre-cleared with rProtein A/G beads (Smart Life Sciences, Cat. No. SA032100). The precleared lysates were incubated for 4 h at 4 °C with Anti-GFP Affinity beads (Smart Life Sciences, SA070005), Anti-Flag M2 Affinity Gel (Sigma, A2220), Anti-Myc Affinity Gel (GNI, GNI4510-MC), or control IgG. After washing with P150 buffer, bound proteins were eluted by boiling in SDS sample buffer for immunoblotting.

Mass spectrometry

For the data presented in Supplementary Fig. 5A and Supplementary Data 1, immunoprecipitated Sororin-GFP were separated by SDS-PAGE (n = 1). The gels were minimally stained with Coomassie brilliant blue. Protein bands were excised according to the molecular weight marker range. The excised protein bands were cut into small pieces, then reduced and alkylated in 10 mM DTT and 55 mM IAA, respectively. For digestion, 0.5 µg of sequencing-grade modified trypsin (Promega, USA) was added, followed by incubation overnight at 37 °C. Prior to mass spectrometry analysis, the peptide preparations were desalted using StageTip and reconstituted with 0.1% formic acid (FA).

Mass spectrometry analysis was performed using a Thermo Scientific Q Exactive HF-X Hybrid Quadrupole-Orbitrap mass spectrometer instrument coupled with an Easy-nLC 1200 system. Mobile phase A and B were water and 80% acetonitrile, respectively, each containing 0.1% formic acid. Protein digests were loaded directly onto an analytical column (75 µm×15 cm, 1.9 µm C18, 1 µm tip) at a flow rate of 450 nl/min. Samples were separated using a linear gradient of 6-8% B for 2 min, 8-27% B for 40 min, 27–40% B for 12 min, 40-90% B for 3 min, 90% B for 3 min.MS/MS data were collected in a data-dependent manner using a top-20 method with a full MS mass range from 400 to 1400 m/z, 60,000 resolution, and an AGC target of 3 × 10⁶. MS2 scans were triggered when an ion intensity threshold of 4.5 × 10⁵ was reached. A dynamic exclusion time of 30 s was used. Ions with charge state 1, 6, 8, and > 8 were excluded. The resulting raw data were uploaded into MaxQuant (version 2.0.3.0) and searched against the UniProtKB Sororin (Q96FF9) sequence database. Trypsin/P was specified as the cleavage enzyme, allowing up to two missing cleavages. The mass tolerance for precursor ions was set to 20 parts per million in first search and 5 parts per million in main search, and the mass tolerance for fragment ions was set to 0.02 Da. Carbamidomethylation (C, +57.0215 Da) was used as the fixed modification. Phospho (STY, +79.9663 Da) was included as variable modifications. The false discovery rate (FDR) was adjusted to <1%, with a minimum score for modified peptides set to >20.

Quantitative RT-PCR analysis

Total cellular RNA was extracted with TRIzol (Thermo Fisher Scientific). The homogenate was incubated on ice for 10 min and centrifuged at 13,000 × g for 10 min. The supernatant was mixed with chloroform, vigorously shake, and incubated at room temperature. After centrifugation, the upper aqueous phase was mixed with an equal volume of isopropanol, followed by centrifuged. The RNA pellet was washed twice with cold 75% ethanol and dissolved in DEPC-treated water. Next, 1000 ng of RNA was reverse-transcribed using PrimeScript RT Master Mix (Takara, RR047A). Quantitative PCR was carried out using SYBR qPCR SuperMix (Vazyme, Q711-03) according to the manufacturer’s instructions. The Esco2 primer sequences were (5’- ATCCAATCTTCAGTGCATCTTCA-3’ and 5’- TGTTTCTGACCAGCGTCGAT-3’).

Structure prediction

The structure of a complex formed between the Sororin C-terminal fragment (residues 191-252), the SMC3-HD (residues 1-253 and 959-1217 connected by a GGGGS linker), and the RAD21 N-terminal fragment (residues 1-110) was predicted using AlphaFold3. All predictions were performed via the public AlphaFold3 web server (https://www.alphafoldserver.com) using the default settings and without applying any constraints. A representative complex was visualization in PyMOL v3.1.3.

Sequence alignment

Multiple sequence alignments of Sororin, RAD21, and SMC3 were performed using UniProt Align (https://www.uniprot.org/), and the results were visualized using ESPript 3.0 (https://espript.ibcp.fr/ESPript/ESPript/).

Statistics

Statistical analyses were conducted using a two-tailed, unpaired Student’s t-test in GraphPad Prism 6. A P-value of less than 0.05 was considered statistically significant. All representative images, including Western blots and micrographs, were obtained from independent biological replicates that yielded consistent results.

Reporting summary

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

Author contributions

F.W. designed research; Q.C., X.Y., M.S., X.Z., S.Z., and H.Y. performed research; Q.C., X.Y., M.S., H.Y., and F.W. analyzed data; W.L. contributed key reagents; Q.C., X.Y., H.Y., and F.W. acquired funding; H.Y. and F.W. wrote the manuscript with the input from all authors.

Peer review

Peer review information

Nature Communications thanks Huqiang Lou, Debananda Pati and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

The mass spectrometry proteomic data generated in this study have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD074092. The SMC3 (residues 1-253 and 959-1217)-RAD21 (residues 1-110)-Sororin-CTR (residues 191-252) model is accessible at ModelArchive [ma-285gc; https://modelarchive.org/doi/10.5452/ma-285gc]. All data supporting the findings of this study are available within the paper and its Supplementary Information. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-70484-2.