Embryonic stem cells (ESCs), which have unique properties such as pluripotency and unlimited self-renewal, exhibit unique expression patterns of genes related to chromosome topology, chromatin modification, and genomic integrity (1, 2). Cohesin is a ring-shaped protein complex composed of several subunits, structural maintenance of chromosome (SMC) family proteins, α-kleisin proteins, stromal antigen (STAG) proteins, and accessory factors (3-11). Furthermore, cohesin has two different combinations of subunits depending on the cell division stage: mitosis-specific subunits (RAD21) and meiosis-specific subunits (REC8 and RAD21L). During mitosis, the cohesin complex includes SMC1α, SMC3, RAD21, STAG1, and STAG2. Unlike the mitotic cohesin complex, SMC1β, which is the homolog of SMC1α, REC8, RAD21L, and STAG3, mainly functions in the chromosomes during meiosis (Fig. 1A) (12). The cohesin complex with REC8 appears on the chromosome during the prophase in meiosis I, and the complex remains until meiosis II, whereas the cohesin complex including RAD21L localizes along chromosome arms from the leptotene to the middle pachytene stage in prophase I and disappears after the late pachytene stage (12-14). RAD21 is also expressed during meiotic cell division (15, 16). The kleisin RAD21 is detected before early prophase I and disappears in early prophase I. Afterward, it reappears on chromosomes after the mid-late pachytene stage during the prophase stage of meiosis I (15, 16).

Our study revealed the role of the α-kleisin factors of the cohesin complex, REC8, and RAD21 in both chromatin modification and maintenance of genome integrity. Unlike in differentiated cells such as mouse embryonic fibroblast (MEF) cells, mitotic cohesin complex subunits and meiotic cohesin complex subunits are expressed in pluripotent states (17). Our findings demonstrated that the α-kleisin factors are constitutively expressed to function as safeguards of genome integrity throughout the entire cell cycle in ESCs and that the involvement of REC8 and RAD21 is important to modulate chromatin remodeling in the ESC cycle.

The α-kleisin subunits RAD21 and REC8 are abundantly expressed in mouse ESCs throughout the cell cycle

To explore the expression dynamics of α-kleisin factors during the cell cycle, we synchronized ESCs and MEFs at the G1/S phase via the double thymidine block and released them from the G1/S phase (Fig. 1B). Our findings indicated that the α-kleisin factors REC8 and RAD21 were constitutively expressed regardless of the cell cycle phase of the ESCs (Fig. 1C). However, in contrast to ESCs, MEFs expressed RAD21, not REC8, during cell cycle progression. Particularly, cohesin builds cohesion during DNA replication because sister chromatids are generated by replication during the S phase (18). Therefore, we further focused on the expression levels of cohesin factors during the S phase (i.e., 2.5 h after the cell cycle arrest is released by the double thymidine block). The expression of RAD21 in ESCs at the protein and RNA levels exhibited 1.71- and 4.57-fold differences relative to MEFs, respectively (Fig. 1D). Additionally, REC8 was expressed at the protein and RNA levels in ESCs but not in MEFs (Fig. 1D-F) (17). Furthermore, the RNA expression levels of the factors in ESCs were different from those in MEFs. RNA sequencing and real-time PCR (RT-PCR) analyses revealed that the α-kleisin factors were expressed in ESCs but not in MEFs (Fig. 1E, F). Additionally, the RNA expression levels of diverse factors involved in meiotic or mitotic cohesin subunits in ESCs were higher than those in MEFs (Fig. 1E, F).

In the Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis, we demonstrated that REC8 or RAD21 depletion could differentially regulate numerous genes in ESCs (Supplementary Fig. 1, 2). Therefore, we classified KEGG pathways according to the number of genes whose expression levels were differentially regulated by REC8 or RAD21 depletion. Furthermore, the depletion of each cohesin factor enables the differentiation of ESCs into specific cell-type lineages (19).

Depletion of cohesin decreases DNA repair efficiency

In ESCs, homologous recombination (HR) is a major DNA repair system, and the factors involved in the HR-mediated repair system are highly expressed compared with those in MEFs, which result from the prolonged S phase and rapid cell cycle progression (17, 20, 21). Furthermore, we found that the replication protein A (RPA) focal number in ESCs was higher than that in MEFs, and the cohesin intensity of ESCs was higher than that of MEFs (Fig. 2A, B). When cohesin factors were depleted, there was an increase in the number of RPA foci, which bind to single-stranded DNA to prevent the strand from forming a secondary structure (Fig. 2C, D, I and Supplementary Fig. 3A-C). We observed that the expression of each factor decreased under the depleted condition. Particularly, REC8 and RAD21 expression decreased by 55%-70% and 70%-80%, respectively (Fig. 2C, D, G, H and Supplementary Fig. 3A, B, D, E). Furthermore, the average number of RPA foci in REC8-depleted ESCs was approximately 27.26, whereas the average number of RPA foci in RAD21-depleted ESCs was approximately 36. The RPA focal number under the REC8 depletion condition was 1.7-2 times that of the control (Fig. 2C, G and Supplementary Fig. 3A, C). Accordingly, the number of RPA foci increased dramatically in the RAD21-depleted ESCs by as much as 2.33-2.8-fold compared with that of the control (Fig. 2D, I and Supplementary Fig. 3B, C). Likewise, siRAD21-treated MEF cells exhibited higher RPA focal numbers than the siControl cells. The RPA focal number was significantly increased to 37 in RAD21-depleted cells (Fig. 2E, F, J). Conversely, the RPA focal number in REC8-depleted cells was 1.2, which was similar to the siControl cells (RPA foci in siControl cell = 1.36).

To confirm the role of the cohesin complex in DNA damage repair and replication stress, we depleted either RAD21 or REC8 in ESCs treated with hydroxyurea (HU). This impedes DNA replication by inhibiting ribonucleotide reductase (RNR), which converts ribonucleotides into deoxyribonucleotides (Fig. 2K) (22). Cohesion occurs in DNA replication to hold sister chromatids (17, 18, 23-26). Thus, we used HU to induce the stalling of DNA replication and impede cell cycle progression to arrest at the S phase. The expression of the kleisin subunits decreased, which resulted from the depletion of the factors after treating the cells with siRNA against REC8 or RAD21 (Supplementary Fig. 3F). In ESCs, RPA focus formation significantly increased upon REC8 depletion with HU and RAD21 depletion with HU (Fig. 2L-N). When the cells were treated with HU only, the RPA focal number was approximately 66.97. Conversely, RPA focal formation increased when both kleisin subunits were depleted (approximately 80.5 foci in REC8-depleted cells and 92.9 foci in RAD21-depleted cells).

Depletion of cohesin decreases cell viability by reducing HR-mediated DNA repair efficiency

To confirm cell viability under cohesin depletion and ectopic expression, cell viability was analyzed with thiazole orange (TO) and propidium iodide (PI) staining. Three siRNA candidates against REC8 or RAD21 were applied to validate their depletion efficiency through protein analysis and cell viability test (Supplementary Fig. 4, siREC8 #1 and siRAD21 #1 were used in this study). We then transfected the cells with pDR-GFP and pCBA-SceI to confirm the efficiency of homologous recombination (HR), which is considered an error-free DNA repair system. The HR efficiency of the siControl cells was 10.1%, whereas those of the REC8- and RAD21-depleted stem cells were 4.42% and 4.28%, respectively (Supplementary Fig. 5A). Furthermore, cell viability was not affected by the depletion or ectopic expression of REC8 in MEFs (Supplementary Fig. 5B-E). Live cells accounted for 91.9% of all cells under the control condition, whereas viability decreased in the REC8- or RAD21-depleted ESCs (Fig. 3A-D). Specifically, when the REC8 level decreased, the live ESC population decreased to approximately 84% of the total cells; when RAD21 was depleted, the normal cell population was 74.4%. Additionally, the live population of HU-treated cells was approximately 67.5% (Fig. 3A-D). However, the ectopic expression of RAD21 or REC8 did not significantly affect cell viability compared with that of depletion because the expression levels of other cohesin subunits likely did not change. We further analyzed RNA sequencing data from REC8- and RAD21-depleted ESCs to confirm the expression levels of DNA repair genes. In REC8-depleted ESCs, 48 genes related to DNA repair were upregulated, whereas 116 were downregulated. In RAD21-depleted ESCs, 73 genes were upregulated, and 154 genes were downregulated. REC8 and RAD21-depleted cells shared 20 upregulated genes and 76 downregulated genes (Fig. 3E, F). Next, the significantly downregulated genes under both conditions were assigned to different functional classifications. The top 10 classifications were then ranked according to the number of downregulated genes. Genes associated with DNA repair, DNA damage stimulus, double-strand break repair, DNA replication, and DNA replication were downregulated in REC8- and RAD21-depleted ESCs (Fig. 3G). Interestingly, many genes involved in DNA repair interacted with RAD21 and REC8. Among the genes that interacted with these α-kleisin factors, DNA repair genes bound more strongly (Supplementary Fig. 6).

The cohesin complex is involved in DNA methylation

In embryonic stem cells, DNA methylation plays an essential role in cell development and differentiation (27, 28). Patients with cohesinopathies (i.e., diseases caused by mutations in the cohesin complex) such as CdLS (Cornelia de Lange Syndrome) and Roberts syndrome exhibit abnormal DNA methylation patterns, resulting in changes in transcriptional regulation (29). Therefore, we investigated the effects of cohesin depletion on the dynamics of DNA methylation. Cells treated with siRNA against cohesin were stained with anti-5-methylated cytosine (5mC) antibodies to investigate the relationship between cohesin and DNA methylation in ESCs. DNA methyltransferases, namely, DNMT1, DNMT3a, and DNMT3b, play an important role in ESC differentiation and cell fate (30-33). The expression level of DNMT3b, the main de novo DNA methyltransferase, was increased in REC8- or RAD21-depleted ESCs (Fig. 4A, B). The average fluorescence intensity of 5mC in the control ESCs was approximately 41.4 ± 18. Under REC8 or RAD21 depletion, the DNA methylation level increased compared with that of the control. The average intensity in REC8-depleted ESCs was approximately 78.2 ± 10.3, whereas the average intensity in RAD21-depleted ESCs was approximately 90.9 ± 19.3 (Fig. 4C, D).

Differentiation was induced with RA in ESCs, and the DNA methylation level was analyzed by staining the cells with anti-5mC antibodies. The intensity of the 5mC signal increased as early as 24 h after the cells were treated with RA. As the ESCs lost their pluripotency, the fluorescence intensity of 5mC gradually decreased accordingly (Supplementary Fig. 7). However, 96 h after the RA treatment, the DNA methylation level was similar to that of differentiated cells (MEFs), suggesting that the DNA methylation pattern could be changed dynamically during ESC differentiation.

Unlike the differentiated cells, ESCs have a unique gene expression landscape that can be attributed to the cell cycle pattern (i.e., prolonged S phase), pluripotency, and self-renewal abilities (17, 20, 21). Interestingly, ESCs express REC8 and RAD21, which are involved in chromosome segregation, cohesion, and topology (17). Furthermore, mutation of cohesin complex leads to tumorgenesis (34). Our study provided novel insights into the involvement of the α-kleisin factors RAD21 and REC8 in the maintenance of genomic integrity and chromatin methylation to manage the self-renewal of ESCs and enhance the efficiency of ESC differentiation.

Our findings demonstrated that REC8 and RAD21 were expressed in ESCs regardless of the cell cycle phase (Fig. 1C). Furthermore, ESCs constitutively express HR proteins and facilitate HR-mediated DNA repair during the cell cycle. Therefore, under the DNA damage condition, RPA foci formation increased, and the expression of the kleisin factors increased significantly (Fig. 2C-N), suggesting that these two factors are closely related to DNA damage repair. Therefore, the depletion of cohesin interfered with HR-mediated DNA repair, and the proportion of apoptotic cells increased in kleisin-depleted ESCs. Cohesin depletion in ESCs led to the downregulation of diverse genes involved in DNA repair, DNA damage responses, and genomic integrity compared with ESCs with the normal cohesin complex (Fig. 3E-G). Moreover, α-kleisin was strongly correlated with the interaction with DNA repair genes (Supplementary Fig. 6). Our results suggested that cohesin could support the maintenance of genome integrity (Fig. 4E).

We further demonstrated that the DNA methylation level was lower in differentiated cells than in undifferentiated ESCs. Furthermore, the abnormal establishment of the cohesin complex resulting from REC8 or RAD21 depletion causes hypermethylation by increasing the expression level of the de novo methyltransferase DNMT3b (Fig. 4A-E). During cell differentiation, the DNA methylation pattern was dynamically changed to control gene expression by regulating gene expression to restrict the lineage (29, 30). Thus, we propose that the depletion of cohesin factors led to DNA hypermethylation, which presumably changed gene expression and chromosome structure.

Materials and methods are available in the supplemental material.

We thank Soojin Lee of Chung-Ang University for helping in the preparation of the REC8 plasmid. This work was supported by grants to K.P.K. from the National Research Foundation of Korea, funded by the Ministry of Science, ICT, & Future Planning (No. 2020R1A2C2011887).

The authors have no conflicting interests.