Checkpoint kinase 1–induced phosphorylation of O-linked β-N-acetylglucosamine transferase regulates the intermediate filament network during cytokinesis

X Zhe Li (李喆), X Xueyan Li (李雪燕), Shanshan Nai (能姗姗), Qizhi Geng (耿奇志), X Ji Liao (廖蓟), X Xingzhi Xu (许兴智), and X Jing Li (李静) From the Beijing Key Laboratory of DNA Damage Response and College of Life Sciences, Capital Normal University, Beijing 100048, China and the Guangdong Key Laboratory of Genome Stability & Disease Prevention, Shenzhen University School of Medicine, Shenzhen, Guangdong 518060, China

O-Linked ␤-N-acetylglucosamine (O-GlcNAc) 4 -transferase (OGT) is the sole enzyme in humans that catalyzes the adding of the O-GlcNAc moiety to the Ser/Thr residues of nuclear and cytoplasmic proteins (1). The reverse reaction is catalyzed by O-GlcNAcase (OGA). The OGT/OGA pair accounts for all the adding and reversible removal of the O-GlcNAc group in human proteins. Despite the significant roles of O-GlcNAc in protein modification and signal transduction, its substrates have just started to unfold. Extensive crosstalk between O-GlcNAcylation and phosphorylation has long been documented; O-GlcNAc either vies with phosphate groups for the same or adjoining Ser/Thr sites, or in some other cases, promotes it (1,2).
Recently, a vital role of O-GlcNAc in cytokinesis has emerged. It has been well known that cytokinesis is an intricate process fine-tuned by a plethora of kinases (3). Now through a proteomic approach, 141 proteins involved in spindle assembly and cytokinesis have been newly shown to be O-GlcNAcylated (4). OGT localizes to the midbody (4,5), an organelle involved in cytokinesis. Moreover, over-expression of OGT induces the polyploidy phenotype, indicative of cytokinesis defects (4). Furthermore, OGT has also been shown to modify vimentin (5), one of the intermediate filament proteins expressed in mesenchymal, cultured, and tumor cells (6).
Phosphorylation of vimentin is regulated at multiple levels during the cell cycle. Vimentin is phosphorylated at G 2 /M by cyclin-dependent kinase 1 (CDK1) at Ser-55 (7), which primes vimentin for subsequent phosphorylation by polo-like kinase 1 (PLK1) at Ser-82 (8), and consequent inhibition of vimentin filament formation. Besides, vimentin is also phosphorylated by Aurora B at Ser-72 and Rho kinase at Ser-71, leading to localization of phosphorylated vimentin to the cleavage furrow from anaphase to cytokinesis (9 -11). pSer-71 inhibits vimentin filament formation (10), and it is enhanced upon either OGT or OGA over-expression (5). When 12 mitosis-related phosphorylation sites were mutated to Ala and the resultant VIM SA/SA knock-in mice were generated, the mice manifested binucleation and aneuploidy in the lens, age-related cataracts, and skin defects, suggestive of premature aging (12,13).
The detailed mechanism of Chk1 in cytokinesis remains unresolved. In an effort to identify more potential Chk1 interacting proteins, we performed mass spectrometry (MS) analysis following immunoprecipitation (IP) and found OGT. We show here that Chk1 interacts with and phosphorylates OGT at Ser-20, which is a prerequisite of OGT stability. Attenuation of OGT protein levels by siRNA treatment or the S20A mutation hampers the phosphorylation of vimentin at Ser-71, resulting in a failure of vimentin bridge disassembly. Our results not only exemplified another crosstalk between O-GlcNAcylation and phosphorylation, but also revealed the molecular underpinning of Chk1's role in cytokinesis.

Chk1 associates with OGT
To investigate potential proteins that interact with Chk1, we performed an IP experiment using anti-Chk1 antibodies followed by MS analysis and identified OGT (Fig. 1A). To validate the interaction, HeLa cells were subject to IP experiments using anti-Chk1 and anti-OGT antibodies. OGT was detected in anti-Chk1 immunoprecipitates (Fig. 1B), suggesting that endogenous Chk1 and OGT co-immunoprecipitate. To exclude the possibility that the endogenous interaction was because of nonspecificity of the antibodies, we transfected HeLa cells with epitope-tagged Chk1 and OGT plasmids and performed co-IP experiments. As shown in Fig. 1, C and D, both HA-tagged and Myc-tagged OGT interacted with FLAG-Chk1. To test whether the interaction between OGT and Chk1 was direct, we transfected HA-OGT into HeLa cells and used recombinant GST-Chk1 to perform pulldown assays (Fig. 1E). GST-Chk1 could pull down HA-OGT (Fig. 1E). Reciprocally, GST-OGT also could pull down FLAG-Chk1 (Fig. 1F). Taken together, these data suggest that Chk1 interacts with OGT.

Chk1 phosphorylates OGT at Ser-20
In a previous investigation (16), OGT was identified in a phospho-proteomic screen of Chk1, and Ser-20 of OGT was identified as one of the phosphorylation sites. We constructed OGT(S20A) plasmids accordingly, and verified this possibility by in vitro kinase (IVK) assays. Commercially available GST-Chk1 was incubated with GST-OGT or GST-OGT(S20A) proteins and subject to IVK assays. Chk1 phosphorylated OGT efficiently but not the OGT(S20A) mutant ( Fig. 2A). The amino acid residues surrounding Ser-20 were examined for a Chk1 kinase consensus motif, and a minimum consensus for Chk1dependent phosphorylation was identified (Fig. 2B).
We then prepared a phospho-specific antibody toward OGT-pSer-20 as described in "Experimental Procedures" and adopted IVK assays with the pSer-20 antibodies. Chk1 efficiently phosphorylated wild-type (WT) OGT, but not the OGT(S20A) mutant (Fig. 2C). We also tested whether OGT is phosphorylated by Chk1 in vivo. The pSer-20 antibodies specifically detected a band in the HeLa cell extracts, which was significantly attenuated when cells were treated with UCN-01, a specific inhibitor of Chk1 (Fig. 2D). Then HeLa cells were transfected with vectors or HA-OGT-WT plasmids, and synchronized in cytokinesis or left unsynchronized. When treated with UCN-01, the signal of the antibody markedly decreased (Fig.  2E). We sought to determine when the phosphorylation event occurs during the cell cycle. HeLa cell extracts were synchronized by thymidine-nocodazole (Noc) block, and then released. Cell extracts were taken at different time points, prepared, and blotted with the pSer-20 antibodies and subject to flow cytometry analysis. A crisp band was discernable only after the percentage of mitotic cells started to decline (Fig. 2F, lane 2). These results suggest that Chk1 specifically phosphorylates OGT at Ser-20.

Chk1-dependent phosphorylation of pSer-20 stabilizes OGT
OGT has been shown to be poly-ubiquitinated and subject to proteasome-mediated degradation (17). Therefore, we examined whether Chk1-dependent phosphorylation of OGT at Ser-20 modulates OGT abundance. HeLa cells were transfected with HA-OGT WT or S20A plasmids, and treated with cycloheximide (CHX), a protein synthesis inhibitor. Cell extracts were collected at different time points, and subject to immunoblotting (IB). OGT(S20A) displayed significantly shorter halflife than WT (Fig. 3, A and B). Moreover, we examined the ubiquitination levels of OGT by transfecting HA-Ub and Myc-OGT WT or S20A plasmids into HeLa cells. We observed elevated ubiquitination levels in OGT(S20A) (Fig. 3C), consistent with more rapid turnover of OGT(S20A). In addition, we examined interplay between Chk1 kinase activity and OGT ubiquitination levels by treating the cells with UCN-01. As shown in

OGT-pSer-20 localizes to the midbody
We then addressed the question concerning in which intracellular compartments the phosphorylated OGT resides. As shown in Fig. 4A, pSer-20 of OGT localizes to the midbody, suggesting that phosphorylated OGT could take part in cytokinesis. This is also in line with previous findings that OGT localizes to the midbody (5). To determine that the localization was not because of fortuitous binding of the phospho-antibody, we adopted two siRNA oligos targeting OGT, both of which efficiently depleted OGT protein levels (Fig. 4B). Upon siOGT, the midbody localization of pSer-20 significantly decreased (Fig. 4, C and D), suggesting that the midbody localization was indeed specific to OGT-pSer-20. Then we assessed whether the midbody localization pattern was dependent on Chk1. When we used siCHK1 to deplete Chk1 (Fig. 4E), the midbody localization of pSer-20 was also attenuated (Fig. 4F). Upon quantitation, the midbody staining was compromised upon siCHK1 (Fig. 4G), suggesting that Chk1-dependent phosphorylation of OGT localizes to the midbody.
HeLa cells were transfected with two distinct oligos targeting OGT, and then synchronized in the cytokinetic stage (Fig. 5A). The resultant cell extracts were subject to IB to examine O-GlcNAc together with vimentin pSer-71 levels. Our results indicate that decreased O-GlcNAc readily correlated with curbed vimentin-pSer-71 levels (Fig. 5A).
Given that suppressed vimentin-pSer-71 may impede vimentin filament severing and subsequent vimentin bridge formation, we reasoned that decreased O-GlcNAc levels may thus lead to vimentin bridges. We verified our hypothesis by examining vimentin staining in siOGT cells (Fig. 5B). Although vimentin staining was lacking in the midbody area in control cells, vimentin formed conspicuous bridges in the siOGT cells (Fig. 5B). Upon quantitation, cells harboring cytokinetic defects increased markedly in siOGT cells (Fig. 5C).

Chk1 depletion induces vimentin bridge formation during cytokinesis
If the Chk1-OGT-vimentin pathway stands true in modulating cytokinesis, then Chk1 should have a direct role in vimentin phosphorylation and polymerization. We utilized siCHK1 in synchronized cytokinetic cells and examined biochemically whether vimentin-pSer-71 was affected (Fig. 7A), and evidently there was a 30% decrease. Cytologically we applied the Chk1 inhibitor UCN-01 in cells, and vimentin bridges were abundant (Fig. 7, B and C). Taken together, the cytokinetic defects observed in Chk1-depleted cells could be attributed to, at least in part, failure of intermediate filaments to disassemble.

Discussion
OGT is the sole enzyme in humans that accounts for all the O-GlcNAcylation reactions. Recently an intricate role of OGT in cytokinesis emerged. Here we show that Chk1 interacts with OGT, and phosphorylates OGT at Ser-20 both in vitro and in vivo. OGT-pSer-20 localizes to the midbody, and underscores the stability of OGT. Vimentin, subject to both phosphorylation and O-GlcNAcylation, mediates the intermediate filament bridge formation and severing during cytokinesis. We discovered that attenuation of O-GlcNAcylation reduces vimentin-pSer-71 during cytokinesis. Because vimentin-pSer-71 is a prerequisite of vimentin filament disassembly, it is conceivable that vimentin-pSer-71 levels provide the molecular underpinning of cytokinesis defects induced by OGT or Chk1 depletion (Fig. 7D).
The modifications of OGT are far from being known. OGT has been shown to be ubiquitinated and deubiquitinated. The E3 ligase of OGT has recently been identified to be the histone demethylase LSD2 (18). LSD2 not only demethylates H3K4me1/ me2, but also ubiquitinates OGT and promotes its subsequent

Chk1 phosphorylates OGT in cytokinesis
Our results do not preclude the possibility that Chk1 is O-GlcNAcylated by OGT. Indeed, in a proteomic screen for proteins involved in DNA damage response, Chk1 was identified to be O-GlcNAcylated (21). Chk1 phosphorylation profiles changed upon OGT deletion (21). Specifically, pThr-113 increased 2.2-fold whereas pSer-151 decreased 0.7-fold. Therefore, it is highly unlikely a simple friend or foe relationship exists between O-GlcNAcylation and phosphorylation, as far as Chk1 is concerned.
Perhaps it came as a bit of a surprise that O-GlcNAcylation of vimentin actually promotes phosphorylation at Ser-71. This is not unprecedented, however. Although O-GlcNAcylation vies with phosphorylation in a great many cases, it may enhance phosphorylation at times. Indeed, when active phosphorylation sites were closely monitored in a study where O-GlcNAcylation was elevated, 148 phosphorylation sites were increased and 280 phosphorylation sites were reduced (2). As mentioned previously, Chk1 pSer-151 decreased upon OGT depletion (21). Hence, it is not a simple relationship of competition between the two modifications.
We speculate that O-GlcNAcylation may augment Rho kinase activity, the kinase responsible for pSer-71 of vimentin. Multiple lines of genetic evidence are congruent with the idea. First, in endothelium-denuded rat aortas, OGT inhibition via PUGNAc (OGA inhibitor) stimulated contraction to phenylephrine, which was abolished by the Rho kinase inhibitor Y-27632 (22). Second, in the same aorta system, glucosamine was utilized to induce O-GlcNAcylation of proteins by increasing the influx of the hexosamine biosynthesis pathway that feeds into the O-GlcNAc pathway. Glucosamine also elevated RhoA activity, which was negated by OGT inhibition (23). Lastly, in SKOV3 and 59M ovarian cells, siRNA targeting OGT suppressed the cellular migration and invasion, whereas Thiamet-G (OGA inhibitor) bolstered it. More importantly, Thiamet-G boosted RhoA activity and the phosphorylation of Rho kinase substrates, whereas siOGT dampened RhoA activity and subsequent Rho kinase substrate phosphorylation (24). Collectively, O-GlcNAcylation may fuel the RhoA/Rho kinase signaling pathway.
Besides cytokinesis, vimentin also plays various roles in various stages of cancer, including tumorigenesis, epithelial-tomesenchymal transition and ultimately metastasis (25), in human hepatocellular carcinoma (26), lung cancer (27), and breast cancer (28), just to name a few. There has been increasing awareness that vimentin expression elevation correlates with more intensive cancer cell migration and invasion, which could be attributed to the interconnection between vimentin and actin (29), as filamentous vimentin controls actin stress fiber assembly and contractile actomyosin bundles, thus promoting metastasis. Alternatively, vimentin filaments could support extension of microtubule protrusions upon tumor cells detaching from the extracellular matrix (28), and in this way assist a successful tumor cell spread. In either case, an intact vimentin filament is indispensable, and decreased vimentin-pSer-71 may well lend a hand to filament formation. We envision that the Chk1-OGT-vimentin pathway may play a role in cell cycle control and beyond.

Cell synchronization
Protocols to synchronize cells in the cytokinetic phase were described before (31). Briefly, cell cultures were first blocked by double thymidine, and collected 9 h after releasing from the second thymidine block.
For plasmid transfection, cells were seeded at 50 -60% confluence/10 cm 2 Petri dish and transfected with 7.5 g of plasmid DNA using FuGENE 6 according to the manufacturer's instructions for immunoprecipitation (IP) experiments.

In vitro kinase assay
Chk1 in vitro kinase assay was performed as described previously . Briefly, recombinant Chk1 kinase was purchased from R&D Systems (catalog no. 1630-KS), incubated with purified GST-OGT with 1 M HEPES (pH 7.4), 1 M MgCl 2 , 1 M dithiothreitol, 0.1 M Na 3 VO 4 , 0.1 mM ATP or 1 Ci of [␥-32 P]ATP. After 20 min at 30°C, reactions were stopped by the sample buffer. Protein samples were separated by SDS-PAGE and phosphate incorporation was determined by phosphorimager.
Author contributions-J. Li wrote the manuscript. J. Li and X. X. designed the project and analyzed the data. Z. L., X. L., S. N., Q. G., and J. Liao performed the experiments. All authors reviewed and approved the manuscript.