Cdk1-Mediated Phosphorylation of Human ATF7 at Thr-51 and Thr-53 Promotes Cell-Cycle Progression into M Phase

Activating transcription factor 2 (ATF2) and its homolog ATF7 are phosphorylated at Thr-69/Thr-71 and at Thr-51/Thr-53, respectively, by stress-activated MAPKs regulating their transcriptional functions in G1 and S phases. However, little is known about the role of ATF2 and ATF7 in G2/M phase. Here, we show that Cdk1-cyclin B1 phosphorylates ATF2 at Thr-69/Thr-71 and ATF7 at Thr-51/Thr-53 from early prophase to anaphase in the absence of any stress stimulation. Knockdown of ATF2 or ATF7 decreases the rate of cell proliferation and the number of cells in M-phase. In particular, the knockdown of ATF7 severely inhibits cell proliferation and G2/M progression. The inducible expression of a mitotically nonphosphorylatable version of ATF7 inhibits G2/M progression despite the presence of endogenous ATF7. We also show that mitotic phosphorylation of ATF7 promotes the activation of Aurora kinases, which are key enzymes for early mitotic events. These results suggest that the Cdk1-mediated phosphorylation of ATF7 facilitates G2/M progression, at least in part, by enabling Aurora signaling.


Introduction
The activating transcription factors (ATFs) belong to the AP-1 family of transcription factors [1]. ATF consists of seven members, ATF1,7. Among these, ATF2 and ATF7 (originally called ATFa) have highly homologous sequences [2][3][4] and are ubiquitously expressed in various tissues [5,6]. Knockout mutations of Editor: Taishin Akiyama, University of Tokyo, Japan ATF2 and ATF7 lead to early postnatal lethality and abnormal behavioral traits reminiscent of isolation-reared wild-type mice, respectively [7,8]. In addition, the ATF2 and ATF7 double knockout mice die during embryogenesis with abnormalities in the developing liver and heart [2]. ATF2 is mainly controlled by stress-activated protein kinases or protein kinase C (PKC). Jun NH 2 -terminal protein kinase (JNK), p38, and Erk that are activated by stress stimuli can phosphorylate ATF2 at Thr-69 and Thr-71, leading to its transcriptional activation [9][10][11][12][13][14]. Moreover, the phosphorylation of ATF2 at Ser-121 by several PKC isoforms plays a role in the c-Jun-mediated activation of transcription in response to 12-O-tetradecanoylphorbol-13-acetate [15]. It is known that ATF7 is phosphorylated by p38 at Thr-51 and Thr-53, which correspond to Thr-69 and Thr-71 in ATF2, also leading to its transcriptional activation [16,17]. In contrast to its transcriptional functions, ATF2 has some functions that are independent of transcriptional activation [18]. ATF2 is phosphorylated at Thr-52 by PKCe, which negatively regulates the outermembrane permeability of mitochondria and inhibits apoptosis during genotoxic stress [19]. In the DNA damage response, the ATM (ataxia-telangiectasiamutated) protein phosphorylates ATF2 at Ser-490 and Ser-498 to stimulate DNA repair [20,21]. Thus, ATF2 and ATF7 play important roles in G1 and S phases. However, it is largely unknown whether ATF2 and ATF7 play any role in G2 and M phases.
In this study, we have investigated whether ATF2 and ATF7 are phosphorylated in G2 and M phases in HeLa cells. We show that ATF2 (at Thr-69/Thr-71) and ATF7 (at Thr-51/Thr-53) are phosphorylated by cyclin-dependent kinase 1 (Cdk1) in M phase. Notably, we find that, similar to knockdown of ATF7, the expression of a mitotically nonphosphorylatable ATF7 mutant protein inhibits entry of cells into M phase. Our results suggest that phosphorylation of ATF7 at Thr-51/Thr-53 in M phase is required for G2/M progression, in part by activating Aurora kinases.

RNA interference
Knockdown of human ATF7 was performed using shRNAs for silencing ATF7 [59-GCTAGATTTGATGACATATTA-39, which is a sequence in the 39UTR (Sigma MISSION shRNA library), and 59-GTCACATTACTACGCAATG-39, which is a sequence in the CDS] (S1A Fig.). Knockdown of human ATF2 was performed with an shRNA for silencing ATF2 (59-GAAGAAGTGGGTTTGTTTA-39, which is a common sequence among all 13 transcript variants) [28,29]. The oligonucleotides used for shRNA were annealed and subcloned into the XbaI and BglII sites of the pENTR4-H1 vector (provided by Hiroyuki Miyoshi). The EBNA1-based episomal pEBMulti-H1 vector, which encodes the H1 promoter and a neomycin-resistant gene, was generated from the pEBMulti vector (Wako Pure Chemical Industries, Osaka, Japan) by replacing the CAG promoter with the H1 promoter. The oligonucleotides used for shRNA were annealed and subcloned into the pEBMulti-H1 vector. To generate ATF2 or ATF7 knockdown cells, HeLa S3/TR cells or HeLa S3/TR expressing inducible ATF7-wt or ATF7-TA were transfected with pEBMulti-neo/shATF2 or pEBMulti-neo/shATF7, and selected in 600 mg/ml G418. Viable parental HeLa S3/TR cells were not detected after a 5-day selection using 600 mg/ml G418.

Flow cytometry
For cell-cycle analysis, cells detached by trypsinization were fixed in 4% paraformaldehyde for 1 h, and permeabilized with 70% ethanol for at least 1 h at 230uC [26,30]. Fixed cells were permeabilized and blocked in phosphate-buffered saline (PBS) containing 0.1% saponin and 3% bovine serum albumin (BSA) for 30 min at room temperature. After washing with PBS containing 0.1% Tween 20, cells were reacted with anti-phospho-histone H3 Ser-10 antibody for 1 h at room temperature, then stained with AF647-conjugated anti-mouse IgG antibody for 1 h. Subsequently, cells were treated with 200 mg/ml RNase A and 50 mg/ml propidium iodide (PI) at 37uC for 30 min to stain DNA. A minimum of 5,000 cells per sample was analyzed by flow cytometry using a Guava easyCyte (Millipore) equipped with a 488-nm blue laser and a 640-nm red laser using liner amplification. Data acquired with a Guava easyCyte apparatus were analyzed using Flowing Software version 2.5.0 (Perttu Terho, Centre for Biotechnology, Turku, Finland). Cell debris was excluded by gating on forward scatter and pulsewidth profiles.

Cell synchronization
The HeLa S3 cell line is a HeLa cell variant that can be highly synchronized at a various stages of the cell cycle [26,27,[31][32][33]. HeLa S3 cells were treated with 4 mM thymidine for 24 h, washed and released into thymidine-free fresh medium. After 11 h of release from S-phase arrest, mitotic cells were collected by mitotic shake-off. Cells were grown for 24 h in medium with 4 mM thymidine, an additional 9 h without thymidine, then an additional 15 h with 4 mM thymidine (DTB), and released into fresh medium for 10 h (G2 phase) and 11,12 h (M phase). Alternatively, to synchronize exponentially growing cells at M phase, cells were treated with 4 mM thymidine for 24 h. After washing with PBS, cells were released into thymidine-free medium and cultured for 5 h. The cells were then incubated with 9 mM RO-3306 for 10 h, to arrest cells at late G2 phase. The G2arrested cells were washed with PBS supplemented with Ca 2+ and Mg 2+ and released into prewarmed fresh medium [26]. Cell-cycle distribution was determined by flow cytometry.

Immunofluorescence
Immunofluorescence staining was performed as described previously [34][35][36]. In brief, HeLa S3 cells were cultured in IMDM containing 1% FBS and 4% bovine serum at 37uC. Cells were fixed in 4% paraformaldehyde for 15 min at room temperature. Fixed cells were permeabilized with 100% methanol for 10 min at 220uC and blocked in PBS containing 0.1% saponin and 3% BSA for 30 min at room temperature. Cells were stained with a primary and a secondary antibody for 1 h each. For DNA staining, cells were subsequently treated with 200 mg/ml RNase A for 30 min and 20 mg/ml PI for 30 min. Stained cells were mounted with an antifade reagent. Confocal and Nomarski differential-interference contrast images were obtained using an FV500 laser scanning microscope (Olympus). To ensure that there was no bleed through from the Alexa Fluor 488 signal into the red channel, Alexa Fluor 488 and Alexa Fluor 546 were independently excited at 488 nm and 543 nm, respectively. Emission signals were detected between 505 and 525 nm for Alexa Fluor 488 and between 560 and 600 nm for Alexa Fluor 543. Composite figures were prepared using Photoshop CS6 and Illustrator CS6 (Adobe).

Involvement of ATF2 and ATF7 in cell-cycle progression
The human ATF2 and human ATF7 genes generate 13 and six transcript variants, respectively. To examine whether ATF2 and ATF7 are involved in cell-cycle progression into G2/M phases, we chose the following knockdown target sequences: the first sequence was common among the human ATF2 variants, and the other two sequences were the sequences in the coding sequence (CDS) and the 39-untranslated region (39UTR) of human ATF7 (see Materials and Methods, S1A Fig.). We tried to knock down endogenous ATF2 and ATF7 in HeLa S3 cells using short hairpin RNAs (shRNAs). After transfection of cells with an EBNA1-based episomal vector, which encoded shRNA against ATF2 (shATF2) or ATF7 [shATF7(CDS) or shATF7(39UTR)], we selected knockdown cells in G418containing medium. Western blot analysis showed that the level of endogenous ATF2 or ATF7 was significantly decreased 5 days post transfection ( Fig. 1A-B S1B-E Fig.). The use of shATF7(CDS) or shATF7(39UTR) resulted in efficient knockdown of ATF7 proteins. Intriguingly, knockdown of ATF2 or ATF7 inhibited the rate of cell proliferation and decreased the number of M-phase cells (Fig. 1C, S1F Fig.). Note that ATF7 knockdown had severe inhibitory effects on cell proliferation compared with the ATF2 knockdown ( Fig. 1C, S1F Fig.). These results suggest that ATF2 and ATF7, in particular ATF7, play a role in cell proliferation, in addition to their well-established role during stress and DNAdamage responses in G1/S phases. Moreover, ATF7 knockdown also strongly affected the level of the ATF2 protein ( Fig. 1B, S1B-E Fig.), although none of the ATF2 transcript variants contain any detectable homology to the shATF7(39UTR) and shATF7(CDS) sequences. These findings suggest that the level of the ATF2 protein is dependent on that of ATF7.
Next, we examined by flow cytometry whether knockdown of ATF2 and ATF7 affected cell viability. We found that the number of cells in subG1 phase was significantly increased upon ATF7 knockdown 7 days post transfection (Fig. 1D). To examine further whether ATF2 and ATF7 play a role in cell-cycle progression, we highly synchronized knockdown cells using a double-thymidine block (DTB) method. After transfection with either vector, shATF2 or shATF7(39UTR), cells that were synchronized in G1/S phase appeared to progress normally into G2/M phase 10 h after release from DTB. It is of interest to note that 12 h after release from DTB, knockdown of ATF7 by shATF7(39UTR) severely decreased the number of G1-phase cells compared with knockdown of ATF2 by shATF2 (Fig. 1E), which indicates that ATF7 knockdown by shATF7(39UTR) drastically inhibits the progression of cells from M phase to G1 phase. Moreover, knockdown of ATF7 by shATF7(CDS) appeared to partially arrest cells at G1 phase (Fig. 1E, S2 Fig.), in addition to inhibiting the progression from M phase to G1 phase (Fig. 1E, compare 10 h with 12 h), indicating that the ATF7 transcript variant 6 (long noncoding RNA) (S1 Fig.) is involved in the progression from G1 phase to S phase. Taken together, these results suggest that ATF7 knockdown primarily affects cell-cycle progression during M phase.
Phosphorylation of ATF2 at Thr-69/Thr-71 and of ATF7 at Thr-51/ Thr-53 in M phase ATF2 and ATF7 are activated by threonine phosphorylation [12,16]. The amino acid sequence surrounding the threonine-phosphorylation sites of ATF2 is identical to that of ATF7. The use of anti-phosphorylated ATF2[pT71] antibody (anti-pT69/71) enabled us to detect the phosphorylation of Thr-69/Thr-71 of ATF2 and Thr-51/Thr-53 of ATF7. In some experiments anti-ATF2[N96] antibody, which recognizes both ATF2 and ATF7 [41], was used. To examine the phosphorylation states of endogenous ATF2 at Thr-69/Thr-71 and endogenous ATF7 at Thr-51/Thr-53 in the cell cycle, we synchronized cells at S, G2, and M phases, and used anti-pT69/71 and anti-ATF2[N96] antibodies. Western blotting with anti-pT69/71 antibody revealed that endogenous ATF2 and ATF7 were strongly phosphorylated in M phase. Strikingly, the appearance of ATF2 and ATF7 protein forms with lower mobility on SDS-PAGE, suggestive of phosphorylation, was only found in M phase ( Fig. 2A). To identify the kinase that phosphorylates ATF2 and ATF7 in M phase, we used inhibitors of various mitotic kinases, such as Cdk1, Aurora A, Aurora B, and Polo-like kinase 1. We also used anti-ATF2 antibody[SS-16] and anti-ATF7 antibody[HPA003384], which recognize specific sequences of ATF2 and ATF7, respectively. Notably, the phosphorylation levels of ATF2 and ATF7 in M phase were drastically decreased in cells treated with RO-3306, an inhibitor of Cdk1 (Fig. 2B, C). In addition, we noticed that the amino acid sequence surrounding the phosphorylation sites of ATF2 and ATF7 complied with the Cdk1 phosphorylation consensus sequence (S/ T*-P-x-K/R). Since MAP kinases and PKC phosphorylate ATF2 and ATF7 in G1 phase [12,15,16], we examined whether MAP kinases and PKC were capable of phosphorylating ATF2 at Thr-69/Thr-71 and ATF7 at Thr-51/Thr-53 in M phase. . Asterisks indicate the significant differences (*P,0.05; **P,0.01), as calculated by Student's t-test. (D) Cells transfected with control vector, ATF2 shRNA, ATF7 shRNA(39UTR), or ATF7 shRNA(CDS) were collected on day 7 and stained with propidium iodide (PI). SubG1 cells were quantitated by flow cytometry. Values are means ¡ SD, n 53 independent experiments. Asterisks indicate the significant differences (*P,0.05; **P,0.01), as calculated by Student's t-test. (E) Schematic depiction of our synchronization method. Cells transfected with control vector, ATF2 shRNA, ATF7 shRNA(39UTR), or ATF7 shRNA(CDS) were synchronized using double thymidine block (DTB) in the presence of 600 mg/ml G418 and released into thymidine-free medium for 10,12 h. Cells were stained with PI for analyzing cell cycle progression by flow cytometry (left panels). The percentages of G1-phase cells were compared between 10 h and 12 h after release from DTB (right graph).  However, treatment with inhibitors of MAP kinases such as p38, JNK, and Erk did not decrease the phosphorylation levels of ATF2 at Thr-69/Thr-71 and ATF7 at Thr-51/Thr-53 in M phase (Fig. 2C). Moreover, inhibition of PKC by Gö6976 did not affect the phosphorylation levels of ATF2 at Thr-69/Thr-71 and of ATF7 at Thr-51/Thr-53, although PKC inhibition partially reduced the mobility shift of ATF2 and ATF7 (Fig. 2D). These results suggest that Cdk1, but not MAP kinases or PKC, phosphorylates ATF2 at Thr-69/Thr-71 and ATF7 at Thr-51/Thr-53 in M phase.
Phosphorylation of ATF2 and ATF7 from prophase to anaphase To examine whether ATF2 was able to form a complex with ATF7, we immunoprecipitated endogenous ATF2 and endogenous ATF7 using the specific antibodies ATF2[SS-16] and ATF7[SAB2500131 or HPA003384], respectively. The results of co-immunoprecipitation experiments showed that ATF2 was associated with ATF7, and that ATF7 was associated with ATF2 (Fig. 4A). In addition, phosphorylated ATF2 and ATF7 could be immunoprecipitated using anti-pT69/71 antibody from cell extracts prepared from M phase cells. Next, untransfected cells were co-immunostained for endogenous, phosphorylated ATF2/ATF7 and endogenous Cdk1. Phosphorylated ATF2/ATF7 colocalized with Cdk1 and cyclin B1 in M phase (Fig. 4B, arrows, Fig. 4C). To compare the localization of ATF2/ATF7 and phosphorylated ATF2/ATF7, we used anti-ATF2[N96], anti-pT69/71, and anti-phospho-histone H3 at Ser-10 [H3pS10] antibodies. Anti-ATF2[N96] antibody, which recognizes both ATF2 and ATF7, reacted with the nucleus in interphase, the cytoplasm from prophase to anaphase, and chromosomes in telophase (Fig. 4D, upper panels). Anti-pT69/71 antibody reacted with the cytoplasm from prophase to anaphase (Fig. 4D, lower panels), and the mitotic phosphorylation of ATF2 and ATF7 preceded the onset of the phosphorylation of histone H3 at Ser-10 (Fig. 4E). Anti-H3pS10 and DNA staining confirmed the presence of cells undergoing prophase, metaphase, or anaphase. The use of the Cdk1 inhibitor RO-3306 and the Aurora kinase inhibitor ZM447439 showed that phosphorylation of ATF2 and ATF7 was not inhibited by ZM447439 and was inhibited by RO-3306, a result that further substantiates the notion that the mitotic phosphorylation of ATF2 and ATF7 takes place upstream of the activation of Aurora kinases (Fig. 4F). Moreover, to examine the relationship between the levels of phosphorylated ATF2/ATF7 and cyclin B1, we examined at the metaphase-anaphase transition. Mitotic slippage was induced by ZM447439 after cells were arrested by the spindle checkpoint-inducer monastrol. We observed that phosphorylation of ATF2/ATF7 was decreased to basal levels in anaphase and telophase, where cyclin B1 are low due to active degradation (Fig. 4G). These results suggest that ATF2 and ATF7 are phosphorylated by Cdk1cyclin B1 from early prophase to anaphase.

Role of the phosphorylation of ATF7 in G2/M progression
Although ATF2 and ATF7 were similarly threonine-phosphorylated in M phase (Fig. 2, 4A), ATF7 knockdown affected cell proliferation and cell-cycle progression more severely compared with ATF2 knockdown (Fig. 1C-E, S1F Fig.). Therefore, we focused on scrutinizing the role of ATF7 in M phase. To examine the role of the mitotic threonine-phosphorylation of ATF7 in cell-cycle progression, we generated three independent, stable HeLa S3/TR cell clones that expressed tetracycline-inducible ATF7-wt or ATF7-TA (see Fig. 3A). Western blot analysis showed that expression of ATF7-wt and ATF7-TA was induced by treatment with doxycycline (Dox), an analog of tetracycline (Fig. 5A, S3A-B Fig.). Upon Dox treatment, we examined the localization of ATF7-wt and ATF7-TA in specific cell-cycle phases and found that ATF7-wt and ATF7-TA were similarly localized to the nucleus in interphase, the cytoplasm from prophase to anaphase, and on mitotic chromosomes in telophase (Fig. 5B). These results suggest that the state of mitotic phosphorylation of ATF7 at Thr-51/Thr-53 does not affect the localization of ATF7. Next, to examine whether the mitotic phosphorylation of ATF7 affected cellcycle progression, we synchronized parental cells, ATF7-wt-and ATF7-TAinducible cells with DTB, and treated them with Dox after release from DTB (Fig. 5C, 0 h). In the case of parental cells, 19% of cells progressed into M phase at 10.5 h as assessed by induction of the phosphorylation of histone H3 at Ser-10. Subsequently, 24% and 62% of cells divided and proceeded to enter G1 phase at 11.5 h and 12.5 h, respectively. Moreover, upon ATF7-wt induction, the rate of cell-cycle progression was similar to that of parental cells. However, upon ATF7-TA induction, only 2,11% of cells progressed into M phase at 10.5 h. Subsequently, only 9,14% and 16,37% of cells divided and proceeded to enter G1 phase at 11.5 h and 12.5 h, respectively (Fig. 5C, S3C-D and S4 Figs.). To investigate in greater detail the effect of the phosphorylation state of ATF7 on the G2/M transition, we synchronized cells in late G2 phase using RO-3306, and released cells into M phase. For induction of ATF7-wt and ATF7-TA, cells were treated with Dox at 5 h before release from G2 arrest. We found that the expression of ATF7-TA, but not of ATF7-wt, inhibited progression into M phase even in the presence of endogenous ATF7 (Fig. 5D, S3E Fig.), which suggests that ATF7-TA acts as a dominant-negative protein. These results indicate that the mitotic phosphorylation of ATF7 is required for normal G2/M progression.

Effect of ATF7 phosphorylation on Aurora signaling
Although G2/M progression was inhibited by ATF7-TA expression (Fig. 5C-D, S3C-E and S4 Figs.), it was unclear how phosphorylation of ATF7 was involved in G2/M progression. To investigate the effect of ATF7 phosphorylation on M phase, we wished to perform rescue experiments using ATF7-knockdown cells. However, considering that ATF7 knockdown inhibited cell proliferation (Fig. 1C, E, S1F  Fig.), we assumed that ATF7-knockdown cells are not capable of undergoing cellcycle synchronization because of poor cell-cycle progression. Accordingly, the expression of ATF7-wt and ATF7-TA was induced by Dox addition before endogenous ATF7 proteins were knocked down, and the cells were subsequently synchronized in the presence of Dox using the DTB method ( Fig. 6A-C, S5C-E  Fig.) or the thymidineRRO-3306 method (Fig. 6A, D, and E, S5B, F, and G Fig.). 2), or HeLa S3/TR/ATF7-TA (cl.1) cells were synchronized using DTB and released into thymidine-free medium containing 1 mg/ml Dox for 12 h. Whole cell lysates were analyzed by WB. Full-length blots are presented in S12A Fig. (B) HeLa S3/TR/ATF7-wt (cl.2) (upper panels) or HeLa S3/TR/ATF7-TA (cl.1) cells (lower panels) were synchronized using single-thymidine block and released into thymidine-free medium containing 1 mg/ml Dox for 11 h. Cells were triply stained with anti-ATF7[SAB2500131] and anti-a-tubulin antibodies and PI (for DNA). Scale bars, 20 mm. (C, D) Cells were stained with anti-histone H3pS10 antibody (for M phase) and PI for analyzing cell-cycle progression by flow cytometry. (C) Parental HeLa S3/TR, HeLa S3/TR/ATF7-wt (cl.2), or HeLa S3/TR/ATF7-TA (cl.1) cells were synchronized using DTB and released into thymidine-free medium containing 1 mg/ml Dox for 10.5,12.5 h. Two-dimensional histograms (DNA vs histone H3pS10) are presented together with DNA histograms, and the percentages of cells in G1 and M phases were measured. Cells in M and and G1 phases were quantitated from the results of S4 Fig. Values are means ¡ SD (three independent clones). Asterisks indicate the significant difference (*P,0.05; NS, not significant), as calculated by Student's t-test. (D) Parental HeLa S3/TR, HeLa S3/TR/ATF7-wt (cl.2), or HeLa S3/TR/ATF7-TA (cl.1) cells were cultured in the presence of 9 mM RO-3306 for 10 h and treated with 1 mg/ml Dox for the last 5 h. The cells that were arrested at G2 phase were released into RO-3306-free medium containing 1 mg/ml Dox and incubated for 0, 10, and 20 min. Two-dimensional histograms (DNA vs histone H3pS10) are presented together with DNA histograms, and M-phase cells were quantitated. Values are means ¡ SD, n 53 independent experiments. An asterisk indicates the significant difference (*P,0.05), as calculated by Student's t-test. We confirmed the successful knockdown of endogenous ATF7 and ATF2 (Fig. 6B,  E, S5A, B, F, And G Fig., see also Fig. 1B). Because transfection of shATF7(39UTR) appeared to act mainly in M phase (Fig. 1E), inducible expression of ATF7-wt and ATF7-TA, none of which contain the 39-untranslated region, was carried out under conditions in which shATF7(39UTR) was transfected (Fig. 6B, S5A-B Fig.). To examine whether ATF7 knockdown affected the number of daughter cells after mitosis, we measured the percentage of G1phase cells using a flow cytometer 12 h after release from DTB (Fig. 6C, S5E Fig., see Fig. 1E). We found that ATF7 knockdown decreased the number of G1 cells and that the inducible expression of ATF7-wt, but not ATF7-TA, in ATF7knockdown cells recovered the number of G1 cells (Fig. 6C, S5E Fig.). Furthermore, ATF7 knockdown decreased the number of viable cells, accompanied by an increase in the number of subG1 cells, and this ATF7-knockdowninduced apoptosis was alleviated by the inducible expression of ATF7-wt, but not ATF7-TA (S5C-D Fig.). These results suggest that phosphorylation of ATF7 is important for cell-cycle progression and cell viability.
To examine whether ATF7-knockdown cells were capable of entering M phase, these cells were synchronized in G2/M phase before completion of ATF7 knockdown using the method depicted in Fig. 6A-(ii) and co-immunostained for both endogenous ATF7 and Ser10-phosphorylated histone H3 as an M phase marker was carried out. It has to be emphasized that phosphorylation of histone H3 at Ser-10 was not detectable in ATF7-knockdown cells (Fig. 6D). Because Aurora kinase phosphorylates histone H3 at Ser-10 [42][43][44][45], we examined whether the phosphorylation of ATF7 played a role in Aurora signaling. Western blot analysis showed that the phosphorylation and protein levels of Aurora A, B, and C, in particular Aurora A, were reduced in G2/M phase after ATF7 knockdown (Fig. 6E, S5F-G Fig.). Importantly, the inducible expression of ATF7-wt, but not ATF7-TA, in ATF7-knockdown cells restored the phosphorylation and protein levels of Aurora kinases to normal M-phase levels. These results indicate that ATF7 phosphorylation plays a role in Aurora signaling, which is known to be essential for G2/M transition.

Discussion
In the present study, we have identified for the first time ATF2 and ATF7 as novel M phase substrates of the Cdk1-cyclin B1 complex. Mitotic phosphorylation of human ATF2 takes place on Thr-69/Thr-71 and on Thr-51/Thr-53 for human ATF7 from early prophase to anaphase. In agreement with a role for ATF2 and ATF7 in enabling passage through M phase, knockdown of both endogenous ATF2 and ATF7 inhibited cell proliferation and cell-cycle progression through M phase, but the effect of ATF7 knockdown was more severe compared with that of ATF2. We have also shown that the inducible expression of mitotically nonphosphorylatable ATF7-TA, but not ATF7-wt, inhibited G2/M progression, despite the presence of endogenous ATF7. Knockdown rescue experiments revealed that the mitotic phosphorylation of human ATF7 by the Cdk1-cyclin B1 complex plays an important role in enabling Aurora signaling and mitotic entry. Our results also suggest that the threonine-phosphorylation of ATF7 and ATF2 in G1/S phases versus G2/M phases play differential functional roles (Fig. 7).
A growing body of evidence demonstrates that phosphorylation of ATF2 at Thr-69/Thr-71 and of ATF7 at Thr-51/Thr-53 in the N-terminal activation domains is required for their transcriptional activation upon exposure of G1/Sphase cells to stress stimuli [9-13, 16, 17]. However, the role of ATF2 and ATF7 in G2/M phases has not yet been reported. Intriguingly, the protein structure of human ATF2 is highly homologous to that of human ATF7 in the N-terminal activation domain, which contains the threonine-phosphorylation sites, and in the DNA-binding domain (Fig. 3A, S6A Fig.). ATF2 is maintained in an inactive form by an intramolecular interaction between the N-terminal activation domain and the DNA-binding domain, and activators of ATF2 disrupt this interaction to activate transcription [46]. The activators of ATF2 include (i) regulatory proteins, such as adenovirus E1A, HTLV-I Tax, HMG1, and c-Jun, which are able to interact with the DNA-binding domain; and (ii) threonine phosphorylation by stress-activated MAP kinases, which increases the activity of the DNA-binding domain, both of which may be coordinately regulated [17,[46][47][48][49][50][51][52]. It is well known that activation of the Cdk1-cyclin B1 complex is required for G2/M entry and progression, and that the Cdk1-cyclin B1 complex is activated from late G2 phase to late anaphase and phosphorylates various mitotic substrates [53,54]. Considering that the two threonine residues on human ATF2 and human ATF7 that are phosphorylated by the Cdk1-cyclin B1 complex are identical to the respective sites that are phosphorylated by MAP kinases (Fig. 2, 3), their mitotic phosphorylation may play a role in regulating dynamic events during M phase. Furthermore, we hypothesized that ATF2 and ATF7 directly or indirectly interact with nucleocytoplasmic or cytoplasmic proteins, rather than chromosomes, during M phase, because mitotically phosphorylated ATF2 and ATF7 do not appear to localize to chromosomal DNA, but are dispersed throughout the nucleocytoplasm/cytoplasm from early prophase to anaphase (Fig. 4D-E). Thus, these novel functions of ATF2 and ATF7 may be found in the nucleocytoplasm/ cytoplasm during G2/M progression, unlike those of AP-1-dependent transcription factor during G1 and S phases [9][10][11][12].
Similar to human ATF2, human ATF7 was phosphorylated by the Cdk1-cyclin B1 complex during M phase (Fig. 2, 3, and 4A). However, knockdown of human ATF7 inhibited cell proliferation and cell-cycle progression much more strongly than did that of human ATF2 (Fig. 1C-E, S1F Fig.). Because human ATF7 exhibits relatively low homology to human ATF2 in the middle and the Cterminal regions (S6A Fig.), it is reasonable to assume that the role of human ATF7 in M phase is somewhat different from that of human ATF2. Moreover, given that double knockdown of ATF2 and ATF7 yielded strong inhibition of mitotic progression compared with single knockdown of ATF2 or ATF7 (S1F Fig.), ATF2 and ATF7 might play a synergistic role in mitotic progression. Notably, the human ATF2 and human ATF7 genes generate 13 and six transcript variants, respectively. Our knockdown construct against human ATF2 (shATF2) targeted the common sequence among all of the 13 transcript variants of human ATF2 [28,29] (see Materials and Methods). Regarding ATF7 knockdown, one of our knockdown constructs against human ATF7 [shATF7(39UTR)] targeted the human ATF7 transcript variant 2, transcript variant 3, and an uncharacterized long noncoding RNA, and the other knockdown construct [shATF7(CDS)] targeted human ATF7 transcript variant 2, transcript variant 3, and long noncoding RNA transcript variant 6 (S1A Fig., see Materials and Methods). Despite the induction of partial arrest in G1 phase by shATF7(CDS) (Fig. 1E), both shATF7(CDS) and shATF7(39UTR) are fully capable of knocking down human ATF7 proteins and of inhibiting cell-cycle progression from M phase to G1 phase (Fig. 1B, S1 Fig.). Considering that the inducible expression of ATF7-TA inhibited G2/M progression, most likely because of a dominant-negative effect (Fig. 5C-D S3 and S4 Figs.), it should be emphasized that the mitotic phosphorylation of human ATF7 proteins, rather than human ATF2 proteins, is important for cell-cycle progression from M phase to G1 phase (Fig. 1E).
Mice bearing a deletion of the DNA-binding domain of the ATF7 gene are viable, and no severe abnormalities are observed in adult animals [2,41]. Regarding ATF2 knockout mice, the deletion of the DNA-binding domain of the ATF2 gene invariably leads to death at birth [8]. Knock-in mutant mice in which the two threonine-phosphorylation sites located in the N-terminal ATF2 activation domain are mutated into alanine also die shortly after birth [2]. In contrast with the single deletions, ATF2/ATF7 double deletion results in severe hypoplasia of the embryonic heart and liver, which leads to death between embryonic days 11.5 and 12.5 [2]. This demonstrates that mouse ATF2 and mouse ATF7 share essential functions during embryonic development [2,17]. Nonetheless, primary fetal liver cells prepared from ATF2/ATF7 double knockout mice are able to grow over 7 days in culture, although the cells exhibit reduced rates of cell proliferation after 2,3 days of culture, as well as an increase in the number of apoptotic cells [2]. Collectively, we assumed that mouse ATF7 does not have a significant effect on cell proliferation compared with human ATF7. Human ATF7 may play a more complex role in cell signaling than does mouse ATF7 because human ATF7 has six transcript variants and the structure of the human ATF7 protein contains an additional C-terminal sequence compared with that of mouse ATF7 (S1A and S6B Fig.). In fact, we have not yet succeeded in establishing stable ATF7 knockdown HeLa S3 cell lines, which may be attributable to the fact that knockdown of human ATF7 strongly inhibits cell proliferation (Fig. 1C). To overcome this problem, we introduced an EBNA1-based episomal vector encoding an shRNA and selected knockdown cells using G418 in a short span of time (Fig. 1A, E, and 6A). Our knockdown method has the advantage of avoiding the preparation of a single-cell clone on the grounds that a single-cell clone is, to some extent, idiosyncratic and may have undergone compensatory adaptations or mutations.
ATF2 and ATF7 can form homodimers and heterodimers with other AP-1 transcription factors [17,18]. In fact, we showed that ATF2 is co-immunoprecipitated with ATF7 from Triton X-100 lysates of asynchronous and M-phase cells (Fig. 4A), which is supported by in vitro data [55]. N-terminal phosphorylation of ATF2 at Thr-69/Thr-71 by MAP kinases protects ATF2 from ubiquitination and degradation [56,57]. Furthermore, ATF7 knockdown affects the level of the ATF2 protein, and vice versa (Fig. 1B, 6B, S1B-E Fig.). Taken together, these results suggest that ATF2 and ATF7 actually interact with each other in vivo to regulate their stabilization.
Intriguingly, we showed that mitotic phosphorylation of ATF7 is involved in Aurora signaling (Fig. 6E, S5F-G Fig.). The Aurora family consists of three members, Aurora A, B, and C. Aurora A is an oncogene in a variety of cancers and plays a role in centrosome maturation for G2/M transition [58]. The disruption of Aurora A function delays mitotic entry [59,60]. Since inhibition of Aurora kinases is also known to induce apoptosis [61], down-regulation of Aurora kinase signaling by ATF7 knockdown (Fig. 6) may explain an increase in the presence of cells in subG1 phase (Fig. 1D) following G2/M-phase arrest of cells in which ATF7 was knocked down (Fig. 1C, S1F Fig.). Inhibition of the kinase activity of Cdk1 induces Aurora A inactivation, even though Cdk1 does not directly phosphorylate Aurora A [62]. Activated Aurora B phosphorylates histone H3 at Ser-10 [42][43][44][45]. However, once Cdk1 is activated, mitotic phosphorylation of ATF7 precedes phosphorylation of histone H3 at Ser-10 (Fig. 4E, F), which suggests that ATF7 is located upstream of Aurora kinases. Thus, we assumed that the mitotic phosphorylation of ATF7 by the Cdk1-cyclin B1 complex promotes the activation of Aurora kinases for mitotic entry via the stabilization of Aurora kinases. In addition, to examine the effect of proteasome inhibition on ATF7-mediated Mphase progression, we treated parental cells, ATF7-wt-or ATF7-TA-inducible cells with the proteasome inhibitor MG132 at the onset of mitosis. Treatment with MG132 almost fully blocked entry of ATF7-TA-induced cells into M phase (S5H Fig.), which suggests that ATF7-TA-mediated inhibition of M-phase progression is augmented by MG132 treatment. Considering that MG132 blocks entry into and exit from mitosis through intricate mechanisms [63,64], we hypothesized that the mitotic phosphorylation of ATF7 promotes Aurora signaling at the onset of M phase in a manner dependent on proteasome activity. Although we have not yet detected any physical associations between ATF7 and the phosphorylated Aurora kinases, INCENP or TPX2, it is now of interest to determine precisely how the mitotic phosphorylation of ATF7 is involved in Aurora signaling.