ATF7 is stabilized during mitosis in a CDK1-dependent manner and contributes to cyclin D1 expression

Abstract

The transcription factor ATF7 undergoes multiple post-translational modifications, each of which has distinct effects upon ATF7 function. Here, we show that ATF7 phosphorylation on residue Thr112 exclusively occurs during mitosis, and that ATF7 is excluded from the condensed chromatin. Both processes are CDK1/cyclin B dependent. Using a transduced neutralizing monoclonal antibody directed against the Thr112 epitope in living cells, we could demonstrate that Thr112 phosphorylation protects endogenous ATF7 protein from degradation, while it has no effect on the displacement of ATF7 from the condensed chromatin. The crucial role of Thr112 phosphorylation in stabilizing ATF7 protein during mitosis was confirmed using phospho-mimetic and phospho-deficient mutants. Finally, silencing ATF7 by CRISPR/Cas9 technology leads to a decrease of cyclin D1 protein expression levels. We propose that mitotic stabilized ATF7 protein re-localizes onto chromatin at the end of telophase and contributes to induce the cyclin D1 gene expression.

Introduction

Activating Transcription Factor 7 (ATF7) belongs to the ATF/cAMP response element-binding protein family, and is composed of at least 3 variants (ATF7–1, ATF7–2 and ATF7–4).^1,2 Although ATF7–2 differs from ATF7–1 by additional 21 residues within the N-terminal domain, their transcriptional activities are undistinguishable.³ The ATF7–4 isoform is a C-terminally truncated variant of ATF7, which down-regulates transcriptional activity of both ATF7 and ATF2.¹

ATF7 shares significant structural and functional homologies with ATF2.⁴ Especially, the N-terminal activation domain and the C-terminal DNA binding/dimerization domain are highly conserved.⁵ Indeed, ATF7, like ATF2, can also associate with c-jun or c-fos through its C-terminal leucine-zipper region.⁶ The ATF7 activation domain is comprised of a critical zinc-binding element and 2 threonine residues, Thr51 and Thr53, related to Thr69 and Thr71 in ATF2.^6,7 Phosphorylation events on both Thr51 and Thr53 are induced by stress such as UV irradiations and stimulate the transcriptional activity of ATF7.^5,8 Altering these 2 threonine residues into alanine residues impairs the direct interaction between ATF7 and TAF12, a subunit of TFIID complex, and subsequently abolishes ATF7-dependent transcriptional activation.³ ATF7 also silences the serotonin 5-HT receptor 5B (Htr5b) gene in mouse by recruiting the ESET to induce the formation of a heterochromatin-like structure.⁹ Lastly, the Lys118 of ATF7 is targeted by Small Ubiquitin-like Modifier proteins (SUMO), which delays its nuclear entry and inhibits its transcriptional activity by perturbing its interaction with TAF12.¹⁰

Recent reports using mass spectrometry experiments have shown that phosphorylation of ATF7 undergoes cell cycle-regulated changes.^11,12 Cell division is a complex process tightly regulated by cyclin-dependent kinases (CDKs) and their associated cyclins.^13,14 CDKs activate or inactivate target proteins by phosphorylation to regulate crucial protein functions in eukaryotic cells.^15-17 For instance, entry into mitosis is characterized by phosphorylation of specific transcription factors leading to their displacement from the condensed chromosomes.^15,18,19 In few cases, phosphorylation protects transcription factors from protein degradation during cell cycle progression.^20,21

In the present study, we show that ATF7 undergoes phosphorylation on its Thr112 residue at the onset of mitosis and is displaced from the condensed chromatin in a CDK-dependent manner. Using a site-specific Thr112 monoclonal antibody in living cells, we demonstrate that phosphorylation on the Thr112 residue plays a crucial for ATF7 stabilization, rather than reducing its DNA-binding ability during mitosis. Lastly, we show that ATF7 activates the cyclin D1 gene expression to the same extends as ATF2.

Results

Thr112 phosphorylation peaks during the cell cycle G2/M phase

While phosphorylation events on both Thr51 and Thr53 have been positively correlated to ATF7 transcriptional activity, similarly to the phosphorylation events on Thr69 and Thr71 of ATF2, the physiological relevance of Thr112 phosphorylation remains unknown.⁵ Recent phospho-proteome screenings have revealed a large number of proteins, including ATF7, whose phosphorylation levels fluctuate during the cell cycle.^11,12

In order to confirm that ATF7 is regulated during cell cycle progression, we examined the expression and the phosphorylation patterns of ATF7 in synchronized HeLa-ATF7 cells, a stable cell line expressing ATF7–1 isoform. Cells were arrested in G1/S phases by a double thymidine block. At different times following block release, cells were exposed or not to UV irradiation and whole cell extracts were analyzed by Western-blotting. UV radiations induced shifted bands due to phosphorylation events on Thr53 and Thr112 residues at all stages of the cell cycle following drug withdrawal (Fig. 1A). Strikingly, in unexposed cells, Thr112 phosphorylation also occurred transiently between 6 hours and 10 hours after drug removal, with a peak at 8 hours, while more than 90% of the cells were in G2/M phases (Fig. 1A, B). Thereafter, phosphorylation on Thr112 rapidly declined as cells entered G1 phase (Fig. 1A). By contrast, Thr53 phosphorylation was not induced in unexposed cells (Fig. 1A). Thus, in the absence of UV treatment, phosphorylation of Thr112 is specifically restricted to cell cycle G2/M phases, whereas UV radiations induce a strong phosphorylation of both Thr53 and Thr112 residues throughout the cell cycle.

Figure 1.

Phosphorylation of ATF7 on Thr112 occurs in mitotic cells. (A) HeLa-ATF7 cells were either synchronized using a double thymidine block or not (AS for asynchronous). After block release, cells were exposed (+) or not (−) to UV radiations (40J/m2) 30 min prior to harvesting at the indicated time points. Whole cell extracts were analyzed by Western-blotting with pThr53, pThr112 and ATF7 mAbs, and an anti-cyclin B1 to visualize the end of mitosis, as indicated. Actin was used as a loading control. (B) To monitor cell cycle progression an aliquot of each sample was taken for FACS analysis. Percentages of mitotic cells are indicated.

Endogenous ATF7 phosphorylated on Thr112 is excluded from the condensed chromatin during mitosis

To assess the Thr112 phosphorylation events of endogenous ATF7, we performed immunofluorescence assays on asynchronous HeLa, Mel501 cells and primary mouse embryonic fibroblasts (MEFs). Thr112 phosphorylation was exclusively observed in mitotic cells in all cell lines indicating that this phosphorylation event is not cell type specific (Fig. 2A). By contrast, no specific staining could be observed with pThr53 mAb (Fig. 2A). The pThr112 staining was detected in prophase, and maintained until anaphase. Thereafter, the pThr112 staining declined rapidly and disappeared at the end of telophase (Fig. S1). While ATF7 signal was mainly localized within the nuclei in interphase cells, it was clearly excluded from the condensed chromatin in mitotic cells (Fig. 2A). Consistent with this observation, ATF7 resisted extraction with CSK-Triton buffer, which leaves proteins bound to the cytoskeleton, in interphasic but not in mitotic cells (Fig. S2). Using an antibody against phospho-Ser10 of core histone H3 (pH3S10), a cell cycle marker of mitosis status,²² we showed that ATF7 staining re-localized over the de-condensing chromosomes as cells entered telophase and thereby co-localized with the pH3S10 immuno-staining (Fig. 2B). Taken together, these findings support the conclusion that Thr112 phosphorylation occurs transiently during mitosis and that ATF7 is displaced from the condensed chromatin.

Figure 2.

ATF7 is phosphorylated on Thr112 during mitosis and displaced from the condensed chromatin. Images captured on confocal microscope (A) from asynchronous HeLa, Mel501 and MEF cells grown on coverslips and stained with either pThr112 mAb (green) and a polyclonal anti-ATF7 (red) antibody or (B) from asynchronous HeLa cells stained with anti-ATF7 mAb (green) and an anti-Histone H3 phospho Ser10 polyclonal antibody (red) followed by DAPI staining (blue). Merged images (green and blue) are shown in the right-hand panels. Scale bar 20 µm. (C) HeLa-ATF7 cells were synchronized using thymidine-nocodazole block. Cells were harvested at different times after block release and fractionated into soluble and insoluble moieties. Fractions were analyzed by Western blot using various ATF7 antibodies, as indicated, and anti-Hsp60 and anti-Histone H4 as fractionation controls and an anti-actin antibody as loading control. Cell cycle progression was monitored by FACS analysis (data not shown).

The correlation between the phosphorylation of the Thr112 residue and the exclusion of ATF7 from the condensed chromatin was investigated by conducting protein fractionation assays with HeLa-ATF7 cells synchronized at G2/M by a double thymidine-nocodazole block. Cells were fractionated into “soluble” and “insoluble” moieties, at different intervals following block release. As shown by immuno-blotting, ATF7 protein was undetectable in “insoluble” fractions within the 2 hours following drug withdrawal, which correlated with Thr112 phosphorylation in mitotic cells (Fig. 2C). Thereafter, Thr112 phosphorylation disappeared abruptly when cells entered interphase, and ATF7 was re-associated with the “insoluble” fractions (Fig. 2C). Importantly, the Thr112 phosphorylated form of ATF7 was exclusively observed in “soluble” fractions (Fig. 2C). As expected, the nucleosomal histone H4 remained tightly associated with “insoluble” chromatin fractions throughout mitosis, while Hsp60 partitioned almost exclusively with the “soluble” fractions.^23,24 Thus, ATF7 distributes between “soluble” and “insoluble” fractions from interphase cells, but is no longer associated with the chromatin in mitotic cells, an effect concomitant with its phosphorylation on Thr112.

Thr112 phosphorylation and ATF7 displacement from the condensed chromatin are both CDK-dependent events

To determine whether CDKs could be implicated in Thr112 phosphorylation and ATF7 chromatin displacement, we examined their sensitivity to roscovitine, a potent inhibitor of CDK2 and CDK1.²⁵ To this end, HeLa-ATF7 cells were blocked in prometaphase by nocodazole treatment in the presence or in the absence of roscovitine. In parallel, asynchronous HeLa-ATF7 cells, either exposed or not to UV radiations, were treated or not with roscovitine. “Soluble” and “insoluble” cell fractions were prepared and analyzed by Western-blotting. In “soluble” fractions from untreated cells, a strong Thr112 phosphorylation signal was detected in mitotic cells and after UV radiations, as expected (Fig. 3A, lanes 2, 3). However, Thr112 phosphorylation declined abruptly in mitotic cells treated with roscovitine compared to the untreated mitotic cells, an effect that was not seen in cells exposed to UV radiations (Fig. 3A, compare lane 5, 3, 4). Phosphorylation of the Thr53 residue after UV radiations was not significantly affected by the roscovitine treatment (Fig. 3A, lanes 2, 3, 5). In “insoluble” fractions from untreated cells, ATF7 was observed in asynchronous cells exposed or not to UV radiations but was strongly reduced in mitotic cells (Fig. 3A, compare lanes 6, 7, 8). Roscovitine treatment restored the chromatin and nuclear matrix occupancy of ATF7 in mitotic cells (Fig. 3A, compare lane 8, 10). Taken together, these data suggest that Thr112 phosphorylation and chromatin displacement of ATF7 are CDK-dependent in mitosis but CDK-independent following UV radiations.

Figure 3.

Thr112 phosphorylation and ATF7 eviction from the condensed chromatin are CDK-dependent. (A) HeLa-ATF7 cells were either exposed or not to UV radiations, or blocked for 12 hours with nocodazole (Noc) in the presence or absence of roscovitine (Ros). Mitotic cells were collected by shake-off. Cells were fractionated, and “soluble” (cytoplasm and nucleoplasm) and “insoluble” (chromatin and nuclear matrix) fractions were analyzed by Western-blotting using the pThr53, pThr112 and anti-ATF7 mAbs. An anti-actin antibody was used as a loading control, and anti-Hsp60 and anti-Histone H4 were used as fractionation controls. (B) WCE from synchronous and mitotic-arrested cells were treated or not with alkaline phosphatase (CIP) before running a DNA Affinity Precipitation Assay (DAPA) with a CRE probe and analyzed by Western-blotting with ATF7 antibodies, as indicated. (C) Purified GST-ATF7 proteins, either wild-type (WT) or mutant (T112A), were incubated with either active CDK1/cyclin B or active CDK2/cyclin A1 for 1 hour at 30°C. Kinase reactions were analyzed by Western-blotting with pThr112 mAb (upper panel) or Coomasie staining as loading control (bottom panel). (D) Myelin basic protein (MBP) was incubated with either active CDK1/cyclin B or active CDK2/cyclin A1 using the same assay conditions, in the presence of [γ−³²P] ATP. Kinase reactions were resolved on SDS-PAGE. The gel was dried and exposed with a Typhoon FLA9500.

We examine the relationship between phosphorylation events and the loss of ATF7 DNA-binding activity during mitosis by a DNA Affinity Precipitation Assay (DAPA) using a CRE-binding site. As shown in Figure 3B, ATF7 DNA-binding activity was markedly reduced in mitotic cell extracts. Alkaline phosphatase treatment restored significantly ATF7 DNA-binding activity (Fig. 3B), suggesting a direct correlation between phosphorylation events on ATF7 and its exclusion from the condensed chromatin during mitosis.

To test whether ATF7 could be directly phosphorylated by CDKs, we performed in vitro kinase assays with either CDK1/Cyclin B1 or CDK2/Cyclin A1 purified complexes and recombinant ATF7_WT or mutant ATF7_T112A as substrates. Western-blot analysis revealed that CDK1/cyclin B1 complex was most efficient to phosphorylate ATF7 on the Thr112 residue, compared to the CDK2/cyclin A1 complex (Fig. 3C). These differing phosphorylation capabilities were not seen with the myelin basic protein (MBP) as a substrate (Fig. 3D), confirming the specificity of the effect observed with ATF7. Thus, ATF7 constitutes an efficient phosphorylation substrate for the CDK1/Cyclin B in vitro.

Thr112 phosphorylation is not a prerequisite for the displacement of ATF7 from the condensed chromatin during mitosis

To further investigate the physiological relevance of Thr112 phosphorylation, we generated site-specific mAbs directed against the unphosphorylated peptide encompassing the ATF7 Thr112 residue and tested them for their ability to inhibit the phosphorylation of Thr112 in an in vitro kinase assay. Only one mAb, referred to as uThr112 mAb, was able to block the in vitro phosphorylation of Thr112 by the CDK1/Cyclin B1 complex (Fig. 4A), whereas an anti-E6 mAb had no effect. The same results were obtained in a parallel experiment using mitotic whole cell extract (WCE Noc) instead of the recombinant CDK1/Cyclin B1 complex (Fig. 4A). The uThr112 antibody did not inhibit by itself the phosphorylation of the unrelated MBP by the CDK1/Cyclin B1 complex (Fig. 4B), clearly demonstrating that inhibition was specific to ATF7.

Figure 4.

The site-specific uThr112 mAb specifically inhibits Thr112 phosphorylation in living cells. (A) Purified GST-ATF7_WT, pre-incubated in the presence or in absence of mAbs as indicated, was used for an in vitro kinase assay with either purified CDK1/Cyclin B1 complex or whole cell extracts from nocodazole-arrested cells (WCE Noc). Kinase reactions were analyzed by Western blotting (LC for light chain). (B) MBP was used in an in vitro kinase assay in the presence or absence of mAbs, as indicated, with purified CDK1/Cyclin B1 complex and [γ−³²P] ATP. Kinase reactions were resolved on SDS-PAGE and exposed with a Typhoon FLA9500. (C–E) HeLa cells were transduced with either anti-E6 or uThr112 mAbs, and grown on coverslips. Immunofluorescence staining experiments were performed 24h or 72h post-treatment. (C) Left-panels: the delivered mAbs were revealed with an Alexa 488-labeled goat anti-mouse antibody. Right-panel: the anti-E6 or the uThr112 mAbs were used as primary antibodies for conventional immuno-staining on non-transduced HeLa cells. (D) Transduced cells were stained with an Alexa 568-labeled pThr112 mAb (pThr112*; red) to visualize the Thr112 phosphorylation event in mitotic cells, and in parallel with a polyclonal-ATF7 antibody (green) and DAPI (blue). (E) HeLa cells were transduced either with the uThr112 or uThr112-NLS mAbs. 24h post-transduction, cells treated or not with CSK-Triton buffer and stained with an Alexa 488-labeled goat anti-mouse and an anti-ATF7 antibody. Images were taken by confocal microscopy. Scale bar 20 µm.

We used an intra-cytoplasmic delivery method of mAbs into living cells ²⁶ to assess the capacity of the uThr112 mAb to bind to its endogenous epitope in a cellular context. The uThr112 and the E6 mAbs, as control, were electro-transferred into HeLa cells. Cells were fixed either at 24h or 72h post-treatment and the mAbs were visualized using a fluorescent anti-mouse antibody. In parallel, both mAbs were used as primary antibodies in conventional in vitro immuno-staining assays. The anti-E6 mAb was detected in the cytoplasm 24h post-treatment and within the entire cell after 72h (Fig. 4C). Since HeLa cells do not contain HPV16 E6 protein, as confirmed by the conventional in vitro immuno-staining (Fig. 4C), the transduced anti-E6 mAb can diffuse within the nucleus only after cell division.²⁶ By contrast, the electro-transferred uThr112 mAb was partially detected in the nucleus after 24h but exhibited a predominant nuclear staining pattern 72h post-treatment, indistinguishable from the pattern provided by the conventional ATF7 in vitro immuno-staining (Fig. 4C). These results indicate that the uThr112 antibody is able to recognize the endogenous unphosphorylated Thr112 epitope of ATF7 in the cytoplasm and once bound to it, is translocated into the nucleus in a piggyback fashion.

We next investigated the ability of the uThr112 mAb to inhibit Thr112 phosphorylation in living cells and its consequence on ATF7 subcellular localization. After electro-transfer of either the uThr112 or the anti-E6 mAbs, cells were fixed at 24h or 72h post-treatment and stained with the pThr112 mAb coupled with Alexa-568 (to avoid interference with the transduced mAb for immunofluorescence detection) and with the anti-ATF7 polyclonal antibody (Fig. 4D). In mitotic cells, the intensity of the pThr112 staining was markedly reduced in uThr112 electro-transfected cells 72h post-treatment compared to either the uThr112 transduced cells at 24h, or to anti-E6 transduced cells (Fig. 4D), indicating that the uThr112 mAb inhibits ATF7 phosphorylation only within the nuclear compartment. By contrast, the transduced uThr112 mAb did not impair the UV-dependent phosphorylation of Thr53, demonstrating its strict specificity (Fig. S3). As cyclin B1 enters into the nucleus at the onset of G2-M, giving rise to an active CDK1/Cyclin B1 complex able to phosphorylate its nuclear substrates, we conclude that the transduced uThr112 mAb bound to its cognate epitope competes with this CDK1/Cyclin B1 complex for the kinase binding site of ATF7. Interestingly, this inhibition did not impair the chromatin displacement of ATF7 as shown by the polyclonal ATF7 antibody staining (Fig. 4D).

To exclude the possibility that the uThr112 mAb bound to its endogenous epitope inhibits by itself the binding of ATF7 on the chromatin, we examined the cytoskeleton-bound fractions of ATF7-uThr112 complexes in interphase cells by immunofluorescence. To be re-localized into the nucleus, the uThr112 mAb was conjugated with a peptide corresponding to SV40 nuclear localization signal (uThr112-NLS) as previously described.²⁶ HeLa cells were transduced with either the uThr112 or the uThr112-NLS antibodies. After 24h, cells were treated or not with CSK-Triton buffer and fixed for immunofluorescence. As expected (Fig. 4E), the uThr112 mAb was predominantly localized in the cytoplasm and did not resist the CSK-Triton extraction. By contrast, the uThr112-NLS mAb was mainly localized in the nucleus and resisted the extraction, indicating that the uThr112 mAb bound to endogenous ATF7 does not compromise its binding to the chromatin (Fig. 4E).

Thus, inhibiting Thr112 phosphorylation in living cells by the uThr112 mAb, does not lead to a re-localization of ATF7 onto the chromatin during mitosis, demonstrating that phosphorylation of the Thr112 residue and ATF7 displacement from the condensed chromatin are two independent events.

Phosphorylation of the Thr112 residue stabilizes ATF7 during mitosis

A computational analysis of ATF7 revealed a putative PEST domain within its N-terminal region, encompassing residues 117 to 130.²⁷ To determine whether ATF7 is stabilized in a CDK-dependent manner, HeLa-ATF7 cells were grown in the presence or in the absence of the protein synthesis inhibitor cycloheximide and the CDK inhibitor roscovitine. Extracts from cells harvested at different intervals following treatment were analyzed by Western-blotting. The level of ATF7 protein dropped abruptly 6 hours after simultaneous treatment with both roscovitine and cycloheximide, whereas it was stable over at least 8 hours in cells treated separately with anyone of these drugs (Fig. 5A and S4A), indicating that ATF7 is stabilized by phosphorylation in a CDK-dependent manner.

Figure 5.

Thr112 phosphorylation stabilizes ATF7 protein. (A) HeLa-ATF7 cells were treated or not (NT) with either roscovitine (Ros) or cycloheximide (CHX) or both (Ros + CHX) and harvested at different intervals (as indicated). Cell extracts were analyzed by Western-blotting using anti-ATF7 mAb and anti-actin antibody as a loading control. (B) HeLa cells were transiently transfected with either ATF7 wild type, Thr112A or Thr112E, as indicated, and arrested in mitosis by Nocodazole followed by CHX treatment. Cells were harvested at different time points and cells extracts were analyzed by Western-blotting using an anti-ATF7 mAb and an anti-actin antibody as a loading control. (C) HeLa cells were transduced with buffer alone (None) or with the uThr112-NLS mAb and grown on coverslips. 24h post-transduction, cells were treated with cycloheximide and roscovitine for 6 hours, and fixed for immuno-staining using an anti-ATF7 polyclonal antibody (red), an anti-mouse antibody (green) and DAPI (blue). Images were captured on immuno-fluorescent microscope. Scale bar 20 µm.

The potential implication of Thr112 phosphorylation in regulating the stability of ATF7 during mitosis was further investigated using ectopic expression of wild type, phosphorylation-deficient Thr112A or phospho-mimetic Thr112E mutants. Twenty-four hours post-transfection, HeLa cells were arrested in mitosis by nocodazole, and treated or not with cycloheximide. The turnover of the Thr112A mutant was significantly enhanced after 2 hours of cycloheximide treatment compared to ATF7 wild type and Thr112E mutant (Fig. 5B and S4B), suggesting that CDK-mediated phosphorylation on the Thr112 residue stabilizes ATF7 protein during mitosis, as does the presence of a phospho-mimetic (Glu) at position 112. Finally, the involvement of the Thr112 phosphorylation on ATF7 stabilization was further explored by transducing the mAb U-Thr112. Surprisingly, 24 hours post-transduction, the steady-state level of ATF7 in uThr112-NLS-transduced cells was markedly higher than in control cells (Fig. 5C), suggesting that the uThr112 mAb is able to stabilize ATF7 protein in living cells may be by inhibiting an interaction with an E3 ligase.

Altogether, these results demonstrate that ATF7 is stabilized in a CDK-dependent manner and that the phosphorylation on the Thr112 residue plays a crucial role in this process.

Cyclin D1 expression is regulated by ATF7 protein

Stabilization and chromatin re-localization of ATF7 during mitosis lead us to speculate its implication in regulating the expression of target genes needed for the progression in the subsequent G1 phase. It has been shown that ATF2 regulates the expression of the cyclin D1 gene in various cell lines.²⁸ Therefore, we investigated the impact of ATF7 on cyclin D1 expression. We generated targeted deletions at either ATF7 or ATF2 loci using CRISPR/Cas9 system in HeLa and Mel501 cell lines. Cells were transfected with bicistronic expression vectors encoding for human codon-optimized Cas9 (Cas9–2A-GFP) and a single-guide RNA (sgRNA) targeting endogenous sites within the ATF7 or the ATF2 loci. Following selection, cell extracts from several independent clones were analyzed by Western-blotting. As shown in Figure 6A and B, knockdown of both ATF7 and ATF2 was very efficient in both cell lines. Cyclin D1 protein level was down regulated to the same extend in ATF7 or ATF2 knockdown cell lines compared to control cell lines (Fig. 6A and B). Decrease in cyclin D1 protein levels fluctuates from 30 to 40% in the different CRISPR clones tested regardless to the parental cell lines (Fig. 6A and B). These results emphasize the implication of ATF7 to regulate cyclin D1 gene expression, as does ATF2.

Figure 6.

ATF7 regulates the cyclin D1 gene. (A and B) Editing of ATF7 or ATF2 loci in HeLa and Mel501 was engineered using CRISPR/Cas9 technology. WCE of several positive clones were analyzed by Western-blotting using anti-ATF7, ATF2, cyclin D1 and Actin antibodies. Remaining expressions of cyclin D1 quantified by Image J and normalized to actin expression, are indicated as a percentage of parental cells. (C and D) CRISPR atf7−/− and HeLa parental cells were synchronized using thymidine-nocodazole block and harvested at different times after block release. The levels of cyclin D1 protein were quantified by Image J and normalized to actin expression. Results are the average of 3 independent experiments and are expressed in arbitrary units relative to the level of cyclin D1 at 2h. (E and F) CRISPR atf7−/− cells were transfected with either ATF7 wild type, Thr112A or Thr112E, and arrested in mitosis by nocodazole treatment. WCE were analyzed by Western blotting at different times after block release. The levels of cyclin D1 protein were quantified by Image J and normalized to actin expression. Results are the average of 3 independent experiments and are expressed in arbitrary units relative to the level of cyclin D1 in GFP samples. Standard deviations and statistical significance by Student t-test are shown: NS = nonsignificant, **P < 0.01, ***P < 0.001.

Considering that ATF7 is stabilized during mitosis, cyclin D1 expression was examined at the boundary of M to G1 phase. CRISPR atf7−/− and parental HeLa cell line were synchronized using thymidine-nocodazole block. After release in fresh medium, cells were harvested at different time points and total cell extracts analyzed by Western-blotting (Fig. 6C). Cyclin D1 expression was observed in both cell lines and peaked between 6 to 8 hours after block released. However, in CRISPR atf7−/− cells, cyclin D1 expression was strongly reduced and never reached the level of the control parental cell line (Fig. 6C and D).

Finally, we investigated the impact of the Thr112 residue on the cyclin D1 expression using ectopic expression of either the wild type, the phosphorylation-deficient Thr112A or the phosphorylation-mimetic Thr112E mutant in a CRISPR atf7−/− background. Cells transfected with ATF7 derivatives were synchronized in M phase using nocodazole treatment and released in fresh medium. Levels of cyclin D1 expression were monitored at 6 and 8 hours following block release. As shown in Figure 6E and F, cyclin D1 was detectable in all cell extracts. However, expression levels were significantly up regulated in the wild type and Thr112E transfected cells compared to control and Thr112A cells. Quantitative analysis indicated that ectopic expression of the ATF7 wild type and Thr112E mutant increased cyclin D1 expression from at least 35% 6 hours after block release. These results indicate that ATF7 contributes to cyclin D1 expression and that Thr112 residue plays a crucial role in this process.

Discussion

The results presented here establish that endogenous ATF7 transcription factor is specifically phosphorylated on the Thr112 residue throughout mitosis and displaced from the condensed chromatin. Our findings validate the high-throughput proteomic screens for mitotic phospho-proteins in which the ATF7 Thr112 residue was identified as a phosphorylated site in mitotic arrested cells.^11,12

We demonstrate that ATF7 is phosphorylated on Thr112 in a CDK-dependent manner, since roscovitine, a potent CDK inhibitor ²⁵ abolishes this phosphorylation in mitotic arrested cells. In vitro kinase assays further point toward the CDK1/Cyclin B1 complex as being responsible for Thr112 phosphorylation. In addition, the roscovitine-mediated inhibition of ATF7 phosphorylation correlates with a re-localization of ATF7 onto the condensed chromatin in mitotic arrested cells but not in UV exposed-cells, demonstrating that phosphorylation of the Thr112 residue and ATF7 displacement from the chromatin are both CDK-dependent.

Phosphorylation of transcription factors at the entry into mitosis has been shown to be a key step in turning off transcription. Components of the basic transcriptional machinery, like the RNA polymerase II and TFIIH, are phosphorylated in mitotic cells and displaced from the condensed chromatin.^29,30 CREB, which belongs to the CREB/ATF transcription factor family, is also phosphorylated at the beginning of mitosis on two serine residues (Ser720/Ser721) in a CDK-dependent manner, correlating with a reduced association of CREB with chromatin during mitosis.³¹ Although the phospho-mimetic substitution of both residues by aspartates reduced CREB association with the chromatin, mutation of these two residues into alanine did not fully abolish the release of CREB from chromatin dissociation, suggesting that phosphorylation of these two residues of CREB is only part of the mechanism leading to its displacement from the chromatin in mitotic cells.³¹ AIB1, a transcriptional co-activator, which promotes pre-neoplastic changes and cancer initiation in animal models, is phosphorylated on Ser728 in mitotic cells in a CDK1-dependent way, resulting in its exclusion from the condensed chromatin.³² However, neither did phosphorylation nor point mutations of the Ser728 into alanine or glutamate impair its recruitment onto a minimal promoter and capacity to co-activate its target genes during mitosis.³² These results highlight the difficulties to establish a strict causality between phosphorylation events and displacement of transcription factors from the condensed chromatin in a cellular context.

To get further insight into the correlation between Thr112 phosphorylation of ATF7 and its exclusion from the condensed chromatin, we generated a site-specific mAb against the unphosphorylated Thr112 residue of ATF7 (uThr112), which was able to prevent Thr112 phosphorylation by the active CDK1/Cyclin B1 complex in an in vitro kinase assay. Using an intra-cytoplasmic antibody delivery,²⁶ we could show that the uThr112 mAb remains fully active and is brought to the nucleus by binding to the endogenous ATF7 epitope. Moreover, this antibody is able to specifically inhibit Thr112 phosphorylation of endogenous ATF7 protein in mitotic cells. We assume that the uThr112 mAb interacts with the kinase-phosphorylation site on ATF7, prior to the nuclear translocation of the corresponding CDK1/Cyclin B1 kinase complex. Intrabodies have been developed to dissect the role of a particular posttranslationally-modified form of a given protein in living cells.^33,34 However, to our knowledge, there is no example of a neutralizing mAb able to interfere in a dynamic cellular environment with a specific phosphorylation event by directly competing with the kinase binding to the target protein.

Interestingly, preventing Thr112 phosphorylation does not allow the re-association of ATF7 with the condensed chromatin, an observation suggesting that Thr112 phosphorylation during mitosis is not a prerequisite for ATF7-chromatin dissociation. It is likely that another, yet to be characterized CDK-dependent modification is responsible for the displacement of ATF7 from the condensed chromatin during mitosis.

Phosphorylation events have also been reported to stabilize protein steady state levels.^20,21 For example, phosphorylation of transcription factor SREBP1 on Ser439 by the CDK1/cyclin B complex stabilizes mature SREBP1, thus preserving a pool of active transcription factor to support lipid synthesis at the exit of mitosis.²⁰ Here, we demonstrate that ATF7 is also stabilized in a CDK-dependent manner. Indeed, ATF7 protein is rapidly degraded in the presence of the CDK inhibitor roscovitine and in the absence of de-novo synthesis. The phospho-mimetic mutant Thr112E is stabilized during mitosis in the presence of CDK inhibitor in contrast to the phospho-deficient mutant Thr112A, which is rapidly degraded, indicating that Thr112 phosphorylation plays a major role in this stabilization process. Similarly, in the presence of the uThr112-NLS mAb, endogenous ATF7 protein accumulates, whereas it is degraded in its absence, suggesting that the uThr112 may inhibit the interaction of an E3-ligase with ATF7 or stabilized a particular conformational state of the protein. The presence of a PEST sequence downstream of the Thr112 phosphorylation site supports these hypotheses.

It has been shown that ATF2 regulates cyclin D1 through a CRE binding site as homodimer or heterodimer with either c-jun or CREB in various cell types such as melanoma, human mammary carcinoma, HeLa cells, and chondrocytes.^35-38 Here, we demonstrate that interruption of either ATF7 or ATF2 loci decreases cyclin D1 protein expression to the same extends in two different cell lines. This strengthens the implication of ATF2 protein in targeting the cyclin D1 promoter^35-37,39 and highlights the contribution of ATF7 in such a mechanism. Moreover, expression of cyclin D1 at the boundary of M to G1 phase was strongly impaired in CRISPR atf7−/−. Finally, ectopic expression of either ATF7 wild type, Thr112A or Thr112E mutants reveals the critical role of the Thr112 residue in maintaining the expression of cyclin D1 in G1 phase. Thus, preserving a pool of active ATF7 by a CDK-dependent phosphorylation of the Thr112 residue may be one mechanism to induce cyclin D1 expression in G1 phase.

In the current study, we underscore a functional redundancy between the closely related ATF2 and ATF7 proteins, which are both able to heterodimerize with c-jun.^4,6 As cyclin D1 overexpression occurs frequently in human cancers^40,41 by gene amplification and alternative transcriptional up-regulation mechanisms ^41–43 a better understanding of the redundancy between ATF2 and ATF7 should be of great interest.

Materials and Methods

Cell culture, transfection and synchronization

HeLa and HeLa-ATF7 cells were grown as monolayer in Dulbecco's modified Eagle's medium (DMEM) supplemented with 4.5 g/L glucose and 10 % fetal calf serum (FCS) and 1 mM pyruvate. Mel501 cells were grown as monolayer in Dulbecco's modified Eagle's medium (DMEM) supplemented with 1 g/L glucose and 10 % fetal calf serum (FCS). Cells were transiently transfected using TransIT-LT1 reagent (Mirus Bio Corporation, Euromedex) according to the manufacturer's protocol. For synchronization, a double thymidine block and thymidine-nocodazole protocols were used as previously described ⁴⁴ using 2 mM thymidine (Calbiochem) and 100 ng/ml of nocodazole (Calbiochem). For enrichment of cells in metaphase, nocodazole was added at 100 ng/ml for 12 hours, and cells were collected by shake-off. Roscovitine (Calbiochem) was used to a final concentration of 20 µM and cycloheximide at 25 µg/ml.

Plasmids, stable cell line and CRISPR-cell lines

ATF7–1 cDNA and derivatives inserted into the pXJ vector, under the control of the cytomegalovirus promoter have been described elswhere.³ To generate HeLa-ATF7 cell line, HA-ATF7–1 coding sequences were introduced into the pOZ-FH-N retroviral vector⁴⁵ to transduce HeLa cells. After several rounds of selection, HeLa-ATF7 cell line was isolated. Generation of atf7−/− and atf2−/− cell lines was adapted from Ran et al.⁴⁶ using the gRNA-CAS9–2A-eGFP encoding plasmid (Addgene N°48138) coding for independent pairs, which are summarized in supplementary data Table 1. The GFP positive clones were sorted by cell sorter flow cytometer (FACS Callibur Becton Dickinson).

Cell extracts

Whole cell extracts (WCE) and fractionation were performed as previously described.⁴⁷ For fractionation, the insoluble fraction was pelleted by centrifugation at 14000 rpm for 25 min and the supernatant (designated as soluble fraction) was precipitated with cold acetone at −20°C. Both fractions were then solubilized in Laemmli buffer and the insoluble fraction (chromatin and nuclear matrix) was sonicated using a Bioruptor™ Next Gen (Diagenode).

Fluorescence microscopy and Fluorescence-activated Cell Sorter analysis

Cells were treated as previously described ¹ and mounted in DAPI-fluoromount (Southern-Biotech). Images were captured with deconvolution (confocal) fluorescence microscopy (Leica DMIRBE inverted microscope and OPENLAB 3.1.4 software) or with fluorescence microscopy (Leica DM5500B and LAS AF software). Image processing was performed with ImageJ 2.0.0. For in situ fractionation, HeLa cells were incubated in ice-cold CSK buffer (10 mM HEPES pH 6.8, 100 mM NaCl, 300 mM sucrose, 3 mM MgCl2, 1 mM EGTA) with 0.5% Triton-X100 for 5 min prior fixation. Asynchronous or synchronous cells were analyzed on fluorescence-activated cell sorter for DNA content (FACS Callibur Becton Dickinson).⁴⁸

Antibodies and electroporation

Mouse monoclonal pThr112, pThr53 specific to ATF7 phosphorylated forms, anti-ATF7 antibodies and a polyclonal anti-ATF7 antibody ⁸, rabbit polyclonal anti-actin (Sigma-Aldrich) and anti-Histone H3 phospho Ser10 (Santa Cruz Biotechnology) antibodies, mouse monoclonal anti-Hsp60 (Cell-signaling), anti-cyclin B1 and anti-histone H4 (Santa Cruz Biotechnology) antibodies, anti-cyclin D1 (Millipore), were used in this study. The pThr112 and pThr53 antibodies were labeled with Alexa568 using the DylightTM Microscale Antibody labeling kit (Thermscientific). Secondary antibodies IRDye800-conjugated anti-mouse or IRDye700-conjugated anti-rabbit antibodies (ScienceTec) were used. Membranes were scanned with the LI-COR Odyssey infrared imaging system and analyzed with the Odyssey v3.0 software (LI-COR). To generate specific non phospho-Thr112 monoclonal antibodies (mAbs), a peptide LPSTPDIKIKEEEPV corresponding to amino acids [109–123] of ATF7–3 was synthesized, coupled to ovalbumin as a carrier protein. Immunization, monoclonal antibody production and screeening were performed as previously described.^8,26,49 Electroporation of mAbs in living cells and antibody-NLS conjugates were performed as previously described.²⁶

In vitro kinase assay and DNA affinity precipitation assay (DAPA)

Cold kinase reactions were performed according to the manufacturer (New England Biolabs) using purified GST-ATF7_WT or mutant GST-ATF7_T112A together with either CDK1/cyclin B1 complex (New England Biolabs) or the CDK2/cyclin A1 complex (SignalChem). For radioactive reactions, MBP (Sigma-Aldrich) was incubated in the same conditions with the respective kinase and [γ-³²P] ATP to a final specific activity of 150 µCi /µmol. Proteins were analyzed by western blot or exposed to a phosphoimager screen and scanned with a Typhoon FLA9500. DAPA was performed as previously described ⁵⁰ using a biotinylated CRE-containing double-stranded oligonucleotide and a non-biotinylated mutated version as a competitor.

Funding

This work was supported by funds from CNRS, University of Strasbourg, French Ministry of Research, Ligue Nationale contre le Cancer, and INCa (PLBIO-2010–127).

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.

Acknowledgments

We thank C. Kedinger, PJ. Hamard and J. Diring for reading our manuscript and helpful discussions.