DEPDC1B coordinates de-adhesion events and cell-cycle progression at mitosis.

Cells entering mitosis become rounded, lose attachment to the substrate, and increase their cortical rigidity. Pivotal to these events is the dismantling of focal adhesions (FAs). How mitotic reshaping is linked to commitment to divide is unclear. Here, we show that DEPDC1B, a protein that accumulates in G2, coordinates de-adhesion events and cell-cycle progression at mitosis. DEPDC1B functions as an inhibitor of a RhoA-based signaling complex, which assembles on the FA-associated protein tyrosine phosphatase, receptor type, F (PTPRF) and mediates the integrity of FAs. By competing with RhoA for the interaction with PTPRF, DEPDC1B promotes the dismantling of FAs, which is necessary for the morphological changes preceding mitosis. The circuitry is relevant in whole organisms, as shown by the control exerted by the DEPDC1B/RhoA/PTPRF axis on mitotic dynamics during zebrafish development. Our results uncover an adhesion-dependent signaling mechanism that coordinates adhesion events with the control of cell-cycle progression.

Introduction

The cell cycle is a sequence of coordinated events leading to genome duplication and its correct segregation into the daughter cells at mitosis. The fidelity of this process is secured by mechanisms that are activated at specific restriction points: the cellular checkpoints (Gérard and Goldbeter, 2009; Hartwell and Weinert, 1989; Tyson and Novak, 2008). The G2/M checkpoint occurs at the onset of mitosis and is in charge of preserving genomic integrity and its inheritance without damage or mutations (Branzei and Foiani, 2008; Löbrich and Jeggo, 2007). The G2/M transition is driven by several mitotic kinases, including the Aurora, Polo, and the cyclin-dependent kinases (CDKs) (Hochegger et al., 2008; Lindqvist et al., 2009; Smits and Medema, 2001). The activation of the CDK1/cyclin B complex (mitosis-promoting factor [MPF]) is key in the control of mitotic entry and depends on multiple mechanisms that modulate the expression and/or localization of cyclin B and the phosphorylation status of CDK1 (Gavet and Pines, 2010; Lindqvist et al., 2009; Nigg, 2001; Norbury et al., 1991; Santos et al., 2012). Once activated, the MPF phosphorylates a series of molecular targets that trigger downstream mitotic events, such as nuclear envelope breakdown and chromosome condensation (Nigg, 2001; Ohi and Gould, 1999).

At mitotic entry, cells also become rounded, lose attachments to the substrate, and display increased cortical rigidity (Cramer and Mitchison, 1997; Kunda and Baum, 2009; Théry and Bornens, 2006). This reshaping is thought to be necessary to set the axes for symmetric or asymmetric partitioning of cell determinants and to establish a correct spindle orientation (Kunda and Baum, 2009; Théry et al., 2005).

Adhesion to the extracellular matrix (ECM) is mainly mediated by structures called focal adhesions (FAs), in which establishment, maturation, and dismantling are tightly controlled (Parsons et al., 2010; Zamir and Geiger, 2001). FAs exert a mechanostructural role by physically connecting the actin cytoskeleton to ECM via integrin receptors, and a signaling role, serving as hubs to assemble signaling complexes (Mitra and Schlaepfer, 2006; Parsons et al., 2010). As cells approach mitosis, they dismantle FAs via inactivation of FA kinase (FAK) and downmodulation of Rap1-GTPase activity (Dao et al., 2009; Kunda and Baum, 2009; Pugacheva et al., 2006; Yamakita et al., 1999). Concomitantly, cells experience mitotic rounding and cortical stiffening caused by actomyosin remodeling through RhoA (Maddox and Burridge, 2003; Matthews et al., 2012), ezrin, radixin, and moesin complex (ERM) proteins (Carreno et al., 2008), and myosin II (Maddox and Burridge, 2003).

A mechanistic picture of how the cell coordinates detachment/rounding and entry into mitosis is, however, still lacking. Here we show that DEPDC1B, a cell-cycle-regulated gene (Nicassio et al., 2005), mediates the interplay between cell-cycle progression and de-adhesion events at the mitotic entry. The DEPDC1B protein specifically accumulates at the G2 phase of the cell cycle and inhibits RhoA recruitment to and activation by the FA-associated receptor protein tyrosine phosphatase, receptor type, F (PTPRF). By this mechanism, DEPDC1B functions as an inhibitor of the RhoA/Rho-associate protein kinase (ROCK)/MLC2 pathway during the G2/M transition, thereby allowing FA dismantling and cell detachment. Ablation of DEPDC1B impaired de-adhesion events and delayed mitotic entry. Similarly, conditions that induced persistent adhesion to the substrate, independently of DEPDC1B, inhibited mitotic entry, suggesting that adhesion per se controls cell-cycle progression. Thus, we have identified a feedback loop in which the nucleus signals to cell periphery the need to initiate mitotic reshaping through the synthesis of DEPDC1B. In turn, adhesion-dependent mechanisms delay progression into the M phase until mitotic reshaping is correctly executed.

Results

DEPDC1B Regulates Mitotic Entry

DEPDC1B is a proliferation-associated gene expressed in a cell-cycle-dependent fashion through an Rb/E2F-dependent transcriptional mechanism (Nicassio et al., 2005). We examined the pattern of expression of DEPDC1B mRNA and protein in HeLa cells synchronized by double-thymidine block (D-THY; Figure S1A available online). As cells entered the G2 phase (4 hr after release), DEPDC1B mRNA was induced, and the protein accumulated until mitosis (M phase, 8 hr), closely resembling the behavior of cyclin B. In addition, similar to cyclin B, DEPDC1B protein was degraded during mitosis in a proteasome-dependent manner (Hershko, 1999) (Figure S1B).

Knockdown (KD) of DEPDC1B with three different short interfering RNA (siRNA) oligos (1B-KD1, 1B-KD2, 1B-KD3; Figures 1A, 1B, and S1C) in HeLa cells synchronized by D-THY reduced the number of cells that reached mitosis (Figures 1A–1C; Movie S1), an effect that could be rescued by the concomitant expression of a siRNA resistant GFP-tagged DEPDC1B (Figures 1B and 1C). Flow-cytometry analysis showed that DEPDC1B-KD cells progressed normally from S to G2 (G2 phase, Figure 1D), while the transition from G2 to mitosis (mitosis, Figure 1D) was inhibited. Silencing of DEPDC1B also inhibited mitotic entry in other cell types, including nontransformed and cancer cell lines (Figure S1D).

Figure 1

DEPDC1B Silencing Delays Mitotic Entry in Human Cells

(A and B) HeLa cells were synchronized in S phase by D-THY and released in fresh medium with nocodazole (100 nM) to follow cell-cycle progression from S phase to mitosis. During synchronization, endogenous DEPDC1B was silenced using different siRNA oligos (1B-KD1, 1B-KD2, or 1B-KD3). A custom non-targeting oligo was used as a control. Mitotic cells were monitored using mitotic-specific phosphorylation of histone H3 (Ser-10) as a marker (mean ± SEM of three experiments).

(B) Cells were transduced with an inducible EGFP-DEPDC1B transgene (EGFP-1B) and synchronized as in (A). Endogenous DEPDC1B was silenced using a 3′UTR-targeting oligo (1B-KD3). Upon release, siRNA-resistant EGFP-1B was induced by doxycycline (100 ng/ml), and mitotic cells were counted. Western blot shows levels of endogenous and exogenous DEPDC1B. Vinculin was used as loading control.

(C) Mitotic events were directly measured by time-lapse microscopy. The percentage of cells entering mitosis upon silencing of DEPDC1B, DEPDC1A, or both genes (KD-1A/KD-1B) is shown.

(D) Cell-cycle transitions were monitored by flow cytometry. DNA content (propidium iodide staining) of control (Ctrl) and DEPDC1B-silenced cells (1B-KD1) are shown. The percentage (mean ± SEM of two experiments) of cells in G2 phase (cyclin B positive) and M phase (pH3/cyclin B double positive) were determined.

(E and F) Lamin B/cyclin B staining and distribution (mean ± SEM of three experiments) were analyzed to monitor mitotic entry dynamics. The distribution of lamin B (membrane versus diffuse [E]) and cyclin B (nucleus versus cytosol, see F) of cells in G2 phase is shown. Asterisks mark nuclear cyclin B or membrane lamin B. Arrows mark cytosolic cyclin B.

(G) Western blot analysis of cyclin B and total/phospho-Cdk1 (Tyr14/15). Vinculin was used as loading control. The densitometry analysis is also reported (mean ± SEM of two experiments). Asterisks in graphs mark significant values (p < 0.05, Student’s t test). Representative images are shown. Scale bars represent 20 μm.

A DEPDC1B-like gene, DEPDC1A, encodes two isoforms (Figure S1E) whose expression is also regulated during the cell cycle (Figure S1F). Silencing of DEPDC1A caused a mitotic phenotype similar to that of DEPDC1B-KD (Figures 1C, S1G, and S1H). Importantly, the simultaneous depletion of both genes had additive and robust effects (Figures 1C, S1G, and S1H), arguing for functional redundancy and tight cooperative control over the G2/M transition.

We investigated the effects of DEPDC1B silencing on the key molecular events of the G2/M transition (Güttinger et al., 2009; Lindqvist et al., 2009). Upon DEPDC1B silencing, the nuclear membrane abnormally persisted in the majority of cells (lamin B staining, Figure 1E), while no differences were found in cyclin B expression (Figures 1F–1G). However, nuclear accumulation of cyclin B was decreased (Figure 1F), suggesting that activation of the MPF could be impaired. Thus, we investigated the phosphorylation status of CDK1 since dephosphorylation on Tyr14/15 is required for progression into mitosis (Hunter, 1995). The phosphorylation of CDK1 in DEPDC1B-KD cells was increased and sustained in time compared with control cells, confirming that MPF activation was delayed (Figure 1G). Together these results indicate that DEPDC1B is a regulator of the G2/M transition, acting upstream of the MPF activation.

DEPDC1B Modulates Adhesion and Actin Cytoskeleton Dynamics in G2

We employed the GFP-tagged version of DEPDC1B to analyze its subcellular distribution. In G2-syncronized cells, we observed a plasma membrane (PM) localization of DEPDC1B that persisted during mitosis (Figures S2A and S2B). In addition, while control cells lost attachment to the substrate and became rounded as they approached mitosis, DEPDC1B-KD cells appeared flattened, more motile, and often failed to detach from the substrate and become rounded (Movie S1). DEPDC1B might, therefore, act at the PM to regulate cellular adhesion. We investigated this possibility by following the dynamics of GFP-paxillin, a marker of FAs (Parsons et al., 2010; Zamir and Geiger, 2001). In control cells, the typical punctuate staining of GFP-paxillin at the ventral membrane, which marks FAs in interphase, quickly disappeared as cells approached mitosis (Figure 2A; Movie S2). Conversely, DEPDC1B-KD cells displayed larger FAs that persisted in G2 (Figure 2A; Movie S2). A quantitative analysis revealed that the absolute number of FAs per cell was unaffected; however, their size was significantly increased and their mitotic dismantling delayed (Figures 2A, 2B, S2C, and S2D).

Figure 2

DEPDC1B Silencing Perturbs Adhesion and Actin Cytoskeleton Dynamics in G2 Phase

(A and B) FA dynamics were observed using GFP-paxillin (A) or vinculin (B) as reporters in control (Ctrl) or DEPDC1B-silenced (1B-KD1) HeLa cells synchronized in early G2 phase (D-THY plus 4 hr release). (A) The mean number, size (area, μm2), and the duration of FAs, determined by total internal reflection fluorescence (TIRF) microscopy, in control (Ctrl, blue) and DEPDC1B-silenced cells (KD-1B, red) are shown. (B) The average number and area (μm2) of FAs per cell, determined by confocal microscopy using vinculin staining, are shown.

(C) The actin cytoskeleton (FITC-phalloidin) and the activation of myosin light chain (phospho-MLC2-Ser19) were analyzed by immunofluorescence and confocal microscopy in HeLa cells synchronized in G2 phase. The percentage (mean ± SEM of three experiments) of cells with high/low phospho-MLC2 staining is reported.

(D) Western blot analysis shows levels of phospho-MLC2, phospho-Cofilin (Ser3) upon DEPDC1B-KD. Vinculin, total MLC2, and Cofilin were used as loading controls. In parallel, RhoA activity was measured by GST-RBD pull-down (right panel). As positive control, the lysate was activated by GTPγS stimulation. An asterisk marks a nonspecific band detected by the anti-RhoA antibody. The densitometry analysis of active RhoA in two experiments is shown.

(E) IMR90 and MCF10A cells were analyzed in G2 phase for actin cytoskeleton and FA dynamics as described in (B) and (C). Western blot analysis shows levels of phospho-MLC2 (Ser19) upon DEPDC1B silencing.

(F) Cell spreading dynamics, on fibronectin substrate, of G2 synchronized HeLa cells was monitored by Real Time Cell Analyzer (Atienzar et al., 2011). Graph shows the normalized cell index, as a measure of cell-covered area after replating. Asterisks in graphs mark significant values (p < 0.05, Wilcoxon test in A and B; Student’s t test in C). Representative images are shown. Scale bars represent 10 μm.

We also detected significant modifications of actin dynamics in DEPDC1B-KD HeLa cells in G2 phase, with cells displaying an altered pattern of actin stress fibers and increased phosphorylation of the actin regulator myosin light chain 2 (MLC2-Ser19), a typical downstream target of the ROCK, and cofilin (pCofilin-Ser3) (Figures 2C and 2D). These observations could be extended to other cell types, including fibroblasts and nontransformed epithelial cells (Figure 2E). Finally, the silencing of DEPDC1B also altered the dynamics of cells spreading, an effect that could be rescued by the ectopic expression of the siRNA-resistant GFP-DEPDC1B (Figures 2F and S2E).

These results point to a role for DEPDC1B in the control of cellular adhesion and actin dynamics during the G2 phase of the cell cycle.

DEPDC1B-KD Induces an Adhesion-Dependent Checkpoint at the G2/M Transition

We investigated the relationship between the mitotic and the adhesion phenotypes caused by DEPDC1B silencing. Initially, we took advantage of a HeLa derivative clone (HeLa-S3) adapted to growth in suspension (Puck et al., 1956) (Figure 3A). In these cells, DEPDC1B-KD did not affect the G2/M transition (Figure 3B), arguing that in the absence of cell adhesion the mitotic phenotype of DEPDC1B-KD cells could be suppressed. If so, it should be possible to abrogate the said phenotype by directly interfering with FAs. Thus, we silenced structural (vinculin, alpha-actinin) and catalytic (FAK) components of FAs in DEPDC1B-silenced cells (Figure 3C). In all cases, the downmodulation of FA components completely rescued the DEPDC1B-KD-dependent mitotic delay (Figure 3D), indicating that, in DEPDC1B-KD cells, the cell-cycle phenotype is linked to the abnormal persistence of FAs at the G2/M transition phase.

Figure 3

DEPDC1B Controls an Adhesion-Dependent Checkpoint at the G2/M Transition

(A and B) HeLa (grown in adhesion) and HeLa-S3 (grown in suspension) cells were treated with control or DEPDC1B siRNAs and synchronized (D-THY) to follow cell-cycle progression. (B) Percentage of cyclin B-positive (G2 phase) and pH3/cyclin B double-positive (mitosis) cells were determined by flow cytometry.

(C and D) Control (Ctrl) or DEPDC1B-silenced (1B-KD1) HeLa cells were treated either with vinculin, FAK, or alpha-actinin siRNA oligos. (D) Mitotic events were measured using pH3 marker (mean ± SEM of three experiments).

(E and F) Control (Ctrl) or DEPDC1B-silenced (1B-KD1) HeLa cells were synchronized as in (A) and treated with the ROCK kinase inhibitor (Y27632, 10 μM) upon release. Mitotic cells were measured using pH3 staining (E) or by time-lapse microscopy (F). Western blot analysis shows levels of phospho-MLC2 on DEPDC1B-KD and Y27632 treatment. Vinculin and total MLC2 were used as loading controls.

(G) HeLa cells were synchronized (as in A) and treated upon release with Mn2+ (1 mM, upper panel). In the lower panel, HeLa cells were transfected with pEGFP-vinculin (VincT12 mutant) during synchronization. Western blot analysis shows levels of phospho-MLC2 and phospho-CDK1 upon Mn2+ treatment and VincT12 expression. RhoA activation by Mn2+ treatment was measured by RBD pull-down. An asterisk marks a nonspecific band detected by the anti-RhoA antibody.

(H and I) Mitotic cells were measured using pH3 staining (H and I) or cyclin B distribution (H). Scale bar graphs report the mean ± SEM of three (C and D) or two (B, E, H, and I) experiments. Asterisks mark significant values (p < 0.05, Student’s t test).

We also analyzed the involvement of actomyosin contractility, which was altered upon DEPDC1B silencing (see Figure 2). As mentioned above, this process is closely linked to cell adhesion mechanisms and is controlled by RhoA/ROCK/MLC2 signaling. Therefore, we treated G2 synchronized DEPDC1B-KD cells with a ROCK inhibitor (Y27632, 10 μM). The treatment normalized the levels of phospho-MLC2 and concomitantly rescued the mitotic phenotype in DEPDC1B-KD cells (Figures 3E and 3F).

If the impairment in G2/M transition observed in DEPDC1B-KD cells were indeed due to the persistence of FAs, then induction of persistent adhesive structures should phenocopy the DEPDC1B silencing. To investigate this, we employed two tools: (1) an autoinhibition-deficient mutant of vinculin (VincT12 mutant), which increases adhesion strength and force transmission (Humphries et al., 2007), and (2) manganese treatment (Mn2+, 1 mM), which induces ανβ3 integrin activation and clustering (Cluzel et al., 2005; Gailit and Ruoslahti, 1988). In G2-synchronized HeLa cells, both treatments induced cell spreading on the substrate, formation of actin stress fibers, and high levels of phospho-MLC2 (Figures 3G and S3A–S3D), while concomitantly inhibiting the G2/M transition (Figures 3H–3I and S3E).

These results suggest the existence of a DEPDC1B-based mechanism that controls the coordination of adhesion and actin cytoskeleton dynamics with entry into mitosis. Of note, the G2/M arrest induced by DNA damage-inducing agents (i.e., doxorubicin) appeared stronger than the cell-cycle arrest induced by persistent adhesion (Figures S3E and 3F).

DEPDC1B Modulates RhoA-Dependent Cell Adhesion at G2/M Transition

Since both actin cytoskeleton and adhesion dynamics are regulated by Rho-GTPases, we silenced the expression of each of the three prototypical members of this family, RhoA, Rac1, and Cdc42, alone and in conjunction with DEPDC1B-KD. The depletion of any of the three Rho-GTPases alone had no major effect on G2/M transition (Figure 4A). However, RhoA silencing, but not Rac1 or Cdc42 silencing, completely rescued the mitotic delay induced by DEPDC1B silencing (Figures 4A and 4B). Furthermore, the significant increase in the size of FAs in DEPDC1B-KD cells (see Figures 2A and 2B) was fully rescued by silencing RhoA (Figures 4C–4E). Finally, the DEPDC1B-KD-dependent spreading defect was rescued by silencing of RhoA, but not of Rac1 or Cdc42 (Figure S4A). These results link the function of DEPDC1B to the control of RhoA activity, likely through inhibition of the latter at the G2/M transition. This notion is further supported by the observations that (1) the DEPDC1B silencing increased the activity of RhoA (RBD pull-down assay, Figure 2D), (2) the levels of downstream targets, such as phospho-MLC2 (Figures 2C–2E), (3) RhoA overexpression in G2 cells phenocopied DEPDC1B KD, causing an increase in the size of FAs and in the levels of phospho-MLC2 (Figures 4F–4G) and a decrease in the mitotic index (Figure 4H).

Figure 4

DEPDC1B Controls RhoA-Dependent Cell Adhesion at the G2/M Transition

(A and B) Control (Ctrl) or DEPDC1B-silenced (1B-KD1) D-THY synchronized HeLa cells were treated with RhoA, Rac1, or Cdc42 siRNA oligos. Mitotic cells were measured using pH3 staining (A) or by time-lapse microscopy (B).

(C–E) HeLa cells were treated with control, DEPDC1B and RhoA siRNA oligos as indicated and synchronized in G2 phase. Staining for pH3 was used to distinguish G2 from mitotic cells. G2 cells were examined for FAs (using vinculin as marker) and actin cytoskeleton (Phalloidin). (D and E) The distribution of FA mean area per cell (D) or percentage of cells displaying enlarged FAs and actin stress fibers (E) are shown.

(F–H) HeLa cells were transfected with RhoA-myc and synchronized in G2 (F and G) or M phase (H). (F) Western blot analysis shows levels of RhoA-myc and phospho-MLC2. Total MLC2 was used as loading control. In parallel, FA dynamics were analyzed by vinculin staining. (G and H) Bar graphs show the percentage of cells with normal/enlarged FAs and actin stress fibers (G) and the percentage of pH3-positive cells (H). Scale bar graphs report the mean ± SEM of three (A, D, and E) or two (F and H) experiments. Asterisks mark significant values (p < 0.05, Student’s t test). Representative images are shown. Scale bars represent 10 μm.

PTPRF Controls RhoA-Dependent Signaling at Mitotic Entry

Despite having a RhoGAP-like (GTPase-activating) domain (Figure S1E), DEPDC1B is most likely not an active GAP since it lacks the catalytic arginine typical of true RhoGAPs (Graham et al., 1999; Rittinger et al., 1997) (Figure S4B). Indeed, we failed to detect RhoGAP activity of the recombinant RhoGAP domain of DEPDC1B (Figure S4C).

To understand how DEPDC1B modulates RhoA signaling, we performed a yeast-two-hybrid (Y2H) screening to identify DEPDC1B-interacting proteins. Most of the hits were represented by PTPRF (Figure 5A; Table S1), a transmembrane receptor that has been suggested to function as a molecular hub at adhesive sites that coordinates adhesion and migration events (Chagnon et al., 2004; Serra-Pagès et al., 1995; Tsujikawa et al., 2002). We confirmed the direct biochemical interaction between the DEPDC1B and PTPRF by glutathione S transferase (GST) pull-down experiments performed with cell lysates (Figures 5B and 5C) or purified proteins (Figure 5D). The interaction was also confirmed in intact cells by coimmunoprecipitation of GFP-tagged DEPDC1B and overexpressed PTPRF (Figure S5A) and by colocalization of the two proteins at the PM (Figure S5B).

Figure 5

The Membrane Receptor PTPRF Interacts Biochemically and Genetically with DEPDC1B and RhoA

(A) An Y2H screen was employed to search for DEPDC1B interactors. The chart shows the distribution of positive clones.

(B and C) GST-PTPRF-c (cytoplasmic fragment, 1 μM) was incubated with total lysates (2 mg) from HeLa cells overexpressing EGFP-1B (B) or from control HeLa cells (C). Western blot analysis was performed using anti-GFP or anti-DEPDC1B antibodies. Asterisk marks a non-specific band detected by the anti-GFP. Ponceau staining is shown.

(D) GST-PTPRF-c (1 μM) was incubated with the purified GAP domain of DEPDC1B (1B-GAP, 3 μM). Proteins were resolved by SDS-PAGE and detected by Coomassie staining to reveal the amount of 1B-GAP pulled down by PTPRF-c.

(E and F) Control (Ctrl) or DEPDC1B-silenced (1B-KD1) HeLa cells were synchronized and treated or not with PTPRF siRNA. (E) Mitotic cells were measured using pH3-staining (mean ± SEM of three experiments). Asterisks mark significant values (p < 0.05, Student’s t test). (F) Western blot analysis shows levels of DEPDC1B, PTPRF, and phospho-MLC2. Vinculin and total MLC2 were used as loading controls.

(G) Mass spectrometry analysis of proteins pulled down using GST or GST-PTPRF-c identified several members of RhoA signaling pathway (listed in the table) as PTPRF interactors.

(H) Coomassie Staining of GST-PTPRF-c pull-down assay (1 μM), using purified RhoA (3 μM).

(I and J) HeLa cells were transfected with RhoA-myc, treated with control or PTPRF-targeting siRNA oligo, and synchronized in S phase to follow mitotic entry. (I) Western blot analysis shows levels of RhoA-myc, PTPRF, and vinculin (loading control). (J) Scale bar graphs report the percentage of cells entering mitosis (mean ± SEM of two experiments) using pH3 staining as a mitotic marker. Asterisks mark significant values (p < 0.05, Student’s t test).

Next, we investigated whether PTPRF is involved in the DEPDC1B-dependent phenotypes. PTPRF-KD impaired cell spreading onto fibronectin, as previously shown (Figure S5C; Asperti et al., 2009), while it did not cause appreciable effects on the rate of mitotic entry (Figures 5E and 5F). However, PTPRF silencing was able to rescue both the mitotic delay and the increase in phospho-MLC2 levels induced by DEPDC1B silencing (Figures 5E and 5F), suggesting its participation in the DEPDC1B-dependent mechanisms that control actin dynamics at the G2/M transition. PTPRF phosphatase activity appeared not to be involved since PTPRF interaction with DEPDC1B was not affected by phosphatase treatment and treatment with inhibitors of PTPRF phosphatase activity did not significantly affect mitotic entry alone or together with DEPDC1B silencing (Figures S6A and S6B).

To gain further insights into the role of PTPRF, we examined the PTPRF interactome by mass spectrometry. The list of PTPRF-interacting proteins was significantly enriched in components of the RhoA/ROCK pathway (p < 0.001; Figures 5G and S5D), including RhoA itself and several RhoA-binding partners, such as its effectors ROCK2 and mDIA1, as well as members of the actin network (Table S2). The direct interaction between PTPRF and RhoA was confirmed in in vitro pull-down experiments, largely independently of the activation status of the latter (Figures 5H and S5E) and by colocalization at the PM (Figure S6C). Importantly, the silencing of PTPRF reversed the mitotic defect induced by RhoA overexpression, albeit not completely (Figures 5I and 5J). Among PTPRF-interacting proteins, we identified several guanine nucleotide exchange factors (GEFs), and most notably GEF-H1, suggesting that the PTPRF/GEF-H1 axis could be involved in RhoA activation at FAs.

This contention is reinforced by the observations that (1) PTPRF localizes at the PM close to FA markers (vinculin, paxillin) and to RhoA (Figures S6C–S6E); (2) the interaction between GEF-H1 and PTPRF was confirmed in a pull-down assay (Figure 6A); (3) silencing of GEF-H1, but not of another GEF, PDZ-GEF, which was pulled down by PTPRF, rescued the mitotic phenotype induced by DEPDC1B silencing (Figure 6B); (4) overexpression of GEF-H1 phenocopied the effect of RhoA activation or DEPDC1B silencing, inhibiting mitotic entry and inducing actin stress fibers (Figure 6C).

Figure 6

PTPRF as a Hub for the Activation of the RhoA Signaling Complex, which Is Dependent on DEPDC1B Protein Levels

(A) GST-PTPRF-c pull-down assay with total lysates (1 mg) from HeLa control cells or cells overexpressing 1B-GFP. Western blot analysis shows levels of GEF-H1 and 1B-GFP.

(B and C) The effects of silencing or overexpression of GEF-H1 has been analyzed on HeLa cells. Bar graphs report the percentage of cells entering mitosis (mean ± SEM of two experiments) using pH3 staining as mitotic marker. Asterisks mark significant values (p < 0.05, Student’s t test). (B) A bar graph on the right shows the relative mRNA expression of another RhoA GEF, PDZ-GEF, upon silencing. (C, below) Western blot shows levels of GEF-H1 upon silencing or overexpression and vinculin (loading control). (Above) Representative images of FAs (vinculin) and actin cytoskeleton (phalloidin) of control HeLa cells or cells overexpressing GEF-H1. Scale bars represent 10 μm.

(D–G) GST-PTPRF-c pull-down assays were performed with total lysates (1 mg) from control cells, cells overexpressing both 1B-GFP and RhoA-GFP (D), cells overexpressing just 1B-GFP (E and F), cells overexpressing RhoA-myc and silenced for endogenous DEPDC1B (G). Western blot analysis was performed using the anti-GFP antibody for both 1B-GFP and RhoA-GFP, the anti-myc antibody for RhoA-myc, the anti-mDIA1 or anti-ROCK2 antibodies (RhoA downstream effectors), or anti-DEPDC1B antibody. The asterisk marks a nonspecific band detected by the anti-GFP antibody (D and F). In (E), protein bound to PTPRF-c were identified and quantified by SILAC in presence or not of DEPDC1B-GFP overexpression (two independent experiments, forward and reverse, with swapping of isotope-encoded amino acid among the two channels). Among PTPRF interactors, 39 proteins were found with binding consistently upmodulated or downmodulated by DEPDC1B overexpression (>1 log2 fold), such as DEPDC1B itself and RhoA.

(H) Western blot analysis shows levels of ROCK2 (RhoA effector) bound to active RhoA (RBD-pull-down) upon DEPDC1B silencing or Mn2+ treatment. As positive control, the lysate was activated by GTPγS stimulation.

(I) GST-PTPRF-c (1 μM) was incubated with purified RhoA (3 μM) and/or 1B-GAP (3 μM). Proteins were resolved by SDS-PAGE and detected by Coomassie staining to reveal the amount of protein pulled down by PTPRF-c. The densitometry analysis of the PTPRF-c/RhoA (and 1B-GAP) interaction in three experiments is shown.

(J) The scheme depicts the working model for DEPDC1B as a competitive inhibitor of the RhoA/PTPRF interaction resulting in inhibition of RhoA-dependent signaling that regulates the adhesion dynamics of G2 cells and mitotic entry. Positive (left) and negative (right) inducers of the G2/M adhesion checkpoint are also reported, according to the findings of this study.

RhoA:PTPRF Interaction Depends on DEPDC1B Levels

The sum of our results strongly supports a model in which RhoA/ROCK signaling in G2 phase is induced by PTPRF/GEF-H1 and is inhibited by DEPDC1B. One mechanism through which this might occur is competition of the interaction of RhoA with PTPRF by DEPDC1B, with ensuing inhibition of the RhoA signaling complex and the consequent dismantling of adhesion structures at the G2/M transition. We tested this hypothesis through a series of experiments. We showed by in vitro pull-down assays performed on total cellular lysates that the interaction between the cytoplasmic domain of PTPRF (GST-PTPRF-c) and RhoA was almost completely inhibited by the simultaneous presence of excess DEPDC1B (obtained by ectopic expression) in the cell lysate (Figure 6D). This effect on PTPRF and RhoA binding was also reproduced in stable isotope labeling by amino acids in cell culture (SILAC) experiments (Figure 6E) and appears specific for RhoA itself since binding of GEF-H1 (RhoA activator), ROCK2, and mDIA1 (RhoA effectors) to PTPRF were unaffected by DEPDC1B overexpression (Figures 6A and 6F). Conversely, silencing DEPDC1B significantly increased the interaction between GST-PTPRF-c and RhoA in the same assay (Figure 6G) and the interaction between RhoA and ROCK2 in the RBD pull-down assay (Figure 6H). Finally, the purified DEPDC1B fragment that interacts with PTPRF-c (Figure 5D) halved the interaction of the latter with RhoA in in vitro pull-down experiments (Figure 6I).

These results suggest a role for the DEPDC1B-PTPRF axis in the control of RhoA signaling at mitotic entry, with DEPDC1B acting as negative regulator of the recruitment of RhoA to PTPRF-based complexes (Figure 6J).

DEPDC1B Modulates Cell Proliferation during Zebrafish Embryogenesis

To analyze the relevance of the DEPDC1B/RhoA/PTPRF axis at the organismal level, we turned to zebrafish, which express a DEPDC1B ortholog (depdc1b) that has an exon/intron organization conserved with the human gene (Figure S7A) and ∼75% similarity to the human protein. By in situ hybridization and RT-qPCR of zebrafish embryos, we showed that depdc1b mRNA has no maternal contribution, first appeared at 3 hr postfertilization (hpf), and gradually accumulated until the end of segmentation period (32 hpf), with more intense expression detected in the head and dorsal regions (Figures 7A and S7B).

Figure 7

DEPDC1B Controls Mitotic Events in Zebrafish Embryo

(A) Zebrafish (ZF) embryos were collected at different times from fertilization (hpf) and levels of depdc1b or depdc1a mRNA expression measured by RT-qPCR. Actin B (Actb) was used as a normalizer.

(B) ZF embryos were injected with control or depdc1b splice-blocking morpholino (MOSB) at the one-cell stage. Pictures show representative images of the resulting morphological defects observed at the late gastrula (10 hpf, left) or late segmentation (24 hpf, right) stages. Scale bars represent 125 μm (left) and 500 μm (right).

(C–I) Zebrafish embryos were treated as in (B), and mitotic figures were monitored in the embryonic dorsal region of the Tg(h2afva:GFP)kca6 zebrafish transgenic line using confocal microscopy. Representative images are shown in (C).The scale bar represents 10 μm.

(D) The graphs show the number of mitotic events (mean ± SD of two experiments) occurring from 4 to 6 hpf. DEPDC1B human mRNA was microinjected together with depdc1b MOSB to rescue embryo defects.

(E) Duration of mitotic phases (mean ± SEM) was measured by confocal time-lapse microscopy, after MBT.

(F) The graphs show the number of mitotic cells at 10 hpf determined by immunofluorescence upon various treatments (mean ± SEM of two experiments).

(G) Cortical actin cytoskeleton was analyzed in the anterodorsal region of embryos treated as in (D), at 10 hpf by FITC-phalloidin staining. Representative images are shown. Red arrows mark thickening and actin protrusions on cell edges. hs-1B, human DEPDC1B mRNA. The scale bar represents 10 μm.

(H) Bar graphs show the number of cell contacts (left panel) and actin protrusions observed in two independent experiments (mean ± SEM). Asterisk marks significant values (p < 0.05).

(I) Embryos were treated as in (D), with different splice-blocking morpholinos (MOSBs for depdc1b, rhoab, or ptprf) at the one-cell stage. The graphs show the number of mitotic events (mean ± SD of two experiments) occurring from 4 to 6 hpf.

The function of Depdc1b during zebrafish development was investigated by taking advantage of a specific splice-blocking morpholino (MOSB) to induce the formation of a truncated protein (Figure S7C). The injection of a depdc1b-MOSB caused a severe morphological defect, already visible at the late gastrula stage (10 hpf; Figure 7B), possibly due to alterations in morphogenetic mechanisms, and finally displaying a phenotype at 24 hpf characteristic of mutations with delayed or incomplete epiboly (Kane et al., 1996; Figure 7B). We investigated the effects of Depdc1b ablation on cell proliferation by measuring the mitotic rate in the anterodorsal side of the embryo, where Depdc1B is preponderantly expressed. We used the Tg(h2afva:GFP)kca6 transgenic line (Pauls et al., 2001), in which the fusion protein histone variant H2A.F/Z:GFP is expressed from the start of zygotic transcription, to monitor mitosis. Real-time analysis in the dorsal region after midblastula transition (MBT, ∼4 hpf; Figures 7C and 7D) scored a defect upon depdc1b-MOSB injection in mitotic events occurring between 5 and 6 hpf. We did not detect alterations in mitotic duration per se (Figure 7E), suggesting that the defect in mitotic peaks was caused by a premitotic defect, possibly in the control of the G2/M transition as observed in mammalian cells. At 10 hpf the developmental defects of depdc1b-morphant embryos were mirrored by an increase in proliferation (number of pH3+ cells) of anterodorsal region (Figure 7F). At this stage, we also observed a series of cytoskeleton related phenotypes, such as cells with much irregular shape, thickening of cortical actin, and frequent appearance of actin protrusions on cell edges (Figures 7G and 7H). All of these defects were rescued by the coexpression of human DEPDC1B mRNA, suggesting that the phenotype was specifically caused by Depdc1b ablation and that Depdc1b function is conserved (Figures 7F–7H and S7D). To corroborate this possibility, we tested whether the DEPDC1B genetic interactions identified in mammals were conserved in zebrafish, by analyzing the effect of ablation of the RhoA and PTPRF zebrafish orthologs (rhoab and ptprf, respectively). KD of either rhoab or ptprf by MOSB injection (Figure S7C) rescued all of the defects of the depdc1b morphants observed at 6, 10, and 24 hpf (Figures 7I, 7F, and S7D), suggesting that the DEPDC1B/RhoA/PTPRF axis is conserved. Of note, neither the isolated DEP domain nor the pseudo-RhoGAP domain of DEPDC1B was sufficient to rescue the MOSB phenotype, suggesting that both domains are required for complementing DEPDC1B functions in vivo (Figure S7E).

Discussion

We have uncovered a feedback mechanism of communication between the nucleus and the cell periphery that is centered on DEPDC1B and that allows the coordination and control of a series of events critical to the correct execution of the mitotic program. Mechanistically, this occurs through the RhoA-dependent regulation of FA clustering and actin dynamics and the PTPRF-dependent regulation of the RhoA signaling complex in proximity to FAs. The impact of the DEPDC1B/RhoA/PTPRF circuitry is not limited to 2D cell culture settings but extends to real 3D situations, as observed in zebrafish development.

Signaling from the Nucleus to the Cell Periphery

The levels of DEPDC1B oscillate during the cell cycle with behavior and mechanisms indistinguishable from those of checkpoint proteins, such as cyclin B. This cyclin-like regulation is critical for mammalian cell proliferation. Indeed, a lack of DEPDC1B produces two clear phenotypes: (1) a cell-cycle effect, consisting in a significant delay in the transition to mitosis (an effect greatly augmented by the concomitant depletion of DEPDC1A), and (2) an adhesion/cytoskeleton phenotype, with cells displaying enlarged and persistent FAs and aberrant actin stress fibers formation. The two phenotypes are intimately connected, since under conditions that promote de-adhesion (or upon growth in suspension) the mitotic defect is completely bypassed. Accordingly, impairment of the de-adhesion process, produced by outside-in (integrin clustering and increased force transmission) or inside-out (RhoA ectopic expression) mechanisms, resulted in a defect in cell-cycle progression that was comparable in all experimental conditions (Figure S3E and summarized in Figure 6J). We conclude that we have identified an “adhesion-dependent checkpoint” that participates in the regulation of the G2/M transition and affects most, but not all the cells (a few still reach mitosis). Thus, the “adhesion-dependent checkpoint” appears a little less strong than a typical G2 checkpoint, such as the one induced upon DNA damage, which usually affects all the cells.

Signaling at the Cell Periphery

At the PM, two major processes contribute to cell shape remodeling of cells that enter mitosis: FA dismantling (de-adhesion) and cortical stiffening (mitotic rounding). The process of de-adhesion was previously shown to be under the control of the Rap1-GTPase, whose cell-cycle-specific downmodulation, by an as-yet-unknown mechanism, is critical to inhibit the formation of new FA sites and to shut down integrin signaling (Dao et al., 2009). Conversely, mitotic rounding does not apparently involve attenuation of Rap1 signaling, as it was shown that cells expressing a Rap1 dominant-negative mutant (Rap1∗) could enter mitosis and displayed some kind of rounding and contractility while remaining attached to the substrate (Dao et al., 2009; Lancaster et al., 2013). Cell rounding, instead, is thought to be due to RhoA- and ERM-dependent control of cortical rigidity (Kunda and Baum, 2009; Kunda et al., 2008; Maddox and Burridge, 2003; Matthews et al., 2012). Therefore, it is likely that two pathways, one regulating FA dismantling (Rap1 dependent) and the other regulating mitotic rounding (RhoA and ERM dependent), exist and cooperate to induce cell reshaping at mitosis. Our data argue that both these pathways are under the control of DEPDC1B.

The control on FA dynamics exerted by DEPDC1B is due to its ability to bind to PTPRF and to compete specifically with RhoA for binding to this hub, thereby inhibiting activation of the RhoA signaling (likely due to GEF-H1) and actomyosin contractility at adhesion sites. It remains to be established how the DEPDC1B/RhoA/PTPRF axis connects with and/or controls Rap1-dependent de-adhesion. The role of DEPDC1B in cortical actin dynamics, on the other hand, presents us with an apparent paradox, since DEPDC1B is a functional inhibitor of RhoA signaling at PTPRF-based hubs, while the increase in cortical rigidity needed for mitotic rounding requires active RhoA. This latter event occurs just before prophase, through the Rho-GEF ECT2 (Matthews et al., 2012). Based on our results, a simple reconciling scenario might be that RhoA, once displaced by DEPDC1B from PTPRF sites during late G2, becomes available for ECT2-dependent activation at other sites of the PM later in the cell cycle when cells are in prophase. The precise order of events should be guaranteed by the fact that lack of DEPDC1B accumulation impacts on MPF activation, which, in turn, controls ECT2 translocation from the nucleus to the PM (Matthews et al., 2012).

Signaling from the Cell Periphery to the Nucleus

Our data show that DEPDC1B not only coordinates FA disassembly with mitotic rounding at the PM but also signals to the nucleus to time entry of cells into mitosis, acting at a point upstream of the MPF complex. This signaling appears dependent on the decrease of RhoA activity at FAs, as suggested by the comprehensive analysis of our molecular genetics and biochemical evidence. The exact nature of the molecular network linking adhesion signaling to nuclear MPF activation at the G2/M transition remains to be elucidated. Possible candidates include members of MAPK-ERK cascade, which are known to have a role in cyclin B activation during mitotic entry (Patel et al., 1999; Wang et al., 2007), as well as to be activated upon FA establishment and stabilization (reviewed in Margadant et al., 2013).

In summary, we uncovered a mechanism coordinating cell adhesion with cell-cycle progression that acts as a cell-cycle “adhesion-dependent checkpoint” and demonstrated the relevance of this mechanism in vivo. Of note, the molecular components of this checkpoint, including DEPDC1B, PTPRF and RhoA, are frequently altered in cancer. Thus, the subversion of the adhesion-dependent checkpoint might also be relevant to human pathology: a possibility that warrants further investigations.

Experimental Procedures

Cell-Cycle Synchronization and Procedures

HeLa, Phoenix, and U2OS cells were cultured in Dulbecco’s modified Eagle’s medium plus fetal bovine serum (10% v/v), l-glutamine (2 mM), and sodium pyruvate (1 mM). MCF10A cells were cultured in DMEM/F12 medium plus horse serum (final concentration: 5% v/v), insulin (10 mg/ml), epidermal growth factor (20 ng/ml), cholera toxin (100 ng/ml), and hydrocortisone (0.5 mg/ml). IMR90 fibroblasts were cultured in Eagle’s minimum essential medium plus fetal bovine serum (10% v/v), l-glutamine (2 mM), and nonessential amino acids (1% v/v). HeLa S3 cells were seeded on poly-HEMA-coated plates (P3932 from Sigma, final concentration: 12 g/l). Synchronization in G2/M phase was performed by D-thymidine treatment (final concentration: 2.5 mM; Sigma) for 16–18 hr followed by release in complete medium with nocodazole (final concentration: 200 ng/ml; Sigma). Synchronization in M phase was performed by nocodazole treatment for 16 hr, followed by mitotic shake-off and replating on poly-D-lysine-coated plates (final concentration: 15 μg/ml). For adhesion experiments, cells were plated on fibronectin-coated coverslips (final concentration: 10 μg/ml) and monitored by immunofluorescence (FITC-phalloidin) or by Real-Time Cell Analyzer Technology (xCELLigence Roche). Integrin activation by Mn2+ was achieved by treating cells with MnCl2 (final concentration: 1 mM). Y27632 was purchased from Merck (688001) and used in mitotic experiments at a final concentration of 10 μM. DNA damage was induced by 1 hr doxorubicin (final concentration: 10 μM) after 2 hr of release from D-thymidine treatment. RWJ-60475 (AM)3 was purchased from Abcam (ab141729) and used at a final concentration of 10 μM. DNA constructs expressing DEPDC1B, RhoA, PTPRF, and GEF-H1 were transfected into Phoenix cells using calcium phosphate or into HeLa cells by Lipofectamine 2000. Details are reported in the Supplemental Experimental Procedures. Custom or predesigned siRNA oligos were used to silence endogenous expression of DEPDC1B, DEPDC1A, vinculin, actinin, FAK, RhoA, Rac1, Cdc42, and PTPRF (see Supplemental Experimental Procedures for siRNA sequence information). A custom nontargeting oligo was used as a negative control. HiPerfect Transfection Reagent from QIAGEN was used for siRNA transfection (final oligo concentration: 50 nM), according to the manufacturer’s instructions.

Microscopy and Data Analysis

Mitotic and adhesion phenotypes were quantified by epifluorescence microscopy under a fluorescent (DM5500B; Leica) microscope using a 20×/NA 0.15 objective lens at room temperature. All antibodies used for immunofluorescence were employed according to manufacturer’s instructions (a complete list is reported in the Supplemental Experimental Procedures). Fluorochromes used were Cy3 (cyclin B, lamin B, vinculin, pMLC2, Flag-tag), Alexa488 (pH3), FITC-/TRITC-phalloidin, or Cy5 (vinculin [pH3], Myc-tag). All images were acquired with a camera (DCF350FX; Leica) and LAS-AF image software (Leica). Images were processed with the same settings (brightness, contrast, crop, image size) using Adobe Photoshop CS5.1, and imaged figures were constructed in Adobe Illustrator.

Confocal analysis of DEPDC1B localization, actin cytoskeleton, and FAs was performed on a Leica TCS SP2 AOBS microscope, using a 40×/NA 1.25 oil-immersion objective and processed in Adobe Photoshop. Images were taken with identical settings, and the number and average area of FAs per cell were determined using ImageJ software with a mask with a fixed threshold that identifies vinculin- or paxillin-GFP positive FAs. Details on live microscopy (time-lapse and TIRF) are in the Supplemental Experimental Procedures.

Protein Purification and Pull-Down Experiments

pGEX-SH3BP1 (GAP domain) and pGEX-RhoA constructs were a kind gift from Dr. Giorgio Scita (IFOM). pGEX-LAR (cytoplasmic tail) was a kind gift from Professor Axel Ullrich (Max Planck Institute). GST-tagged proteins were expressed in pLysS BL21 bacterial strains upon IPTG induction (0.5 mM) and purified by glutathione sepharose beads (GE Healthcare). For GST pull-down experiments, RhoA was further solubilized by 3C-Prescission Protease treatment. The pseudo-RhoGAP domain (aa 177–400) of DEPDC1B was amplified by PCR using specific primers and cloned into pFastBac1-HisMBPTEV, a custom-made vector derived from pFastBac1 (Life Technologies), expressed in High-5 insect cells with the MultiBac expression system (Berger et al., 2004) and purified by MBP affinity followed by size exclusion chromatography. Pull-downs with total cell lysate were performed in complete JS buffer, whereas those with purified proteins were performed in HEPES 10 mM (pH 7.5), NaCl 100 mM, glycerol 5%, Tween 0.1%. Details on labeling and quantitation of protein bound by SILAC are reported in the Supplemental Experimental Procedures. GST-RBD assays were performed by incubating 1 mg of protein lysate with 50 μg of GST-RBD beads for 1 hr at 4°C in a final volume of 0.5 ml. Beads were washed three times with RBD Wash Buffer (50 mM Tris-HCl [pH 7.6], 150 mM NaCl, Triton X-100 1% v/v, 10 mM MgCl2 freshly added) and resuspended in 30 μl of SDS-PAGE sample buffer.

Zebrafish Morpholinos and Time-Lapse Microscopy

A depdc1b splice-blocking morpholino was synthetized by Gene Tools against sequence GGTAAGAGCTGCGGGTAAAGCCTGC and used at a maximum final concentration of 0.7 mM. rhoab ATG-morpholino was previously described (MO1-rhoab; Zhu et al., 2008). To inhibit ptprf genes, a mixture 1:1 of ptprfa and ptprfb morpholinos was used (MO1-ptprfa and MO1-ptprfb; Wang et al., 2012). All morpholinos were injected in one-cell stage zebrafish embryos. Human DEPDC-1B mRNA was coinjected at a final concentration of 50 ng/μl. The specificity of morpholinos was checked by RT-PCR on total cDNA (details are in the Supplemental Experimental Procedures and Figure S7C). Time-lapse analysis was performed using embryos from the Tg(h2afv:GFP) zebrafish transgenic line (Pauls et al., 2001). Embryos at gastrula stage were embedded in a matrix of 1% low-melting agarose and analyzed by SP2 confocal microscope using argon 488 laser. Images were collected every 10 min with a 20× water-immersion objective. Actin cytoskeleton was analyzed by FITC-phalloidin (Sigma, P5282) staining (2 hr, final concentration: 5 μg/ml, 1% goat serum) on 10 hpf embryos fixed with 4% PFA in PBS (overnight at 4°C) and permeabilized with Triton X-100 2% for 2 hr. Stained embryos were mounted and analyzed as above.

Statistical Analysis

Microsoft Excel was used to generate bar graphs and to perform statistical analyses. Dot plots were produced and analyzed with Prism 6 software. Fitting curves of mitotic time-lapse experiments were generated using JMP 10 (SAS) software.