SMIFH2 has effects on Formins and p53 that perturb the cell cytoskeleton

Formin proteins are key regulators of the cytoskeleton involved in developmental and homeostatic programs, and human disease. For these reasons, small molecules interfering with Formins’ activity have gained increasing attention. Among them, small molecule inhibitor of Formin Homology 2 domains (SMIFH2) is often used as a pharmacological Formin blocker. Although SMIFH2 inhibits actin polymerization by Formins and affects the actin cytoskeleton, its cellular mechanism of action and target specificity remain unclear. Here we show that SMIFH2 induces remodelling of actin filaments, microtubules and the Golgi complex as a result of its effects on Formins and p53. We found that SMIFH2 triggers alternated depolymerization-repolymerization cycles of actin and tubulin, increases cell migration, causes scattering of the Golgi complex, and also cytotoxicity at high dose. Moreover, SMIFH2 reduces expression and activity of p53 through a post-transcriptional, proteasome-independent mechanism that influences remodelling of the cytoskeleton. As the action of SMIFH2 may go beyond Formin inhibition, only short-term and low-dose SMIFH2 treatments minimize confounding effects induced by loss of p53 and cytotoxicity.

processes  . Yet, observed effects, applied concentrations and treatment duration are quite variable and highlight lack of a thorough characterization of SMIFH2  .
In this study we find that SMIFH2 causes alternated depolymerization-repolymerization cycles of actin and microtubules, as well as scattering of the Golgi complex, whose onset correlates with posttranscriptional downregulation of Formin mDia2 and p53. While SMIFH2 attenuates p53-transcriptional activity independently of mDia2, it may induce cytoskeletal remodelling as a result of both Formin and p53 inhibition.
As we also report that SMIFH2 triggers cytotoxicity at high doses, we recommend current and future SMIFH2 users to employ short incubation (, 1 hour) and moderate concentrations (, 25 mM) to minimize confounding effects induced by loss of p53 and cytotoxicity. Under this provision, SMIFH2 represents a useful tool to study Formins both at the cellular and organismal levels.

Results
Single dose of SMIFH2 induces alternated depolymerizationrepolymerization cycles of actin and tubulin. SMIFH2 is frequently exploited as pharmacological Formin inhibitor in loss-offunction studies. Consistent with Formins being key actin-regulatory proteins, many reports showed that SMIFH2 perturbs the actin cytoskeleton and gives rise to a constellation of actin-dependent phenotypes (Table 1). However, the broad range of employed concentrations and incubation times prevent comparative analyses and raises concerns about target specificity of SMIFH2 (Table 1). Despite Formins' ability to regulate also microtubule dynamics, no information is available as to whether and how SMIFH2 affects microtubules. Together, these considerations show that a thorough and systematic characterization of SMIFH2 is missing.
We treated U2OS and HCT116 cells with SMIFH2 and monitored the remodelling of the actin cytoskeleton and microtubules over time  56 10 mM ,63 minutes To study the role of mDia Formins in CXCL12-induced bleb formation in MDA-MB-231 human breast carcinoma cells Oakes et al., 2012 47 10 -15 mM .4 hours To confirm a previous study that formation of radial stress fibres formation in U2OS cells is Formin Dia1-dependent Miklavc et al., 2012 45 25 mM not specified To show that actin coat formation on lamellar bodies in alveolar type II cells are Formin-dependent Chin et al., 2012 60 10 mM 4 hours To study the role of Formin-mediated cytoskeletal signaling in Chlamydia bacterial inclusion and extrusion from host cells (HeLa cells) Rosero et al., 2012 51 5 -30 mM 72 hours To study the plant cell growth and morphogenesis of Arabidopsis plants 2013 Sandbo et al., 2013 52 3 -30 mM 30 minutes To study myofibroblast differentation in human lung fibroblast cells Fritzsche et al., 2013 65 40 mM 30 minutes To analyse the contribution of Formin-mediated actin polymerization on actin cortex homeostasis Goldspink et al., 2013 40 10 mM 40 minutes To show that microtubule reorganization of EB-2 depleted ARPE-19 cells are restored to normal upon Formin inhibition Wilson et al., 2013 55 40 mM up to 490 seconds To study the contribution of Formin-mediated actin polymerization at the leading edge of polarized HL60 neutrophil-like cells during 3D migration Rao  using confocal microscopy. In order to compare different time points of treatment, we stained cells in parallel and used identical settings for image acquisition. Control U2OS and HCT116 cells showed well-organized actin and microtubule networks ( Fig. 1A and 2A). Cells incubated for one hour with SMIFH2 displayed reduced phalloidin signal accompanied by an increase in that of tubulin ( Fig. 1A and 2A). As actin and tubulin protein levels did not vary in either U2OS or HCT116 cells ( Fig. 1B  and 2B), there appears to be a drop in cellular filamentous actin and a concomitant increase in polymeric tubulin. After two hours of treatment, stress fibres and apical filopodia-like protrusions formed in U2OS and HCT116 cells, respectively ( Fig. 1A and 2A). Notably, the actin cytoskeleton of U2OS cells consisted primarily of lamellipodia and shaped into filopodia-like protrusions after four and eight hours of treatment with SMIFH2, respectively (Fig. 1A). Although HCT116 cells did not show lamellipodia at four hours, they formed basal filopodia-like protrusions after eight hours of treatment with SMIFH2 ( Fig. 2A).
Two hours after addition of SMIFH2, microtubules were barely detectable in both U2OS and HCT116 cells ( Fig. 1A and 2A, respectively). Lack of microtubules persisted up to the eight-hour time point in U2OS cells (Fig. 1A). Conversely, HCT116 cells showed some microtubules resistant to SMIFH2-induced depolymerization and also underwent a full repolymerization-depolymerization cycle between four and eight hours ( Fig. 2A). Both U2OS and HCT116 cells reacquired a morphologically normal cytoskeleton when exposed to SMIFH2 for more than sixteen hours (Fig. 1A, 2A and not shown). The finding that SMIFH2 induces alternated depolymerization-repolymerization cycles of actin and microtubules was confirmed by live-cell imaging (Supplementary Movie S1) and has three possible explanations: i) SMIFH2 undergoes intracellular breakdown and/or inactivation, which progressively lowers SMIFH2's active concentration below that required for full inhibition of all Formins, or ii) SMIFH2 has different and Formin-specific inhibitory effects on the actin and microtubule-regulatory activities, or iii) it has additional unknown targets. To gain insight into this issue, we treated U2OS cells with SMIFH2 and, every two hours, replaced the medium containing the inhibitor. Under this regimen, SMIFH2 addition caused a progressive and persistent depolymerization of both the actin cytoskeleton and microtubules ( Supplementary Fig. S1A). As the SMIFH2-containing medium was prepared at the beginning of the time course, these results suggest that the depolymerization-repolymerization cycles of actin and microtubules are due to intracellular SMIFH2 breakdown/inactivation rather than its instability.
SMIFH2 increases cell migration and prevents mitosis. The formation of SMIFH2-induced lamellipodia in U2OS cells prompted us to analyse cell migration. We manually tracked individual DMSO-or SMIFH2-treated U2OS cells and calculated total displacement, directionality and migration speed. SMIFH2treated cells showed enhanced motility compared to DMSOtreated control cells, whereas speed and directionality did not significantly change ( Fig. 3A-C, and Supplementary Movie S2). Interestingly, migration speed and displacement of SMIFH2treated U2OS cells increased between three and five hours ( Fig. 3D-E), temporally correlating with lamellipodium formation. As directionality was not concomitantly affected (data not shown), these data suggest that SMIFH2 affects cell motility by transiently modulating actin-based protrusions.
In-depth analysis of these time-lapse experiments evidenced that while the DMSO-treated cell population underwent mitosis, this was instead a rare event in the cells exposed to SMIFH2 ( Fig. 3F and Supplementary Movie S2). These results suggest that SMIFH2 delays (or abrogates) cell division and are consistent with Formins being implicated in cytokinesis.
SMIFH2 perturbs the architecture of the Golgi complex. The integrity of the Golgi apparatus strictly depends on microtubule 66 and actin dynamics, 67 INF2 and mDia1 Formins 68,69 . Thus, we stained the Golgi complex in control and SMIFH2-treated cells using Giantin, a bona fide marker for this organelle. The Golgi complex underwent a dramatic remodelling induced by SMIFH2 in both U2OS and HCT116 cells, as illustrated in Figure 1A and 2A, respectively. After one hour of treatment, the intensity of Giantin staining started to decrease and the Golgi showed a scattered morphology. At four hours, most of the Giantin signal disappeared, whereas protein levels remained stable ( Fig. 1B and  2B). Longer treatment with SMIFH2 resulted in gradual recovery of the normal Golgi structure, which was restored at sixteen hours ( Fig. 1A and 2A). We confirmed that SMIFH2 causes scattering of the Golgi complex using the cis-Golgi marker GM130 and the trans-Golgi marker TGN46 ( Supplementary Fig. S2A). As U2OS and HCT116 cells express different Formins ( Supplementary  Fig S3A-B), this effect seems to support that SMIFH2 is a general Formin inhibitor 39 . Strikingly, SMIFH2 failed to cause disappearance of Giantin and only resulted in a moderate scattering of the Golgi complex in mouse embryo fibroblasts (MEFs) ( Supplementary Fig.  S4). Moreover, MEFs also started to die at 8 hours post addition of 25 mM SMIFH2 ( Supplementary Fig. S4A). These discrepancies may be due to either inter-species differences (H. sapiens vs. M. musculus) or the nature of the analysed cells (cancer cells vs. immortalized cells). Nevertheless, MEFs showed alternated depolymerizationrepolymerization of both actin and tubulin ( Supplementary Fig. S4).
SMIFH2 reduces p300, mDia2 and p53 levels in a proteasomeindependent manner. Pilot dose-response experiments unmasked dramatic cytotoxic effects and cell death at SMIFH2 concentrations higher than 25 mM (Supplementary Movie S3), in agreement with previously reported observations 39 . As p53 is a master regulator of cellular apoptotic programs, we selected a few cells lines to cover all fundamental aspects of p53 biology. We employed 293T, A375, and assessed the protein levels of p53 and its transcriptional coactivator p300. Expression of Diaphanous-related Formins mDia1, mDia2 and mDia3 was monitored as a control. Remarkably, we found that SMIFH2 reduced the protein level of mDia2, p53 and p300, while that of mDia1 and mDia3 were not affected ( Fig. 4A-E). Notably, mDia1 showed a reproducible SDS-PAGE mobility shift suggesting a post-translational modification ( Fig. 4A-E). Downregulation of p53 was clearly independent of p53 status, expression levels, and transcriptional activity and occurred in both immortalized (293T) and cancer (A375, U2OS, MDA-MB-231 and HCT116) cells of human origin (Fig. 4F). In agreement with this conclusion, SMIFH2 triggered downregulation of mDia2 and p53 also in MEFs ( Supplementary Fig. S4B). Finally, we exploited two different anti-p53 antibodies to verify that SMIFH2 truly decreases p53 levels rather than causing post-translational modifications of p53 that prevent epitope recognition (Fig. 4G).
Collectively, these data show that downregulation of mDia2, p53 and p300 is a general and proteasome-independent effect of SMIFH2. Yet, the involvement of Formins in regulating gene expression post-transcriptionally remains unclear.
SMIFH2 attenuates p53 transcriptional activity. p53 is a wellknown transcription factor regulating expression of genes related to cell cycle, DNA damage repair and apoptosis 15 . Given that SMIFH2 reduces the expression of p53 and p300, we assessed the functional consequence of it on p53 transcriptional activity. To this end, we transfected 293T cells with a p53-responsive, HDM2 promoter-driven luciferase reporter plasmid, treated them with either SMIFH2 or DMSO and then measured the luciferase activity. These experiments revealed that addition of SMIFH2 attenuated p53-transcriptional activity compared to control samples (Fig. 5A).
Although SMIFH2 decreased both p53 and mDia2 expression, three observations suggest that reduced p53-transcriptional activity is independent of mDia2: knockdown of mDia2 did not affect i) either p53-transcriptional activity or ii) p53 protein levels ( Fig. 5B-D), and iii) SMIFH2 reduced p53 levels also in HCT116 cells (Fig. 2B), where mDia2 is below detection limit ( Supplementary Fig. S5C). Similarly, reduced mDia2 expression is independent of p53 levels as shown by the two following observations: i) silencing of p53 did not alter mDia2 protein levels ( Fig. 5E-G), and ii) time-course experiments revealed the SMIFH2induced dowregulation of mDia2 occurred with similar kinetics in both control and p53 knockdown U2OS cells ( Supplementary  Fig. S6A-B).
SMIFH2-induced cytoskeletal remodelling and downregulation of p53 temporally overlap. Analysis of total cell extracts matching the time-course of Figures 1 and 2 revealed that p53 and mDia2 levels started to decline in U2OS one and two hour after SMIFH2 addition, respectively (Fig. 1B). In U2OS cells, expression of p53 was dramatically decreased at the eight-hour time point, whereas that of mDia2 was below detection limit already after four hours of treatment. Remarkably, normal p53 levels were fully rescued after sixteen hours, whilst mDia2 downregulation remained complete. Conversely, p53 remained low when SMIFH2 treatment involved replacing the inhibitor every two hours ( Supplementary Fig. S1B). This observation corroborates the conclusion that SMIFH2 is broken down or inactivated within cells. In HCT116 cells, SMIFH2-induced downregulation of p53 became evident only at two hours of treatment and was less dramatic than in all other tested cell lines ( Fig. 2B and Fig. 4A-E). Nevertheless, SMIFH2-induced remodelling of the cytoskeleton and p53 downregulation occurred as temporally correlated events also in HCT116 cells. It is worth noting that effectiveness of SMIFH2 treatment is not related to basal p53 expression as SMIFH2 strongly reduced p53 levels in 293T cells, which have more p53 than HCT116 cells (Fig. 4A-B and Supplementary Fig. S5C).
In addition to regulating gene transcription, p53's action extends to the cytoskeleton 17-21 . As SMIFH2 reduced p53 expression and transcriptional activity, we wondered whether the effects of SMIFH2 on p53 could explain, at least partially, those on the cytoskeleton. Side-to-side comparison of cytoskeletal remodelling evoked by SMIFH2 in wild-type and p53 -/-HCT116 cells showed that the decrease in phalloidin and concomitant increase in tubulin signals observed at the one-hour time point were linked to p53 expression ( Fig. 2 and Fig. 6). Conversely, the two cell lines responded similarly to SMIFH2 from two hours onwards, namely since reduction of p53 levels in the p53 wild-type cells. Control and p53 knockdown U2OS cells showed that p53 expression levels regulate the actin cytoskeleton, whereas they do not affect either microtubules or Golgi organization ( Supplementary Fig. S6A). Under growing conditions, we observed that the cortical actin cytoskeleton and stress fibres were more prominent in either p53 knockdown mass populations than in the corresponding control knockdown mass population ( Supplementary Fig. S6A). Live-cell imaging of control and p53 knockdown U2OS cells expressing EGFP-LifeAct and mCherry-a-Tubulin unveiled that p53 is needed for SMIFH2 to induce the protrusion of lamellipodia (Supplementary Movie S4). Conversely, these experiments showed that SMIFH2's effect on microtubule dynamics are independent of p53, at least in U2OS cells (Supplementary Movie S5).
As SMIFH2 reduces p53 expression and activity and p53 levels affect the cytoskeleton, the observed interplay between p53 and the cytoskeleton makes it difficult to ascribe any cytoskeletal effects induced by SMIFH2 treatments longer than one hour solely to Formin inhibition. In spite of that, SMIFH2's effects on the Golgi complex did not seem to be modulated by p53 in either HCT116 or U2OS cells (Fig. 2 and Fig. 6 and Supplementary Movie S6).

Discussion
In this study we found that the general Formin inhibitor SMIFH2 causes alternated depolymerization-repolymerization cycles of actin and microtubules, as well as scattering of the Golgi complex in human cells ( Fig. 1 and Fig. 2). Surprisingly, SMIFH2 decreased the protein levels of p300, mDia2 and p53 without affecting the abundance of their messenger RNA. As proteasome inhibitors failed to restore the expression of p300, mDia2 and p53, we suggest that SMIFH2 regulates expression of these proteins post-transcriptionally and independently of the proteasome. Although it is formally possible that SMIFH2 might affect protein translation as such, we regard this possibility as unlikely since expression of other proteins (e.g. mDia1, mDia3, actin, tubulin and Giantin) remained unchanged (Fig. 4).
Consistent with the observed reduction in p53 and p300 levels, SMIFH2 attenuated p53 transcriptional activity independently of mDia2. The finding that expression of p53 and mDia2 were not interlinked confirms that there is no causal relationship between mDia2 and p53 regulation by SMIFH2 (Fig. 5).
Nonetheless, depolymerization of F-actin can activate p53 transcription 25 and polymerization of G-actin inhibits p53 function 26 . Given that Formins are actin nucleators, it is reasonable to speculate that they directly or indirectly impact on the p53 pathway. Interestingly, a recent study implicated Formin FMN2 in stabilization of cyclin-dependent kinase inhibitor p21 during oncogene/ stress-induced cell cycle arrest 72 .
Confocal imaging confirmed that SMIFH2 modulates the actin cytoskeleton as previously reported 39 , although observed phenotypes depend on both the cell line being tested and treatment duration (Fig. 1, Fig. 2 and Supplementary Fig. S4). We hypothesize that this is most likely due to Formins having cell-type-specific expression profiles (Supplementary Fig. S3) and different sensitivity to SMIFH2.
Under the condition that SMIFH2 has Formin-specific affinities and its active concentration drops relatively fast, the alternated depolymerization-repolymerization cycles of actin and microtubules triggered by SMIFH2 would agree with mDia1, mDia2, INF2, and possibly some other Formins, co-regulating the dynamics of F-actin and microtubules and either cytoskeletal network modulating Formins' action on the other one 29,73 .
Most importantly, we discovered that SMIFH2 reduces both the expression levels and the transcriptional activity of p53 and that this property contributes to SMIFH2-induced cytoskeletal remodelling (Fig. 5, Fig. 2, Fig. 6 and Supplementary Fig. S6). The link between p53 and cytoskeletal remodelling is further strengthened by our reanalysis of a recent study identifying a set of 198 genes upregulated one hour after stabilization of p53 in HCT116 cells 74 . Among them, 77 were novel and previously uncharacterized p53 targets 74 . Using Ingenuity Pathway Analysis, we found that 43 out of the 192 mapped genes are early p53 target genes significantly associated with functions related to cellular movement (Table 2). In keeping with this notion, we noted that wild-type and p53 -/-HCT116 cells responded differently to SMIFH2 at early (, 1 hour), but not late time points. We established that p53 modulates the cellular responses to SMIFH2 both in HCT116 cells and U2OS cells by using syngenic wild-type and p53 knockout (or knockdown) cell lines ( Supplementary Fig. S6 and Supplementary Movie S4).
Although SMIFH2 has been previously reported to exert antimigratory effects, 39 we found that it increased cell migration between three and five hours of treatment (Fig. 3). This discrepancy might be due to Rizvi and colleagues exposing NIH3T3 cells to a sub-lethal concentration of SMIFH2, NIH3T3 and U2OS cells differing in Formin-protein expression, or the experimental setup.
At any rate, SMIFH2 perturbs the architecture of the Golgi complex independently of p53 expression with a kinetics that differs from those of the actin and the microtubule networks (Fig. 1, Fig. 2, and Supplemental Movie S6). In this regard, loss of INF2 and activation of Rho-mDia1 pathway have been shown to result in partial dispersal of the Golgi complex in U2OS cells 68,69 . As virtually complete loss of visible Golgi structures occurs in U2OS and HCT116 cells treated with SMIFH2 ( Fig. 1A and 2A), one or more Formin(s) may cooperate with INF2 in regulating Golgi architecture. The fact that perturbation of the Golgi complex and centrosomal microtubules are out of synchrony suggests that SMIFH2 might interfere with the dynamics of non-centrosomal microtubules nucleated at the Golgi, which are crucial for proper assembly and functionality of the Golgi complex 75 . Finally, why SMIFH2 does not elicit loss of the Golgi in MEFs warrants future investigation.
Remarkably, SMIFH2-treated cells recovered their original morphology after sixteen hours of treatment. Given that intracellular SMIFH2 decays in a few hours, this implies that inhibition of Formins by SMIFH2 is transient and reversible. In light of these considerations, it is notable that mDia2 remains fully silenced also when both the Golgi complex and the cytoskeleton have regained a normal morphology. Overall, our observations reinforce the idea that SMIFH2 may have different binding affinities for and ensuing inhibitory effects on Formins, and that the vast collection of perturbations induced by SMIFH2 (Table 1) results from different subsets of Formins being inhibited at any analysed time point.
High SMIFH2 concentration (50 mM) triggered rapid cell death in all tested cell lines (Supplementary Movie S3), consistent with previous observations and IC 50 (IC 50 5 28.0 mM) 39 . As our data also show that this SMIFH2-induced event does not require p53 (Supplementary Movie S3), further studies should address the mode whereby SMIFH2 promotes cell death.
In summary, we showed that the general Formin inhibitor SMIFH2 has profound effects on F-actin, microtubules and integrity of the Golgi complex and influences important cellular processes, such as cell migration and cell division. Unexpectedly, we found that SMIFH2 also reduces the expression and the transcriptional activity of p53 and that this latter property may contribute to SMIFH2induced cytoskeletal remodelling. Yet, it remains to be established    www.nature.com/scientificreports whether p53 downregulation is caused by SMIFH2 inhibiting Formins or uncharacterized off-target effects of SMIFH2. As SMIFH2 affects both Formins and p53, we advise current and future users to administer SMIFH2 at moderate concentrations (, 25 mM) and employ short treatments (, 1 hour) to minimize confounding effects induced by loss of p53 and cytotoxicity. Under this provision, SMIFH2 remains a useful tool to study Formins both at the cellular and organismal levels.

Methods
Chemicals and Reagents. High-glucose DMEM supplemented with pyruvate and GlutaMaxH was from Invitrogen. Dual-LuciferaseH Reporter Assay System was from Promega. Lactacystin was from Cayman Chemicals and dissolved as 10 mM stock in DMSO and used at 10 mM. SMIFH2 was from Sigma-Aldrich and dissolved as 50 mM stock in DMSO and stored at -80uC in single-use aliquots. Thus, freeze-thaw cycles were limited to one since we noticed decreased activity upon additional cycles. SMIFH2 was used at 25 mM throughout this study, unless specified otherwise. Retroviral expression plasmid pMX-EGFP-LifeAct was generated by polymerase chain reaction and sequence verified. Retroviral expression plasmid pCX-mCherrya-Tubulin was a kind gift from R. Wolthuis. pECFP-Golgi was from Clontech. pGL3-HDM2-luc reporter plasmid was a kind gift from R. Bernards. X-tremeGene9 was from Roche. All other reagents were purchased from Sigma-Aldrich.
Total RNA Isolation and RT-qPCR Analyses. Total RNA from adherent cells was extracted using RNeasy Mini kit (Qiagen) according to manufacturer's instructions. Complementary DNA synthesis was performed using 1-2 mg of mRNA with SuperScript-IIH reverse transcriptase according to the manufacturer's instructions (Invitrogen). Real-time qPCR reactions were set up using 5-10 ng of cDNA as a template and gene specific primers (200 nM) in a StepOnePlus TM Real-Time PCR system (Applied Biosystems). All reactions produced single amplicons (100-200 bps), which allowed us to equate one threshold cycle difference. RT-qPCR primers are listed in Table 3 and have been previously validated 76 .
Luciferase Activity Assay. 293T cells (5 x 10 4 cells/well) were plated in 24-well plates. One day after, cells were transfected with pGL3-HDM2-luc (100 ng) using calcium phosphate method. All experiments were carried out in triplicates and the Firefly Luciferase activity was measured 24 hours post-transfection with the Dual-LuciferaseH Reporter Assay System (Promega) using an EnVisionH Multilabel Reader (PerkinElmer). Initially, we added HDM2-Firefly luciferase-based reporter and Renilla luciferase-based co-reporter in a 1:10 ratio, which prevented trans effects between the two promoters (not shown). Incomplete quenching of the Firefly luciferase and low Renilla luciferase activity affected the normalization of the samples thereby causing misrepresentation of the effects (not shown). As total DNA amount was kept constant, we obtained very similar transfection efficiencies for different conditions and independent samples. Thus, the Luciferase activity was expressed as arbitrary units/mg of total proteins. Results are normalized against the control samples and represented as relative HDM2-promoter activities (mean 6 s.d.), as obtained from three independent experiments carried out in technical triplicates.
Immunofluorescence and Imaging. Cells were plated on gelatin-coated glass coverslips (#1.5) and fixed with 4% paraformaldehyde in PIPES buffer (80 mM PIPES pH 6.8, 5 mM EGTA, 2 mM MgCl2) for 10 minutes. Fixed cells were permeabilized in PBS containing 0.5% BSA (w/v) and 0.1% Triton-X100 (v/v) for 10 minutes, and stained with primary and secondary antibodies in blocking buffer (2.5% BSA (w/v) in PBS). Coverslips were mounted using Mowiol. Images were acquired on a CLSM Leica TCS SP5 operated with Leica Confocal Software (LAS-AF; Leica) and equipped with a HCX PL APO CS 63.0x (N.A. 1.40) oil objective. All channels were acquired sequentially.
Live cell confocal images were acquired on a CLSM Leica TCS SP5 equipped with a humidified climate chamber with 5% CO 2 at 376C. Single basal sections with a widened pinhole (1.5 Airy) were acquired every 20 minutes. Cells were left in the humidified climate chamber for at least one hour to obtain steady-state conditions prior to the beginning of each experiment. All channels were acquired sequentially. Images were corrected for photobleaching through estimation of the baseline intensity level by curve fitting the background intensities using ''Exponential with offset'' selection in ImageJ. These estimates were subsequently used for bleach correction in ImageJ with the simple ratio method 81 . Minor drift was corrected using the TurboReg and StackReg plugins in ImageJ 82 .
Random Cell migration assay and quantification of cell motility and mitosis. Cells were plated on gelatin coated 6 or 12 wells plate and cells were allowed to adhere for at least sixteen hours. Experiments were performed in a humidified chamber with 5% CO 2 at 37 6C in the presence of DMSO or SMIFH2. Cells were imaged every five Table 3 | List of primers used for RT-qPCR.

Gene
Primer #1 (5'-3') Primer #2 (5'-3') TGCATTTTGAGAAGAACAAAGTG CCAGCTTATCTTGATCTTTGCAG www.nature.com/scientificreports minutes on a Zeiss Axio Observer Z1 microscope (Carl Zeiss) equipped with a LD Plan-Neofluar Ph2 20x (N.A. 0.40) objective, operated with Zeiss Microscope Software ZEN 2012. Individual cells were tracked using Manual Tracking plugin for ImageJ. Average distance, speed and directionality of movement were computed using the Chemotaxis Tool plugin for ImageJ provided by ibidi GmbH (http://www. ibidi.com). Cells entering mitosis were scored manually and defined as follows: a cell entered mitosis when its flat and spread appearance changed to a round-up, yet adhesive state. Initial number of cells was counted manually in each field of observation and percentage of cells entering mitosis every hour was calculated using this initial number of cells as reference.