EMT induces characteristic changes of Rho GTPases and downstream effectors with a mitosis-specific twist

Epithelial-mesenchymal transition (EMT) is a key cellular transformation for many physiological and pathological processes ranging from cancer over wound healing to embryogenesis. Changes in cell migration, cell morphology and cellular contractility were identified as hallmarks of EMT. These cellular properties are known to be tightly regulated by the actin cytoskeleton. EMT-induced changes of actin-cytoskeletal regulation were demonstrated by previous reports of changes of actin cortex mechanics in conjunction with modifications of cortex-associated f-actin and myosin. However, at the current state, the changes of upstream actomyosin signaling that lead to corresponding mechanical and compositional changes of the cortex are not well understood. In this work, we show in breast epithelial cancer cells MCF-7 that EMT results in characteristic changes of the cortical association of Rho-GTPases Rac1, RhoA and RhoC and downstream actin regulators cofilin, mDia1 and Arp2/3. In the light of our findings, we propose that EMT-induced changes in cortical mechanics rely on two hitherto unappreciated signaling paths—i) an interaction between Rac1 and RhoC and ii) an inhibitory effect of Arp2/3 activity on cortical association of myosin II.


Introduction
Epithelial mesenchymal transition (EMT) is a cellular transformation of epithelial cells that entails the loss of apical-basal cell polarity and intercellular adhesion in combination with a gain of mesenchymal cell traits, see figure 1(a) [1][2][3][4]. EMT was linked to the initiation of metastasis and bad cancer prognosis through the acquisition of aggressive traits in cancer cells of epithelial origin [1][2][3][4]. In particular, EMT was reported to be connected to enhanced cell migration and cell proliferation in metastatic cancer cells [1][2][3][4][5].
The actin cytoskeleton is a major regulator of cell mechanics, cell shape and cellular force generation. Thereby, the actin cytoskeleton constitutes a key player in cancer-related changes of cell migration and cell division [6][7][8]. Consistent with this, it was found that EMT causes major changes in the actincytoskeleton [5,9,10].
Rho GTPases are known to be essential regulators of the actin cytoskeleton. We and others showed that EMT is associated with characteristic changes in the activation of Rho GTPases as judged by the abundance of its active GTP-bound forms [5,[11][12][13]. In particular, we reported a decrease of total RhoA-GTP and an increase of total Rac1-GTP upon EMT in MCF-7 breast epithelial cells. Furthermore, the increased expression of the Rho GTPase RhoC was associated to enhanced metastasis in several cancer types [14]. In addition, RhoC signaling was featured to be essential for EMT [13][14][15][16].
Previously, we reported characteristic EMTinduced changes of actin cortex mechanics in rounded cells of diverse epithelial cancer cell lines originating from breast, lung, prostate and skin tissue indicating that this EMT-induced cell-mechanical change is a widely conserved feature in cells of diverse tissue origin [5,17,18]. Cortex-mechanical changes were entailing cortical softening and contractility reduction in interphase but a cortical stiffening and contractility increase upon EMT in mitosis. Concomitantly, we found EMT-induced changes of cortical actin and myosin II with reduced cortical myosin in interphase and increased cortical actin in mitosis [5].
While associated EMT-induced changes in Rho GTPases signaling provide a viable hypothesis for downstream changes of cortical actin and myosin, the details of EMT-induced changes in cortical signaling remain elusive. In particular, it is unclear how cortical mechanics is affected in an opposite way in interphase and mitosis, as none of the downstream actomyosin effectors of Rho GTPases are known to induce mechanical changes in the cortex that depend on the cell cycle stage [19,20].
With this work, we aim to deepen our understanding of EMT-induced changes in cortical signaling, cortical composition and cortical mechanics with a focus on the differences between interphase and mitosis. To this end, we quantify EMT-induced changes of Rho GTPases RhoA, RhoC and Rac1 which were previously linked to EMT. Furthermore, we investigate as actin-regulating downstream targets formin, Arp2/3 and cofilin. In particular, we provide a quantitative analysis of EMT-induced changes of cortical protein localization in non-adherent cells in combination with changes of cortical mechanics and protein expression. In light of our results, we propose that two hitherto unappreciated signaling mechanisms at the cortex are at the heart of EMT-induced cell-mechanical changes-i) an interaction between Rac1 and RhoC and ii) an inhibitory effect of Arp2/3 activity on myosin II cortical localization.

Results
To investigate the effects of EMT on actin-cortical signaling and mechanics, we chose to work with the breast epithelial cancer cell line MCF-7 which exhibits epithelial cell traits in control conditions. We induced EMT in these cells via an established method (see e.g. [5,[21][22][23][24][25]) that entails a 48 h treatment with the tumor promoter 12-Otetradecanoylphorbol-13-acetate (TPA) at 100 nM, see section 5. We and others showed previously that in response to this treatment, MCF-7 cells display an EMT-characteristic protein expression change, corresponding cell-morphological changes towards a mesenchymal-like phenotype, as well as increased proliferation and migration, see e.g. [5,24,25]. In particular, we showed earlier that the epithelial marker E-cadherin is downregulated through the treatment, while the mesenchymal markers N-cadherin and Vimentin are upregulated, see figures S1(b) and (c) in the supporting information of [5]. Further, cells acquire a more spindle-shaped morphology and grow more isolated from each other, see figure S1(a) in the supporting information of [5]. We note that EMT-transformed cells will be referred to as modMCF-7 cells throughout this manuscript.
Previous research has shown that activation of Rho GTPases is connected to their association with the plasma membrane [26]. In addition, their immediate downstream signaling is inherently local as direct downstream effectors such as mDia1, Rock and Wasp and Wave require persistent binding for activation [27][28][29]. Furthermore, the activation of downstream actin effectors cofilin, formin and Arp2/3 was shown to be linked to f-actin binding [28,[30][31][32]. Correspondingly, association of these proteins to cortical f-actin is a measure of their activity at the actin cortex. Therefore, one prevalent strategy of this study is to quantify changes of the relative amount of cortex-associated cortical regulators upon EMT as a readout of EMT-induced changes in cortical signaling. Following previous studies [5,18,19,33,34], we worked with rounded, non-adherent cells since this has the advantage that cell shapes are spherical in both epithelial and EMTtransformed conditions with a largely uniform actin cortex. In this way, a meaningful comparative analysis of cortical protein association between epithelial and the mesenchymal-like cells becomes possible.
For the measurement of cortical protein association, immunostaining of the cortical regulator under consideration was combined with fluorescent DNA staining (DAPI or Hoechst) which allowed to identify cells to be in an interphase or mitotic stage, see section 5. For the measurement of cortical regulators in mitotic cells, the fraction of mitotic cells was enriched through mitotic arrest induced by co-incubation with S-trityl-L-cysteine (STC), see section 5. Using a previously established image analysis scheme, we analyzed confocal images of immunostaining fluorescence intensities to infer the cellular outline and the averaged cortical fluorescence profile along the radial coordinate, i.e. orthogonal to the cell boundary, see figure 1(b) and [5,18]. The averaged radial fluorescence intensity was then used to derive the cortex-to-cytoplasm ratio of protein localization in the cells by calculating the ratio of the integrated cortical fluorescence normalized by the cytoplasmic fluorescence intensity, see figure 1(c), section 5 and [5,18].

Rho GTPases change their cortical association upon EMT in interphase and mitosis
To investigate whether cortical association of Rho GTPases changes through EMT, we quantified the cortex-to-cytoplasm ratio of RhoA, RhoC and Rac1 in cells with and without EMT induction. To this end, we performed confocal imaging of the equatorial cross section of suspended interphase or STC-arrested mitotic cells which were immunostained for either of the Rho GTPases under consideration, see figure 1(d) and section 5. Quantitative analysis shows that the cortex-to-cytoplasm ratio of Rac1 increases upon EMT both in interphase and mitosis (figures 1(d) and (e)). By contrast, the cortex-to-cytoplasm ratio of RhoA decreases through EMT (figures 1(d) and (f)). We conclude that cortical association of Rac1 and RhoA follows the EMT-induced quantitative change of GTP-bound Rac1 and RhoA in whole-celllysates of MCF-7 cells [5]. For RhoC, EMT-induced changes of the cortex-to-cytoplasm ratio are distinct in interphase and mitosis. While cortical RhoC goes down in interphase, we see an increase of cortical RhoC in mitosis (figures 1(d) and (g)). Our results on the effect of EMT on cortical signaling of Rho GTPases is summarized in figure 1(h).
We further asked about the influence of Rho GTPases on cortical mechanics. For this purpose, we relied on cortex-mechanical measurements with an established cell confinement setup based on oscillatory cell-squishing with the cantilever of an atomic force microscope (AFM). We chose a deformation frequency of 1 Hz. This assay was previously shown to provide a readout of cortical stiffness, cortical tension as well as a characterization of the viscoelastic nature of the cortex quantified by the phase shift between stress and strain [5,17,18,35]. (The phase shift takes values between 0 • -90 • with lower values corresponding to a more solid-like response). In particular, we previously showed that Rac1 signaling was linked to a decrease in cortical stiffness and contractility in interphase cells but to an increase of cortical stiffness and contractility in mitotic cells with a stronger effect in post-EMT cells [5]. This is in agreement with our here reported finding of increased cortical Rac1 association post-EMT (figure 1(e)). On the other hand, we previously found that RhoA signaling increases cortical stiffness and contractility in particular in pre-EMT cells [5]. Again, the bigger mechanical effect pre-EMT is in agreement with our current observation of higher cortical RhoA association pre-EMT (figure 1(f)).
The effect of RhoC on cortical mechanics has to our knowledge not been reported previously. Using the AFM-based cell confinement assay, we measured cortical mechanics with and without RhoC knockdown via RNA interference in pre-and post-EMT conditions, see figures 1(i)-(l), S1 and section 5. We find that similar to RhoA, RhoC signaling increases cortical contractility and stiffness, see figures 1(i)-(l). However, this effect is restricted to pre-EMT interphase cells and post-EMT mitotic cells. It is plausible that the absence of an effect in these conditions is linked to low abundance of RhoC at the cortex (figure 1(g)). Furthermore, RhoC knock-down increases the phase shift in pre-EMT interphase conditions indicating that RhoC signaling contributes to the solid-like nature of the cortex in interphase, see figure S1(b).

Rac1 and RhoC mutually affect their cortical association
The previously reported finding that Rac1 activity affects cortical mechanics opposite in interphase and mitosis provides a clue that the signaling of Rac1 might be at the heart of the cell-cycle dependence of cytoskeletal changes upon EMT. However, currently it is unclear how cortical Rac1 signaling can act in a manner that is qualitatively different in interphase and mitosis. In particular, it surprised us that Rac1 would make a strong contribution to cortical contractility in mitosis in post-EMT conditions given that Arp2/3 activity increase downstream of Rac1 is expected to diminish cortical contractility, see figure 6 and [20,36]. Furthermore, previous reports showed that RhoA is at the heart of cortical contractility in mitosis [37]. While RhoA activity and cortical association is low in post-EMT MCF-7 cells (figure 1(f) and [5]), we note that RhoC signaling is similar to RhoA. Therefore, RhoC signaling might step in for RhoA signaling after EMT during mitosis.
Following this line of thought, we asked whether Rac1 might increase cortical contractility in post-EMT mitosis via (direct or indirect) activation of RhoC. To test this hypothesis, we monitored changes in cortical association of RhoC upon knock-down of Rac1 in pre-and post-EMT conditions judged by fluorescence intensity of RhoC immunostaining (figure 2(a)). Obtained confocal images of equatorial cross-sections were used for image analysis in all conditions. We find that inferred cortex-to-cytoplasm ratios of RhoC increase upon knock-down of Rac1 in rounded interphase cells with a stronger effect in post-EMT conditions (figures 2(a) and (c)). By contrast, the cortex-to-cytoplasm ratio of RhoC decreases upon knock-down of Rac1 in mitosis with a stronger effect in post-EMT conditions (figures 2(a) and (d)). We conclude that Rac1 signaling increases cortical association of RhoC in mitosis but diminishes cortical association in interphase in MCF-7 cells (figure 2(g)).
To test whether in turn also RhoC signaling influences Rac1, we performed immunostaining of Rac1 with and without RhoC knock-down in pre-and post-EMT conditions in interphase and mitosis, see figure 2(b). We find that inferred cortex-to-cytoplasm ratios of Rac1 decrease upon knock-down of RhoC in pre-EMT interphase cells and in post-EMT mitotic cells (figures 2(e) and (f)). In all other conditions, there is no significant effect on Rac1 cortical association (figures 2(e) and (f)). We conjecture that the signaling from RhoC to Rac1 is restricted to pre-EMT interphase and post-EMT mitosis due to the low cortical representation of RhoC in post-EMT interphase and pre-EMT mitotic conditions, see figures 1(g) and (h). With this explanation approach, our data are consistent with an in general activating effect of RhoC on Rac1 (figure 2(g)).
In previous work, the active forms of RhoA and Rac1 were shown to affect each other through mutually inhibitory interactions in breast epithelial cells [38]. This is consistent with the EMT-induced switchlike change from a state of high RhoA and low Rac1 activation to a state of low RhoA and high Rac1 activation [5]. The interaction between Rac1 and RhoC has been to our best knowledge unknown so far.

Cortical cofilin association increases through EMT
We went on to ask how EMT-induced changes in Rho GTPases signaling affect downstream cortical regulators. We first investigated EMT-induced changes of cofilin cortical association. Cofilin is known to promote the depolymerization of the actin cortex through severing of actin fibers [39]. In the context of cancer, cofilin activity at the cortex has been suggested to be a main factor in f-actin turnover thus playing a key role in cancer cell migration and invasion [40]. Cofilin becomes deactivated through phosphorylation mediated by Lim kinases [40] and phosphorylated cofilin was shown to not interact with f-actin [30]. Correspondingly, cortex-bound cofilin can be interpreted as active cofilin. We quantified total amounts of cofilin (CFL1) and phospho-cofilin (phospho-CFL1 (Ser3)) via western blotting from lysates of adherent cells with or without EMT-induction, see figures 3(a), (b) and section 5. Calculating fold changes upon EMT, we find a trend of a shallow increase of total cofilin (only interphase) and a decrease of phospho-cofilin upon EMT, see figure 3(c). Taken together, this points at an increase of the active non-phosphorylated form of cofilin upon EMT in MCF-7.
To assess cortical association of cofilin, we performed also cofilin-immunostaining of rounded cells, see figure 3(d). We find that the cortex-to-cytoplasm ratio of cofilin is elevated upon EMT indicating an EMT-mediated increase of cortical cofilin activity, see figures 3(d) and (e). This finding is in agreement with an increase of active cofilin as suggested by western blotting as described above, see figure 3(c).
Immunostaining of the cofilin upstream regulator phospho-Limk1 (phospho-LIMK1 (Thr508)) shows no cortical association but cytoplasmic localization in agreement with previous findings, see figure S3(a) and [41]. Quantifying phospho-Limk1 abundance in whole-cell lysates via western blotting, we find an EMT-induced decrease in interphase (asynchronous cell population) in accordance with the observed concomitant decrease of phosphocofilin, see figures 3(a)-(c) and section 5. In mitosis, phospho-Limk1 amounts are very low and show a trend of decrease upon EMT which is, surprisingly, in disagreement with the EMT-induced trend of phospho-cofilin (figures 3(a)-(c)). We speculate that this apparent inconsistency may be attributed to the previously reported modified activation scheme of Limk1 in mitosis, where hyperphosphorylation rather than phosphorylation at Thr508 is at the heart of Limk1 activation [42]. This observation features phospho-Limk1 (Thr508) as an unsuitable readout of cofilin phosphorylation activity in mitosis.
Investigating the effect of cofilin on cortical mechanics, we find that cofilin knock-down through RNA interference changes cortical mechanics, see figures 3(f)-(i) and S3(e)-(h). Both cortical tension and stiffness increase in interphase and mitosis, see figures 3(f)-(i). In addition, the phase shift and therefore the fluid-like nature increased mildly upon knock-down in interphase cells, see figure S3(f). We conclude that increased cortical association of cofilin in EMT-induced cells contributes to a trend of decreased cortical contractility and stiffness.
We note that Chugh et al [19] previously reported a tension increase upon CFL1-knockdown in interphase HeLa cells in agreement with our findings. However, the authors reported by contrast a tension decrease upon CFL1 knock-down in mitosis opposite to our findings in MCF-7. This apparent discrepancy might be rooted in the non-monotonous dependence of cortical tension on actin filament length as was proposed by the same study [19]. According to this idea, increased actin filament length through cofilin knock-down can either increase or decrease cortical tension depending on the initial state of the cortex. Large differences in cortical tension values between mitotic MCF-7 cells and mitotic HeLa cells make different cortical configurations in mitosis for the two cell lines additionally plausible.

The actin nucleator mDia1 shows distinct changes of cortical association upon EMT in interphase and mitosis
In order to further understand EMT-induced changes of cortical composition and mechanics [5], we addressed how actin nucleators are affected upon EMT downstream of Rho GTPases. Cortical actin is polymerized by formins and Arp2/3. We will first focus on the influence of the former. Previous studies have shown that formin activity has a major influence on cortical mechanics [19,20,43,44]. We confirmed this finding in rounded MCF-7 cells with our AFM-based cell confinement setup showing that formin inhibition via 40 µM SMIFH2 reduced cortical stiffness and contractility in MCF-7 cells in interphase and mitosis, see figure S4.
To further investigate how formin-mediated polymerization changes at the cortex upon EMT, we decided to focus on the formin representative mDia1 (also Diaph1) which together with the actin nucleator Arp2/3 polymerizes the majority of f-actin in the actin cortex [45]. mDia1 is activated downstream of RhoA, RhoB or RhoC through binding to the active form of the respective Rho GTPase [28]. The active form of mDia1 associates with f-actin [28,32] and thus with the actin cortex.
Performing quantification of protein amounts in whole-cell lysates via western blots, we find that there is no significant change of expression of mDia1 upon EMT (figures 4(a)-(c)). We then went on to monitor cortical association of mDia1 via immunostaining and quantification of the cortex-to-cytoplasm ratio, see figures 4(d) and (e). We find clear EMT-induced changes; in interphase cells, the cortex-to-cytoplasm ratio of mDia1 is reduced, see figure 4(e), blue boxes. We conclude that cortical association of mDia1 is decreased upon EMT in agreement with our finding of reduced presence of RhoA and RhoC at the cortex. In mitotic cells, on the other hand, the cortex-tocytoplasm ratio of mDia1 is increased, see figure 4(e), yellow boxes. This finding points at an increase of cortical mDia1 activity upon EMT in mitosis. We suggest that this effect is due to an EMT-induced rise of cortical activity of RhoC in mitosis overcompensating the effect of reduced RhoA activity in post-EMT mitotic cells, see figure 1(h).
Taken together, our results suggest that the observed characteristic EMT-induced changes of cortical mDia1 likely make an essential contribution to EMT-induced changes of cortex-associated actin and cortical mechanics in interphase and mitosis.

The actin nucleator Arp2/3 increases its cortical association upon EMT
To further increase our understanding of changes in cortex-associated actin upon EMT as reported in [5], we also investigated EMT-induced changes of the second major actin nucleator beyond mDia1, namely the Arp2/3 complex [6,45]. The Arp2/3 complex becomes activated by the Wasp family of proteins [31,46]. Activated Wasp proteins promote binding of the Arp2/3 complex to f-actin [31,47,48] and thus its association to the actin cortex. Wasp proteins (WAVE) are activated downstream of Rac1 [29]. We therefore expect that our finding of EMT-induced increase of cortical Rac1 association (figure 1(h)) should give rise to a downstream increase of cortical Arp2/3 association. Previous studies reported that Arp2/3 signaling reduces cortical tension in interphase and mitosis [33,36]. This is counter-intuitive as Arp2/3 mediates actin polymerization and, thereby, could be expected to increase the amount of cortical myosin II substrate. To test the effect of Arp2/3 signaling on cortical tension in MCF-7 cells, we measured actin cortical mechanics with and without the Arp2/3 inhibitor CK666, see figures 5(f)-(i) and S5. We find that in agreement with previous results, cortical tension and stiffness increases upon Arp2/3 inhibition in interphase and mitosis, see figures 5(f)-(i). By contrast, the phase shift did not change significantly upon Arp2/3 inhibition, see figures S5(b) and (d).
We conclude that our data suggest that increased Arp2/3 activity upon EMT downstream of increased Rac1 activity tends to reduce cortex stiffness and contractility independent of the cell cycle state.

Arp2/3 activity enhances cortical actin but reduces cortical association of myosin II
To deepen our understanding of cortical response to EMT-induced changes of cortex-associated Arp2/3, we investigated how cortex-associated actin and myosin change in response to Arp2/3 inhibition. For this purpose, we transfected MCF-7 cells with constructs expressing fluorescently labeled myosin regulatory light chain (MLC2) or fluorescently labeled actin (ACTB) as fluorescent reporters of cellular myosin II and actin localization, see figure 6(a) and section 5. Cell transfection was performed for suspended cells in interphase and mitosis as well as in preand post-EMT conditions, see section 5. Quantifying cortical association of myosin II and actin via confocal imaging of transfected live cells, we determined the cortex-to-cytoplasm ratio of actin and myosin II in conditions with and without Arp2/3 inhibition via the inhibitor CK666. As expected, we find that cortical actin association goes down upon Arp2/3 inhibition in all conditions, see figure 6(b). By contrast, we observe that Arp2/3 inhibition increases myosin II association to the cortex in all conditions in spite of reduction of cortical actin, see figure 6(c).
To further corroborate these findings, we doublechecked the observed effect of Arp2/3 signaling in fixed cells using Phalloidin as a fluorescent reporter of f-actin and immunostaining of MYH9 as a fluorescent readout of myosin II localization, see figure S6. Again, we see that Arp2/3 inhibition leads to diminished cortical f-actin in combination with an increase in cortical myosin II confirming our results obtained from transfected live cells, see figures S6(b) and (c).
We conclude that cortical Arp2/3 in addition to expected allocation and polymerization of cortical actin leads to diminished cortical association of myosin II. Increased cortex-associated Arp2/3 downstream of EMT-induced enhanced Rac1 signaling may therefore, at least in part, account for the observation of emergent reduced cortical tension and stiffness as well as reduced myosin II cortex-to-cytoplasm ratios upon EMT in interphase cells [5]. We note that a negative effect of Arp2/3 on myosin activity was previously reported within mouse oocytes [49].

Discussion
In this study, we investigated how actin cortex regulation is changed upon EMT in MCF-7 breast epithelial cancer cells. As previous work suggested that activity changes of the Rho GTPases are a game changer of cytoskeletal regulation upon EMT [5,13,38,[50][51][52], we focused on modulations of cortical signaling of RhoA, RhoC and Rac1 as well as on selected downstream effectors.
Signaling of cortical regulator proteins was assessed through immunostaining and subsequent confocal imaging of fixed MCF-7 breast epithelial cells in a rounded, non-adherent state with a largely uniform cortex. The magnitude of the cortex-tocytoplasm ratio of the regulator protein under consideration was used as a quantitative readout of cortical signaling strength. In particular, we compared cortexto-cytoplasm ratios in control and EMT-induced conditions in interphase and mitosis. Furthermore, for cortical regulators RhoC, formin, Arp2/3 and cofilin, we identified their influence on cortical mechanics which was quantified by an established AFMbased cell confinement setup [5,17,18,35]. Red boxes indicate proteins whose cortical signaling increases through EMT. Blue boxes indicate proteins whose cortical signaling decreases through EMT. Grey boxes indicate proteins that show no net change in their cortical signaling upon EMT. In the signaling network, pointed black arrows represent activating signaling while flat black arrows indicate inhibiting signaling.
In summary, we found that EMT reduces cortical association of RhoA but enhances cortical association of Rac1, while EMT-induced changes of cortical RhoC are different in interphase and mitosis, see figures 1(h) and 7. Interestingly, we discovered a hitherto unappreciated interaction between RhoC and Rac1 that likely contributes to RhoC activation in mitotic EMT-induced cells, see figure 2(g). This interaction entails in particular a reduction of cortical RhoC in non-adherent interphase cells but an increase of cortical RhoC in mitosis through Rac1 signaling.
Downstream of Rho GTPases, we found that also the cortical signaling of the formin mDia1 is affected in a cell-cycle-dependent manner by EMT. The corresponding decrease of mDia1 at the EMTtransformed interphase cortex can be attributed to decreased RhoA and RhoC signaling. On the other hand, EMT-related increase of mitotic RhoC signaling can account for increased mDia1 at the mitotic post-EMT cortex, see figure 7. Furthermore, we find an EMT-induced increase in Arp2/3 and cofilin signaling at the cortex in both interphase and mitosis, see figure 7.
Taken together, our study indicates that actin nucleation at the interphase cortex is promoted upon EMT through upregulation of Arp2/3, but diminished through downregulation of mDia1 and enhanced cofilin signaling at the cortex. All in all, these signaling changes may give rise to the observed absence of a net change of cortex-associated actin upon EMT in interphase, see figure 6(b) and [5]. Furthermore, myosin activity at the interphase cortex is diminished through reduced RhoA and RhoC signaling (via Rock) and via increased Rac1 signaling (mediated downstream by Arp2/3, see figure 6(d). The combined effects can account for the observed net decrease of cortex-associated myosin upon EMT and can be causative to cortical stiffness and contractility reduction, see figure 7.
In mitotic cells, EMT increases cortical actin nucleation through enhanced cortical RhoC signaling (e.g. via mDia1) and Rac1 signaling (via Arp2/3). On the other hand, EMT decreases cortical actin through reduced cortical RhoA signaling and enhanced cofilin signaling. The integration of all signaling changes can account for the observed net increase of cortical actin upon EMT in mitosis, see figure 6(d) and [5].
Further, EMT promotes myosin activity at the mitotic cortex through enhanced RhoC signaling, but diminishes it through reduced RhoA signaling and via increased Rac1 signaling (mediated downstream by Arp2/3 signaling, see figure 6(d). Through the combination of these opposite effects, the net change of cortical myosin II may vanish as was observed in MCF-7 cells, see figure 7.
In conclusion, we find that EMT induces complex modifications in actin-cytoskeletal signaling through a combination of changes in the signaling of Rho GTPases and downstream effectors such as cofilin, Arp2/3 and mDia1. The integration of all partly opposing effects give rise to an emergent change of actin and myosin at the cortex. In particular, our findings shed further light on how differences emerge in cortical composition and mechanics that are distinct in interphase and mitosis. Finally, we note that our study provides a cellular EMT fingerprint of rounded cells that may be relevant for cancer diagnostic approaches in particular for those that rely on isolated cells such as FACS-related assays or deformability flow cytometry approaches [53][54][55].

AFM measurement of cells Experimental setup.
To prepare mitotic cells for AFM measurements, approximately 10 000 cells were seeded in a cuboidal silicon cultivation chamber (0.56 cm 2 area, from cutting ibidi 12-well chamber; ibidi, Gräfelfing, Germany) that was placed in a 35 mm cell culture dish (fluorodish FD35-100, glass bottom; World Precision Instruments, Sarasota, FL) 1 day before the measurement so that a confluency of ∼30% was reached on the measurement day. Mitotic arrest was induced by supplementing S-trityl-L-cysteine (Sigma-Aldrich) 2-8 h before the measurement at a concentration of 2 µM. For measurement, mitotic-arrested cells were identified by their shape. Their uncompressed diameter ranged typically from 18 to 23 µm.
The experimental setup included an AFM (Nanowizard I; JPK Instruments, Carpinteria, CA) that was mounted on a Zeiss Axiovert 200M optical, wide-field microscope using a 20x objective (Plan Apochromat, NA = 0.8; Zeiss, Oberkochen, Germany) along with a CCD camera (DMK 23U445 from The Imaging Source, Charlotte, NC). Cell culture dishes were kept in a petri-dish heater (JPK Instruments) at 37 • C during the experiment. Before every experiment, the spring constant of the cantilever was calibrated by thermal noise analysis (builtin software; JPK) using a correction factor of 0.817 for rectangular cantilevers [58]. The cantilevers used were tipless, 200-350 µm long, 35 µm wide, and 2 µm thick (CSC37, tipless, no aluminum; Mikromasch, Sofia, Bulgaria). The nominal force constants of the cantilevers ranged between 0.2 and 0.4 N m −1 . The cantilevers were supplied with a wedge, consisting of UV curing adhesive (Norland 63; Norland Products, East Windsor, NJ) to correct for the 10 • tilt [59]. The measured force, piezo height, and time were output with a time resolution of at least 500 Hz.
Dynamic AFM-based cell confinement. Preceding every cell compression, the AFM cantilever was lowered to the dish bottom in the vicinity of the cell until it touched the surface and then retracted to ≈14 µm above the surface. Subsequently, the free cantilever was moved and placed on top of the cell. Thereupon, a bright-field image of the equatorial plane of the confined cell was recorded to evaluate the equatorial radius R eq at a defined cell height h. Cells were confined between dish bottom and cantilever wedge. Then, oscillatory height modulations of the AFM cantilever were carried out with oscillation amplitudes of 0.25 µm at a frequency of 1 Hz.
During this procedure, the cell was on average kept at a normalized height h/D between 60 and 70%, where D = 2(3/(4π)V) 1/3 and V is the estimated cell volume. Using molecular perturbation with cytoskeletal drugs, we could show in previous work that at these confinement levels, the resulting mechanical response of the cell measured in this setup is dominated by the actin cortex (see figure 4 in [35] and figure S7 in [5]). This is further corroborated by our observation from previous work that the smallest diameter of the ellipsoidal cell nucleus in suspended interphase cells is smaller than 60% of the cell diameter (see figure S4 in [17]). Additionally, it has been shown that for a nucleus-based force response, when measuring cells in suspension with AFM, a confinement of more than 50% of the cell diameter is needed [60].
Data analysis. The data analysis procedure was described in detail in an earlier work [35]. In our analysis, the force response of the cell is translated into an effective cortical tension γ = F/[A con (1/R 1 + 1/R 2 )], where A con is the contact area between confined cell and AFM cantilever and R 1 and R 2 are the radii of principal curvatures of the free surface of the confined cell. Here R 1 is estimated as half the cell height h and R 2 is identified with the equatorial radius R eq [17,35,56]. Cell height h and equatorial radius R eq were estimated from the AFM readout and optical imaging, respectively [35]. For the determination of the radius of the contact area A con , see also supplementary section 5 in [56].
Oscillatory force and cantilever height readouts were analyzed in the following way: for every time point, effective cortical tension γ and surface area strain ϵ(t) = (A(t) − ⟨A⟩)/⟨A⟩ were calculated. Here, A(t) is the total surface area of the confined cell. It is estimated as the area of a rotationally symmetric body with semi-circular free-standing side walls at cell height h and equatorial radius R eq , i.e. A = π(2h(R eq − h/2)(π/2 − 1) + h 2 /2 + 2R 2 eq ) [17]. An amplitude and a phase angle associated to the oscillatory time variation of effective tension γ and surface area strain are extracted by sinusoidal fits. To estimate the value of the complex elastic modulus at a distinct frequency, we determine the phase angles φ γ and φ ϵ as well as amplitudes A γ and A ϵ of active cortical tension and surface area strain, respectively. The complex elastic modulus at this frequency is then calculated as A γ /A ϵ exp(i(φ γ − φ ϵ )).
Statistical analyses of cortex mechanical parameters were performed in MATLAB using the commands 'boxplot' and 'ranksum' to generate boxplots and determine p-values from a Mann-Whitney U-test (two tailed), respectively.

Imaging of transfected cells
Transfected cells were placed on PLL-g-PEG coated fluorodishes (FD35-100) with CO 2 -independent culture medium (described before). Cellular DNA was stained with Hoechst 33342 solution (PN:62249, Invitrogen) in order to distinguish between mitotic and interphase cells. During imaging, cells were maintained at 37 • C using an Ibidi heating stage. Imaging was done using a Zeiss LSM700 confocal microscope of the CMCB light microscopy facility, incorporating a Zeiss C-Apochromat 40x/1.2 water objective. Images were taken at the equatorial diameter of each cell at the largest cross-sectional area (see figure 6(a)).

Calculation of cortex-to-cytoplasm ratios
This has been described before [5]. In short, using a MATLAB custom code, the cell boundary was identified ( figure 1(b) shows an exemplary cell, the cell boundary is marked in red). Along this cell boundary, 200 radial, equidistant lines were determined by extending 1.5 µm to the cell interior and 2.5 µm into the exterior ( figure 1(b), red lines orthogonal to cell boundary, only every tenth line was plotted out of 200). The radial fluorescence profiles corresponding to these lines were averaged over all 200 lines ( figure 1(c), blue curve). This averaged intensity profile is then fitted by a linear combination of an error function (cytoplasmic contribution) and a skewed Gaussian curve (cortical contribution), see figure 1(c), orange curve. The respective fit formula is given by (1) Here p = {µ, σ 1 , σ 2 , I cyt , I cort , I BG , α} are fit parameters, which determine the position of the cortex (µ), the slope of the error function decay (σ 1 ), the width of the cortical Gaussian peak σ 2 , the amplitudes of the error function and the Gaussian peak (I cyt and I cort ) and the skewness of the Gaussian peak (α). To calculate the cortex-to-cytoplasm ratio, the fitted skewed Gaussian is integrated and the obtained integral is then normalized by the cytoplasmic intensity I cyt [5].

Immunostaining and confocal imaging of cells
Immunostaining of suspended cells (interphase or STC-arrested mitotic) was performed as described previously [62]. Briefly, before fixation, cultured cells were detached by the addition of 0.05% trypsin-EDTA (Invitrogen) and resuspended in a glassbottom dish (e.g. ibidi; #80826) at a density of ≈3 × 10 5 /cells per cm 2 . In order to not be washed away during washing steps, cells were left to incubate and weakly adhere for ≈10 min. Then, cells were fixed with 3.7% PFA/PBS for 10 min (10% TCA for  Fluor 488 conjugate at a concentration of 1:1000 in 5%BSA/PBS for 2 h at room temperature. At the same time, cells were treated with 5 µg ml −1 DAPI (2 min) and 0.2 µg ml −1 Phalloidin-iFluor-647 (10 min) in 5% BSA/PBS solution. Images were taken with a Zeiss LSM700 confocal microscope of the CMCB light microscopy facility, incorporating a Zeiss C-Apochromat 40x/1.2 water objective. Images were taken at the equatorial diameter of each cell showing the largest cross-sectional area.

Western blotting
Protein expression in MCF-7 cells before and after EMT was analyzed using western blotting. Cells were seeded onto a six-well plate and grown up to a confluency of 80%-90% with or without EMT-inducing agents. Thereafter, cells were lysed in SDS sample/lysis